Hydrodynamics

The HYDRODYNAMICS
of CAVITATING FLOWS
Alexey G. Terentiev
Ivan N. Kirschner
James S. Uhlman
THE HYDRODYNAMICS OF CAVITATING FLOWS
Alexey G. Terentiev, Ivan N. Kirschner, James S. Uhlman
Edited by Alfred Tunik
Cover design by Jarek Kupsk
The image of cloud cavitation in the flow past a fixed hydrofoil in the front cover
background is courtesy of Roger Arndt
Printed in the United States of America

by G&H SOHO Inc., Hoboken, NJ
ISBN-13 978-09742019-5-5
ISBN-10 0-9742019-5-2
© 2011 Backbone Publishing Company

All rights reserved. No parts of this book may be used or reproduced in any manner
whatsoever without written permission except brief quotations in critical articles or reviews
Library of Congress Control Number: 2011924385
Backbone Publishing Company

P.O. Box 562, Fair Lawn, NJ 07410, USA
tel: 201 447 1834
FAX: 201 670 7892
www.backbonepublishing.com
alfred@backbonepublishing.com
alfred.tunik@gmail.com
ABSTRACT
The Hydrodynamics of Cavitating Flows by A. Terentiev, I. Kirschner, J. Uhlman
© 2011 Backbone Publishing Co. ISBN 978-09742019-5-5, Trade Cloth, 7.5"x10", 598 p;
430 figs; 638 ref.
This monograph is intended to introduce the reader to the physics and mathematics of
cavitation and provide an in-depth review of the broad range of important effects that occur
in gas-fluid multiphase areas under various circumstances. Most importantly, the text
provides an expansive overview of the many modeling approaches that have proven
successful in investigating cavitating flows, with sufficient detail that the advanced reader
can implement and apply the techniques to help answer the new cavitation-related questions
that arise as the science and technology of fluid mechanics advances. Though the emphasis
of the book is put on physical modeling and computational fluid dynamics, the numerous
modeling and computational plots are richly complimented by experimental photographs
courtesy of several leading laboratories and individual experts. The list of chapter titles is
self-explanatory for outlining the content of the book.
Ch.1. Physical Description of Cavitation (17p). Ch.2. Main Boundary-Value
Problems of Cavitating Flow (34p). Ch.3. Method of Conformal Mapping (60p). Ch.4.
Method of Expansion in Series (36p). Ch.5. Asymptotic Methods in Cavitating Flow (54p).
Ch.6. Unsteady Cavitating Flows (47p). Ch.7. Approximation Methods (31p). Ch.8.
Supercavity as a Dynamical System (42p). Ch.9. Numerical Methods (67p). Ch.10. Method
of Singularities in Cavitating Flow (36p). Ch.11. Selected Effects in the Modeling of
Cavitating Flows in Real Fluids (49p). Ch.12. Cavitating Flow of Marine Propellers (30p).
Ch.13. Applications of Cavitation Theory to Continuum Mechanics. (23p) Appendices.
References. Subject Index. Name Index.
Authored by three principal co-authors with tangible contributions from many
leading experts, the monograph brings together the perspectives of Russian and American
hydrodynamicists, and thus provides a unique and useful treatment of independently
developed physics models.
The monograph is addressed to professionals specializing in hydrodynamics who are
involved both in theoretical and experimental research and practical applications of their
studies for improving the performance of marine vehicles. It is a comprehensive reference
source on cavitating flows which can be very helpful for university professors and post-
graduates, and a unique reference source for research institutions, technical libraries,
universities, design offices and consulting experts.
Table of Contents
Editor's Preface ……………………………………………………………… 7
Acknowledgements .....………………………………………………………. 8
Abbreviationd and Acronyms ………………………………………………. 10
Ch. 1. A Physical Description of Cavitation ………………………………… 11
Sec.1.1.Introduction …………………………………………………………… 11
Sec.1.2. Dimensionless Quantities Describing Cavitating Flows ………. 13
Sec.1.3. .Main Assumptions in the Theory of Cavitating Flow …………. 24
Sec.1.4 .Summary …………………………………………………………… 26
Ch. 2. Main Boundary-Value Problems of Cavitating Flow ………………… 28
Sec.2.1. Mathematical Modeling of Cavitating Flow ………………………. 28
Sec.2.2. Simple Examples of Cavitating Flows …………………………….. 33
Sec.2.3. Fully Cavitating Flat Plate ………………………………………… 37
Sec.2.4. Partially Cavitating Flow …………………………………………. 45
Sec.2.5. Review of Cavitating Flow Models ………………………………. 51
Sec.2.6. Open Models of Cavitating Flow …………………………………. 56
Ch. 3. Method of Conformal Mapping ……………………………………….. 62
Sec.3.1. Mathematical Aspects of Conformal Mapping ………………….. 62
Sec.3.2. Cavitating Flow around a Wedge ………………………………… 66
Sec.3.3. A Wedge in a Channel and in a Jet ……………………………….. 73
Sec.3.4. Cavity with a Negative Cavitation Number ……………………… 79
Sec.3.5. Flat Plate near a Free Surface ……………………………………. 83
Sec.3.6. Flat Plate near a Flat Bottom …………………………………….. 88
Sec.3.7. Cavitating Flow around a Plate near a Free Surface …………. 91
Sec.3.8. Applying Tulin’s Model to Cavitating Flows ……………………. 95
Sec.3.9. Cavitating Flow in a Channel with Flat Sides ………………….. 103
Sec.3.10. Cavitating Flow of a Cascade of Flat Plates …………………. 107
Sec.3.11. Equilibrium Cavitating Vortices in a Stream ……….……….. 114
Ch. 4. Method of Expansion in Series ………………………………….. 122
Sec.4.1. Cavitating Flow around a Wedge with Curved Sides …………… 122
Sec.4.2. Cavitating Flow around a Circular Cylinder ……………………. 127
Sec.4.3. Cavitating Flow in a Longitudinal Gravity Field ……………….. 131
Sec.4.4. Supercavitating Flat Plate in a Transverse Gravity Field ……… 136
Sec.4.5. Axisymmetric Cavitating Flow past a Sphere ……………………. 141
Sec.4.6 Surface Tension in Free Boundaries ……………………………….. 145
Ch. 5. Asymptotic Methods in Cavitating Flow ………………………………... 158
Sec.5.1. Asymptotic Approaches in Hydrodynamics ……………………. 158
Sec.5.2. Boundary-Value Problem Statements in Linear Theory …..…… 164
Sec.5.3. Steady Flow around Cavities …………………………………… 170
Sec.5.4. Cavitating Flow around a Curvilinear Foil …………………… 179
Sec.5.5. Small Nonlinear Flow Perturbations …………………………….. 185
Sec.5.6. Matched Asymptotic Expansion ………………………………… 190
Sec.5.7. Asymptotic Models of Axisymmetric Cavitating Flows …………. 197
Sec.5.8. Asymptotic Models of Axisymmetric Supercavitating Flows ……. 203
4
Ch. 6. Unsteady Cavitating Flows ……………………………………………….. 212
Sec.6.1. Unsteady Movement of a Slender Hydrofoil ……………………. 212
Sec.6.2. Unsteady Movement of a Plate with a Kirchhoff Cavity ………. 221
Sec.6.3. Water Entry of a Thin Hydrofoil ……………………………….. 230
Sec.6.4. Analytic Extension Method for a Free Surface between Two Fluids 239
Sec.6.5. Momentum and Energy of Impact Upon or Entrance Through a Free
Surface 243
Sec.6.6. Green’s Function in Linear Theory ……………………………. 247
Sec.6.7. Unsteady Movement of a Wing of Finite-Span ……………………. 252
Ch. 7. Approximation Methods …………………………………………………… 259
Sec.7.1. Dropping of a Plate on a Flat Bottom – Exact Solution ………. 259
Sec.7.2. Gravity and Air Effects during Water Entry …………………… 262
Sec.7.3. Evaporation from a Cavity Boundary …………………………. 269
Sec.7.4. Perturbed Flow around a Foil …………………………………. 273
Sec.7.5. Approximation Method for a Compressible Fluid Flow ……….. 277
Ch.8. The Supercavity as a Dynamical System …………………………… 280
Sec.8.1. Logvinovich’ Principle and Its Formalization …………………… 281
Sec.8.2. A Simple Approach to Estimating 3D Supercavitating Flow …… 287
Sec.8.3. Mathematical Model of an Unsteady Supercavity ……………… 293
Sec.8.4. Linear Theory of Cavity Stability and Oscillations ………………. 295
Sec.8.5. Development of the Linear Theory ………………………………… 299
Sec.8.6. Approximating Cavity Initiation as a Rayleigh Bubble ………… 303
Ch.9. Numerical Methods …………………………………………………. 322
Sec.9.1. Boundary Element Method …………………………………………… 322
Sec.9.2. Penetrable Hydrofoil …………………………………………………. 332
Sec.9.3. Numerical Modeling of Cavitating Flow ………………………… 337
Sec.9.4. Non-Symmetric Cavitating Flow …………………………………… 342
Sec.9.5. Numerical Modeling of Perturbed Flows …………………………. 349
Sec.9.6. Numerical-Analytical Method for Arbitrarily-Shaped Foils ……… 354
Sec.9.7. Complex Variable BEM for Plane Stationary Problems ………… 358
Sec.9.8. Unsteady Flow with Free Boundaries ……………………………… 365
Sec.9.9. Generation of Waves on Free Boundaries ………………………… 369
Sec.9.10. Deformation of Gas Bubbles ……………………………………….. 372
Sec.9.11. Three-Dimensional Problem of Bubble Deformation ………… 374
Sec.9.12. Finite Particle Method in Entry Problems ……………………. 378
Sec.9.13. Smoothed Particle Hydrodynamics ……………………………… 382
Ch.10. Method of Singularities in Cavitating Flow …………………………… 389
Sec.10.1. Velocity-Based Method for Cavitating Flow past Thin Foils ……. 389
Sec.10.2. Nonlinear Velocity-Based Method for Cavitating Hydrofoils …… 395
Sec.10.3. Velocity-Based Method for 3D Cavitating Flow past Thin Foils ... 400
Sec.10.4. Nonlinear Potential-Based Method for Cavitating 2D Hydrofoils 402
Sec.10.5. Nonlinear Potential-Based Method for Supercavitating Flow past
Axisymmetric Bodies …………………………………………………….. 407
Sec.10.6. Nonlinear Potential-Based Method for Partially Cavitating Flow past
Axisymmetric Bodies …………………………………………………. 412
Sec.10.7. Nonlinear Potential-Based Method for 3D Hydrofoils 415
5
Sec.10.8. Computation of the Added Mass and Damping of Supercavitating
Axisymmetric Bodies ………………………………………………….. 419
Ch.11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids 426
Sec.11.1. A First-Principles Model of Real Effects on Cavity Detachment 429
Sec.11.2. High-Speed Motion of a Body through a Bubbly Liquid …….. 443
Sec.11.3.Micro-Scale Effects ………………………………………………. 473
Ch.12. 4. Cavitating Flow of Marine Propellers ……………………………. 475
Sec.12.1. Historical Notes and Testing Facilities ……………………….. 475
Sec.12.2. Supercavitating and Highly Cavitating Screw Propellers ….. 477
Sec.12.3. Mathematical Modeling …………………………………………. 487
Sec.12.4. Simple Mathematical Statements and Numerical Algorithms 498
Ch.13. Applications of Cavitation Theory to Continuum Mechanics ……… 505
Sec.13.1. Impulsive Impact on the Ground …………………………….… 505
Sec.13.2. Mathematical Modeling of Underground Sprinkler System …. 511
Sec.13.3. Series Expansion for Flows of Limited Reynolds Number .. ….. 517
Sec.13.4. Application of BEM to Small Reynolds Number Flows ……….. 523
Appendices ………………………………………………………………… 528
App 1. Boundary-Value Problems for Singular Functions ……………….. 528
App 2. Integral Relationships ………………………………………………… 531
App 3. Singular Functions ……………………………………………………. 536
App 4. Generalized Analytic Functions ………………………………. 539
App 5. Calculation of Integrals and Special Functions …………………. 542
App 6. Velocity Field as a Distribution of Vortices Over the Boundary 549
App 7. 2D Mass and Momentum Control Volume Analyses ……………. 550
App 8. Axisymmetric Mass and Momentum Control Volume Analyses … 552
App 9. Axisymmetric Green Functions ………………………..…………. 553
References …………………………………………………………………... 557
Subject Index ……………………………………………………………….. 585
Name Index ……….………………………………………………………… 594
6
Editor's Preface
“There is nothing more practical than a good theory” – a saying attributed to Niels Bohr – is
fully applicable to this monograph. Its main emphasis is to comprehend the accumulated
theoretical background and computational techniques and to provide advanced analytical
models and computational tools capable of solving numerous practical problems. And all
that is for a most complicated field in modern hydrodynamics. The complexity is due to the
fact that the cavitating flow is a mixture of two fundamentally different media: liquid and
gas (usually water and air) separated by one or more free boundaries. In other words, the
practicality of this physics and math theory is in providing a foundation for not merely
improving but rather real break-through in performance of marine vehicles, projectiles and
propulsive systems.
It was a pleasure and a privilege to work with such distinguished team of authors
that includes Prof. Alexey G. Terentiev, a peer in this field in Russia, and his American co-
authors Dr. Ivan N. Kirschner and Dr. James S. Uhlman, prominent experts in the fast-
growing pool of American specialists in this area.
The undersigned is grateful to Dr. Kirschner for selecting Backbone Publishing to
present this advanced “by-lingual” book to English-speaking engineers and for his helpful
advices in editing it. Thanks are also due to Dr. Victor Mishkevich for reviewing chapter 12,
and to Dr. Eduard Amromin for, besides his considerable contributions to the book,
discussing and explaining some topics in the book. The fine work of cover design by Jarek
Kupsk is gratefully acknowledged. And last, but far, far from least, thanks are due to my son
Leonid Tunik, the Backbone’s founder, for his never-ending techsupport in computer
applications.
Alfred Tunik, PhD

Editor
Backbone Publishing
Fair Lawn, NJ
7
Acknowledgements
The authors gratefully appreciate the valuable contributions and the help willingly provided
for this monograph by many individuals from Russia and the United States of America. The
information presented in this book represents the contributions not only of the three primary
co-authors, but also direct contributions from several colleagues. The book also incorporates
background knowledge from the technical literature that was contributed by many experts in
the field over decades, and – in some cases – centuries. We are grateful to all of these
individuals, to the sponsors who supported our research, and to various colleagues who
provided either direct support for the publication or invaluable technical correspondence
over many years.
We are especially grateful to those colleagues whose direct contributions appear in
this book, as follows:
Sec. 5.6 - written jointly with Kirill V. Rozhdestvensky;
Sec. 5.8 - written by Vladimir V. Serebryakov;
Sec. 8.3 - written by Emil V. Paryshev;
Sec. 9.13 - prepared using results by Konstantin E. Afanasiev and Sergey V. Stukolov;
Sec. 11.1 - written jointly with Éduard L. Amromin;
Sec. 11.2 - written with contribution by John R. Grant; and,
Sec. 12.1 to Sec.12.3 - written by Alexander Sh. Achkinadze.
Thanks are also in order to the following individuals and organizations for providing
various images presented in this book, as indicated throughout the text: Prof. Roger Arndt
(University of Minnesota/Saint Anthony’s Falls Laboratory), Prof. Christopher Brennen
(California Institute of Technology), and Prof. Tadd Truscott (Brigham Young University).
Many of the images of supercavitating projectiles were made at the Naval Undersea Warfare
Center (NUWC) Division Newport, Rhode Island, with the contributions of an extensive
scientific team, including Test Director J. Dana Hrubes, who led all aspects of photography,
cinematography, and videography.
Many organizations in Russia and the U.S.A. provided financial support for the
research that underlies this monograph. In Russia, special thanks are due to the Cheboksary
Polytechnic Institute of Moscow State Open University and Chuvash State University. In
the U.S.A., special thanks are due to the following individuals who sponsored research and
development at one time or another through their organizations: James Fein, Sharon
Beermann-Curtin, Kam Ng, Teresa McMullen, Ronald Joslin, and Scott Hassan (currently or
formerly of the Office of Naval Research); William Carey, Theo Kooij, and Khine Latt
(currently or formerly of the Defense Advanced Research Projects Agency); John Sirmalis,
James Meng, Bernard Myers, Robert White, Charles Elste, Clifford Curtis, and Thomas
Gieseke along with the In-house Laboratory Independent Research Program, then managed
by Stuart Dickinson and later by Richard Philips (all currently or formerly of the of NUWC
Division Newport); and Franz Edson, Jennifer Panosky, John Chapman and Joseph Porcaro
(General Dynamics Electric Boat).
The authors have also benefited from fruitful technical collaboration and discussions
with many different colleagues over many years. They are too numerous to mention, but
8
without the benefit of their technical insight and generous suggestions and criticism, the
publication of this book would not have been possible.
Nadezhda A. Dimitrieva supported type-scripting of portions of this book, along
with contributions to the development of the subject and name indices (in addition to her
own technical contributions to the field of cavitation). Crystal Henry also provided
invaluable support for various administrative and editorial processes.
Finally, each of us is indebted to our publisher, Alfred Tunik for his ongoing and
diverse help in preparing the book, including support for translation of the technical text,
securing permission to publish certain photographs and material, and – most importantly –
for his great patience during the completion of this project.
9
Abbreviations and Acronyms
AN Ukr. SSR = Academy of Sciences of the Ukrainian SSR
AN USSR = Academy of Sciences of the USSR
BEM = boundary element method
CAHI = TsAGI= Central Aero-Hydrodynamic Institute
CCVL = Concept of Cavitation within Viscous Layers
CFD = Computational Fluid Dynamics
Ch. = Cheboksary
CSU = Chuvash State University, Cheboksary, Russia
DAN USSR = Doklady AN USSR
FML = GIFML = Physical-Mathematical Literature Press
GTTI = State Press for Technical and Theoretical Literature
HCP = highly cavitating propeller
IDE = integro-differential equation
J. = Journal (of)
JSR = Journal of Ship Research
JFM = Journal of Fluid Mechanics
KSRI = Krylov Shipbuilding Research Institute
L. = Leningrad
LIWT = Leningrad Institute of Water Transport
LKI = Leningrad Shipbuilding Institute (now MTU)
M. = Moscow
MAE = Matched Asymptotic Expansion (methods)
MGU = Moscow State University
MT = Marine Technology
MTU = Marine Technical University (formerly Leningrad Shipbuilding Institute=LKI)
MZG = Mechanics of Fluids and Gases, AN USSR (or RAN)
NANI = (Russian) National Academy of Sciences and Arts
NAS of Ukraine = National Academy of Sciences of Ukraine
NKI = Nikolaev Shipbuilding Institute
NTOSP = Society of Shipbuilding Engineering (in USSR or Russia)
NUWC = Naval Undersea Warfare Center
ONR = Office of Naval Research
PMM = Journal of Applied Mathematics & Mechanics
RAN = Russian Academy of Sciences
RINA = Royal Institution of Naval Architects
RKPM = Reproducing Kernel Particle Method
PMTF = Applied Mechanics and Technical Physics, AN USSR (or RAN)
SBT = Slender Body Theory
SCP = CK = supercavitating propeller
SNAME = Society of Naval Architects and Marine Engineers
SPb = St. Petersburg,
SPbSU = St. Petersburg State University
TsAGI = see CAHI
UTCC = Uniform Turbulent Cavitation Concept
ZIAM = Central Institute of Aviation Machinery
10
Ch. 1. A Physical Description of Cavitation

Sec. 1.1. Introduction
Cavitation is a broad phenomenon that may be defined in many different ways. In general,
however, it involves the presence of gas in a flow of fluid that is predominantly composed of
the liquid phase. Cavitation in fluid mechanical systems of interest to engineers can appear
in several forms, depending on the geometry of the flow and the boundary conditions. Its
occurrence can drastically affect the performance of a system, either degrading it (as in the
case of the loss of lift of a hydrofoil intended for fully-wetted operation), or improving it (for
example, by reducing the drag of an axisymmetric body when properly designed).
Effects that are usually considered undesirable include reducing the lift or thrust of
machinery, cavitation-excited noise and vibration and the rapid and severe erosion that can
occur when an object such as a marine propeller is exposed to sustained cavitation collapse.
On the other hand, these effects are occasionally put to practical use, as for example, when
applied to underwater boring. Below some physical descriptions of cavitation are presented.
Cavities appear in a liquid at high flow speeds due to phase changes from liquid to
gas, followed by displacement discontinuity and the appearance of cavities first as nuclei
growing to form large cavities. As an example of various forms of cavitation appearance,
consider a two-dimensional flow past several identical hydrofoil sections operating at
various conditions. In particular, consider a fluid particle traversing a streamline from
upstream infinity past each of the foil sections presented in Fig.1.1.1.
Cavitation can appear in several other forms as well. For example, a sudden drop in
the ambient pressure of a quiescent, nucleated liquid can lead to cloud cavitation – a
dispersed collection of bubbles. Such a circumstance could be produced by putting tap water
under tension by the action of a piston in a cylinder. Back bubble cavitation is a highly
unsteady form of cloud cavitation, see Fig.1.1.1a,d-f.
So far, the cavitation being discussed is solely associated with a phase change of a
liquid into its vaporous form, and the process is therefore referred to as vaporous cavitation
or natural cavitation. The extent of a cavity can be increased by supplying air to its interior,
in which case the process is known as ventilation or ventilated cavitation. Ventilation can
occur naturally, for example, the bleeding of air from the ocean surface to a rudder, or a
ventilation system may be incorporated as part of a design, such as is sometimes provided on
supercavitating high-speed undersea vehicles.
A common characteristic of all forms of cavitation is the presence of one or more
free boundaries. In the case of cloud cavitation, the surface of each individual bubble is a
free boundary that deforms according to the local dynamics of the flow. For sheet
cavitation, as shown in Fig.1.1.1c-e, the free boundary originates at the cavity detachment
point, and persists downstream to the region of cavity closure. Under certain conditions, the
free boundaries interact or even merge with a free surface bounding part of the fluid domain
under consideration, as in the case of a surface-piercing propeller, where a ventilated cavity
forms as each blade enters the water and evolves as the blade traverses its rotation.
It is the presence of free boundaries that makes the mathematics of cavitation more
interesting than that associated with fully-wetted flow. The free surface boundary conditions
are quite different from those on fully wetted surfaces. Moreover, unless some stringent and
very limiting simplifications are made, such problems are inherently nonlinear, since the
11
geometry of the free boundary must be determined as part of the problem solution. Also due
to the boundary conditions on the free surface, for a broad range of circumstances, cavitation
is an unsteady effect (Fig.1.1.1b) at least in local regions of the flow. E.g. sheet cavitation
over a hydrofoil can remain relatively steady, except near the cavity end (Fig1.1.1d,e). As
the free boundary conditions are incompatible with a steady stagnation point, the flow there
becomes unstable, at least locally, and leads to one of several unsteady processes.
e f
Fig.1.1.1. Types of cavity formation: a) bubble cavitation; b) unsteady cavitation; c) full
(sheet) cavitation, d) partial cavitation; e) partial cavitation of finite span wing; d) cloud
formation in sheet and tip cavitation. (a-e are reproduced from the cover of Proc. 3rd Int’l
Symp. on Cavitation, 1998, Grenoble;and f is reproduced from [Bark, 1998]).
The physics of cavitating flows emphasizes macro-scale effects. The micro-physics

of nucleation, intermolecular interaction, and the fluid chemistry associated with such effects
as the modification due to surface tension are beyond the scope of this book; more can be
found in [Brennen, 1995] or other references as an introduction to such topics. The
12
remainder of this chapter provides an overview of the physics of cavitation sufficient to
ground the reader in the first principles on which the mathematical models are based. In
order to help identify those physical quantities that are of greatest importance to those
models, a set of dimensionless quantities is introduced.
Sec. 1.2. Dimensionless Quantities Describing Cavitating Flow

This treatise is primarily concerned with the prediction of two basic aspects of cavitation: its
extent and the forces – local or global – that it produces on the surface of the particular body
under consideration. The extent of cavitation is characterized by the principal dimensions of
the cavity or cavities that are produced. Both the extent of cavitation and the associated
forces may be time-dependent in general. For most flows of interest, time-dependent effects
are unavoidable (especially near cavity closure), although under certain conditions they may
be negligible for predicting many quantities of engineering interest.
The effect of cavitation on the global forces acting on a body, such as the lift and
drag on a hydrofoil, are typically characterized as coefficients of the components of the
generalized force (that is, force or moment) acting on a body: drag, lift, and hydrodynamic
moment. Frequently, interest is focused on how cavitation modifies the forces that would be
produced under conditions for which cavitation did not occur – a loss of lift on a foil, or the
reduction of drag on a projectile – due to the generation of a cavity. Since cavities are
inherently unsteady in most cases of interest, the general problem must account for time-
dependent effects. Moreover, as will be demonstrated, even if the nominal flow is globally
steady, the interaction between a body and a cavitating fluid often comprises a system whose
behavior is not – just as a flat stone spun from the hand skips off the surface of a still lake, a
cavitating body may draw energy from its forward motion to respond in an oscillatory or
even a chaotic mode.
Local effects are more properly characterized as a stress distribution. One example
is the pressure field produced near cavity closure on the surface of a partially cavitating
body, which under almost all practical conditions features a nearby stagnation point.
Another is the highly transient stress field – both in the fluid and in the adjacent body –
associated with the collapse of an individual cavitation bubble or cloud of bubbles.
Neither the extent of cavitation nor the forces or the stress field it generates has any
meaning without a characterization of the tendency of the flow to cavitate in the first place.
For many problems of engineering interest, the most fundamental dimensionless quantity for
the investigation of cavitation is the cavitation number, or cavitation index:
σ 2( p∞ − pc ) ρV 2 , where p∞−pc is the difference between pressures at infinity and in
the cavity; ρ is the density of liquid and V is the free stream speed. This parameter is the
ratio of static pressure forces to dynamic pressure forces, or (from an alternative perspective)
to the kinetic energy per unit volume. For vaporous cavitation, the relevant cavity pressure is
the vapor pressure of the ambient liquid. For ventilated cavity flows, the cavity pressure is a
function of the ventilation rate, which in turn is characterized by the ventilation coefficient.
In either case, when the ambient pressure is not significantly greater than the
pressure required to sustain a gaseous state in the flow mixture (in comparison with the flow
kinetic energy per unit volume), the working fluid tends to feature one or more regions
characterized by high void fractions of gas – that is, localized pockets of fluid where the
volumetric proportion of gas to liquid is large. For vaporous cavitation, such regions are
13
filled with vapor (or populated by vapor bubbles) produced via a phase change of the
ambient liquid to its gaseous form. For ventilated cavitation, the cavities may be primarily
composed of the ventilation gas supplied to the system (although under many conditions of
engineering interest the vaporous phase of the ambient liquid may contribute a non-
negligible fraction of the total gaseous mixture in the cavity).
Table 1.2.1 presents many of the dimensionless parameters familiar from the study
of fluid mechanics, and discusses them in the context of cavitation. As can be deduced from
the table, for many problems of engineering interest some of these parameters are clearly
more important than others. Specifically, it is shown that viscous and surface tension effects
are relatively unimportant to many flows of interest, such that the physics are governed by
the competition between static and dynamic pressure forces, as characterized by the
cavitation number. For many common problems, gravitational effects and the effects of a
ventilation gas supply compete with those due to the static and dynamic pressure forces, and
the cavity Froude number (the ratio of inertial forces to gravitational forces) and the
ventilation coefficient (the ratio of the volumetric ventilation rate to the nominal advection
rate) become important. The relationship among the cavitation number, the cavity Froude
number, and the ventilation coefficient is complicated and important, especially for
nominally axisymmetric cavities, and serves to categorize such flows into several regimes of
cavity closure: the re-entrant jet regime, the twin vortex regime, and the intermediate
toroidal shedding regime, e.g. [Brennen, 1969]. Fig.1.2.1 presents a comparison between the
re-entrant jet regime and the twin vortex regime of axisymmetric cavitating flows
Thus, a physical understanding of cavitation may be enhanced by examining within
that specific context some of the key dimensionless parameters that are commonly employed
in the study of general hydrodynamics. Several of the parameters commonly used in the
technical literature are presented in Table 1.2.1, including their names, physical definitions,
and a brief description of their relative importance to the study of cavitation.
In any branch of physics, two processes have identical characteristics (at least
statistically) if all of the relevant dimensionless parameters coincide. In experimental
practice, this is often difficult to achieve, as exemplified by the well-known challenge of
scaling to estimate the drag of a full-scale surface ship based on model tests. In that case,
viscous drag components depend mainly on Reynolds number, whereas the wave drag
depends on Froude number only. With the established practice of tow tank model tests at the
same Froude number, the tow speed is proportional to L1/2, where L is the model length.
Then the model Reynolds number of a sub-scale model is proportional to L3/2. Thus the
water-filled tow tank tests cannot be matched simultaneously to that of a full-scale ship
operating in a lake or ocean. Sub-scale testing with all of the parameters corresponding to
the full-scale ship is impossible. Such challenges are pervasive in hydrodynamics
experiments, and the sub-topic of cavitation is not immune to this issue. It thus becomes the
task of the hydrodynamicist to identify the parameters that are of most relevance to the
problem at hand, and provide a justification for neglecting the others, or at least relegating
them to a basic empirical assessment of the sensitivity of the main results to their variation.
Sophisticated experimental techniques may be used to match the most relevant
parameters. For example, the effects of gravity on cavitating flows can be studied by
simulating gravitational acceleration using fluid acceleration effects due to a converging or
diverging channel, or by performing experiments in a centrifuge, thereby matching the
Froude number at model and full scales.
14
d
Fig.1.2.1. A comparison of the re-entrant jet and twin vortex regimes of supercavitating
flow: a – re-entrant jet flow, which occurs at a relatively high value of the cavity Froude
number with respect to the cavitation number; b – twin-vortex flow, which occurs at a
relatively low value of the cavity Froude number; c – cavity closure on a downstream sting
in the twin vortex regime; and, d – cavity closure in the twin vortex regime with the cavitator
mounted on an upstream sting. The topology of the flow is a function of the flow regime: the
re-entrant jet drives liquid in the upstream direction inside the cavity, where some sort of
interaction with the cavity boundary (or the cavitating body) must occur, whereas the twin
vortex system forms a different streamline pattern in the liquid that confines such interaction
effects to the closure region. The theory presented in Ch. 8 shows that gas loss along the
twin vortex system stabilizes the cavity dynamics. (All images courtesy of R.. Arndt.)
15
Table.1.2.1. Dimensionless parameters relevant to the study of cavitating flows.
Parameter Definition Physical Quantities Compared
Name Symbol Numerator / Denominator
Parameters Relevant to All Cavitating Flows
cavitation p∞ − pc static pressure forces / dynamic pressure forces
number σ
1 2 ρV 2 (or kinetic energy per unit volume)
The most important parameter for the study of cavitation, the cavitation
number characterizes the tendency of the flow to cavitate. For vaporous
cavitation, the relevant cavity pressure is the vapor pressure of the ambient
liquid. For ventilated cavity flows, the cavity pressure is a function of the
ventilation rate, which in turn is characterized by the ventilation coefficient.
The relationship among the cavitation number, the cavity Froude number, and
the ventilation coefficient is complicated and important, especially for
nominally axisymmetric cavities, and serves to categorize such flows into
several regimes of cavity closure: the re-entrant jet regime, the twin vortex
regime, and the intermediate toroidal shedding regime, as is described more
fully elsewhere in this chapter.
Euler Δp
number Eu static pressure forces / inertial forces
1 2 ρV 2
The Euler number is a generalization of the cavitation number involving
pressure differences that may be of interest other than the difference between
ambient and cavity pressure. Frequently the factor of 1 2 is eliminated in the
definition, and occasionally the parameter is defined as the reciprocal of that
form. When the high pressure of interest to the system is much greater than the
low pressure, the numerator is sometimes specified as simply the high
pressure. This is the case for some cavitation problems, for example, for a
vaporous cavitating flow in water at any depth below a free surface subject to
atmospheric pressure, such that the ambient pressure is much greater than
vapor pressure. In that case the Euler number as defined above is nearly equal
to the cavitation number and is sometimes used as a surrogate.
Parameters Characterizing the Importance of Gravity
cavity Fr V2 g inertial forces / gravitational forces
Froude
number The cavity Froude number describes the importance of gravity to the flow
field. The characteristic length scale, , is usually taken as one of several
quantities: the cavitator base diameter or transverse dimension, the maximum
cavity diameter or transverse dimension, the cavity length, or (occasionally)
the maximum diameter or transverse dimension of a body enveloped in a
supercavity. At relatively high values of the cavity Froude number, gravity is
of little importance to the flow, and neither a nominally two-dimensional nor a
nominally axisymmetric cavitating flow will show significant deformation due
to gravity, whereas the converse is true for low and moderate values of the
parameter. It is the relative value of the cavity Froude number that is
important: for example, for nominally axisymmetric flows, the value of the
16
parameter must be compared with the cavitation number to judge the relative
importance of gravity to the problem. See the entry for cavitation number and
the text of this chapter for a discussion of the important relationship among the
cavitation number, the cavity Froude number, and the ventilation coefficient.
cavity Ri g vertical V 2 potential energy / kinetic energy
Richardson
number The Richardson number is the reciprocal of the square of a Froude number,
where the characteristic length scale is based on a representative vertical
dimension. As opposed to problems in oceanography and atmospheric
sciences, where the product of gravitational acceleration and the vertical length
scale is often very large relative to the flow speed of interest, for most of the
cavitating flows of interest in this book the vertical scales are small, and it
would appear at first glance that the Richardson number is insignificant, and
the kinetic energy of the flow is much more important than changes in
potential energy. However, it should be noted that, because the boundary
between the cavity and the ambient liquid is nearly a constant pressure surface
for most cavitating flows of interest, it may be shown that the cavity
Richardson number sets a lower bound on the cavitation number for any flow
in which the component of gravitation acceleration normal to the free stream is
non-zero. Thus, although the cavity Richardson number is small, its value sets
a critical bound on another parameter – the cavitation number – which,
although also small, is the parameter that is the main predictor of the tendency
of the flow to cavitate. Obviously, since the cavity Richardson number is
directly related to the cavity Froude number, the same conclusions can be (and
typically are) deduced from that more commonly employed parameter,
although the definition of the Richardson number emphasizes the energy
conservation principle of interest.
Parameters Involving Cavity Ventilation and Heat Transfer
ventilation CQ volumetric ventilation rate / nominal advection
Q V 2
coefficient rate
The ventilation coefficient characterizes the rate at which gas must be supplied
to a nominally steady cavitating flow in order to maintain the cavity. The
characteristic length, , is usually taken as one of several quantities: the
cavitator base diameter or transverse dimension, the maximum cavity diameter
or transverse dimension, or (occasionally) the maximum diameter or transverse
dimension of a body enveloped in a supercavity. Alternatively, the actual area
corresponding to the square of each of these length scales may be employed
via the introduction of the factor π . See the entry for cavitation number and
the text of this chapter for a discussion of the important relationship among the
cavitation number, the cavity Froude number, and the ventilation coefficient.
Eckert Ec V 2 c p ΔT kinetic energy / enthalpy
number
The Eckert number would be of theoretical interest in the study of heat transfer
and dissipation in supercavities that are ventilated from a gas generation
system, since it characterizes the heat transfer from the gas to the ambient
liquid, a cooling process that reduces the specific volume of the gas and
17
therefore its effectiveness in supporting the cavity. If such cooling is

associated with a phase change from a vaporous to a liquid phase, such as
would be the case for a steam ventilation system, the effectiveness of the
ventilation system may be extremely limited.
Prandtl ν cp μ
number Pr = viscous diffusivity / thermal diffusivity
α k
The Prandtl number is a material property, and is theoretically useful in
characterizing the relative importance of convection with respect to conduction
of heat within both the cavity gas or vapor and the ambient liquid. Low values
indicate conduction dominated heat transfer. For cavitating flows in water, the
Prandtl number of the ambient liquid is 7, whereas typical values are less than
unity for the gas within the cavity. If the heat transfer process of interest is that
associated with ventilation of a cavity by hot gas, it may therefore be expected
that convection dominates the heat transfer away from the cavity boundary into
the ambient liquid, whereas the effects of convection and conduction tend to be
more in balance within the gas layer, provided the flow is laminar. For
turbulent flow, the turbulent Prandtl number must also be considered (see the
separate entry below).
turbulent ν t c p μt momentum eddy diffusivity / thermal eddy
Prandtl Prt =
αt kt diffusivity
number
The turbulent Prandtl number is a property of the flow, being the ratio that
quantifies the relative importance of transport of momentum and heat by
turbulence. Since momentum transport by turbulence tends to dominate the
effects of viscosity at the high Reynolds numbers characteristic of most of the
cavitating flows of interest in this book, it is anticipated that the turbulent
Prandtl number is more important in characterizing the heat transfer process
than is the Prandtl number, for regions of the flow where turbulence is fully
developed.
Biot heat transfer resistance in cavity / heat transfer
Bi h kc
number resistance at cavity boundary
In general, the Biot number is important to transient problems of heat transfer,
and its value determines the spatial variation of temperature within a body
immersed in a fluid of a different temperature. The length scale is often
selected as the volume of the body divided by its surface area. For cavitation,
the Biot number would be of theoretical importance in determining the spatial
variation of temperature within a cavity ventilated by hot gas during a
transient, as the gas flows from a source near the cavitator downstream toward
cavity closure, where it escapes into the ambient liquid through a mixing
process. In this case, the appropriate length scale would be the volume of
the cavity divided by its surface area, and the appropriate heat transfer
coefficient h is the convective heat transfer coefficient. Biot numbers much
less than unity would be associated with relatively uniform temperature
distributions within the cavity. Heat transfer in cavities often involves
significant convective effects at the interface, from the hot gas within the
18
cavity that is typically characterized by relatively low thermal conductivity to a

much more conductive liquid that may be flowing at a somewhat different
speed. Under such conditions, it can be seen that the Biot number would take
on values significantly larger than unity, suggesting rapid temperature
variation spatially on either side of the interface. The problem is more
complicated if the cavity envelopes a body, which would be characterized by
yet another thermal conductivity.
Parameters Involving Compressibility of the Ambient Liquid
Mach inertial forces / isentropic compressibility forces
number M V c
in a perfect gas
The Mach number is important to very high-speed cavitating flows, such as
might be the case for a gun-launched supercavitating projectile. (See for
example Fig.1.2.2.) As for an object traveling at high speed in air, the drag of
such a projectile depends on the value of the Mach number, and increases
dramatically near values of M=1 at which point a strong normal shock
develops at the nose of the object. Many of the same compressible flow
characteristics that have been observed at high speeds in air may be expected
in water. There are some differences however, including the fact that the sound
speed in liquids such as water is much higher than that in air (approximately
1500 m/s for water versus approximately 340 m/s in air under typical
conditions). Also, the equation of state in water (Tait’s equation) differs from
that in air (often approximated as the ideal gas law). Finally, since cavitating
flows involve a dynamic condition on the cavity boundary, the differences in
the pressure field between high-Mach-number flows in water and air are
associated with effects on the shape of the cavity.
Cauchy Ca ρV 2 K inertial forces / compressibility forces
number
The Cauchy number may be considered to be a generalization of the square of
the Mach number for non-isentropic flows.
Parameters Involving Cavity Stability and Dynamics
Silberman- vaporous cavitation number / ventilated
Si σv σ
Song- cavitation number
Logvinovich In addition to its theoretical importance in predicting the boundary between
number stable and auto-oscillating flows past nominally axisymmetric supercavities
(where it is actually a proxy for the Euler number), this quantity appears in
Logvinovich’ semi-empirical law for the gas loss from supercavities in the re-
entrant jet regime.
Campbell- Fr σ (inertial forces · static pressure forces) /
Cm
Hilbert (gravitational forces · dynamic pressure forces)
number This parameter is important in an empirical formula that provides an estimate
of the boundary between the re-entrant jet and twin vortex regimes of
nominally axisymmetric supercavitating flow.
Buivol (inertial forces · static pressure forces) /
Bu Fr σ σ
number (gravitational forces · dynamic pressure forces)
This parameter is important in a simple albeit first-principles formula that
19
provides an estimate of the boundary between the re-entrant jet and twin
vortex regimes of nominally axisymmetric supercavitating flow.
Atwood A Δρ 2 ρ density difference / average density
number
Theoretically the Atwood number is of theoretical importance to the growth of
local Rayleigh-Taylor instabilities on the upper boundary of long cavities
(especially sheet cavities) at low Froude numbers due to the density inversion
between the external liquid and the gas or vapor within the cavity.
Strouhal ω Length / traversed path at characteristic time, or
number St or vortex shedding frequency / characteristic
VT V lengths traveled per unit time
The Strouhal number is important to the description of oscillating flow
mechanisms especially in nominally single-phase flow. For large Strouhal
numbers (of order 1), viscosity dominates the flow, resulting in a collective
oscillation of the fluid. For low Strouhal numbers (of order 10−4 and below),
the high-speed, quasi-steady state portion of the movement dominates the
oscillation. Oscillation at intermediate Strouhal numbers is characterized by
the buildup and rapidly subsequent shedding of vortices. At nominally
moderate cavitation numbers, the low pressures that develop in the vortex
cores can lead to local cavitation. The oscillation of developed cavities tends to
be governed by other parameters, such as the Silberman-Song-Logvinovich
number described above.
Roshko Ro St Re = ω 2 ν density difference / average density
number
The Roshko number is a dimensionless number describing oscillating flow
mechanisms, where St is the Strouhal number; Re is the Reynolds number; ω
is the frequency of vortex shedding; l is the characteristic length (for example
hydraulic diameter) and ν is the kinematic viscosity of the fluid.
Parameters Characterizing the Importance of Viscosity
Reynolds Re V ν inertial forces / viscous forces
number
Since in many applications cavitation is associated with high flow speed in
fluids characterized by low values of the kinematic viscosity, the Reynolds
numbers of interest for many flows is very high, and the flow is dominated by
inertial rather than viscous effects. Consequently, the forces acting on
cavitating bodies are often dominated by pressure forces, rather than friction
drag, a feature that is sometimes put to use in reducing the drag of a body
designed to travel at high speeds. The characteristic length scale, , is usually
taken as one of several quantities: the cavitator base diameter or transverse
dimension, the cavitator length, or (occasionally) the maximum diameter or
transverse dimension or the length of a body enveloped in a supercavity. When
considering transition of laminar to turbulent flow, the Reynolds number based
on distance from the leading edge of the cavitator is important.
Galileo Ga g 3 ν2 gravitational forces / viscous forces
number
Similar to the case of the Reynolds number, the Galileo number is typically
very high for the cavitating flows of interest, and the flow is dominated by
20
gravitational rather than viscous effects. This is not as clear for small-scale
structures in cavity flows, such as those that appear near cavity closure, or in
the case of wetting of a cavity enveloped body by a re-entrant jet.
Parameters Characterizing the Importance of Surface Tension
Weber We ρV 2 γ inertial forces / surface tension forces
number
The Weber number is one of several parameters characterizing the relative
importance of surface tension to the behavior of the flow, and is the most
important such parameter for flows that are dominated by inertial effects, such
as is the case for most cavitating flows of interest in this book. High values
indicate that surface tension is relatively unimportant, as is the case for most
regions of high-speed cavitating flow in liquids such as water that are
characterized by low surface tension coefficients. However, since the surface
tension itself depends not only on that coefficient, but also on the local
curvature of the free surface separating the liquid from the gaseous or vapor
phase of the fluid, surface tension effects can become very important when the
curvature is high. Such is the case, for example, at the downstream end of a
supercavity near the formation of the re-entrant jet or twin-vortex closure
system, where the free surface takes on significant curvature. Free surface
effects are also important to the break-up of the re-entrant jet within the cavity,
to the extent of wetting of a cavity enveloped body by the re-entrant jet, and to
the shape and evolution of droplets that form within the cavity and bubbles that
develop downstream of cavity closure. Finally, for cases of water entry and
planing on a cavity boundary, the spray jet that forms can be strongly affected
by surface tension.
Bond or Bo or Eo Δρ g 2 γ body forces / surface tension
Eötvös
number Here the acceleration associated with the body force has been selected as
gravitational acceleration, although more general cases can be considered. For
cavitating flows at high Bond (Eötvös) numbers, surface tension effects are of
relatively small importance. However, for cavitation at very small scales, or
for portions of the flow that involve small dimensions (such as the re-entrant
jet at the closure of cavities at low cavitation numbers, or the vortex tubes that
appear in the twin vortex regime), the Bond number is low, and surface tension
effects can be very important. Low Bond numbers should be considered when
studying the break-up of the larger structures in the flow, such as the formation
of the mixing region when a re-entrant jet interacts with the cavity boundary or
the body.
Capillary Cl μV γ viscous forces / surface tension
number
Since the flow speeds of interest are typically great, whereas the molecular
viscosity is typically small, the value of the capillary number ranges in value
depending on the local value of the surface tension.
Laplace γρ Re 2 surface tension forces / viscous momentum
(Suratman) La (Su) =
μ 2
We transport
number
For most fluid flows, this parameter is especially important when the
21
momentum transport is that due to dissipation. For most cavitating flows of

interest in this book, the Reynolds number is very large and the Weber number
is large except in regions of high curvature of the free surface separating the
cavity from the ambient liquid, as described above, where it is reduced. The
fact that the Reynolds number is squared in this formula suggests that surface
tension forces are far more important than viscous momentum transport
everywhere in the flow, especially where the curvature of the cavity boundary
is large. This represents additional evidence that viscous effects are among the
least important of all the physical effects considered in the study of cavitation.
Ohnesorge μ We
= viscous forces / square root of the product of
number Oh
ργ Re inertia and surface tension forces
Since the Laplace number is very large for most cavitating flows of interest in
this book, the square root of its reciprocal is very small, further indication that
viscous effects can be neglected for all but the most detailed computations.
Parameters Important to the Multiphase Nature of a Flow System
Volumetric Vgas
void fraction α volumetric gas content / total fluid volume
Vfluid
The void fraction defines the volume fraction of a multiphase fluid that is
occupied by the gas phase. A gas-filled supercavity is characterized by a very
high void fraction, whereas the void fraction of the bubbly flow near its
closure region (or for flows such as those near the ocean surface away from the
shoreline) is much lower. When the gas content of a multiphase fluid is in the
form of bubbles, the sound speed of the mixture can be much lower than either
the pure liquid or the pure gas components of which it is comprised. A chapter
in this book is devoted to high-speed motion of an object through a bubbly
liquid, for which the void fraction is of fundamental importance. The void
fraction of the ambient fluid is also important to cavitation inception.
Fig.1.2.2 provides examples of cavitating flows involving extreme values of two of

the key parameters presented in Table 1.2.1, specifically, a very low value of the cavitation
number and over a range of Mach numbers from moderate values up to a high value. The
projectiles presented in the figures were launched from a submerged gun at very high speeds
[Kirschner, 2001; see also Kirschner, 1997; Hrubes, 2001]. The tip of each projectile was
blunted to form a very small cavitator at the nose, thus generating a cavity that enveloped the
remainder of the body. Speeds ranged from approximately 900m/s through supersonic speed
in water. The top speed recorded for the projectiles shown was approximately 1549m/s,
corresponding to the three frames presented in the left-hand part of the figure. At these
speeds and at the rather shallow test depths, the cavitation number was extremely low – with
values ranging between 1e-4 and 3e-4. The resulting cavity was extremely long and slender,
extending for many, many body lengths downstream of the projectile. Under the
environmental conditions of these tests, the sound speed in water ranged from approximately
1482m/s to 1496m/s, such that the Mach number ranged from a moderate subsonic value of
approximately 0.61 at projectile velocities to a supersonic value of approximately 1.03-1.05
22
for the supersonic tests. In supersonic flight, the projectile and its cavity generated a bow
shock wave, which is clearly seen in the upper left-hand image of Fig.1.2.2. The behavior of
cavitating flows at high Mach number is the subject of ongoing research. (See for example
[Poruchikov, 1973; Kuznetsov & Manevich, 1979; Nishiyama & Khan, 1981; Terentiev &
Chechnev, 1985; Filippov, 1987; Serebryakov, 1992;2007; Savchenko et al., 1993; Vasin,
1996; Varghese et al., 1997; Zigangareieva & Kiselev, 1998; Saurel et al., 1999; Vlasenko,
2002; Serebryakov & Schnerr, 2003; 2004; Serebryakov et al., 2008].)
Fram e 1
Shock
P rojectile Fram e 2
base +100 µs
Shock
Glassy
Fram e 3 cavity Base
+200 µs
Tail-slap scar Tip
Shock
Fig.1.2.2. Free-flying projectiles launched at very high speed from a submerged gun
[Kirschner, 2001; Hrubes, 2001]. The images on the left are three successive frames of a
high-speed film that illustrate rather extreme values of two of the key parameters discussed
in Table 1.2.1, namely a very low cavitation number ~1.2e-4 and a very high Mach number
~1.03. The projectile speed was approximately ~1549 m/s. The high-speed shadow-graph
film was made at a rate of 10,000 frames/s. The shock wave ahead of the projectile is
clearly evident in frame 1. The images on the right were captured using 35 mm still frame
cameras, and show the details of the projectiles flying within their cavities, including
interaction of the projectile base with the cavity boundary.
It should also be noted that, because each of the relevant parameters listed in Table
1.2.1 is dimensionless, various alternative definitions are equally valid, and the choice
becomes a matter of convention. For example, unless otherwise specified, the Froude
number used in this book is defined as Fr=V2 /gl, where V is the characteristic velocity; g is
the gravitational acceleration; and l is the characteristic dimension of the problem; however,
another common definition for the Froude number is the square root of this quantity.
Of particular interest is the magnitude of each parameter relative to quantities of
order unity, since it is frequently straightforward to anticipate the key physical effects by
23
posing the problem at hand in such a way that parameters that are much greater or less than
unity may be neglected. For example, very high-speed supercavitating horizontal flows
involving moderately sized cavitators may be characterized by Froude numbers that are
much greater than unity, indicating that the effects of gravity are negligible. Under such
conditions, it is possible to predict the cavity shape and the forces acting on the cavitator to
very acceptable accuracy by neglecting gravitational acceleration. On the other hand,
problems of ship drag reduction by a partially ventilated cavitaty depend substantially on Fr,
see [Butuzov et al, 1990; Thill, 2010; Amromin et al., 2010] for more detail.
Finally, it is traditional in fluid mechanics to reduce the basic equations to their
dimensionless forms by scaling all densities by the characteristic density (for example, the
density of the ambient fluid at upstream infinity), all lengths by a characteristic length (such
as cavitator base diameter), and all velocities by a characteristic velocity (such as the free
stream speed). If calculations are carried out in dimensional units, this is easily accomplished
by setting each of these characteristic quantities to values of unity, upon which all of the
dimensionless quantities reduce to simple forms. For example, the Froude number as defined
in Table 1.2.1 becomes Fr = g −1 , where g is the gravitational acceleration, and the Reynolds
number becomes ν −1 , where ν is the fluid kinematic viscosity.
Sec. 1.3. Main Assumptions in the Theory of Cavitating Flow

1.3.1. Pressure inside a Cavity
Cavitating flow of a real liquid can be modeled neither theoretically nor experimentally
taking into account all the above-mentioned quantities. Therefore depending on the intended
task, flow conditions, and many other factors it is possible to ignore some and use only those
parameters which are most essential to the given process. The cavitation number is usually
the most essential parameter in cavitating flow. Natural cavitation in a real fluid is possible
when pressure in the liquid is reduced to vapor pressure due to a speed increase. However,
the same cavitation number can be obtained by supplying air to the cavity, thus increasing
the internal pressure. Thus, by supplying air to the cavity one can study the effect of
cavitation number on hydrodynamic characteristics of a body without increasing the flow
speed. As mentioned above, such cavitation is named ventilated cavitation. However
observations show that natural cavities in a real fluid and ventilated cavitation are different
things. Fig.1.3.1 [Batchelor, 1970] presents artificial and vaporous cavities.
As seen in Fig.1.3.1a, the surface of a ventilated cavity is smooth and clear, while
the vaporous cavity in Fig.1.3.1b is opaque. Fig.1.3.1c made at a small exposure time of
about 10–4 s can explain this opaqueness; the real cavity is filled with a mix of water and gas
with an indistinguishable cavity surface. The pressure measured in the cavity has a non-
stationary character and changes slightly along the cavity, with its average value remaining
constant. In essence, the process of cavitation is unsteady, at least in some parts of the flow.
Although in Fig.1.3.1c, most of the cavity boundary is unsteady at small scale, the most
practically important characteristics of the cavity can be often described by time-averaged
values. In this sense, the quasi-steady description of the unsteady behavior of cavitation is
24
somewhat analogous to the
Reynolds averaging of turbulent
flow.
Fig.1.3.1.
Cavities behind a disc at
σ =0.19 [Batchelor, 1970]: (a)
cavity; (b) vapor cavity; (c)
vapor cavity at an exposition of
10−4 sec.
Plotted in Fig.1.3.2 is
the pressure distribution on the
surface of a cylinder in a
channel over conditions ranging
from fully-wetted flow through various levels of cavitation. One can see that, over a range of
values of the cavitation number (here denoted using Epshteyn’s notation, λ) the pressure on
the wetted part of the boundary is almost the same as that on a cylinder flowing
continuously, and then it becomes constant though slightly greater than the pressure obtained
from the given cavitation number. This discrepancy may be explained by an inner flow of
vapor-fluid mix. To simplify the problem,
assume that the pressure has a constant
character and does not change along the
cavity. The same is assumed for an artificial
or ventilated cavity. Then, from the
mathematical viewpoint, there are no
distinctions between a vaporous and
ventilated cavity that both flows can be
considered within the framework of one
model.
Fig.1.3.2.
Pressure distribution along the surface of a
cylinder in a channel at cavitation number,
in this case denoted as λ = p1 − pd ρ v12 / 2 ,
where (following Epshteyn’s notation) pd is
the pressure in the cavity, p1 and ν1 are the
pressure and speed far upstream in the
channel, p2 is the pressure on the cylinder,
after Epshteyn [1959].
1.3.2. Flow at Cavity Closure

Most problematic is a description of the
flow at the downstream end of a cavity
where the real fluid flow is usually highly unsteady. The traditional flow model, however, is
the steady flow of an ideal fluid, and common sense suggests that the cavity should be
convex and hence cannot be closed [Sedov, 1950]. In overcoming this complexity, many
25
authors have offered various simplifications concerning flows past a cavity. Of the many
models proposed by various researchers, one consists of a circuit with a re-entrant jet, which
reflects most realistically the physical process; in experiments this has been observed as a jet
which breaks up into a spray inside the cavity, then exits the cavity into the main stream in
the form of droplets of liquid that form on the cavity boundary. This cavitating model was
offered independently by Efros [1946], Kreisel [1946], and Gilbarg & Rock [1946]. It should
be noted that, considering the unsteady behavior of cavity closure under most conditions,
this is a rather simplistic model. Moreover, reflecting a real flow, it cannot be applied for
describing asymmetrical cavitating flow because the direction of the jet is unknown and
there are no conditions to define it. The same uncertainty arises in cavitating flow with a
trailing artificial plate (or Riabouchinsky wall) discussed in [Riabouchinsky, 1920; Cisotti,
1921; and Weinig, 1932]; the plate can be oriented in any arbitrary direction. It seems that
the only model of ideal steady cavitating flow which can be based on an exact mathematical
assumption is the model with single spirals proposed in [Tulin, 1964] and formulated
mathematically in [Terentiev, 1976]. This model will be most commonly used in the book.
Unlike a real flow, the cavity boundary terminates in a single spiral; therefore this model
should be named ideal cavitating flow. Some other cavitating models will be discussed in
Chapter 2.
1.3.3. Detachment Point
One interesting problem is that of detachment of a cavity from the surface of a body. If the
body has sharp edges then boundary layer separation generating relatively large vortices
takes place behind them. Cavitating bubbles are formed first in the cores of these vortices
and then cavitation grows to the edges. Thus, the cavity finally detaches directly from the
body at its sharp edge. However, on a body with a smooth boundary, the cavity detachment
point is difficult to distinguish.
An equality condition for smooth body-cavity boundaries was offered in [Brillouin,
1911] to determine the cavity detachment point from the body, but the physical reality is
more complex. Due to viscosity, laminar separation in real fluids can precede actual cavity
detachment. Nevertheless, comparison of theoretical calculations with experimental data
shows that – at least in laminar flow – the separation point nearly coincides with the cavity
detachment point. Thus, in real flow in which cavitation occurs along a smooth portion of
the body boundary, a balance exists among three main processes: cavitation (natural or
artificial, the thermo-dynamical process in which the gas and liquid phases of the flow
compete for dominance); flow separation (either laminar or turbulent, a process that
determines the local direction – and hence the topology – of the flow); and capillary tension
(the process that determines the balance of forces along the boundary that separates the gas
from the liquid in the flow) [Amromin, 2007]. A more detailed physical description of these
competing processes and a method for predicting cavity inception along a smooth boundary
are presented in Chapter 11.
Sec. 1.4. Summary

The intent of this book is to provide the reader with insight into two perspectives of modern
hydrodynamics: the broad physics of cavitation, and successful approaches to modeling
cavitation processes based on sound mathematical and numerical techniques. There are
many aspects of cavitation that are not probed in detail in this text, for example, those
26
processes that link the macro- and micro-scales of cavitation with those at the molecular
scale. There are many valuable references that present such information, for example,
[Brennen, 1995] and those of the many authors cited therein.
Rather, this book attempts to guide the reader toward tractable models that –
although limited to the more human scale – are useful to engineering applications and to
those laboratory experiments that are bounded at the smaller scales by the limits of
continuum mechanics. Through this approach to the presentation of cavitation, the authors
hope not only to provide the reader with a useful set of tools, but also endeavor to capture
the natural beauty of the hydrodynamics of cavitation. It is a topic of study that –
paraphrasing C.-S. Yih – never fails to engage the curious and to delight the mind.
27
Ch. 2. Main Boundary Problems of Cavitating Flow
Ch. 2. Main Boundary-Value Problems of Cavitating Flow

Sec. 2.1. Mathematical Modeling of Cavitating Flow
2.1.1. Fundamental Assumptions of Cavitation Theory
Theoretical investigation consists of two steps: problem statement and solution. While the
second step is mathematical only, the statement involves some credible conditions based on
physical experiments for choosing the fluid model, initial and boundary conditions, and
additional conditions for the correctness of mathematical problems. This is due to very
complicated real hydrodynamic processes whose exact description is hardly possible.
Therefore, the obtained theoretical results reflect the real process approximately. However,
by choosing appropriate assumptions, a sufficiently good agreement of theoretical and
physical data can be achieved. For instance, the fluid viscosity has usually little if any effect
on hydrodynamic forces so that it can be ignored. But for a flow near the body surface or in
the wake, its viscosity should be taken into account. The compressibility of a fluid can be
neglected at speeds much slower than the sonic speed, but should be taken into consideration
for high-speed processes, such as bubble collapce, impacts, etc. If the wing span is mush
greater than the foil chord and is near normal to inflow, the problem can be considered to be
two-dimensional. Incompressible and inviscid fluid is mostly considered here and the flow is
assumed to be vortex-free. The velocity field is determined by a scalar function ϕ ( x, y , z , t )
such that v = ∇ϕ which satisfies the Laplace equation
∂ 2ϕ ∂ 2ϕ ∂ 2ϕ
∇ 2ϕ = + + = 0. (2.1.1)
∂x 2 ∂y 2 ∂z 2
The pressure, velocity and mass force are related by Bernoulli’s equation
∂ϕ v 2 p
+ + + F = C(t) . (2.1.2)
∂t 2 ρ
Turning back to cavitation, it should be noted that the cavitating flow is actually
unsteady even for a non-moving body, and a cavity has a non-smooth boundary. Such a
process cannot be described theoretically as a whole, but fortunately unsteady effects exert
influence not much on the entire steady flow and the pressure inside cavity is virtually
constant ( p = p0 = const ). This allows consideration of a flow under assumption of steady
state and smoothness of the cavity boundary. Viscosity affects cavitation rather mildly due to
presence of free cavity boundaries, so that it can be ignored in many cases. Instead of
dimensional pressure inside cavity p0 and pressure at infinity p∞ , a non-dimensional
cavitation number is considered as σ = ( p∞ − p0 ) / p∞ . (2.1.3)
For the inviscid potential flow, the cavitation number is expressed as
σ = v0 2 − 1 , (2.1.4)
where v0 is the ratio of the speed at cavity boundary to the inflow speed. The inflow speed,
fluid density, and characterized length – each of them – are assumed to be equal to unity.
2.1.2. A General Problem of Cavitating Flow
A problem of cavitating flow under the above-mentioned assumption is simulated by a
mathematical problem for a harmonic function satisfying the Laplace equation in flow
28
domain D, kinematical condition ( vn = ∂ϕ / ∂n = 0 ) at the total domain boundary and

additional dynamic condition ( v = v0 = const ) at an unknown free boundary. This problem
has been formulated long ago by Helmholtz [1868] in a form, of course, unrelated to
cavitation. Helmholtz’s problem should be supplemented with the Brillouin [1911] condition
on minimum pressure in the cavity, i.e. the speed in the flow domain should be less than that
at the cavity boundary: v ≤ v0 . (2.1.5)
Simple observations show that an isolated body in unbounded flow around a closed
boundary stationary cavity is mathematically impossible [Brillouin, 1911]. Indeed, the cavity
boundary is convex due to the minimum pressure inside. If the boundary is closed then the
stagnation point would be on free boundary. But this is impossible because the velocity is v0
at all points of the cavity. This contradiction is usually overcome by formal assumptions
about flow patterns at the cavity end. Numerical calculations and experimental studies
confirm that all cavitation models for small cavitation numbers give almost the same results
that agree with experiment results [Knapp et al., 1970; Terentiev, 1977a; Gurevich, 1979;
Ivanov, 1980]. Hence, selecting a cavitation model is primarily determined by subjective
considerations and the possibility of getting comparatively simple solutions. Some of main
cavitating models will be applied below.
2.1.3. Singularity of Solution
Let us consider first the plane problem of cavitating flow around a body, Fig.2.1.1.
Respective points on the physical z-plane and complex potential w-plane are marked by the
same letters - see Figs.2.1.1 & 2.1.2. The trailing points of cavity boundaries are unknown
but should be different. The planar flow is completely determined by complex potential
w( z ) = ϕ (x, y ) + iψ (x, y ) which is an
analytic function of complex variable
z = x + iy in flow domain D.
Logarithmic function of complex
velocity on the w-plane,
ω ( w) = ln(dw / v0 dz ) , is often used
instead of potential w( z ) . The w-plane
Fig.2.1.1. Schematics of cavitating flow
and the non-cavitating foil are multi-
sheet domains with a horizontal open-
ended slit, Fig.2.1.2; the lengths of the slit faces differ by circulation Γ . Taking horizontal
cut as shown in Fig.2.1.2 one obtains a single-plane domain. Function ω ( w) is also
analytical in the flow domain and satisfies the following boundary conditions:
• on the slit faces γ corresponds to wetted boundary Im ω = −θ (ϕ ) , (2.1.6a)
• on the slit faces γ 1 corresponds to cavity boundary Re ω = 0 , (2.1.6b)
• on the cut faces γ 2 ω + (ϕ ) = ω − (ϕ + Γ) . (2.1.6c)
Then the flow at infinity should be undisturbed, i.e. the additional condition is to satisfy
1
∫
z →∞
dz =
v ∫
0 w→∞
e −ω ( w) dw = 0 . (2.1.6d)
29
For simple reasons, solving the
value problem (2.1.6) cannot be limited at
point C. Indeed, if function ω ( w) is limited
on the slit faces then it is a continuous Fig.2.1.2. The w-plane.
function because of continuity of boundary
conditions. Therefore, the cavity boundaries pass to trailing streamlines continuously. Since
the cavity is open, a semi-infinite wake with congruent boundaries forms behind it and the
flow continuity breaks. In that case, instead of equation (2.1.6) we have
1
∫
z →∞
dz = ∫
v0 w→∞
e −ω ( w) dw = z 2 − z1 . (2.1.7)
Equation (2.1.7) corresponds to the open cavity model proposed by Wu [1962]. In order to
find a desirable solution we should require a singularityat the trailing point of the cavity, i.e.
at point w = 0. The type of singularity is described by the theorem below.
Theorem: Brillouin’s condition (2.1.5) allows for function ω(w) only the following possible
singularity at the trailing point: ω ( w) ≈ −bw−1/ 2 , (2.1.8)
1/2
where b is a positive real coefficient, a branch of the multi-value function w on the real
ϕ axis is chosen such that ϕ1/2 >0 at ϕ >0.
Proof. According to Brillouin’s condition (2.1.5), inequality ln v ≤ ln v0 has to be satisfied,
i.e. the domain of variation of function ω = ln v − iθ can be the exterior of the half circle of
infinite radius while the potential changes in the whole disk of infinitesimal radius. Hence
singularity (2.1.8) is valid.
This singularity yields an additional unknown parameter, b, which allows to satisfy
the closure condition (2.1.6d).
Note 1. The proof includes an infinitesimal neighborhood at w = 0 so that the Brillouin
condition can be valid in the certain infinitesimal neighborhood at the cavity end only.
Note 2. The theorem is also valid for partial cavitating flow. In this case the cavity ends at
the solid boundary, thus the zero streamline corresponds to the cavity boundary at ϕ < 0, or
solid boundary at ϕ > 0, i.e. the infinitesimal neighborhood at point w = 0 is a half-disk
while the conformed neighborhood of function ω is a quarter-circle of infinite radius.
Singularity (2.1.8) was first proposed by Tulin [1965] as a possible cavitating model
with single spirals. Larock & Street [1965] applied this model to calculate the cavitating
flow around a plate without circulation. The last assumption makes it impossible to satisfy
closure condition (2.1.6d) and, furthermore, an infinite wake stretches downstream of the
cavity. An exact analytical solution of cavitating flow around a flat plate was derived
without additional assumptions in [Terentiev, 1976] where the theorem was proved. Below
in the following singularity (2.1.8) is used as the main cavitation model.
2.1.4. Streamlines near Cavity End
An asymptotic expression of the derivative of function z(w) is obtained from equation (2.1.8)
dz 1 b / w
as = e . (2.1.9)
dw v0
If potential ϕ approaches zero from right along the streamline (ψ = 0 ), then
dx 1 b / ϕ dy
≈ e → ∞ and → 0.
dϕ v0 dϕ
30
Since differential dϕ is negative in this case, differential dx should be negative as

well, and then the separating streamline behind a cavity reaches infinity to the left directions
( x → −∞ ); but its ordinate approaches a constant value.
If potential ϕ approaches zero from the right along the upper bank of the cut (upper
cavity boundary) then ω = ib / −ϕ → i∞ , i.e. tangent angle θ = −b / −ϕ → −∞ as ϕ → −0.
Since the arc element of the cavity boundary is equal to ds = dϕ / v0 then the radius of
1 dϕ 2b 2 −3
curvature of the cavity’s upper boundary asymptotically approaches R = ≈ θ .
v0 dθ v0
Hence, the upper boundary ends as a fast convergent single spiral. The lower bounder of the
cavity ends as a fast convergent single spiral with the opposite orientation.
Consider now a next streamline closed to the upper boundary (ψ is a positive small
constant). Substituting w = ϕ + iψ into (2.1.8), one can find the
tangent angle in the form
b sin α / 2 ⎧arctan(ψ / ϕ ), if ϕ ≥ 0,
θ =− , where α = ⎨
4
ϕ 2 +ψ 2 ⎩π + arctan(ψ / ϕ ), if ϕ < 0.
dθ b ψ cos α / 2 + ϕ sin α / 2
Hence = .
dϕ 2
( )
3
4 ϕ 2 +ψ 2
From this it follows that the derivative vanishes at

ϕ0 = −ψ / 3 , positive at ϕ > ϕ0 and negative at ϕ < ϕ0 .
Fig.2.1.3. Streamlines
Therefore, the tangent angle decreases first to a certain value and at the cavity end.
then increases to zero.
It is simple to imagine the streamlines at the cavity end at
the multi-sheet plane, Fig.2.1.3. If two spiral centers are connected with a cut and then a cut
side is laminated with another side of different sheets, then a streamline on the sheet can go
over the cut to the other sheet and then to the next one and return back. The cavity boundary
ends on infinite sheet so that the steady flow at the cavity end occurs at the multi-sheet
plane.
2.1.5. Detachment Point on Foil
Properties of currents with free bounders have been considered mostly in the last century and
are well described in [Birkhoff & Zarantonello, 1957; Logvinovich, 1969; Gurevich, 1979].
Therefore there is no need to repeat them again. Here we shall stop only on a detachment
point of a stream on a body surface. Without losing generality one can assume that the origin
on the z-plane coincides with the detachment point, and complex potential w is zero at that
point. Besides, the value of speed on the free boundary is the unity. Consider first continuous
flow along a corner of angle πμ, the conformal transformation from the z-plane to the w-
plane in a small neighborhood of this point, Fig.2.1.4, has dominant term as z=Awμ, hence
the complex speed dw / dz = w1− μ / Aμ . Thus, if the angle is greater than the flat angle, μ > 1 ,
the speed approaches infinity and the pressure should theoretically be indefinitely negative,
what contradicts cavitation conditions (2.1.5). Thus this point should be a detachment point,
Fig.2.1.5a. At such a flow the detachment point is given and conditions (2.1.6) are sufficient
for the solution of a value problem.
31
Let the tangent angle from the left side be equal to πν ; the free boundary attaches to
the foil from the right side at point O. If the body's wetted boundary is straight line AO, then
the small neighborhood of logarithmic
function ω = ln v − iθ at detachment point
bounds by straight lines θ = θ 0 = −πν and
ln v = ln v0 = 0 which form a right angle at
the detachment point, Fig.2.1.5c. Then
the dominant term is ω − ω0 = Bw1/ 2 so
that the tangent angle at the free boundary
is equal to angle θ 0 at the detachment.
Let's consider the derivative of
logarithmic function of complex velocity. Fig.2.1.5. Small neighborhoods: flow with
Since the stream function on a boundary is free boundary (a), w-plane (b), ω -plane.
constant, hence the value of speed is equal
to v = dϕ / ds where s is the curvilinear abscissa. Derivative of the logarithmic function
d ω 1 dv dθ B −1/ 2
with respect to the ϕ-axis can be written as = −i = ϕ , (2.1.10)
dϕ v 2 ds ds 2
whence the acceleration, vdv/ds, at the body's straight boundary and the curvature of free
boundary dθ / ds approach infinity as the curvilinear
abscissa approaches the detachment point.
Consider now a boundary of a body given by
continuous double-differentiable functions x = x( s )
and y = y ( s ) , Fig.2.1.6a,c. In that case, instead of a
straight boundary the angle is a function of s, and the
boundary of a small neighborhood form the angle is
0 < π v ≤ π , so that the dominant term is
ω − ω0 = Bwν , and
d ω 1 dv dθ B − (1−ν )
= −i = ϕ . (2.1.11)
dϕ v 2 ds ds 2
Fig.2.1.6. Small neighborhoods: That is also right for this case if ν ≠ 1 , e.g. for
fixed detachment (a), ω -plane (b), curvilinear arc AO, but the cavity boundary intersects
smooth detachment(c), ω -plane(d). the body with closed surface. Therefore, the value of
ν should be equal to unity, and from equation
1 dv
(2.1.11) the following condition is valid: − iκ b = −iκ c , (2.1.12)
v 2 ds
where κb and κc are the curvatures of the body and free boundary, respectively. Hence,
velocity function v(s) has an extreme at the smooth detachment point, and curvature κ(s) is a
continuous function. Also, the condition for detachment point is
dv / ds = 0 . (2.1.13)
These results have been obtained first by Villat [1915] and confirmed in other works.
Finally, it should be noted that equations (2.1.10) – (2.1.11) are valid under certain
assumptions about the boundaries of the body and attachment; theoretically, the attachment
32
can have another type, e.g. as singularity (2.1.8). Therefore, in a certain sense, condition
(2.1.13) can be considered as an assumption about the detachment. Besides, the conditions of
detachment can essentially differ under the assumption of surface tension or viscosity.
Namely, the detachment point in reality should be located a bit farther along the stream than
the point determined by equation (2.1.13). Nevertheless, condition (2.1.13) is simple to
apply and gives sufficiently good results consistent with experiments, though condition
(2.1.5) may be used directly by computing (see Sec. 9.4).
2.1.5. Resultant Force and Torque
Pressure acts on an infinitesimal arc of the boundary as a force that can be presented in the
complex form: dX + idY = −i ( p − p0 )dz . Since the stream function on a steady boundary
satisfies condition ψ = const then dw = dw and v 2 dz = ( dw / dz ) dw . Hence
iρ ⎛ 2 dw ⎞
dX + idY = − ⎜ v0 dz − dw ⎟ . (2.1.14)
2⎝ dz ⎠
Integrating with respect to curvilinear abscissa, one can calculate the force: X + iY . Let's
consider an integral along any closed contour being equal to zero because of an analytical
property of the right-hand term of (2.1.14). Deform this closed contour up to an infinite
radius circle, which describes the foil with a cavity, and into boundary of the foil and the
iρ ⎛ ⎞
cavity. The first integral along the circle is − ⎜ ∫ v0 2 dz − ∫ ( dw dz ) dz ⎟ = i ρ v∞ Γ ,
2
2 ⎝ z =∞ z =∞ ⎠
i ρ v0 2
the second one is X + iY − ( z2 − z1 ) ,
2
where z1 and z2 are the lower and upper centers of spirals; Γ is the circulation.
Equating these two expressions, one can obtain the complex resultant force in the form
i ρ v0 2
X + iY = i ρ v∞ Γ − ( z2 − z1 ) . (2.1.15)
2
Torque around the origin is calculated from differential
dM = xdY − ydX = − Re ⎡⎣iz ( dX + idY ) ⎤⎦ .
Substituting expression (2.1.14) into this equation yields the differential of torque
ρ⎛v 2 ⎛ dw ⎞ ⎞
dM = − ⎜ 0 d ( z z ) − Re ⎜ z ⎟ dw ⎟⎟ . (2.1.16)
2 ⎜⎝ 2 ⎝ dz ⎠ ⎠
Let the x-axis be determined as the direction of inflow speed when the fluid remains
on the left. Then components X and Y coincide with the drag, D, and the lift, L, of a foil.
Sec. 2.2. Simple Examples of Cavitating Flows

2.2.1. Flow along a Downstream-Facing Step
Consider the simplest problem of cavitating flow along a downstream-facing step,
Fig.2.2.1a. The defining domain of the complex potential is in the upper half-plane. Without
losing generality, three points A, B and O can be given as depicted on Fig.2.2.1b, and the
speed at the cavity boundary is assumed to be equal to the unity. Then the speed at infinity is
33
v∞ = 1 / 1 + σ where σ is a given cavitation number.

Under these assumptions all of linear parameters as step
height H, cavity length L, and distance h between the spiral
center and the bottom should be calculated.
Function ω = ln dw dz on the w-plane satisfies the
mixed boundaries on the ϕ -axis:
• within interval (−1<ϕ <0) conformed to the cavity
boundary, Re ω = 0 , (2.2.1a)
• beyond this interval conformed to step bottom,
Fig.2.2.1. The sketch of
Im ω = 0 on (− ∞ < ϕ < −1) and (0 < ϕ < ∞ ) , (2.2.1b)
flow (a), the w-plane (b).
• at infinity, Re ω = ln v∞ = −0.5 ln(1 + σ ) . (2.2.1c)
The solution of the mixed value problem with singularity (2.1.8) is
w +1
ω ( w) = A , (2.2.2)
w
where parameter A is determined from condition (2.2.1c)
A = − ln 1 + σ . (2.2.3)
Transformation from the w-plane to the z-plane can be found from differential
dz
equation = e −ω (w) . (2.2.4)
dw
The step height is calculated integrating the last equation along the circle of an
infinite radius as residue of function at infinity
1 ⎛ ⎞ π
H = Im⎜ ∫ e −ω ( w) dw ⎟ = 1 + σ ln 1 + σ . (2.2.5)
2 ⎜⎝ w→∞ ⎟ 2
⎠
Integrating equation (2.2.4) within interval (-1, 0) one can obtain cavity length L and
distance h. Then all geometric parameters are expressed in as dimensionless ratios to H. The
cavity boundary is found by equations
ϕ
1 ⎛ 1 + s ⎞⎟
⎜
H −∫1 ⎜⎝
x(ϕ ) = cos ln 1 + σ ds,
− s ⎟⎠
(2.2.6)
ϕ
−1 ⎛ 1 + s ⎞
sin ⎜ ln 1 + σ ⎟ds .
H −∫1 ⎜⎝
y (ϕ ) =
− s ⎟⎠
Namely, dimensionless cavity length L and distance h are calculated by
L = x(0), h = 1 – y(0). (2.2.7)
Distance h can also be found simply by using the balance of forces. Since a semi-infinitive-
length body has no drag, so that the force obtained from (2.1.11) should be balanced by the
force acting on step H ρhv02 / 2 = ( p∞ − p0 ) H .
Hσ
Hence, h= . (2.2.8)
(1 + σ )
Equations (2.2.7) and (2.2.8) can be used for verification of accuracy by computation.
The calculated data are plotted in Fig.2.2.2. Fig.2.2.2a shows the streamlines for five
values of the stream function (0, 0.003, 0.01, 0.1, 0.2) and σ =1. The cavity boundary (ψ =0)
34
has a finite length ending as a fast converging single spiral, Fig.2.2.2b. Streamlines are self-
crossing near the cavity boundary (dotted line in Fig.2.2.2a). Thus, as noted in Sec. 2.1, the
flow near the cavity end is realized theoretically on a multi-sheet plane [Terentiev, 1981a].
Fig.2.2.2.
Cavitating flow
along a
rectangular
step bottom:
a) streamlines;
b) spiral at the
cavity end.
2.2.2. Flow around a Cavity

The value problem statement is as follows:
• within interval (-1, 1) of the ϕ -axis Re ω = ln v0 = const , (2.2.9a)
• otherwise on the ϕ -axis Im ω = 0 , (2.2.9b)
• at infinity ω = ln v∞ = ln v0 − ln 1 + σ . (2.2.9c)
Dividing the velocity by v0 yields the unit velocity on the cavity, i.e., as before, it can be
assumed without losing generality a unit speed on the cavity boundary, v0 = 1 . Then the
solution with singularity (2.1.8) at the ends of interval (−1, 1) is expressed as
w
ω=− ln 1 + σ . (2.2.10)
w2 − 1
Whence the complex potential as well as the transformation from the w-plane to the z-plane
dz
can be written in a differential form as follows = e −ω (w) (2.2.11)
dw
The cavity boundary is calculated integrating this function along the ϕ -axis within (0, 1):
ϕ
⎛ s ⎞
x(ϕ ) = ∫ cos ⎜ ln 1 + σ ⎟ ds,
⎝ 1− s
2
0 ⎠
ϕ
(2.2.12)
⎛ s ⎞
y(ϕ ) = H − ∫ sin ⎜ ln 1 + σ ⎟ ds.
⎝ 1− s
2
0 ⎠
Cavity length L and width H are determined as the maximum distance between vertical and
horizontal tangents to cavity, respectively. The vertical tangent point at the boundary is
obtained from equation Im ω = π / 2 which can be solved analytically. Another point with
vertical tangent is calculated from equation Im ω = 3π / 2 . As a result, infinite sequence of
π (2.n + 1)
solutions can be found ϕ n = , n = 0, 1, 2, . (2.2.13)
(π (2.n + 1)) 2 + ln 2 (1 + σ )
35
The integrands in equation (2.2.12) are fast oscillating functions in the vicinity of extremity
± 1 . The spiral can preferably be calculated by integrating separately within intervals
( ϕ n , ϕ n+1 ) and then summing with respect to number n. Sequence (2.2.13) is fast
converging, for instance, for cavitation number σ = 1 the first four terms are equal to
ϕ 0 = 0.9765 , ϕ1 = 0.9973 , ϕ 2 = 0.9990 , ϕ 3 = 0.9995 . Because the integrands are bounded,
their integrals decrease as quickly as M (ϕ n +1 − ϕ n ) Summing the integrals allows one
calculating the spiral with any accuracy though this is unimportant for practical applications.
Let the cavity length be defined as the maximal distance between two perpendicular
tangents, then it is expressed as follows L = 2 x(ϕ 0 ) . (2.2.14)
Cavity width H can be found by contour integral of function (2.2.11) along the circular arc
which connects any point on the ϕ -axis ( ϕ > 1 ) and the origin, namely,
π
H = 2 Im ∫ z w (1 + eis )eis ids . (2.2.15)
0
The distance between two centers of spirals at the right (left) end of the cavity, h, is found as
h = 2 y (1) . (2.2.16)
Distance h can also be calculated by integrating along a circle which is locked on the ϕ -axis
π
at point ϕ = 1 − δ h = Im ∫ z w ( 2 + (δ + 1)e is )(δ + 1)e is ids , (2.2.16a)
−π
where δ is positive and sufficiently small. The drag coefficient with respect to cavity width
h
is determined as C DH = (1 + σ ) (2.2.17)
H
2.2.3. Asymptotic Expressions for Small Cavitation Numbers
Consider a small cavitation number. It is clear from equation (2.2.10) that function ω is
infinitesimal of the same order as cavitation number σ . Beyond small neighborhoods of
points w = −1 and w = 1, derivative (2.2.11) can be expanded in power series as
dz ω2
= 1− ω + +… , (2.2.18)
dw 2
where function ω becomes an imaginary expression within interval (-1, 1); its square is a
real-value function with singularity as a single pole at the points. It follows from (2.2.18)
that the linear approach yields z 0 ( w) ≈ 1 − ω . Thus the real part, x0 , is equal to speed
potential ϕ , while ordinate y 0 is obtained from differential equation dy0 / dx0 = − Im ω .
(All asymptotic variables are marked by subscript 0). Function (2.2.10) takes the form
σ z0
ω0 = − . (2.2.19)
2 z 02 − 1
Thus after integrating, the cavity becomes elliptical
σ
y0 = 1 − x02 (2.2.20)
2
with length L0 = 2 and width H0 = σ. The results concur with the linear theory [Tulin, 1953].
36
From equation (2.2.16a) the distance between two spirals is calculated by integrating
equation (2.2.18) along the circle of infinitesimal radius around the point of z 0 = 1 :
ω 02 πσ 2
h0 = Im ∫ 2
dz =
8
. (2.2.21)
z =1
Now, one can find the resulting force by equation (2.1.15) and the drag coefficient with
respect to this distance as C D0 = 2 X 0 / ρv∞2 h0 = 1 . (2.2.22)
Dividing cavity length and width by h0 gives asymptotic expressions obtained by Tulin for
L0 8 H0 4 0
arbitrary thin foils as = C D0 , = Cd . (2.2.23)
h0 πσ 2 h0 πσ
2.2.4. Numerical Results

Figs.2.2.3 and 2.2.4 show a comparison of nonlinear and asymptotic expressions. Plotted in
Fig.2.2.3 are width H and drag coefficient C DH versus cavitation number (lines and dots
correspond to nonlinear and asymptotic expressions, respectively). Cavity shapes for some
cavitation numbers are presented on Fig.2.2.4: the dot-lines correspond to ellipses (2.2.20). It
is seen that the asymptotic results are in a good agreement with the nonlinear solution even
for
cavitation
numbers
greater
than 0.75.
Fig.2.2.3. Comparison nonlinear Fig.2.2.4. Cavity shapes for nonlinear

and asymptotic results. and asymptotic expressions.
Sec. 2.3. Fully Cavitating Flat Plate

Flat plate is the only cavitating body which can be investigated analytically. Consider
cavitating flow around a flat plate, Fig.2.3.1a. The plate extends between endpoints A and B,
from each of them the two legs of the cavity detach. Endpoint A is located at the origin. A
stagnation streamline originating at upstream infinity (point D) intersects the plate at
stagnation point O. There is a second stagnation point C in the region of cavity closure,
bound here by points C1 and C2. The domain of complex potential w=ϕ+iψ is the multi-
sheet plane with an open horizontal cut as shown in Fig.2.3.1b.
A solution of the problem can be obtained in parametric form by conformal mapping
onto the first quadrant of the auxiliary ζ-plane, Fig.2.3.1c. Conformity of the points is shown
in Fig.2.3.1. Stagnation point O maps to coordinate a on the ζ-axis, while point D at
upstream infinity maps to ζ D = c + id . Coordinates a and ζ D = c + id are unknown. Points
O' and C' in the ζ-plane are symmetric to points O and C respectively. The speed on the
37
cavity boundaries as well as fluid density is assumed to be unity. As follows from
Bernoulli’s equation, the speed at infinity in the z-plane is related to the cavitation number
according to v∞ = 1/ 1 + σ .
2.3.1. Complex Potential

The far field of the w-domain corresponds to
an infinitesimal neighborhood about point
ζ D = c + id , so the function w(ζ ) must have
a simple pole and a logarithmic singularity in
the first quadrant; i.e. this function
represents the potential of a doublet and a
concentrated vortex at point ζ D . Let the
leading part of the potential be
A Γ
+ ln(ζ − ζ D ) . Since the
ζ − ζ D 2π i
Fig.2.3.1. Cavitating flow around a plate:
streamline ψ = 0 composed of the plate and
(a) plane z; (b) plane w; (c) plane ζ.
the cavity’s two legs map to the positive
portions of the real and imaginary axes, the
potential is real on the ξ - and η -axes; therefore function w(ζ ) can be analytically
continued through these axes over the whole ζ-plane, so that additional singularities appear
at symmetrical points −ζ D , − ζ D , ζ D with factors − A, − A, A and −Γ, Γ, − Γ , respectively.
Hence, the potential can be presented in the form
2 Aζ 2 Aζ Γ ζ 2 − ζ D2 ⎛ A⎞
w(ζ ) = 2 D 2 + 2 D 2 + ln 2 + 4 Re ⎜ ⎟. (2.3.1)
ζ − ζ D ζ − ζ D 2π i ζ − ζ D 2
⎝ζD ⎠
From this, the derivative of the potential is found as
⎛ Aζ Aζ ⎞ Γζ ⎛ 1 1 ⎞
wζ (ζ ) = −4ζ ⎜ 2 D 2 2 + 2 D 2 2 ⎟ + ⎜ 2 − 2 2 ⎟
. (2.3.2)
⎝ (ζ − ζ D ) (ζ − ζ D ) ⎠ π i ⎝ ζ − ζ D ζ − ζ D ⎠
2
It can be seen from the latter equation that the derivative vanishes at the origin. Moreover, it
should vanish at points a and i as well. These yield two conditions, respectively:
wζ (a) = 0 and wζ (i ) = 0 . (2.3.3)
2.3.2. Derivative of the Potential

Derivative dw / d ζ = wζ can also be easily obtained from analysis of its poles and zeros. It
becomes a real value on the horizontal axis and has a pure imaginary value on the vertical
axis. Also, it has simple zeroes at points ζ = 0, a, i and a pole of second order at point ζ D .
Also, function wζ (ζ ) can be continued analytically to the whole ζ -plane so that simple
zeroes should be added at symmetrical points -a and -i, while poles of second order should
(ζ 2 − ζ D2 ) 2 (ζ 2 − ζ D2 ) 2
be added at points ζ D , − ζ D , − ζ D . Hence the expression wζ is
ζ (ζ 2 − a 2 )(ζ 2 + 1)
analytical on the whole ζ-plane without any singularities. Therefore, to satisfy Schwartz’s
38
ζ (ζ 2 − a 2 )(ζ 2 + 1)
theorem, it must be identically constant. Thus wζ = N . The constant
(ζ 2 − ζ D2 ) 2 (ζ 2 − ζ D2 ) 2
value N is determined by the specified length of the plate. Conversely, constant factor N can
be given, namely it can be supposed, without losing generality, to be unity and the plate
length must then be computed. The transformation from the w-plane to the ζ-plane can be
ζ (ζ 2 − a 2 )(ζ 2 + 1)
written in differential form: wζ = 2 (2.3.4)
(ζ − ζ D2 ) 2 (ζ 2 − ζ D2 ) 2
Separating fraction (2.3.4) into simple fractions and substituting the result into Eq. (2.3.2),
the factors A and Γ may be solved in terms of coordinates in the auxiliary plane, as follows:
[(c + id ) 2 + 1][(c + id ) 2 - a 2 ]
A= , (2.3.5)
64(c + id )c 2 d 2
π
Γ= 3
32c d 3 ( (c 2
+ d 2 ) 2 − a 2 + (a 2 - 1)(d 2 - c 2 ) ) . (2.3.6)
2.3.3. Logarithmic Function of Complex Velocity

Function ω (ζ ) = ln v (ζ ) = ln wζ (ζ ) is purely imaginary on the imaginary η -axis, so that it
can be analytically continued through the η -axis, whereupon the boundary conditions are
applied on the ξ -axis as follows:
⎧α , ξ ∈ [−a, a ];
Im ω = ⎨ (2.3.7)
⎩α − π , ξ ∈ (−∞, − a ) ∪ (a, ∞).
Moreover, function ω (ζ ) has a simple pole at point C, ζ = i. Using the generalized
Schwartz operator (see Appendix 1), the solution is obtain in the form
ζ − a 2bζ
ω = ln + + (α − π )i , (2.3.8)
ζ + a ζ 2 +1
where b is an unknown real factor.
The transformation from the z-plane to the ζ -plane is determined from functions (2.3.4) and
(2.3.8) in differential form dz / dζ = (dz / dw)(dw / dζ = wζ e −ω ) , or
2 bζ
dz ζ (ζ + a) 2 (ζ 2 + 1) − ζ 2 +1
= zζ = ei (π −α ) 2 e . (2.3.9)
dζ (ζ − ζ D2 ) 2 (ζ 2 − ζ D2 ) 2
Functions (2.3.1) or (2.3.4), (2.3.8) and (2.3.9) form a general solution. The four unknown
real parameters a, b, c, and d can be calculated from the flow conditions as follows:
• the given velocity at infinity, ω (ζ D ) = − ln 1 + σ
ζD −a ζ 1
ln + 2b 2 D + (α − π )i + ln(σ + 1) = 0 ; (2.3.10)
ζD + a ζ D +1 2
• and the closed-cavity condition ∫ zζ (ζ )dζ = 0 .

ζD
(2.3.11)
Integral (2.3.11) can be calculated analytically using the method of residues. Since function
(2.3.9) at point ζD has a Taylor series
39
B−2 C−1 d
zζ (ζ ) = + + C0 + , then B−1 = B−2 lim ln zζ (ζ )(ζ − ζ D ) 2 .
(ζ − ζ D ) 2
(ζ − ζ D ) ζ →ζ D dζ
Integral (2.3.11) is equal to 2πiB1, so the closed-cavity condition is

d 2 2ζ 2b(ζ D −1) 4ζ
lim ln zζ (ζ )(ζ − ζ D ) 2 = + 2 D + − 2 D 2 = 0 . (2.3.12)
ζ →ζ D dζ ζ D + a ζ D + 1 (ζ D + 1)
2 2
ζD −ζD
Equations (2.3.10) and (2.3.12) form a closed system of four real equations with four
parameters a, b, and ζD=c+id.
2.3.4. Asymptotic Solution of Equations
It can be shown from equation (2.3.10) that small values of angle of attack α and cavitation
number σ correspond to small values of parameters a−1 and b. Expanding equations (2.3.10)
and (2.3.12) in power series of small parameters a−1 and b, one can obtain an asymptotic
solution for small angle of attack α→0 and cavitation number σ→0 (denoting κ=σ /α):
4 σ 1 1
a= 4 + κ 2 − 2, b = − 4 + κ 2 − 2, c = 4 + κ 2 − 2, d = 4 + κ 2 + 2 . (2.3.13)
σ 8 2 2
A considerable simplification occurs for a plate perpendicular to the free stream,
α=π/2. The flow is symmetric in that case and the symmetric axis in the z-plane transforms
into a circular arc of unit radius in the ζ-plane. Points A and C fall also on that arc, so that
a = 1, c2 +d2=1 and equations (2.3.10) and (2.3.12) transform to two equations of the form
2b (1 + σ )(1 − c)
c 2 + b(1 − c 2 ) = 0, + ln = 0. (2.3.14)
c 1+ c
Hence, for a small cavitation number ( σ << 1 ), one obtains the following asymptotic
σ σ2 σ2 σ3
expansions for c and b: c= − + …,
+… b=− +
(2.3.15)
4 8 16 16
Asymptotic solutions (2.3.13) and (2.3.15) can be used as initial values for the unknown
parameters for numerical solutions of equations (2.3.10) and (2.3.12) and also for finding
asymptotic expressions of hydrodynamic and geometrical characteristics.
2.3.5. Cavity Shape and Flow Characteristics
As noted above, functions (2.3.4), (2.3.8) and (2.3.9) determine the cavitating flow around a
plate of unknown length. The length is calculated by integrating function (2.3.9) along the
whole positiveξ-axis, i.e. within (0, ∞). It is convenient to change variables ξ=u/(1−u) so
that the range of integration transforms to range (0,1). Let us consider the function:
t
s (t , a, b) = ∫ s′(u , a, b)du , (2.3.16)
0
u (1 − u ) ⎡⎣u 2 + (1 − u ) 2 ⎤⎦ [u + a(1 − u ) ]
2 2 bu (1− u )
−
where s′(u , a, b) = 2 2
e . u 2 + (1− u )2
(2.3.17)
⎡⎣u 2 − (c + id ) 2 (1 − u ) 2 ⎤⎦ ⎡⎣u 2 − (c − id ) 2 (1 − u ) 2 ⎤⎦
Then the length of the plate is calculated as l = s (1, a, b) (2.3.18)
Henceforth, all linear parameters are made dimensionless with respect to the plate length.
40
The cavity shape, length and maximum width can be obtained by integrating Eq.
(2.3.9) along the η -axis; the cavity's lower and upper boundaries correspond to ranges (0, 1)
and (1, ∞) respectively. For the latter range, it is convenient to change variables η = 1/u.
Dimensionless length of cavity Lc is determined as the abscissa of the center of
spirals either C1 or C2 divided by plate length (2.3.12). The difference of these centers is
very small. For example, for α = 300 and σ = 0.5 the abscissa of point C1 is x1 = 3.931
while that of point C2 is x2 = 3.935 . Dimensionless maximum width of cavity H c is
determined as the distance between horizontal tangents to both free boundaries divided by
plate length l. The points of contact with the horizontal tangents are calculated numerically
from β (η ) = − Im ω (iη ) . (2.3.19)
Using complex potential (2.3.1) the streamlines and equipotential lines can be
calculated, first in the ζ -plane using complex potential (2.3.1), and then in the z-plane. The
stagnation streamline upstream of the plate joins stagnation point O, which has coordinate a
on the real axis in the auxiliary plane, with the point at upstream infinity, ζ D = c + id .
Given an ordinate η in the range (0, d), corresponding abscissa ξ on the stagnation
streamline is calculated from equation ψ (ξ ,η ) = Im w(ξ + iη) . Then coordinate z is
determined by integrating function (2.3.9) along the straight-line segment joining points a
and ζ = ξ + iη . Determination of the streamlines and equipotential lines is reduced mainly
to numerical solution of equations ψ (ξ ,η ) = constant and ϕ (ξ ,η ) = constant . Since these
problems are mostly theoretical in motivation, hereinafter we will not pursue them.
Fig.2.3.2.
Cavitating flow
around the inclined
flat plate:
(a) the plate, cavity,
and mesh of
streamlines and
equipotential lines;
(b) spiral cavity
streamlines at
cavity closure;
(c) streamlines in
the near-field of one
of the spirals at
cavity closure.
The cavity, the

mesh of stream and
equipotential lines
drawn in incremental dimensionless values of a half are shown in Fig.2.3.2a. A near-field
view of the two spirals at the cavity end are shown in Fig.2.3.2b. Some streamlines very
41
close to the cavity closure region are plotted in Fig.2.3.2c. At the first glance, it appears that
some streamlines intersect others, which is impossible. But the flow at cavity closure must
be considered in the proper context of its existence on a multi-sheet surface, and then no
streamlines intersect.
Cavity shapes for angle of attack α = 10° and various cavitation numbers are shown
in Fig.2.3.3a. Fig.2.3.3b shows the single spirals at the cavity end. The upper cavity
boundaries close to the plate leading edge for different cavitation number values are shown
in Fig.2.3.3c. It can be seen that the cavity boundary bulges slightly as cavitation number
increases.
Fig.2.3.3.
Cavity geometry for a flow
past an inclined flat plate at
angle of attack α=10°:
(a) cavity geometry for
different values of
cavitation number;
(b) two single spirals at
cavity closure for cavitation
number σ=0.2;
(c) cavity geometry near the
plate leading edge for
various values of cavitation
number.
2.3.6. Hydrodynamic Characteristics

The pressure is determined by application of Bernoulli’s equation
p = p0 + (1 − wz wz ) / 2 . (2.3.20)
The resultant force and torque about the origin are calculated by integrating equations
(2.1.14) and (2.1.16) along the plate, which can be written on the parametric ζ -plane as
i⎛ ⎞
∞ ∞
X + iY = − ⎜ ∫ zζ (ξ )d ξ − ∫ wz (ξ ) wζ (ξ )d ξ ⎟ , (2.3.21)
2⎜ 0 ⎟
⎝ 0 ⎠
1⎛1 ⎞
∞
M = − ⎜ − Re ∫ wz (ξ ) wζ (ξ ) z (ξ ) d ξ ⎟ . (2.3.22)
2 ⎜⎝ 2 0
⎟
⎠
Since complex functions zζ (ξ ) and wz (ξ ) have the same imaginary arguments as e(π −α ) i ,
the resultant force as expected is directed inside the cavity, normal to the plate. Taking into
account equation (2.3.16), coefficients of resultant force Cn = 2 | X + iY | / ρ v∞2 l and torque
about the plate’s trailing edge CM = 2M / ρ v∞2 l 2 are obtained in the forms
⎛ s (1, − a, −b) ⎞
Cn = ⎜ 1 − ⎟ (1 + σ ) (2.3.23)
⎝ l ⎠
42
⎛1 1 1 ⎞
and CM = − ⎜ − 2 ∫ s′(t , − a, −b) s (t , a, b)dt ⎟ (1 + σ ) . (2.3.24)
⎝2 l 0 ⎠
The lift and drag coefficients are calculated from (2.3.23) multiplying by cos α and sin α
respectively. They can be also found using equation (2.1.15):
1
CL = [2Γ − (1 + σ )(xC2 − xC1 )] , (2.3.25)
l
1
CD = (1 + σ )( yC2 − yC1 ) . (2.3.26)
l
Since the difference between abscissas xC1 and xC2 is very small (see Fig.2.3.2b and
2.3.3b), the lift is well approximated using the circulation only, as for flow without
cavitation. The distance of the pressure center from the plate trailing edge is obtained as the
ratio s0 =| CM | / Cn . (2.3.27}
As expected, the torque and its coefficient are negative. This fact implies that the torque acts
clockwise. The torque relative to the leading edge is equal to
CMo = (1 − s0 )C n . (2.3.28)
As follows from asymptotic solution (2.3.15), the asymptotic expansion of drag coefficient
for the perpendicular plate, α = π / 2 , is found as
CD 2π ⎛ σ2 σ3 ⎞
= ⎜ 1 + − + …⎟ . (2.3.29)
1 + σ π + 4 ⎝ 8(π + 4) 8(π + 4) ⎠
This expansion coincides with the expression obtained in [Geurst, 1956], who applied an
end-plate model of cavity closure, a.k.a. the Riabouchinsky wall. The same expression can
be obtained if the cavity closure is modeled using a re-entrant jet.
2.3.7. Kirchhoff Problem
D’Alembert’s paradox of ideal flow is well known, namely that the resultant force due to
steady ideal flow around a body is equal to zero. To overcome this paradox, Kirchhoff in
1869 proposed a model of an infinitely long cavity behind a body, filled by stagnant fluid. A
similar problem for the inclined plate was later considered by Rayleigh in 1876. In this study
the Kirchhoff’s model has special significance as it corresponds to zero cavitation number.
The solution for Kirchhoff flow past an inclined flat plate can be found from the
above-obtained solution taking the limit as σ → 0 . Since the cavity closure region moves
away to infinity but point C in the ζ -plane remains fixed at ζ = i , then point D must
approach C. Also, by equation (2.3.13), the factor b in equation (2.3.8) must vanish as the
cavitation number approaches zero for any non-zero angle of attack. As parameter b
approaches zero and coordinate ζ D approaches i, equations (2.3.4) and (2.3.8) become
ζ (ζ 2 − a 2 ) ζ −a
wζ = , ω = ln + (α − π )i . (2.3.30)
(ζ + 1)
2 3
ζ +a
Hence, the complex velocity and the derivative of the transformation are expressed as
ζ −a ζ (ζ + a) 2
wz = −eα i , zζ = −e −α i . (2.3.31)
ζ +a (ζ 2 + 1)3
The only parameter a is found by substituting ζD = i, b = 0 and σ =0 in to equation (2.3.10):
43
α
a = cot . (2.3.32)
2
Now, all kinematic and dynamic characteristics can be calculated from equations (2.3.30) –
(2.3.32). As σ → 0 , coefficients (2.3.23) and (2.3.26) turn into Raleigh’s expressions
2π sin α C0 ⎛ 3 cos α ⎞
Cn0 = , CM0
= n ⎜⎜1 − ⎟. (2.3.33)
4 + π sin α 2 ⎝ 2(4 + π sin α ) ⎟⎠
2.3.8. Numerical Calculations

Results of numerical calculations for nonlinear cavitating flow around the plate are plotted in
Figs.2.3.4 – 2.3.6. Figs.2.3.4 and 2.3.5 show a dependence of normal force coefficient and
pressure center on angle of attack for six values of cavitation number.
Due to Tulin’s asymptotic expressions, the cavity's reduced length and width are
πσ 2 πσ
expressed as L0 = Lc , H0 = Hc , (2.3.34)
8C D0 4C D0
where C D0 = Cn0 sin α is the drag coefficient; length Lc is determined as the abscissa of low
spiral center C1; L0 and H0 are plotted in Fig. 2.3.6. It should be noted that the abscissas C1
and C2 of spiral centers are close but not coincident, and so the curves calculated for C2 will
differ from Fig. 2.3.6a, especially if angle of attack is small.
Fig.2.3.4. Coefficient of resultant Fig.2.3.5. Center of pressure for

force for flat plate flat plate
2.3.9. Asymptotic Expressions for Small Angle of Attack and Cavitation Number
Using asymptotic solution (2.3.13), one can obtain the known expressions from linearized
problem [Tulin, 1953] as follows:
4α 2
• cavity length lc = 2 ; (2.3.35)
σ
π (4α 2 + σ 2 )
• lift coefficient cL = cn = ; (2.3.36)
2(2α + 4α 2 + σ 2 )
44
0 π
• torque coefficient cM = 4
(4α 2 + σ 2 )(2 4α 2 + σ 2 + α )( 4α 2 + σ 2 − 2α ) 2 . (2.3.37)
8σ
Fig.2.3.6.
Geometric
measurement
of cavity
length (a)
and width
(b).
For comparison of linear and non-linear theories, some results of the lift coefficient
are given in Table 2.1. In the asymptotic approach the linear theory determines rather
accurate values of main flow parameters (see Table 2.1), therefore, using the linear theory is
acceptable for engineering purposes. In the next section some other models are presented.
Table 2.1. Lift coefficients by the linear approach, c L , and by the non-linear theory, C L .
σ = 0.1 α = 5°
α cL CL σ cL CL
1 0.1251 0.1244 0.1 0.1692 0.1683
2 0.1218 0.1219 0.2 0.2516 0.2524
3 0.1320 0.1325 0.3 0.3628 0.3612
4 0.1488 0.1490 0.4 0.4897 0.4831
10 0.2908 0.2726 0.5 0.6257 0.6123
Sec. 2.4. Partially Cavitating Flow

2.4.1. Problem Statement
Cavitating flow around a foil is called partially cavitating if the cavity length is less than the
foil chord. Both the partially cavitating and fully cavitating, discussed in Sec.2.3, flows are
supercavitation. Below the value problem of partially cavitating flow is considered in
conformity with flow of a foil with its trailing edge as a wedge of the given angle μ,
Fig.2.4.1a. As before, the stream function on the cavity boundary and the sides of angle are
constant. Without losing generality the definitional domain of complex potential can be
presented as shown in Fig.4.2.1b. For solving the problem analytically let's again consider
the first quadrant of the auxiliary ζ -plane, Fig.2.4.1c. As usually, the corresponding points
are marked by the same letters. The speed at the boundary is assumed to be equal to unity;
logarithmic function ω = ln(dw / dz ) has a singularity at point C as w −1 / 2 . Now, both w(ζ )
and ω (ζ ) can be found by their singularities.
45
Fig.2.4.1. (a) Partially cavitating flow; (b) the w-plane; (c) the parametric plane.
2.4.2. Complex Function Derivative

It is preferable to consider a single-value derivative, wζ = dw/dζ , instead of multi-value
complex function w(ζ ) . Since Im w is constant at the boundary then the derivative wζ (ζ )
is totally real on the ξ -axis and pure imaginary on the η -axis. Besides, it has a simple zero
at points C, A, O, and the pole of second kind at point ζ D = c + id . When analytically
continued through datum lines these nulls and singularity at symmetrical points remain. As
ζ (ζ 2 − 1)(ζ 2 − a 2 )
a result wζ = 2 . (2.4.1)
(ζ − ζ D 2 ) 2 (ζ 2 − ζ D 2 ) 2
The right part of this equation can be presented as a sum of simple fractions and then
A2 A1
integrating. E.g. the leading part at point ζ D is + ,
(ζ − ζ D ) 2 ζ − ζD
2
(ζ D2 − 1)(ζ D2 − a 2 )
where A2 = lim wζ (ζ − ζ D ) = , (2.4.2)
ζ →ζ D 2ζ D (ζ D2 − ζ D 2 ) 2
⎛ 2ζ 2ζ 4ζ ⎞
A1 = A2 ⎜ 2 D + 2 D − 2 D 2 ⎟ . (2.4.3)
⎜ ζ −1 ζ − a ζ − ζ ⎟
⎝ D D D D ⎠
Similar to potential (2.3.1) the potential can be recreated. Namely, the circulation of velocity
around foil is determined by Γ = 2π iA1 . (2.4.4)
As the sign convention, clockwise circulation is positive.
2.4.3. Complex Velocity and Transformation
Complex velocity wz = ve −iθ has a constant value | wz |= 1 at the free boundary BC, and a
step-wise argument at the wetted foil surface, i.e.:
θ = −πμ on CA, θ = 0 on OA, θ = −π on OB. (2.4.5)
Therefore, function w2(ζ) has a simple null at point O(a) and null of the μ-order at point
A(1). It can be extended analytically over the entire ζ-plane, then function w2(ζ) will have
another single pole at symmetrical point (-1) and a pole of the μ-order at point (-a).
Moreover, logarithmic function ln[w2(ζ)] has a simple pole at the origin, so that w2(ζ) has
μ b
ζ − a ⎛ ζ −1 ⎞ ζ
an exponential singularity and can be expressed as wz = − ⎜ ⎟ e . (2.4.6)
ζ + a ⎜⎝ ζ + 1 ⎟⎠
46
Transformation from the z –plane to the ζ- plane is written in a differential form as

b
ζ (ζ + a ) 2 (ζ − 1)1− μ (ζ + 1)1+ μ − ζ
zζ = − e . (2.4.7)
(ζ 2 − ζ D 2 ) 2 (ζ 2 − ζ D 2 ) 2
Functions (2.4.6) – (2.4.8) have four unknown real parameters: a, b, c and d. At infinity
z → ∞ or at corresponding point D(ζD), the following conditions should be imposed:
e − iα
• the given cavity number and the angle of attack: wz (ζ D ) = , (2.4.8)
1+σ
• function z(ζ) at point D is one-dimensional
2 1− μ 1+ μ b 4ζ D
+ + + − =0. (2.4.9)
ζ D + a ζ D −1 ζ D +1 ζ D 2
ζ D − ζ D2
2
Unknown parameters a, b, c, d can be calculated from four real equations obtained from
(2.4.8) and (2.4.9). It should be noted that the equations set to given parameters σ , α and
μ can have either one or two solutions or none. Any numerical solution is sensitive to the
initial values. Therefore, additional asymptotic investigation is needed. The trailing edge
angle should practically be small, thus it can initially be ignored. An asymptotic solution of
equations (2.4.8) and (2.4.9) for small σ and α may be expressed as follows [Terentiev,
1981]: a = 4d / α (3 + 2d 2 ), b = −α (1 + 2d 2 ) 2 / 2d , c = 1 + d 2 1 − 2α d 1 + d 2 , (2.4.10)
3
while parameter d can be found from equation d / 4(1 = 2d 2 ) 2 = α / σ . (2.4.11)
These asymptotic parameters can be used as initial values for solving equations (2.4.8) and
(2.4.9) numerically. However, only one solution can be found, while the other solution can
be lost. Therefore, it is advisable to pre-assign parameter d and then calculate other
parameters including ratio κ = sin α / σ . The lower limit of parameter d is zero and
corresponds to a continuous flow but the upper limit is unknown and is to be calculated.
2.4.4. Geometric and Dynamic Parameters
Integrating equation (2.4.7) can yield all necessary geometric parameters, namely,
τ (1 + aτ ) 2 (1 + τ )1+ μ (1 − τ )1− μ
1
• plate length l = ∫ e − bτ dτ , (2.4.12)
0 [1 − 2(c 2
− d 2 )τ 2 + (c 2 + d 2 ) 2τ 4 ]
2
1 /η 1/η
• cavity shape x(η ) = ∫ f (τ ) cos β (τ )dτ , y (η ) = ∫ f (τ ) sin β (τ )dτ , (2.4.13)
0 0
τ (1 + τ 2 )(1 + a 2τ 2 )
where f (τ ) = ,
[1 + 2(c 2
− d 2 )τ 2 + (c 2 + d 2 ) 2τ 4
2
]
2μ
β (τ ) = π − 2 arctan(aτ ) − arctanτ + bτ .
π
The cavity length is determined as the abscissa of the spiral center: Lc=x(1).
The resultant force and torque can be found using general integral expressions
iρ iρ
∫ wz 2 dz = −
2 ζ =∫ζ
R = X − iY = wz (ζ ) wζ (ζ ) dζ , (2.4.14)
2 z →∞
D
47
ρ ρ
M =− Re ∫ wz 2 zdz = Re ∫ wz (ζ )wζ (ζ ) z (ζ )dζ . (2.4.15)
2 z →∞
2 ζ =ζ D
Both integrals can be calculated analytically as residues of their integrand but can be
calculated numerically integrating along the circle of center ζ D and radius d. This circle
touches the ξ-axis at point ξ = c, thus function z(ζ) should be found numerically first by
integrating along the x-axis within interval ( ∞,c) , then along the circle ( ζ = ζ D + de it ' )
within interval ( − π 2, t ).The lift coefficient is analytically expressed as follows
π a 2 + (c 2 + d 2 ) 2 − (1 + a 2 )(c 2 − d 2 )
CL = 1+ σ . (2.4.16)
l 16c 3 d 3
For a plate ( μ = 0 ) with a small angle of attack and a small cavitation number, asymptotic
expressions can obtained from (2.4.10) and (2.4.11) as follows:
4d 2 (1 + d 2 )
Lc = , CL = 2πα (1 + d 2 ),
(1 + 2d )
2 2
(2.4.17)
⎡ 1⎛ 6 1 4 ⎞⎤
CM = πα ⎢1 + d − ⎜ 3 +
2
− − ⎟⎥ .
⎣ 8 ⎝ 1 + 2d 2 (1 + 2d 2 ) 2 (1 + 2d 2 )3 ⎠ ⎦
From first equation of (2.4.17) one can find: 1 + 2d 2 = 1 / 1 − Lc 2 . Then coefficients CL, CM
and ratio α/σ are expressed as functions of cavity length as
(
CL = πα 1 + 1/ 1 − Lc ) (2.4.18)
CM =
C L πα
2
−
8
(
2 + Lc + 2(1 + 2 Lc ) 1 − Lc , ) (2.4.19)
α Lc 1 − Lc
= . (2.4.20)
(
σ 1+ 1− L 2
c )
Asymptotic expressions (2.4.18) – (2.4.20) coincide with the linearized problem solution
derived in [Geurst, 1959].
Some results of calculations are plotted in Figs.2.4.2 – 2.4.4. The cavity shape on a plate
with attack angle α = 5° and cavity length Lc =0.4 is shown in Fig.2.4.2a. The part of cavity
boundary near the leading edge and a spiral at the end of cavity are shown in a larger scale in
Fig.2.4.2b. It is seen that the boundary near the leading edge goes beyond the plate, while at
the cavity end it comes to the end of a quickly converging spiral. These deviations are so
insignificant, that at any approximation cannot be taken into account. The cavity boundary
can be represented by an open-ended curve connecting the separation point (here the leading
edge) with a point outside the foil.
Distribution of speed on a plate and cavity is shown in Fig.2.4.3. It is interesting to
notice a curious but expected result: speed along the plate drops sharply and becomes zero at
the cavity while pressure reaches its maximum. This contradicts the assumption about the
minimum pressure in the cavity. The contradiction is explained by flow in the multi-sheet
plane; the pressure on all the sheets except the basic one takes maximal but constant values.
48
Therefore it will not render
any influence on the lift and
torque.
Geometric and
dynamic characteristics are
plotted in Fig.2.4.4. The
curves for α → 0 are
calculated using equations Fig.2.4.2.
(2.4.18) – (2.4.20). Fig.2.4.4a The cavity shape on the
shows that the cavity length plate: (a) total boundary;
can double for the same fixed (b) parts of cavity
ratio k = sinα / α . boundary near the leading
Theoretically, the full edge and spiral at the end
cavitation is possible as well. of cavity
Thus three modes of
cavitating flow are possible
for a small fixed angle of attack and the same cavitation number. Physically, only a flow
with a short cavity, which has the lowest lift, seems to be realistic. The extreme points, k 0 ,
on the curves in Fig.2.4.4a separate the partially cavitating flow
from the full one. The cavitating flow near k0 should be an acute
unsteady flow. It has been observed experimentally. Fig.2.4.4c
shows that the load point moves first to the leading edge then to
the middle of the plate. Fig.2.4.5 shows two possible cavity
shapes for fixed cavitation number σ = 4.5 and angle of attack
α = 10°. The cavity length can theoretically be greater than the
plate length.
Fig.2.4.3. Speed distribution.
Returning to the foil with thickness (Fig.2.4.1a) one should consider first the
cavitating flow with a smooth touch of the cavity boundary on the inclined side of trailing
angle, i.e. calculate for b = 0. The solution of that case has four unknown parameters a, c, d
and ratio κ = sin α / σ which are calculated from (2.4.8) and (2.4.9).
Fig.2.4.4. Partially cavitating plate: (a) cavity length as a function of attack angle and
cavitation number, (b) lift coefficient, (c) resultant force application point S 0 = M / Ll sin α .
49
Fig.2.4.5. Cavity shapes for α=10° and σ = 4.5 : (a) shape at the left from the curve
maximum; (b) shape at the right side.
Fig.2.4.6.
Plate with inclined upper side
and smooth touch of the cavity
end.
Fig.2.4.6a shows that

the possible cavitation number
of this model has lower limit
σ m ( μ ) for cavitation numbers
greater than σm, two sorts of
cavitating flow smoothly
touching the cavity end can be
realized. Cavity shapes for
those flows are shown in
Fig.2.4.7a. For a fixed trailing
wedge angle (e.g. μ=0.0672)
two conditions of the flow are
realized at cavitation numbers σ<1.5,
but for a greater angle (here
μ=0.4406) the cavitating flow is
possible at cavitation numbers σ >
1.5 only.
Fig.2.4.7.
Cavity shapes with partial
cavitation: two possible cavities with
smooth (a)
and singular (b) touch of the
boundary at the cavity end;
(c) – a single possible cavity.
The results presented here

show a variety of partially cavitating
50
flows. Infinity at the multi-sheet surface makes it virtually impossible to apply this model as
well as the model with re-entrant jet to investigate partially cavitating hydrofoils with
curvilinear boundaries. Furthermore, it is difficult to apply the cavitation model with
singularity at the cavity end to calculate the cavitating flow using the finite-difference
numerical methods. For those purposes, it is quite convenient to apply a cavitation model
with an end plate. Some possible models will be considered in Sections 2.5 and 2.6 below.
Sec. 2.5. Review of Cavitating Flow Models

Cavitating flows have been modeled using various assumptions about the flow at the cavity
end. Some of those models are discussed below. All models can be sorted in two groups:
closed models in which condition (2.1.6d) is satisfied and open models in which semi
infinite wake appears. The Kirchhoff’s flow at zero cavitation number is a simplified open
model. It can be considered open model at zero cavitation number if the speed distribution
on the wetted boundary is the same as that in the Kirchhoff's flow but with the inflow speed
v0 = v∞ 1 + σ . In that case the coefficients of resultant force and moment differ from those
in Rayleigh’s equations (2.3.33) by factor ( 1 + σ ), i.e.
Cn = Cno (1 + σ ), CM = CMo (1 + σ ) . (2.5.1)
These expressions are also obtained from (2.3.23) and (2.3.24) by expansion in series for a
small cavitation number. The geometric sizes of cavity can also be found using Tulin’s
formulas (2.3.37). Below some closed models are considered first.
2.5.1. Riabouchinsky Model with a Perpendicular Trailing Plate
Historically, the model was first considered as a mathematical problem of possible flow
around a perpendicular plate with free boundaries closed at the plate [Riabouchinsky, 1920;
Cisotti, 1921]. It is called the Riabouchinsky model, see Fig.2.5.1a. Weinig [1932] obtained
numerical data for cavitating flows. Cavitating flow around an inclined plate was considered
in [Terentiev, 1964] assuming the central symmetry, which is relatively easy for analytical
and numerical studies. Results very similar to those in Sec.2.3 were obtained using the
cavitation model with a
normal trailing plate
[Terentiev, 2003].
Fig.2.5.1.
a) The Riabouchinsky
model, b) the ζ -plane.
The solution can be obtained parametrically for a rectangle with corner points ζ = 0 ,
π / 2 , π (1 + τ i ) / 2 , πτ i / 2 on the auxiliary ζ -plane, Fig.2.5.1b. The inclined plate is the
lower base of the rectangle, the perpendicular end plate is the upper base and the cavity
boundaries are the rectangle's sides; the stagnation points are represented by points ζ=0 and
ζ = b + πτ i / 2 . It is clear that the complex velocity has simple zero at points a and
b + πτ i / 2 ; its absolute magnitude is constant at cavity boundaries and its argument is
piecewise constant on the plates. Therefore, it can be extended analytically throughout the
sides of rectangle on the entire ζ--plane, so that the complex velocity becomes simple poles
at symmetrical points –a and −b + πτ i / 2 within the fundamental rectangle (dotted lines).
51
Thus, complex velocity wz(ζ) becomes an elliptical function with the given poles and zeros
which can be written by theta-functions [Whittaker & Watson, 1927] in the form (see also
ϑ (ζ − a)ϑ4 (ζ − b)
Appendix 5 for theta-functions) wz = −eiα 1 , (2.5.2)
ϑ1 (ζ + a)ϑ4 (ζ + b)
where unknown parameters a and b satisfy the equation:
3π α
a+b= − . (2.5.3)
4 2
The last equation is obtained by satisfying the condition at point D. Thus one of the
unknown parameters, b, is expressed via another parameter, a.
The derivative of the complex potential has a simple zero at the stagnation points
and at its symmetric points, and also at the corners of the rectangle, besides it has poles of
second order at the inner points, ζ = ± c ± id , corresponding to infinity, z → ∞ , of the z-
ϑ1 (2ζ )ϑ1 (ζ − a)ϑ1 (ζ + a)ϑ4 (ζ − b)ϑ4 (ζ + b)
plane. Then, wζ = . (2.5.4)
[ϑ1 (ζ − c − id )ϑ1 (ζ + c + id )ϑ1 (ζ − c + id )ϑ1 (ζ + c − id )]
2
Functions (2.5.2) and (2.5.4) represent a general solution of the problem. The four remaining
parameters a, c, d and τ should be found satisfying two complex conditions as follows:
• the cavity closure condition (2.1.6d) which can be reduced to
d
lim ln zζ (ζ )(ζ − c − id ) 2 = 0 , (2.5.5)
ζ → c + id dζ
• the inlet velocity at infinity wz (c + id ) = −1/ 1 + σ ; (2.5.6)

Instead of complex speed it is preferable to use its logarithmic function, ω , then the real part
is a single-value function while the imaginary part is a multi-value function. It is quite
difficult to clarify the desirable branch of this function, but it could be equal to zero or to a
number multiple to π . In any case, sinus of it should be equal to zero. Now, instead of
equation (2.5.5) it is much better to consider two real equations as
ln | wz (c + id ) |= −1/ ln 1 + σ and sin(Im ω (c + id )) = 0 . (2.5.6a)
Adding here two scalar equations from (2.5.5) one obtains a closed loop system for unknown
parameters a, c, d and τ . The equation set can be easily solved numerically. Due to the
trailing plate perpendicularity, the lift is equal to ρ v∞ Γ , so that the lift coefficient is
2π
2 1+σ
CL = ∫ wζ (c + id + δ e )δ ei t idt .
it
(2.5.7)
L 0
The cavity shape for angle of attack α = π / 6 and cavitation number σ = 0.5 is
shown in Fig.2.5.1a; while all geometric ratios of this parametric rectangle are plotted in
Fig.2.5.1b as calculated data.
2.5.2. Re-Entrant Jet Model
Re-entrant jets have been observed in real cavitating flows, which are dispersed and run
away into mainstream. This model was virtually simultaneously introduced by Efros,
Kreisel, and Gilbarg & Rock in 1946. The re-entrant jet model has been used in many works
mainly for studying the symmetric cavitating flow [Gurevich, 1947; Kuznetsov, 1962;
Pyhteev, 1956]. An inclined flat and the effect of the re-entrant jet’s direction were
considered in [Gusev & Terentiev, 1970].
52
For a flat plate, Fig.2.5.2, the problem can be solved in a parametric form in the
domain shown in Fig.2.3.1c. The boundary conditions are almost the same as for the model
with single spirals but here there should be an inner point E (ζ = b1 + ib2 ) at which the
velocity must be zero. Let the infinite point of re-entrant jet map onto point C (ζ = i ) , then
the derivative of the complex potential will have a simple pole at this point as well as at
symmetric point C ′(−i ) while the complex velocity at point C is equal to e − iγ where γ is the
angle between the velocity in the re-entrant jet and the positive x-axis. Hence, taking into
account all poles and zeros, the following desirable functions can be found:
ζ (ζ 2 − a 2 )(ζ 2 − ζ E2 )(ζ 2 − ζ E2 )
• derivative of the complex potential wζ = (2.5.8)
(ζ 2 + 1)(ζ 2 − ζ D2 ) 2 (ζ 2 − ζ D2 ) 2
(ζ − a)(ζ − ζ E )(ζ − ζ F )
• complex velocity wz = ei (α −π ) . (2.5.9)
(ζ + a)(ζ + ζ E )(ζ + ζ F )
Functions (2.5.8) and (2.5.9) have five unknown real parameters: a, ζ D = c + id ,
ζ E = b1 + ib2 . To determine these parameters there are four non-linear equations as follows:
• two equations obtained from the given velocity at infinity wz (ζ D ) = 1 ; (2.5.10)
• two equations obtained from “closure” condition (2.1.6d)
1 1 1 ζ 2ζ
+ + − 2 D − 2 D 2 = 0. (2.5.11)
ζ D + ζ E ζ D + ζ E ζ D + a ζ D +1 ζ D + ζ D
Generality, the problem is indeterminate, and an auxiliary condition should also be given.
This condition could be a given direction of the re-entrant jet, γ,
Im(ln wz (i )) = γ . (2.5.12)
Theoretically, equations (2.5.10) –
(2.5.12) can be solved for arbitrary
angle γ. Fortunately, numerous
calculations convince that the results
do not depend much on this angle. In
Fig.2.5.2. Cavity shapes of re-entrant model for practice, it is desirable γ=π. Fig.2.5.2
different angels of the jet; α = 30 , σ = 0.5 . shows the cavity shapes calculated
for angle of attack α=30°, cavitation
number σ = 0.5 and three values of the angle, γ=π, 5π/6, 4π/3.
2.5.3. Re-Entrant Straight-Lines Model
Another closed cavitating model was offered by Kuznetsov [1964]. Two re-entrant straight
lines are included in the model instead of a jet, Fig.2.5.3. As a result, the fluid flows first
inside then returns back in the main stream. The model, like the Riabouchinsky’s one, has
finite-length cavity boundaries that are convenient for numerical calculations. The model has
been used [Kotlyar & Troepolskaya,
1975] for studying a cavitating flow in a
transverse gravity field.
Fig.2.5.3.
Cavity shapes of re-entrant straight
lines model: here α=30°, σ = 0.5.
53
The value problem was solved in a parametric rectangle, Fig.2.5.1b. All singularities
and zeros hold out as in the Riabouchinsky model except point B(b+πt/2) at which the
complex velocity has multiple zeros (zero of second order). Thus, the expression of
derivative of complex potential coincides with equation (2.5.4) but the complex velocity is
ϑ (ζ − a)ϑ42 (ζ − b)
wz = −eiα 1 . (2.5.13)
ϑ1 (ζ + a)ϑ42 (ζ + b)
All conclusions made about trailing-plate model are valid for this case as well.
2.5.4. Numerical Comparison of Closed Cavitating Models
The cavity shapes behind inclined
flat plate with respect to different
models are shown in Fig.2.5.4.
Fig.2.5.4. Cavity shapes of

different closed models.
The reduced
cavity width and length
calculated using
different models are
plotted in Fig.2.5.5.
These plots and many
others confirm that the Fig.2.5.5.
single-spiral and Reduced cavity width
trailing-plate models, and length as
like the re-entrant functions of cavitation
plates, give the results number.
identical up to 10-4. The
re-entrant jet model
gives different results
for cavity length only. Therefore, all these models could be used for studying cavitating
flows. But only the cavitating model with singularity (2.1.8) which generates a single spiral
is valid theoretically. Below, this model will mainly be used.
2.5.5. Partial Cavitating Models with Closed Cavity
The above-mentioned close models could also be used for describing a flow with partial
cavity. Fig.2.5.6a,b shows the partial cavity on a flat plate corresponding to the trailing plate
[Terentiev, 1969] and re-entrant jet [Kuznetzov & Terentiv, 1970]. Both models in
conformity with the flat plate can be investigated on the auxiliary ζ -plane by conformal
mapping onto the first quadrant, Fig.2.5.6c. Corresponding points are denoted by the same
letters. Also the derivative of complex potential wζ (ζ ) takes a real value on the real axis
and an imaginary value on the other axis while the complex velocity, wz (ζ ) , takes a real
value on both axes and can be found by singularities and zeros.
The derivative of potential has a dipole singularity at inner point M(c+id) and at
symmetrical points M ′, M ′′, M ′′′ , and a simple zero at stagnation point B(b) and at trailing
54
edge C(1), as well as at symmetric points B′(−b) and C ′( −1) . Besides, the derivative for
trailing plate model has a zero at the origin, O(0), while it has a simple pole for the re-entrant
model and additional zeros at points A(a) and A′( −a) . The complex velocity of both models
has a simple zero at stagnation
point B(b), but different values at
point A(a): it is zero ζ − a for
the first model but simple zero for
the other model. A zero changes
to a pole at symmetrical points.
Now, the derivative of complex
potential and complex velocity
are:
Fig.2.5.6.
Partial cavity:
(a) trailing plate model;
(b) re-entrant model,
(c) auxiliary ζ -plane.
The cavity shapes are shown for
cavity length Lc=0.75 and angle
of attack α=5°. The values of cavitation number are shown inside the cavities
ζ (ζ 2 -b 2 )(ζ 2 -1)
• for trailing plate model: wζ = 2 , (2.5.14)
(ζ − (c + id ) 2 ) 2 (ζ 2 − (c − id ) 2 ) 2
ζ −b ζ −a
wz = − ; (2.5.15)
ζ +b ζ +a
(ζ 2 − 1)(ζ 2 -b 2 )(ζ 2 -1)
• for re-entrant model: wζ = , (2.5.16)
ζ (ζ 2 − (c + id ) 2 ) 2 (ζ 2 − (c − id ) 2 ) 2
(ζ − a)(ζ − b)
wz = − . (2.5.17)
(ζ + a)(ζ + b)
The transformation from the z-plane to the ζ -plane is realized, as above, integrating
function zζ = wζ / wz . The unknown parameters, a, b, c and d, are calculated as in the full
cavitating flow using the same conditions: cavity closure condition (2.5.5) and inlet velocity
at infinity (2.5.6). Because the cavity length is an ambiguous function of cavitation number,
it is preferable to give the cavity length but the cavity should be calculated. All curves are
similar to the corresponding lines obtained in Sec. 2.4. Fig.2.5.7 shows two cavity shapes
corresponding to the trailing plate and to the re-entrant jet. The cavity and all curves for the
singular model coincide with the trailing plate model.
Fig.2.5.7.
Cavity shapes for trailing
plate and re-entrant
models.
55
2.5.6. Cavitating Models with Rotational Fluid Motion

Some other cavitating models with closed cavity have been proposed since 1930, but all of
them were to a certain extent inconvenient for modeling an asymmetrical flow though they
are mathematically valid. The two cavitating models in [Karlikov & Tolokonnikov, 2004]
appear very attractive. They provide closed domains of rotating fluid at the end of the cavity
induced by the dipole and the vortex at point C, Fig.2.5.8a & b, respectively. Their
computation results are plotted in Fig.2.5.8. These cavities correspond to cavitation number
σ = 0.9 ; for comparison, the cavities were computed also using Riabouchinski and Efros –
Gilbarg models. As stated above, all these models give almost the same results.
Fig.2.5.8.
Cavities with
closed domain of
rotating fluid at
cavitation number,
σ = 0.9 :
(a) the dipole at
point C;
(b) the vortex.
Sec. 2.6. Open Models of Cavitating Flow

The above-mentioned cavity flow models satisfy closure condition (2.1.6d) without any
wakes past the foil. In reality a turbulent wake filled up by a water-gas mix is formed behind
a cavity. This phenomenon is very difficult to investigate, though there are attempts to so.
We shall return to this problem later. Here other cavity models are considered which involve
semi-infinite wakes past the cavity but they cannot be identified with any real wake, though
some of these models were applied to clarify a real flow. In any case, these models have
been used for studying the complicated problem of cavitating flow of inviscid fluid, for
instance, marine propeller, influence of free surface or given boundary. One of such models
is the above-mentioned Kirchhoff’s flow considered in Sec. 2.5.
2.6.1. Second Tulin Model of Cavitating Flow
As mentioned above, Tulin [1965] also suggested another
possible nonlinear model for simulating cavitating flows. A
value of speed at the free boundaries is assumed to be piecewise
constant; namely, the speed is constant, v0 , at the cavity
boundary and is equal to another constant, v∞ , at the wake
boundary. Both boundaries form a double spiral at the attached
point (Fig.2.6.1) and function ω ( w) has a logarithmic Fig.2.6.1. Streamlines
singularity only [Birkhoff & Zarantonello, 1957]. The Tulin’s of double spiral.
model allows the multi-connected flow domain to be reduced to
a simply connected one and this simplifies the flow problem. Larock & Street [1967] applied
this model for a nonlinear problem solution of a fully cavitating hydrofoil beneath a free
56
surface. Then Terentiev & Lasarev [1969] used the Tulin’s model to investigate numerous
problems of cavitating flow in a bounded domain (see Sec. 3.8). Here we apply the model in
an unbounded domain only.
The flow problem of an
inclined plate with an attached
Tulin’s cavity has been solved
by conformal mapping the flow
domain on a physical domain in
the first quadrant of the ζ - Fig.2.6.2.
plane, Fig.2.6.2. The flow (a) Tulin’s model
problem is reduced to boundary with double spirals,
value problems for the (b) the parametric
derivative of complex potential ζ -plane.
wζ , and for the logarithmic
function of complex velocity ω . Boundary conditions of function wζ are the same as for the
Kirchhoff’s problem, i.e. its value is real on the ξ -axis and pure imaginary on the η -axis,
and has simple zero at point a, at symmetrical point –a and at the origin; besides, it has a
ζ (ζ 2 − a 2 )
pole of third order at points i and –i. Hence, wζ = . (2.6.1)
(ζ 2 + 1)3
Since the imaginary part of function ω has a discontinuity of the first kind at stagnation
point A(a) by increment π , while its real part has a discontinuity of the first kind at points bi
and ci by increments ln(v∞ ) and ln(1/ v∞ ) , respectively. Hence, the derivative of that
function will have simple poles at these points and also at symmetrical points – a, – ib and
– ic. In view of the principle of analytic extension, the factors of poles at symmetrical points
differ by signs only so that the desirable function can be written in the form
dω 1 1 ln(v∞ ) ln(1/ v∞ ) ln(v∞ ) ln(1/ v∞ )
= − + + − − .
d ζ ζ − a ζ + a ζ − bi ζ − ci ζ − bi ζ − ci
ζ −a i (ζ − ci )(ζ + bi )
Hence, ω = ln + ln v0 ln + ln C . (2.6.2)
ζ +a π (ζ + ci )(ζ − bi )
Constant C is found by satisfying the following condition at the trailing wedge:
C = ln v0 + iα − π i .
Functions (2.6.1) and (2.6.2) have three unknown parameters: a, b and c. But there is only
scalar equation Im ω (i ) = 0 . (2.6.3)
Thus the solution of flow problem due to this cavitating model is twice indeterminate. Two
additional equations are needed, but plausible and well-grounded conditions are lacking. The
following conditions have been used in [Terentiev & Lasarev, 1969]:
Re[ w(bi) − w(ci)] = 0 , (2.6.4)
d
Imω (η ) η =1 = 0 . (2.6.5)
dη
Condition (2.6.4) has been used in many works but in this case the circulation around cavity
and body is always zero. Eq. (2.6.4) can be replaced with equality of abscissas of double
spirals xM = xN . (2.6.6)
57
In this case, circulation Γ differs from zero and determines the lift by the well known rule,
L = ρ v∞ Γ . For considering a real cavitating flow, the use of a relationship from viscous
theory was suggested in [Tulin, 1964], i.e. the trailing wake has to imitate the downstream
behavior of displacement thickness of real wake t. As a result, the asymptotic thickness of
the trailing wake in infinity should be t / l = CD / 2 (2.6.7)
Numerous calculations show that only the above-mentioned conditions (2.6.4) and
(2.6.5) are realized easily and the wake boundaries at infinity are not intersected, as is the
case in other conditions. The cavity and trailing wake past a perpendicular plate for
cavitation number σ = 1 are plotted in Fig.2.6.2. The spiral center’s ordinate, yM=−0.3875,
was obtained using conditions (2.6.3) and (2.6.5). Condition (2.6.4) is similarly fulfilled due
to symmetry. Other two cavity shapes and wakes were calculated by given spiral ordinates
( yM = −0.25, yM = 0 ) at the intersections of the wake boundaries. Cavity shapes are plotted
in Fig.2.6.2. The calculations were done for the given thickness of wake at infinity. The
cavity-wake boundary is practically independent of the value of finite width. The drag
coefficient for the perpendicular plate was the same, Cn = 1.775, for all cases considered.
Therefore, equation (2.6.7) is preferable to use for calculating the displacement thickness of
trailing wake at infinity instead of an additional condition for unknown parameters.
Fig.2.6.3. The cavity shapes and wakes for different ordinates of spiral center yM..
The calculated cavity shape past an inclined plate at α = 10 and σ = 0.2 is shown in
Fig.2.6.3. More detailed investigation using this model will be given in Sec. 3.8 below.
In conclusion it should be noted about the Tulin’s model:
• the model with double spirals is easy to apply to sufficiently complicated problems with
free boundaries because of the simple domain of the complex potential and simple
expressions of conditions;
• the model can be helpful for simulating real flows taking into account a viscous wake
past a cavity or for designing hydrodynamic vehicles [Stepanov, 1998;Tulin 2003];
• the model often has indeterminacy, thus making its broad application problematic.
2.6.2. Joukowski Model
Theoretical problems of a flow with free boundaries were considered in [Joukowski, 1890].
One of them has been used later to describe a cavitating flow [Roshko, 1955; Eppler, 1954;
Wu, 1956; Terentiev, 1970]. The cavitating model for an inclined flat plate is shown in
Fig.2.6.4a. The free boundaries terminate at two straight lines, DB and BE, which stretch to
infinity. The velocity decreases along these lines from 1 to v∞ = − ln 1 + σ . The parametric
domain for this model can be taken as a rectangle with sides π / 2 and πτ / 2 , as seen in
Fig.2.5.3b.
58
The derivative of complex
potential has a multiple pole of third order
at point B, and a simple zero at point A, as
well as at corners of the rectangle, while
the complex velocity has a simple zero at
point A; besides, its value is constant at
Fig.2.6.4. the vertical sides and its argument is step
(a) Joukowski model; constant at the horizontal sides. Both
(b) parametric functions can be continued through the
rectangle. rectangle sides and then reestablished by
theta-functions
ϑ (2ζ )ϑ1 (ζ − a)ϑ1 (ζ + a)
wζ = 1 (2.6.8)
[ϑ4 (ζ − b)ϑ4 (ζ + b)]3
ϑ (ζ − a )
and wz = e − (π −α ) i 1 . (2.6.9)
ϑ1 (ζ + a)
The complex velocity at the corner point D is wz = 1, therefore
a = (π − α ) / 2 . (2.6.10)
Functions (2.6.8) and (2.6.9) have two unknown parameters, b and τ . For that two
equations are obtained from given inlet velocity v∞ , and the extreme condition at point B
ϑ (b − a ) d ϑ4 (b − a)
ln 4 = ln 1 + σ and ln = 0. (2.6.11)
ϑ4 (b + a) db ϑ4 (b + a )
Geometric behavior is calculated as before integrating function f (ζ , a, b,τ ) = wζ / wz . As a
result, the coefficient of resultant force is determined as
Cn = (1 − J 2 / J1 )(1 + σ ), (2.6.12)
π /2 π /2
where J1 = ∫ | f (ξ , a, b,τ ) | dξ ,
0
J2 = ∫ | f (ξ , −a, b,τ ) | dξ .
0
The calculated cavity for
attack angle α=30° and cavitation number σ = 0.5 is shown in Fig.2.6.4a.

2.6.3. Wu – Fabula Model
As noted in Sec. 2.3, theoretically a flow is possible
without “closing” condition (2.1.6d); then equation
(2.1.7) is valid and the flow involves a semi-infinite
trailing wake between two congruent stream lines. In
Fig.2.6.5. Wu – Fabula model
this case the solution of flow problem can be found both
with singularity (2.1.8) and without it. The latter was
considered as a possible cavity flow model in [Wu, 1962; Fabula, 1962]. The flow problem
statement is described by equations (2.1.6a) – (2.1.6c) and a solution should be found in a
class of bounded functions.
All expressions from Sec. 2.3 for the flat plate are also valid for this model if
parameter b is supposed to be zero. The solution will have superfluous parameters, i.e. the
problem as in all the above-considered models is indeterminate and should involve an
auxiliary condition for obtaining all unknown parameters. In many works [Wu, 1962;
Fabula, 1962; Terentiev, 1967; Achkinadze, 1983; and others] condition (2.6.4) has been
59
used. Another condition can be used instead of equation (2.6.4), e.g. condition (2.6.6). The
cavity shape of the Wu – Fabula model for inclined flat plate (α =10°, σ =0.5) using
condition (2.6.6) is shown in Fig.2.6.5. Comparison of open models shows that the double
spiral model can describe a real cavitating flow of viscous fluid [Tulin, 2003; Stepanov,
1998] while other two models have a remarkable influence on cavity boundary and involve
thicker wakes than that of the frontal projection of the body. They can be used mostly for
calculating hydrodynamic forces only. The model with a wake of congruent lines is
preferable due to its mathematical simplicity.
2.6.4. Remark on Partial Cavitation Models
It was shown in previous sections that all cavity
models can be applied to flow with a partial cavity.
Among open models only the Wu – Fubula model
represents the partial cavity without any problems,
Fig.2.6.6, [Wu, 1962, Fabula, 1962]. Using the Fig.2.6.6. (a) Partial cavity by
parametric first quadrant of the ζ -plane, a solution of Wu; (b) parametric domain.
the problem is determined by two functions
ζ (ζ 2 − a 2 )(ζ 2 − 1)
wζ = , (2.6.13)
[(ζ 2 − ζ E2 )(ζ 2 − ζ E2 )]2
(ζ − 1)
wz = − , (2.6.14)
(ζ + 1)
where ζ ∞ = a + ib – is unknown coordinate of point E obtained analytically from condition,
1 − v∞ e − iα
wz (ζ E ) = v∞ e − iα as ζ E = c + id = . (2.6.15)
1 + v∞ e − iα
Because of congruence of stream lines AE, the increment of function z (ζ ) round point E
becomes a pure imaginary number. Thus Re ∫ wζ d ζ = 0 .
ζ∞
Hence, Re[2π iA( a, c, d ) B (a, c, d )] = 0 , (2.6.16)

((c + id ) − a )(c + id + 1)
2 2 2
A(a, c, d ) = ,
64c 2 d 2 (c + id )
where
2(c + id ) 2 i (c + id )
B ( a , c, d ) = + + .
(c + id ) − a
2 2
(c + id + 1) cd
Equation (2.6.16) determines only unknown parameter a.
The Wu – Fabula models (solid lines), close models from Sections 2.3 and 2.4
(broken line) and experimental points [Achkinadze, 2005; Numachi & Chida, 1958] are
compared in Fig.2.6.7. All curves on this figure were calculated for angle of attack α=5°
though they are valid for any small angle. One can see that the model without trailing wake
and the model with congruent boundaries wake differ much in the partial-to-full cavitation
transition interval, σ / α ∈ (6,14) . Though, the experimental points are closer to the open
model, one can hardly say that the model imitates a real flow correctly. Some experiments
give results that better agree with the closed model [Numachi & Chida, 1958]. Thus, the
60
transient state from partial to full cavitation
should be investigated in more detail. In spite of
that contradiction the model could be used for
investigating complicated problems of cavitating
flow.
Fig.2.6.7.
Comparison of close and partial cavity models
with experiments.
The Tulin’s double-spiral model can also be applied to partial cavitation where two
possible models are available. Its mathematical statements are clear from Fig.2.6.8; the
solution can be found in a parametric form on the above-mentioned rectangle for the model
in Fig.2.6.8a, and on the first quadrant for the model in Fig.2.6.8b.
Fig.2.6.8. Two possible

models: without (a) and
with (b) double spiral.
61
Ch. 3. Method of Conformal Mapping

Sec. 3.1. Mathematical Aspects of Conformal Mapping
As shown in Ch. 2, the conformal mapping is an effective tool of analytical study of flow
problems. The conformal technique can be found in many books, e.g. [Lavrent’ev & Shabat,
1965]. Some basic results are presented below.
3.1.1. Analytic Functions of Complex Variable
Complex variables and analytic functions have been studied since the 18th century and there
is a well established theory with many applications. A function ζ ( z ) = ξ ( x, y ) + iη ( x, y ) that
is differentiable at each point in a complex plane z = x + iy is called analytical. This
definition yields the so-called Cauchi1 conditions
∂ξ ∂η ∂ξ ∂η
= , =− . (3.1.1)
∂x ∂y ∂y ∂x
The velocity components of a 2D flow, u and v, for incompressible and inviscid fluid satisfy
the Cauchy conditions for the function χ ( z ) = u − iv , which is called conjugate complex
velocity, or simply complex velocity or velocity. For an unbounded flow with inflow speed at
infinity v∞ e − iα the velocity – as an analytic function – can be expanded into series
a a
χ = v∞ e − iα + 1 + 22 + … , (3.1.2)
z z
whose integral is an analitic function as well and has a series expansion as
a
w( z ) = c + v∞ e − iα z + a1 ln z − 22 − … The derivative of the latter is equal exactly to the
z
velocity, so that function w(z) is called complex potential. It is simple to show that the real
part, Re w = ϕ ( x, y ) , is speed potential while the imaginary part, Im w = ψ ( x, y ) , is stream
function. Both of them are defined with accuracy up to arbitrary values.
Due to the logarithmic term the increment of potential by roundabout infinity
anticlockwise is equal to 2π ia1 = Γ + iQ , where Γ is circulation round infinity, Q is quantity
of fluid flowing through a large-radius circle. Series expansion of complex potential at
Γ + iQ c
infinity is w( z ) = v∞ e − iα z + ln z + 1 + … + const . (3.1.3)
2π i z
Consider ordered increment of variable Δz = 2π ic1Δs /(Γ + iQ) , then the derivative of
the second expansion term is equal to the third term and so forth. The third term of (3.1.3) is
called dipole, the next term – quadrupole and so on but those are no longer important.
Function w(z) is multi-valued because of logarithmic term so that its definitional domain is
on a multi-sheet plane.
Another property of analytic functions is that an analytic function of an analytic
function is an analytic function. Let w(z) and z(ζ) are analytic functions, then W(ζ)=w(z(ζ))
is an analytic function of ζ. Let now z(ζ) be a single-value function in a domain with
1
Conditions (3.1.1) were derived first by D'Alambert in 1752 and then by Euler in 1755 [Lavrent’ev
& Shabat, 1965].
62
b1 b2
infinity so that its expansion at infinity is z (ζ ) = b0ζ + + +… , then expansion
ζ ζ2
Γ + iQ d
(3.1.3) is transformed to the form w(z(ζ )) = v∞ e − iα b0ζ + ln ζ + 1 + … + const .
2π i ζ
Note: The latter shows that function w( z (ζ )) is complex potential of flow on the ζ -plane
with the same circulation Γ and the same fluid consumption Q.
3.1.2. Principal Theorems of Analytic Functions
Theorem 1. If χ ( z ) = u − iv is an analytic function in domain D bound by closed curve C,
then integral along this curve is equal to zero, i.e. ∫ χ ( z )dz = 0 .
C
(3.1.4)
Proof. Choose the real and imaginary parts of integral as follows

∫ χ ( z )dz = ∫ udx + vdy + i ∫ udy − vdx
C C C
and then apply Green’s formula to each part:
⎛ ∂v ∂u ⎞ ⎛ ∂u ∂v ⎞
∫C χ ( z )dz = ∫∫D ⎜⎝ ∂x − ∂y ⎟⎠dxdy + i ∫∫D ⎜⎝ ∂x + ∂y ⎟⎠ dxdy
Due to Cauchy conditions (3.1.1) both integrands vanish. The theorem is thus proved.
Corollaries:
1. Integral (3.1.4) along any curve connecting two points is equal to the difference between
the values of primitive function
CMN
∫ χ ( z )dz = w( zN ) − w( zM ) (3.1.5)
2. An analytic function at inner points is expressed as contour integral (Cauchy integral)

1 w(ζ )d ζ
w( z ) = ∫
2π i C ζ − z
(3.1.6)
Equation (3.1.6) is derived using integral (3.1.4) to function w(ζ ) /(ζ − z ) and then
deforming the close curve to curve C and a small circle centered at point z. The latter is
equal to 2π iw( z ) because of pole singularity at point z. This point can be at the
boundary as well, but in this case the integral along an small-circular arc should be equal
to 2πε iw( z ) . Also, Cauchy integral (3.1.6) is generalized by factor ε to w(z). Then ε=1
for inner point, ε =1/2 for the point at smooth curve and ε=β/2π for a corner of angle β.
3. The analytic function at point z is equal to the average value on a circle centered at this
2π
1
2π ∫0
point, i.e. w( z ) = w( z + reiϕ )dϕ (3.1.7)
Some theorems are given below without proof.

Theorem 2. If function χ ( z ) is analytical but non-constant in domain D and continuous at
boundary B, then its module cannot have a maximum within the domain. If the analitic
function also differs from zero at any inner point then its module cannot have a minimum
within the domain.
According to this theorem velocity can have a maximum (or minimum) only at a boundary.
Theorem 3 (Cauchy - Liouville). If function f(z) is analytical over the whole z-plane and
bounded, then it should be constant.
63
The Cauchy - Liouville theorem is very useful for determining a function by given leading
part on the whole plane or by given singularities and zeros (see previous chapter).
3.1.3. Singularities of Transformation by Conformal Mapping
Analitic function z (ζ ) determines boundary Cz of definitional domain Dz of complex
variable z for given boundary Cζ of definitional domain Dζ in the ζ -plane, and vice versa.
Conformity of points z and ζ is conformal map one domain onto another one. The main
problem of conformal mapping is in determining transformation z (ζ ) by given boundaries
Cz and Cζ − see Cauchy theorem on transformation given below without proof.
Theorem 4 (Cauchy). Transformation z (ζ ) from the z-plane to the ζ -plane is determined
identically by satisfying three scalar conditions about the given inner points:
z (ζ 0 ) = z0 , arg z ′(ζ 0 ) = α 0 . (3.1.8)
Instead of conditions (3.1.8), three points on Cz or one point on the boundaries and one
point in domain Dz should be conformed to three given points on Cζ or to one point on the
boundary and one point in domain Dζ or some other three conditions. Transformation z (ζ )
is an analitic function inside domain Dζ , and vice versa, but its analytical property on the
boundary can be broken.
Consider some singularities on the boundary. Let point za ∈ Cz be a corner with
inner angle π μ , Fig.3.1.1a. As the role of signs, moving along the boundary is supposed
always to be positive if surrounding of this point is found on the left side. Let now
corresponding point ζ a ∈ Cζ be a vertex of angle πν , Fig.3.1.1d. The argument of vector
z − za , when passing clockwise around the point changes to −π μ while the argument of
vector ζ − ζ a changes to −πν. Thus, the asymptotic behavior in a small neighborhood is
z − za ≈ (ζ − ζ a ) μ /ν or ζ − ζ a ≈ ( z − za )ν / μ .
In case of Fig.3.1.1a, the sign of angle μ coincides
with the sign of angle ν , then the ratio μ /ν > 0 , and
transformation z (ζ ) − z0 has zero of the μ /ν -order.
In this case the conformity breaks. If angles μ and ν
coincide, then conformity holds out. From that reason
the parametric domains have been chosen in previous
sections. If point z0 is at infinity, Fig.3.1.1b, then ratio
μ/υ<0, and transformation z(ζ) has a singularity of the
| μ /ν |-order, i.e. if μ = −ν then the transformation
Fig.3.1.1. Boundary corners at:
has a simple pole at that point. If boundary C
(a) finite location); (b) infinity;
becomes a stripe of width h at infinity, then the
(c) a stripe at infinity; (d)
transformation has a logarithmic singularity as
parametric domain.
h
z≈ ln(ζ − ζ a ) , or ζ − ζ a ≈ eπν z / h . (3.1.9)
πν
64
3.1.4. Analytic Extension through Boundary

Let's at first determine symmetric complex points concerning a straight line and a circular
arc. Two points z and z ∗ are symmetric relative to a straight line if they are located at the
both sides and equally spaced from the line. For instance, complex number z = x + iy and its
conjugate z = x − iy are symmetrical numbers relative to real axis z ∗ = z . By fractionally
linear transformation one can transform the real axis to any strait line or to a circular arc and
obtain the following symmetrical points:
⎧ z symmetry relative to real axis,
∗ ⎪
z = ⎨− z symmetry relative to imeginary axis, (3.1.10)
⎪ 2
⎩ R / z symmetry relative to circle with radius R .
Let's show the symmetric point relative to the circular arc. Function
z = R (i − ζ ) /(i + ζ ) transforms an inner circle of radius R onto the upper half-plane of the
ζ -plane. Symmetric points on the ζ -plane are ζ and ζ , so that the symmetric point on
the z-plane relative to the circle is z ∗ = R(i − ζ ) /(i + ζ ) . The last fraction is equal to R / z .
Thus the validity of equation (3.1.10) is proven.
Theorem 5 (Rimann – Schwartz principle of symmetry). If part γz of boundary Cz is a
straight or circular segment, and the corresponding boundary γ w is also a straight or circular
segment then the analytic function at Dz can be analytically extended via γz to symmetric
⎧⎪ w( z ), if z ∈ Dz ∪ γ z ,
domain Dz∗ as w( z ) = ⎨ ∗ ∗
. (3.1.11)
⎪⎩ w ( z ) if z ∈ Dz .
Fig.3.1.2 shows boundary and points correspondence. Symmetric domains, D1∗ and D2∗ , as
well as primary domain D represent
quarter of a circle, while symmetric
domain D3∗ relative to the circle is an
outer sector with infinite point, which
is symmetric to the origin. The
definition domain of function w = f ( z )
is chosen as a quarter of circle to show
more possible cases. It should be noted
that the principle of symmetry was
proven for analytic functions only, but Fig.3.1.2. Symmetric points on the z-plane (a) and
it can also be used for singular correspondence of definition domain of function
functions [Gurevich, 1979]. f(z) by extension (b)
Let the function w( z ) inner domain has zero as ( z − za )λ , i.e. the whole circle of
small neighborhood at point za corresponds to a sector of angle πλ on the w-plane; its
vertex is at the origin. If appropriate boundaries of definition domain on the w-plane are
straight lines (here AB and AO), then the symmetrical domain will be also a sector of angle
πλ , thus the extended function will have also a zero as ( z − za∗ )λ . If the boundary as BO is a
circular arc then the symmetrical domain to this sector will be a sector of the same angle but
with vertex at infinity and the extended function will have at symmetrical point a singularity
65
as ( z − za∗ ) − λ . The negative values of λ are considered similarly.

Let meromorphic function w(z) has a simple pole at point z0 with complex factor A,
so the leading part is A /( z − z0 ) . If the boundary γz has a straight part of the real or
imaginary axis over which a real or imaginary part of function becomes constant then it will
have a pole ± A /( z − z0 ) at symmetrical point z0 , or ± A /( z + z0 ) at symmetrical point
− z0 . The symmetry extension together with Cauchy - Liouville theorem is very useful in the
hydrodynamics. It was also used in the previous sections for determining the derivative of
complex potential and complex velocity.
Sec. 3.2. Cavitating Flow around a Flat Sided Wedge

3.2.1. Introduction and Problem Statement
Cavitating flows around non-lifting two-dimensional bodies are of practical interest.
Surface-piercing struts, rudders of high-speed hydrofoils, and brake flaps of ships may be
operated in a ventilated cavitation. Similar flow can be observed by air smearing beneath
ship bottom. Therefore, it is important to understand the cavitation to consider in detail a
simple case of a symmetric cavitating flow. The cavitating flow around a wedge is a simple
problem for studying while gives insight into the processes and phenomenon. Many authors
have considered a wedge for explaining some effects in flow mechanics or some methods in
the theory [Tulin, 1953; Birkhoff, 1957; Gurevich, 1979].
Let's consider a cavitating flow past a wedge with flat sides with an attached cavity,
Fig.3.2.1a. The opening angle 2πμ and cavitation number σ are given. The speed at the
cavity is supposed to be equal to unity.
Fig.3.2.1. a) Cavitating flow past a wedge; b) complex w-plane; c) parametric ζ-plane.
Due to the symmetry, the flow domain on the z-plane corresponds to the w-plane with a
horizontal cut of finite length (Fig.3.2.1b). The logarithmic function of complex velocity
ω = ln(dw / dz ) satisfies following boundary conditions at cut sides:
Re ω = 0 on BC and AC ,
⎧πμ on OA, (3.2.1)
Im ω = ⎨
⎩−πμ on OB.
Outside the cut, it is analytical; at the cut's right end it has a singularity (2.1.8). The
boundary-value problem can be solved directly at the w-plane, but to consider arbitrary
66
symmetrical bodies we will use another parametric domain, a semi-disc of unit radius
(Fig.3.2.1c). Conformal mapping of the domain in complex potential onto the right semi-disc
of parametric ζ -plane (Fig 3.2.1b and c) reduces the problem of Eq. (3.2.1) to the
Schwartz’s problem for meromorphic function ω (ζ ) with simple pole at point of origin O.
Conformity of points is shown in Fig.3.2.1.
3.2.2. Parametric Solution
The transformation from the w-plane to the ζ-plane can be written in differential form by
dw ζ (1 − ζ 4 )
using all zeros and poles = wζ = 2 . (3.2.2)
dζ (ζ − c 2 ) 2 (c 2ζ 2 − 1) 2
Transformation w(ζ) can be found from here, but it can also be recreated directly from its
singularities. Indeed, the function w(ζ) has simple poles at point D(c) and also at
symmetrical points D′(1/ c ) , D′′(−c) and D′′′(−1/ c) . Therefore, it can be written as
A A A3 A4
w= 1 + 2 + + +B.
ζ − c ζ + c ζ − 1/ c ζ + 1/ c
The complex potential is totally real on the axes and on the circular arc and so the following
conditions are valid: w(ζ ) = w(ζ ) , w(−ζ ) = w(ζ ) , w(1/ ζ ) = w(ζ ) .
Hence, A1 = − A2 = m, A3 = − A4 = M = − m / c 2 .
The complex potential takes the form
⎧ 1 1 1 1 ⎫
w = m⎨ − − + ⎬. (3.2.3)
⎩ ζ − c ζ + c c(cζ − 1) c(cζ + 1) ⎭
The factor in (3.2.3) can be found by comparing both equations (3.2.2) and (3.2.3)
m = −1/ 4c(1 − c 4 ) . (3.2.4)
Function w(ζ) is totally real on the diameter AB, and so it can be extend over the whole disc
at that the Schwartz’s condition at the unit circle should be satisfied
⎧πμ on β ∈ [ - π ,0],
Im ω (eiβ ) = ⎨ . (3.2.5)
⎩−πμ on β ∈ [0, π ].
Besides, it has a simple pole at the origin ( ω ≈ w−1/ 2 ≈ ζ −1 ). Analytical solution inside the
circle were obtained using the Schwartz’s operator [Lavrent’ev & Schabat, 1977]
2π
i eiβ + ζ 1−ζ
ω0 =
2π 0 ∫ Im ω
e −ζ
iβ
d β = 2μ ln
1+ ζ
.
Consider auxiliary function χ = ω − ω0 which is analytical inside the circle except the origin
where it has a simple pole. Therefore, it can be expanded into Taylor series
b
χ = −1 + b0 + b1ζ + b2ζ 2 +
ζ
Since function χ becomes a pure imaginary value at diameter AB and a totally real value at
radius CO and also at circular arc AOB, thus the following symmetric conditions are valid:
χ (ζ ) = χ (ζ ) = − χ (−ζ ) = χ (ζ −1 ) .
Satisfying these conditions one obtains b−1 = b1 = b , b0 = b2 = = 0
67
1−ζ ⎛ 1⎞
Hence, ω = 2μ ln + b⎜ζ + ⎟ . (3.2.6)
1+ ζ ⎝ ζ⎠
The transformation from the z-plane to the ζ-plane can be found from differential equation
dz
= zζ = wζ e −ω (ζ ) . (3.2.7)
dζ
3.2.3. Determination of Parameters c and b
The functions (3.2.2) and (3.2.7) depend on two unknown parameters c and b. An equation
can be obtained from the condition for inlet velocity in infinity
ω (c) = ln V∞ = − ln 1 + σ ,
1− c ⎛ 1⎞ 1
or 2 μ ln + b ⎜ c + ⎟ = − ln(1 + σ ) (3.2.8)
1+ c ⎝ c⎠ 2
Another equation is obtained by “closed” condition (2.1.6d) which can be written similar to
⎡d d ω (ζ ) ⎤ b(1 − c 2 ) 4μ
(2.3.12) lim ⎢ ln wζ (ζ )(ζ − c) 2 − ⎥ = + =0. (3.2.9)
ζ → c dζ
⎣ dζ ⎦ c 2
1 − c2
Equations (3.2.8) and (3.2.9) make a set of closed equations for parameters b and c.
2
⎛ c ⎞
Parameter b is expressed as b = −4μ ⎜ 2 ⎟
. (3.2.10)
⎝1− c ⎠
Then equation (3.2.8) yields an equation for parameter c only
1 − c 2c(1 + c 2 ) ln(1 + σ )
ln − = . (3.2.11)
1 + c (1 − c 2 ) 2 4μ
It is seen that equation (3.2.11) at σ > 0 is solvable and has the only solution because the
left-hand function is continuous within interval ( 0 < c < 1 ) and has different signs at the ends
of the interval. Since its derivative −2(1 + 4c 2 + 3c 4 ) /(1 − c 2 )3 is negative over the whole
interval, hence equation (3.2.10) has a single solution. That solution can be calculated by any
ln(1 + σ )
numerical methods. Denote χ= (3.2.12)
16μ
For a small parameterχ, the desired solution can be expanded in asymptotic series. Then
parameter c should be small, and so equation (3.2.11) can be expanded in asymptotic series
⎛ 5 13 ⎞
−4c ⎜ 1 + c 2 + c 4 + ⎟ + 4 χ = 0 .
⎝ 3 5 ⎠
Substituting parameter c with a power series and equating coefficients at the same power of
⎛ 5 26 4 ⎞
χ , one obtains power series as c = χ ⎜1 − χ 2 + χ ⎟, (3.2.13)
⎝ 3 15 ⎠
⎛ 4 176 4 ⎞
and from equation (3.2.10) b = −4μχ 2 ⎜1 − χ 2 + χ + ⎟. (3.2.14)
⎝ 3 45 ⎠
Equation (3.2.13) shows that abscissa c approaches zero as either parameter χ or cavitation
number σ approaches zero, and the trailing points C1 and C2 in the z-plane approach infinity.
This implies that the cavity becomes of infinite length by σ=0 and the pressure in cavity, p0,
becomes the same as the pressure at infinity, p∝,. As was noted above that model is named
68
Kirchhoff’s flow. Asymptotic series (3.2.13) and (3.2.14) verify possibility of continued
conversion from cavitating flow to Kirchhoff’s flow.
3.2.4. Geometric Characteristics of Cavity
Compatibility of points on the wedge side OB ( z = seiπ μ ) and at the correspondent circular
arc ( eiβ ) is found integrating function (3.2.7) along the arc OB
β
cos β sin1+ 2 μ β (1 + cos β ) 2 μ −2b cos β
s ( β , b, μ ) = 4 ∫ e dβ . (3.2.15)
0
(1 − 2c 2 cos 2 β + c 4 ) 2
The length of the wedge side is l = s (π / 2, b, μ ) (3.2.16)
All numerical results are presented dimensionless as ratios of the wedge length.
d = 2lsinπμ (3.2.17)
Integrating equation (3.2.7) along the imaginary axis one obtains the cavity boundaries. The
upper boundary is determined by functions
1 1
x(η ) = ∫ f (τ ) cos θ (τ )dτ + l cos πμ , y(η ) = ∫ f (τ )sin θ (τ )dτ + l sin πμ , (3.2.18)
η η
τ (1 − τ 4 ) ⎛1 ⎞
where f (τ ) = , θ (τ ) = 4μ atanτ + b ⎜ − τ ⎟ (3.2.19)
(τ + c ) (c τ + 1)
2 2 2 2 2 2
⎝τ ⎠
The coordinates of spiral center C2 are x2 = x(0), y2 = y (0) . Determine the cavity's length
and width, respectively, as L = x2 − l cos πμ , H = 2y max = 2 y (η0 ) , (3.2.20)
where η0 is the solution of equation θ (η0 ) = 0 .
Consider small value of variable η , then the variable τ will be small also, and then
f (τ ) ≈ τ / c 4 , θ (τ ) ≈ b / τ . Let the origin of axes ( x, y ) is at the spiral center, i.e.
η η
x = x − x2 , y = y − y2 . Then x(η ) = − ∫ τ cos(b/τ )dτ , y (η ) = − ∫ τ sin(b/τ )dτ
0 0
Substituting τ = −b / t and integrating along interval ( -b/η, ∞ ) we obtain the spiral with its
b2 ⎛ cos (b / η ) sin (b / η ) ⎞
x(η ) = 4 ⎜
−ci(−b / η ) − + ⎟,
2c ⎝ (b / η ) 2 b /η ⎠
center C2 (3.2.21)
b2 ⎛ sin( b / η ) cos (b / η ) ⎞
y (η ) = 4 ⎜ si(−b / η ) − − ⎟,
2c ⎝ (b / η ) 2 b /η ⎠
where ci(t) and si(t) are integral cosine and sine [Gradstein &Ryzhik 1962]. Its power
∞
t 2k
ci(t ) = C + ln t + ∑ (−1) k ,
k =1 2k (2k )!
series are (3.2.22)
π ∞ ( −1) k +1 t 2 k −1
si(t ) = − + ∑ ,
2 k =1 (2k − 1)(2k − 1)!
where C = 0.5772 is the Euler constant.
Equations (3.2.21) as distinct from (3.2.18) allow calculating a spiral with very high
precision. For exact description of the cavity's total boundary, functions (3.2.18) and (3.2.21)
should be connected at any point, e.g. at point η with θ(ηc)=−π. The trailing spiral at point
C2 for cavitation number σ=0.75 and wedge angel μ=1/4 is plotted in Fig.3.2.2 where the
69
dotted line was calculated by functions (3.2.18).
Fig.3.2.2.
Trailing spiral for
σ = 0.75 and
μ=1/4 in different
scales.
Fig.3.2.2 shows a
very high
convergence of the trailing spirals. It cannot be drawn in the normal scale. For practical
purposes it suffices using equations (3.2.18) only; equations (3.2.21) are mostly of
theoretical interest.
Consider now the cavity boundary close to the detachment point B. Substituting
τ = 1 − t and then expanding equations (3.2.18) into power series with respect to variable t
yields the following expressions:
4t ⎡ ⎛3 ⎞ ⎤
x′(t ) = 2 4 ⎢
cos πμ + ⎜ cos πμ − (b − 2 μ )sin πμ ⎟ t ⎥ ,
d (1 + c ) ⎣ ⎝2 ⎠ ⎦
(3.2.23)
4t ⎡ ⎛ 3 ⎞ ⎤
y ′(t ) = sin πμ + ⎜ sin πμ + (b − 2μ ) cos πμ ⎟ t ⎥ .
d (1 + c 2 ) 4 ⎢⎣ ⎝2 ⎠ ⎦
(b − 2 μ )(1 + c 2 )d
Hence, the curvature is asymptotically expressed as κ (t ) = , t →0.
4t
From (3.2.23) the curvature at detachment point B(t=0) approaches infinity. Fig.3.2.3 shows
cavity shapes at cavitation number σ =0.75 and different open wedge angles of 2πμ. It is
seen from Fig.3.2.3b that all open angles, μ > 3/4, give almost the same results for the cavity
as well as for the drag. As the special cases the Kuznetsov flow is resulted at μ = 1 and the
single spiral is arrived at μ→∞.
Fig.3.2.3.
Cavity shapes for
the same unit bases
and different
wedge angles if
attachment points
are jointed.
3.2.4. Wedge Drag
The wedge drag can be found integrating the pressure along the wedge sides or using
equation (2.1.15) which takes the form of D=y2, where y2 is the ordinate of spiral center C.
Instead of integrating along the cavity boundary it is more accurate to integrate along a circle
of a small radius δ ( ζ = δ + δ eit , −π < t < π ); the circle touches the imaginary axis at the
origin. The drag coefficient with respect to the wedge base and inflow speed is
70
π
2D (1 + σ )
CD = = Im ∫ zζ (δ + δ eit )δ eit idt , δ = 0.1 (3.2.24)
ρV∞ d
2
d −π
The drag can be found also by integrating the pressure along the cheeks similar to equations
(2.3.21) or (2.3.23) using function (3.2.15)
⎛ s (π / 2, −b, − μ ) ⎞
CD = ⎜1 − ⎟ (1 + σ ) (3.2.24a)
⎝ s (π / 2, b, μ ) ⎠
Inasmuch as integrals of both expressions (3.2.24) and (3.2.24a) should be computed
numerically, one can expect that numerical results may differ, but numerous calculations
show that both expressions give the same results of sufficiently high accuracy.
3.2.6. Asymptotic Expressions
Consider now a small cavitation number assuming σ << μ, so that given parameter χ and
parameters b and c are also small. But simple expressions cannot be obtained for an arbitrary
angle. Aspiring to obtain simple expressions we shall consider angle μπ = π/2 which
corresponds to the plate normal to inflow speed. Due to equations (3.2.13) and (3.2.14)
parameters b and c 2 are small of the same order, so that the integrands in equation (3.2.24)
can be expanded into Taylor series and then integrated as
4+π ⎡ 16 + 5π 2 ⎤
s (π / 2, b, μ ) = 2 2 ⎢
1+ σ (1 − σ ) + O(σ 4 ) ⎥ ,
(1 + c ) ⎣ 32(4 + π ) ⎦
4 −π ⎡ 16-5π ⎤
s (π / 2, −b, − μ ) = 2 2 ⎢
1+ σ 2 (1 − σ ) + O(σ 4 ) ⎥ .
(1 + c ) ⎣ 32(4 − π ) ⎦
Thus the drag coefficient (3.2.24) is reduced to
2π ⎡ σ 2 (1 − σ ) ⎤
CD = ⎢1 + + O(σ 4 ) ⎥ (1 + σ ) (3.2.25)
π + 4 ⎣ 8(π + 4) ⎦
Substituting c = 0 into functions (3.2.19) yields a singularity as τ −3 so that the cavity length
1
τ dτ 1 1
(3.2.20) should be found as L = x(0) ≈ ∫ 2 ≈ 2 = (3.2.26)
0
(τ + c )
2 2
2c 2χ 2
Since the asymptotic solution of equation θ (η0 ) = 0 is η0 ≈ χ , then, as follows from
(3.2.20), the asymptotic expression for cavity width is
H = 2 y (η0 ) ≈ 2 / χ (3.2.27)
In view of asymptotic value χ≈σ/8 and equation (3.2.25), the dimensionless length and width
of cavity are expressed asymptotically as
L 16 8 H 8 4
≈ 2 = CD0 , ≈ = CD0 , σ →0,
d σ (4 + π ) πσ 2
d σ (4 + π ) πσ
where CD0 = 2π /(4 + π ) is the drag coefficient of the vertical plate by Kirchhoff flow. A
similar asymptotic expression for slender bodies has been obtained by Tulin [1954]. Due to
these asymptotic expressions the cavity length and width could be expressed as
πσ 2 L πσ H
L0 = o
, H0 = , (3.2.28)
8C D d 4CDo d
71
where CDo is the drag coefficient of a wedge for zero cavitation number (Kirchhoff flow).
That problem has been considered by DK Bobylev in 1881 and M Rethy in 1895, see
[Gurevich, 1979]. All the required expressions can be obtained from the above-mentioned
equations by formally substituting b = 0 and c = 0.
Simple reasons show that the drag coefficient approaches σ or σ + 1 as the angle
πμ approaches zero or π , respectively. In the first case, the speed on the wedge cheeks is
equal to the inflow speed, v∞ , while in the second case it is zero. Consider in detail a thin
wedge if its angle is πμ and cavitation number σ are small parameters of equal order. In this
case parameter χ = σ /16μ can be of an arbitrary value, parameter c ∈ (0,1) is a solution of
equation (3.2.11) while parameter b is of a small value of the same order as μ due to
(3.2.10). The logarithmic function ω (ζ ) is of a small value over the whole domain except
small neighborhood at points ζ = 0, ± 1 . Expanding complex velocity wz = eω (z) into series
about function ω (ζ ) , the derivative of transformation (3.2.7) can be written as
⎛ ω2 ⎞
zζ = ⎜1 − ω + − ⎟ wζ . (3.2.29
⎝ 2 ⎠
The transformation from the z-plane to the ζ -plane keeping the first term takes the form
z (ζ ) = w(ζ ) + const (3.2.30
Hence, wedge cheek length l and the L/l ratio are expressed, respectively,
l = z (i ) − z (1) = 2 /(1 − c 4 ) 2 , (3.2.31
L z (0) − z (i ) (1 − c 2 ) 2
and = = . (3.2.32
l l 4c 2
From the latter, equation (3.2.11) can be expressed as a function of this ratio
L / l +1 +1 2 L / l +1 σ
ln + = . (3.2.33)
L / l +1 −1 L/l 2μ
For calculating the cavity boundary, one more term in (3.2.28) should be kept so that the
ordinate of cavity shape is calculated from the set of differential equations and the abscissa
dy = − Im ω (η i ) wζ (η i )idη , x(η ) = w(iη ) . (3.2.34)
The distance between two spiral centers and the drag can be obtained by contour integral of
function (3.2.29) about the origin. In this case the third term in (3.2.29) should be kept to
1 b2
calculate the drag of a thin wedge D = Im ∫ ω 2 wζ d ζ = π 4 . (3.2.35)
4 ζ =0 c
CD = 4μ ⎡1 + ( L / l ) ⎤ .
−1
Hence the drag coefficient is (3.2.36)
⎣ ⎦
Fig.3.2.4 shows distribution of velocity V over the wedge cheek at various open angles
β = 2πμ . Plotted in Fig.3.2.5 is drag coefficient CD as a function of open angle for some
cavitation numbers σ . It shows minimums of the curves for small angles. Relative lengths
and widths of cavity are plotted in Fig.3.2.6.
72
Fig.3.2.4. Velocity distribution over Fig.3.2.5. Dependence of drag

the wedge at different angles. coefficient on wedge angle.
Fig.3.2.6. Dependence of cavity length and width on cavitation number and wedge angle.
Sec. 3.3. A Wedge in a Channel and in a Jet

Cavitating flows around a wedge in a channel and in a jet are considered as well as some
asymptotic cases. The flow problem can be solved using a parametric domain similar to that
in Sec. 3.2 [Terentiev, 1981a], but here the problem will be solved in a domain of complex
potential. This presentation seems prima facie complicated but it is actually more convenient
for asymptotic transformations and a numerical calculations. Inasmuch as the previous
problem has been discussed in detail, here there will be mostly a consideration of problem
statement and analytical solution in general, as well as calculation results.
The cavitating flow in a channel around a wedge is shown in Fig.3.3.1a, b. Without losing
generality, the width of stripe of the defining domain of complex potential w is assumed to
be equal to π / 2 (Fig.3.3.1c); unlike the preceding section, the wedge base, the cavity width,
and the channel (and jet) width are denoted as 2d, 2H and 2h, respectively; the other notation
retains. All linear parameters (cavity length L, width H, and drag D) are made dimensionless
with respect to base 2d. Also, the open angle of wedge β=2πμ and ratio h/d are given, the
73
others are to be found. Intervals (0, a)
and (a, b) correspond to the side of
wedge and cavity, respectively;
parameters a and b are to be determined.
The value problem of the
logarithmic function of complex velocity
ω=lnwz is represented by the following
boundary conditions at the stripe.
For the channel:
• within interval (0, a) at a wedge side
Im ω = −πμ , w ∈ (0, a ) , (3.3.1)
• at the cavity boundary (speed at
cavity boundary is assumed to be
unity) Re ω = 0, w ∈ (a, b) , (3.3.2)
• in infinity of the channel
Fig.3.3.1. Cavitating flow in a channel (a) ω (−∞) ≡ ω (∞) = −0.5ln(1 + σ ) , (3.3.3)
and in a jet (b); the plane of complex
• at a channel wall C1C2 and on the
potential (c).
axis beyond interval (0, b)
Im ω = 0 , (3.3.4)
For the jet:
at the cavity boundary (speed at the jet boundary is assumed to be unity)
Re ω ( w) = ln(1 + σ ) / 2, w ∈ (a, b) , (3.3.5)
at the free boundary C1C2 Re ω ( w) = 0 , (3.3.6)
at a wedge side as (3.3.1) and on the axis as (3.3.4).
The free boundary at infinity should have the same width h both at the left and right sides,
i.e. function ω should satisfy the following additional condition
∞
∫ Im ω (ϕ + π i / 2)dϕ = 0 .
−∞
(3.3.7)
Besides, function ω ( w) has the only singularity as ( w − b) −1/ 2 at point B(b).
3.3.2. Analytical Solution of the Problems

Using the solution of mixed value problem for a stripe given in Appendix 1 (Eqs. A1.3,
A1.5, A1.9 and A1.10), in which constant B vanishes and parameter c = (b − a ) / 2 , the
solution is obtained as follows:
a
sinh( w − a) sinh(t − b) sinh(t − w + c)
sinh( w − b) ∫0 sinh(t − a ) sinh(c)sinh(t − w)
for the channel ω = −μ dt . (3.3.8)
Integral (3.3.8) can be calculated analytically

⎛ sinh b sinh( w − a ) − sinh a sinh( w − b)
ω = μ ⎜ ln +
⎜ sinh b sinh( w − a ) + sinh a sinh( w − b )
⎝
(3.3.9)
1 sinh( w − a) sinh b − sinh a − 2sinh c sinh a sinh b ⎞
+ ln ⎟
2sinh c sinh( w − b) sinh b − sinh a + 2sinh c sinh a sinh b ⎟⎠
74
for the jet
1 sinh( w − a )
ω= ×
π sinh( w − b)
(3.3.10)
⎛ b
sinh(b − t ) dt
a
sinh(b − t ) dt ⎞
× ⎜⎜ 0.5ln(1 + σ ) ∫ − πμ ∫ ⎟
⎝ a
sinh(t − a ) sinh(t − w) 0
sinh( a − t ) sinh(t − w) ⎟⎠
The transformation from the w-plane to the z-plane, as ever, is written in differential form
dz = e −ω ( w) dw . (3.3.11)
It should be noted that the function (3.3.9) and (3.3.10) is multi-valued; the respective
branches could be chosen considering flow character at the boundaries. Namely, the first
term in (3.3.9) should have a positive imaginary part within interval (a, b) while square root
by the second term and in (3.3.10) should be negative.
3.3.3. Determination of Interval Extremities a and b
Integrating equation (3.3.11) in interval (0, a) one finds the wedge cheek length
a
l = ∫ | e − ω (ϕ ) | d ϕ (3.3.12)
0
Since the quantity of fluid flowing in the channel and jet is equal to π , or to 2hv∞ , then
h = π / 2v∞ (3.3.13)
Given ratio d / h = l sin πμ / h together with (3.3.13) and (3.3.12) yields an equation in
a
π d
∫| e
− ω (ϕ )
integral form | dϕ = (3.3.14)
0
2v∞ sin πμ h
Due to the above-assumption the inlet velocity is
⎧⎪1/ 1 + σ for the channel,
v∞ = ⎨ (3.3.15)
⎪⎩1 for the jet.
The second equation for the channel is obtained satisfying condition (3.3.3)
1 ⎧⎪ c sinh b + ec sinh a −c sinh b + e − c sinh a ⎫⎪ ln(1 + σ )
⎨ e ln − e ln ⎬= . (3.3.16)
sinh c ⎩⎪ sinh b − ec sinh a sinh b − e − c sinh a ⎭⎪ μ
Equations (3.3.14) and (3.3.16) constitute the required equation set for unknown parameters
a and b. In case of a jet, condition (3.3.7) yields another equation, where imaginary part is
1 cosh(ϕ − a )
Im ω (ϕ + π i / 2) = ⋅
4 cosh(ϕ − b)
(3.3.17)
⎛ b
sinh(b − t ) dt
a
sinh(b − t ) dt ⎞
⋅ ⎜⎜ ln(1 + σ ) ∫ − 2πμ ∫ ⎟
⎝ a
sinh(t − a ) cosh(t − ϕ ) 0
sinh(a − t ) cosh(t − ϕ ) ⎟⎠
As one can see, equation (3.3.7) with (3.3.17) is expressed by double integrals, indeed,
without any singularities. The integrand of equation (3.3.10) has a Cauchy singularity within
interval (0, b) and integrable singularities at points a and b, so that numerical calculations
present some difficulties. There are numerical methods applicable to singular integrals, but
they need considerable manipulations which may not be efficient. The difficulties due to
singularities can be bypassed by calculating at inner point ϕ + iδ close to the ϕ -axis
75
( |δ | << 1 ). For instance, the wedge side length, drag, and the cavity boundary can be
calculated along the straight line w = ϕ + iδ instead of the ϕ -axis. A comparison of
calculated data by δ = 10−4 with exact boundary values confirms a very high accuracy.
3.3.4. The Drag Coefficient
The drag can be obtained integrating the pressure along the wedge cheeks, then the drag
coefficient is calculated by
2D ⎛ 1 ⎞
a
CD = 2 = ⎜ 1 − ∫ | eω (ϕ ) | dϕ ⎟ (1 + σ ) for the channel, (3.3.18)
ρ v∞ d ⎝ l 0 ⎠
a
1 ω (ϕ )
or CD = 1 + σ − ∫ | e | dϕ for the jet. (3.3.19)
l0
The drag coefficient in both cases can also be transformed to
CD = y B (1 + σ ) / d (3.3.20)
It should be noted that at the given ratio d/h and open angle β, cavitation number for the jet
can be arbitrarily small and positive while for the channel the cavitation number cannot be
smaller than respective value σ0, which corresponds to flow around a cavity of semi-infinite
length. As b approaches infinite ( b → ∞ ) then another parameter can be found analytically
1 ⎡ ⎛ (1 + σ 0 )1/ 2 μ − 1 ⎞ ⎤
2
from (3.3.16) a0 = − ln ⎢1 − ⎜ ⎟ ⎥. (3.3.21)

2 ⎢ ⎝ (1 + σ 0 )1/ 2 μ + 1 ⎠ ⎥
⎣ ⎦
e w/ 2
sinh( w − a0 ) − sinh a0
The function (3.3.9) is transformed to ω0 ( w) = μ ln w / 2 . (3.3.22)
e sinh( w − a0 ) + sinh a0
The critical cavitation number, σ0, is calculated from equation (3.3.15) using
(3.3.21) and (3.3.22). Equation (3.3.18) for drag coefficient is valid in this case as well.
Obviously, the drag can be obtained from the law of momentum [Gurevich 1969], and the
h
( )
2
drag coefficient is expressed as CDo = 1+ σ 0 −1 . (3.3.23)
d
It is interesting to notice the ordinate yB of spiral center becomes certain value which is
determined from (3.3.18) based on the liquid conservation law ( h = (h − H ) 1 + σ 0 ).
Comparing both equations (3.3.15) and (3.3.23) yields a relation between asymptotic values
of distance yB and cavity width H , as yB h = H 2 . This implies the flow around semi-
infinite cavity in a channel which approaches asymptotic from the right part of the channel.
The jet flow also produces two asymptotic flows as cavitation number approaches zero: one
is Kirchhoff’s flow around the wedge in a jet, another is the Kirchhoff’s flow of two single
spirals.
Solution of the flow of unbounded fluid can be obtain from function (3.3.9)
approaching parameters a and b to zero. Denoting b = b′a , w = w′a and approaching
a → 0 , one can obtain from (3.3.9) the solution of the unbounded domain problem as
⎛ b( w − 1) − w − b 2 b w − 1 ⎞
ω ( w) = μ ⎜ ln − ⎟. (3.3.24)
⎜ b( w − 1) + w − b b − 1 w − b ⎟⎠
⎝
76
The only unknown parameter, b, is calculated from equation
b −1 b ln(1 + σ )
ln −2 + =0. (3.3.25)
b +1 b − 1 2μ
Eqs. (3.3.24) and (3.3.25) constitute another solution of the problem discussed in Sec. 3.1.
Some results for the Kirchhoff’s model of a wedge in a channel are plotted in Fig.3.3.2. The
left-hand plot shows critical cavitation number vs. ratio d/h for some open angle, while the
effect of critical cavitation number on drag coefficient is shown in the right-hand plot.
Calculation results for a wedge with opening angle β = 30 in a channel (solid lines) and in
a jet (dash lines) are plotted in Figs.3.3.3 and 3.3.4
Fig.3.3.2. Cavitation number (a) and drag coefficient (b) for a wedge with semi-infinite
length cavity in a channel.
Fig.3.3.3.
Cavity
length
(left) and
width
(right).
77
Fig.3.3.5. Cavity shapes for a

wedge in a jet.
Fig.3.3.4. Drag coefficient of a wedge of opening angle β = 30

in channel (solid lines) and jet (dash lines).
Comparisons of flows in channel and jet show that the channel enlarges a cavity while the jet
decreases it. The drag is decreased in channel, while in jet it is slightly increased at high
cavitation numbers and decreased at low numbers. On the whole, jet has little effect on drag.
Fig.3.3.5 shows configurations of cavity and free boundary for different cavitation
numbers and jet widths. The main part of cavity and its free boundary have a form of
circular arcs and correspond to flow of point vortex. The flow over the wedge side is similar
to the undisturbed flow at speed v∞ parallel to the side, so the drag coefficient approaches the
cavitation number.
3.3.6. Asymptotic Approaches
Let the wedge angle πμ and cavitation number σ be infinitesimals of the same order, so
that the ratio σ / πμ may be an arbitrary finite number. Functions (3.3.9) and (3.3.10)
becomes infinitesimal of the same order over the whole w-plane except small neighborhoods
at points O and B. Excluding these neighborhoods the complex velocity for infinitesimal
function ω ( w) can be expanded as eω = 1 + ω + O ( μ 2 ) and then the z-plane due to
(3.3.11) differs from the w-plane by an infinitesimal function. In this case the function
ω ( w) ≈ ω (z ) is the perturbation speed; it is determined by function (3.3.9) or (3.3.10) for
channel or jet, respectively, in which the variable z should replace the potential w. The
width of channel and jet are assumed to be equal to π / 2 , and parameter a is the wedge
length. Parameter b is the only one to be calculated from equations (3.3.16) or (3.3.17).
Plotted in Fig.3.3.6 are numerical results [Terentiev, 1971a] including the unknown
parameter b versus given ratio σ / πμ (Fig.3.3.6b).
78
Fig.3.3.6. Linearized problem of cavitating flow of a thin wedge in a channel (solid lines)
and a jet (dashed lines).
Sec. 3.4. Cavity with a Negative Cavitation Number

When streamlining a profile with a blunt trailing part a cavity can form behind it with
pressure inside the cavity exceeding the pressure in the ambient environment. As the liquid
moved away occupies small volume and is inactive, the pressure can be assumed constant,
i.e. velocity at the cavity boundary and in the flow is constant and reaches a minimal value.
The cavitation number in this case is negative. Flow of a body with a final cavity of the
maximal pressure was considered by Kolscher [1940]. Earlier similar flow was investigated
by Chaplygin [1899] who replaced a forward critical point with a final cavity. Similar flows
can be found in many models of continuum mechanics (hydrodynamics, filtration, the theory
of explosion, etc.) by many authors. More information can be found in [Gurevich 1979] and
in special literature. Here a cavity with maximal pressure will be considered from the
viewpoint of cavitating flow. As the pressure gradient is directed towards the cavity, its
boundary is concave and has an isolated cuspidal edge.
3.4.1. Problem Statement of Longitudinal Flow around a Wedge
Consider a flow along a wall with a wedge-shaped jut
and an adjacent cavity; pressure in the cavity is
assumed maximal and constant (see Fig.3.4.1a). The
w-plane in this case is an upper half-plane as in
previous sections (Fig.3.4.1b), and the problem can be
solved as a mixed-value problem, see Appendix 1.
Fig.3.4.1.
Flow in the z-plane (a) and in the w-plane (b).
79
Let the values of potential at points D and O be 0 and 1,and at points B and C be unknown
parameters –b and –c, respectively. The boundary values of logarithmic function
⎧0 on AB and OA,
⎪
ω = ln v − iθ are as follows: θ = ⎨πμ on BC , (3.4.1)
⎪π ( μ + v − 1) on CD,
⎩
ln v = 0 on DO .
Besides, function ω ( w) is bounded at all points of flow domain and at the boundary.
3.4.2. Solution of Mixed-Value Problem

From equation (A1.8), the boundary value problem solution is expressed as
w ⎛ ⎞
−c 0
t − 1 dt t − 1 dt
ω= ⎜⎜ ∫
μ + ( μ + ν − 1) ∫ + B ⎟⎟ . (3.4.2)
w − 1 ⎝ −b t w − t −c
t w−t ⎠
Equation (3.4.2) determined in such a way has a singularity at point O(1). This singularity
can be eliminated if the expression in parenthesis vanishes at w = 1. Then the constant B is
determined as a function on unknown parameters b and c
( 1+ c + c ) + 2(μ + ν )ln
B (b, c) = 2μ ln ( )
1+ c + c . (3.4.3)
( 1+ b + b)
When variable w approaches infinity, constant B will be equal to the value of function
(3.4.2) at infinity B(b, c) = ω (∞) = − ln(1 + σ ) / 2 . (3.4.4)
Transformation from the w-plane to the z-plane can be obtained by integral
w
z (w) = ∫ e −ω ( w′ ) dw′ , (3.4.5)
−b
which should have a single value, i.e. the residue of its derivative at infinity should be zero.
Expanding integrand for large variable w in to Taylor series yields a factor to term w−1 :
c b
t +1 t +1 1
A(b, c) = ( μ + ν − 1) ∫ dt + μ ∫ dt + B (b, c) (3.4.6)
0
t c
t 2
Then the single-value condition is written as A(b, c ) = 0 . (3.4.7)
Functions (3.4.4) and (3.4.7) form a set of equations for unknown parameters b and c.
3.4.3. Notes on Computation
The integrals in equation (3.4.2) can be expressed analytically but the function (3.4.2) can
become a complicated multi-value function of logarithms, fractional degree and so on.
Choosing the appropriate branches can sometimes be a tricky business. Since the upper half-
plane is the defining domain of complex potential w, the arguments of functions w, w – 1
and w – t change in interval ( 0, π ). Then the integrand singularities can be eliminated if
w + δ i is substituted instead of w, where δ is assumed to be a positive and sufficiently
small value (see App. 5). As a result, analitic function ω ( w + iδ ) will differ from function
ω ( w) by a value about δ over the entire flow domain except the infinitesimal neighborhood
of points b and c. Then the following expressions may be used.
80
−c
∫e
− ω (ϕ + iδ )
Length of side BC: l= dϕ . (3.4.8)
−b
0
∫e
− ω (ϕ + iδ )
Length of side CD: d= dϕ . (3.4.9)
−c
ϕ
x(ϕ ) = l cos(πμ ) + d cos [π ( μ + ν − 1) ] + Re ∫ e −ω ( t + iδ ) dt ,
0
Cavity boundary: ϕ
. (3.4.10)
y(ϕ ) = l sin(πμ ) + d sin [π ( μ + ν − 1) ] + Im ∫ e −ω ( t + iδ )
dt
0
1 1
Cavity length and thickness: L = Re ∫ e − ω ( t + iδ )
dt , H = − Im ∫ e−ω ( t + iδ ) dt . (3.4.11)
0 0
The additional item, iδ , allows to avoid the singularities of Cauchy type and at corner C and
interval edges. Indeed this direct approach takes a bit more computing time but the results
can be obtained reliably. All numerical results have been obtained by δ = 10−6 .
The conformal mapping was used for the above-mentioned problems. Here the first quadrant
as parametrical domain of the auxiliary ζ -plane is used
(Fig.3.4.2). The conformity of points is shown in
Figs.3.4.1a and 3.4.2. The derivative of complex
potential wζ becomes a real value on the real axis and a
pure imaginary value on the imaginary axis so it can be Fig.3.4.2. The auxiliary ζ -plane.
extended analytically over the whole ζ -plane, at that the
double pole at point ζ = a shall cause another double pole at symmetrical point ζ = − a . In
view of simple zero at the origin, the derivative of complex potential is expressed as
ζ
wζ = (3.4.12)
(ζ − a 2 )
2 2
Complex speed wz = dw / dz can be found from logarithmic function ω satisfying boundary

condition (3.4.1). As a result, function ω has logarithmic singularity at point B(b) because
of discontinuity of the first kind with increment πμ i , i.e. ω ≈ μ ln(ζ − b) in infinitesimal
neighborhood of point B. Hence the complex velocity at that point has the factor
wz = eω ≈ (ζ − b) μ . On the other hand, the real part of function ω vanishes on the imaginary
axis, η , so that it can be extended analytically through that axis over the whole upper half-
plane of the ζ -plane. Function ω should have at symmetric point B′(−b) logarithmic
singularity as ω ≈ − μ ln(ζ + b) , so that complex speed should have another factor
wz ≈ (ζ + b) − μ . Similarly, the complex speed should have factors (ζ − 1)ν −1 and (ζ + 1)1−ν .
These factors can be figured out from changes the complex velocity argument in going
around points ζ = ±1 anticlockwise. For instance, the increment of argument (ζ − b) by
anticlockwise round is equal to π , but increment of argument of complex velocity is equal
81
to πμ . Hence the speed function can have only factor (ζ − b) μ . Since the absolute
magnitude is constant, additional factor at symmetrical point should be only (ζ − b) − μ
[Gurevich 1979]. Collecting all factors on whole ζ -plane yields the complex speed as
wz = N (ζ − b) μ (ζ + b) − μ (ζ − 1)ν −1 (ζ + 1)1−ν .
Constant factor N is determined satisfying a known value of speed at any pint, e.g. at point
D(ζ → ∞) : wz = eπ (1− μ −ν ) i . Therefore,
μ 1−ν
⎛ ζ − b ⎞ ⎛ ζ +1⎞
π (1− μ −ν ) i
wz = e ⎜ ⎟ ⎜ ⎟ (3.4.13)
⎝ ζ + b ⎠ ⎝ ζ −1 ⎠
The inflow speed is determined substituting ζ = a
μ 1−ν
⎛ b − a ⎞ ⎛1+ a ⎞
V ( a, b) = ⎜ ⎟ ⎜ ⎟ . (3.4.14)
⎝ b + a ⎠ ⎝1− a ⎠
Transformation from the ζ -plane to the z-plane is obtained from the differential equation
dz wζ
= . (3.4.15)
dζ wz
As was shown in Sec. 2.3, the single-value condition for transformation z (ζ ) is represented
by equation (2.3.12) which requires to vanish the function
d 2μ b 2(1 − ν )
f (a, b) = lim ln zζ (ζ )(ζ − a ) 2 = 2 − (3.4.16)
ζ → a dζ b −a 2
1 − b2
Also, unknown parameters a and b for the given cavitation number are calculated from
1
equations V ( a, b) − = 0, f (a, b) = 0 . (3.4.17)
1−σ
And then all geometric characteristics are found integrating equation (3.4.15), namely,
Lengths of sides BC and CD are as follows
μ 1−ν
1
ξ ⎛ ξ + b ⎞ ⎛1−ξ ⎞
l=∫ 2 2 2 ⎜ ⎟ ⎜ ⎟ dξ (3.4.18)
b
(ξ − a ) ⎝ ξ − b ⎠ ⎝ 1 + ξ ⎠
1 μ 1−ν
t ⎛ 1 + bt ⎞ ⎛ 1 − t ⎞
d=∫ 2 2 2 ⎜ ⎟ ⎜ ⎟ dt (3.4.19)
0
(1 − a t ) ⎝ 1 − bt ⎠ ⎝ 1 + t ⎠
The integrand in the second integral is obtained substituting ξ = 1/ t because of infinite
interval ξ ∈ (1, ∞) . The cavity boundary corresponds to imaginary axis η ∈ (0,∞) so that the
substitution of η = (1 − u ) / u has been used. Then the tangent angle to cavity boundary is
⎛1− u ⎞ ⎛1− u ⎞
θ (u ) = 2μ atan ⎜ ⎟ − 2(1 − ν )a tan ⎜ ⎟ (3.4.20)
⎝ bu ⎠ ⎝ u ⎠
and the derivative of the potential with respect to variable u is determined by function
u (1 − u )
g (u ) = . (3.4.21)
[(1 − u ) 2 + a 2 u 2 ]2
Hence, the cavity boundary is calculated by integrals
82
u
x(u ) = ∫ g(u)cosθ (u)du + xD ,
0
u
(3.4.22)
y (u ) = ∫ g (u )sin θ (u )du + yD
0
Namely, the cavity thickness is equal to the ordinates of point D ( H = yD ); the cavity length
is obtained as L = x(1) − xD . (3.4.23)
In conclusion it should be noted that calculations using conformal map are much faster than
the mixed value problem.
Results of computations are plotted in Figs.3.4.3 and 3.4.4. Geometric parameters x, y, L and
H are made dimensionless as ratios to length l. All computed results except the points in
Fig.3.4.3a were obtained using conformal mapping; the points are calculated by equations
(3.4.10). Fig.3.4.3 shows cavity boundaries past the wedge (a) and the perpendicular plate
(b). Length L and thickness H of cavity are plotted in Fig.3.4.4a past a perpendicular plate.
Fig.3.4.4b shows the speed distribution at the outer face of the plate without cavity ( σ = −1 )
and with Kirchhoff’s cavity past interior side.
Fig.3.4.3. The cavities past a

wedge (a) and plate (b); the points Fig.3.4.4. Cavity length L and thickness H
are calculated using the w-plane. behind the plate (a), speed distribution (b).
Sec. 3.5. Flat Plate near a Free Surface

A nonlinear problem of a flat plate near the free surface was first simultaneously in
[Cherepanov, 1963; Epshteyn, 1963; and Kuznetsov, 1969] though the latter was published
with a few years delay due to technical problems. While the first two authors obtained
general solutions only, AV Kuznetsov [1969] analyzed some asymptotic expressions and
made calculations for deeper submergence. Some numerical data have been also calculated
in [Capodano, 1969]. Much more results have been obtained in [Kuznetsov & Terentiev,
[1980] using theta-function, which is much more convenient for calculations.
3.5.1. Complex Function of Speed
Let the plate be under the free boundary (Fig.3.5.1). Let also place the origin at the leading
edge, and the x-axis to be parallel to stream speed at infinity. The angle of attack is denoted
as α . The following conditions should be satisfied.
83
• dynamic and kinematic conditions at the free boundary,

| wz ( z ) |= 1, ψ = ψ 1 = const., z ∈ (CC ) (3.5.1)
kinematic condition on the flat plate, ψ = ψ 0 = const , z ∈ ( A1OBA2 ) . (3.5.2)
The distance between two streamlines at infinity h = ψ 1 − ψ 2 determines the depth of plate.
Instead of condition (3.5.2), another condition for velocity argument could be used as
⎧α , z ∈ ( A1O ) ∪ ( BA2 ),
arg( wz ) = ⎨ (3.5.3)
⎩α − π , z ∈ (OB )
Mapping the flow domain in the
complex z-plane onto an interior
of a rectangle with vertexes at
points 0, π , π + πτ i / 4, πτ i / 4
of the auxiliary ζ -plain
(Fig.3.5.1b), one can obtain
two-value problems of complex
velocity wz (ζ ) and complex Fig.3.5.1. Plate under free boundary (a); parameter
potential w(ζ ) with given rectangle on ζ -plane (b).
conditions (3.5.1) – (3.5.3). The
vertical sides of the rectangle correspond to the opposite faces of the slit, which joints the
trailing edge with a point on the free boundary.
Consider first the complex velocity, which satisfies the single-value condition
wz (ζ + π ) = wz (ζ ) . (3.5.4)
Due to condition (3.5.3), function wz (ζ ) should have a single pole at point B (ζ = b) and a
single zero at point O(ζ = a) , and due to condition (3.5.1) it can be extended analytically
trough the horizontal upper side. Due to the principle of symmetry, function wz (ζ ) should
have a single zero at symmetric point B ∗ (ζ = b + πτ i / 2) and a single pole at
point O∗ (ζ = a + πτ i / 2) . As all singularities and zeros on fundamental rectangle with sides
ϑ (ζ − a)ϑ4 (ζ − b)
π and πτ are known, then wz = eα i 1 , (3.5.5)
ϑ4 (ζ − a)ϑ1 (ζ − b)
where constant factor eα i is obtained to satisfy condition (3.5.3) on the plate. The velocity
angle at the free boundary could be found from (3.5.5) by substituting ζ = ξ + πτ i / 4 ,
⎡ ⎛ πτ i ⎞ ⎛ πτ i ⎞ ⎤
β (ξ ) = − arg wz == 2 ⎢arg ϑ4 ⎜ ξ − a + ⎟ − arg ϑ4 ⎜ ξ − b + ⎟ + b − a −α . (3.5.6)
⎣ ⎝ 4 ⎠ ⎝ 4 ⎠ ⎥⎦
The velocity is directed at infinity along the x-axis, so that the condition should be satisfied
as β (c ) = 0 . (3.5.7)
3.5.2. Complex Potential

Instead of an ambiguous complex potential, let's consider its single-value derivative wζ (ζ ) ,
which has an imaginary part vanished at the horizontal sides of the rectangle and a second-
order pole at point C (ζ = c + πτ i / 4) . Thus it can be extended analytically through the
rectangle's horizontal sides. The fundamental rectangle obtained in such a way has its sides
84
π and πτ i / 2 , so that function wζ (ζ ) in the rectangle has a second-order pole at point

C ∗ (ζ = c + 3πτ i / 4) with the same coefficient. Hence, the derivative can be expressed as
⎡ ⎛ πτ i ⎞ ⎛ πτ i ⎞ ⎤
wζ = A ⎢ B − ln ′′ϑ4 ⎜ ζ − c − ⎟ − ln ′′ϑ4 ⎜ ζ − c + ⎟ . (3.5.8)
⎣ ⎝ 4 ⎠ ⎝ 4 ⎠ ⎥⎦
Factors A and B are real constants in this case. The derivative vanishes at the trailing edge
⎛ πτ i ⎞ ⎛ πτ i ⎞
A1 (ζ = 0) . Whence, B = ln ′′ϑ4 ⎜ c + ⎟ + ln ′′ϑ4 ⎜ c − ⎟. (3.5.9)
⎝ 4 ⎠ ⎝ 4 ⎠
Besides, function wζ (ζ ) vanishes at stagnation point O ( ζ = a ) . This condition yields
a = 2c (3.5.10)
Integrating derivative (3.5.8), one can obtain complex potential as
⎡ ⎛ πτ i ⎞ ⎛ πτ i ⎞ ⎤
w = A ⎢ Bζ − ln ′ϑ4 ⎜ ζ − c − ⎟ − ln ′ϑ4 ⎜ ζ − c + ⎟ + const (3.5.11)
⎣ ⎝ 4 ⎠ ⎝ 4 ⎠ ⎥⎦
The above-mentioned depth of plate is expressed from (3.5.11) in the form
h = Im( w(ξ + πτ i / 4) − w(ξ )) = A (1 + π Bτ / 4 ) (3.5.12)
3.5.3. Transformation Function
Transformation from the z-plane to the ζ -plane can be calculated integrating function
dz wζ
= (3.5.13)
d ζ wz
Function zζ (ζ ) = dz / d ζ is an elliptic function with periods π and πτ i. In the fundamental
rectangle it has second-order poles at points C (c + πτ i / 4) and C * (c − πτ i / 4) , and a single
pole at point B ∗ (b + πτ i / 2) . The leading part can be found expanding functions (3.5.5) and
(3.5.8) at those points. Function (3.5.5) has the following expansions:
1 ⎛ πτ i ⎞
• at point C (c + πτ i / 4) = 1 + ik ⎜ ζ − c − ⎟ +… ; (3.5.14)
wz ⎝ 4 ⎠
1 ⎛ πτ i ⎞
• at point C * (c − πτ i / 4) = e −2α i − ike −2α i ⎜ ζ − c + ⎟ +…; (3.5.15)
wz ⎝ 4 ⎠
1 ne −α i
• at point B * (c + πτ i / 2) , = +… ; (3.5.16)
wz ζ − b − πτ i / 2
⎡ ⎛ πτ i ⎞ ⎛ πτ i ⎞ ⎤ ϑ4 (0)ϑ1 (b − a )
where k = 2 Im ⎢ln ′ϑ4 ⎜ c − a + ⎟ − ln ′ϑ4 ⎜ c − b + ⎟⎥ , n = ′ . (3.5.17)
⎣ ⎝ 4 ⎠ ⎝ 4 ⎠⎦ ϑ1 (0)ϑ4 (b − a )
Function (3.5.8) has the second-order poles at points C and C* with the same coefficients A,
and is bounded at point B*. Let the value of the function at that point be presented as
wζ (b + πτ i / 2) = Am . (3.5.18)
The sum of all coefficients at single poles in the rectangle should vanish because of two-
periodicity of derivative (3.5.13). Thus mn = 2k sin α . (3.5.19)
Now, the leading part of singularities in the fundamental rectangle is known and
function zζ (ζ) could be presented as a linear combination of logarithmic derivatives of
theta functions [Whittaker & Watson, 1927]. Let us denote the expression as
85
⎛ πτ i ⎞ −α i ⎛ πτ i ⎞
f (ζ ) = −eα i ln ′ϑ4 ⎜ ζ − c + ⎟ − e ln ′ϑ4 ⎜ ζ − c − ⎟+
⎝ 4 ⎠ ⎝ 4 ⎠
(3.5.20)
⎡ αi ⎛ πτ i ⎞ −α i ⎛ πτ i ⎞ ⎤
+ik ⎢e ln ϑ4 ⎜ ζ − c + ⎟ − e ln ϑ4 ⎜ ζ − c − ⎟ + 2k sin α ln ϑ4 (ζ − b)
⎣ ⎝ 4 ⎠ ⎝ 4 ⎠ ⎥⎦
the derivative of which has the same singularities as function zζ (ζ ) but the coefficients
⎛ df (ζ ) ⎞
Hence, zζ = Ae −α i ⎜
differ by factor Ae −α i . − K⎟. (3.5.20a)
⎝ dζ ⎠
Constant K should be zero because of single-value condition z (ζ + π ) − z (ζ ) = 0 . Thus
z (ζ ) = Ae −α i [ f (ζ ) − f (b) ] (3.5.21)
The derivative of transformation (3.5.21) vanish at point B (b) corresponding to the leading
edge. Thus the additional condition should be satisfied as
df (b)
=0 (3.5.22)
db
Now, we have three conditions – (3.5.7), (3.5.12) and (3.5.22) – for calculating three
unknown parameters b, c and τ; parameter a can be excluded by (3.5.10). Constant factor A
is determined from the plate length condition
l = A [ f (0) − f (b)] . (3.5.23)
It should be noted that the length was assumed to be unity (l = 1), but instead of length,
factor A or another fixed number can be chosen as unity. In this case the plate length should
be calculated by (3.5.23) and then all hydrodynamic parameters should be converted to
dimensionless quantities relative to this length.
3.5.4. Lift and Torque
Using Blasius-Chaplygin integrals along a loop enveloping the plate give the lift and torque.
On the ζ-plane in the fundamental rectangle, integration along a horizontal straight line gives
π
i
R = − ∫ wz (ξ + iη0 ) wζ (ξ + iη0 )d ξ
20
π
(3.5.23)
1
M = Re ∫ wz (ξ + iη0 ) wζ (ξ + iη0 ) z (ξ + iη0 )d ξ
2 0
where η0 is a constant value from interval (0<η0 <πτ /4). The first integral, as well as
(3.5.21), can be presented analytically. Instead of that integral, the lift can be calculated
using the concentrated force at the leading edge as
i
P = ∫ wz (ζ ) wζ (ζ )d ζ . (3.5.24)
4 ζ =b
Analytic expression of value | P | can be obtained using the technique of residue and then lift
L =| P | / sin α . The lift coefficient is CL = 2π k A . (3.5.25)
Three equations (3.5.7), (3.5.12) and (3.5.22) allow to calculate the hydrodynamic
parameters for a wide range of submergence depth of plate (0.05<h<∝). The starting values
of unknown parameters can be calculated from asymptotic expressions, which can be
86
obtained using parameter τ → 0 , or the angle of attack α → 0 [Kuznetsov & Terentiev,
1980]. The first case corresponds to the plate close to free boundary; the other corresponds
to linear problem.
The free boundaries for angle of attack α=10° and different submergence depths are plotted
in Fig.3.5.2. The free boundary ordinate at infinity has an asymptotic expression as
y = −(CL ln | x |) / 2π , i.e. the free boundary at infinity is reduced to a logarithm singularity.
This phenomenon was first revealed by Kuznetsov in 1963. Submergence smaller than h
<0.1 would perhaps be impossible to realize practically because the flow could vary and a
jet can be formed near the leading edge and the plate will glide along the free surface. It
should be noted that a model of plate planning could not be obtained asymptotically by
approaching h → 0 . Fig.3.5.2b shows the free boundary calculated for h = 0.01, which is
close to approaching h → 0 . The flow splits in two parts asymptotically: a jet flow around
the semi-infinite plate's edge and planning of the semi-infinite plate. The latter differs from
the known problem of planning [Gurevich, 1961]; the jet of a fluid flows in front of the
semi-infinite plate and returns along its upper side to the main flow. The jet along the plate
does not affect pressure but generates a force directed along the plate and equal to 2h , so
that the resultant force is vertical. For the asymptotic case, all hydrodynamic parameters are
related to length l0 calculated as a distance from the trailing edge to the point of its
intersection with the tangent line to the free boundary, which is normal to the plate
(Fig.3.5.2b). The lift coefficient for the planning plate is expressed as
2π cos α
CL∞ = . (3.5.26)
tan (α / 2) + π + tan(α / 2) ln[tan −2 (α / 2) − 1]
−1
At α=10° the coefficient is CL∞ = 0.4126 . The corresponding length in Fig.3.5.2b is

l0 = 0.533 and the lift coefficient is CL cos 2 α / l0 = 0.4191 which is close to 0.4126.
Approaching h→0
allows one to obtain the
exact value of coefficient
CL∞ .
Fig.3.5.2.
(a) Free boundaries for
different submergence
depths; (b) almost
asymptotic case.
Using the exact

analytical solution one
can calculate all
necessary values
including various
asymptotic cases as distinct from numerical investigation. Lift coefficient
87
CL0 = CL / 2π sin α and the center of pressure δ = CM / CL cos α (a distance from the leading
edge) for different angles of attack and depths are presented in Table 3.5.1.
Table 3.5.1. Lift coefficient and pressure center.

C L0 at α = δ at α =
h
0 2 5 10 15 0 2 5 10 15
0.1 0.6164 0.5918 0.5558 0.4978 0.4412 0.2253 0.2308 0.2404 0.2612 0.2901
0.2 0.6963 0.6714 0.6351 0.5777 0.5234 0.2213 0.2246 0.2304 0.2428 0.2593
0.3 0.7621 0.7370 0.7005 0.6430 0.5891 0.2239 0.2258 0/2295 0.2375 0.2483
0.4 0.8144 0.7898 0.7540 0.6972 0.6440 0.2283 0.2294 0.2317 0.2370 0.2445
0.5 0.8545 0.8309 0.7965 0.7415 0.6897 0.2324 0.2330 0.2345 0.2382 0.2435
0.6 0.8845 0.8624 0.8298 0.7774 0.7277 0.2358 0.2362 0.2371 0.2397 0.2436
0.7 0.9070 0.8865 0.8560 0.8066 0.7592 0.2385 0.2387 0.2393 0.2411 0.2441
0.8 0.9241 0.9050 0.8767 0.8303 0.7854 0.2405 0.2407 0.2411 0.2424 0.2447
0.9 0.9371 0.9195 0.8931 0.8497 0.8074 0.2421 0.2422 0.2425 0.2435 0.2453
1 0.9473 0.9309 0.9064 0.8658 0.8260 0.2434 0.2434 0.2436 0.2444 0.2458
2 0.9850 0.9762 0.9628 0.9400 0.9168 0.2481 0.2481 0.2481 0.2482 0.2485
∞ 1 1 1 1 1 0.25 0.25 0.25 0.25 0.25
Sec. 3.6. Flat Plate near a Flat Bottom

Many authors have considered the ground effect on lift of a monoplane aerofoil. First
theoretical results on ground effect as a special case of a biplane were obtained in
[Chaplygin & Golubev, 1935; Sedov, 1950]. In the absence of modern computers, authors of
the early works had focused
mainly at theoretical studies.
First numerical results were
obtained in [Prosnak &
Kucharzyk, 1959] using
expansion of analytical solution
into power series of small
parameters. Later, Galanin
[1974] defined more exactly the
asymptotic expansions of
solution. Another solution and Fig.3.6.1. (a) Plate near the bottom;
numerical calculation were realized in [Terentiev & (b) the parametric
Kartuzova, 1996]. rectangle.
3.6.1. Transformation from the z-Plane to the Parametric ζ-Plane
The flat plate in the z-plane and the parametric rectangle in the ζ-plane are shown on the
Fig.3.6.1a. The slit in the z-plane is chosen as a circular arc of radius R; edges A and B are
symmetric relative to the circle. If the length is unity and the distance from the trailing edge
to the bottom is denoted as h, then the slit radius is expressed as
88
h ⎛ h ⎞
R= ⎜ + 1⎟ . (3.6.1)
sin α ⎝ sin α ⎠
Points O and D in the z-plane are symmetrical and they are mapped into
symmetrical points O( ζ = a ) and D( ζ = π − a ) within the rectangle (Fig.3.6.1b). The plate
edges are also mapped into points B and A placed on equal distances b from the rectangle
vertical sides. Stagnation point C on the plate is mapped into point C on the rectangle upper
side (distance c shown n Fig.3.6.1b). The logarithmic derivative of transformation z(ζ) has
a real value in the horizontal sides and single poles at points O and D, so that it can be
extended over the rectangle's sides and found as elliptic function of periods π and πτ i . It
can be expressed by a logarithmic derivative of theta-functions as
d
ln z (ζ ) = A ln ′ϑ1 (ζ − a ) − A ln ′ϑ1 (ζ + a ) (3.6.2)
dζ
Hence, map the flow domain in the z-plane onto the rectangle of the ζ -plane by the
ϑ (ζ − a )
transformation: z (ζ ) = A 1 . (3.6.3)
ϑ1 (ζ + a )
The logarithmic derivative could also be found with regard to simple zeros at points
A and B, and single poles at points O and D as
zζ (ζ ) ϑ (ζ − b)ϑ4 (ζ + b)
=B 4 . (3.6.4)
z (ζ ) ϑ1 (ζ − a )ϑ1 (ζ + a)
From (3.6.4) and (3.6.3) the derivative of the transformation is found as
ϑ (ζ − b)ϑ4 (ζ + b)
zζ = AB 4 (3.6.5)
ϑ1 (ζ + a ) 2
3.6.2. Derivative of Complex Potential and Complex Speed
The derivative of complex potential has the second-order pole at point D and simple zeros at
stagnation point C and at point B corresponding to the trailing edge, and real value at the
horizontal sides of the rectangle. Therefore, it can be extended over the whole ζ -plane and
ϑ (ζ − b)ϑ4 (ζ + c)
expressed as wζ = C 4 . (3.6.6)
ϑ1 (ζ + a ) 2
The complex velocity is obtained from (3.6.5) and (3.6.6) as

wζ Cϑ4 (ζ + c)
wz = = (3.6.7)
zζ ABϑ4 (ζ + b)
All the functions from (3.6.3) to (3.6.7) have unknown factors A, B, C, parameters a, b, c,
and aspect ratio of the fundamental rectangle τ . Symmetry condition R 2 = z (b) z (π − b)
yields A= R (3.6.8)
If a distance between a point on the plate and the origin is denoted s, then the point is
determined by the transformation as z (ξ + πτ i / 2) = s exp(π − α )i . Satisfying this condition
yields from (3.6.3) parameter a and function s (ξ ) as follows:
π −α
a= ; (3.6.9)
2
89
ϑ3 (ξ + α / 2 )
s (ξ ) = R . (3.6.10)
ϑ3 (ξ − α / 2 )
Function (3.6.10) has an extreme at point A so that its derivative vanish at that point.
The extreme condition together with an expression for the plate unit length yields two
equations relative parameters b and τ as follows:
s (−b) − s (b) = 1 , (3.6.11)
ln ′ϑ3 ( b + α / 2 ) − ln ′ϑ3 ( b − α / 2 ) = 0 . (3.6.12)
Factor B in (3.6.4) and (3.6.5) could be determined by comparing the leading parts of
function (3.6.5) and derivative obtained directly from function (3.6.3), i.e.
ϑ (0)ϑ3 (0)ϑ4 (0)ϑ1 (α )
B= 2 . (3.6.13)
ϑ3 (b + α / 2)ϑ3 (b − α / 2)
Satisfying the condition for the known velocity argument in the plate yields
c = b −α . (3.6.14)
Finally, factor C is calculated from the condition for velocity at infinity,
ϑ (b + α / 2)
C = BRv∞ 3 . (3.6.15)
ϑ3 (c + α / 2)
3.6.3. Lift Coefficient and Computation
The lift as in Sec. 3.5, could be calculated by concentrated force at the plate leading edge
π C 2ϑ1 (2b)ϑ12 (α )
CL = . (3.6.16)
BRϑ2 (0)ϑ3 (0)ϑ4 (0)ϑ32 (b + α / 2)sin α
The lift and torque can also be calculated integrating pressure along the plate. The lift
coefficient and pressure center are given in Table 3.6.1 as a function of the distances and
angle of attack. Results of calculations show that the bottom effect is positive for small angle
of attack, but negative at the angle of attack α ≥ 15°.
Table 3.6.1. The lift coefficient and center of pressure.

CL /2π sinα δ
h α° α°
2 5 10 15 20 2 5 10 15 20
0.01 4.2101 2.1672 1.3222 1.0144 0.8546 0.3974 0.3890 0.3566 0.2370 0.3023
0.05 2.6648 1.8021 1.2498 1.0032 0.8634 0.3383 0.3430 0.3304 0.3126 0.2950
0.1 1.9956 1.5492 1.1791 0.9854 0.8663 0.3152 0.3184 0.3115 02997 0.2869
0.2 1.5019 1.3018 1.0907 0.9586 0.8683 0.2938 0.2945 0.2905 0.2835 0.2755
0.3 1.3046 1.1843 1.0417 0.9430 0.8710 0.2819 0.2817 0.2787 0.2739 0.2683
0.4 1.2022 1.1188 1.0130 0.9349 0.8751 0.2740 0.2735 0.2712 0.2677 0.2633
0.5 1.1419 1.0791 0.9956 0.9311 0.8801 0.2685 0.2680 0.2662 0.2635 0.2604
0.6 1.1036 1.0536 0.9849 0.9300 0.8854 0.2645 0.2641 0.2626 0.2605 0.2581
0.7 1.0780 1.0365 0.9782 0.9304 0.8908 0.2616 02612 0.2601 0.2584 0.2564
0.8 1.0601 1.0248 0.9741 0.9318 0.8960 0.2594 0.2591 0.2582 0.2568 0.2552
0.9 1.0473 1.0165 0.9717 0.9336 0.9010 0.2578 0.2575 0.2568 0.2557 0.2543
1.0 1.0377 1.00105 0.9703 0.93356 0.9057 0.2565 0.2563 0.2557 0.2547 0.2536
2.0 1.0063 0.9933 0.9731 0.9546 0.9376 0.2518 0.2518 0.2516 0.2514 0.2510
∝ 1 1 1 1 1 0.25 0.25 0.25 0.25 0.25
90
Sec. 3.7. Cavitating Flow around a Plate near a Free Surface

An influence of free surface on supercavitating flow around a foil is of practical interest, in
particular for high-speed craft. First theoretical results in a linear problem statement were
obtained in [Johnson, 1958; Auslaender, 1962; Yim, 1964]. Then this problem has been
considered using other methods but also in a linear statement in [Efremov, 1969; Vishnevsky
& Galanin, 1975]. The authors of the latter work obtained asymptotic expressions for the
plate in a channel and in a jet, as well as close to a free surface. Larock & Street [1967] in
USA and independently Terentiev & Lasarev [1969] in Russia applied the Tulin’s cavitating
model with double spirals to a nonlinear problem of cavitating flow around a flat plate near
free surface, but the problem had indefinite solution and an additional condition should be
used. Later, YV Kuznetsov & AG Terentiev in [Terentiev, 1981] obtained an exact solution
of the nonlinear problem using the cavitating model with single spirals. An effective
numerical method for nonlinear problems of flow with free boundaries was developed in
[Maklakov, 1997], where numerous problems including cavitating flow near free boundaries
were analyzed numerically. An analytical solution different from the above-mentioned
solutions is discussed below.
Consider a supercavitating plate under a free
boundary, Fig.3.7.1a. For solving the problem,
the double-connected flow domain should be
turned into a simply connected one by cuts as
shown in Fig.3.7.1. For simplicity the cut
should be made in such a way that the domain
of complex potential would become
symmetric, Fig.3.7.1b. Obviously, the cut's Fig.3.7.1. (a) Flat plate under a free
sides (1-4) and (2-3) in the w-plane should, in
surface; (b) the w-plane; (c) the ζ-plane.
general, not coincide – the distance between
them is equal to circulation along the flat and cavity boundary. Now, map the domain of
complex function onto the inner of rectangle, Fig.3.7.1c. The corresponding points on all
planes are denoted with the same letters: a, b and c are coordinates of points A, B, and C on
the ζ−plane; coordinates of points A2 and D are equal to π − a and π (1 + iτ ) / 2 ,
respectively. The cavity boundary speed is assumed to be unity, and thus the speed of free
boundary ( D1 D1 ) is given: v∞ = −0.5ln(1 + σ ) . Besides, angle of attack α and the plate-to-
free boundary distance h are known. As above, submergence depth is defined as a distance
at infinity between the free boundary and a stream line passing through stagnation point on
the plate.
3.7.2. Derivative of Complex Potential
The derivative of transformation wζ (ζ ) is real at the rectangle's horizontal sides and has a
second-order pole at point D(π / 2 + πτ i / 2) . Analytic extension over the whole ζ -plane
yields an elliptic function with periods π and πτ i , so that the derivative can be written in
the form wζ = A ln ′′ϑ3 (ζ ) + B , (3.7.1)
Hence, the transformation from the w-plane to the ζ-plane is presented as
91
w = A ln ′ϑ3 (ζ ) + Bζ + D (3.7.2)
Coefficient B can be obtained from the condition for circulation Γ as B = −Γ / π . The origin
can be chosen to make the term D to be equal to zero. If the plate submergence depth (h)
would, as before, be considered as the distance between two streamlines at infinity, then the
length of section (2-3) in the w-plane is equal to v∞ h and coefficient A = −v∞ h − Γτ / 2 .
Thus the transformation is expressed in the form
w = −(v∞ h + Γτ / 2) ln'ϑ3 (ζ ) − Γζ / π . (3.7.3)
Here, circulation Γ and parameter τ are unknown. If L is the length of section A1A2 and a
and π − a are coordinates of points A1 and A2 , respectively, then two conditions should
2Γ
be satisfied ln ′′ϑ3 (a) = − , (3.7.4)
π (Γτ + 2v∞ h)
(π − 2a )Γ − π L
ln ′ϑ3 (a ) = . (3.7.5)
π (2v∞ h + Γ)
Therefore, transformation (3.7.3) contains two unknown parameters a and τ or unknown
circulation Γ and length L. They could be determined after solving the whole problem.
3.7.3. Logarithmic Function of Velocity
The problem of function ω (ζ ) = ln wz / v0 is described by following boundary conditions:
• at the upper side of the rectangle corresponding to the free surface
Re ω (ϕ + πτ i / 2) = ln(v∞ / v0 ) = −0.5ln(1 + σ ) ; (3.7.6)
• at intervals (c, π ) and (0, b) of the bottom side of the appropriate cavity boundary
Re ω (ϕ ) = 0 ; (3.7.7)
• at interval ( b, π − a ) corresponding to section BA2 of the plate
Im ω (ϕ ) = α ; (3.7.8)
• at interval ( π − a, c ) corresponding to section A2C
Im ω = α − π . (3.7.9)
Besides, function ω (ζ ) should have a simple pole at point A1 (ζ = a) .
The desired function can be expressed using functions (A1.1) – (A1.5) of Appendix 1
⎧ ln(1 + σ ) π ⎫
1
⎪
⎪ 2 ∫ g 4 (ξ ) A4 (ξ − ζ )d ξ − ⎪
⎪
ω= ⎨ c
0
⎬ + iα , (3.7.10)
2π ig (ζ ) ⎪ ⎪
⎪π ∫ g1 (ξ ) A(ξ − ζ ) d ξ + C A( a − ζ ) ⎪
⎩ π −a ⎭
ϑ1 (ζ − c) ϑ (c − ξ ) ϑ4 (ξ − c)
g (ζ ) = , g1 (ξ ) = 1 , g 4 (ξ ) = ,
ϑ1 (ζ − b) ϑ1 (ξ − b) ϑ4 (ξ − b)
ϑ2 (0)ϑ3 (0)ϑ4 (0)ϑ1 (ξ − ζ + (c − b) / 2)
where A(ξ − ζ ) = , (3.7.11)
ϑ1 ((c − b) / 2)ϑ1 (ξ − ζ )
ϑ (0)ϑ3 (0)ϑ4 (0)ϑ4 (ξ − ζ + (c − b) / 2)
A4 (ξ − ζ ) = 2 .
ϑ1 ((c − b) / 2)ϑ4 (ξ − ζ )
92
3.7.4. Transformation z(ζ) and Obtaining Unknown Parameters

The transformation from the z-plane to the ζ -plane can be calculated using differential
dz
equation = zζ (ζ ) = e −ω (ζ ) wζ (ζ ) . (3.7.12)
dζ
Functions (3.7.3), (3.7.10) and (3.7.12) have five unknown parameters: a, b, c, τ and Ñ0 ,
which could be obtained satisfying following conditions:
• speed at infinity ( z → ∞ ) is directed toward the x-axis, i.e.
Im ω (π / 2 + iπτ / 2) = 0 , (3.7.13)
• speed should be limited at point C(c)
π
ln(1 + σ )
c
2 ∫0 g 4 (ξ ) A4 (ξ − c ) d ξ − π ∫ g1 (ξ ) A1 (ξ − c)dξ + C A(a − c) = 0 , (3.7.14)

π −a
• curve enveloping the plate and cavity has to be closed,
π
∫ zζ (ξ + iη )dξ = 0,
0
η = const , (0 < η < πτ / 2) . (3.7.15)
The real and imaginary parts of the last condition yield two real equations. The fifth equation
is obtained from the given length of the plate
c
l = ∫ | zζ (ξ ) | d ξ . (3.7.16)
b
Conditions (3.7.13) – (3.7.16) constitute a set of nonlinear equations that could be rather
difficult to solve. For successful solving, the initial value of unknown parameters should be
taken as close as possible to the exact one. This task could be achieved considering
asymptotic approaches for small cavitation numbers and the angle of attack.
3.7.5. Asymptotic Expressions
At σ → 0 and α → 0 , point C approaches point A2 , ( c → π − a ) so that function (3.7.10) is
also expressed like Eq. (3.7.10) but without the second integral. Since function ω (ζ ) takes
small values, then derivative zζ coincides with wζ . With regard to circulation infinitesimal,
the domain on the z-plane, hereafter denoted ( z0 ), as well as the domain on the w-plane are
the lower half-plane with the horizontal slit. The cut's sides (1-4) and (2-3) coincide and
transformation from the z0 -plane to the ζ -plane can be written in the form
z0 (ζ ) = −h ln ′ϑ3 (ζ ) . (3.7.17)
If the ratio of cavity length to depth of submergence, L/h, is given, then conditions (3.7.4)
and (3.7.5) are to be satisfied by vanished circulation. Adding to them a condition for the
plate length, one can obtain a set of equations for a, b and τ
ln ′′ z0 (a) = 0, ln ′ z0 (a ) = − L / 2h, ln ′ z0 (b) = (2l − L) / 2h . (3.7.18)
Though equations (3.7.18) are nonlinear, they could be easily solved numerically
using equations (A5.14) – (A5.17). These solutions may be used as initial values for the
numerical procedure by solving the above-mentioned equations set. Asymptotic expressions
(3.7.17) and (3.7.18) correspond to the linear theory. The ordinate of free boundaries
including the cavity boundary could be calculated integrating function (3.7.10) as
93
y (ξ ) = Im ∫ ω (ξ ) zζ (ξ )d ξ . (3.7.19)
Namely, integral (3.7.19) along the bottom side of the rectangle must be zero due to the
cavity closure condition. Instead of the bottom side, integrating can be realized along any
straight horizontal line between two sides of the rectangle. Having obtained coefficient Ñ0
from condition (3.7.15) and substituting it in function (3.7.10) gives the asymptotic solution:
σ g (ζ ) ⎛ π Im F1 ⎞
ω (ζ ) = ⎜ − ∫ ( g 4 (ξ ) A4 (ξ − ζ ) d ξ + A(a − ζ ) ⎟ , (3.7.20)
2π i ⎝ 0 Im F2 ⎠
ϑ (ζ − b) ϑ (0)ϑ3 (0)ϑ4 (0) ϑ1 (ζ − (a + b) / 2)
where g (ζ ) = 1 , A(ζ ) = 2 ,
ϑ1 (ζ + a) ϑ1 ((a + b) / 2) ϑ1 (ζ )
ϑ1 (ζ − b) ϑ (0)ϑ3 (0)ϑ4 (0) ϑ1 (ζ − (a + b) / 2)
g (ζ ) = , A(ζ ) = 2 ,
ϑ1 (ζ + a ) ϑ1 ((a + b) / 2) ϑ1 (ζ )
ϑ4 (ξ + a ) ϑ (0)ϑ3 (0)ϑ4 (0) ϑ4 (ζ − ( a + b) / 2)
g 4 (ξ ) = , A4 (ζ ) = 2 , (3.7.21)
ϑ4 (ξ − b) ϑ1 ((a + b) / 2) ϑ4 (ζ )
π π
F1 = ∫ g ( s + iη0 ) f ( s + iη0 )z0 ( s + iη0 )ds, F2 = ∫ g ( s + iη0 ) A( a − s − iη0 )z0 ( s + iη0 ) ds,
0 0
π
f (ζ ) = ∫ g 4 (ξ ) A4 (ξ − ζ )d ξ , η0 ∈ (0, πτ / 2) .
0
The lift and toque relative to the origin are calculated as before by integrating the
pressure along the flat plate. Dependences of normal force on submergence depth and on
cavitation number were calculated by YV Kuznetsov for a nonlinear problem of the plate
with angle of attack α=10°, see Fig.3.7.2. It is seen that the resulting force, and
consequently, the lift increase when approaching the free boundary. However, at small
cavitation numbers the increase is slowing down. It
is easy to find that the resulting force coefficient,
when coming to the free surface, approaches
cavitation number Cn→σ while the streamlines near
the leading edge degenerate to concentric circular
arcs similar to stream lines of concentric vortex;
width of the ring formed by the free surface and
cavity boundary approaches
h( 1 + σ − 1)
hc = (3.7.22)
1 + σ ln 1 + σ
Indeed, when submergence depth approaches zero,
an infinitely thin jet is formed along the plate
wetted side. At one side of it the velocity is equal to
that at infinity, and at the other side – to the velocity
at the cavity boundary, what is equivalent to a
circular flow from a point vortex. Dependence
Fig.3.7.2. Dependence of total between the ring radiuses and velocities is given in
force coefficient on σ and l / h. Eq. (3.7.22).
94
3.7.6. Results of Calculations

Results of asymptotic problem solution calculated using formulas (3.7.17) – (3.7.21) are
plotted in Figs.3.7.3 and 3.7.4. The cavity shapes and free surfaces are shown in Fig.3.7.3
where the points in Fig.3.7.3b correspond to nonlinear problem calculated by Maklakov &
Naborova [Maklakov, 1997] using another numerical spline-method. A good agreement
between linear and nonlinear theories is seen in the bottom plot. Fig.3.7.4 shows lift
coefficient versus cavity length given as ratios by lift coefficient CL0 and cavity length L
corresponding to the plate in unbounded fluid flow, as well as ratio α / σ . It is seen that the
lift changes slightly but cavity length shortens by approaching free surface.
The lift coefficient as a function of submergence depth is calculated by
σ Im F1
CL = Re(− F1 + F2 ) . (3.7.23)
π Im F2
Fig.3.7.4. Lift coefficient,

Fig.3.7.4.
cavity Liftratio
length and coefficient,
α/σ forcavity
length and ratio α / σ for
a plate near free surfacea plate
and near
free surfacein
andunbounded fluid.fluid.
in unbounded
Fig.3.7.3. Cavity shapes under free surface; points
Fig.3.7.3. Cavity
in the bottom shapes under
correspond free surface;
to nonlinear points
problem.
in the bottom correspond to nonlinear problem.
Sec. 3.8. Applying Tulin’s Model to Cavitating Flows

The Tulin’s model with two spirals, Sec. 2.6.1, allows one to reduce the multi-connected
flow domain to a single-connected one, thus to simplify the cavitating flow problem
significantly. This model was applied in [Larock & Street, 1967] for solving a nonlinear
problem of a fully cavitating hydrofoil beneath a free surface. The Tulin’s model was also
used in [Terentiev & Lasarev, 1969] to investigate numerous problems of cavitating flow in
bounded domain.
3.8.1. Cavitating Plate in a Flow out of a Channel
As shown in Fig.3.8.1a, the stream flows out from a channel of width L. Flow speed at the
stream free boundary is constant and equal to v1; the speed in the channel at infinity is v∝.
The speed value at the cavity boundary is constant and equal to v0. In compliance with the
95
double vortex model, a wake past the cavity is
assumed to exist with constant speed v1 at its
boundary. The attached points M and C are centers of
double spirals with unknown locations.
Fig.3.8.1.
(a) - flow in the z-plane;
(b) - parametric rectangle in the ζ -plane
It is given: coordinates of points E and G and

width L relative to plate length l and velocities v0 and v1. As usually, the problem can be
solved parametrically by conformal mapping of flow domain onto the inner rectangle with
vertexes 0, π / 2, (π + iπτ ) / 2, iπτ in the ζ -plane. A conformity of points are shown in
Fig.3.8.1a and b. The derivative of complex potential ωζ (ζ) is real at the horizontal sides
and purely imaginary at the vertical sides of the rectangle, so that it can be extended
analytically over the whole ζ-plane and become an elliptic function with periods π and iπτ .
On the fundamental rectangle, it has single poles only at points F (b + iπτ / 2), K (im) ,
D (π / 2 + in) , at symmetrical points −b + iπτ / 2, − im, π / 2 − in and at single zeros at
points 0, π / 2, (π + iπτ ) / 2, iπτ , a, -a. Hence, the derivative can be expressed in the form:
ϑ1 (2ζ )ϑ1 (ζ − a)ϑ1 (ζ + a)
wζ = . (3.8.1)
ϑ1 (ζ − im)ϑ1 (ζ + im)ϑ2 (ζ − in)ϑ2 (ζ + in)ϑ4 (ζ − b)ϑ4 (ζ + b)
Integrating this function round the points F, K and D one can obtain the amounts of flowing
fluid in the channel, in the jets, K, and, D, respectively,
ϑ4 (b − a )ϑ4 (b + a )
QF = N , (3.8.2)
| ϑ4 (b + im)ϑ3 (b + in) |2
| ϑ1 (a + im) |2
QK = − N , (3.8.3)
ϑ2 ((m − n)i )ϑ2 ((m + n)i ) | ϑ4 (b + im) |2
| ϑ2 (a + in) |2
QD = − N , (3.8.4)
ϑ2 (( m − n)i )ϑ2 ((m + n)i ) | ϑ3 (b + in) |2
where N = π / ϑ2 (0)ϑ3 (0)ϑ4 (0) .
Hence, the complex potential is presented as
1 ⎧QF ln ϑ4 (u − b)ϑ4 (u + b) + QK ln ϑ2 (u − in)ϑ2 (u + in) + ⎫
w(ζ ) = ⎨ ⎬. (3.8.5)
π ⎩QD ln ϑ2 (u − im)ϑ2 (u + im) ⎭
Instead of complex velocity w(ζ), let's first consider the derivative of its logarithm
ωζ (ζ)=dlnwz /dζ which is real at the rectangle's sides, so that it also can be extended
analytically over the whole ζ-plane. The fundamental rectangle has single poles at points
O(a), M(id), C(π/2+id) and at symmetrical points −a, −id, −π/2−id. In view of changes of
function ω by moving round these points, one can obtain a leading part as
1 1 ln(v0 / v1 ) ⎛ 1 1 1 1 ⎞
− + − + − (3.8.6)
u−a u+a π i ⎝ u + id u − id u − ic − π / 2 u + ic − π / 2 ⎟⎠
⎜
Hence, derivative ωζ can be presented as a sum of logarithm derivative of theta-functions
[Whittaker & Watson 1927]. After integration, one can obtain the complex velocity as
96
ln( v0 / v1 )
ϑ (ζ − a) ⎛ ϑ1 (ζ + id )ϑ2 (ζ − ic) ⎞ iπ
wz (ζ ) = −v1eiα 1 ⎜ ⎟ (3.8.7)
ϑ1 (ζ + a ) ⎝ ϑ1 (ζ − id )ϑ2 (ζ + ic) ⎠
2(d − c) ⎛v ⎞
with the requisite condition α = π − 2a + ln ⎜ 0 ⎟ . (3.8.8)
π ⎝ v1 ⎠
The last equation is obtained satisfying the conditions at all edges of the rectangle.
For choosing the appropriate branch of multi-value function (3.8.7) it is necessary to raise to
power each multiplier in both numerator and denominator separately. Transformation from
the z-plane to the ζ -plane can be written, as before, in a differential form
dz / d ζ = zζ (ζ ) = wζ (ζ ) / wz (ζ ) . (3.8.9)
Functions (3.8.1) and (3.8.7) as well as transformation (3.8.9) contain seven
unknown parameters (a, b, c, d, m, n, τ); one of them could be found from equation (3.8.8)
as a function of other three. For the remaining six unknown parameters there are only four
conditions for the given coordinates of channel edges E and G (to avoid singularities it is
desirable to integrate function (3.8.9) along diagonals AE and BG). The fifth condition is a
dependence of the speed at infinity in the channel on the given cavitation number,
wz (b + πτ i / 2) = v0 / 1 + σ . (3.8.10)
And the sixth condition could be a plausible assumption on equality of directions of both
jets, EDC and GKM, though these jets can have different directions,
arg wz (im) − arg wz (in + π / 2) = 0 . (3.8.11)
A channel width can be found as a difference of ordinates of points, G and E, or by amount
QF H = yE − yG = QF / v∞ . (3.8.12)
Of course, all values should be dimensionless, i.e. all coordinates should be related to plate
length, which is obtained by integrating differential equation (3.8.9) along the rectangle's
π /2
lower side l = −e − iα ∫
0
zζ (ξ )d ξ . (3.8.13)
The resulting force can be obtained by integration of pressure along the plate. Coefficient of
π /2
ei (α −π )
v∞ l ∫0
normal force is then Cn = 1 + σ − 2 wz (ξ ) wζ (ξ )d ξ . (3.8.14)
In conclusion it should be noted that computing the set of equations is very sensitive to
initial unknown parameters. Some research is needed for computing these equations. Results
of one solution are plotted in Fig.3.8.2.
Fig.3.8.2.
Cavitating
plate in a jet
of fluid
flowing out
from channel
with partial
(a) and full (b)
cavitation.
97
3.8.2. A Cavitating Flat Plate in a Sluice

Consider a plate near a bottom in a stream flowing out of a sluice (Fig.3.8.3a). The problem
solution can be obtained substituting m = πτ / 2 , i.e.
having moved point G in the z-plane away to infinity
and having it coincided with point K. The complex
velocity is expressed by the same relationship (3.8.7). If
a sluice is inclined on the angle, πγ , than the speed
function (3.8.7) will differ by multiplier
Fig.3.8.3. Plate in a stream
from under horizontal Sluice. [ϑ1 (ζ − a ) / ϑ 1 (ζ + a )]γ
only. The derivative of complex
potential (3.8.1) is taken as
ϑ1 (ζ )ϑ2 (ζ )ϑ3 (ζ )ϑ1 (ζ − a )ϑ1 (ζ + a )
wζ (ζ ) = . (3.8.15)
ϑ4 (ζ )ϑ2 (ζ − in)ϑ2 (ζ + in)ϑ4 (ζ − b)ϑ4 (ζ + b)
The new constant factor in (3.8.15) was assumed to be unity, so that the sluice width is
ϑ (b − a )ϑ4 (b + a)
Lv∞ = π 2 4 . (3.8.16)
2ϑ1 (b)ϑ3 (b − in)ϑ3 (b + in)
The direction of speed at infinity in jet D should be horizontal:
ln(v0 / v1 ) ϑ2 [i (n + d )]ϑ1[i (n − c)]
α − π − 2arg ϑ2 (a + in) − ln = 0 (3.8.17)
π ϑ2 [i (n − d )]ϑ1[i (n + c)]
The last two equations together with conditions for two coordinates of points, E, and
equations (3.8.8) and (3.8.10) form a set of six equations that uniquely determine the six
parameters, a, b, c, d, n and τ . The force coefficient can be calculated, as before, by
equation (3.8.14).
3.8.3. A Cavitating Flat Plate in a Channel
Substituting b = π / 2 yields a solution of the cavitating flow around a flat plate in a channel,
Fig.3.8.3. The derivative is taken from (3.8.15) as
ϑ (ζ )ϑ2 (ζ )ϑ1 (ζ − a )ϑ1 (ζ + a )
wζ (ζ ) = 1 . (3.8.18)
ϑ3 (ζ )ϑ4 (ζ )ϑ4 (ζ − b)ϑ4 (ζ + b)
Equation (3.8.16) is transformed to
ϑ (b − a)ϑ4 (b + a )
Lv∞ = π 4 2 . (3.8.19)
Fig.3.8.4. Cavitating plate in 2ϑ1 (b)ϑ22 (b)
a channel. The speed at boundaries CE and MG should be equal to
the speed at infinity at the left, v1 = v∞ , or
ϑ4 (b − a) ⎧2 v ⎫
exp ⎨ ln 0 [ arg ϑ4 (b + ci ) − arg ϑ4 (b + di ) ]⎬ − 1 = 0 . (3.8.20)
ϑ4 (b + a) ⎩ π v∞ ⎭
Conditions (3.8.8) and (3.8.10) are valid in that case as well. Approaching n → πτ / 2 , one
can obtain from (3.8.17) and (3.8.8) the equality, d arg wz (πτ / 2) / d ζ = 0 , or
i ln(v0 / v∞ )
ln'ϑ3 (a ) + [ln'ϑ3 (d ) − ln'ϑ3 (c)] = 0 . (3.8.21)
π
Conditions (3.8.19), (3.8.20) and (3.8.21) together with (3.8.8) and with the given distance h
98
between trailing edge A and the lower side of channel allows computing all five unknown
parameters a, b, c, d, τ . Distance h may be computed integrating function zζ (ζ ) along the
inclined straight line, η = (πτ / b)ξ . One should keep in mind that the cavitation number, like
in Sec. 3.3, cannot be less than any critical value which corresponds to Kirchhoff’s flow in a
channel, i.e. c = πτ / 2, d = πτ / 2 . It should be noted that instead of (3.8.21), another form
of the condition can be obtained approaching first n → πτ / 2 and then m → πτ / 2 :
i ln(v0 / v∞ )
ln'ϑ4 (a ) + [ln'ϑ4 (d ) − ln'ϑ4 (c)] = 0 (3.8.21a)
π
This means that a different succession of approaches would yield different conditions in the
model. However, equations (3.8.21) and (3.8.21a) describe the flow far from the plate and so
the hydrodynamic characteristics could differ little.
3.8.4. Cavitating Flow around a Plate between a Free Surface and a Bottom
Substituting b = π / 2 and v1 = v∞ in equations (3.8.15) – (3.8.17), one can obtain a solution
of flow problem around a cavitating plate between the
free surface and the bottom, Fig.3.8.3.
The derivative of complex potential is
expressed as
ϑ (ζ )ϑ2 (ζ )ϑ1 (ζ − a)ϑ1 (ζ + a )
wζ (ζ ) = 1 . (3.8.22)
ϑ3 (ζ )ϑ4 (ζ )ϑ2 (ζ − in)ϑ2 (ζ + in)
Fig.3.8.5. Plate between the
Speed function (3.8.7) remains valid. Equations (3.8.8)
free surface and the bottom.
and (3.8.17) after substituting v1 = v∞ also remain valid.
πϑ32 (a )
The stream width can be found from (3.8.22) as Lv∞ = . (3.8.23)
2ϑ22 (0)ϑ42 (in)
Integrating derivative (3.8.9) along the straight line connecting point A with an arbitrary
point at the rectangle's upper side (e.g. with the middle point), one can obtain the distance of
the plate trailing edge from the bottom:
⎛ π (1 + i 2τ )ξ ⎞ π (1 + i 2τ )
1
h = − Im ∫ zζ ⎜ ⎟ dξ . (3.8.24)
0 ⎝ 4 ⎠ 4
Also, there are four conditions (3.8.23), (3.8.24), (3.8.8) and (3.8.17). It is not enough to
determine five parameters (a, c, d, n, τ ), i.e. the problem is indeterminate. Equation (3.8.21)
can be used as the fifth condition may. Assuming v0 = v1 = v∞ , one can obtain the flow with
Kirchhoff’s model for which the cavitation number vanishes. The derivative of complex
potential (3.8.22) is remained while speed function (3.8.7) is reduced to
ϑ (ζ − a)
wz (ζ ) = −v∞ eiα 1 . (3.8.25)
ϑ1 (ζ + a)
Equation (3.8.8) gives a = (π − α ) / 2 , so the solution will have only two unknown
parameters n and τ , which can be calculated from equations (3.8.23) and (3.8.24) taking into
account the plate length (3.8.13). The solution has two asymptotic cases, one is transformed
to a symmetric angle flow (trailing edge A touches the bottom), the other corresponds to a
plate gliding over the bottom (liquid flows along the plate into infinity). The first problem
follows from substituting ζ = uτ + π / 2, n = n′τ and approaching τ → 0 . The rectangle is
99
then transformed into a half-string on the u-plane. The second problem is obtained taking
n = 0 . The free jet direction is β = − arg wz (π / 2 + ni ) . Asymptotic cases were considered by
many authors: at the end of the 19th century, symmetric flow around a wedge was considered
by Bobylev, Gerlach, Meshersky, Rethy, and in the middle of the last century, glides
problems were investigated by SA Chaplygin, Gurevich & Yanpolsky, Green, Sedov, YS
Chaplygin. These and others works were outlined in [Gurevich, 1979].
Some results for depth L = 1 and angle of attack α = π / 6 are plotted in Figs.3.8.6
and 3.8.7. The latter plot shows that the flow is possible just for a distance of the trailing
edge exceeded the stream depth. This phenomenon has been shown by numerous
calculations in [Green, 1935]. The force coefficient first decreases then slightly improves
and then abruptly drops down. The latter corresponds to the gliding as shown in Fig.3.8.6
(line 4); the distance is about h ≈ 1.07 . The jet direction above plate is an increasing
function; its maximum value is β = π − α and corresponds to gliding.
Fig.3.8.6. A plate under the free

boundary above bottom at depth L=1
and different distances h=0.2(1),
0.4(2), 0.65(3), 1.0695(4). Fig.3.8.7. Dependences of force
coefficient and jet angle on distance.
3.8.5. Plate in a Free Jet

A picture of free jet can be obtained from Fig.3.8.1a by approaching points E and G to F, i.e.
moving the top of rectangle shown in Fig.3.8.1b far away into infinity, see Fig.3.8.8.
Fig.3.8.8.
A plate in a free
stream.
a)
When parameter τ → ∞ and velocity v1 = v2, equations (3.8.1) and (3.8.7) are transformed
sin 2ζ sin(ζ − a )sin(ζ + a )
to wζ (ζ ) = , (3.8.26)
sin(ζ − im)sin(ζ + im) cos(ζ − in) cos(ζ + in)
100
ln( v0 / v∞ )
iα sin(ζ − a) ⎡ sin(ζ + id ) cos(ζ − ic) ⎤ iπ

wz (ζ ) = −v∞ e . (3.8.27)
sin(ζ + a ) ⎢⎣ sin(ζ − id ) cos(ζ + ic) ⎥⎦
Condition (3.8.8) is valid in that case as well, so the jet direction F coincides with the x-
axis, but jets В and K can get any directions. As before, it can be assumed that those
directions should be the same, i.e. equation (3.8.11) is valid. A position of plate in a stream
could be as in Sec. 3.7 determined by distance h0 between the upper and neutral stream lines.
Expressions for stream width L and distance h0 can be found from derivative (3.8.26):
cosh 2n + cos 2a
Lv∞ = π , h0 v∞ = π . (3.8.28)
cosh 2n + cosh 2m
The plate length is obtained by integral (3.8.13). For given ratios L / l and h0 / l , equations
in (3.8.28) are insufficient for determining four unknown parameters c, d, m, n. The
problem is twice indeterminate. Equation (3.8.11) for equality of directions of jets D and K
can be used as an additional condition. Another condition can be found assuming about non-
rotational flow, i.e. integral on function (3.8.26) along straight line MC should be zero,
w(d i ) − w(ci + π / 2) = 0 . Instead of the latter condition, another assumption can be used, e.g.
equality of abscises of points M and C, or vanishing the wake thickness at infinity past
cavity, or others.
3.8.6. Cavitating Plate beneath Free Surface and in a Semi-Infinite Domain
For the case of a plate beneath the free surface (Fig.3.8.9), parameter m should approach
infinity. Obviously, as m approaches infinity, function (3.8.26) approaches zero. To
overcome this uncertainty, the right-hand term of (3.8.26) should first be multiplied by factor
0.5cosh 2m . It should be noted that, without losing
generality, the derivative could be multiplied by any,
but the same for all linear parameters, factor.
Approaching m → ∞ while taking into account that
factor, yields from (3.8.26) the following expression
sin 2ζ sin(ζ − a )sin(ζ + a )
Fig.3.8.9. Plate beneath the free wζ (ζ ) = . (3.8.29)
surface. cos(ζ − in)cos(ζ + in)
After substituting, v1 = v∞ , speed function
(3.8.27) and equation (3.8.8) will remain valid for this case as well. Consequently, functions
(3.8.26) and (3.8.27) are a parametric solution of the problem. As expected, width L in
(3.8.28) approaches infinity because due to this factor while distance h0 is converted to
π
h0 v∞ =( cosh 2n + cos 2a ) . (3.8.30)
2
The speed direction at point D should be horizontal, thus (3.8.17) will change as follows:
ln(v0 / v∞ ) cosh(n + d )sinh(n − c)
α − π − 2arctan(tan a tanh n) − ln = 0 . (3.8.31)
π cosh(n − d )sinh( n + c)
There are three equations (3.8.5), (3.8.7) and (3.8.31) for determining four unknown
parameters a, c, d, n. The additional condition for vanishing circulation along curve CBAM
cosh 2n − cosh 2c cosh 2c + cosh 2d
is expressed as ln + = 0. (3.8.32)
cosh 2n + cosh 2d cos 2α + cosh 2n
101
With all parameters obtained, all kinematic and dynamic characteristics can be calculated.
With ordinate n of point D approaching infinity, one can obtain a solution for a
cavitating plate in unbounded stream discussed in Sec. 2.6. The derivative of complex
potential (3.8.29) could be written as wζ = sin 2ζ sin(ζ − a )sin(ζ + a ) . (3.8.33)
The speed function (3.8.27) is still valid in this case. Condition (3.8.8) is expressed by the
2(d − c) ⎛ v0 ⎞
same function after substituting v1 = v∞ : α = π − 2a + ln ⎜ ⎟ . (3.8.34)
π ⎝ v∞ ⎠
Approaching n → ∞ , one can obtain from (3.8.31) the same equation (3.8.34), i.e. there is a
single equation for determining three unknown parameters a, c, d. Again the problem is
twice indeterminate. An additional condition could be found from Eq. (3.8.32) with (3.8.33)
as cosh 2d − cosh 2c − 2cos 2a = 0 . (3.8.35)
Another equation can be obtained from (3.8.31) expanding it into series by power of small
parameter δ = e −2 n and equating the coefficient at δ to zero:
ln(v0 / v∞ )
sin 2a − (sinh 2c + sin 2d ) = 0 . (3.8.36)
π
Instead of equations (3.8.35) and (3.8.36), other conditions could be applied, e.g. abscissas
of spiral centers can be given (Fig.3.8.10), or the wake width at infinity should vanish.
Numerous calculations show that only the above-mentioned conditions are realized
easily and the wake boundaries past the cavity at infinity do not intersect, while at other
conditions the wake boundaries can intersect. Fig. 3.8.10 shows cavity shapes at different
distances from the free boundary by satisfying equality of abscissas of spiral centers. Some
dependences of coefficient of normal force Ñn and cavity length Lc are shown in Fig.3.8.11.
The dashed line corresponds to cavitating flow if circulation vanishes.
As mentioned above numerical investigations are associated with indeterminate problems
and are sensitive to the input data. The ragged curves inn Fig. 3.8.11 are evidence of that.
Fig.3.8.10. Cavity shapes and wakes at different depth of submergence by equality of

abscissas of spiral centers.
102
Fig.3.8.11. Cavitating flow beneath a free surface: (a) coefficient of normal force Ñn (solid
line) and cavity length Lc (dotted line); (b) Ñn vs. cavitation number σ at three values of
submergence depth h; dotted line corresponds to vanished circulation.
Sec. 3.9. Cavitating Flow in a Channel with Flat Sides

The Joukowski model (Sec. 2.6) can imitate neither cavity nor thickness of the displaced real
wake. But it allows calculating dynamic characteristics of cavitating flows and
configurations of some flow boundaries in a channel. The latter is used in the hydraulics.
3.9.1. Problem Statement and General Solution
As shown in Fig.3.9.1, the stream flows around inclined flat plate OAB, then forms two free
boundaries OE and BD, which are transformed in two horizontal lines EC1 and DC3 . The
kinematic conditions are satisfied at the inclined plate and horizontal straight lines, while
dynamic conditions are satisfied at the free boundaries. At detachment points O, E, D and B
the hodograph wz has right angle vertices so that they are desirable to map on the vertices of
rectangle, Fig.3.9.1b. Then transformation wz(ζ) from the hodograph to the rectangle is an
analytic function in the rectangle including its vertices. Furthermore, the horizontal sides
correspond to the straight lines of hodograph while the vertical sides conform to circular arc,
so function wz(ζ) can be extended through all the sides of rectangle over the whole ζ-plane.
Then the transformation becomes an elliptic function with simple zero at point a and simple
pole at symmetrical point -a on the fundamental
rectangle of sides π and πτ i , and its expression
coincides with equation (2.6.9) by satisfied condition
(2.6.10).
Fig.3.9.1.
Cavitation flow in a channel (a) - the z-plane view;
(b) - parametric rectangle on the ζ -plane
The definition domain of complex potential is a

stripe with a cut-out on the multi-sheet w-plane. Instead
103
of the multi-value potential it is better to consider, as usually, single-value derivative wζ (ζ ) ,

which can also be extended double-periodically through the rectangle's sides on the whole
ζ -plane; then its singularities are simple poles at points c1 , c2 , c3 and at symmetrical points
−c1 , − c2 , − c3 ; furthermore, it has simple zeros at vertices of rectangle 0, π / 2 , π (1 + τ i ) / 2 ,
πτ i / 2 , at stagnation point a, and at symmetrical point −a . Now, the derivative of potential
ϑ (2ζ )ϑ1 (ζ − a)ϑ1 (ζ + a)
can be written as wζ = 13 . (3.9.1)
∏ϑ4 (ζ − ck )ϑ4 (ζ + ck )
k =1
The latter function together with (2.6.9) is the general solution of the problem. The amount
of fluid flowing in the channels, Ck (k = 1, 2, 3) , is obtained by residue of function (3.9.1)
ϑ4 (ck − a )ϑ4 (ck + a)
Qk = π 3
. (3.9.2)
ϑ2 (0)ϑ3 (0)ϑ4 (0)∏ϑ1 (ck − c j )ϑ1 (ck + c j )
j =1
j ≠k
Substituting coordinates ζ = ck + πτ i / 2 (k = 1, 2, 3) into equation (2.6.9) yields the speed in

ϑ (c − a )
infinity of the channels, Ck , or at points ck vk = 4 k . (3.9.3)
ϑ4 (ck + a )
The channel width in infinity Ck is hk =| Qk | / vk , (k = 1, 2, 3) . (3.9.4)
As before, all geometric parameters are determined integrating differential equation
zζ = wζ / wz . It is shown that derivative zζ (ζ ) is a second order double-periodic function
and can be presented as a sum of fundamental functions, see App. 1, Eq. (1A,2) and then
integrated by expanding into series. For computing, it is better to find the plate length by
π /2
ϑ (2ξ )ϑ1 (ξ + a) 2 d ξ
integral l= ∫ 3 1 . (3.9.5)
0
∏ 4
ϑ (ξ −
k =1
c k )ϑ4 (ξ + c k )
3.9.2. Determination of Unknown Parameters

Functions (3.9.1) and (2.6.9) depend on five unknown parameters: a, c1 , c2 , c3 , t . Parameter
a is found from equation (2.6.10). For the rest of parameters, there are conditions:
• the given ratio of main channel width to the plate length,
h2 / l = H ; (3.9.6)
• the given ratio of distance between the plate and the lower side of channel to the plate
h1 1 πτ / 2
length, − Im ∫ zζ (iη )idη = h ; (3.9.7)
l l 0
• the given cavitation number, v2−2 − 1 = σ . (3.9.8)

These three conditions are usually given, but they do not suffice for determining all the
parameters. In reality, neither velocities v1 or v3, nor widths h1 or h3 could be given, so the
problem is generally undetermined. It should be noted that theoretically each of these values
could be given but they do not have any physical basis. As a plausible hypothesis, speeds at
points c1 and c3 can be assumed equal v1 − v3 = 0 . (3.9.9)
104
Equations (3.9.6) – (3.9.9) and (2.6.10) form a closed loop of a set of equations for
unknown parameters. However, parameters H, h and σ cannot be arbitrary. Let width H be
given as an arbitrary positive number. Distance h should be positive and less than H − sin α .
Instead of h it is desirable to take the ratio
h = h /( H − l sin α ) , 0 < h < 1 . (3.9.10)
As mentioned above for flows in a channel, the cavitation number is limited from
below by minimal number σ0, which corresponds to Kirchhoff flow in a channel, v1 =v2 =1,
and can be found from equations (3.9.6) – (3.9.8). Special cases, such as cavitating flow of a
plate close to a wall, or cavitating flow in an unbounded domain, can be obtained
approaching c3 → c2 , or c1 → c2 , c3 → c2 , respectively [Terentiev, 1965].
3.9.3. Hydrodynamic Force

Again, hydrodynamic force can be calculated integrating the pressure distribution along the
plate. The coefficient of normal resultant force is determined by
Cn = (1 − I / l )(1 + σ ) , (3.9.11)
where I is integral (3.9.5) in which the sign of parameter a is changed.
For theoretical analysis it is expedient to find a drag from the momentum theorem. Let v0
and p0 be the speed and pressure at free boundaries OE and BD. The drag of plate is equal to
D = ∫ ( p − p0 )dy . The projection on the x-axis of resultant force of all pressure is
EOBD
− D + h 2 ( p2 − p0 ) − h1 ( p1 − p0 ) − h3 ( p2 − p0 ) , while increment of pulse is equal to

− ρ mv2 + ρ mv1 , where m= h2 v2 is amount of liquid flowing through any vertical section.
Equating the resultant force with pulse increment and then using Bernoulli integral gives the
v2 ⎛ v ⎞⎛ v 2 ⎞
drag as D = ρ 2 h 2 ⎜ 1 − 1⎟⎜ 0 − 1⎟ . (3.9.12)
2 ⎝ v2 ⎠⎝ v1v2 ⎠
The latter equation is valid for the contour of any form. The drag coefficient is expressed by
⎛ 1+σ ⎞
CD = H ⎜⎜ (
− 1⎟⎟ (1 + σ )(1 + σ 1 ) − 1 , ) (3.9.13)
⎝ 1 + σ1 ⎠
where σ 1 = 2(p1 − p0 )/ρ v12 = v1−2 − 1 . (3.9.14)
3.9.4. Vertical Plate

Let's consider a special case of a plate normal to the channel sides, α = π / 2 . This case has
been described in detail in [Terentiev, 1965]. Skipping all intermediate manipulations, it
π (1 + σ )
gives for the channel width H= , (3.9.15)
P+Q
π ⎛ 1+σ ⎞
where P= ⎜⎜
2 ⎝ 1 + σ1
− 1⎟⎟ ( (1 + σ )(1 + σ 1 ) − 1 ,)
⎠
⎛π 1 ⎞ 1+σ ⎛ π 1 ⎞
Q = σ ⎜ − 2arctan ⎟ − σ 1 ⎜ − 2arctan ⎟.
⎝2 1+σ ⎠ 1 + σ1 ⎝ 2⎜ 1 + σ1 ⎟
⎠
105
2P
The drag coefficient is reduced to CD = (1 + σ ) . (3.9.16)
P+Q
Equation (3.9.14) allows obtaining σ 1 by given cavitation number σ and channel width H.
The Kirchhoff’s flow in a channel corresponds to σ1=0, drag coefficient in this case follows
π ( 1 + σ − 1)(1 + σ )
from (3.9.16) CD = . (3.9.17)
2(1 + 1 + σ )(arctan 1 + σ ) − π
Cavitation number approaching zero yields from the latter expression the drag
coefficient for a plate normal to the wall; the coefficient coincides with Kirchhoff’s formula.
Equations (3.9.15) – (3.9.17) are independent of location of normal plate, h. This
phenomenon for the Kirchhoff’s flow in a channel was noted in [Birkhoff, 1957] and for a
normal plate close to a wall in [Bonder, 1936].
3.9.5. Numerical Analysis
Calculation results for normal plate are plotted in Fig.3.9.2. Cavitation number σ, for the
flow in a channel cannot be less than the minimal cavitation number σ0 for the Kirchhoff’s
flow. Fig.3.9.2a shows minimal cavitation number as a function of channel width for the
Kirchhoff’s flow in semi-logarithmic scale, ln σ 0 ( H ) . Shown also in Fig.3.9.2a is pressure
number, σ1 as a function of cavitation number σ at width H=10. Fig.3.9.2b shows the
relative drag coefficient, CD∗ = CD / CD0 (1 + σ ) , versus cavitation number at different channel
widths, where CD0 = 2π /(π + 4) is the Kirchhoff’s drag coefficient. The curves in Fig.3.9.2
shows that the minimal cavitation number σ0 drastically increases with the channel width
decreasing. This means that reaching a small cavitation number in a cavitation tunnel is
virtually impossible. As seen in Fig.3.9.2b, the drag coefficient depends mostly on cavitation
number and is approximately equal to CD ≈ CD0 (1 + σ ) . The vertical scale in Fig.3.9.2 is very
small so the channel width affects drag coefficient via cavitation number; as mentioned
above, plate location has no effect at all.
Fig.3.9.2.
Right:
lnσ0 (H) for
Kirchhoff’s flow;
left: σ1 (σ) for
H=10; CD∗ (σ ,H )
106
Some numerical results for an inclined plane
with Kirchhoff’s cavity near the wall (upper curve)
and in a channel (two low curves) are plotted in
Fig.3.9.3. It shows the dependence of coefficient Cn
and cavitation number σ0 on relative distance
h = h / H − l sin α . Extreme cases h → 0 or h → 1
correspond to a wedge in the channel of width 2H at
opening angle 2(π − α ) or 2α , respectively.
Fig.3.9.3.
Inclined plate with Kirhhoff’s cavity near a wall and in
a channel.
The Kirchhoff’s cavity and the Joukowski model for inclined plate in a channel are
shown in Fig.3.9.4 where the input data are shown in the top while the calculated data are
show in the
bottom.
Fig.3.9.4.
Inclined plate in a
channel:
(a) - Kirchhoff’s
flow,
(b) - Joukowski
model.
Sec. 3.10. Cavitating Flow of a Cascade of Flat Plates

Investigation of cavitating flow of a cascade is very important for turbo-machines since its
speed can be very high and cavitation can occure. Cascades can be considered as simple
models of axial hydro-turbines and propellers (Fig.3.10.1).
Papers by Kutta [1911] and Chaplygin [1914] were the
pioneering theoretical studies of flow through a plate
cascade. Foundations of hydrodynamic theory of turbines
and cascades have been made by Weinig [1935] and Kochin
[1949], and essentially advanced by Sedov [1950] and
Stepanov [1962].
Fig.3.10.1.
Axial turbine (a) and its evolvement (b)
A complete solution of the cascade flow problem

was presented in [Chaplygin & Minakov, 1930] using the
107
Kirchhoff’s model with infinite length of cavity. Betz & Petersohn [1931], also using the
Kirchhoff’s model, considered a flow throughout a flat plate cascade.
Almost all cavitating models have been applied for considering cavitating flow of
plate cascade: the open Wu-model in [Terentiev, 1967], the re-entrant jet model in [Gusev,
1974], and the model with single spiral in [Vasiliev, 1977; Kuznetsov, 1977], the double
spiral model in [Gubriy, 1969]. See Ch. 14 for more detailed review of cascade flow. Below
the cascade of plate is briefly considered using a modified model with a single spiral.
3.10.1. Fully Cavitating Flow
Consider a flat plate cascade,
Fig.3.10.1a. Each plate is shifted
relative to the lower plate by vector
Teiβ . Velocities in front of and behind
the cascade are denoted v1e − iα1 and
v2 e − iα 2 , respectively. Velocity at the
cavity boundary is assumed to be equal
to unity. For solving the hydrodynamic
problem, the flow domain in the Fig.3.10.2.
physical z-plane is mapped onto the first (a) Cascade of plates,
quadrant of the infinite-sheet ζ -plane (b) first quadrant of
(Fig.3.10.2b). The logarithmic function the parametric ζ-
of complex velocity has the same plane.
boundary conditions as it has for a
single plate with cavity, i.e.
⎧Im ω (ξ ) = 0 on ξ ∈ (0, a ),
⎪Im ω (ξ ) = −π on ξ ∈ (a, ∞),
⎪
⎨ (3.10.1)
⎪Re ω (iη ) = 0 on imaginary axis,
⎪⎩ω ≈ 1/(ζ − i ) at point C (i ).
ζ − a 2bζ
Hence, it can be obtained as ω = ln+ − iπ . (3.10.2)
ζ + a ζ 2 +1
The complex potential has logarithmic singularity at points D1 (ζ 1 ) and D2 (ζ 2 ) , so that its
derivative wζ (ζ ) has simple poles at these points. Furthermore, function wζ (ζ ) has a
simple zero at points 0, A(a) and C(i); it is real on the real axis and imaginary on the
imaginary axis. After analitic extension, the function has singularities and zeros of the same
order at symmetric points, so that it can be expressed in the form:
ζ (ζ 2 + 1)(ζ 2 − a 2 )
wζ = 2 2
. (3.10.3)
(ζ 2 − ζ 12 )(ζ 2 − ζ 1 )(ζ 2 − ζ 22 )(ζ 2 − ζ 2 )
The derivative of mapping the function was, as before, obtained from differential equation
(2.3.9), zζ (ζ ) = wζ e −ω .
All geometric parameters and hydrodynamic characteristics could be expressed by
functions (3.10.2) and (3.10.3). Namely, plate length is calculated by integrating function
(3.10.3) along the positive ξξ- axis, or after substituting ξ=t/(1-t) along segment [0, 1]:
108
1 t (1− t )
t (1 − t )[t 2 + (1 − t ) 2 ][t + a(1 − t )]2 −2 b 2
L=∫ 2 e t + (1− t ) 2
dt . (3.10.4)
0
[t − ζ 1 (1 − t ) ][t − ζ 1 (1 − t ) ][t − ζ 2 (1 − t ) ][t − ζ 2 (1 − t ) ]
2 2 2 2 2 2 2 2 2 2 2
The normal hydrodynamic force is P = ( L − L1 ) / 2 , (3.10.5)

where L1 is expressed like integral (3.10.4) by changing sign of parameters a and b.
Functions (3.10.1) – (3.10.5) contain six unknown real parameters: a, b and two by
two real and imagine parts of complex coordinates ζ 1 and ζ 2 . The period of cascade is
Teiβ = ∫ zζ (ζ )dζ ,
ζ1
while integral about point ζ 2 differs from the first integral only by its
sign; therefore: F ( a , b, ζ 1 , ζ 2 ) + F ( a , b, ζ 2 , ζ 1 ) = 0 , (3.10.6)

iβ
T e = F ( a , b, ζ 1 , ζ 2 ) , (3.10.7)
− ω (ζ 1 , a ,b )
iπ (ζ + 1)(ζ − a )e 2 2 2
where F ( a , b, ζ 1 , ζ 2 ) = 1 1
(ζ 12 − ζ 12 )(ζ − ζ 22 )(ζ 12 − ζ 22 )
1
2
Equations (3.10.6) and (3.10.7) include four real functions. Two other equations could be
obtained by the given speed of the incoming or outgoing flows:
ω (ζ 1 ) = ln v1 − iα1 or ω (ζ 2 ) = ln v2 − iα 2 (3.10.8)
Both speeds in (3.10.8) are assumed given, then the period and angle β could be calculated.
Results of calculations at inlet angle α1=10° and cascade front angle β=100° are
plotted in Figs.3.10.3 to 3.10.5. Fig.3.10.3 shows the domain of solution existence; like for a
flow in a channel, the incoming speed should be less than v1∗ which corresponds to the
velocity at the infinitely long cavity boundary. This extreme flow can be calculated using the
above-obtained formulas if the replacement of ζ2 = i is made in equation (3.10.6) and b=0
in the first equation (3.10.8). All other parameters including incoming speed v1 = v1∗ are to be
calculated. This extreme flow was investigated by [Betz & Petersohn, 1931] who obtained
an analitic expression for the asymptotic curve in Fig.3.10.3, which could be written as
v1as = sin β /(sin( β − α ) + sin α ) . (3.10.9)
Fig.3.10.3. Inflow speed Fig.3.10.4. Normal force

of Kirchhoff Flow. coefficient.
Fig.3.10.4 shows the normal force coefficient as a function of cascade period and
inflow speed. The hydrodynamic force of a cascade is less than that of an isolated plate, but
109
the cavity in cascade is much longer. For the inflow speed v1 < v1as = 0.839 , the flow exists
theoretically for arbitrary period T, but since at some value of T ∗ the cavity boundary
intersects a plate, the flow could not be realized. Thus, the cavity for that flow has always a
finite length.
Some cavity
configurations
are plotted in
Fig.3.10.5.
Fig.3.10.5.
Cavity shapes
in the plate
cascade.
3.10.2. Partially Cavitating Flow

A partially cavitating cascade of plates is shown in Fig.3.10.6a. A parametric solution of the
flow problem can also be found mapping the flow domain in the z-plane onto the first
quadrant of the ζ -plane (Fig.3.10.6b). Since the logarithmic function of complex velocity
has a single pole at point ζ = 0 , the known imaginary part of the plate boundary is
⎧0, ζ ∈ (0, a )
Im ω = ⎨ and its given real part is Re ω = 0, ζ ∈ (0, i∞ ) .
⎩−π , ζ ∈ (a, ∞)
ζ −a b
Hence, ω = ln + − iπ . (3.10.10)
ζ +a ζ
Fig.3.10.6. Partially cavitating cascade:

(left) – flow schematics; (right) - the parametric ζ -plane.
110
Unlike function (3.10.3), derivative of the complex potential has simple zero at the trailing
edge O(ζ = 1) instead of point ( ζ = i ), so that it can be expressed in a similar form
ζ (ζ 2 − 1)(ζ 2 − a 2 )
wζ = 2 2
. (3.10.11)
(ζ 2 − ζ 12 )(ζ 2 − ζ 1 )(ζ 2 − ζ 22 )(ζ 2 − ζ 2 )
All unknown parameters ( a, b, ζ 1 , ζ 2 ) are calculated from the same conditions. Function
F (a, b, ζ 1, ζ 2 ) in conditions (3.10.6) and (3.10.7) should be written as
iπ (ζ 12 − 1)(ζ 12 − a 2 )e −ω (ζ1 , a ,b )
F ( a , b, ζ 1 , ζ 2 ) = . (3.10.12)
(ζ 12 − ζ 12 )(ζ 12 − ζ 22 )(ζ 12 − ζ 22 )
The plate length is calculated integrating function zζ (ζ ) along the ξ -axis in interval (1, ∞ ):
t (1 − t 2 )(1 + at ) 2 e− bt
1
L=∫ dt . (3.10.13)
0
(1 − ζ 12t 2 )(1 − ζ 12 t 2 )(1 − ζ 22t 2 )(1 − ζ 22 t 2 )
The hydrodynamic force can be calculated by
iρ −i ρ ⎛ ⎞
R = X − iY =
2 C∫ wz 2 dz = ⎜ ∫ wz (ζ ) wζ (ζ )d ζ + ∫ wz (ζ ) wζ (ζ )d ζ ⎟ .
2 ⎜⎝ ζ 2 ⎟
ζ1 ⎠
The integrals in the latter can be calculated analytically as residuals of single poles at points
ζ1 and ζ2. After dividing by ρ Lv02 / 2 ( v0 = 1 is the velocity at the cavity boundary), the
resulted force coefficient is:
−i
C x − iC y = CR = ⎡⎣ v1e − iα1 (Γ1 + iQ1 )i + v2 e − iα 2 (Γ 2 + iQ2 ) ⎤⎦ (3.10.14)
L
where Γ and Q are circulation and intensity of a vortex and source at points ζ1 and ζ2:
iπ (ζ 12 − 1)(ζ 12 − a 2 )
Γ1 + iQ1 = G (ζ 1 , ζ 2 ) = , Γ 2 + iQ2 = G (ζ 2 , ζ 1 ) . (3.10.15)
(ζ 12 − ζ 12 )(ζ 12 − ζ 22 )(ζ 12 − ζ 22 )
Since Q1 = −Q2 , the hydrodynamic force in a cascade can be expressed a
Γ + Γ2 Γ + Γ2
Cx = − 1 (v1 sin α1 + v2 sin α 2 ), C y = 1 (v1 cos α1 + v2 cos α 2 ) . (3.10.16)
L L
Expressions (3.10.16) can also be obtained directly from the theorem of momentum
[Sedov & Stepanov, 1962]. The calculated cavity shapes and force coefficients are plotted in
Fig.3.10.7. The short
cavity for the fixed inlet
velocity decreases in a
cascade, but the long
cavity increases while
thickness decreases.
Fig.3.10.7.
Shapes of partial cavity
and some numerical
data.
111
3.10.3. Asymptotic Solution

Consider a large parameter a and a small b for partial cavitating. With velocity function
(3.10.10) taken into account the entry and exit velocities (3.10.8) and their arguments are as
c ⎛ ab ⎞ d ⎛ ab ⎞
follows: vk = 1 − k ⎜ 2 − 2 2 ⎟
, αk = k ⎜ 2 + 2 ⎟ , (k = 1, 2) , (3.10.17)
a⎝ ck + d k ⎠ a ⎝ ck + d k2 ⎠
Denote inlet and outlet cavitation numbers as σ k = vk−2 − 1 , then,
ck ⎛ ab ⎞
σk = 2⎜2− 2 ⎟, . (3.10.18)
a⎝ ck + d k2 ⎠
As seen from equations (3.10.17) – (3.10.18), parameters α1 , α 2 , σ 1 , σ 2 and ω are
small of the same order as parameters b and a −1 , i.e. the asymptotic transformations are
valid for small perturbations of a main horizontal stream. Taking into account the smallness
of function ω (ζ ) except a small neighborhood of the origin, differential equation (3.10.3) is
transformed to dz / d ζ = dw / d ζ . In this case, function w(ζ ) may be found directly due to
logarithmic singularities at points ζ 1 and ζ 2 with factors ±Teiβ , respectively, and so the
transformation from the z-plane with periodic cuts to the ζ -plane may be written as
T ⎡ iβ ζ 2 − ζ 12 − iβ ζ 2 − ζ 12 ⎤
z (ζ ) = ⎢ e ln − e ln ⎥. (3.10.19)
2π i ⎣ ζ 2 − ζ 22 ζ 2 − ζ 22 ⎦
Whence, the derivative is
T ζ ⎡ iβ ⎛ 1 1 ⎞ − iβ ⎛ 1 1 ⎞⎤
zζ (ζ ) = ⎢e ⎜ 2 − 2 2 ⎟
−e ⎜ 2 − 2 2 ⎟⎥
. (3.10.20)
π i ⎣⎢ ⎝ ζ − ζ 1 ζ − ζ 2 ⎠
2
⎝ ζ − ζ 1 ζ − ζ 2 ⎠ ⎦⎥
2
⎛ T 1 − ζ 12 ⎞
Hence, the plate length is Im ⎜ eiβ ln
L = z (1) − z (∞) = ⎟ (3.10.21)
π ⎝ 1 − ζ 22 ⎠
T ⎛ ζ2⎞
and the cavity length is Lc = z (0i ) − z (i∞) = Im ⎜ eiβ ln 12 ⎟ . (3.10.22)
π ⎝ ζ2 ⎠
The derivative (3.10.22) should vanish at the trailing edge, zζ (1) = 0 , or
2c1d1 cos β + (1 − c12 + d12 )sin β 2c2 d 2 cos β + (1 − c22 + d 22 )sin β
− =0. (3.10.23)
(1 − c12 + d12 ) 2 + 4c12 d12 (1 − c22 + d 22 ) 2 + 4c22 d 22
At infinity, the derivative tends to zero as ζ −5 , i.e. the factor by ζ −3 should vanish,
2(c1d1 − c2 d 2 ) cos β + (c12 − d12 − c22 + d 22 )sin β = 0 . (3.10.24)
Besides, an amount of fluid flowing between two neighboring congruent lines should be
constant, 2(α1 − α 2 )cos β + (σ 1 − σ 2 )sin β = 0 . (3.10.25)
One can see from (3.10.17) – (3.10.19) that ratios σ 1 / α1 , σ 2 / α1 , α 2 / α1 depend only
on ζ 1 , ζ 2 and the product, m = ab . Dividing all linear parameters by the length of plate, or
taking L=1 in Eq. (3.10.21), one can obtain a set of 5 equations (3.10.22) – (3.10.26) for the
given cavity length. All other values can be calculated using Eqs. (3.10.17) and (3.10.18).
The force coefficients (3.10.16) are reduced to
112
T ⎡ σ1 ⎛ σ 2 ⎞ ⎛ σ ⎞ ⎤
C y = 2α1 ⎢ ⎜ − 1⎟ cos β + ⎜ 1 − 2 ⎟ sin β ⎥ , (3.10.26)
L ⎣ 2α1 ⎝ σ 1 ⎠ ⎝ σ1 ⎠ ⎦
Cy ⎛ α ⎞
C x = − α1 ⎜ 1 + 2 ⎟ . (3.10.27)
2 ⎝ α1 ⎠
Asymptotic expressions (3.10.17) – (3.10.27) are identical to the respective
asymptotic formulas obtained using the cavitating model with trailing plate [Terentiev &
Vishnevsky, 1970] and agree with the data in [Wade, 1967].
Similar asymptotic expressions could also be obtained for a fully cavitating flow of a
plate cascade. For that it is easier to interchange the cavity and the plate in Fig. (3.10.6).
Then function ω (ζ ) will be pure imaginary on the ξ -axis, and its imaginary part will be
partly constant on the η -axis. It will also have a simple pole at trailing point ζ = 1 , what
ζ − ai ζ
may be written in the form: ω = ln + bi 2 . For a large value of coordinate a and a
ζ + ai ζ −1
small factor b, the asymptotic expansion is ω (ζ ) = 1 + 2iζ / a + biζ /(ζ 2 − 1) + … . Hence,
ck ⎡ ck2 + d k2 − 1 ⎤ dk ⎡ ck2 + d k2 + 1 ⎤
α k = −2 ⎢1 + m 2 ⎥ ; σ k = − 4 ⎢1 − m 2 2 ⎥
. (3.10.28)
a⎣ (ck − d k − 1) + 4ck d k ⎦
2 2 2 2
a ⎣ (ck − d k − 1) + 4ck d k ⎦
2 2 2
The formula for plate length L of partial cavitating (3.10.21) coincides with the cavity length
Lc for fully cavitating flow, also cavity length (3.10.22) coincides with plate length L of full
cavitation. Also, for given L and Lc , the unknown parameters m, c1 , d1 , c2 and d 2 are
calculated from (3.10.21) – (3.10.25) and then angles α1 and α 2 and cavitation numbers σ 1
and α 2 can be found from (3.10.17) and (3.10.18), or from (3.10.28), Finally, hydrodynamic
coefficients (3.10,26) and (3.10.27) are calculated, see Figs3.10.8 – 3.10.9. It should be
noted that the lift coefficient asymptotically coincides with C y ; a fully cavitating flow could
be inverse with respect to the x-axis, then angle β will correspond to π − β .
Fig.3.10.8. Partially cavitating flow through a plate cascade:

ratios α1 / σ 1 (a) and CL / 2πα1 (b) as functions of Lc / L and T0 = T / L .
113
The calculation results in Figs.3.10.8 – 3.10.9 are for partially and fully cavitating
flows. The dashed line in Fig.3.10.8a stands for T / L = 1000 which coincides with results for
a single plate. The dashed lines in Fig.3.10.9 are for ratio σ 1 / α1 , while the solid lines – for
α1 / σ 1 . All curves in Fig,3.10.9 are for fixed ratio T0 = T / Lc ; the curves for T0 → ∞ were
calculated for T0 = 1000 which coincide also with those for a single plate.
Fig.3.10.9. Fully cavitating flow through a plate cascade:

ratios, α1 / σ 1 (a) and 2CL / πα1 (b) as functions of L / Lc and T0 = T / Lc .
Sec. 3.11. Equilibrium Cavitating Vortices in a Stream

Experimental observations show that cavity vortices can appear in a stream. For instance,
entrapment of gas from ventilated cavity is realized through two whole vortices of semi-
infinite length or by ring whole vortices [Epshteyn, 1970]. The ring whole vortices appear
also in jet stream against a wall. A problem statement of whole vortices in a channel and its
general solution ware carried out long ago [Cisotti, 1921]. Cavity vortices connected with
gas lost from cavities was considered by [Cox & Clayden, 1955]. General solutions of some
flow problems connected with whole vortices were obtained in [Hoperskov, 1963]. Many
problems of cavity vortices have been considered in [Terentiev, 1998; Terentiev & Makarov,
1999].
Let's denote the whole vortex with constant pressure as the cavity vortex. Streamline
of a point vortex in unmoved and unbounded fluid is just the whole (cavity) vortex. Because
circulation around a vortex differs from zero, no equilibrium of the vortex in an unbounded
stream is possible. A vortex can be in equilibrium only if it is in a system of vortices, or
close to a solid boundary. Obviously, a cavity vortex approaches a point vortex as the
cavitation number approaches infinity. It can be expected that whole vortices behave similar
to point vortices and many problems could be investigated using point vortices. Generally,
the free boundary differs from a circle but it should be close to that, and a velocity should be
constant at the boundary if the flow is stationary.
114
3.11.1. Vortex near a Flat Bottom

Flow of a whole vortex near a bottom is shown in
Fig.3.11.1a. The vortex is located inside the
circulating flow domain bounded by streamline ACB
and part BA of the bottom. Besides, the value of
velocity at the cavity boundary is constant and can be
assumed to be unity ( v0 = 1 = const ). Stagnation
points A and B belong to the bottom and their
abscissas are known as ±1. Two dimensionless
parameters, cavitation number σ = v∞−2 − 1 and
circulation Γ, determine uniquely a flow of the whole
vortex near a wall or equivalently the flow around
two vortices in an unbounded domain.
The solution, as before, can be obtained Fig.3.11.1. Whole vortex near a
parametrically by conformal mapping a flow domain bottom; (a) – flow schematics; (c)
onto an inner rectangle with vortexes at points parametric rectangle.
0, π , π + iπτ / 4, iπτ / 4 on the auxiliary ζ -plane
(Fig.3.11.1b). After double analytical continuation across horizontal sides, the complex
velocity and the derivative of complex potential are determined by their singularities as
ϑ (ζ − a )ϑ1 (ζ + a )
follows: wz = − 1 , (3.11.1)
ϑ4 (ζ − a )ϑ4 (ζ + a )
ϑ (ζ − a )ϑ1 (ζ + a)ϑ4 (ζ − a )ϑ4 (ζ + a )
wζ = − 1 . (3.11.2)
ϑ12 (ζ )ϑ42 (ζ )
Hence, transformation from the z-plane to the ζ -plane can be written in differential form as
dz ϑ42 (ζ − a)ϑ42 (ζ + a )
= . (3.11.3)
dζ ϑ12 (ζ )ϑ42 (ζ )
The last function is a two-periodic elliptic function and can be presented as a sum of
logarithmic derivatives of theta-functions in its poles ( ζ = 0, ζ = iπτ / 2 ):
zζ (ζ ) = − A ln ′′ϑ1 (ζ ) − B ln ′′ϑ4 (ζ ) , (3.11.4)
ϑ44 (a) ϑ14 (a)
where A= , B= . (3.11.5)
ϑ22 (0)ϑ32 (0)ϑ44 (0) ϑ22 (0)ϑ32 (0)ϑ44 (0)
The derivative of transformation should be zero at symmetric points ζ = ± a + iπτ / 2 , that is
equivalent to the only equation as A ln ′′ϑ4 (a) + B ln ′′ϑ1 (ζ ) = 0 . (3.11.6)
Another equation follows from the condition at infinity for the given velocity, i.e.
ϑ12 (a ) 1
= v∞ = . (3.11.7)
ϑ4 (a )
2
1+σ
Equations (3.11.6) and (3.11.7) allow to calculate two unknown parameters a and τ .
Integrating equation (3.11.4), one can obtain the transformation as
z (ζ ) = − A ln ′ϑ1 (ζ ) − B ln ′ϑ4 (ζ ) . (3.11.8)
115
It is easy to see that z (π / 2) ≡ 0 , i.e. the origin of the z-plane corresponds to point ζ = π / 2
and its imaginary axis coincides as shown in Fig.3.11.1. Using equation (3.11.8), one can
calculate all geometric parameters of the flow including the cavity configuration. The
abscissa of stagnation point B is equal to
xB = z (π − a) = A ln ′ϑ1 (a) + B ln ′ϑ4 (a) . (3.11.9)
Function (3.11.8) transforms the similar domain with abscissa xB . Thus all geometric
parameters should be divided by xB . The cavity boundary is determined parametrically by
1 ⎛ iπτ ⎞ 1 ⎛ iπτ ⎞
functions x(ξ ) = Re ⎜ ξ + ⎟, y (ξ ) = Im ⎜ ξ + ⎟ , ξ ∈ [0, π ] . (3.11.10)
xB ⎝ 2 ⎠ xB ⎝ 2 ⎠
π
1 ⎛ iπτ ⎞
And its length by integral as l=∫
xB 0 ⎝
zζ ⎜ ξ + ⎟ dξ .
2 ⎠
(3.11.11)
Inasmuch the speed at the free boundary is constant, v0 = const , circulation around the
cavity is determined as a product of length and speed. Its dimensionless factor divided by
flow velocity is equal to Γ = l 1+ σ . (3.11.12)
The circulation could also be calculated using complex potential obtained from (3.11.2) as
w(ζ ) = AB ⎡⎣( ln ′′ϑ1 (a ) + ln ′′ϑ4 (a) ) ζ − ln ′ϑ1 (ζ ) − ln ′ϑ4 (ζ ) ⎤⎦ .
−π 1 + σ
Hence Γ= AB [ ln ′′ϑ1 (a ) + ln ′′ϑ4 (a) ] . (3.11.13)
xb
Here the circulation is positive if moves clockwise. The stream function at the cavity
boundary is calculated by ψ c = Im w(iπτ / 4) . The flow line separating the main stream from
circulation flow could be calculated by function (3.11.8) over segment ξ ∈ [0, a ] where
dependence η (ξ ) is determined from equation Im w(ξ + iη ) = 0 . Solving equation
Im w(ξ + iη ) = ψ n , ψ n ∈ (0, ψ c ) for given
abscissa ξ ∈ [0, π / 2] , we can calculate the flow
loop in circulation domain. Some of streamlines
are plotted in Fig.3.11.2. The stream function
values are obtained dividing ψ c = −0.225 into
five equal parts for the upper lines (Fig.3.11.2a)
and ψ c = −0.539 into ten parts for the lower
lines (Fig.3.11.2b). The latter plot shows that the
cavity takes a circular form like streamlines
around a point vortex. This case can also be
formed by asymptotic transfer t → ∞ at that the
rectangle is ultimately transformed to a half-strip
and all functions should be expressed by
trigonometric functions.
Fig.3.11.2.
Streamlines of circulation flow for two
cavitation numbers: (a) σ =4, Γ =6.906;
(b) σ =20, Γ =7.223 .
116
The asymptotic problem of a vortex flow could be solved directly on the z-plane.
Consider a vortex at point D (di ) with circulation Γ 0 . To satisfy a kinematic condition at the
bottom, another vortex should be considered at symmetric point D(− di ) with circulation
Γ z + id
−Γ 0 . The flow is determined by complex potential w0 = z + 0 ln . (3.11.14)
2π i z − id
The vortex is in equilibrium, i.e. the local velocity at point D(di) should be zero. Thus,
Γ 0 = 4π d . Satisfying the condition at stagnation point ( ±1 ), one obtains the ordinate d and
1 4π
circulation Γ 0 : d= , Γ0 = . (3.11.15)
3 3
Ordinate d is marked by black point in Fig.3.11.2. Circulation Γ=7.255 is close to Γ=7.223
calculated for cavity vortex for cavitation number σ =20. Therefore, instead of cavity vortex
in complicated problems, one can consider simpler problem with a classical vortex.
3.11.2. Equilibrium Vortices in a Jet Stream against a Wall
Unlike for a stream with an ordinary vortex, the problem of a whole vortex in a stream
(Fig.3.11.3) is rather complicated. Let a stream flowing out from a channel runs on to a
plate. A vortex is formed in the stream. Let the following is given: the position of the vertical
channel edge, E(1+ih), and intensity of vortex Γ . The half width and the speed are assumed
to be equal to unity. Coordinates of vortex and stagnation points are to be calculated.
Mapping the flow domain in the z-plane onto the first quadrant of the auxiliary ζ -plane, as
shown in Fig.3.11.3, can yield complex potential and complex velocity as follows:
1 Γ ζ 2 − ζ D2
w(ζ ) = ln(ζ 2 − c 2 ) − ln 2 , (3.11.16)
π 2π i ζ − ζ D2
(ζ − a )(ζ − b)(ζ + ζ D )(ζ + ζ D ) ζ − 1
wz (ζ ) = . (3.11.17)
(ζ + a )(ζ + b)(ζ − ζ D )(ζ − ζ D ) ζ + 1
Fig.3.11.3.
Impinging of a jet
against a wall: (a)
the z-plane; (b) the
quadrant on the
ζ -plane
Functions (3.11.16) and (3.11.17) contain five unknown scalar parameters: a, b, c and
ζ D = re β . Two first parameters could be expressed via the rest if conditions at stagnation
points satisfied wζ (a ) = 0, wζ (b) = 0 : a = p − p 2 − q, b= p+ p 2 − q , (3.11.18)
Γr 2
where p = r 2 cos 2β + sin 2β , q = r 4 + Γc 2 r 2 sin 2β . (3.11.19)
2
The remaining three parameters are determined from the vortex equilibrium condition
117
∫ w (ζ )wζ (ζ )dζ = 0 ,
ζD
z (3.11.20)
which yields two scalar functions, and from a condition of the known speed in a channel
Im Wz (c) = V . (3.11.21)
Calculations show that the equilibrium vortex can be located at three different places
close to: the y-axis, corner O, and the horizontal wall. Generally, all three frames can exist
simultaneously. Fig.3.11.4a & b shows the third frame of the vortex and circulation domain
for two values of vortex intensity (Γ=2 and Γ=5). We can assume that the cavity vortex will
be located at the same place where the point vortex is. A cavity near a solid surface makes
damage to it. Cavitation erosion of an aluminum pattern by impinging of jet with diameter
d=2 mm is shown in Fig.3.11.4c, obtained experimentally [Soyama et al., 1998]. Though
cavitation erosion was obtained for the jet filled with bubbles, its mechanism can begin with
formation of cavitation vortex. As shown in Fig.3.11.4c, the jet of radius 1 mm makes
erosion of the ring of radiuses 3mm and 12mm, i.e. the ring cavity vortex should have a
radius about 7.5mm. The distance between the vortex and symmetric axis is exactly 7.5 mm
for circulation Γ = 2 (Fig.3.11.4a) and 8.2 mm for Γ =5 (Fig.3.11.4b), that coincides with
experiments.
Fig.3.11.4. Free boundary and circulation domain by a jet impinging on a wall, V=0.97,
h=3.79: (a) Γ=2, xD =7.49 ; yD =0.16 ; (b) Γ=5, xD =8.12 ; yD =0.45 ; (c) erosion pattern
on an aluminum specimen (cylindrical nozzle, d=2mm, σ =0.03) [Soyama et al., 1998].
3.11.3. Karman Double Vortex Street – Cavitating Vortices

Real flows are usually periodic; this periodic character could mostly be ignored, although in
some cases it should be considered. For example, periodic flow separation at bluff bodies
leads to periodic spatial wake patterns, as modeled by Karman’s double vortex street in a
viscous liquid [Karman & Rubach, 1912]. Pressure close to a vortex point could be negative
what contradicts to real fluids. Similar to the vortex street is the periodic wake after vertical
patterns in a cavitating flow [Tulin, 2003] but in this case the vortices are filled by gas or
cavity cloud. Cavity vortices in a Karman Street are considered below.
Mapping a period of a whole vortices street onto the inner of rectangle with vertexes
at points 0, π , π + iπτ / 2, iπτ / 2 on the auxiliary ζ -plane, we find that the complex velocity
118
wz (ζ ) has single zeros at inner central symmetric points ζ 1 = a + ib and π + iπτ / 2 − ζ 1 , but
the derivative of complex potential should have a simple pole at central symmetric points
ζ 2 = c + id and π + iπτ − ζ 2 corresponding to infinite points on the z-plane. With regard to
its boundary conditions on horizontal sides of the rectangle ( | wz (ζ ) |= 1, Im wζ = 0 ), one
can assume analytically that sides and represented as elliptic functions with periods π and
ϑ (ζ − ζ 1 )ϑ4 (ζ + ζ 1 )
iπτ as follows: wz = − 1 , (3.11.22)
ϑ1 (ζ − ζ 1 )ϑ4 (ζ + ζ 1 )
ϑ1 (ζ − ζ 1 )ϑ1 (ζ − ζ 1 )ϑ4 (ζ + ζ 1 )ϑ4 (ζ + ζ 1 )
wζ = . (3.11.23)
ϑ1 (ζ − ζ 2 )ϑ1 (ζ − ζ 2 )ϑ4 (ζ + ζ 2 )ϑ4 (ζ + ζ 2 )
The transformation from the w-plane to the ζ -plane is written in differential form as
ϑ1 (ζ − ζ 1 ) 2 ϑ4 (ζ + ζ 1 ) 2
zζ = − . (3.11.24)
ϑ1 (ζ − ζ 2 )ϑ1 (ζ − ζ 2 )ϑ4 (ζ + ζ 2 )ϑ4 (ζ + ζ 2 )
Elliptic functions (3.11.23) and (3.11.24) can be written as a sum of logarithmic derivatives
of theta functions according to their singularity. The derivative of complex potential is
wζ = A [ ln ′ϑ1 (ζ − ζ 2 ) − ln ′ϑ4 (ζ − ζ 4 ) ] + A ⎡⎣ ln ′ϑ1 (ζ − ζ 2 ) − ln ′ϑ4 (ζ − ζ 4 ) ⎤⎦ + B , (3.11.25)
ϑ1 (ζ 2 − ζ 1 )ϑ1 (ζ 2 − ζ 1 )ϑ4 (ζ 2 + ζ 1 )ϑ4 (ζ 2 + ζ 1 )
where A= .
ϑ2 (0)ϑ3 (0)ϑ4 (0)ϑ1 (ζ 2 − ζ 2 )ϑ4 (ζ 2 + ζ 2 )ϑ4 (ζ 2 + ζ 2 )
Function (3.11.25) should vanish at stagnation point ζ 1 and an additional constant B is found
as B = − A[ ln ′ϑ1 (ζ 1 − ζ 2 ) − ln ′ϑ4 (ζ 1 − ζ 4 ) ] − A ⎡⎣ ln ′ϑ1 (ζ 1 − ζ 2 ) − ln ′ϑ4 (ζ 1 − ζ 4 ) ⎤⎦ .
Integrating equation (3.11.25), one obtains the complex potential
ϑ (ζ − ζ 2 ) ϑ (ζ − ζ 2 )
w(ζ ) = A ln 1 + A ln 1 −ζ B . (3.11.26)
ϑ4 (ζ + ζ 2 ) ϑ4 (ζ + ζ 2 )
The potential, stream function and circulation along the cavity could be calculated as
ϕ (ξ ,η ) = Re w(ξ + iη ), ψ (ξ ,η ) = Im w(ξ + iη ), Γ = π ⎡⎣i ( A − A) + B ⎤⎦ (3.11.27)
It is evident that the value of circulation is equal to the cavity length.
Equation (3.11.24) could be expressed similarly and the transformation can be obtained
ϑ (ζ − ζ 2 ) ϑ (ζ − ζ 2 )
z = C1 ln 1 + C2 ln 1 +ζ D, (3.11.28)
ϑ4 (ζ − ζ 2 ) ϑ4 (ζ − ζ 2 )
where coefficients C1 and C2 similar to A are determined as residues of function at its poles
−ϑ12 (ζ 2 − ζ 1 ) 2 ϑ42 (ζ 2 + ζ 1 )
C1 = ,
ϑ2 (0)ϑ3 (0)ϑ4 (0)ϑ1 (ζ 2 − ζ 2 )ϑ4 (ζ 2 + ζ 2 )ϑ4 (ζ 2 + ζ 2 )
ϑ12 (ζ 2 − ζ 1 ) 2 ϑ42 (ζ 2 + ζ 1 )
C2 = .
ϑ2 (0)ϑ3 (0)ϑ4 (0)ϑ1 (ζ 2 − ζ 2 )ϑ4 (ζ 2 + ζ 2 )ϑ4 (ζ 2 + ζ 2 )
Additional constant D is, as before, obtained equating function (3.11.24) to zero
D = C1 ⎡⎣ln ′ϑ4 (ζ 1 + ζ 2 ) − ln ′ϑ1 (ζ 1 − ζ 2 ) ⎤⎦ + C2 ⎡⎣ ln ′ϑ4 (ζ 1 + ζ 2 ) − ln ′ϑ1 (ζ 1 − ζ 2 ) ⎤⎦ .
Function (3.11.28) for arbitrary parameters is non-periodic. Its increment roundabout cavity
is calculated by Δ = π [i (C2 − C1 ) + D ] . (3.11.29)
119
Since the arguments of all theta-functions are expressed as a sum or difference of complex
numbers ζ 1 and ζ 2 , then they are transformed to theta-functions on imaginary numbers only
if both numbers have real parts π / 2 or π / 4 . In that case coefficients C1 and C2 are pure
imaginary numbers but term D is a real number, so that increment (3.11.29) is real only. For
this reason the increment of complex function (3.11.26) could also be real. Complex velocity
(3.11.22) at point ζ 2 in that case is real only: v∞ = wz (ζ 2 ) . (3.11.30)
Therefore, we could assume that two possible flows are valid: one corresponds to axial
symmetric flow (real parts a = π / 2, c = π / 2 ), and the other corresponds to axial
symmetric flow ( a = π / 4 , c = π / 4 ). The same results exist for point vortices.
Three unknown parameters b, d and τ could be obtained from given cavitation
number σ = v∞−2 − 1 , circulation Γ and condition for closing cavity Δ = 0 . After numerical
calculations of non-linear equations (3.11.27), (3.11.29) and (3.11.30), streamlines including
the cavity shape could be found, as before, using stream function ψ (ξ , η ) = qn = const and
transformation (3.11.28). The period of the vortices street is calculated as a residue of
function (3.11.24) at point ζ 2 : T = 2π | C1 | . (3.11.31)
Further, coordinates x and y are calculated a ratio of function z (ζ ) to period T, and
circulation is divided by v∞T , i.e. x + iy = z (ζ ) / T , Γ 0 = Γ 1 + σ / T .
Fig.3.11.5.
Cavity shapes of
axisymmetric
double vortices
street:
(a) different
cavitation
numbers,
(b) different
circulations.
Fig.3.11.5 shows cavity shapes in a symmetric double vortices street for different
cavitation numbers (σ =1, 2, 3, 5, 10) and a fixed circulation (Γ0=1.85) (left plot) and for
different circulations (Γ0=1.75, 1.85, 1.95) and a fixed cavitation number ((σ =5) (right plot).
The points in Fig.3.11.6a correspond to locations of the point-vortices in a Karman vortices
street [Kochin et al., 1955]. A distance between two vortices in the symmetric Karman street
is calculated from equation v∞ = (Γ / 2T ) coth π h / T . Calculations show that the distance
between points in Fig.3.11.5a is equal to 0.424 even for cavitation numbers σ ≥ 10 .
Fig.3.11.5b shows that at a fixed cavitation number and circulation increasing, the cavity
move up. Streamlines for a central symmetric flow of Karman vortices street are plotted in
Fig.3.11.6. Streamlines darkened in Fig.3.11.6a separate the main stream from the
circulation one, which round each cavity vortex and flow in the same direction as the main
stream. It should be noted that the circulating streams in symmetric flow (Fig.3.11.5)
separate the main stream from the inner non-circulating stream, which flows in the opposite
120
direction to the mainstream.
Fig.3.11.6.
Streamlines in cavity vortices street:
(a) σ =10, Γ=2.629, Γ0=5;
(b) σ =5, Γ=2.676, Γ0=5.
Fig.3.11.7 shows a cavitating

Karman double vortex street downstream
of a lifting supercavitating foil. Here the
flow is modified from the simpler case
presented above due to three effects: (1)
the foil is operating in a supercavitating
condition, and therefore the cavitation is
not confined to the vortex cores; (2) the
circulation about the foil that is associated
with the lift introduces an asymmetry;
and, (3) the figure shows only the upstream portion of the vortex street, in the transient
region just downstream of the body. Finally, the reader should note that the laboratory
photograph presented in Fig.3.11.7 is from the perspective of a camera fixed to the foil,
whereas the results presented in Figs.3.11.5-6, are depicted in a frame of reference fixed to
the vortex street, which moves relative to the bluff body that generates it. The streamlines of
Fig.3.11.6 are stationary in the frame that moves with the idealized vortex street
(representing the asymptotic behavior far downstream of the body), whereas the image
shown in Fig.3.11.7 is a snapshot of a flow that is unsteady with respect to the camera.
Fig.3.11.7.
Vortex shedding
in the wake of a
supercavitating
hydrofoil. (Image
courtesy of
California
Institute of
Technology and
C.E. Brennen).
121
Ch. 4. Method of Expansion in Series

Sec. 4.1. Cavitating Flow around a Wedge with Curved Sides
4.1.1. Short Review on Expansion in Series
Expansions of analytic function in series are useful for numerical solving of flow problems
with complicated boundary conditions. Some of those problems are discussed in this chapter
applying series expansion. Information on the mathematical problem statement can be found
in many books on analitic functions, e.g. [Lavrent’ev & Shabat, 1965].
Let's present, without proof, some fundamental theorems which will be used further.
Cauchy theorem. An analytic function on any open circle-domain with center "a"
∞
f ( n ) (a)
can be expanded into power series1 f ( z) = ∑ ( z − a)n , (4.1.1)
n=0 n !
which converges uniformly in any closed region from this open circle-domain.
If an analytic function is given then it can be calculated expanding numerically into Taylor
series with a finite number of terms with precision determined by
( z − a) N +1 f ( z ′)dz ′
RN =
2π i ∫
C
( z ′ − z )( z ′ − a) N +1
. (4.1.2)
For computing it, behavior of series of analytic functions should be considered.

∞
Weierstrass’s theorem. If series F ( z) = ∑ fn ( z) (4.1.3)
n =0
converges uniformly in domain D including boundary C in which functions f n ( z ) are

analytical in domain D and continuous across boundary C then the sum, F (z ) , is analytical
in D and continuous at C. Series (4.1.3) can be differentiated, term by term, any times in D.
Also, the power series (4.1.1) should converge uniformly both in the domain and at
the boundary. Let the convergence domain of power series be a unit circle, then its real and
imaginary parts on the circle, z = exp( β i ) , are represented as Fourier series
∞ ∞
U (β ) = ∑ ( an cos nβ − bn sin nβ ), V ( β ) = ∑ ( bn cos nβ + an sin nβ ) . (4.1.4)
n =0 n =0
Now, the Fourier series theory [Smirnov, 1957] can be applied.
Dirichlet’s theorem. If a periodic function together with its derivatives of k-s order
is continuous and its k+1-s derivative has finite numbers of discontinuity of the first kind
then Fourier series converges and the coefficients approaches zero as
M M
| an |< k +1 , | bn |< k +1 for n → ∞ . (4.1.5)
n n
Also, if function f ( β ) is discontinuous of the first kind, then the Fourier series
converges very slow as 1/n, therefore applying the Fourier series to numerical computing is
questionable. Convergence of Fourier series can be increased. For that an auxiliary
function, ω0 ( β ) , the same discontinuity is to be found. Then the difference
f1 ( β ) = f ( β ) − ω0 ( β ) will be continuous at all points within interval [0, 2π ] thus the
1
Series (4.1.1) was derived for real functions first by B. Taylor and systematically used by Maclaurin.
122
Fourier series coefficients of the function approach zero not slower than quantity, 1/ n 2 does.
Turning back to function ω ( z ) of complex variable in a unit circle, an auxiliary function,
ω0 ( z ) , of the same singularities can be obtained from the value problem. Then the
difference ω1 ( z ) = ω ( z ) − ω0 ( z ) (4.1.6)
can be expanded in power series with the coefficients decreasing not slower than n −2 does.
4.1.2. Parametric Expressions of Potential and Its Derivative
Consider a cavitating flow around a curved symmetrical wedge with opening angle 2πμ ,
Fig.4.1.1a. Mapping the flow domain in the z-plane
onto a half inner circle of unit radius on the ζ -domain,
Fig.4.1.1b, and then extending analytically on another
half inner circle, the flow problem can be reduced to
value problems on the whole circle.
Fig.4.1.1.
Cavitating flow around a wedge:
(a) flow on the z-plane;
(b) parametric circle on the ζ -plane.
The boundary condition of the complex

potential is known; its imaginary part is constant at the
boundary, so that it can be extended analytically through the boundary onto the whole ζ -
plane and expressed by known singularities as simple poles at point D(c) and at symmetric
points D′(1/ c) , D′′(−c) , D′′′(−1/ c) . Without losing generality the constant item can be
⎛ 1 1 1 1 ⎞
zero, thus w = m⎜ − − + ⎟, (4.1.7)
⎝ ζ − c ζ + c c(cζ − 1) c(cζ + 1) ⎠
where m is an unknown real factor. As usually, it is convenient to obtain derivative of
complex potential, which has a second-order pole at points D(c), D′(1/ c) , D′′(−c) ,
D′′′(−1/ c) and a simple zero at points O(1) , , A(−i ) , B (i ) , C (0) and at symmetric point
ζ (1 − ζ 4 )
O′(−1) . Hence, wζ = N . (4.1.8)
(ζ 2 − c 2 ) 2 (c 2ζ 2 − 1) 2
A connection between factors m and N is established by factorizing the last function on
common fractions: m = − N / 4c(1 − c 4 ) . (4.1.9)
As before, factors N or m can be assumed to be equal to unity.
4.1.3. Representation of Logarithmic Function of Speed by Power Series
It is evident that the real part of function ω = ln wz (ζ ) is constant on the vertical diameter
corresponding to the cavity boundary. As before, velocity can be assumed to be unity at the
free boundary, thus function ω(ζ) is pure imaginary function on the η-axis and can be
extended analytically through this axis over the whole circle. Then we get an analytic
function ω(ζ) on ring 0 < ζ < 1 with single pole at point C (ζ = 0) . Therefore, it can be
expanded in Taylor series as
123
c−1
ω= + c0 + c1ζ + c2ζ 2 + + ck ζ k + . (4.1.10)
ζ
Generally, all coefficients are complex numbers, but in this case because of
symmetrical flow they should be real and vanish for even subscripts. Further, one can find a
derivative of transformation zζ = wζ e −ω (ζ ) and calculate coefficients of series (4.1.10),
unknown parameter c and factor N satisfying appropriate geometric conditions on the
wetted boundary of the wedge connecting with circle | ζ |= 1 and closure condition (2.1.6d).
However, calculated coefficients converge very slowly due to discontinuity of velocity at
stagnation point O (ζ = 1). This discontinuity and a simple pole can be considered by
additional function (logarithmic function of velocity for cavitating flow of a wedge) as
1−ζ ⎛ 1⎞
ω0 = 2 μ ln + b⎜ζ + ⎟ . (4.1.11)
1+ ζ ⎝ ζ⎠
Then the difference, F (ζ ) = ω (ζ ) − ω0 (ζ ) , is a continuous function over the whole circle
ζ ≤ 1 , and ω and ω0 are real on the real axes. Thus, the coefficients of power series
∞
F (ζ ) = ∑ A2 k −1ζ 2 k −1 (4.1.12)
k =1
should have odd subscripts only and converge much better than the coefficients in (4.1.10).
If the coefficients are given, then unknown parameters c and b could be obtained
from conditions at infinity:
• for velocity ω (c) = v∞ = − ln 1 + σ
1− c 1 + c2 ∞ 1
2 μ ln +b + ∑ A2 k −1c 2 k −1 + ln(1 + σ ) = 0 ; (4.1.13)
1+ c c k =1 2
• for single-value function z (ζ )
4μ 1 − c2 ∞
+ b 2 − ∑ A2 k −1 (2k − 1)c 2 k − 2 = 0 . (4.1.14)
1− c 2
c k =1
Transformation z (ζ ) and all geometric and dynamic characteristics can be calculated using
functions wζ (ζ ) and ω (ζ ) = ω0 (ζ ) + F (ζ ) . Unknown factor N is determined if a length is
given, e.g. a wedge base, d. If factor N is given then all linear dimensions can be calculated.
Small parameter c in equation (4.1.13) corresponds to a small cavitation number if ratio b/c
is as small as c. Equation (4.1.14) confirms this assumption and yields b / c 2 = A1 − 4 μ . Then
the asymptotic solution of equations (4.1.13) and (4.1.14) can be found as
σ σ2
c= + O (σ 2 ), b=− + O (σ 2 ) (4.1.15)
4(4 μ − A1 ) 16(4μ − A1 )
Hence, parameters c and b approach zero as cavitation number σ approaches zero only if
coefficient A1 < 4μ because parameter c must be positive. This implies that Kirchhoff flow
could be obtained by continuous limited conversion if A1 < 4μ . Nevertheless, equations
(4.1.13) and (4.1.14) could have another solution by σ = 0 but it yields a solution different
from Kirchhoff flow and parameter c differs from zero.
124
4.1.4. Boundary Condition on the Wedge

The wedge wetted boundary corresponds to half circle ζ = eit , Fig.4.1.1b. The angle of
inclination to the boundary is calculated as
∞
β (t ) = − Im ω (eit ) = ±πμ − ∑ A2 k −1 sin(2k − 1)t , (4.1.16)
k =1
where positive sign is assigned to the wedge's upper side and negative sign to the lower side.
∞
2μ
2 b cos t + ∑ A2 k −1 cos(2 k −1) t
Speed is expressed as v(t ) = tan (t / 2)e k =1
. (4.1.17)
The differential of complex potential over the circle is obtained from (4.1.8) as
4sin t cos t
dϕ = Nf (t )dt , f (t ) = . (4.1.18)
(1 − 2c 2 cos 2t + c 4 ) 2
Now, the wetted shape of wedge could be calculated from integral as parametric functions
f (t ′)cos β (t ′) f (t ′)sin β (t ′)
t t
x(t ) = N ∫ dt ′, y (t ) = N ∫ dt ′ . (4.1.19)
0
v(t ′) 0
v(t ′)
Functions (4.1.19) could be used as a condition for calculating unknown coefficients
if the wedge boundary is given. Similar integral equations for Kirchhoff model were
obtained by Levi-Civita [1907], Villat [1911] and Nekrasov [1947]. These equations are
very inconvenient for numerical calculations because of a nonlinear nature of integral
equations. The same model had been calculated numerically by [Brodetsky, 1923] and then
[Schmieden, 1929] using a relationship for boundary curvature. The curvilinear abscissa
along the boundary could be written from equations (4.1.19) in integral form as
f (t ′)
t
s (t ) = N ∫ dt ′ . (4.1.20)
0
v(t ′)
Function (4.1.20) and equation (4.1.16) can be used for calculating the coefficients. Instead
of angle (4.1.16), the coefficients can be calculated by satisfying the condition for curvature
dβ f (t )
κ(s)=βt/st or the following equation: = Nκ (s) , t ∈ ( − π 2, π 2) . (4.1.21)
dt v(t )
Equations (4.1.13) and (4.1.14) together with one of conditions (4.1.20) or (4.1.21)
are sufficient for uniquely determining all unknown parameters including the coefficients but
only if detachment points A and B are given. If the points are unknown then the finite
curvature conditions [Willat, 1911] at the detachment points should also be satisfied:
d
Re ω ±π / 2 = 0 . (4.1.22)
dt
4.1.5. Collocation and Iteration Methods
Because of the analytic nature of function F(ζ), coefficients of series (4.1.12) are
determinable uniquely by any numerical approaches. But numerical methods allow
calculating only a finite number of coefficients. Thus, generally, the coefficients could differ
from each other depending on the method applied. Nevertheless, we can expect that the
difference should be insignificant if the number of coefficients is large.
The collocation method contemplates to satisfying the above-mentioned equations
(4.1.19) and (4.1.21) at a finite number of points on the circular arc connected with the
125
wedge wetted surface. The number of points should correspond to the number of coefficients
to be used for solving a set of non-linear equation for unknown parameters and coefficients.
For calculating the coefficients in an iterative manner, taking into account equations (4.1.16)
– (4.1.17), equation (4.1.21) can be written as
∞
∑ (2k − 1)c
k =1
2 k −1 cos(2k − 1)t = −8 N k ( s ) g (t )cos te − β ( t ) , (4.1.23)
sin1− 2 μ (t / 2)cos1+ 2 μ (t / 2) ∞
where g (t ) = , β (t ) = −2b cos t − ∑ A2 k −1 cos(2k − 1)t .
(1 − 2c 2 cos 2t + c 4 ) 2 k =1
π
2
32 N
Thus,
π (2k − 1) ∫0
A2 k −1 = − (κ (s) g (t )eβ (t) cos t cos(2k − 1)t ) dt , (4.1.24)
If factor N and parameters a and b are known, the coefficients could be calculated iteratively
beginning from initial values A2 k −1 = 0 . It should be noted that convergence of iterative
process depends on the integrand and would not necessary be high, so that its application can
be problematic. Nevertheless, the iterative method allows reaching desired accuracy.
Condition (4.1.22) for smooth detachment points becomes algebraic as follows:
∞
2μ − 2b + ∑ (−1) k (2k − 1) A2 k −1 = 0 . (4.1.25)
k =1
Taking into account (4.1.24), the sum in (4.1.25) is represented as integral with the sum of
terms (−1) k cos(2k − 1)t , which could be written as
∞
1 ⎛ cos 2mt ⎞
lim ∑ (−1) k cos(2k − 1)t = − + lim ⎜ (−1) m ⎟.
m →∞
k =1 2cos t m →∞
⎝ 2cos t ⎠
Since integral (4.1.24) of the integrand with factor cos 2mt vanishes as m approaches
infinity, hence condition (4.1.25) is transformed to integral:
π
2
16 N
π ∫0
κ ( s ) g (t )e β (t ) dt = 0 .
2 μ − 2b + (4.1.26)
This equation helps to find factor N on each iterative step. The common iterative process
should be realized as follows: initial values of coefficients are assumed to be zero, then
parameters b and c are calculated from equations (4.1.13) and (4.1.14), then factor N is
calculated from (4.1.26) and coefficients ck are determined from (4.1.24). The cycle of
iterative process is repeated to the desirable accuracy. The number of coefficients could be
changed depending on the accuracy required. If the length of wetted boundary is given, then,
instead of equation (4.1.26), a condition for that length should be satisfied integrating
equation (4.1.20) within interval ( 0, π / 2 ). All other characteristics as cavity shapes and
drag are calculated using functions (4.1.8), (4.1.11) – (4.1.19). Drag is obtained by integral:
π
2
⎛ 1 ⎞
D=N∫⎜ − v(t ) ⎟ f (t )sin β (t )dt . (4.1.27)
0 ⎝
v (t ) ⎠
Some numerical results of cavitating flow of circular lens calculated iteratively are plotted in
Fig.4.1.2. The edge half angel is denoted πμ = α . As seen in the plot, cavity size decreases
very fast with angle α decreasing, at certain angles (here α<40°) the cavity intersects the
lens. This implies that cavitating flow at these angles and cavitation numbers σ ≥ 1 is
126
impossible. The dashed lines represent shapes of lens and dot straight lines show sides of
angle α . Detachment points are jointed by the broken curve. Main characteristics of these
cases are presented in the table in Fig.4.1.2: CD = D / v∞2 l is drag coefficient ( l = 2 R sin πμ is
the chord of a lens), H is the width of cavity, xB and yB are the coordinates of detachment
point B, xC is the abscissa of spiral center. It is to remind that the drag depends on the
distance between spiral centers as D = ρ v02 yc so that yc = CD / 2(1 + σ ) .
Fig.4.1.2. Cavity shapes for cavitating flow of circular lenses at different edge angels 2α
at σ=1.
Fig.4.1.3.
Cavity
shapes for
cavitating
ellipse.
Cavity shapes for cavitating flow around ellipses of different aspect ratio δ
calculated in a collocation manner for cavitation number σ=1 and a fixed detachment point
of 60° are plotted in Fig.4.1.3. The ellipse curvature is not constant but depends on abscissa.
The cavity for elliptical arc of relative thickness 0.25 is located fully within the ellipse. That
means that no such flow can really exist. Using Levi-Civita power series expansion followed
by iteration one can calculate many other problems in the hydrodynamics. Cavitating flow of
a circle cylinder will be discussed below in detail using both collocation and iteration.
Sec. 4.2. Cavitating Flow around a Circular Cylinder

Many authors have theoretically studied the cavity flow past a circular arc with a detachment
and a cylinder with smooth one. The first numerical results were obtained by Brodetsky
[1923] and Schmieden [1929]. They calculated a flow past a circular cylinder with an
attached Kirchhoff cavity by retaining three and five coefficients in the power series
expansion. Kawaguti [1953], Roshko [1954], Eppler [1954] calculated the cavitating flow
using the model with two parallel straight lines. Gurevich [1953] applied the cavitating
model with re-entrant jet and calculated two steps of iterative process. The most detailed
numerical results on cavitating flow past a circular and an arc have been obtained in
127
[Terentiev, 1981; Dimitrieva, 1996; Terentiev & Dimitrieva, 1998].
4.2.1. Numerical Results for Circular Cylinder
For circular cylinder, the curvature is constant and is reciprocal to the radius of cylinder.
Thus the curvature in all formulas of Sec. 4.1
should be replaced with κ ( s ) = −1/ R . The
problem was calculated in iterative and
collocation manners both by given detachment
points and applying the condition of smooth
separation. The first of them employs a fixed
number of coefficients while in the second
approach the number of terms in series
(4.1.12) depended on the required accuracy.
The number of coefficients for both
approaches increases with an increase of
detachment angle γ. Convergence of
coefficients to accuracy of ε =10−5 is shown in
Table 4.2.1. Plotted in Fig.4.2.1 is drag
Fig.4.2.1. Drag coefficient as a function of
coefficient CD = D / ρ R v∞2 as a function of
detachment angle and cavitation number.
detachment angle γ at different cavitation
numbers. The curves have two maxima corresponding to
two smooth detachment points. In reality only the first Table 4.2.1. Coefficients for
smooth detachment can be realized, the second one is of two smooth detachment points
theoretical interest only. For this case and for all greater γ 60°7' 138°54'
detachment angles, the cavity boundaries are very small CD 1.0751 0.9630
and are located inside the cylinder. c 0.1692 0.7422
Stepanov [1998] gave reasonable consideration b -0.0299 -0.3917
for the behavior of these curves connected with A1 1.0450 2.8386
experimental data for a viscous fluid: a boundary A3 0.0030 0.5454
laminar layer separates from the body at the point near A5 0.0009 0.1800
the first maximum, and the turbulent separation A7 -0.0001 0.0709
corresponds to the minimum of the curve. An A9 - 0.0303
explanation of this phenomenon may be that the stream A11 - 0.0137
line curvature at the detachment point among two A13 - 0.0064
maxima has negative infinite value, besides, all the A15 - 0.0034
cavity boundary near the first maximum is outside of the A17 - 0.0027
body, and therefore both the potential stream and the A19 - 0.0007
laminar boundary layer separate from the body at any
A21 - 0.0004
point. For the detachment points lying near the
A23 - 0.0002
minimum, the cavity boundary is very short so that it can
A25 - 0.0001
contact the body at the cavity end and generate bubbles,
which can cause turbulent separation. In addition, we
calculated the cavity flow of the circular arcs with zero cavitation number (Kirchhoff’s
model) and also with a negative cavitation number. Dependence of drag coefficient on arc
size at σ = 0 and σ = −0.5 is also shown in Fig.4.2.1.
128
A flow with cavitation number σ = 0 and angles 0 ≤ γ ≤ 124.2o corresponds to the

Kirchhoff model while angles γ >124.2° yield finite cavities located within the circle. The
values of separation angle γ for which max CD is obtained within the range of γ >124.2°
also corresponds to the regime of
smooth cavity separation. Fig.4.2.2
shows cavities for some negative
cavitation numbers; all these cavities
have smooth detachment points
determined by angle γ o and a
Fig.4.2.2. Cavity shapes at negative cuspidal point at the cavity end. The
cavitation numbers. flow with fixed detachment points is
executed for the arcs with sizes
γ > γ o , where γ o is the size of the arc, for which the flow with a given negative cavitation
number goes according to the Chaplygin–Kolscher model. One of those curves for σ = −0.5
and angle γ ∈ (γ 0 , π ) , γ 0 = 148.25 is plotted in Fig.4.2.1. It should be noted that the
Chaplygin-Kölcher flow takes place at cavitation numbers σ ∈ (−1, 0) .
4.2.2. Method of Double Series Expansions

Another solution of the plane problem associated with solution of a relevant axisymmetric
problem is presented below. Consider the outside of a unit circle as a parametric domain,
Fig.4.2.3. The arc and the cavity boundary are mapped at the circle so that point t = eiα 0 is
relevant to the detachment point, where α 0 is initially unknown. The corresponding points
on Figs.4.2.1a and 4.2.3 are denoted likewise.
Fig.4.2.3. Parametric t-plane.
In this case the complex potential is expressed by

w(t ) = a ( t + 1/ t ) . (4.2.1)
Instead of function ω (t ) , the mapping
transformation, z (t ) , is expanded into power series.
For that a singularity of the function should be
segregated. Let function z0 (t ) has the same singularity at the cavity end as the mapping
function z (t ) , so that the derivative of mapping function is expressed as the sum
dz dz0 dz1
= + . (4.2.2)
dt dt dt
The first term considers the particularity of the stream at the cavity end, Sec. 2.1, and can be
dz0 t 2 − 1 t2−c1
written as =a 2 e , (4.2.3)
dt t
The second tern represents a continuous function, it can be expressed as the following series
dz1 ∞ an
=∑ . (4.2.4)
dt n = 0 t n
129
Here a, c, an (n = 0,1, ...) are unknown real constants.

The expression for the conformal mapping function on the parametric region circle,
t = eiα , is determined integrating equation (4.2.2) with respect to parametric angle α
z (α ) = z0 (α ) + z1 (α ) , (4.2.5)
π
e2iβ − 1 ⎛ 2c ⎞
where z0 (α ) = −ia ∫ iβ
exp ⎜ iβ ⎟dβ , (4.2.6)
α e ⎝ e −1⎠
∞
a
z1 (α ) = a0 eiα + ia1α + ∑ n e − i ( n −1)α + C1 + iC2 , (4.2.7)
n=2 n − 1
∞
a
C1 = a0 − ∑ ( −1) n −1 n , C2 = −π a1 .
n=2 n −1
The spiral center coordinates at cavity end are determined integrating (4.2.2) with respect to
∞
α from 0 to π :
⎣ ( )
xC + = 2ae − c ⎡ 2 − c Ei (c) − ec Ei (−c) ⎤ + C1 + a0 − ∑ n ,
⎦
a
n=2 n − 1
yC + = 2acπ e + C 2 ,
−2 c
where Ei ( x) is an integral exponential function.

2a sin α
Velocity distribution on the arc is V = , α0 ≤ α ≤ π . (4.2.8)
2 2
⎛ dx ⎞ ⎛ dy ⎞
⎜ ⎟ +⎜ ⎟
⎝ dα ⎠ ⎝ dα ⎠
All other kinematic and dynamic characteristics are found from the above functions.
Namely, the drag of the arc is determined by
⎛ ⎛ dw ⎞
2
⎞ ⎛ 2 π
dy ⎞
D = ρ ⎜ V0 2 ∫ dz + ∫ ⎜ ⎟ dz ⎟ == ρ ⎜ V y + ∫ V2 dα ⎟ , (4.2.9)
⎜ dz ⎠ ⎟ ⎜ 0 B
dα ⎟
⎝ AOB AOB ⎝ ⎠ ⎝ α0 ⎠
where yB is the detachment point ordinate.
4.2.3. Numerical Solution

In numerical solution of the problem, the number of terms in the series (4.2.4) is limited by a
certain value, N . Unknown parameters a, c, an (n = 0,1, ...) , N could be determined
satisfying the following conditions:
• condition of uniqueness of function z(t) at infinity,
dz
∫t =∞ dt dt = 0 , or 2ac + a1 = 0 ; (4.2.10)
a
• condition of velocity at infinity wz (∞) = v∞ , 1 + 0 = σ + 1 ; (4.2.11)
a
N
• condition at the “closure” point z1t (1) = 0 , ∑a
n=0
n =0. (4.2.12)
Equations (4.2.10) - (4.2.12) allow finding only three unknown parameters. Other N+1
equations can be obtained satisfying the wetted boundary equation at M1 points α∈(α0 ,π)
and dynamic condition at M 2 = N + 1 − M 1 points on the cavity boundary, α ∈ (0, α 0 ) . Thus
the problem is reduced to solving a system of N+4 nonlinear equations, which can be solved
130
numerically. Of course, the presented numerical approach is more complicated than that
described in Sec. 4.1, but it is preferable over others for solving axisymmetric problems.
Fig.4.2.4.
Cavity shapes calculated for different
collocation numbers; squares
correspond to first method.
Cavity shapes at γ = 60°, σ =1

and various M 1 and M 2 calculated by
Dimitrieva are shown in Fig.4.2.4. For comparison, the calculated results from Fig.4.1.3 are
plotted as small squares. When the amount of collocation points changes at M 1 from 4 to 8
and at M 2 from 6 to 20 the cavity shape does not change significantly. In the numerical
solution the amount of collocation points, their position on the arc and the cavity boundary
vary depending on the arc size: as γ increases, M 1 grows and M 2 decreases.
Sec. 4.3. Cavitating Flow in a Longitudinal Gravity Field

A flow in a gravity field as a flow of an imponderable fluid is determined by the same
equations; the only difference is in the free boundary condition presented as
v 2 + 2 g y = const , (4.3.1)
where v is the velocity, g is the gravity acceleration directed opposite to the y-axis. Condition
(4.3.1) is valid for constant pressure. Differentiating with respect to velocity potential, one
can write the boundary condition in differential form as
d ln v = −v −3 g sin β dϕ , (4.3.2)
where β is the tangent angle to the free boundary, dy / dϕ = v −1 sin β . Obviously, the
gravity can influence the flow only if there are free boundaries. But conditions (4.3.1) or
(4.3.2) cannot be, in general, satisfied exactly and no analytical solutions can be obtained.
Only a few problems of flows with free boundaries of special forms have been solved
analytically.
4.3.1. Joukowski’s Analytical Solution
Joukowski [1891] was the first who obtained an exact solution of a problem of a flow with
stagnation point on the free boundary, Fig.4.3.1. He proposed an original way of a satisfying
the exact condition at the free boundary in the gravity field using auxiliary analytic
functions. If the logarithmic function of complex velocity is sum of two analitic functions
ω = ω1 + ω2 , (4.3.3)
which at free boundary C0 satisfies conditions Im ω1 = 0, Re ω2 = 0 , then the first
function can be expressed vie the other one as

1
( )
ω1 = ln −3g ∫ sin(iω2 ) dw .
3
(4.3.4)
It is easy to find that function ω(w) satisfies condition (4.3.2). Choosing appropriate function
ω2, one can obtain a flow of gravity fluid with free boundary. The Joukowski flow in
Fig.4.3.1 corresponds to function ( ) (
ω2 = i arcsin w , ω1 = ln 3 2 g w . ) (4.3.5)
131
The complex velocity is wz ( w) = eω ( w) = 3 2 g wei arcsin w

. (4.3.6)
Hence, the inlet velocity at infinity is obtained as v∞ = 3 2 g 2 . (4.3.7)
The free boundary from (4.3.6) and (4.3.7) is determined by functions
ϕ
x(ϕ ) =
1
2v∞
( )
arcsin ϕ + t (1 − t ) , y (ϕ ) = −
2v∞
. (4.3.8)
Fig.4.3.1. Joukowski flow.
The calculated curve is plotted in Fig.4.3.1. The

distance between two vertical plates is equal to
h = 2 x(1) = π / 2v∞ , thus the Froude number
obtained relatively to this distance and the
stream velocity is equal to
Fr = v∞2 / gh = 1/ 2π . (4.3.9)
The Joukowski flow is valid for a single value
of Froude number only. Exact analytical solutions of some special cases were obtained in
[Sautreaux, 1893; Bervy, 1894; Rudzki, 1898; Villat, 1915].
4.3.2. Review of Gravity Problems
Generally, the gravity effect should be studied using different approaches. One of them is an
expansion in power series. This approach has been applied to the non-linear problem of
gravity effects using different cavity models: a double rectilinear model in a longitudinal
gravity field [Lenau, 1963], Tulin’s second model in a longitudinal gravity field [Larock &
Street, 1967], Riabouchinsky model of partially cavitating flow in a gravity field [Kotlyar &
Terentiev, 1970], a double rectilinear model in an inclined gravity field [Vishnevsky et al.,
1974]. Cavitating flows in a transverse gravity field have been considered by using the re-
entrant double rectilinear model [Kotlyar & Troepolskaya, 1975], Riabouchinsky model
[Terentiev at all, 2005] and single spiral model [Terentiev, 2004]. In this section, the value
problem of cavitating flow around a wedge in a longitudinal gravity field is considered using
the cavitation model with single spiral.
Consider a 2-D steady cavitating flow past a
wedge with opening angle 2πμ, as shown in
Fig.4.3.2a.
Fig.4.3.2.
A cavitating wedge in longitudinal gravity field:
(a) flow in the z-plane; (b) parametric circle.
Assuming an inviscid, incompressible

fluid and non-rotational flow, the complex
potential can be defined as analytic function
w=ϕ+iψ. To determine w(z), the following boundary conditions are to be satisfied:
• on the boundary including the wedge: ψ = const or ∂ϕ / ∂n = 0 , (4.3.10)
132
• on the cavity boundary – to conditions (4.3.2).

Furthermore, function ω ( w) = ln(dw dz ) is the solution of a mixed value problem and, in
general, has a singularity of the ( w − wc ) −1 2 type at point wc . Equation (4.3.2) is
dimensionless so that the speed at detachment point A or B is unity. The complex velocity
should satisfy the supplementary condition at infinity wz ( z ) z →∞ = −iv∞ . (4.3.11)
Also, the stream at infinity should be undisturbed, i.e. condition (2.1.6d) should be satisfied.
To uniquely determine the solution, the following dimensionless parameters are given:
• cavitation number σ = 2( p∞ − p0 ) / ρ v∞2 = (vB2 − v∞2 ) / v∞2 , (4.3.12)
where p∞ is the pressure at infinity along the x - axis, y = 0 , p0 is the cavity pressure, v∞ -
is the stream velocity at infinity, ρ is the fluid density. If the velocity at point A or B is
assumed to be unity, vB = 1 , then the cavitation number will be expressed as σ = v∞−2 − 1 .
• Froude number Fr = v∞2 / gd , (4.3.13)
where g is the gravity acceleration, d = 2 L sin πμ is the wedge base, L is the length of wedge
side.
The value problem formulated above can be solved parametrically by conformal mapping
the physical z-plane onto the half unit circle of the auxiliary ζ -plane as shown in Fig.4.3.2a,
b respectively. Transformation from the w-plane to the ζ -plane can be written in differential
ζ (1 − ζ 4 )
form as wζ = . (4.3.14)
(ζ 2 + a 2 ) 2 (ζ 2 + a −2 )
2
Eliminating all singularities, function ω (ζ ) can be expanded to a power series as:

1+ μ b(ζ 2 − 1) ∞
ω (ζ ) = −π
i + μ ln ζ − + ∑ Cn (ζ 2 n − 1) . (4.3.15)
2 2(ζ 2 + 1) n =1
As before, the mapping function z (ζ ) is determined from the differential equation
dz / du = wζ e −ω (ζ ) . Equations (4.3.14) and (4.3.15) are the general solution of the above-
mentioned problem. They include unknown parameters a and b and undetermined
coefficient Cn . Condition (4.3.11) at point D (ai ) together with the “closed condition”
produce two real equations as
b(1 + a 2 ) N 1
+ ∑ Cn ⎡⎣(−1) n a 2 n − 1⎤⎦ + μ ln a + ln(1 + σ ) = 0 , (4.3.16)
2(1 − a ) n =1
2
2
μba 2 N
+ + ∑ (−1) n nCn a 2 n = 0 , (4.3.17)
2 (1 − a )2 2
n =1
which can determine a and b for given coefficients Cn. Namely at Cn =0 function (4.3.14)
and equation (4.3.15), together with conditions (4.3.17) and (4.3.19), uniquely determine the
problem solution for a flow past a wedge in weightlessness.
4.3.5. Determining the Coefficients
The real and imaginary parts of equation (4.3.15) on the arc ζ = eiτ are defined as:
133
∞
ln v = ∑ Cn (cos 2nτ − 1) , (4.3.18)
n =1
1+ μ b ∞
− μτ + tan τ − ∑ Cn sin 2nτ .
β (t ) = π (4.3.19)
2 2 n =1
Equation (4.3.2), taking into consideration functions (4.3.14), (4.3.18) and (4.3.19), can be
∞
−3∑ Cn (cos 2 kτ −1)
∞
4a 4 sin 2τ
written as ∑n =1
Cn 2n sin 2nτ = g
(1 + 2a 2 cos(2τ ) + a 4 ) 2
e n =1
sin β (τ ) . (4.3.20)
Coefficients Cn. can be found using either the collocation approach or an iterative
process. Multiplying equation (4.3.20) by sin2nτ and integrating over interval [0, π], one
∞
π −3 ∑ Ck (cos 2 kτ −1)
g 4a 4 sin 2τ
π n ∫0 (1 + 2a 2 cos(2τ ) + a 4 ) 2
can obtain: Cn = e k =1
sin β (τ )sin 2nτ dτ . (4.3.21)
Hence, coefficients Cn. and parameters a and b can be determined iteratively by given g. It
should be noted that the value of g differs from gravity acceleration because the functions
are dimensionless. Its exact connection with Froude number is given by
1
Fr = . (4.3.22)
2 g (1 + σ ) L sin(πμ )
Integrating pressure p = −0.5(v 2 − 1) − gy along the wedge sides and dividing by 0.5v∞2 d ,
can yield drag coefficient as CD = CDi + CDh , (4.3.23)
i h
where C and C are the inertial and the hydrostatic forces:
D D
1 1
C = (1 − F / L)(1 + σ ), F = ∫ | wζ e |d ζ , L = ∫ | wζ e −ω |d ζ ,
i
D
ω
(4.3.23)
0 0
sin(2πμ )
CDh = . (4.3.24)
4 Fr sin 2 πμ
4.3.6. Numerical Examples
Cavity shapes for a wedge with 2πμ =30° and for a flat plate are shown in Fig.4.3.3 and
4.3.4. Negative Froude numbers correspond to coincidently directed speed and gravity.
Calculations show that the cavities for certain σ and Fr have a closed boundary with a
cuspidal point at the cavity end, as was determined first in [Acosta, 1961] and then in
[Lenou, 1963]. The lift for that case coincides with buoyancy force (Archimedes force). For
negative Froude number, the lift can be zero at certain σ and Fr (dotted line in Fig.4.3.3).
Fig.4.3.4 shows large cavity shapes for a flat plate with small cavitation numbers. The plate
length is small relative to the cavity thus the flow should be similar to the flow obtained by
the linearized or first order theory [Tulin, 1953]. Instead of the blunt forebody a similar
singularity could be placed at the cavity leading edge. Calculated results in Fig.4.3.4 agree
perfectly with conclusions in [Tulin, 1964]: the cavity becomes squashed elliptic in shape
with its bulbous forebody in the case of a field pointing in the same direction as the flow.
When the field points in a direction opposed to the flow, the cavity is lengthened and its after
part becomes less blunt than its forward. Calculations show that the gravity influences much
on the drag only for negative Froude number for very short cavities only. But it should be
appreciated that flow in that case could not be realized in the nature; it is obtained
134
mathematically only and the drag
theoretically can be zero or even
negative by small negative Froude
numbers.
Fig.4.3.3.
Cavity shapes in longitudinal field.
Fig.4.3.4.
Large cavity past a perpendicular plate.
135
Sec. 4.4. Supercavitating Flat Plate in a Transverse Gravity Field

Unlike for a longitudinal field, cavitating flow in a transverse gravity is rather complicated.
The first results of the flow had been obtained under certain assumptions [Parkin, 1957;
Street, 1963; Ivanov, 1961; Tulin, 1964]. In the latter paper the shape of cavities was
calculated using the linear theory. Many authors considered the flow in a non-linear problem
statement but mostly assuming a unique existence of the flow and a free boundary only
[Gurevich, 1979]. Cavitating flow of a flat plate in the transverse gravity field was briefly
outlined in Sec. 4.3. To discuss all problems in a consistent format, we shall consider below
the cavitating flow in a transverse gravity field using the cavitation model with single spirals.
4.4.1. Supercavitating Flow Past a Plate
Fig.4.4.1a shows the flow in the z-plane. All conditions are the same as in Sec. 4.3 except
the cavity boundary where conditions (4.3.1) or (4.3.2) should be satisfied. As before, the
flow domain on the z-plane is conformal mapped onto the inner of a unit circle (Fig.4.4.1b).
Using singularity of functions, Sec. 3.1, one can determine the desirable functions for a
weightless fluid, e.g. derivative of complex potential
ζ (ζ 2 − 1)(ζ − eia )(ζ − e− ia )
wζ = − ; (4.4.1)
(ζ − d eic ) 2 (ζ − d e − ic ) 2 (ζ − eic / d ) 2 (ζ − e − ic / d ) 2
and logarithmical function of complex velocity ω = i (α − π ) + ln ζ + ω0 (ζ ) , (4.4.2)
where function ω0 (ζ ) should have a single pole at point C (eia ) , vanishes at points ( u = ±1 ),
and has real values on diameter AB, so that it can be
b(ζ 2 − 1)
written as ω0 = . (4.4.3)
(ζ − eia )(ζ − e − ia )
Equations (4.4.1) and (4.4.2) constitute the general
solution of the problem of cavitating flow around a
flat plate in weightlessness. The gravity field, as
before, does not change the derivative of complex
potential and also singularities of complex velocity
but the velocity must change due to boundary
condition (4.3.1). As before, the unknown function
additional to (4.4.2) could be determined by power
series expansion
N
ω3 = ∑ Cn (ζ n − 1) . (4.4.4)
Fig.4.4.1. Conformity of domains n =1
on the z- and ζ - planes Coefficients Cn are chosen from the condition for
speed at the plate leading edge vB = 1 , which is equivalent to condition ω3 (1) = 0 . But
another condition at the trailing edge A should be satisfied, i.e., vA = 1 + 2 gL sin α .
Moreover, the speed at spiral centers C1 and C2 is equal to vC1 = 1 − 2 g yC1 and
vC 2 = 1 − 2 g yC 2 , respectively. Thus, function ω (ζ ) for cavitating flow has a discontinuity
of the first kind at point C that may worsen the convergence of the series. Therefore, it is
better to use two other functions that satisfy these conditions. Let functions y1(t) and y2(t) be
136
ordinates of the cavity upper and lower boundaries that depend on angle t determining the
point on arc BCA, Fig.4.4.1b. Then the ordinates of the spiral centers at the cavity end are
calculated as yC1 = y1 ( a − 0) and yC 2 = y2 (a + 0) . Function ω 2 (ζ ) , which has a piecewise
constant real part on arc BCA and is real on diameter AOB, can be written as
π −a i ⎛ ζ − eia ⎞
ω1 = ln δ − ln δ ln ⎜ − ia ⎟
, (4.4.5)
π π ⎝ζ −e ⎠
1 − 2 g yC1
where δ= .
1 − 2 g yC 2
Another function should be real on the x-axis and its real part vanishes at point B(1) and is
⎛ 1 + k + (1 − k )ζ ⎞
equal to ln v A at point A(−1). It can be chosen as ω2 = ln ⎜ ⎟, (4.4.6)
⎝ 2 ⎠
where k = 1 + 2 gL sin α e −ω1 ( −1) −ω3 ( −1) .
Now, we have the expression for logarithmical function of velocity as follows:
ω = i (α − π ) + ln ζ + ω0 (ζ ) + ω1 (ζ ) + ω2 (ζ ) + ω3 (ζ ) . (4.4.7)
The general solution of the flow problem given by functions (4.4.1) and (4.4.7) in addition to
parameters a, b, c, d has unknown coefficients Cn , n = (1, N ) . If these coefficients are
given then the parameters are determined, as before, from the conditions at infinity for the
given velocity and undisturbed flow:
1
ω (deic ) = ln and ∫ wζ (ζ )e −ω (ζ ) d ζ = 0 . (4.4.8)
1+σ ζD
The integral in (4.4.8) can be calculated analytically to satisfy an equality like

f ( a , b, c, d ) = 0 , (4.4.9)
d
where f ( a, b, c, d ) = limic ln ⎡⎣ wζ (ζ )(ζ − deic ) 2 e −ω (ζ ) ⎤⎦ =
ζ → de d ζ
2deic 1 1 2 2
= 2 i 2c + ic ia + ic − ia
− ic − ic
− ic ic − (4.4.10)
d e − 1 de − e de − e de − de de − e / d
2 d
− ic − [ω0 (ζ ) + ω1 (ζ ) + ω2 (ζ ) + ω3 (ζ )] ζ = deic
de − e / d d ζ
− ic
Satisfying condition (4.3.2) at the cavity boundary, one can calculate coefficients Cn .
Derivative of the real part of function (4.4.7) on arc BCA is given by
d ln wz N
= −∑ Cn n sin nt + R (t ), (4.4.11)
dt n =1
d (k 2 − 1)sin t
where R (t ) = Re ω2 (eit ) = .
dt 4[cos 2 (t / 2) + k 2 sin 2 (t / 2)]
From condition (4.3.2) taking into account the last trigonometric sequence, one can obtain,
similar to (4.3.20), relationships as
π
2 ⎡
∫ R (t ) − gwζ (eit )iei t sin(Im ω (eit ))e −3Re ω ( e ) ⎤ sin ntdt . (4.4.12)
it
Cn =
πn 0 ⎣ ⎦
137
Equation (4.4.12) allows determining the coefficients iteratively. Plate length L and cavity
shapes ( y1 (t ), x1 (t ) ) and ( y2 (t ), x2 (t ) ) are calculated by integrating differential equation
zζ (ζ ) = wζ (ζ )e −ω (ζ ) . As long as length L is calculated after determining unknown
parameters and coefficients, Froude number will also be calculated later. In this case,
coefficient g is not more equal to the gravity acceleration, but inversely proportional to
Froude number. Now, iteration by given g and σ could be made in as follows:
• the initial values of the coefficients are assumed to be zero (Cn =0), and parameters
δ=1, k=1. This assumption corresponds to flow without gravity;
• parameters a, b, c, d are calculated from equations (4.4.8) and (4.4.9), then
coefficients Cn are obtained from (4.4.12);
• length L, ordinates of the spiral centers (yC1, yC2), and values of δ in (4.4.5) and k in
(4.4.6) are calculated.
The last two actions are repeated to satisfy desirable validity for all unknown parameters and
coefficients. The validity may be verified with Froude number calculated in two ways. The
v2
first way is the common way of obtaining Froude number as Fr = ∞ . (4.4.13)
gL
Another way is in calculating from the speeds at plate edges. As v A2 − 2 gL sin α = 1 , or
2sin α
v A2 − 2sin α /(1 + σ ) Fr = 1 . Thus Fr = . (4.4.14)
(1 + σ )(v A2 − 1)
Iteration was repeated till fulfilling the desirable validity as 1 − Fr / Fr ≤ 10−4 .
As the above-obtained equations are non linear their solving is sensitive to input
values. The input data can usually be as a (0) = π − α , b(0) = 0 , c (0) = 0.5π , d (0) = 0.5 , but
for small cavitation numbers the initial value of d (0) should be close to 1. The smallness of
cavitation number yields another hurdle associated with behavior of function wζ (ζ ) : if
parameter d is close to 1, function wζ (ζ ) on the circle may be of a high value and the
integrand of (4.4.12) may oscillate causing slow convergence of the coefficients. Therefore,
calculations are usually made for σ ≥ 0.3 only. Small cavity numbers in a transverse gravity
field can be considered by the first order theory [Tulin, 1964]. The hydrodynamic force can
be calculated by integrating the pressure due to Bernoulli equation, see Figs.4.4.2 and 4.4.3.
It is seen that Froude number greatly affects cavity length and moderately affects
hydrodynamic forces; the cavity shortens without transverse deformation. These results
differ from Tulin’s [Tulin, 1964]. This is probably the difference in assumptions in the
problem statement; Tulin considered small cavity numbers assuming that both boundaries of
the cavity are fixed at the same point. The table in Fig.4.4.2 shows that the drag coefficient
increases slowly with Froude number decreasing. Dependence of the abscissa of spiral
centers (the length of upper and low boundaries of cavity) is shown on Fig.4.4.3.
138
Fig.4.4.2. Cavity shapes in a transverse

gravity field at σ =0.5.
Fig.4.4.3.
Relationship between boundary
length and Froude number.

Let the direction of gravity relative to plate be arbitrary. Angle between the negative gravity
direction and the y-axis is marked γ , see Fig.4.4.4a for symbols. Mapping conformal the
flow domain on the z-plane onto inner of semicircle on the ζ-plane (Fig.4.4.4b), the
derivative of complex potential according to its singularities and nulls can be written as
ζ (ζ 2 − 1)(ζ − a )(ζ − a −1 )
wζ = . (4.4.15)
(ζ − deic ) 2 (ζ − de − ic ) 2 (ζ − d −1eic ) 2 (ζ − d −1e− ic ) 2
The logarithmic function of complex velocity for
weightlessness fluid is
⎛ ζ −a ⎞ ζ −1
ω0 = ln ⎜ −1 ⎟
+b − ln a . (4.4.16)
⎝ζ −a ⎠ ζ +1
Fig.4.4.4.
P partially cavitating flow:
(a) flow on the z-plane;
(b) parametric circle on the ζ-plane.
Free boundary in this case is a continuous curve so that

the speed in gravity field is also continuous on the
curvilinear abscissa. Therefore, the logarithmic function of velocity in gravity field can be
expressed as a sum ω = ω0 + ω1 , (4.4.17)
where the additional function could be determined in the form of a power series as
N
ω1 = ∑ Cn (ζ n − 1) . (4.4.18)
n =1
139
If the coefficients are given then the unknown parameters can be calculated satisfying
conditions at infinity as written in (4.4.8). Function (4.4.10) in this case is expressed as
1 2deic 2 2
f ( a, b, c, d ) = ic + 2 i 2 c + ic −1
− ic −
de d e − 1 de − a de − de − ic
N
. (4.4.19)
2 2 2b
− ic ic − − − ∑ Cn nd e
n −1 ic ( n −1)
de − e / d deic − e − ic / d (deic + 1) 2 n =1
The transformation from the ζ-plane to the z-plan can be obtained as before from differential
equation zζ (ζ ) = wζ (ζ )e−ω (ζ ) . Namely, the plate length is equal to
1
L = − ∫ wζ (ξ )e −ω (ξ ) d ξ . (4.4.20)
0
Now, all dimensional coordinates should be made dimensionless dividing by length
(4.4.20). For instance, the cavity shape is calculated by parametrical functions as
t
1
x(t ) + iy (t ) = ∫ wζ (eit )e −ω ( e ) eit idt .
it
(4.4.21)
L0
Obviously, both length L and the cavity length Lc = x(π ) depend on parameters a, b, c and
d, or on angle of attack α and cavitation number σ . But in case of partial cavitation, the
problem has ambiguous solution for a fixed cavitation number. Moreover, the cavitation
number cannot be less than a certain magnitude. Therefore, it is useful to calculate all
unknown parameters including cavitation number by given cavity length Lc ∈ (0,1) .
Numerical calculations show that the set of non-linear equations can be solved numerically
using arbitrary initial values of unknown parameters. Hydrodynamic force and torque can be
obtained integrating the pressure with respect to ξ -axis in interval (0, 1). Hydraulic pressure
should also be calculated to obtain additional normal force coefficient Ch = L2c sin γ / Fr .
Coefficients Cn , as before, are calculated by satisfying condition (4.3.2) in the x ', y ' axis
(Fig.4.4.4a) iteratively using expressions as follows:
π
2
∫
it
Cn = e −3Re ω ( e ) wζ (eit )ieit sin[γ − Im ω (eit )]sin ntdt . (4.4.22)
π (1 + σ ) Fr L 0
Some numerical results are plotted in Figs.4.4.5 and 4.4.6.
Fig.4.4.5. Shapes of partial cavity on the plate.
Fig.4.4.5 shows cavity shapes for angle of attack α = 5 , angle γ = 10 , cavitation number
σ = 1.25 , and three values of Froude number ( Fr = ∞, − 1,1 ). Fig.4.4.6 shows dependence of
lift and cavity length on Froude number for the same parameters. The lift and length are
140
attributed to the values for a weightless fluid ( Fr = ∞ ). These figures also show that gravity
enlarges the cavity and the lift if it acts opposite to the ordinate axis, and decreases at
negative Froude numbers. The gravity effect is greater at
positive Froude numbers than at negative ones. Earlier,
this problem was studied using a cavitation model with a
trailing plate [Kotlyar & Terentiev, 1970]; both models
gave the same results.
Fig.4.4.6.
Llift and cavity length versus Froude number.
4.4.3. Simulation of Cavitating Flow Using Other Models

The method of expansion in series can be also used for numerical calculathons of
supercavitating flow in a transverse gravity field using Riabouchinsky as well as Kuznetsov
models, Figs.2.5.1&2.5.3. In this case, the rectangle in Fig.2.5.1b, should be conformal
mapped onto a ring and then function ω may be expanded in Loran series summing from
−∞ to ∞ . All other conditions remain the same. The cavity shapes calculated in [Terentiev
et al., 2005] using Riabouchinsky model, as well as calculated in [Kotlyar & Troyepolskaya,
1975] using Kuznetsov model are plotted in Fig.4.4.7.
Fig.4.4.7. Cavities in gravity field simulated by Riabouchinsky (a&b) and

Kuznetsov (c&d) models for perpendicular (a&c) and inclined (b&d) plates.
It is seen that the gravity miniaturizes cavities while the centerline is bowed down as
seen in Fig.4.4.2. Similar results are obtained by M.Tulin in the bounds of the linear theory
(see Sec. 5.3). This contradiction could be explaned by differences of flow near the cavity
end. Thus the different models yield in this case different results. The transverse gravity field
has also been considered in [Larock & Street, 1967a] using models with double spirals.
Sec. 4.5. Axisymmetric Cavitating Flow Past a Sphere

The above-considered problems demonstrate great efficiency of power series expansion in
planar flow problems that can be solved analytically. In this section we will show that the
141
analitic functions and power series expansion can be effectively applied for axisymmetric
cavitating flow as well. Axisymmetric cavity problems in a non-linear formulation have been
solved either analytically using differential equations and various approaches [Brennen,
1969; Shepelenko, 1968], or by transforming differential equations into integral equations
for some auxiliary functions and solving them numerically. There are different methods of
transforming the boundary problems into integral equations and its numerical solving, e.g.
boundary source layers [Krylov, 1963; Terentiev, 1994] or vortexes [Guzevsky, 1975, 1979;
Ivanov, 1980; Kojouro, 1980; Todorashko, 1977], or the boundary elements method
[Krasnov & Kuznetsov; 1989, Subkhanculov & Khomyakov, 1990; Yas’ko, 1993]. They
used mostly cavitation model with artificial disc or cone, Guzevsky [1975] used also the
cavitation model with a semi-infinite cylinder. The methods have been tested mainly on flow
problems of disks or cones. In solving the cavity flow of a sphere, the authors assume a
smooth separation of jets [Brennen, 1969; Ivanov, 1980; Kojouro, 1980]. Results of these
calculations for the cavity flow of spherical segments smaller than a semi-sphere are given in
[Guzevsky, 1979].
In this section, generalized analitic functions of complex variables introduced by
Pologiy [1965] are applied to investigate a cavitating flow of sphere. As mentioned in Ch.
14, the integral transformation has a decisive role here and it connects the generalized
analitic function of characteristic p=y with the analitic function of complex variable. This
allows solving the boundary problem by this function using the methods that have been well
developed in the theory of analitic functions. Earlier, a similar approach in solving various
axisymmetric problems have been used in [Goman & Popov, 1980; Zhitnikov, 1993;
Terentiev, 1993; Terentiev & Dimitrieva, 1998]. Results of the latter are included here.
Consider a cavitating flow of a spherical segment with given detachment ring line at the
segment edge. Location of this ring is determined by angle γ, see Fig.4.5.1a. The cavity ends
with single spiral rings. Wile the single spiral in the
case of planar flows has a strong mathematical basis
in case of the axisymmetric flow the single spiral ring
at the cavity end should be considered as assumption.
Studying a flow in the close proximity to cavity end is
the problem to be considered in the future.
Fig.4.51. Cavitating flow past a spherical arc;

(a) flow on the z-plane; (b) parametric circle.
The velocity potential and stream function of

axisymmetric flow can be expressed as
Φ ( x, y ) = ϕ ( x, y ) + v∞ x,
, (4.5.1)
Ψ ( x, y ) = ψ ( x, y ) + v∞ y 2 / 2
where ϕ(x,y) and ψ(x,y) are the velocity potential and stream function of the disturbed flow,
which satisfy the following conditions:
∂ϕ 1 ∂ψ ∂ϕ 1 ∂ψ
• the differential equations in a flow domain: = , =− ; (4.5.2)
∂x y ∂y ∂y y ∂x
142
• the kinematic condition at a boundary ∂ϕ / ∂n = v∞ cos γ ; (4.5.3)

• the dynamic condition on the cavity shape ∂ϕ / ∂s + v∞ sin γ = 1 (4.5.4)
Functions ϕ ( x, y ) and ψ ( x, y ) vanish at infinity.
Let's consider generalized analytic function (see App. 4) F ( z ) = ϕ ( x, y ) + iψ ( x, y ) of
characteristic factor 1/y. Applying Pologiy integral relationship (A4.3), one can express it by
an unknown but analitic function f ( z ) = u ( x, y ) + iv ( x. y ) , as
1 f (τ )[−1 + i (τ − x)]
F ( z) = ∫ dτ , (4.5.5)
2Γ (τ − z )(τ − z )
where line Γ joints self-conjugate variables z = x − iy and z = x + iy . Functions v( x, y ) and
ψ(x,y) on the real x-axis are assumed to be equal to zero, i.e.
v( x,0) = 0, V ( x,0) = 0 . (4.5.6)
If the analitic function for z → ∞ satisfies condition f ( z ) ≈ z − (1+ε ) where ε is an arbitrary
positive value, then the integrating boundary in (4.5.5) can be deformed to a quarter circle of
an infinite large radius and a semi infinite negative x-axis and an elementary part of surface
of revolution, so that the real and imaginary parts of (4.5.5) may be written in two real
integral transformations as
f (τ ) f (τ )(τ − x)
z z
ϕ ( x, y ) = − Im ∫ dτ , ψ ( x, y ) = Im ∫ dτ . (4.5.7)
0 (τ − z )(τ − z ) 0 (τ − z )(τ − z )
For the derivatives of functions ϕ(x,y) and ψ(x,y) the following equations are derived:
∂ϕ df (τ ) dτ ∂ψ df (τ ) (τ − z )dτ
z z
1
= − Im ∫ , = − Im ∫ . (4.5.8)
∂x 0
dτ (τ − z )(τ − z ) ∂y y 0
dτ (τ − z )(τ − z )
Satisfying the above-mentioned kinematic condition at boundary ABC and dynamic
condition at free surface BC can yield combined equations in integral form. Also, equation
(A4.10) is to be satisfied, which is expressed at the critical point as
π
u ′x (0,0) + v∞ = 0 . (4.5.9)
2
4.5.2. Solving the Equations Numerically
Since the problem consists of determining an analitic function f ( z ) in the flow domain,
which satisfies the above-mentioned boundary conditions, the power series expansion could
be applied mapping the flow domain in the z-plane onto the outer unit circle of the auxiliary
t-plane (Fig.4.5.1b). Since potential ϕ vanishes at infinity as r −2 , analytic function f ( z )
also vanishes, due to equation (A4.10), as r −2 . Thus the unknown function can represent the
∞
c
outer of unit circle t > 1 as a Laurent series f (t ) = ∑ k k+1 . (4.5.10)
k =1 t
It is assumed that transformation z(t) has the same singularities at cavity end as in the planar
flow. Then the relationships (4.2.2) – (4.2.7) for the unknown transformation are valid in this
case as well. Unknown coefficients ck in (4.5.10) and coefficients an in (4.2.7) as well as
factor a in (4.2.8) should be calculated satisfying the kinematic and dynamic conditions on
the unit circle and equation (4.5.9). Conditions on the boundary are expressed as follows:
143
• kinematic condition on 0 < α < π

⎛ π df ( β ) dβ 1 ⎞ dy
⎜ Im ∫ + ⎟ −
⎜ α dβ ( z ( β ) − z (α ))( z ( β ) − z (α )) 1+σ ⎟ dα
⎝ ⎠
(4.5.11)
⎛ 1 π
df ( β ) z ( β ) − x(α ) ⎞ dx
−⎜ Im ∫ dβ ⎟ =0
⎜ y (α ) α d β ( z ( β ) − z (α ))( z ( β ) − z (α )) ⎟ dα
⎝ ⎠
• dynamic condition on the arc 0 < α ≤ α0
2
⎛ π df ( β ) dβ 1 ⎞
⎜ Im ∫ + ⎟ +
⎜ α dβ ( z ( β ) − z (α ))( z ( β ) − z (α )) 1+σ ⎟
⎝ ⎠
2
, (4.5.12)
⎛ 1 π
df ( β ) z ( β ) − x(α ) ⎞
+⎜ Im ∫ dβ ⎟ =1
⎜ y (α ) α d β ( z ( β ) − z (α ))( z ( β ) − z (α )) ⎟
⎝ ⎠
iα iβ
where functions e and e are denoted for simplicity as α and β .
Furthermore, equation (4.5.9) and conditions (4.2.10) – (4.2.12) are to be satisfied. It
is problematic to calculate iteratively but in this case the collocation method can be used.
Retaining a finite number of coefficients in the series (4.5.10) and (4.2.7) and then satisfying
conditions (4.5.11) and (4.5.12) at points M1 on the cylinder's wetted boundary and M2 on
the cavity, one can obtain a closed set of non-linear equations.
4.5.3. The Drag Coefficient
Drag of the body can be calculated by integrating the pressure along the wetted surface:
yB
D = 2π ∫ ( p − pc ) ydy ,
0
where yB is the cavity radius in its separation plane. The drag coefficient is
D ⎛y 2 2 π dy ⎞
C D= = (σ + 1) ⎜ B2 − 2 ∫ V 2 ( β ) y ( β ) dβ ⎟ , (4.5.13)
ρ 2 2 ⎜ b b αo dβ ⎟
π b V∞ ⎝ ⎠
2
2 2
⎛ ∂Φ ⎞ ⎛ ∂Φ ⎞
where V 2 (β ) = ⎜ (β ) ⎟ + ⎜ (β ) ⎟ .
⎝ ∂x ⎠ ⎝ ∂y ⎠
4.5.4. Calculated Results
Results of computing [Terentiev & Dimitrieva, 1998] are given in Figs.4.5.2 – 4.5.4. Drag
coefficient of spherical segments at cavitation numbers σ =0.2; 0.5; 1.0 is plotted in
Fig.4.5.2. Function CD(y) has two maximums as in the plane case. Also plotted for
comparison are the results of calculations in [Brennen, 1969; Ivanov, 1980; Kojouro, 1980]
marked by symbols , o and Δ respectively. The calculations, using Riabouchinsky model,
show a smooth separation of cavity from the sphere, with a good agreement. The maximums
of drag coefficients correspond to smooth separation at the detachment point. After the
second maximum, the whole cavity boundary is located inside the sphere and so the cavity
length is zero. Cavity length L/R and the cavity's greatest diameter, H/R, as a function of
144
detachment angle are plotted in Fig.4.5.3. The dashed line marks the case when the cavity
diameter is less than the detachment ring's diameter, thus the line is a sinusoid (sin y).The
results of
Brennen
[1969] and
Kojouro
[1980] are
marked as
in Fig.4.5.2.
It is seen
that the
results for
Fig.4.5.3. Relative length L/R and maximal
thickness H/R of cavity at cavity
Fig.4.5.2. Drag coefficient of cavitation number, σ =0.5: radius
spherical segment: Brennen, Brennen 1969; Δ Kojouro 1980 coincide
1969;
with ours
though the cavity length is different.
Cavity shapes at fixed cavitation number σ = 0.5 and
different detachment angles are plotted in Fig.4.5.4.
Angle γ = 70° corresponds to smooth separation of free
boundary from the sphere. The cavity boundary has
inflection for detachment points between two maximums.
Other axisymmetric bodies can be calculated similarly, as Fig.4.5.4. Cavity shapes for different
for example did Dimitrieva [2002] for an ellipsoid at detachment points.
different aspect ratios.
Sec.4.6. Surface Tension in Free Boundaries

Surface tension takes place only on free boundaries but its account can considerably change
the free boundary configuration and influence hydrodynamic characteristics of the
deformable surface. Capillary forces are most interesting from theoretical point of view.
They can affect the detachment point position on the body wetted boundary in cavitating
flow. These problems were considered in [Amromin & Ivanov, 1981]. Studying the capillary
tension in supercavitating flow is coupled with certain difficulties for the following reasons:
1) the real cavity boundary is not a continuous curve but an interrupted and bubbly surface;
2) due to tension, the cavity surface must be fastened at the boundary edge, but in reality the
cavity is free at its end; 3) capillary tension conflicts with the minimal pressure in the cavity:
its convex boundary is directed to the pressure gradient, but the convex of boundary
capillary surface is directed opposite to the gradient ∇p. This contradiction introduces a lot
of complexity for configuration of cavity: borders can self-intersect or sharply change at
separation sites and so on. Most interesting practical applications are currents with free
surfaces when the pressure gradient is directed towards the free surface. Such conditions
arise in problems about flow of bubbles or a flexible buckled tube, and also jet spray. Some
of such problems will be considered below in this section.
Joukowski [1891] was the first who derived an exact solution for a bubble moving in
145
the channel with capillary forces taken into account. Unfortunately, the solution is valid only
for some assigned ratios of a bubble size and the channel, thus it is impossible to receive a
flow of a bubble in a boundless domain by limiting transition from the analytical solution.
McLeod [1955] derived an exact solution for a special case of planar flow of a flexible tube.
An asymptotic solution for its small deformations was obtained in [Slezkin, 1951], an iterative
calculation method based on expansion of the unknown functions in a power series in terms of a
small parameter was proposed in [Kiselev, 1969]. Detailed studies of a flow past flexible tubes and
bubbles, and a cavitating flow with capillary tension on cavity shapes were made in [Zhitnikov &
Terentiev, 1982, 1984, 1988]. Some results of the latter are included in this section.
The kinematic condition on the boundary with surface tension is the same as the condition
without tension but the dynamic condition results from Laplace’s law
p0 − p = T (1/ R1 + 1/ R2 ) , (4.6.1)
where p and p0 are pressure in fluid at one side and the other side of the free surface or inside
a tank with flexible surface T is tension, respectively; R1 and R2 are the principle radiuses of
curvature. Plahe and axisymmetric flows only are considered. Dimensionless parameter is
2( p0 − p∞ )
μ= , (4.6.2)
ρ v∞2
which differs from cavitation number by its sign only, then condition (4.6.1) due to the
Bernoulli integral can be expressed as follows:
d β cos β ρ v∞2 ⎛ v2 ⎞
• for axisymmetric flow along the x-axis: ∓ + = ⎜ μ −1+ 2 ⎟ ; (4.6.3)
ds y 2T ⎝ v∞ ⎠
d β ρ v∞2 ⎛ ⎞ v2
• for planar flow: ∓ = ⎜ μ − 1 +
⎟, (4.6.4)
ds 2T ⎝ ⎠ v∞2
where β is the angle between the x-axis and the fluid velocity vector, s is curvilinear
abscissa of boundary; signs “ – “ and “ + “ correspond to the positive and negative direction
along the boundary, respectively.
4.6.2. Joukowski’s Analytical Solution for Planar Flow
For a 2D planar flow, exact analytical solutions can be derived as described in Sec. 4.3.
Substituting velocity v = dϕ / ds in (4.6.4), the last condition is expressed as
ρ ⎛ ( μ − 1)v∞2 ⎞
dβ = − ⎜v + ⎟ dϕ . (4.6.5)
2T ⎝ v ⎠
Let's consider a logarithmic function of complex velocity as ω = ln( wz / v∞ ) and present it as
a sum of two functions ω1 ( w) and ω2 ( w) which satisfy following conditions: Im ω1 = 0 ,
Re ω2 = 0 in interval (–1, 1) corresponding to bubble boundary as in Sec. 4.3. If function
ω1 ( w) is given then the function should be found from integral
ρ
ω2 = i ∫ (eω0 ( w) + ( μ − 1)v∞2 e −ω0 ( w) )dw . (4.6.6)
2T
146
Sum ω = ω1 + ω2 will be a solution of non-linear problem (4.6.5). Instead of the w-plane,

one can use some auxiliary domain of parametric ζ - plane. Joukowski [1981] used the
upper half-plane of the auxiliary ζ -plane and the transformation from the w-plane to the ζ -
⎛ 1 1 ⎞
plane written in differential form wζ = N ⎜ − ⎟. (4.6.7)
⎝ζ + a ζ −a ⎠
The first function has been chosen as ω0 = ln 1 − ζ 2 . (4.6.8)
ρ aN ( μ − 1)v + 1 − ζ
2 2
dζ
Then the other function can be found from ω1 = i∫ ∞
.
T 1−ζ 2 a −ζ 2
2
Satisfying additional conditions, ( μ − 1)v∞2 + 1 = a 2 and ρ aN / T = 1 , (4.6.9)
(
one can obtain the last integral in a simple form, ω1 = − ln ζ + i 1 − ζ 2 . Since function)
(4.6.7) transforms the stripe of width π N on the w-plane to the upper half ζ -plane, and
imaginary parts of the sum of functions (4.6.8) and (4.6.9) outside interval (−1, 1) are equal
to −π / 2 , then these functions determine a vertical flow around a flexible-shape cylindrical
1−ζ 2
bubble in a channel. The complex velocity is wz = . (4.6.10)
ζ + i 1− ζ 2
a2 − 1
Thus the speed at infinity in the channel is equal to v∞ = −iwz (a ) = . Substituting
a + a2 − 1
μ
the latter into (4.6.9) yields an exact value of parameter, a = , (4.6.11)
2 μ −1
And so the speed at infinity is transformed to v∞ = ( μ − 2) / 2( μ − 1) . The channel width is
h = 2π N / v∞ , whence N = hv∞ / 2π . The bubble shape is calculated from differential
equation dz / d ξ = e −ω (ξ ) wζ (ξ ) . Integrating it along interval (1, ξ ), one can obtain parametric
formulas of the shape:
μ 2 (1 − ξ 2 )( μ − 1) μ −2 μ − 2ξ μ − 1
x(ξ ) = h arctan , y (ξ ) = h ln . (4.6.12)
2π ( μ − 1) μ −2 4π ( μ − 1) μ + 2ξ μ − 1
Whence, the distances between two horizontal and two vertical tangents are calculated from
(4.6.12) as b = 2 y (−1) and c = 2 x(0) , respectively.
Weber number is the governing dimensionless parameter in capillary or flexible
shells. It can be written in this case as We = ρ v∞2 h / T = π ( μ − 2) / μ μ − 1 . Analyses of the
obtained function show that distance b has a maximum at μ0 = 3.382 ( bmax = 0.285 ); the
maximum Weber number is at μ1 = 5.236 ( We max = 0.943 ). The results of calculations are
plotted in Fig.4.6.1. The bubble shapes are shown in Fig.4.6.1a. The dotted line shows the
boundary of maximum distance bmax . Fig.4.6.1b shows a dependence of main parameters on
number μ . It is worth noting that the Weber number approaches zero (tension approaches
infinity) at two limits of μ: one at μ → 2 when the shell becomes a plate and another at
147
μ → ∞ when the shell becomes a circle, but in the last case the bubble dimensions approach
zero. This is caused by the applied solution method if the Weber number depends on number
μ only. Generally, contrary to this case, the flow is determined by two dimensionless
parameters μ and We . Some of these cases following [Terentiev et al., 2005] are discussed
below.
Fig.4.6.1. Bubble shapes in the channel (a),

dependence of main parameters on μ (b).
4.6.3. Flexible Cylindrical Shell in Unbounded Flow

Let a soft cylindrical shell of length 2L is fixed at two points spaced at b(b < L) from each
other, Fig.4.6.2а. Only the bottom half of flow is considered below. The flow is symmetric
relative to the vertical straight line passing through the middle point C of the shell. As stated
above, the Laplace’s condition (4.6.4) with sign “+” on the shell surface should be satisfied.
The flow problem can be considered in two statements: by given angle βA or by given ratio
l / L, where l is the distance between two stagnation points A and B. The forces acting on
point A and B are equal to | X A |=| X B |= 2T cos β A .
If angle β A = π / 2 , then forces X A and X B vanish like for a single body, i.e. the
shell becomes a free single bubble (see Fig.4.6.1a). In this case the flow is determined by
parameter μ only. If the shell is fixed at two points A and B, then angle β A is unknown and
differs from −π / 2 . In this case, β A is to be found, while ratio b/L should be given.
To solve the problem, the flow domain on the physical z-plane is mapped on the
interior of a unit circle of an auxiliary ζ -plane, Fig.4.6.2b. Conformity of points is clear
from Fig.4.6.2. The complex potential is determined by function w = ia (ζ − ζ −1 ) where a is
a real factor. On the arc of circle ( ζ = eiσ ) the potential is w = ϕ = −2a sin σ , thus the
derivative dϕ / dσ and the differential of curvilinear abscissa are expressed by the following
dϕ 2a
equations: = −2a cos σ , ds = − cos σ dσ . Due to the first equality, condition
dσ v
dβ ⎡v v ⎤ ρ v∞ a
(4.6.4) is transformed as follows: = −λ ⎢ + ( μ − 1) ∞ ⎥ cos σ , λ = . (4.6.13)
dσ ⎣ v∞ v⎦ T
Let’s denote γ = −2β A / π ; then the logarithmic function of complex velocity for a flow
148
along a wall with a cusp in the form of a circular segment can be obtained analytically as
⎛ (1 + iζ ) 2 −γ − (1 − iζ ) 2−γ ⎞
ω0 = 2i ln ⎜ 2 −γ / 2 ⎟
. (4.6.14)
⎝ 2i (2 − β )ζ (1 + ζ ) ⎠
Fig.4.6.2.
Flexible shell with two fixed
points (left), parametric
circle (right)
For two fixed points, the function's behavior at stagnation points of the shell should be taken
into account. As velocity vanishes at the vertex of an angle as | σ − π / 2 |γ and
| σ − π / 2 |≈ cos σ at σ → π / 2 (σ < π / 2) , then condition (4.6.13) yields asymptotic
1+ γ 1−γ
dβ ⎡ ⎛π ⎞ μ −1⎛ π ⎞ ⎤
behavior of the angle derivative as ≈ −λ ⎢C ⎜ − σ ⎟ + ⎜ − σ ⎟ ⎥ , σ →π /2,
dσ ⎣⎢ ⎝ 2 ⎠ C ⎝2 ⎠ ⎦⎥
where C is a constant. Integrating and substituting infinitesimal function cos σ instead
equivalent difference π / 2 − σ , we obtain the following asymptotic relationship:
⎡ C μ −1 ⎤
β ≈ βA + λ ⎢ cos 2+ β A / π σ + cos 2− β A / π σ ⎥ .
⎣ 2 + βA /π C (2 − β A / π ) ⎦
According to the last equality, one can chose the auxiliary function as
⎡ ⎛ ζ 2 + 1 ⎞ 2 −γ 2 +γ
⎛ ζ 2 +1⎞ ⎤
ω1 = iζ 2 ⎢ B1 ⎜ ⎟ + B2⎜ ⎟ ⎥. (4.6.15)
⎢⎣ ⎝ 2 ⎠ ⎝ 2 ⎠ ⎥⎦
Now, unknown function ω (ζ ) could be present as a sum of three functions, one of
them is determined from the flow around a symmetric circle segment (4.6.14), the second
from the behavior of d β / dσ (4.6.15), and the third function is an expansion in power
series, i.e. ω = ω0 + ω1 + ω2 , (4.6.16)
∞
where ω 2 = i ∑ C 2 mζ 2 m . (4.6.17)
m =1
Unknown coefficients B1 and B2 in (4.6.15) are determined by satisfying condition (4.6.13),
1 λ ( μ − 1) λχ
B1 = 1−γ − , B1 = − . (4.6.18)
2 γ (2 − γ )sin( β A ) γ (2 + β A / π )sin( β A )
Finally, constant χ can be calculated by taking σ → π / 2 which leads to
∞
⎛ v ⎞ 2−γ ∑ ( −1)m C2 m
γ = lim ⎜ cos − β A / π σ ⎟ = e m =1
. (4.6.19)
⎠ (1 − γ / 2)
σ →π / 2 v 2
⎝ ∞
The unknown coefficients are calculated by satisfying condition (4.6.4) in collocation
manner [Zhitnikov & Terent'ev, 1984, 2005]. Results calculations are plotted Figs.4.6.3 –
4.6.5. Fig.4.6.3 shows the flexible shell shapes for different numbers μ . The shape for μ = 1
coincides with the exact result obtained in [McLeod, 1955]. The free boundary, as in the
149
Joukowski and McLeod problems, deforms in a speed direction for decreasing μ , and both
stagnation points A and B join at μ0=0.546. The shape could be calculated for all values of μ,
but the boundaries self-intersect at μ<μ0. One of them corresponds to μ=0, see Fig.4.6.3. A
fluid particle in a streamline on the x-axis first reaches the stagnation point on the right,
makes a loop and passes into the second sheet of the z-plane, rounds the loop, moves to the
left stagnation point and then moves again along the x-axis. The limit μ→−∝ corresponds to
“unscrewed” circle. Of course, such a flow is impossible at μ<μ0, but can be considered
theoretically and computed. Fig.4.6.4 presents
the shapes with two fixed points.
Fig.4.6.3. Single flexible shapes. Fig.4.6.4. Flexible shapes with two fixed
points at different number of waves.
The distance between two fixed points
is equal to l=L/π where l and L are the diameter and circumference of circle, respectively
The shape may be considered as a flexible shell fixed at the edges of a slit in an infinite
horizontal plate; the flow is under a half of the plate. Then, creating a curtain pressure p0 on
the other side of plate, one can obtain the shapes plotted in Fig.4.6.4. If pressure p0
approaches infinity, i.e. μ→∞, then the shape becomes a circle. This corresponds to a flow
around a circular cylinder, but for μ→−∞, the case is equivalent to flow along a wall with a
hollow circle. At μ=−0.92 the angle at points A and B vanish. Computations show that
continuously changing μ from positive to
negative infinity is impossible. Approaching zero
from positive values differs from the negative
approach. To clarify this phenomenon, the shell
must be calculated as a function of μ:
2
C y (μ ) = ∫ ( p − p∞ )dx . Fig.4.6.5 shows two
ρ v∞2 l AB
ways of changing this function from μ=∞ and
from μ=−∞. The curves form a deformed double
spiral with its center at point μ=0. If angle βy Fig.4.6.5. Two ways of taking the
vanishes, then force coefficient Cy is equal to μ. limit μ → ±∞
Hence, the intersection points on straight line
150
Cy=μ in Fig.4.6.5 correspond to angle βA=0, and the sign of βA alternates when moving along
curves 1 or 2 to the spiral center. Fig.4.6.4 shows first three shapes of the flexible shell, all at
μ=0, (curves n1, n2, n3 and m1, m2, m3 correspond to the points on lines 1 and 2 denoted by
the same letters, respectively). Closer to zero more waves of the shell surface appear, which
can be considered as capillary waves on a free boundary. Curves n3 and m3 show that the
capillary waves predicted by the nonlinear theory have slight troughs and steep but rounded
crests. The flow around the whole flexible shell with two fixed points is possible
theoretically at μ>μ* only. Otherwise the shape is a self-intersecting line. The magnitude of
μ* depends on ratio l/L and can be determined by calculations.
4.6.3. Influence of Capillarity on Cavities
Due to tension on the free surface, the boundary should be fixed at both ends, so that the
cavity flow can be investigated using a cavitation model with an artificial solid boundary at
the cavity end. Consider a flow around a perpendicular plate with a cavity attached to the
perpendicular end plate, Fig.4.6.6. The problem statement of the cavitating flow using a
cavitation model with a plate end is similar to that considered in the previous section.
Problem solution in the parametric ζ -plane can be presented as sum (4.6.16) with the same
expansion (4.6.17) for function ω2(ζ). The solution for a cavitating flow without capillary
tension can be found with all the
singularities taken into account:
i ζ 2 + b2
ω0 = ln 2 2 2 , (4.6.20)
2 b (b ζ + 1)
where points ζ = ±bi correspond to the
solid boundary corners.
Further, an assumption should be
Fig.4.6.6. Capillary waves on cavity surface. made about the angle of free boundary at
points of detachment. Generally, it should
be given as a capillary angle between the free boundary and the boundary of another
material. Then another problem emerges concerning the detachment point location; it can be
on the front face or the rear face, even if the solid boundary is a plate with a sharp edge. This
problem is discussed in [Amromin & Ivanov, 1981] for the detachment point location on a
curvilinear surface. Below, we assume that the flow breaks away from the straight plate
along its tangent. Then, auxiliary function (4.6.15) cannot be valid. In [Zhitnikov &
Terentiev, 1988] a function was suggested determining a solution of a boundary-value
problem with the given imaginary part on the arc σ ∈ [0, π / 2] , namely β1 = B1 cos 2 σ sin σ
which conforms to behavior of function β (σ ) near σ = π / 2 . The odd additional function is
determined by the Schwartz integral
π /2
4ζ 2 sin 2σ dσ
ω1 = − B1
π 0 ∫ cos 2 σ sin σ
1 − 2ζ 2 cos 2σ + ζ 4
=
(4.6.20a)
2 ⎡ 1 ⎛ 1 + ζ 2 ⎞ ⎛ ζ 2 − 1 1 + iζ ⎞ ⎤
2
= B1 ⎢ − ⎜ ⎟ ⎜1 + ln ⎟⎥ .
π ⎢⎣ 3 ⎝ 2ζ ⎠ ⎝ 2iζ 1 − iζ ⎠ ⎥
⎦
Satisfying condition (4.6.5) at detachment point σ = π / 2 , one can obtain the factor
151
λ⎛ 1− μ ⎞
B1 = ⎜γ − , (4.6.21)
2⎝ γ ⎟⎠
∞
1 + b2
and the additional equation 2∑ m(−1) m C2 m = − . (4.6.22)
m =1 1 − b2
Parameter b was chosen in accord with solution of the cavitation problem without tension
b = 1/ 1 − μ . Moreover, the condition at infinity for velocity and condition (4.6.13) should
be satisfied. Coefficients C2 m were calculated by a collocation method as in the previous
problem. Numerical analysis shows that the free boundary for given μ (or for cavitation
number σ = − μ ) could self-intersect (see Fig.4.6.4). Three non-self-intersecting cavity
boundaries for fixed cavitation number σ = 1 are plotted in Fig.4.6.7. It is seen that the
deformed free boundary approaches the boundary without capillary tension if the number of
half waves, n, tends to infinity (then Weber number We = l ρ v∞2 / T goes to infinity too).
Comparing the shapes of
free surfaces in cavitating flows
with the shell shapes presented in
Figs.4.6.3 and 4.6.4, one can see
that every solution of a cavitation
problem is induced by an
appropriate shell shape with
β A = −π / 2 . Choosing such a shell
with some half waves on it and
continuously increasing the vertical Fig.4.6.7. Cavity shapes with capillary tension
plate length from 0 to b = 1/ 1 − μ , at μ=−1 and different numbers of waves.
one can obtain a cavity shape as
shown in Fig.4.6.7. Moreover, it should be noted that the complex shell shapes with
intersections transform into non-self-intersecting free boundaries. So, real capillary flows at
large Weber numbers have free surfaces with large numbers of small-amplitude capillary
waves.
4.6.4. Kirchhoff Flow along a Flexible Shell
Consider flow DA along the x-axis, then curvilinear
wetted part of the shell AC and free semi-infinite cavity
boundary CD, Fig.4.6.8a. Another part of the shell CB is
a circular arc with radius R = T /( p0 − p∞ ) , where p0 is
inner pressure; p∞ is the pressure at infinity and also in
the cavity. Capillary tension on the cavity boundary is
neglected, so that the velocity at the boundary has a
constant value v∞ . Condition (4.6.5) should be satisfied
on the wetted soft shell AC. Finally, the streamline
coincides with the negative x-axis. Extending
symmetrically over the x-axis, we can get a flow around a Fig.4.6.8. Kirchhoff flow
soft cylinder. Conformal mapping the flow domain on the around a flexible shell.
152
z-plane onto the upper half of unit disc of the ζ-plane, Fig.4.6.8, yields a value problem
similar to that in Sec. 4.1. The opening angle at fixed point A, πy=−2β, has to be calculated.
Due to singularities, the complex potential is
2
1⎛ 1⎞
w = ⎜ζ + ⎟ . (4.6.23)
4⎝ ζ⎠
Thus, the potential on the circle, ζ = ei σ , is equal to cos 2 σ and its derivative is
dϕ dσ = − sin 2σ . As a result, the differential of curvilinear abscissa is
ds = −(1/ v)sin 2σ dσ . (4.6.24)
Hence, the condition on flexible shell differs from (4.6.13) by factor sin σ only.
Logarithmic function of complex velocity can be expressed as before by sum (4.6.16), where
1 + iζ
ω0 = γ i ln , (4.6.25)
1 − iζ
⎡ ⎛ ζ 2 + 1 ⎞ 2 −γ 2+γ
⎛ ζ 2 +1⎞ ⎤ ρ v∞ a
ω1 = λζ ⎢ B1 ⎜ ⎟ + B2 ⎜ ⎟ ⎥, λ = , (4.6.26)
⎢⎣ ⎝ 2 ⎠ ⎝ 2 ⎠ ⎥⎦ T
∞
ω2 = ∑ C2 mζ 2 m +1 . (4.6.27)
m=0
Function (4.6.25) describes a flow around a wedge with Kirchhoff cavity attached; function
(4.6.26) considers a behavior of the complex velocity near point A(σ=π/2). Factors B1 and B2
are determined satisfying the condition by approaching point A:
μ −1 γ
B1 = − , B2 = . (4.6.28)
(2 − γ ) χ sin πγ / 2 (2 + γ )sin πγ / 2
Unknown parameter χ can be calculated by χ = lim ( v / v∞ cos χ σ ) . With regard to
σ →π / 2
∞
transformation z (σ ) it can be expressed as ln χ = −γ + ∑ (−1) m C2 m +1 . (4.6.29)
m=0
Besides, the Brillouin condition at detachment point C (σ = 0) should be satisfied as

∞
follows: ∑ (2m + 1)C 2 m +1 + λ [ B1 (3 − γ ) + B2 (3 + γ ) ] − γ = 0 . (4.6.30)
m =0
The unknown coefficients are calculated satisfying condition (4.6.13) at N points on the arc
σ ∈ [0, π / 2] in the same manner. The drag of the shell is expressed as
2
πρ v∞2 a ⎡ ⎛ B B ⎞⎤
D= ⎢ −2γ + C1 + λ ⎜ 2−1γ + 2 +2γ ⎟ ⎥ . (4.6.31)
4 ⎣ ⎝2 2 ⎠⎦
If a distance between two fixed points is given as ratio l/L, the tangent angle at point
B is to be determined for fixed point A, then the tangent angle at point B is equal to π/2;
distance l/L is to be found.
The shapes of flexible shell for various μ are plotted in Fig.4.6.9. Its flexible
boundary changes a little for μ > 1 , and becomes a multi-value function when parameter μ
decreases. Namely, a shell with two fixed points can have five different contours for the
same μ = 0.6 (curves a – e in Fig.4.6.9a), and if a point is fixed it can have four different
153
shapes for the same μ = 0.5 (curves a – d in Fig.4.6.9b). If the shell has a closed cylindrical
shape then the boundary cannot intersect the dashed straight line, so that curve a is the only
possible shape.
Fig.4.6.9. Shapes of flexible shell: (a) two fixed points (lines a – e correspond to μ = 0.6 ),
(b) single fixed point (lines a – d correspond to μ = 0.5 ).
4.6.5. Axisymmetric Flexible Bubble

Using the Pologiy integral transformation (A4.5) from App. 4, one can also calculate an axisymmetric
flexible bubble using power series expansion as in the previous section. Using dimensionless parameter
μ by equation (4.6.2) and Weber number as We = ρ v∞2 L / 2T (L is the length of generatrix line of
axisymmetric surface), condition (4.6.3) may be presented as follows:
⎛ d β cos β ⎞ ⎛ v2 ⎞
L⎜ − + ⎟ = We ⎜ μ − 1 + 2 ⎟ . (4.6.32)
⎝ ds y ⎠ ⎝ v∞ ⎠
Also, the kinematic condition on the bubble surface should be satisfied for the stream function as follows
y2
Ψ = v∞ +ψ = 0 . (4.6.33)
2
In the planar case, the problem solution can be found more easily in a parametric form by
conformal mapping the region of flow in the axial z-plane onto the exterior of the unit disk in the
auxiliary plane ζ . With allowance for the symmetry about the coordinate axes, transformation
z (ζ ) and required analitic function in (4.5.7) f (ζ ) can be expressed in the form of series:
⎛ ∞
a ⎞ ∞
b
z = L ⎜ a0ζ + ∑ 22mm++11 ⎟ , f = Lv∞ ∑ 22mm (4.6.34)
⎝ m =0 ζ ⎠ m =1 ζ
with real coefficients a0 , a2 m +1 , b2 m . At stagnation points to both sides of the bubble in accordance
with (A.4.10), condition df / dx = ±2v∞ / π must be fulfilled. Substituting functions z and f from
(4.6.34) and setting ζ = ±2v∞ / π we obtain the identical condition:
∞
2⎛ ∞
⎞
∑ 2mb
m =1 π⎝
2m ⎜ a0 − ∑ (2m + 1)a2 m +1 ⎟ = 0 .
+
m=0 ⎠
(4.6.35)
On circle ζ = eiσ the real and imaginary parts of functions z and f are equal to, respectively
⎡ ∞
⎤ ⎡ ∞
⎤
x = L ⎢ a0 cos σ + ∑ a2 m +1 cos(2m + 1)σ ⎥ , y = L ⎢ a0 sin σ − ∑ a2 m +1 sin(2m + 1)σ ⎥ , (4.6.36)
⎣ m=0 ⎦ ⎣ m=0 ⎦
154
∞ ∞
u = Lv∞ ∑ b2 m cos 2mσ , v = − Lv∞ ∑ b2 m sin 2mσ . (4.6.37)
m =1 m =1
The length of generatrix for the axisymmetric bubble is equal to
π 2 2
⎛ dx(σ ) ⎞ ⎛ dy (σ ) ⎞
L = ∫ s′(σ )dσ , s′(σ ) = ⎜ ⎟ +⎜ ⎟ . (4.6.38)
0 ⎝ dσ ⎠ ⎝ dσ ⎠
In order to determine coefficients a2 m +1 and b2 m , it is necessary to satisfy simultaneously
equation (4.6.32) and kinematic condition (4.6.33), which should be expressed as a function of
variable σ using integrals (4.5.7) and (4.5.8) and the following expressions:
x′(σ ) dβ y ′′(σ ) x′(σ ) − x′′(σ ) y′(σ )
cos β = , =− . (4.6.38)
s′(σ ) ds ( s′(σ ))3
When σ = 0 at the stagnation point, condition (4.6.32) takes the form:
2 Lx′′(0)
+ We( μ − 1) = 0 . (4.6.39)
y ′2 (0)
For numerical calculations the collocation method may be used, i.e. conditions (4.6.32) and (4.6.33)
are satisfied at discrete points σ k = π k / 2 N , k = 1, N and, in addition, equalities (4.6.35) and
(4.6.39) are fulfilled. The system of 2N+3 nonlinear equations was solved numerically with
respect to Weber number We, coefficient a0 , N first coefficients a2 m +1 , and N coefficients
b2m. All other coefficients of the sums are assumed to be zeros. Parameter μ is assumed to be
given. In the dimensionless form, length of generatrix L and velocity v∞ are eliminated.
Numerical results are shown in Fig.4.6.10, where solid and dashed lines stand for
axisymmetric and cylindrical (planar flow) shapes, respectively. Curves 1 – 4 correspond to
μ= 5, 1, 0.681 and 0, respectively, and curves 5 – 7 correspond to μ= 0.546, 1 and 5. Lines 3
and 5 are in contact at the stagnation points. In a real fluid this could lead to the formation of
toroidal bubbles or to collapse of a bubble and formating new bubbles. Curve 6 for planar
flow coincides with exact line obtained in [McLeod, 1955]. Curve 4 as in planar flow
(Fig.4.6.3) is a self-intersecting surface and, of
course, cannot realize. It is of interest
theoretically only.
Numerical analyses [Zhitnikov &
Terentiev, 1993; Terentiev et al.., 1997] show that
the change in the axisymmetric bubble
configuration is similar to that of planar
configuration. As parameter μ decreases,
contraction in the flow direction and expansion
in the transverse direction are observed. When μ
< 0.681 (in the plane case μ < 0.546) the bubble
boundaries self-intersect, which corresponds to a
flow on a two-sheet surface. In particular, when
μ = 0.548 (for the plane bubble when μ = 0.441) Fig.4.6.10. Shapes of a flexible
the bubble volume calculated theoretically is bubble and a flexible cylinder.
equal to zero. More details on a flexible shell in a
channel and in an axisymmetric tube, made by Zhitnikov [Terentiev & Zhitnikov, 2005].
155
4.6.6. Conclusions
It should be noted that all these problems could be calculated with no auxiliary functions, but
the coefficients converge much slower. Investigations show that all these problems, in spite
of their differences, have similar features.
Every described problem has a range of solutions. If the angle in the front stagnation
point does not change, a sequence of different types of solutions exists for μ < 1 with an
increasing number of half waves on the shell. If this angle can change, as in the case of
shells with fixed points, the sequence of solutions splits in two branches: one can be
extended to μ→∞, the other toμ→−∞.
For a majority of problems only a half-wave solution is physically realizable. But in
cavitating flow the cavity shape corresponding to this solution is self-intersecting. Shapes
with more than 4-5 half-waves do not have self-intersections.
In flows with separation from the shell surface, the dependencies of tension, drag,
etc. on detachment point position have an extreme when the Brillouin condition is satisfied
(when the free surface curvature at the detachment point has a finite value).
When a flow is restrained by a wall, a sucking force influences the shell. The shell
stretches in the transverse direction and the gap between the shell and the wall can vanish.
However, flow closure does not occur, because a sink appears at the point of contact. The
shell shape and flow properties are governed mainly by detachment point position, distance
to the wall, and parameter μ.
Multi-value solutions for the same physical parameters signify that surface tension
affects stability of flexible shells and continuity of cavity boundary; in real cavitating flows
the boundary continuity is not visible. Capillary tension can be the main cause of
discontinuity of cavity shape.
Fig.4.6.11 highlights a detailed effect of surface tension on flow past a
supercavitating vehicle, in this case the residual body wetting that can take place even inside
the cavity. In this case, the droplets along the lower part of the body are likely residual from
the transition from fully-wetted flow to the supercavitating condition. However, such surface
tension effects also govern the wetting of a vehicle in a partially cavitating condition or in a
condition of afterbody planing on the cavity boundary.
Fig.4.6.11. The importance of surface tension to the wetting of a body enveloped in a

supercavity. In this case, the droplets along the bottom of the body are likely the end result
of the process of cavity initiation from fully-wetted flow to a ventilated supercavitating flow.
However, surface tension effects are also important to the flow over the afterbody of a
partially cavitating or planing supercavitating vehicle, where inertial forces and (depending
on the conditions) the effects of gravity ultimately compete with surface tension. (Image
courtesy of R. Arndt.)
156
The jet of liquid that forms along the body boundary in the downstream part of the
cavity – either a strong re-entrant jet for nominally axisymmetric flows, or a more
complicated feature for flows in which three-dimensional effects are important – can have a
significant effect on the afterbody forces. Since surface tension can influence the shape of
those downstream flow features, one can conclude that its effects should be included in
models intended to estimate the afterbody forces.
157
Ch. 5. Asymptotic Method in Cavitating Flow
Ch. 5. Asymptotic Methods in Cavitating Flow

Sec. 5.1. Asymptotic Approaches in Hydrodynamics
5.1.1. Introduction
The above-considered problems show that certain assumptions relative to main differential
equations or boundary conditions can be very helpful in obtaining many practically
interesting results. Assignably, the linear and approximate approaches have been applied
from the beginning of the hydrodynamics and especially at the early 1900’s for investigation
of flow around airfoils (Prandtl and his numerous colleagues, Joukowski and his school). In
the 1930’s the linear theory was used for studying impact and penetration of a body into a
fluid with a free surface [Karman, 1929; Wagner, 1932; Sedov, 1934, 1935]. Later in that
century, Tulin [1953] used the linear theory to study cavitating flow around thin foils. The
linear and approximate theories remain popular to the present day due to their relative
simplicity and a possibility to investigate analytically [Galanin & Terentiev, 1984].
Here we dwell briefly on some particulars of the linear theory and approximate
approaches, which are used below.
5.1.2. Potential Flow of Impact Problems
The velocity potential satisfies the linear Laplace equation, therefore one should linearize the
boundary and initial conditions only. The easiest problem statement is for impact problems if
the boundaries are moving during an infinitesimal time. These processes can describe ground
explosions or high-speed impact interactions of solid bodies, etc. The integral of Bernoulli’s
equation with respect to time on infinitesimal interval (0, τ) for inviscid incompressible fluid
τ τ τ
∂ϕ ⎛ v2 ⎞ 1
is ∫0 ∂t ∫0 ⎝ 2 ⎠ ρ ∫0 pdt = 0 .
dt + ⎜ − F ⎟ dt +
One can notice that the middle integral approaches zero as time approaches zero
since the integrand is bounded. The first integral is potential ϕ(x,y,z) and the last integral is a
pulse of pressure P. Hence, P = − ρϕ ( x, y , z ) . (5.1.1)
If the pulse is applied to a part of the domain boundary, and the initial normal
velocity is applied to another part, than the potential can be obtained from a mixed value
problem. Such problems have been mostly considered in the 1930’s for hydroplanes landing
on water and for high-speed craft. Later, in the 1960’s, similar problems were considered for
explosions in soil and excavation. Many impact value-problems have been solved in
[Wagner, 1932; Sedov, 1934; Il’insky & Potashev, 1986].
5.1.3. Weakly Perturbed Flow
Consider now a weakly perturbed potential flow. Partial derivatives of potential with respect
to coordinates and time are assumed to be small values of the same order. Then expression
( ∇ϕ ) 2 is an infinitesimal of a higher order than partial time-derivative so that it can be
2
⎛ ∂ϕ ⎞
ignored in Bernoulli integral of flow. Thus p = −ρ ⎜ + gy ⎟ , (5.1.2)
⎝ ∂t ⎠
where g is the gravity acceleration directed opposite to the y-axis; arbitrary function C(t) is
included to potential ϕ ( x, y, z , t ) . If the fluid is weightless then equation (5.1.2) becomes
158
similar to dependence (5.1.1) as p = − ρϕt ( x, y , z , t ) . (5.1.3)

Let there be other coordinates x′, y ′, z ′ which moves opposite to the x-axis with
constant velocity U. Due to relations x′ = x + Ut , y ′ = y, z ′ = z , equation (5.1.3) results in
⎛ ∂ϕ ∂ϕ ⎞
p = −ρ ⎜ +U ⎟, (5.1.4)
⎝ ∂t ∂x′ ⎠
namely, the time derivative for steady flow ∂ϕ / ∂t = 0 , and so the pressure is
∂ϕ
p = − ρU . (5.1.5)
∂x′
It should be noted that function ϕ ( x′, y ′, z ′, t ) is denoted as potential of disturbed velocity
and is a linear item of an expansion of the non-linear potential:
Φ = Ux + ϕ + ... . (5.1.6)
5.1.4. Potential Flow of Compressible Fluid

Consider first a small disturbance of unmoving weightless gas. Then momentum equation
and continuity equation can be written in linear form
∂v ∇p ∂ρ
=− , + ρ 0∇ 2ϕ = 0 . (5.1.7)
∂t ρ0 ∂t
First equation in (5.1.7) yields, similar to (5.1.4), the following dependence
∂ϕ
p = − ρ0 . (5.1.8)
∂t
Denoting the sound speed in undisturbed fluid as a0 2 = (dp d ρ )0 , the partial time-derivative
∂ρ ⎛ d ρ ⎞ ∂p ρ ∂ 2ϕ
of density is =⎜ ⎟ = − 02 2 . Substituting the letter into the second equation
∂t ⎝ dp ⎠0 ∂t a0 ∂t
∂ 2ϕ
of (5.1.7) yields an equation for the potential = a0 2∇ 2ϕ . (5.1.9)
∂t 2
For the moving coordinates (x’, y’, z’) the pressure is calculated by equation (5.1.4)
with density ρ=ρ0 but equation (5.1.9) is transformed to a relatively complicated form
1 ⎛ ∂ 2ϕ ∂ 2ϕ ⎞ ⎛ U 2 ⎞ ∂ 2ϕ ∂ 2ϕ ∂ 2ϕ
⎜ + 2U ⎟ = ⎜1 − ⎟ + + . (5.1.10)
a0 2 ⎝ ∂t 2 ∂x′∂t ⎠ ⎝ a0 2 ⎠ ∂x′2 ∂y ′2 ∂z ′2
For stationary flows the last equation is simplified to
2 ∂ ϕ ∂ 2ϕ ∂ 2ϕ
2
(1 − M ) ∂x′2 ∂y′2 + ∂z′2 = 0 ,

+ (5.1.11)
where M = U / α 0 is Mach number.

Equation (5.1.11) is prevailing in the linear theory of the aerodynamics. Using affine
transformation ( x = x1 , y = y1 / 1 − M 2 , z = z1 / 1 − M 2 ), one can reduce equation (5.1.11)
to Laplace equation, i.e. the compressible fluid can be considered as incompressible fluid
flow by changing the cross profile due to this transformation.
5.1.5. Compressible Effects in Cavitating Flow
Considering an incompressible fluid and a thin body given by function y = ε f ( x) , where ε
159
is a small parameter, we find that the pressure as well as coefficients of lift and torque
depend linearly on ε and on small cavitation number σ , while the drag coefficient is a
homogeneous function of second order, so that they can be written as
ρ v2 ⎛σ ⎞ ⎛σ ⎞ ⎛σ ⎞ ⎛σ ⎞
p = ∞ σ p ⎜ ⎟ , CL = σ CL ⎜ ⎟ , CM = σ CM ⎜ ⎟ , CD = σ 2CD ⎜ ⎟ . (5.1.12)
2 ⎝ε ⎠ ⎝ε ⎠ ⎝ε ⎠ ⎝ε ⎠
Functions p, CL , CM , CD are determined by solving the problem of cavitating flow
of incompressible fluid. For a compressible fluid the body changes according to the above-
mentioned equations so that the small parameter ε should be replaced with ε / 1 − M 2
only. For instance, the flow of compressible fluid around a fully-cavitated plate with small
attack angle α determines, due to equations (2.3.35) – (2.3.37), the following expressions:
4α 2 π α Lc
Lc = − 1 , CL = ,
(1 − M )σ
2 2
(
1 − M Lc − 1 + Lc ( Lc − 1)
2
)
π Lc ⎛ ⎞
Lc
( )
2
Cm = α ⎜ 4 + 1⎟ Lc − Lc − 1 . (5.1.13)
8 1 − M 2 ⎝⎜ Lc − 1 ⎠⎟
It is evident that the cavity length increases in compressible fluid. Similar formulas
can be obtained from (2.4.18) – (2.4.20) by substituting α / 1 − M 2 instead of α for
partially-cavitated plate. In that case the cavity length is calculated from
α Lc 1 − Lc
= . (5.1.14)
( )
2
σ 1− M 2 1 + 1 − Lc
Thus, at fixed values of α and σ , the cavity length increases with Mach number increasing
but the α / σ ratio cannot be greater than 0.1925 1 − M 2 .
5.1.6. Vortex-Type Flow of Incompressible Fluids
Consider a planar flow with an undisturbed stream being parallel to the x-axis, and then the
velocity should be given as a function of y, i.e. u0 = U ( y ) . The vorticity of undisturbed flow
1 dU
is obtained as ω0 = − . (5.1.15)
2 dy
Consider small disturbances of a vortex flow. The velocity and the stream function
differ from their undisturbed values U ( y ) and ψ 0 = ∫ U ( y ) dy by infinitesimal variables
( u ′, v′ ) and ψ ′ . Substituting stream function ψ = ψ 0 + ψ ′ and then retaining linear terms,
dU d ω (ψ 0 )
equation (5.1.15) can be written as + ∇ 2ψ ′ = −2ω (ψ 0 ) − 2 ψ ′. (5.1.16)
dy dψ 0
Since ω (ψ 0 ) ≈ ω0 = − dU / 2dy , dψ 0 = Udy , linear differential equation can be obtained for
1 d 2U
infinitesimal function ψ ′ ∇ 2ψ ′ = ψ′. (5.1.17)
U ( y ) dy 2
If the vorticity is constant, the second term of equation (5.1.17) is zero so that the stream
function is harmonic.
160
5.1.7. Planar Flow of Heterogeneous Fluid

Consider a stationary planar flow of an incompressible and weightless fluid. Continuity
condition ∇ ⋅ v = 0 yields stream function ψ ( x, y , t ) , which determines velocity components
u = ∂ψ ∂y , v = −∂ψ / ∂x . Let the pressure and density for undisturbed fluid are given as
functions p0(x, y, t) and ρ0(x, y, t). In unmoved coordinates, functions u and v as well as
differences p’=p−p0 and ρ’=ρ−ρ0 are infinitesimal functions of the same order. Then Euler
∂u 1 ∂p′ ∂v 1 ∂p′
equations can be linearized =− , =− . Whence the differential equation
∂t ρ0 ∂x ∂t ρ0 ∂y
∂ ⎛ ∂ψ t ⎞ ∂ ⎛ ∂ψ t ⎞
for partial time-derivative is as follows ρ0 + ⎜ ρ0 ⎟=0. (5.1.18)
∂x ⎜⎝ ∂x ⎟⎠ ∂y ⎝ ∂y ⎠
Pressure can be obtained from Euler equation: ∇p = − ρ 0 ∂v / ∂t .
Considering the moving coordinates with constant velocity U, one can find that
stream function ψ satisfies the same equation (5.1.18) if stream is undisturbed in infinity.
Equation of continuity should be solved separately for obtaining the density distribution.
5.1.8. Boundary Conditions
Let’s x, y and x’, y’ stand for fixed and moving coordinate systems, respectively. These
systems are connected by x = s(t ) + x′, y = y ′ , and the moving system moves parallel to the
x-axis. Deformation of the contour is given by equation y ′ = f ( x′, t ) . It is assumed that
partial derivatives ∂′f / ∂x′ and ∂′f / ∂t are small of the same order. Symbol ∂′ is used for
partial derivatives calculated relative to moving coordinates. The movement and deformation
of the contour in unmoving coordinates are determined by equation
y = f ( x − s, t ) . (5.1.19)
Substituting function (5.1.19) into kinematic condition and keeping the first order
infinitesimals one can obtain the linear kinematic condition as
∂′f ( x, t ) ∂′f
v = − s (t ) + , (5.1.20)
∂x ∂t
which is satisfied on a part of straight line which has a small difference from a slender foil.
The flow relative to the moving x′, y ′ - axis has at infinity velocity U = − s . Thus
equation (5.1.20) is the streamline condition for the unmoving foil as well. If there is a free
boundary, a certain linear condition for the pressure should be fulfilled. If a body is much
small than another one, then the first body will be in a fluid disturbed by the large body. Let
w0(z) be a given complex potential of the flow without a small body, then the complex
velocity in the vicinity of an analytical point, e.g. at point z = 0, can be expanded into a
dw0
power series = c0 + c1 z + c2 z 2 + … . (5.1.21)
dz
Expansion (5.1.21) can be regarded as an asymptotic behavior of the complex
velocity at infinity for the flow around a small body. Due to convergence of the series it is
possible to keep in expansion (5.1.21) a finite number of terms only.
The first item of series (5.1.21) corresponds to a flow of undisturbed fluid. The
linear term corresponds to flow in a gravity field. Indeed, if c0 = v0 is a real number which
corresponds to stream speed at infinity, the linear transformation of Bernoulli’s integral of
161
such a flow is v0 2 Re(c1 z ) + p / ρ = 0 . (5.1.22)

Denote g = v0 c1 2
and γ = arg c1 , then equation (5.1.22) is identical to the
hydrodynamic pressure in a gravity field with acceleration geiγ . For instance, a source
located at point S(-R, 0) has dw0 / dz ≈ v0 (1 + z / R) , i.e., the flow of the source in the vicinity
of the origin is equivalent to the flow in a gravity field with acceleration on a value v0 2 / R in
the opposite direction of the x-axis. The vortex simulates a normal gravity field being normal
to the flow velocity. Therefore, disturbing a flow with some bodies or changing boundaries,
one may simulate a gravity field or something else.
5.1.9. Steady Flow of Heterogeneous Fluid around a Thin Airfoil
Consider a flow with unit speed around a thin foil
of a unit chord (Fig.5.1.1). Let’s assume the fluid
is imponderable and its density depends linearly
on vertical coordinate y:
ρ0 = 1 − λ y , (5.1.23)
where λ is a heterogeneity parameter; the density
on the foil level is unity. These assumptions do
not invalidate generality of the problem statement
Fig.5.1.1. but all values are dimensionless. The kinematic
Hydrofoil in heterogeneous fluid conditions on the foil for steady flow are obtained
from (5.1.20)
∂ψ df1,2
= = β1,2 , y = ±0, x ∈ (−1/ 2,1/ 2) , (5.1.24)
∂x dx
where functions f1,2(x) determine the upper and lower boundaries of the foil accordingly.
The stream function, according to (5.1.18), should satisfy in the flow domain the equation
∂ ⎛ ∂ψ ⎞ ∂ ⎛ ∂ψ ⎞
⎜ ρ0 ⎟ + ⎜ ρ0 ⎟ = 0. (5.1.25)
∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎠
Equation (5.1.25) is similar to the main equation of filtration theory [Charny, 1963]
and some methods from that theory can be used. Due to Charny transformation, equation
(5.1.25) can be transformed with respect to new variables
r = ρ 0 , z = −λ x (5.1.26)
to Laplace equation in cylindrical coordinates, i.e.
∂ 2ψ ∂ 2ψ 1 ∂ψ
+ + =0. (5.1.27)
∂z 2 ∂r 2 r ∂r
Transformation (5.1.24) converts cut [–½, ½] of the x-axis on the lateral surface,
−λ / 2 ≤ z ≤ λ / 2 , of the circular cylinder of unit radius. Due to conditions (5.1.24) and
(5.1.26), the partial derivatives on outer and inner surfaces are expressed as
∂ψ 1 ∂ψ 1 ⎛ λ λ⎞
=− = − β1,2 (− z / λ ), r = 1, z ∈ ⎜ − , ⎟ . (5.1.28)
∂z λ ∂x λ ⎝ 2 2⎠
Pressure is calculated similar to homogeneous fluid as p = p∞ − ρ0u = p∞ − ρ0 ∂ψ / ∂y , or
∂ψ
or p = p∞ + λρ 0 . (5.1.29)
∂r
162
Since the difference, p2 − p1 = λρ 0 [(ψ r ) 2 − (ψ r )1 ] = −λρ 0 q( z ) , the lift and toque could be
calculated by integrating along the foil.
Thus the problem determined by equations (5.1.23) – (5.1.25) is reduced to the
potential flow of a ring wing, which can be solved by allocation of ring singularities on the
cylinder [Weissinger, 1956; Terentiev, 1962]. Denoting the potentials of continuous ring
vortexes and ring sources as Ω( z , r ) and Q ( z , r ) , respectively, we present the stream
function as ψ = Ω + Q . An intensity of vortexes on the cylinder surface is obtained by
1
difference of partial derivatives with respect to x: γ = (ψ z ) 2 − (ψ z )1 = ( β1 − β 2 ) . Thus
λ
function Ω( z , r ) is determined by given boundary conditions (5.1.24). The limiting values of
its axial derivative on the outer and inner sides of cylinder are expressed as
⎛ ∂Ω ⎞ γ ∂Ω0
⎜ ⎟ =± + . (5.1.30)
⎝ ∂z ⎠1,2 2 ∂z
After substituting z = λξ / 2 the second term is expressed by full elliptic integrals
∂Ω0
= g (ξ ) , (5.1.31)
∂z
λ ⎛ ⎛ 16 ⎞ 1 1 1
⎞
where g (ξ ) = ⎜ ⎜ ⎟∫
ln γ (ξ ) d ξ − ∫ γ (ξ ′) ln | ξ ′ − ξ | d ξ ′ + ∫ γ (ξ ′)U (ξ ′, ξ )d ξ ′ ⎟ ,
8π ⎝ ⎝ λ e ⎠ −1 −1 −1 ⎠
e 1− k2 4
U (ξ ′, ξ ) = k[ K (k ) − E (k )] + ln , k= .
4k λ (ξ ′ − ξ ) 2 + 16
2
The axial derivative of the source layer on both sides of cylinder has equal limiting
values and also can be expressed by full elliptic integrals:
∂Q 1 q(ξ ′)d ξ ′ λ
1 1
∂z 2π −∫1 ξ ′ − ξ 8π −∫1
= + q(ξ ′)W (ξ ′, ξ )d ξ ′ , (5.1.32)
k
where W (ξ ′, ξ ) = ( k 2 E (k ) − 1) sign[λ (ξ − ξ ′) / 2] .
1− k 2
Summing up the limiting values of the axial derivative of stream function on both
sides of the cylinder, we obtain a singular integral equation for the source layer q (ξ ) :
∂Q 1
=− ( β1 + β 2 ) + g (ξ ), ξ ∈ (−1,1) . (5.1.33)
∂z 2λ
The known intensity γ , as well as unknown intensity q, can be presented as
trigonometric sequences after substituting ξ = − cosθ , ξ ′ = − cosθ ′ . Then the infinite linear
equations set is obtained for unknown coefficients of trigonometric series of function q. The
lift and torque with respect to the origin of coordinate (x, y) can be calculated by integrating
the pressure difference p2 − p1 = λρ 0 [(ψ r )1 − (ψ r ) 2 ] = λρ 0 q :
ρ0 λ 1
ρ0λ 1
L=−
2 −1 ∫ q(ξ )dξ , M =
4 −∫1
q (ξ )ξ d ξ . (5.1.34)
For example, let’s consider a flows of stratified fluid around a flat plate with angle
of attack α and around the symmetric Joukowski airfoil of thickness ε . Numerical results
163
obtained in [Nikitin & Terentiev, 1982] are plotted in Fig.5.1.2: curves (a) correspond to flat
plate, curves (b) – to longitudinal flow around symmetric Joukowski airfoil with ordinates
2
y1,2 = ±ε (1 − 2 x ) 1 − 4 x 2 , (5.1.35)
3 3
where ε is the maximal thickness (d) of the foil; L = πρ 0α , M = −πρ 0α / 4 are the lift and
torque of the plate in homogeneous fluid of density ρ0 . As shown in Fig.5.1.2a, the lift and
torque for the flat plate decrease then increases (linearly). Fig.5.1.2.b shows an increase of
both values for the Joukowski symmetrical foil.
Fig.5.1.2. Dependences of the lift and toque: the flat plate (a);
Joukowski hydrofoil (b).
Sec. 5.2. Boundary-Value Problem Statements in Linear Theory

5.2.1. Short History of Linear Theory
The above-considered problems and methods show that the analytically nonlinear theory has
a limited scope for investigating any problems, more precisely, computing any complicated
problems is very difficult. At the same time the first terms of asymptotic series development
have been obtained in previous chapters which are much easier for calculating and give
satisfactory numerical results. All those expressions can be obtained also from linearized
value problems. The asymptotic theory or the linear theory has attracted attention of many
scientists and engineers since the beginning of the last century.
The linear theory was evolved in the hydrodynamics first by L. Prandtl in his
scientific school at the beginning of the last century in connection with impetuous
development of the aviation. At that time, both non-linear and linearized theories have been
also evolved in many countries. Value problems have been considered by given boundaries
of flow domain which appear in the wing and propeller theory.
In the 1930th the linear theory included just the problems with free boundaries,
associated with the entry of bodies in the water for hydroplanes and speed boats. Here the
research of Wagner [1932] should be mentioned. Closed to Wagner’s approach are problems
of impact of a body against water considered by Sedov [1934; 1936]. The linear theory of
cavitating flow is connected first of all with Tulin [1953-1956], then many problems were
considered in [Acosta, 1961; Oba, 1964; Geurst, 1959 -1960; Ivanov, 1960; Auslaender,
1962; Wade, 1967; Efremov, 1974] and others. Below the linear theory is considered as
asymptotic in combination with a nonlinear theory [Terentiev, 1971].
164
5.2.2. Asymptotic Basis of Linear Theory

Let in a physical z-plane the flow domain is bonded by a system of solid surfaces γ 1 and free
boundaries γ 2 . Let these curves differ little from a system of rectilinear parallel cuts
γ = γ 1 ∪ γ 2 in some initial z-plane. Here boundaries γ 1 and γ 2 correspond to γ 1 and γ 2
respectively. Systems of coordinates x, y and x, y get out so that the real axes were parallel
to cuts γ .
Let the parameters of perturbation, ε k (k = 1, n) , are small of the same order. Its can
be quantities describing angle of attack, cavitation number, relative thickness of a hydrofoil,
distance of a stagnation point from a leading edge of the foil, a distance between two spiral
centers at cavity end, etc. The last two distances are small of second order so that its square
root means to be small parameters. The perturbation stream by ε k → 0 becomes a straight
parallel one with constant speed v∞ .
We now map the flow domain in the z -plane onto the z-plane with the above
mentioned cuts. If the parameters determine the geometrical sizes in a plane, the
transformation and the solution of a nonlinear flow problem depend analytically on small
parameters. Hence, these functions can be expanded in the power series as
z ( z ) = z + ∑ ε kν k ( z ) + ∑ ε k ε mν km ( z ) + ,
k k ,m
(5.2.1)
ω ( z ) = ∑ ε k ωk ( z ) + ∑ ε k ε mωkm ( z ) +
k k ,m
The series converge in the z-plane for all points except an infinitesimal neighborhood of
singular points as cut edges, separate singular points, etc. The second series in (5.2.1) is the
expansion of logarithmic function of complex velocity ω = ln(dw / v∞ dz ) , so that the first
term of the series is infinitesimal of first order.
The linear theory is based on linear terms of series (5.2.1) only. Flow problems in
the linear theory are determined completely by function ω ( z ) = ∑ ε k ωk ( z ) , which is
infinitesimal of first order. Namely, the derivative of complex potential is
dw / dz = (dw / dz )(dz / dz ) = v∞ dz / dz + O (ε ) ,
i.e., the complex potential differs from transformation z ( z ) by a factor and a constant of
integration, and also by a linear infinitesimal term. Since the cuts γ are parallel splits of
finite magnitude, then the differential of transformation along them is determined as
dz = (1 + O(ε ))dx . Whence, the real part of it is equal to dx = dx + O (ε ) , and the derivative
of imaginary part dy / dx = O (ε ) is infinitesimal of first order. On the other hand the
differential is equal to dz = e −ω ( z ) dw / v∞ = (1 − ω )(1 + δ (ε )) dx , where δ (ε ) is real
infinitesimal of the first order. Denote further function ω = u − iv , then the derivative of
dy
ordinate of hydrofoil is expressed as = v( x), x ∉ γ . (5.2.2)
dx
It should be noted that differentials dx and dx coincide to finite values; the linear term
( δ (ε ) − u ) could not be found in first order approach. Equation (5.2.2) is a kinematic
condition on boundary γ . Whence, the derivative of the ordinate should be small of the first
order all over outside an infinitesimal neighborhood of singular points.
165
It follows from expansion (5.2.1) that components of function ω are infinitesimal of
the first order, and then the pressure in linear approach due to Bernoulli integral looks like
p = p∞ − ρ v∞2 u . (5.2.3)
The last equation for pressure at the free boundary is a dynamic condition complementary to
(5.2.2).
5.2.3. Hydrodynamic Characteristics in Linear Theory
All behavior of the flow could be found using equations (5.2.2) and (5.2.3). Let the trailing
point of cavity of the kth foil be at point z=lk, then the complex velocity near cavity
termination takes the form ω ∼ A( z − zk ) −1/ 2 and dynamic characteristics are determined by
the following integrals:
• lift of the kth hydrofoil L = − ρ v∞2 Re ∫ ω dz ; (5.2.4)
γ (k )
• torque relative to the origin M = − ρ v∞2 Re ∫ ω zdz ; (5.2.5)

γ (k )
ρ v∞2
• drag D= Im ∫ ω 2 dz . (5.2.6)
2 zk
The integrals are integrated counterclockwise along a closed line including both
banks of cut and infinitesimal circles at edges. All these integrals are usually obtained by
linear equations (5.2.2) and (5.2.3), but they can be obtained also linearizing of non-linear
expressions. Namely, the resulting force for the kth body is determined by contour integral,
which could be linearized as
i ρ v∞2 i ρ v∞2
2 γ ∫( k ) 2 γ ∫( k )
2ω
L + iD = − e dz ≈ − [1 + 2(u + iv)][1 + c( x)ε + i y ′]dx , (5.2.7)
where c( x)ε is a real but unknown function (infinitesimal of first order), integral of that
vanish. Thus the lift is determined by (5.2.5) as infinitesimal of the first order, and the drag
is determines by (5.2.7) exactly as infinitesimals of the second order. If function ω at edges
has singularity as ( z − zk ) −1/ 2 then its square will have simple pole, and the integral at edge
is calculated as residue of function and is determined a concentric longitude force at edge:
D0 = ρ v∞2 π Re lim ( z − zk )ω 2 ( z ) . (5.2.8)
z → zk
Value problems in linear theory can be divided into three groups: (1) a continuous
flow around a given profile, when only condition (5.2.2) is satisfied; (2) a design of a
hydrofoil with given pressure distribution on the boundary, when condition (5.2.3) is
satisfied and the ordinate of the boundary γ is found; and (3) it is given partly the boundary
and partly the pressure, like in the cavitating flow (mixed value problems). The first group
both in non linear and linear statements are completely investigated in scientific literature.
The second and third groups of value problems (in non-linear theory) have been investigated
in [Tumashev & Nuzhin, 1965; Elizarov et al., 1994]. The third group contains a wide range
of value problems connected with propagation of waves on free boundary, and also
cavitating flow.
166
5.2.4. Singularity of Solution of Linear Problems

For solving value problems, besides boundary conditions (5.2.2) and (5.2.3) one also needs
to know its behavior near to special points. The basic singularities ω ( z ) are well known.
They also can be easily established by second expansion in (5.2.1) on infinitesimal
neighborhood at the singular points. Due to singularity at the cavity end (2.1.8), function
ω ( z ) at that point has the same singularity ( z − zc ) −1/ 2 . The same singularity is obtained also
by enclosing a cavity on artificial plate or on reentrant
jet and then approaching respective parameters to zero.
Consider the leading edge of a hydrofoil. Let
the edge be of an open angle πμ , Fig.5.2.1a. Let the
infinitesimal neighborhood at point A is mapped onto
some infinitesimal neighborhood at point O on the
parametric ζ -plane, Fig.5.2.1. Function
1− μ
ω ≈ ln((ζ − ε ) / ζ ) , and transformation ζ ≈ z . Fig.5.2.1. Infinitesimal
Whence the linear term relative to ε and μ is neighborhoods at leading edge.
ω ≈ z −1/ 2 + μ ln z . Hence, if the opening angle vanishes then function ω has the singularity
as z −1/ 2 ; but if the angle differs from zero then it has additionally logarithm singularity. The
logarithmic singularity appears at trailing edge of hydrofoil as
well so that is impossible in generally to satisfy the Kutta-
Joukowski condition in the linear theory.
Singularity at the leading edge for cavitating flow is
determined similarly considering infinitesimal neighborhoods
in the z -plane and in the z-plane and in the parametric ζ -
plane, Fig.5.2.2. The singularity is valid for the following
dw ζ + ε
asymptotic behaviors: ≈ , z ≈ ζ 4, z ≈ z .
dz ζ − ε
Fig.5.2.2. Infinitesimal
neighborhoods at leading edge for cavitating flow.
Hence, function ω ( z ) at the leading edge has the singularity: ω ≈ 2ε / ζ ≈ z −1/ 4 . (5.2.9)
5.2.5. Sedov’s Solution

For example, presented below is the solution given by Sedov [1950] for a continuous flow
around a thin wing foil. Let the lower and upper boundaries of the foil are determine by the
functions, respectively, y = f1 ( x), y = f 2 ( x), x ∈ [ − a, a ] . (5.2.10)
Function ω should satisfy condition (5.2.2) on the banks of cut and vanishes at infinity.
Following [Sedov, 1950] the solution is presented as a sum of two functions
ω ( z ) = ω1 ( z ) + ω2 ( z ) , (5.2.11)
where ω1 ( z ) = [ω ( z ) + ω ( z )]/ 2, ω2 ( z ) = [ω ( z ) − ω ( z )]/ 2 . (5.2.12)
The functions in (5.2.12) satisfy equality ω1 ( z ) = ω1 ( z ), ω2 ( z ) = −ω2 ( z ) , and so the
conditions on the x-axis should be satisfied as follows:
167
Im ω1 ( x) = −[ f 2′( x) − f1′( x)]/ 2, x ∈ [− a, a ], Im ω1 ( x) = 0, | x |> a,

(5.2.13)
Im ω2 ( x) = −[ f 2′( x) + f1′( x)]/ 2, x ∈ [− a, a ], Re ω1 ( x) = 0, | x |> a,
where f ′( x) is the derivative df / dx .
As seen from (5.2.13), function ω1 ( x) determines a longitudinal flow around a
symmetrical body with boundary y1 = ±0.5[ f 2 ( x) − f1 ( x)] , and function ω2 ( x) determines a
flow around an arc with boundary y2 = 0.5[ f 2 ( x) + f1 ( x)] . The value problem solutions are
f 2′( x′) − f1′( x′)
a
1
obtained in integral forms as ω1 ( z ) =
2π ∫
−a
x′ − z
dx′ , (5.2.14)
1 z − a f 2′( x′) + f1′( x′) a + x′

a
2π i z + a −∫a
ω2 ( z ) = − . (5.2.15)
x′ − z a − x′
The functions are determined taking into account the above-mentioned singularity and
condition at infinity ω1 (∞) = 0, ω2 (∞) = 0 . The lift and torque are calculated accordingly
a
a+x
L = −ρv ∫ [ f 2′( x) + f1′( x)]dx ,
2
from integrals ∞ (5.2.16)
−a
a−x
a
M = − ρ v∞2 ∫ a 2 − x 2 [ f 2′( x) + f1′( x)]dx . (5.2.17)
−a
5.2.6. Design of Thin Hydrofoils

The Sedov’s solution can be easy applied to designing hydrofoils by the given pressure
distribution on the boundary. Let the pressure be given as
p1,2 ( x) = p∞ + ρ v∞2 C1,2 ( x) , (5.2.18)
where subscripts 1 and 2 correspond to the lower and upper boundaries. Functions C1 ( x)
and C2 ( x) are assumed to be infinitesimal of first order all over apart from infinitesimal
neighborhood at edges x = ± a , where an integrable singularity can appear. In a general case,
the solution should have a singularity at both edges of the cut. The solution in that case is:
C1 ( x′) − C2 ( x′) 1 z − a ⎪⎧ a + x′ C1 ( x′) + C2 ( x′) A ⎪⎫
a a
1
ω ( z) = ∫
2π i − a x′ − z
dx ′ + ⎨ ∫
2π z + a ⎪⎩ − a a − x′ x′ − z
dx′ + ⎬ . (5.2.19)
z − a ⎭⎪
The arbitrary constant, A, in (5.2.19) could be found from closure condition (2.1.6d) which is
reduced in linear theory to Im ∫ ω ( z )dz = 0 . (5.2.20)
∞
a
a+x
Satisfying the last equation yields constant A = ∫ [C1 ( x) + C2 ( x)]
dx (5.2.21)
−a
a−x
Singling out the imaginary part of solution (5.2.19) and then integrating with respect to x
along both banks of cut, one can calculate the hydrofoil boundary.
5.2.7. Curvilinear Arc
Let the functions (5.2.18) be as C1 ( x) = −C2 ( x) = C ( x) . (5.2.22)
Then the solution is expressed by first integral in (5.2.19) only, and the arc is obtained from
168
C ( x′)
x a
1
π ∫ ∫ x′ − x
integral y= dx′dx . (5.2.23)
0 −a
Namely, if the pressure distribution is given by function
a−x
C ( x) = ε (1 + a1 x + a2 x 2 ), | ε |<< 1 , (5.2.24)
a+x
then the arc is expression as a cubic parabola
⎡ a (aa2 − a1 ) 2 ⎛ a 2 a2 ⎞ ⎤
y = ε ⎢ − 2 x3 + x − ⎜1 − aa1 + ⎟ x⎥ . (5.2.25)
⎣ 3 2 ⎝ 2 ⎠ ⎦
The lift and torque could be found directly from the given distribution as follows:
a
⎛ a 2 a2 − aa1 ⎞
L = 2 ρ v∞2 ∫ C ( x)dx = 2πρ v∞2 aε ⎜1 + ⎟, (5.2.26)
−a ⎝ 2 ⎠
a
⎛ 3a 2 a2 ⎞
M = 2 ρ v∞2 ∫ C ( x) xdx = πρ v∞2 a 2ε ⎜ aa1 − 1 − ⎟. (5.2.27)
−a ⎝ 4 ⎠
5.2.8. Symmetric Thin Body

Pressure on the lower and upper boundaries is the same, thus
C1 ( x) = C2 ( x) = C ( x) . (5.2.28)
The solution is determined by equation (5.2.19) without first integral. Let function C(x) be
parabola C ( x) = −ε (a2 x 2 + a1 x − 1) . (5.2.29)
Condition (5.2.21) yields the constant A = −π aε ( aa1 + a 2 a2 − 2 ) . (5.2.30)
The imaginary part of the solution on the upper bank of cut is expressed as
ε ⎡ 3 ⎛ a 2 a2 ⎞ a 2 a1 ⎤
v2 = a
⎢ 2 x + a1 x 2
− ⎜ 1 + ⎟ x − ⎥.
a2 − x2 ⎣ ⎝ 2 ⎠ 2 ⎦
Integrating then in interval (−a, x), we find the upper boundary of the hydrofoil in
⎡a a a 2 a2 ⎤
the form: y2 = ε a 2 − x 2 ⎢ 2 x 2 + 1 x + − 1⎥ . (5.2.31)
⎣3 2 6 ⎦
The lower boundary differs from the last equation by the opposite sign only. The ordinate
(5.2.31) can be also negative, that is impossible for a hydrofoil but simulates a flow along a
curvilinear wall with hollow. The hydrofoil generally has a vertical tangent at the edges. A
cusp-shaped edge at point ( x = a ) can be obtained by satisfying equality 2a2a2+aa1=2. Some
hydrofoils of length ( l = 2a = 2 ) and thickness (d = 2 ymax / l = 0.25) for different
coefficients are plotted in Fig.5.2.3. Longitudinal force can be found by integrating along the
1 1
boundary D1 = 2∫ ( p − p∞ ) y ′( x)dx = ρ v∞2 Im ∫ ω 2 dz ≠ 0, (5.2.32)
1 1
then it will differ from zero. A part of the force acts as a concentrated force at point (x=−1),
it can be calculated by equation (5.2.8), then the sum D0+ D1 in this case vanish.
If a1=0, a2=0, equation (5.2.30) yields an ellipse with semi-axes 1 and ε or single
cavity of cavitation number σ=2ε for positive ε, and elliptic hollow at the wall for negative
ε. The linear theory cannot simulate a real flow near stagnation points and the velocity
169
differs much more from zero.
This disparity can be
corrected by the higher
approximation method [Van
Dyke, 1964; Lighthill, 1951]. If a
body differs not much from a
slender ellipse, the Riegels factor
J = 1/ 1 + y ′2 applied to a linear Fig.5.2.3. Configuration of hydrofoils.
solution gives very good results
[Riegels, 1948]. Fig.5.2.4 shows curves calculated by equations (5.2.29) and (5.2.31) with
V(x) and P(x) calculated using Riegels factor:
V ( x) = (1 + u ) / 1 + y ′2 , P ( x) = (1 − V 2 ) / 2 . (5.2.32)
Fig.5.2.4. Comparison of calculations by linear and nonlinear theories and Riegels factor.
Sec. 5.3. Steady Flow around Cavities

5.3.1. Notes on on the Results of Linear Cavitation Theory
Cavity shapes have been studied first by Tulin [1964] who noted that the linearized theory
produces general results about cavity shapes in supercavitating flows, though cavity length L
is unknown, it is connected with cavity drag (5.2.8). In the absence of gravity, the dynamic
condition on cavity shape for disturbed velocity ω = u − iv has a simple form of
u = ( p∞ − p0 ) / ρ v∞2 = σ / 2 and u = 0 at infinity. The disturbed velocity can be determined
relative to speed value at a cavity boundary, so the following conditions should be met:
• on a cavity boundary u = 0 ;
• at infinity the cavity should be closed, i.e. u = −σ / 2 ; (5.3.1)
As usually, density, pressure and inflow speed are assumed to be unity, while the
cavity length is equal to two. Function ω ( z ) for a single cavity may be found as a solution
170
of the boundary-value problem with u = 0 at x ∈ [−1,1] and v = 0 at x ∈ (−∞, − 1) ∪ (1, ∞)
of the x-axis. Besides, the function has the same singularities at the leading and trailing
points. Hence, it can be written ω = Az / z 2 − 1 . (5.3.2)
The factor is determined from the second condition of (5.3.1), A = −σ / 2 . That suffices for
obtaining some conclusions on cavity shapes and drag.
The drag coefficient of forebody in cavitating flow, according to equations (5.2.8)
and (5.3.2), is CD0 = π Re lim[( z − 1)ω 2 ( z )] = πσ 2 / 8 . Integrating equation
z →1
yc′ = ∓σ x / 2 1 − x , one can obtain an elliptic shape of cavity with aspect ratio
2
H c / Lc = σ / 2 . The drag coefficient can be determined relative to any linear parameter d

characterizing the forebody, then CD = CD(0) d / Lc . his result leads to Tulin’s well-known
expressions for cavity length and width, which are similar to (2.2.23).
The boundary value problem for cavity can be solved using any other parametric
domain instead of the z-plane. Namely, the z-plane with a cut is mapped conformably onto a
unit disc by the Joukowski transformation, z = (ζ + ζ −1 ) / 2 . Substituting this in equation
(5.3.2) one may obtain the speed as: ω = A(ζ 2 + 1) /(ζ 2 − 1) . (5.3.3)
In that case, velocity ω becomes a single-value function which is preferable in some
cases. Besides, the cavitation number function is included in ω as a factor, so it can be
assumed to be unity. Below, the value problems are solved mostly in parametric form.
By the present time, a wide range of linearized problems has been studied. Below
only a few of them are considered – those which are original and not well known.
5.3.2. Flow through a Cascade of Cavities
Consider a cascade of cavities with period T0 = Teiβ , Fig.5.3.1a. The inlet and outlet
velocities are denoted as 1 + ω1 and 1 + ω2 , respectively. The velocity at cavity and cavity
length are assumed to be unity. As before, function ω = u − iv is pure imaginary at the cuts
and has integrated
singularities at edges. The
problem may by solved
parametrically by conformal
mapping the z-plane with
periodic cuts onto the unit
circle of a multi-sheet ζ -
plane, Fig.5.3.1b.
Fig.5.3.1. Cavitating flow through cavity cascade:
Transformation z (ζ ) can
(a) the z-plane with periodic cuts and cascade of cavities; (b)
be easily determined by its unit circle on the ζ -plane.
singularities. A derivative
of transformation dz / d ζ = zζ (ζ ) becomes a pure real value at the circle, and so it can be
extended analytically through the circle over the entire ζ -plane. There are simple poles on
the ζ -plane at points ζ = ± aeiγ and ζ = ∓ a −1eiγ only. An increments of transformation
around the first two points are ∓Teiβ , and ±Te − iβ around other points. Hence,
171
T ⎡ iβ ⎛ 1 ⎞ − iβ ⎛
1 1 1 ⎞⎤
zζ ( z ) = ⎢e ⎜ iγ
−⎟+e ⎜ − −1 iγ ⎟ ⎥
. (5.3.4)
2π i ⎣ ⎝ ζ + ae ⎠ζ − aeiγ −1 iγ
⎝ζ −a e ζ + a e ⎠⎦
1 − a2
Derivative (5.3.3) should vanish at point ζ = ±1 , that gives γ = cot β . (5.3.5)
1 + a2
Excluding parameter γ and integrating function (5.3.4) along circle ζ = eiτ , one can obtain
⎡ 1 − 2a 2 cos 2β + a 4 + 2a sin β ⎤
⎢ sin β ln +⎥
T⎢ 1 − 2a 2 cos 2β + a 4 − 2a sin β ⎥
the cavity length Lc = ⎢ ⎥. (5.3.6)
π⎢ 2a cos β ⎥
+ 2cos β arctan
⎢ 1 − 2a 2 cos 2 β + a 4 ⎥⎦
⎣
If aspect ratio T / Lc is known, the only unknown parameter a is calculated
numerically from the latter equation. Assuming the cavity length to be unity, aspect ratio
T/L=T and all linear dimensions should be determined as ratios of that length.
Transformation from the z-plane to the ζ-plane can be found from (5.3.4) numerically or
analytically. In the later case, transformation z(ζ)will be a multi-value function what causes
some difficulties for choosing the branch of function.
Perturbed velocity ω(ζ) should satisfy the same boundary condition as for a single
cavity, thus it should be expressed by the same equation (5.3.3) but with an additional
⎛ ζ 2 +1 ⎞
constant ω (ζ ) = A ⎜ 2 + Bi ⎟ . (5.3.7)
⎝ ζ −1 ⎠
The latter equation should be equal to −σ / 2 at points ζ = ± aeiγ , which corresponds to
infinity on the z-plane; the cavitation number is determined with respect to inflow speed.
σ a 4 − 2a 2 cos 2γ + 1 2a 2 sin 2γ
Thus the constants are determined as A = − , B = .
2 a4 − 1 a 4 − 2a 2 cos 2γ + 1
τ
Now, speed at cavity v = A(cot τ − B) and cavity shape yc (τ ) = ∫ v(t ) z (eit )eit idt can be
π
calculated. Since the speed squared at point ζ = 1 has a pole of second order,
ω ≈ A2 (ζ − 1) −2 + … while derivative (5.3.4) should be simple zero, zζ ≈ C (ζ − 1) + … , then
the drag is
1 π
D = Im ∫ ω 2 zζ d ζ = A2C , (5.3.8)
4 ζ =1 2
and constant C can be calculated numerically as C = ( dzζ (ζ ) / d ζ )ζ =1 .
Some numerical data are plotted in Fig.5.3.2: Cavity shapes for angle β=45° and
different cascade spacings are shown in Fig.5.3.2a. It is seen that cavities deform very much
with cascade spacing T decreasing – cavity width decreases and the camber line bends down.
As mentioned above, the outlet speed is the same as the inflow speed. Fig.5.3.2b shows a
dependence of ratio D / D0 on reciprocal of cascade spacing, 1/T , where D is drag of cavity
in a cascade; D0 is drag of a single cavity. It is seen that drag for low angle β first increases
172
with a decrease in cascade spacing then decreases. In that case, the lift should behave
similarly.
Fig.5.3.2.
Computation results
of cavities cascade:
(a) cavity shapes
for different lattice
spacings;
(b) drag
dependences.
5.3.2. Cavity in a Bounded Flow Domain

A cavity beneath free and solid boundaries has been considered by Isayev in [Egorov et al.,
1971], and in more detail, including a cavity in a jet and in a channel, in [Terentiev, 1974].
Consider first a cavity in a domain which is slightly different from a strip. Then the
cavity adds to strip a horizontal cut of cavity length l =1. The width of the strip and the
distance between the cut and the
low side are denoted as H and h,
respectively. The linearized domain
Fig.5.3.3 in the z-plane is shown in
(a) - Strip with a cut Fig.5.3.3a. Transformation from the
on the z-plane; z-plane to the parametric rectangle
(b) - rectangle on the on the ζ -plane, Fig.5.3.3b, can be
ζ -plane. found by singularities. Derivative
zζ on the fundamental rectangle of
sides π and πτ has simple poles
L L
only at points B and B′ with leading part − . Since
π (ζ − b − πτ i / 2 ) π (ζ − π + b − πτ i / 2 )
it is totally real at the horizontal sides, they can be extended through that sides on the whole
ζ -plane, and then double periodic function which is expressed by logarithmic derivative of
theta-function can be obtained. Hence, the transformation is
L ϑ (ζ − b) π ⎛ h⎞
z = ln 4 , b = ⎜1 − ⎟ . (5.3.9)
π ϑ4 (ζ + b) 2⎝ L⎠
Derivative zζ should vanish at the cut edges. Thus,
ln ′ϑ4 (a − b) − ln'ϑ4 (a + b) = 0 . (5.3.10)
2 L ϑ4 (a + b)
The cut length (cavity length) is Lc = z (π − a ) − z ( a ) = ln . (5.3.11)
π ϑ4 (a − b)
Conditions (5.3.10) and l=1 form a set of equations for unknown parameters a and τ .
173
5.3.3. Cavity in a Channel

As usually, the speed at the cavity, v0 , and density ρ are assumed to be unity. Function
ω (ζ ) = u − iv satisfies the following conditions:
• at the lower base of the rectangle corresponding to the cavity boundaries
u=0; (5.3.12)
• at the upper base of the rectangle corresponding to the channel boundaries
v=0. (5.3.13)
Besides, it has simple poles at the cut ends, so that function ω (ζ ) can be written in the form:
⎛ ϑ2 (ζ − a ) ϑ2 (ζ + a) ⎞
ω = −iA ⎜ + ⎟. (5.3.14)
⎝ ϑ1 (ζ − a ) ϑ1 (ζ + a) ⎠
Real coefficient A is obtained from the condition at infinity, ω (b + iπτ / 2) = −σ / 2 ,
−1
σ ⎛ ϑ3 (b − a ) ϑ3 (b + a) ⎞
A= ⎜ + ⎟ . (5.3.15)
2 ⎝ ϑ4 (b − a) ϑ4 (b + a) ⎠
ξ
dx L ϑ (ξ − b)
Cavity shapes can be calculated by equations y = ∫ v d ξ , x = ln 4 . (5.3.16)
a
dξ π ϑ4 (ξ + b)
It is evident that the integrand in (5.3.16) coincides with the real part of function
Ω(ζ ) = ω zζ which is double periodic with factors 1 and –1 so that it can be presented as
power series and integrated analytically. Hence,
4σ L ∞ (−1) n e −πτ n sin(π nh / H )sin n(a + ξ )sin n(a − ξ )
y= ∑
π n =1 n(1 + e −2πτ n )
. (5.3.17)
Ordinate y could be also calculated directly from the integral in (5.3.16). Function v(ξ ) has
a simple pole at point ξ = a by integrating (5.1.2) but derivative dx / d ξ vanishes at that
point. One can single out a zero of the derivative as
dx ϑ (ξ − a)ϑ1 (ξ + a ) H ϑ2 (0)ϑ3 (0)ϑ4 (0)ϑ1 (2b)
=N 1 , where N = (5.3.18)
dξ ϑ4 (ξ − a)ϑ4 (ξ + a ) πϑ4 (b − a)ϑ4 (b + a)
then theta-function ϑ1 (ξ − a) of the integrand in (5.3.16) will be eliminated. Using functions
(5.3.14) and (5.3.17) one can obtain a concentrated longitudinal force at the singularity at the
cut edges. The concentrated force at singularity points of cavity is derived using equation
LA2ϑ1 (2b)ϑ1 (2a)ϑ22 (0)
(5.1.5), so the drag coefficient is CD = . (5.3.19)
ϑ4 (b − a )ϑ4 (b + a)ϑ2 (2a)ϑ4 (0)
5.3.4. Cavity in a Jet
The statement of the value problem is determined by the following boundary conditions:
• at the lower base of rectangle, Fig.5.3.3b u = σ / 2, (η = 0) ; (5.3.20)
• at the upper base of rectangle u = 0, (η = πτ / 2) . (5.3.21)
Besides, function ω (ζ ) should have simple poles at points ζ = a and ζ = π − ζ .
Due to the above-mentioned poles the value problem solution is expressed as
σ σ
ω= −i [ln'ϑ1 (ζ − a) + ln'ϑ1 (ζ + a)] . (5.3.22)
2 4
174
The ordinate of cavity shape is found from integral
ξ
σ dx
y (ξ ) = ∫ [ln'ϑ1 (ξ − a) + ln'ϑ1 (ξ + a)] d ξ , (5.3.23)
4a dξ
where derivative dx / d ξ is calculated as before by
equation (5.3.18). The drag coefficient is
Lσ 2 (ϑ2 (0)ϑ3 (0)ϑ4 (0)) 2 ϑ1 (2b)ϑ1 (2a)
CD = . (5.3.24)
16lϑ4 (b − a)ϑ4 (b + a )ϑ4 (2a )ϑ4 (0)
Ratio CD0 / CD depends on geometric
parameters of strip as plotted in Fig.5.3.4; where solid
lines correspond to the channel, broken curves – to the
jet. Numerous calculation show (see Secs. 3.2 and 3.3)
Fig.5.3.4. Drag coefficient
that drag depends mostly on cavitation number only,
as a function of strip width
and so the latter ratio describes simultaneously the ratio
and distance.
of cavity length
to that of single cavity, Lc / L0c . Cavity shapes for H =
1 and different distances h/H are plotted in Fig.5.3.5.
Fig.5.3.6 shows a ratio of cavity’s maximum
upper ordinate to cavity length, as well as cavitation
number. The lower ordinate can be easy calculated by
− ym (1 − h / L) , so that the cavity thickness is
determined as H = ym (h / L) + ym (1 − h / L ) and the
cavity middle line is obtained by
ymid = 0.5[ y (ξ , h / L) − y (ξ ,1 − h / L)] . Note the curves
for channel are non-monotonous.
The examples considered show that the linear
theory makes it possible to get more information on
cavitating
flow even
if a single
cavity is
studied.
Fig.5.3.5.
Fig.5.3.6. Cavity shapes vs. distance h:
Maximum (a) in a channel; (b) in a jet.
upper
ordinate of cavity.
5.3.5. Cavity in a Longitudinal Gravity Field

Behavior of cavity in a longitudinal gravity field discussed in Sec. 4.3 can by easily studied
considering a linearized value problem statement. Let the gravity acceleration be directed
opposite to the x-axis; the inlet velocity is always directed along the positive x-axis. Then the
175
σ x
dynamic condition at cavity boundary is expressed as u= + , (5.3.25)
2 Fr
where Fr = v∞2 / gLc is Froude number, positive in the gravity acceleration direction opposite
to the inflow speed, and negative if the directions coincide. A reciprocal of Froude number is
infinitesimal of the same order as the cavitation number.
σ z
Consider auxiliary function Ω( z ) = ω − − . (5.3.26)
2 Fr
This function satisfies the same condition for imaginary part as function ω
Im Ω = Im ω = − y′ ; but its real part vanishes at the cavity boundary. Besides, behavior of the
function at infinity is determined by equation (5.3.25), Ω → − z / l Fr → ∞ . Hence, it can be
z −1 ⎛ B z ⎞
written as Ω= ⎜ A+ − ⎟. (5.3.27)
z ⎝ z − 1 Fr ⎠
z −1 ⎛ B z ⎞ σ z
Thus, the decision function is ω= ⎜ A+ − ⎟+ + . (5.3.28)
z ⎝ z − 1 Fr ⎠ 2 Fr
Coefficients A and B are obtained from ordinary conditions
σ 1 σ 3
A=− − , B=− − . (5.3.29)
2 2Fr 4 8Fr
Now, desired solution ω ( z ) is determined, as before, within an arbitrary cavity length only.
The cavity becomes cusped at its right end (no singularity and no upstream force) if
coefficient B vanishes; we obtain Tulin’s condition of the cavity’s cusped end [Tulin, 1964]
σ Fr = −3/ 2 . (5.3.30)
The cavity becomes also cusped at its left end (no forebody) for positive Froude number but
the cavitation number should be negative such that σ Fr = −1/ 2 . (5.3.31)
The effect of gravity on cavitating flow around a wedge has been studied in [Acosta,
1961; Galanin & Gusev, 1978]. Practically, many problems of cavitating flow considered
earlier without accounting gravity of fluid could be solved in the longitudinal gravity field
by auxiliary function as (5.3.26). Fig.5.3.7
shows the effect of Froude number on cavity
shape.
Fig.5.3.7.
Cavity shapes in longitudinal gravity fields.
Following Tulin’s remark, one can

conclude that stationary flow in the longitudinal
gravity field can be realized in water tunnels
with vertical sections and, artificially, through interference field due to neighboring bodies.
The real cavitating flows during vertical motions of bodies are in most cases unsteady and
the effects of unsteadiness can predominate. Some of those problems will be considered in
the following sections.
5.3.6. Cavity in a Transverse Gravity Field
Let the gravity acceleration be directed opposite to the y-axis, then the dynamic conditions at
the low and upper boundaries, y1 ( x) and y2 ( x) , respectively, are expressed as
176
σ
(Re ω ) n =
− kyn , n = 1, 2 , (5.3.32)
2
where k = gl / 2v∞2 , variable z = x + iy is dimensionless based on l/2, where l is the cavity
length as denoted by Tulin. Besides, function ω ( z ) should have singularities ( z ± 1) −1/ 2 at
both ends of the cavity at z = ±1 .
Tulin’s solution. Using the value problem statement in terms of complex potential, Tulin
solved this problem precisely [Tulin, 1964]. He reduced it in two ordinary differential
equations of second order for cavity thickness tc and camber ym :
d 2 tc Cd x2
+ k 2
t = −2 , (5.3.33)
π (1 − x 2 )3 / 2
c
dx 2
d 2 ym σk
2
+ k 2 ym = , (5.3.34)
dx 2
and derived its solutions in an analytical form
C x ⎛ s kJ ( k ) ⎞
tc ( x) = −2 d ∫ cos k ( x − s) ⎜⎜ + 1 (arccos s − π ) ⎟⎟ ds , (5.3.35)
π −1 ⎝ (1 − s ) J 0 (k )
2
⎠
C 1
ym ( x ) = − d (cos kx − cos k ) , (5.3.36)
π kJ 0 (k )
π
1
where Cd = 4 D / ρ v∞2 l , J n (k ) =
π ∫ cos(ns − k sin s)ds
0
(cylindrical function [Gradstein &
σ 2 cos 2 k ρ v∞2 l
Ryzhik, 1962]). Drag is obtained as D =π . (5.3.37)
8 J 02 (k ) 2
Whence, when k = 0 (no gravity), drag is D = πσ 2 ρ v∞2 l0 /16 . Comparing the latter with
equation (5.3.37) for the same drag one can obtain: l / l0 = cos 2 k / J 02 (k ) . (5.3.38)
When k = π / 2 , the right-hand term is equal to zero, or l / l0 = 0 . This implies, that l0 → ∞
(the Helmholtz flow) while length l can be finite even when σ = 0 . The Tulin’s analytical
results are very remarkable and could be easily used for computing. The cavity shape for
k = π / 2 and σ = 0 calculated in [Tulin, 1964] is shown in Fig.5.3.8.
Iterative method. Turning back to value problem (5.3.32), one can write the solution in
integral relationship as
⎡ 1 1 + s k ( y1 + y2 ) − σ ⎤
⎢∫ ds + ⎥
1 z − 1 ⎢ −1 1 − s s−z
1
k y1 − y2 ⎥
2π i −∫1 s − z
ω ( z) = ds + (5.3.39)
2π z + 1 ⎢ 1 1 1 + s ⎥
⎢+ (k ( y1 + y2 ) − σ )ds.⎥
⎢⎣ z − 1 −∫1 1 − s ⎥⎦
Imaginary parts at the low and upper sides of the cut determine the derivatives of cavity
ordinates, yn′ = −(Im ω ) n . For finding these imaginary parts, Couchy type singular integrals
should be computed. There are many numerical methods for computing Cauchy integral.
Here another numerical procedure is used, namely, the value of ω is found on straight lines
very close to the x-axis ω ( x) n ≈ ω ( x + (−1) n δ i ) , (5.3.40)
177
where δ should be a sufficiently small value; here it was assumed δ = 10−5 . Then the
Couchy singularity is avoided and integrals can be calculated using ordinary algorithms. The
cavity ordinates and its derivatives are unknown but they are connected by equations
yn′ ( x) = − Im ω[ x + (−1) n δ ; y1 ( x), y2 ( x)], x ∈ [−1, 2] . (5.3.41)
The latter can be computed iteratively. In that case the iterative integration encounters some
difficulties which can be avoided by spline-approximation at each iteration.
Iteration algorithm. Let the cavity shape for weightless flow be the initial cavity shape:
σ σ
y1(0) = − 1 − x2 , y2(0) = 1 − x2 . (5.3.42)
2 2
Iteration algorithm is as follows:
1. calculating derivatives y1′ and y2′ at N nodal points;
2. spline-approximation of functions y1′ ( x) and y2′ ( x) ;
x
3. calculating functions yn ( x) = ∫ y′ (s)ds,
−1
n (n = 1, 2) at N nodal points;
4. spline-approximation of functions y1 ( x) and y2 ( x) .

The iteration is repeated up to achieving the required accuracy. The calculations in Fig.5.3.8
show a good agreement of
analytical and numerical
results.
Fig.5.3.8.
Cavity shape in a transverse
gravity field.
5.3.7. Cavity along a Flat Wall in a Shear Flow of Constant Vorticity

A flow close to a wall or foil surface cannot be potential because of viscosity. A cavity
however can be formed directly at the wall where the vorticity is not zero. In the near field,
the local value of vorticity can be approximated by applying the principles of laminar
boundary-layer theory. The formulation presented in this section may be considered a model
problem for the effects of local shear on the local behavior of the cavity.
Let the vorticity of steady flow be constant, and velocity at infinity is a linear
function of ordinate Vx = v∞ (1 − 2ε y ), Vy = 0 . (5.3.43)
Thus the vorticity of stream is ∇ × v = ε v∞ . The fluid speed components can be expressed as
Vx = v∞ (1 − ε y + u ), Vy = v∞ v . (5.3.44)
Then, function ω = u − iv is analytical and satisfies mixed value conditions on the x-axis
Im ω = 0 for x ∈ (−∞, − 1) ∪ (1, ∞) , (5.3.45)
Re ω = σ / 2 + ε yc for x ∈ [−1,1] , (5.3.46)
Im ω = − yc′ for x ∈ [−1,1] , (5.3.47)
where yc(x) is the cavity ordinate to be found.
178
The solution of the mixed value problem of (5.3.45) and (5.3.46) is expressed similar
to the second term of equation (5.3.39) as
1 z − 1 ⎡ 1 + s 2ε yc − σ ⎤
1 1
1 1+ s
ω= ⎢∫ ds + ∫ (2ε yc − σ )ds ⎥ . (5.3.48)
2π z + 1 ⎣ −1 1 − s s − z z − 1 −1 1 − s ⎦
As before, cavity ordinate yc ( x ) is unknown and should be found from differential equation
yc′ = − Im ω or yc′ ( x) = − Im ω[ x + δ ; yc ( x)] . (5.3.49)
Numerical results can be found iteratively. Fig.5.3.9 shows cavity shapes for three numbers
of vorticity ε . According the Bernoulli equation, which is valid along a stream line, the
pressure can be found from linearized equation as
p = C + ρ v∞2 (ε yc − u ) . (5.3.50)
The horizontal force on elementary length dx is
dX = pyc′ dx = Cdyc + ρ v∞2 dy 2 / 2 − uvdx .
The integrals of two first terms along a cavity
vanish, while those of the third term are reduced
to expression (5.2.6) at points x = ±1 . As before,
the effect on cavity length can be calculated by
Fig.5.3.9. Cavity shapes on a wall in Lc / L0c ≈ CD0 / CD . (5.3.51)
a rotational flow. Calculations show that the ratio within interval
−0.5 ≤ ε ≤ 0.5 at an arbitrary small cavitation
number is approximated by straight line Lc / L0c ≈ 1 − (2 / 5)πε . (5.3.52)
5.3.8. Single Cavity in a Rotational Fluid Flow

Let the inlet velocity, as before, be a linear function of ordinate
Vx = v∞ (1 − ε y) . (5.3.53)
Contrary to the previous problem, in this case there are two free boundaries within interval
(−1,1) where kinematic and dynamic conditions should be satisfied:
′ , p = pc = const .
Vy = y1,2 (5.3.54)
As in previous problem, consider analytic function ω (z ) , which also satisfies conditions
(5.3.46) and (5.3.47). Those conditions should be met at both sides of the cut: dynamic
σ
condition (5.3.46) becomes (Re ω ) n =
+ ε yn , n = 1, 2. (5.3.55)
2
One can see that conditions (5.3.55) and (5.3.32) differ only by factors – k and ε .
Therefore, the above-mentioned Tulin’s solution is valid for this problem as well.
Sec. 5.4. Cavitating Flow around a Curvilinear Foil

5.4.1. Cavitating Flow around Slender Foils
Flow around a slender foil with a cavity can be represented as a body of closed boundaries
containing the wetted boundaries of the foil and cavity. Boundary conditions are given by
equations (5.1.2) and (5.1.3) which should be satisfied at the sides of one or several parallel
cuts. Thus, cavitating flow around slender bodies is generally a mixed-value problem.
Neither length nor detachment points of the cavity boundary with foils are given, so some
179
auxiliary conditions should be fulfilled. The linear theory provides little if any tools for
determining the detachment points since the points are located near the leading edge where
linear assumptions are invalid. In this section detachment point locations are supposed to be
given. The mixed-value problem for one or two foils as well as cascades can be solved
analytically by conformal mapping the z-plane with cuts on some auxiliary domain, usually
on the upper half plane or a rectangle. The desirable function can then be obtained using
solutions from Appendix 1. Some simple problems are discussed below.
5.4.2. Cavitating Flow around a Curvilinear Arc
Cavitating flow around a slender body is defined by a mixed value problem for the given
imaginary part on the wetted boundary and the real part at the cavity boundary, and for the
given value of the solution at infinity.
Let a curvilinear arc be determined by rational function f(x) while its derivative
df/dx=g(z) is meromorphic function g(z) on the z-plane. Consider the auxiliary function
Ω( z ) = ω ( z ) + ig ( z ) , (5.4.1)
which is purely real on the wetted boundary of hydrofoil and purely imaginary on the cavity
boundary; its value at infinity should be equal to −σ/2=−ia. Function (5.4.1) due to given
function g(z) is meromorphic with the given leading part. Thus the solution could be
obtained analytically without any integrals [Terentiev, 1972a]. For example, cavitating flow
around a parabola with bending h is y = −4hx 2 + 4hx . (5.4.2)

The flow is shown in Fig.5.4.1a. The value conditions are satisfied on the banks of cut on the
z-plane, Fig.5.2.1b, or on the ξ-axis of the auxiliary ζ-plane, Fig.5.2.1c. Transformation from
the z-plane to the ζ -plane is ζ = z /(l − z ) , (5.4.3)
where l is unknown length of the cavity. The curvilinear
arc is mapped into interval (-b, 0) with b = 1/ l − 1 .
Function (5.4.1) looks like
Ω(ζ ) = ω (ζ ) + i 4h − i8hz (ζ ) . (5.4.4)
It should satisfy the following conditions:
Im Ω = 0, ξ ∈ (−b, 0), η = 0
. (5.4.5)
Re Ω = 0, ξ ∈ (−∞, − b) ∪ (0, ∞)
Fig.5.4.1. Fully cavitating Function z (ζ ) is determined from transformation (5.4.3)
foil: (a) z-plane, (b) lζ 2
as z (ζ ) = (5.4.6)
auxiliary ζ -plane. ζ 2 +1
Then, function Ω(ζ ) has a simple pole at point ζ = i
corresponding to infinity on the z-plane, a singularity of type ζ −1/ 2 at point ζ = 0 and
bounded at point ζ = −b . Due to the singular solution in Appendix 1, function Ω(ζ ) can be
ζ +b ⎛ C + iD C − iD ⎞
expressed as Ω(ζ ) = i ⎜ Aζ + B + + . (5.4.7)
ζ ⎝ ζ −i ζ + i ⎟⎠
If function g ( x) is a polynomial of N order then two terms in (5.4.7) should be written as
180
N
⎛ Ck + iDk
Ck − iDk ⎞
the sum ∑ ⎜ (ζ − i)
k =1 ⎝
k
+ ⎟.
(ζ + i ) k ⎠
Decision function ω (ζ ) = Ω(ζ ) − i 4h + i8hz (ζ ) (5.4.8)
has five unknown constants A, B, C, D, b which can be found from the following conditions:
• bounded condition at infinity ( z → ∞ ): lim {ω (ζ )(ζ − i )} = 0 ; (5.4.9)
ζ →i
σ
• given value condition at infinity ω (i ) = − − iα ; (5.4.10)
2
• closure condition Im ∫
ζ =i
ω (ζ ) zζ (ζ )d ζ = 0 . (5.4.11)
γ −π γ⎛ γ −π ⎞
The first condition yields C + iD = −4hl sin
⎜ sin + i cos ⎟, (5.4.12)
2⎝ 4 4 ⎠
where Geurst’s symbols are used: b = cot(γ / 2), l = cos −2 (γ / 2) .
Coefficients A and B are determined from conditions (5.4.10)
γ ⎡ γ −π σ γ −π ⎛ γ −π γ γ −π ⎞⎤
A = sin ⎢ (α − 4h)sin + cos + 2hl ⎜ 2sin + cos cos3 ⎟ ⎥ , (5.4.13)
2⎣ 4 4 4 ⎝ 4 2 4 ⎠⎦
γ ⎡ γ −π σ γ −π ⎛ γ −π γ γ − π ⎞⎤
B = sin ⎢ (4h − α ) cos + sin − 2hl ⎜ 4cos − cos sin 3 ⎟ . (5.4.14)
2⎣ 4 2 4 ⎝ 4 2 4 ⎠ ⎥⎦
Condition (5.4.11) yields an equation for unknown parameter, γ :
cos(γ / 2)
σ = 2α cot(γ / 2) + 4h . (5.4.15)
1 + sin(γ / 2)
The lift is calculated by the real part of integral (5.1.5), so the lift coefficient is expressed as
πα 3 + 4sin(γ / 2)
CL = +πh . (5.4.16)
sin(γ / 2) (1 + sin(γ / 2) ) [1 + sin(γ / 2)]
2

In this case the cavity length is smaller than the chord of hydrofoil, thus the cut’s length on
the z-plane is given as unity. The transformation is expressed, as above, by functions (5.4.3)
and (5.4.6) where length l = 1. The value problem for function Ω(ζ ) is presented by
Re Ω(ξ ) = 0, ξ ∈ (0, b), η = 0,
conditions as follows: (5.4.17)
Im Ω(ξ ) = 0, ξ ∈ (−∞, 0) ∪ (b, ∞), η = 0,
where parameter b depends on cavity length l as b = l /(1 − l ) . Besides, the solution for
partial cavitation should have a simple pole at point ζ=i and a singularity of type (2.1.8) at
ζ −b ⎛ A C + iD C − iD ⎞
point (ζ=b). The singular solution is Ω(ζ ) = ⎜ +B+ + . (5.4.18)
ζ ⎝ζ −b ζ −i ζ + i ⎟⎠
As before, unknown constants A, B, C, D, b are calculated by satisfying conditions (5.4.9) –
(5.4.11). After manipulations, the final equations for cavitation number and lift coefficient
1 + sin(γ / 2)
are, respectively, σ = 2α cot(γ / 2) + 4h (1 + sin(γ / 20 ) cos(γ / 2) , (5.4.19)
1 − sin(γ / 2)
181
CL = απ (1 + 1/ sin(γ / 2) ) + hπ (1 + sin(γ / 2) )( 3 − sin(γ / 2) ) . (5.4.20)

Cavity length l = cos (γ / 2) can be found from equation (5.4.19).
2
The last two expressions were obtained differently in [Geurst & Verbruch, 1959]. The lift
coefficient for a non-cavitating arc is obtained from (5.4.20) at γ = π : CL = 2πα + 4π h .
Some calculation results are plotted in Figs.5.4.2 and 5.4.3.
Fig.5.4.2. Fully cavitating flow around an arc: left - cavity length; right - lift coefficient.
Fig.5.4.3. Partially cavitating flow around an arc: left - cavity length; right - lift coefficient.
The curves in Figs.5.4.2a and 5.4.3a show that the upper flexures miniaturize the cavity
length by full cavitating flow and the ratio α / σ for partial cavity. The cavity length on
large interval is independent of the ratio (lines 3 – 5); this implies that partial cavitating flow
of a curvilinear arc can be unstable or even impossible. The lift for that ratio is almost the
same as for non-cavitating flow, Figs.5.4.2a and 5.4.3b.
5.4.5. Computational Algorithm for Arbitrary Foil
The problem statement for an arbitrary curvilinear foil is similar to the latter two problems
but solution ω ( z ) can be obtained numerically only. Below, one of the numerical methods
182
of direct computations of singular integrals is presented.
Let the wetted boundary has a leading edge so that the solution at this point becomes
a single pole singularity but bounded at detachment points. As before, the singularity is of
type (2.1.8) at trailing points of cavities. The mixed value problem solution for the foil can
be written for the similar ζ -plane according to the above singularities and boundary
conditions in integral form. Full and partial cavitation is considered in both cases.
Full cavitation. Let detachment points xa and xb be on the low and upper boundary, y1 and
y1 , respectively, which correspond to abscises a and b on the auxiliary ζ -plane; they are
defined by equation (5.4.3). From Appendix 1, one can write the desirable function as
ζ − a ⎛ 1 b s − b v( s) ds C⎞
ω= ⎜⎜ ∫ + Aζ + B + ⎟⎟ , (5.4.21)
ζ −b ⎝πi a s − a s −ζ ζ ⎠
where v( s ) = y x [ x( s)] , a = − xa /(l − xa ) , b = xb /(l − xb ) .
Function (5.4.21) should be limited at point a, i.e.
b
1 v( s )ds C
∫
π a ( s − a)( s − b)
+ Ab + B + = 0 .
b
(5.4.22)
Besides, the function should satisfy conditions (5.4.10) and (5.4.11) too. The latter (cavity
closure condition) is preferable to present equating the contour integral of function ω dz / d ζ
near point ζ = i to zero. The integral can be calculated as
2π
∫ ω (i + 0.5e
iτ
g= )( z ′(i + 0.5eiτ )0.5eiτ idτ , (5.4.23)
0
so that the closure condition is Im g = 0 . (5.4.24)
Equations (5.4.22) and (5.2.24) together with condition (5.4.10) arrange a set of four
equations for unknown parameters A, B, C and l.
Like it was above, the lift may be calculated integrating the pressure distribution
which is reduced to contour integral (5.4.23), i.e. the lift coefficient is equal to
CL = 2 Re g . (5.4.25)
′ ≈ ∓x
If the leading part of a foil is parabolic then derivative y1,2 −1/ 2 −1
or v ≈ s , s → 0 . Thus,
integral (5.4.21) is converged in the Couchi principle and in infinitesimal interval ( −δ , δ ) is
infinitesimal of the same order as δ , and thus (5.4.21) can be rewritten as
ζ − a ⎡ 1 ⎛ −δ s − b v1 ( s )ds b s − b v2 ( s )ds ⎞ C⎤
ω= ⎢ ⎜⎜ ∫ +∫ ⎟⎟ + Aζ + B + ⎥ . (5.4.26)
ζ − b ⎣⎢ π i ⎝ a s − a s − ζ δ s − a s −ζ ⎠ ζ ⎦⎥
Integrals of Couchi type would appear when calculating the pressure distribution or
cavity boundary. The Couchi singularity can be avoided by calculating along the straight line
close to the ξ -axis, e.g. the function of low or upper side of the cut can be calculated to the
same infinitesimal δ , as ω1,2 (ξ ) = ω (ξ ∓ iδ ) . (5.4.27)
As a numerical example, the Joukowski hydrofoil is considered to be determined on a unit
chord by linearized equation y1,2 = 4hx(1 − x) ∓ 2d (1 − x) x(1 − x) , (5.4.28)
where h is the bending deflection of the camber; d is the cavity thickness at the middle of
chord. All computations have been made mostly by δ=10−5; some results were calculated by
183
δ=10−(4÷7). All results coincide well. In the spatial case h = 0, d = 0 (flat plate), all
expressions are reduced to a simple form [Terentiev, 1971],
1 + cos(( β + γ ) / 2)
σ = 2α tan(( β − γ ) / 2), CL = πα , (5.4.29)
cos 2 (γ / 2) cos(( β − γ ) / 2)
where n = xa / l = cos 2 (γ / 2), m = xb / l = cos 2 ( β / 2), a = − cot(γ / 2), b = cot( β / 2) .
Results are plotted in Fig.5.4.4 – solid lines correspond to equation (5.2.29), points – to
integral solution (5.4.26) – by v1 = v2 = −α . Numerical results coincide very well.
Unlike full cavitation, in partial cavitation both detachment point xa and cavity
trailing point xb are located at one side of the foil. The real part of function ω is zero in this
interval. The imaginary part of ω v = y ′( x) is given beyond interval ( xa , xb ) or in intervals
(b, ∞) and (−∞, 0) and at (0, a ) of the ξ -axis on the ζ -plane. As follows from
transformation (5.4.6) where l = 1 (the chord = unity), the transformation’s asymptotic
dependence on large distance is z ≈ 1 − ξ −2 , and thus, function v differs from a constant by
infinitesimal of order ξ −2 . Radical ( s − b) /( s − a) differs from a constant by infinitesimal
Δ
s − b vds
of order ξ −1 , and so the integral
−Δ
∫ s − a s −ζ
≈ O(Δ −1 ), Δ >> 1 . Hence, the
mixed value problem solution for partial cavitating can be written in the integral form as
ζ − a ⎪⎧ С 1 ⎛
−δ
s − b v1 ( s )ds
a
s − b v2 ( s )ds
Δ
s − b v2 ( s )ds ⎞ ⎪⎫
ω= ⎨ B + + ⎜⎜ ∫ +∫ +∫ ⎟ ⎬ . (5.4.30)
ζ − b ⎩⎪ ζ π i ⎝ −Δ s − a s − ζ δ s − a s −ζ b
s − a s − ζ ⎠⎟ ⎭⎪
Choosing small δ and large Δ, one can compute all behaviors of both full and partial
cavitating flows even for more than one cavity. The partial cavity on Joukowski hydrofoil is
shown in Fig.5.4.5.
Fig.5.4.4. Cavitating
flow around a plate:
(a) - cavity length l vs.
σ /α ratio and
distance to detachment
point; (b) lift
coefficient; points in
both plots correspond
to solution (5.4.26).
Fig.5.4.5.
Flow around Joukowski foil
with partial cavity.
184
Sec. 5.5. Small Nonlinear Flow Perturbations

5.5.1. Brief Review of Small Perturbation Methods
The first-order asymptotic theory shows that the linear theory results in simple formulations
of hydrodynamic problems and their solutions. For example, in frameworks of the linear
theory one can consider curvilinear hydrofoils, systems of foils and other problems which
are complicated to investigate using nonlinear theories. However, it is possible to receive full
information on flow within the bounds of the linear theory, e.g. to calculate the cavity
boundary near the detachment point or at the cavity end, etc. Below, a linearized method is
presented. It is based on the known solution of a nonlinear problem. This method allows
calculating the cavity boundary near the leading edge or flow of slightly curved arches under
arbitrary angle of attack, and also flow at the cavity end. The problem is associated with
solving the closed area problems, but is mathematically different from the problem of small
changes of contour considered in [Lavrent’ev, 1965; Tumashev & Nuzhin, 1965, Salimov,
1970]. The approaches, based on small perturbation of nonlinear problems, were also applied
to cavitating flow; e.g. Ivanov [1960] had considered a flow of foils with partial cavity. The
fullest research of cavitating flow using small perturbation approaches were given in [Tulin
& Hsu, 1977]. Imposing small perturbation on stationary nonlinear flow, Kuznetsov [1975]
developed methods of solving some problems of non-stationary flow with free boundaries.
Their problem statements had been discussed in [Gurevich & Haskind, 1953; Woods, 1955].
5.5.2. Small Perturbation Methods
Let's consider flow around a system of foils with solid boundaries γ 1 and free boundaries γ 2
on the plane of complex variable z = x + iy . Let the flow around a system of relative
contours with γ 1 with free boundaries γ 2 is known, i.e. complex potential w( z ) and complex
velocity wz ( z ) are given. It is assumed that boundaries γ 1 and γ 1 , as well as speeds V0 and
V0 at free boundaries γ 2 and γ 2 differ little from each other. For solving a flow problem,
the flow domain on the z -plane is conformal mapped onto flow domain on the z-plane. Then
complex velocity dw / dz and complex potential derivative dw / dz differ little from known
function wz ( z ) so they can be expressed as power series on degrees of small parameters ε n :
dw ⎛ ⎞
= wz ( z ) ⎜ 1 + ∑ ε n f n ( z ) + … ⎟ , (5.5.1)
dz ⎝ n ⎠
dw ⎛ ⎞
= wz ( z ) ⎜ 1 + ∑ ε nωn ( z ) + … ⎟ . (5.5.2)
dz ⎝ n ⎠
Let's remind, that parameters of perturbation can be taken as some characteristic values
(maximal deviation of tangents to γ 1 and γ 1 , or the distance between stagnation points, etc.),
which satisfy condition lim γ = γ . Keeping linear terms in series (5.5.1) and (5.5.2), we
ε n →0
dw
shall consider functions f ( z) = ∑ ε n fn ( z) = −1, (5.5.3)
n dw
dw
ω ( z ) = ∑ ε nωn ( z ) = wz −1 −1 . (5.5.4)
n dz
185
Since boundary γ is a stream line along which imaginary parts Im w and Im w take
constant values, then equation (5.5.3) yields a boundary condition for function f ( z ) ,
Im f ( z ) = 0, z ∈ γ . (5.5.5)
Boundary conditions for function ω ( z ) = u − iv can be found directly from the right part of
equation (5.5.5) substituting dw / dz = Ve − iθ and wz = Ve − iθ and taking into account obvious
relationships (V / V )γ1 = 1 + O (ε ), (V / V )γ 2 = 1 + δ + O(ε 2 ), (θ − θ )γ1 = O (ε ) , one can get
( )
u − iv = [V / V − 1 − i θ − θ ]γ Thus u = δ on γ 2 , (5.5.6)
v = θ ( s ) − θ ( s ) on γ 1 . (5.5.7)
where δ is a given small size; s and s are curvilinear abscises of boundaries γ 1 and γ ,
respectively, positive when the flow domains remain at the left from curves γ 1 and γ 1 ;
difference θ ( s ) − θ ( s ) is the angle between two tangents to curves γ 1 and γ 1 at the nearest
dθ ( s )
points. Condition (5.5.7) can be written as v = θ ( s ) − θ ( s ) + ( s − s ) on γ 1 , (5.5.8)
ds
where θ ( s), θ ( s), dθ / ds should be given as functions of s. Function s ( s ) is unknown,
but can be found from a linear integral equation derived only after solving the value
problems. According to (5.5.3) and (5.5.4), the transformation from the z -plane to the z-
dz
plane can be written in differential form as = 1 + f ( z) − ω ( z) . (5.5.9)
dz
ds
Whence, = 1 + Re [ f ( z ) − ω ( z )] . (5.5.10)
ds γ1
Equality (5.5.10) is the linear integral equation for unknown function s ( s ) .

Let's define singularities of functions f ( z ) and ω ( z ) . Singular points can be all
angular points in a plane of complex potential w and speed hodograph, including stagnation
points. Bringing them in compliance with three points of the boundaries, one can eliminate
singularities at those points. All other singularities of functions f ( z ) and ω ( z ) could be
easily determined for an auxiliary domain. Let the flow domain on the z-plane is
conformably mapped onto the auxiliary domain with a smooth boundary of the ζ-plane.
Consider now a singularity of functions near to angular points of contour γ 1 . Let z1 and z1
be coordinates of close angular points at boundaries γ 1 and γ 1 , during whose circumvention
the argument of speed gets increments πχ and πχ , respectively, with the difference of
χ − χ being a first order infinitesimal ε ; b and b + δ are matching points at the smooth
boundary on the ζ-plane (δ is first order infinitesimal). Derivatives wζ (ζ ) and wζ (ζ ) at
dw dw
= (ζ − b ) g (ζ ) = (ζ − b − ε1 ) g (ζ ) , where
χ χ
these points can be expressed as
dz dz
g (ζ ) and g (ζ ) in the vicinity of point b are continuous functions and are not equal to zero.
χ g (ζ ) g (ζ )
Then ω (ζ ) ≈ or ω (ζ ) ≈ ln (ζ − b ) . (5.5.11)
ζ − b g (ζ ) g (ζ )
186
From these equalities one can conclude that function ω (ζ ) has a simple pole if χ is not
zero while it has a logarithmic singularity if χ ≠ χ . The same approach can be used for
describing behaviors of functions near other singular points. Namely, function f (ζ ) near the
attachment point of free and solid boundaries is limited, but function ω (ζ ) has a singularity
as ζ −1/ 2 ; at the stagnation point both functions have simple poles. Thus, the problem of flow
around a modified contour is reduced to solving value problems of boundary conditions
(5.5.5) – (5.5.7) in an auxiliary domain with above-mentioned singularities and, generally, to
solve integral equation (5.5.10).
5.5.3. Flow Past Arc of Small Curvature
Let the arc in the z -plane be defined by ordinate
y = y ( x), x ∈ (0, l ) with y (0) = 0, y (l ) = lδ ,
Fig.5.5.1a; it is assumed that max {| y / l , y ′} does not
exceed a small ε . Due to the last restriction the initial
curve can be chosen as a flat plate of length l. If the
auxiliary domain is a semi-infinite strip of width π / 2 ,
Fig.5.5.1b, the Kirchhoff problem solution is
determined by functions
wz = − sin (ζ − a ) / sin (ζ + a ) , (5.5.12)
wζ = sin ( 2ζ ) sin (ζ − a ) sin (ζ + a ) , (5.5.13)
zζ = − sin ( 2ζ ) sin (ζ + a ) .
2
(5.5.14)
Fig.5.5.1. The sketch of two where a = α / 2 is abscissa of stagnation point at the

closed flow domains (a) and strip boundary, α is the angle of attack; the speed in
the auxiliary semi-strip (b) infinity is assumed to be unity. The transformation from
the flow domain of a flat plate on the z -plane to the ζ -
plane is found from equation (5.5.14) in analytical form
1⎡ π 1 ⎤
z = ⎢1 + sin α + cos 2ζ − ζ sin α + ( cos(4ζ − α ) − cos α ) ⎥ . (5.5.15)
4⎣ 2 4 ⎦
As before the plate length for simplicity is not given, it can be found by an expression as
l = ( 4 + π sin α ) / 8 . (5.5.16)
The resulting force coefficient is perpendicular to the plate and obtained by formula
Cn = 2π sin α 4 + π sin α . (5.5.17)
Let's define functions f (ζ ) and ω (ζ ) . The first function satisfies (5.5.5) and has a simple
pole at point (a). After analytical continuation through vertical boundaries it becomes an
analytical periodic function, so that it can be restored on its singularities as
f (ζ ) = D [ cot(ζ − a ) − cot(ζ + a )] + C . (5.5.18)
The last term in equation (5.5.8) is infinitesimal of a higher order than the others so that it
can be neglected. Then conditions (5.5.6) and (5.5.8) can be written as
u = 0, η ∈ (0, ∞), 0 ≤ η ≤ ∞, ξ = 0, π / 2,
(5.5.19)
u = β (ξ ) , ξ ∈ (0, π / 2), η = 0 ,
187
where
β (ξ ) = y x ( x(ξ )) , x (ξ ) =
{
2l 1 − 2cos 2ξ + ξ sin α + ⎡⎣ cos ( 4ξ + α ) − cos α ⎤⎦ / 4
(5.5.20)
}.
4 + π sin α
Besides, it has a simple pole at point a. Continuing analytically through vertical boundary,
boundary value problem (5.5.19) can be solved in periodic functions by (10.16) as
π /2
1
ω = − ∫ β (t ) ( cot(t − ζ ) − cot(ζ + t ) ) dt + B [ cot(ζ − a) + cot(ζ + a)] . (5.5.21)
π 0
The real constant B is determined satisfying the condition for velocity at infinity:
π /2
1
B = − ∫ β (t )dt . (5.5.22)
π 0
Integrating equation (5.5.9), one can obtain transformation z (ζ ) . Since derivatives zζ (ζ )

and zζ (ζ ) are bounded at point ζ = a , factors D and B should be equal (D = B). The other
factor in (5.5.17) is calculated from a condition for the length of the arc γ 1 :
⎧⎪π / 2 ⎫⎪
l l Cl Re ⎨ ∫ zζ (ξ ) ( f (ξ ) − ω (ξ ) − C ) d ξ ⎬ .
= + + (5.5.23)
⎪⎩ 0 ⎭⎪
Whence taking into account functions (5.5.18) and (5.5.21), one finds:
π B cos (α ) 1 π / 2 ⎛ α⎞
C=− − ∫ sin ( 2ξ ) sin 2 ⎜ ξ + ⎟ J (ξ ) d ξ , (5.5.24)
4l l 0 ⎝ 2⎠
π /2
1
where J (ξ ) = ∫ β (t ) ( cot(t − ξ ) − cot(t + ξ ) ) dt . (5.5.25)
π 0
The resulting force can be calculated by integrating a pressure along the arc as
π /2
i ρ ⎛ dz dw dw ⎞
2 ∫0 ⎝ dz dz dz ⎠
X + iY = − ⎜ − ⎟zζ (ξ )d ξ
From here, allocating the real and imaginary parts and dividing by ρ l / 2 gives coefficients
π /2
1
Cx = ∫ β (ξ ) sin 2ξ sin (ξ − a ) d ξ − δ ,
2
of force components (5.5.26)
l 0
π /2
4 − π sin α 2π cos α 1
4 + π sin α l ∫0
C y = Cn − C +B + J (ξ )sin 2ξ sin 2 (ξ − a) d ξ . (5.5.27)
4 + π sin α
For numerical calculations it is expedient to introduce variable μ = 2ξ and expand function
∞
β (ξ ) in a Fourier cosines-series β (ξ ) = ∑ an cos ( nμ ) (5.5.28)
n =0
π π
1 ⎛μ⎞ 2 ⎛μ⎞
with coefficients a0 = ∫β ⎜ ⎟ d μ , an = ∫ β ⎜ ⎟ cos ( nμ ) d μ . Then function
π 0 ⎝2⎠ π 0 ⎝2⎠
∞
(5.5.24) is presented as J (ξ ) = −∑ an sin ( nμ ) (5.5.29)
n =1
and all integrals in (5.5.26) and (5.5.27) can be expressed analytically:
188
1 ⎛ (2a0 − a2 )π sin α ∞ a2 k ∞
a2 k +1 ⎞
Cx = − ⎜ +∑ + ∑ ⎟ −δ , (5.5.30)
l⎜ k = 0 2(4 k − 1) k = 0 2 ( (2k + 1) − 4 ) ⎟
2 2
16
⎝ ⎠
π π sin 2 α ∞ a2 k +1
C y = Cn −
16l 2
( 2a0 cos α + 2a1 + a2 cos α ) − ∑
4l 2 k = 0 (2k + 1) ( (2k + 1) 2 − 4 )
. (5.5.31)
The coefficients of drag and lift are calculated by transformations

CD = C x cos α + C y sin α , CL = C y cos α − C x sin α . (5.5.32)
Let's consider a special case if a0 ≠ 0, an = 0 (n = 1, ∞) that corresponds to the
flat plate revolving relative to other plate at angle a0 = δ . The last two expressions yield
2π sin α
C x = −δ , (5.5.33)
4 + π sin α
2π sin α 8π cos α
Cy = −δ . (5.5.34)
4 + π sin α ( 4 + π sin α )
2
The second term in (5.5.31) is a linear member of expansion in power series on infinitesimal
2π sin (α − δ )
δ of Reley’s expression Cy = − cos δ .
4 + π sin (α − δ )
5.5.4. Design of Cavitating Hydrofoil

Unlike the linear theory, this approximation allows calculating cavity boundaries at all points
of a definite domain including infinitesimal vicinity at the leading edge. Choosing
coefficients an in the appropriate way, we can calculate a contour that would only slightly
differ from a plate, and we can also calculate the upper boundary of a cavity and all
hydrodynamic characteristics of a hydrofoil. Solid curve γ 1 and upper boundary of cavity γ 2
form a supercavitating hydrofoil. Design of supercavitating hydrofoil can be reduced to a
variation problem of selecting the highest lift coefficient CL and qualities K = CL / CD while
complying additional conditions connected with the foil rigidity, stream speed, angle of
attack, non self-crossing boundaries condition, etc.
Not discussing anymore the foil design methods, we shall note only that in flow
around a foil with a small angle of attack, the cavity boundary can theoretically approach to
the foil’s lower boundary or even cross it.. To avoid the self-crossing near the leading edge,
the lower boundary should have a negative ordinate, i.e. the foil should be calculated as a
N
multi-parametric function y = x( x − b)(l − x)∑ bk x k , x ∈ (0, l ) . (5.5.35)
k =0
Parameters b, b1, b2, …, bN could be calculated by conditional extreme methods for foil
quality. Some forms of foils with cavity boundaries and their hydrodynamic characteristics
are compared in Fig.5.5.2. All the foils have maximum ordinate h=0.05. It is seen that the
flat plate has the worst characteristics. The circular arc in Fig.5.5.2c has the best qualities,
but its lift is less than that of the foil shown in Fig.5.5.2d; moreover, near the leading edge
the distance between the circular arc and the cavity boundary is too small for providing
sufficient rigidity to the foil. Supercavitating foils of high performance are shown in
Fig.5.5.3. It is seen that the foil in Fig.5.5.3b has lift grater than that in Fig.5.5.3a, but its
189
thickness at x≈0.6 is too small.
Fig.5.5.2. Different configurations of foils and cavity boundaries.
Fig.5.5.3.
High
performance
foils.
In conclusion, it should be noted that special supercavitating hydrofoils are widely

used now in high-speed craft. Pozdunine [1945] was the first who offered to use specially
designed profiles for high-speed propellers. Tulin [1954] offered the TMB profile similar to
circular arc. Johnson [1961] investigated many other forms of supercavitating hydrofoils,
namely, one of them (five terms profile) was similar to Fig.5.5.3.
Sec. 5.6. Matched Asymptotic Expansion*

The asymptotic method presented in Sec. 5.5 makes it possible accounting the flow near
points as stagnation, attached points, or others, whereas the linear theory (Secs. 5.2 - 5.4) has
singularities at these points. As a result, the data obtained by linear theory deviate from the
exact data in such localities, but provide asymptotically accurate integrated characteristics.
Approximate but close to accurate results can be obtained using the so-called Matched
Asymptotic Expansion (MAE) methods.
Prandtl was the first who offered a similar method considering a flow of viscous
fluid. He split the viscous flow into two: the external flow of an ideal fluid and internal flow
of a viscous fluid in a narrow boundary layer. The flows are described by different
*
This section was prepared jointly with Кirill V. Rozhdestvensky.
190
asymptotic equations; their solutions are matched at their boundary. The most detailed
description of the method with numerous applications was given in [Van Dyke, 1964].
Further substantial development of the method with reference to the theory of wing and
cavitating flow was given in monograph [Rozhdestvensky, 1979], and papers
[Rozhdestvensky & Fridman, 1991; Fridman, 2002]. Numerical techniques in this section
differ from those in the above references but the Rozhdestvensky’s concept is well
preserved.
5.6.1. Description of the Method
To simplify the problem statement, only one singular point is considered. Beyond a small
vicinity of the point the sought function, ω ( z ) , has expansion (5.2.1) or
ω ( z ) = ω0 ( z ) + ω1 ( z )ε + ω2 ( z )ε 2 + … .
(5.6.1)
Let the singularity be at the origin, then introducing local (stretched) coordinates inside
small vicinity | z |< δ yields x = δ X, y = δY, z = δ Z , (5.6.2)
each of function-coefficients of series (5.6.1) can be expanded in the series too,
ω ( z ) = ∑ ωn , k ( Z )ε nδ k , (5.6.3)
n,k
where ε and δ are small parameters, generally of different orders. Series (5.6.3) has
singular coefficients so that it may be valid for | Z |≥ 1 only. In that vicinity, function ω ( z ) is
approximated by another function Ω( Z ) , which is found from a simple value problem with
the same singularity as ω ( z ) ; it can be presented as a series
Ω( Z ) = ∑ Ω m ( Z )δ m , (5.6.4)
m
or in the z-variable, Ω( z ) = Ω( z / δ ) = ∑ Ω m , s ( z )δ mε s . (5.6.5)

m,s
Series (5.6.1) and (5.6.4) are called outer and inner expansions, respectively, so that series
(5.6.3) is an inner expansion of an outer expansion, while series (5.6.5) is an outer expansion
of an nner expansion.
Auxiliary function (5.6.4) can have unknown parameters including the dependence
of small parameter δ . All unknown parameters can be found using the Van Dyke principle
[Van Dyke, 1975] which may be prescribed by equations as follows: if products ε nδ k and
δ mε s are small values of the same order, then the following conditions should be satisfied:
ε nδ k ⎛ ωnk ( z ) ⎞ ε nδ k ⎛ ωnk ( Z ) ⎞
lim ⎜ ⎟ = m s lim ⎜ ⎟ =1 . (5.6.6)
δ ε
m s
⎝ Ω ms ( z ) ⎠ δ ε ⎝ Ω ms ( Z ) ⎠
z →0 Z →∞
Therefore, the method of matched asymptotic expansion in cavitating flow consists of three
items: 1) linear problem of the whole flow; 2) nonlinear problem of the flow in small
vicinity at singular point, 3) satisfying matched conditions (5.6.6). Following for simplicity
[Rozhdestvensky, 1991], the Kirchhoff’s flow around flat plate is considered below.
5.6.2. Kirchhoff Flow around Flat Plate
Suppose the plate length is unity, angle of attack α is small and cavitation number is
zero σ = 0 . The linearized problem on the z-plane, Fig.5.6.1, is presented by the boundary
conditions as follows:
191
Re ω1 = u = 0 on BC1 and AC2 ,

(5.6.7)
Im ω1 = −v = 1 on AB .
The problem solution can be obtained as the asymptotic case, σ → 0 , from (5.4.8), but here
it will also be found by conformal mapping of the z-pane with the horizontal semi-infinite
cut onto an auxiliary lower half-plane, Fig.5.6.1b. The proper transformation is
ζ =− z . (5.6.8)
It is obvious that the value problem on the ζ -plane is
similar to the gliding problem of the plate, or to the
lifting problem of a fully wetted plate. The latter case
follows using the Schwartz reflection principle to
continue ω (ζ ) across the real ξ -axis (excluding the cut
ξ ∈ [0,1] ). That is valid also for Kirchhoff flow around
any slender curvilinear foil which was first indicated in
[Tulin, 1953].
The solution of the ensuing boundary problem
for ω1 which vanishes in infinity can be obtained using
Fig.5.6.1. Kirchhoff flow:
(a) flow on the z-plane, the Keldysh – Sedov formula (App. 1) as
(b) - the ζ -plane ⎛ ζ −1 ⎞
ω1 = i ⎜⎜ 1 − ⎟, (5.6.9)
⎝ ζ ⎟⎠
⎛ z +1 ⎞
or in the z-variable, ω1 = i ⎜1 − ⎟. (5.6.10)
⎜ z ⎟
⎝ ⎠
The right-hand term of (5.6.10) is a multi-value function and a single value branch of this
function is determined by inequality 0 ≤ arg z < 2π . Namely, on the upper boundary of
(
cavity, ω1+ = i 1 − ( x + 1) / x , ) x ∈ (0, ∞) , and the cavity boundary is found integrating
equation dyc+ / dx = − Im(ω1+ ) ,

⎧1 ⎫
yc+ ( x) = α ⎨ ⎡ 4 x 1 + x (2 x + 1) − ln( 4 x + 1 + x ) ⎤ − x ⎬ . (5.6.11)
⎩ 2 ⎢
⎣ ⎥
⎦ ⎭
The lower cavity boundary is obtained similarly integrating the function
⎛ x −1 ⎞
ω1− = i ⎜1 − ⎟ , x ∈ (1, ∞) ; (5.6.12)
⎜ x ⎟
⎝ ⎠
and the cavity boundary,
⎧1 ⎫
yc− ( x) = α ⎨ ⎡ 4 x x − 1(2 x − 1) − ln( 4 x + x − 1) ⎤ − x ⎬ . (5.6.13)
⎢
⎩2 ⎣ ⎥
⎦ ⎭
The speed distribution on the wetted plate is determined from function (5.6.1) as
1− x
u ( x) = 1 + α Re(ω1− ) = 1 − α , x ∈ (0,1) . (5.6.14)
x
Lift and torque coefficients with respect to leading edge are calculated integrating pressure along
192
1 1
1− x π
the plate, respectively, CL = −2α ∫ Re ω1− dx = 2α ∫ dx = α, (5.6.15)
0 0 x 2
5π
1 1
1− x
CM = −2α ∫ Re ω1− xdx = 2α ∫ xdx =α. (5.6.16)
0 0 x 32
It follows that the lift coefficient for flat plate by Kirchhoff flow is four times less
than that of the fully wetted plate and two times less than that for the case of a flat plate
gliding on the surface. In comparison to the fully wetted flat plate, here the centre of
pressure is shifted from the quarter-chord point to position xm = CM / CL = 5 /16 of the chord
downstream from the leading edge; the drag coefficient is equal to CD = α CL , i.e. the total
force is normal to the plate.
5.6.3. Uniformly Valid Asymptotic Solution
As follows from the preceding estimates, the linear solution is not valid in the vicinity of the
leading edge of supercavitating plate, as a result of this non-uniformity, pressure on the
wetted side at the leading edge becomes unrealistically infinite. This deficiency can be
corrected by a special consideration of the local region of the flow near the leading edge, as
first done in [Plotkin, 1978].
Fig5.6.2.
Flow near the leading edge of a semi-
infinite flat plate:
(a) - inner region;
(b) – the auxiliary ζ -plane;
(c) - the W-plane of speed potential.
Consider a small neighborhood at

the leading edge, | z |< δ << 1 . To describe the flow near this edge in more detail, introduce
local (stretched) coordinates (5.6.2). In these inner variables for δ→0, the trailing edge of the
plate recedes to infinity and the inner flow becomes as shown in Fig.5.6.2a. All variables
and functions on the stretched domain are denoted by capital letters. The problem is solved
in parametric form by mapping the flow domain, Fig.5.6.2a, onto the first quadrant of the
auxiliary ζ-plane, Fig.5.6.2b. Using the reflection principle to continue across the axis on the
whole ζ-plane, as before, one finds the derivative of complex potential (mapping of the W-
dW ζ 2 −1
plane with semi-infinite cut, Fig.5.6.2c), Wζ = =N , (5.6.17)
dζ ζ5
and the complex velocity, WZ = −(ζ − 1) /(ζ + 1) . (5.6.18)
Due to these functions, transformation from the ζ -plane to the z-plane is obtained
N⎛ 6 8 3 ⎞
integrating derivative Zζ = Wζ / WZ : Z (ζ ) = ⎜ 2 + 3 + 4 ⎟. (5.6.19)
12 ⎝ ζ ζ ζ ⎠
5.6.4. Matching of Outer and Inner Expansions
Matching the outer and inner expansions should be done for a large value of variable Z or for
a small value of ζ which can be expanded from the latter equation as
193
1/ 4
⎛ 3N ⎞
ζ =⎜ ⎟ +… . (5.6.20)
⎝ 12Z ⎠
Then the logarithmic function of speed (5.6.18) has the inner expansion
1/ 4
⎛ 4N ⎞
Ω( Z ) = ln WZ = −2ζ = − ⎜ ⎟ +… , (5.6.21)
⎝ Z ⎠
or in z-variable, Ω( z ) = −(4 N δ )1/ 4 z −1/ 4 + … , (5.6.22)
On the other hand series (5.6.1) with first non-vanishing coefficient (5.6.14) in small vicinity
may be written in both z- and Z-variables as
ω ≈ −(1/ z )1/ 4 α ≈ −(1/ Z )1/ 4 αδ −1/ 4 . (5.6.23)
Comparing equations (5.6.21) and (5.6.22) with (5.6.23) one finds the equality
4N δ = α 4 (5.6.24)
Turning back to z-variable equations (5.6.19) in view of (5.6.24) and (5.6.18) can be
α 4 ⎡ ⎛1+U ⎞ ⎛1+U ⎞ ⎤
2 3 4
⎛1+U ⎞
rewritten as x= ⎢6 ⎜ ⎟ + 8⎜ ⎟ + 3⎜ ⎟ ⎥, (5.6.23)
48 ⎢⎣ ⎝ 1 − U ⎠ ⎝1−U ⎠ ⎝ 1 − U ⎠ ⎥⎦
where U is the speed on the wetted plate; negative value U < 0 corresponds to the part of
plate from leading edge to the stagnation point, a = 17α / 48 . It is evident that the
4
stagnation point is very close to the leading edge. Dependence of velocity on abscissa U ( x)
can be obtained numerically from equation (5.6.23). Then the velocity distribution on the
⎧U ( x) for x < b,
V ( x) = ⎨
wetted plate is determined as ⎩u ( x) for x ≥ b , (5.6.24)
where b can be an arbitrary small number but interval (0, b) must include the stagnation
point; it can be calculated also from equation U(b)−u(b)=0, namely b=0.022377 for α = 10 .
C p = 1 − V 2 ( x)
Fig.5.6.3 shows distributions of pressure coefficient calculated using
different methods. Fig.5.6.3b shows a good coincidence of MAE (dotes) with exact data
(solid line) in small vicinity at the leading edge. A good coincidence is observed also on the
whole plate but this conclusion cannot be correct asymptotically since the pressure
coefficients were calculated taking into account a higher order term (aReω1)2, while speed
has been obtained from the linear problem. The dashed line obtained by linearized formula
Cp = –2aReω1 is asymptotically correct. Nevertheless, Fig.5.6.3a shows that the results can
be improved by formal use of nonlinear expressions in the linear theory.
Fig.5.6.3. Pressure
distribution along the flat plate (a) and
close to leading edge (b) for Kirchhoff
flow: solid lines – accurate nonlinear
solution, dashed line – linear theory,
dotted line — the corrected function,
points – uniformly valid MAE, chain
line, x=b, shows the small vicinity of
leading edge.
194
5.6.5. Flow around a Plate with a Spoiler

Spoiler is an outfit which hinders flow on the bottom side of foil, so that pressure and lift rise
considerably. Such an outfit can be a small rectilinear edge at the end of foil. Fig.5.6.4а
shows a plate of unit length with a spoiler and the Kirchhoff’s cavity; the spoiler length is
equal to c<<1 and makes angle β with the x-axis. Angle β is considered to be not small,
otherwise the boundary condition can be linearized and
solved by the previous formulas.
Fig.5.6.4. Flow around a plate with interceptor (a); z-plane

with the cut (b); parametric ζ - plane (c)
Logarithmic function ω(z) satisfies the following

linearized conditions:
• at the banks of cut, Fig.5.6.4b,
Re ω = 0 on CD1 and OD2 ,
(5.6.25)
Im ω = 0 on BA;
• at infinity, ω = −α i ; (5.6.26)
• singularities as simple poles at the leading and
trailing edges ω ≈ z −1/ 4 and ω ≈ ( z − 1) −1/ 2 . (5.6.27)
The type of singularity is determined by approaching spoiler length AB to zero. An
analytic solution can be obtained by conformal mapping the z-plane with semi infinity cut
onto half plane and using the Keldysh – Sedov equation. But here it is preferable mapping
onto first quadrant of the parametric ζ – plane, Fig.5.6.4c, and then recreating by its
singularities z = (ζ 2 + 1) −2 . (5.6.28)
According to boundary conditions (5.6.25), function ω accepts only real values on the x-
axis and only imaginary on the y-axis. Besides, according to features (5.6.27), it has simple
poles at the origin and at infinity. Hence, ω = Mζ + N /ζ . (5.6.29)
Unknown coefficients M and N are bounded by equation
(5.6.26), M − N = −α . (5.6.30)
Therefore, function (5.6.29) is determined to parameter N
only. It can be calculated by the method of matched
asymptotic expansion. For this purpose we shall consider
stretched variables ζ = ετ , where ε is an unknown but
small factor. In the variables, τ , the flow along a spoil is
equivalent to a flow along a semi infinite plate with a break
at the end (flow round a corner), Fig.5.6.5a. The solution of
hydrodynamic problem on the τ -plane is determined by
two functions:
• complex potential W = − Aτ 2 ; (5.6.31)
Fig.5.6.5. Inner problem: • logarithmic function of complex velocity,
(a) – flow schematics; β τ −1
(b) the parametric τ - plane Ω = ln . (5.6.32)
π τ +1
The length of spoil is
195
0 1 β /π
dW ⎛1+ s ⎞
Lint = ∫ e −Ω ds = AJ , J = 2∫ s ⎜ ⎟ ds . (5.6.33)
1
dτ 0 ⎝
1− s ⎠
If the ratio of length of spoil to length of cord λ = Lint / L foil is given, then factor A is found
from (5.6.33) A=λ/J . (5.6.34)
As the external problem is linear but internal one is nonlinear, the hydrodynamic problem
contains two unknown parameters, N and ε . To determine them the matched asymptotic
expansion, or the principle of asymptotic conformity for functions ω(ζ) and Ω(ζ) and for
derivatives of transformations dz (ζ ) / dτ and dZ (τ ) / dτ should be fulfilled, i.e.
lim[ω (ζ ) / Ω(τ )] = 1 ; lim[dz (ζ ) / dZ (τ )] = 1 . (5.6.35)
τ →∞ τ →∞
ζ →0 ζ →0
Expanding functions of variable ζ of external problem in Taylor series and then turning into
N / ετ πN
variable τ , one can obtain from the first Eq. of (5.6.35), lim =− = 1 , or
τ →∞ Ω(τ ) 2 βε
2β −4τε 2 2ε 2
N =− ε ; and from the other equation, lim = = 1 , or
π τ →∞ −2 Aτ (ζ 2 + 1)3 A
ε = A/ 2 . (5.6.36)
β 2λ ⎛ 1⎞
Thus, function (5.6.29) takes the form: ω = −αζ − ⎜ζ + ⎟ , (5.6.37)
π J ⎝ ζ⎠
where J is calculated by integral (5.6.33). For comparison,
consider analytic solution of the problem in a non-linear statement
which can be easily found in the parametric ζ-plane, Fig.5.6.6:
ζ (ζ 2 − b 2 )
wζ = , (5.6.38)
(ζ 2 + 1)3
ζ −b β ζ −a
ω = ln + ln −πi , (5.6.39)
Fig.5.6.5. Parametric ζ +b π ζ +a
ζ -plane for nonlinear β /π
ζ (ζ + b) 2 ⎛ ζ + a ⎞
flow problem. zζ = − 2 ⎜ ⎟ . (5.6.40)
(ζ + 1)3 ⎝ ζ − a ⎠
All hydrodynamic and geometric parameters are calculated from the latter equations by
integrating along the real and imaginer axes. Infinite interval of integration for plate length,
( a, ∞ ), is transformed to finite interval (0, 1) by changing variable ξ = (a + t ) /(1 − t ) ,
β /π
(t + a)(1 − t )[t + a + b(1 − t ) 2 ] ⎛ t + a + a (1 − t ) ⎞
1
J (a, b) = (1 + a ) ∫ ⎜ ⎟ . (5.6.41)
0
[(t + a) 2 + (1 − t ) 2 ]3 ⎝ t + a − a(1 − t ) ⎠
β /π
a
ξ (ξ + b) 2 ⎛ a + ξ ⎞
The length of spoiler is L( a, b) = ∫ 2 ⎜ ⎟ dξ . (5.6.42)
0
(ξ + 1)3 ⎝ a − ξ ⎠
Unknown parameters a and b are obtained from conditions for the angle of attack and ratio
λ of spoiler length and plate length
⎛1⎞ β ⎛1⎞ α +β
arctan ⎜ ⎟ + arctan ⎜ ⎟ − =0, (5.6.43)
⎝b⎠ π ⎝a⎠ 2
196
λ J ( a , b ) − L ( a, b ) = 0 . (5.6.44)
Equation set (5.6.44) can be computed for any initial values of parameters satisfying
inequalities 0<a<b<∞.
Results are presented in Table 5.6.1 where CL, CD and their ratio (foil quality)
κ=CL/CD calculated by accurate analytical formulas and the method of matched asymptotic
expansion are shown as numerator and denominator, respectively.
The results in Table 5.6.1 are calculated for angle of attack α=10°. and two values of angle
of spoil β. An asymptotic matching was made for the external linear solution with the
internal nonlinear solution. Then the common nonlinear formulas for speed ν=exp(Reω),
pressure coefficient Cp =1 – ν2, , and hydrodynamic characteristics were used.
Table 5.6.1. Numerical results by exact solutions (numerator) and MAE method
(denominator) for α=10°.
β β=90° β=45°
λ 0 0.001 0.005 0.01 0.05 0.01 0.05 0.1
0.236 0.295 0.362 0.410 0.582 0.357 0.466 0.521
C1
0.222 0.277 0.336 0.375 0.507 0.342 0.468 0.554
0.042 0.053 0.068 0.081 0.147 0.067 0.105 0.139
CD
0.039 0.050 0.064 0.075 0.134 0.065 0.105 0.143
5.671 5.575 5.300 5.047 3.946 5.284 4.417 3.748
κ 5.671 5.569 5.274 4.997 3.782 5.271 4.440 3.863
Pressure distribution is plotted in Fig.5.6.6. Solid lines correspond to exact values,

dotted line is used for the MAE method. Calculations show a satisfactory agreement for both
methods, which improves with a reduction of spoil length. As seen from Table 5.6.1, a spoil
considerably raises the lift, at the same time, raises
resistance and reduces hydrodynamic quality.
Fig.5.6.6.
Distribution of pressure coefficient on a plate with
interceptor: solid lines correspond to accurate solution;
dotted lines – to the asymptotic matched method.
In summary, it is necessary to note, that the method

of matched asymptotic expansion can be applied to many
hydrodynamic problems, as has been demonstrated in the
above mentioned works by Rozhdestvensky and Fridman.
Sec. 5.7. Asymptotic Models of Axisymmetric Cavitating Flows

5.7.1. Preliminary Notes on Axisymmetric Flow
The above-mentioned results show that many problems of planar flow around slender bodies
can be solved analytically applying within the bounds of the theory of analytic functions of
complex variable which is very effective for such problems. As a cavity in the linear theory
is elliptic with aspect ratio of σ / 2 ; the concentrated force acting on an elliptical vertex is
197
represented by equation (5.3.3). If the origin coincides with the left elliptical vertex then the
maximum ordinate of its boundary is y = σ x / 2 , and eliminating the cavitation number by
(5.3.3) yields an asymptotic expansion of the cavity boundary at infinity for vanishing
cavitation number as parabola (2.2.24), y 2 = 2 D x / π (the density and inflow speed are
unity). Analytic functions can be also used to investigate an axisymmetric flow (Sec. 4.5),
but convergence of series worsens for slender bodies and small cavitation numbers, thus
making computing problematic in that case. The speed potential and the stream function can
be represented as a generalized analytic function only, and so they cannot be expanded in a
series like (5.2.1) even for a slender body. Moreover, axial cavitating flows cannot be
described exactly by any source distribution along axis [Birkhoff & Zarantonello, 1957]. All
these reasons make exact analytical study of axisymmetric flows difficult. Therefore, one
can use asymptotic and approximate approaches based on conservation theorems and
experimental observations as well.
The first results about asymptotic expansion for an axisymmetric cavity have been
obtained by Levinson [1946] and Gurevich [1947] independently. The first author proceeded
from an integral identity for harmonic function assuming constant velocity on the streamline
and obtained asymptotic expansion in dimensionless form as
cx1/ 2 ⎛ ln(ln x) ⎛ 1 ⎞⎞
y = 1/ 4 ⎜ 1 − + O⎜ ⎟⎟, x → ∞ . (5.7.1)
(ln x) ⎝ 8ln x ⎝ ln x ⎠ ⎠
π c4 π y4
Factor c is connected with drag by Petrov as D = = ln x . As shown in [Gurevich,
8 8x2
1947a] the first term of expansion (5.7.1) corresponds to a semi-infinite body of finite drag.
In [Petrov, 2005] the expansion of a free boundary is found in a general form as:
−1
x ⎛ 1 x 1 ⎞⎛ x ⎞
y 2 = 4Rx ln ⋅ Σ, Σ = 1 + ⎜⎜− ln ln + ⎟⎟⎟⎜⎜⎜ln ⎟⎟⎟ +
R ⎝ 4 R 2⎠⎝ R ⎠
(5.7.2)
⎛3 x 1 x 5 ⎞⎛ x ⎞⎟
−2 ⎡⎛ x ⎞−3 ⎤ D
⎟
+ ⎜⎜ ln ln − ln ln + ⎟⎟⎜⎜⎜ln ⎟⎟ + O ⎢⎜⎜ln ⎟⎟ ⎥ , R = ⎟ 2
⎝ 32 R 2 R 4 ⎠⎝ R ⎠ ⎢⎝ R ⎠ ⎥ 2πρv∞2
⎣ ⎦
As shown in [Petrov, 2005], all expansion terms are independent of body form; its
remainder term is infinitesimal order of O(R/x) and is independent of body form as well. The
first term can be also taken out heuristically from the kinetic energy law for slender body
[Logvinovich, 1969] as well as from asymptotic expansions of a solution of source
distribution along the x-axis [Yakimov, 1968; Serebryakov, 1998; Petrov, 1986], or by
variation methods [Petrov, 1981]. Detailed references of the results of cavitating flow of
axisymmetric body can be found in [Birkhoff & Zarantonello, 1957; Gurevich, 1979], as
well as in Serebryakov, 2002]. Here only slender bodies of revolution and small cavitation
numbers will be considered. Before discussing the asymptotic models, it is desirable to first
present some results for an ellipsoid which are useful for comparison of different approaches
and obtaining simple expressions.
5.7.2. Ellipsoid
The exact analytical solution can be found using elliptical variables [Loyzyansky, 1959].
1 − x2
The speed distribution over ellipsoid y = ε 1 − x 2 is V ( x) = V0 , (5.7.3)
1 − (1 − ε 2 ) x 2
198
3
1− ε 2
where V0 = . From exact solution (5.7.3) one can obtain
( )
1 − ε 2 − ε 2 ln ⎡ 1 + 1 − ε 2 / ε ⎤
⎣⎢ ⎦⎥
asymptotic expansions of various forms. E.g. in [Van Dyke, 1964, p106] it is given as an
⎡ 2 1⎛ 1 ⎞⎤
expansion of high order: V ( x) = 1 + ε 2 ⎢ln − ⎜ 1 + 2 ⎟⎥
+ O ( ε 4 ln 2 ε ) (5.7.4)
⎣ ε 2 ⎝ 1 − x ⎠⎦
For a slender ellipsoid (its vertical axis is small, ε << 1 , the speed on the major part
in the ellipsoid midbody is almost constant V0 = V (0) , and so the ellipsoid can be considered
as a cavity with cavitation number σ = V (0) 2 − 1 . Although this assumption is rough but it
gives the way to take some simple expressions for small cavitation numbers. Suppose the
pressure inside cavity is equal to p0 = p∞ − (V02 − 1) / 2 , then the drag of a half of ellipsoid
can be calculated integrating p − p0 along the surface,
0
π V02 1 + 2ln ε − ε 2
D∗ = −2πε 2 ∫ ( p − p0 ) xdx = − ε 4 . (5.7.5)
−1
2 (1 − ε 2 ) 2
Due to a constant value of speed at the cavity boundary, the drag of any body having
the same cavity is determined by Eq. (5.7.4). The drag coefficient with respect to maximal
area of section СD* = 2 D / πε 2 is marked “+” in Fig.5.7.1; other marks correspond to
experimental data obtained in [Egorov et al., 1971]. Solid line shows numerical calculations
by Riabouchinsky model in [Guzevsky, 1979]. It shows a good agreement between all
presented values.
Fig.5.7.1.
Ratio of drag coefficient per
largest sectional area vs.
cavitation number: solid line is
nonlinear numerical results
[Guzevsky, 1979]; +++ results
for the ellipse; points (1-5) are
experimental results for different
configurations of forebody of
cavitators, [Egorov at al., 1971].
Approaching parameter ε to zero yields asymptotic expansions of drag. If the origin

of X and Y coincides with the right vertex of the ellipse then the ellipsoid approaches
paraboloid Y 2 = 2ε 2 X , and the drag (5.7.5) approaches as D∗ ≈ −πε 4 ln ε , so that the
asymptotic expression of drag is
Y4 X
D∗ ≈ π 2
ln 2 (5.7.6)
8X Y
The letter differs from (5.7.2) by the logarithmic function only. Thus in spite of the
fact that the cavity looks like ellipsoid it cannot completely describe the cavity.
Nevertheless, the ellipsoid can be useful for studying a cavitating flow, e.g. using the method
of matched asymptotic expansion. Some axisymmetric problems connected with the method
of MAE have been presented in [Cole, 1968].
199
Denoting the speed components on the ellipsoid by Vx and Vx one can describe its
behavior. The asymptotic expansion of speed on ellipsoid for small ε is given by (5.7.4).
εx
The speed components should satisfy kinematic condition Vy = −Vx . Thus it is valid
1 − x2
⎛ ε 2 x2 ⎞
for all points beyond a small vicinity of the ellipse top: Vx ⎜1 + 2 ⎟
= V . Whence,
⎝ 2(1 − x ) ⎠
1
Vx ≈ 1 + ε 2 ln + O (ε 2 ) ,
Vy ≈ O(ε ) . (5.7.7)
ε
Although numerical methods allow computing many flow problems in a nonlinear statement
(Ch. 7), slender bodies and small cavitation numbers can hardly by treated by those methods.
Here only slender bodies of revolution and small cavitation numbers will be considered.
5.7.3. Source Distribution
Like in the planar linear theory, the source distribution is mostly used for simulation of an
axial slender body though it cannot describe exactly the flow around any body of revolution
[Birkhoff & Zarantonello 1957]. Let the source intensity at the x-axis be q(ξ )d ξ then the
potential and stream functions of perturbed flow are dϕ = − q (ξ )d ξ / 4π r and
dψ = − q(ξ )( x − ξ ) d ξ / 4π r , respectively. If a half-length of body and inlet velocity are
unities then the speed-potential and stream function are presented as
1 q (ξ )d ξ
1
Φ = x + ϕ, ϕ = − ∫
4π −1 r
, r = ( x − ξ )2 + y 2 , (5.7.8)
1 q (ξ )( x − ξ )d ξ
1
y2
4π −∫1
Ψ= +ψ , ψ = − . (5.7.9)
2 r
Assume the ordinate of axial section of a body (cavity), y = f ( x) , and its derivatives are
small of the same order as ε . Since distance r approaches | x − ξ | as the ordinate
approaches zero hence integral (5.7.8) is expressed as
1 ⎛ ⎞
1 x 1
1
ψ =− ∫
4π −1
q(ξ ) sign( x − ξ )d ξ = − ⎜ ∫ q (ξ )d ξ − ∫ q(ξ )d ξ ⎟
4π ⎝ −1 x ⎠
The stream function vanishes on the body of revolution, i.e. f 2 ( x) − 2ψ ( x) = 0 . Thus, one
can find after differentiating the strength of source distribution
q( x) = 2π f ( x) f ′( x) = dS ( x) / dx , (5.7.10)
where S = π f 2 ( x) is the cross-sectional area.
The speed components are calculated differentiating potential (5.7.8). Since integrals (5.7.8)
and its derivatives u = ϕ x and v = ϕ y on a slender body are small, the pressure coefficient
with respect to inflow speed is, as in a planar case, equal to С p = 2( p − p∞ ) / ρ v∞2 ≈ −2u ( x).
5.7.4. Comparison for an Ellipsoid

The pressure coefficient with respect to the maximal speed is as fallowing:
• for nonlinear solution (5.7.3), С1 ( x) = 2( p − p0 ) = V (0) 2 − V ( x) 2 ; (5.7.11)
200
• for linear solution (5.7.8) without keeping terms of higher order,

С2 ( x) = 2[u (0) − u ( x)] ; (5.7.12)
• for linear solution keeping terms of higher order,
С3 ( x) = C2 ( x) + C2 ( x)[u (0) + u ( x)]/ 2 . (5.7.13)
Fig.5.7.2.
Comparison of nonlinear
and linear results for
ellipsoid: (a) on the left
half-ellipsoid quarter; (b) in
a small vicinity of the
leading edge.
Distributions of pressure
coefficients at the left half-
ellipse with aspect ratio 1/b
are plotted in Fig.5.7.2; the
lines in Fig.5.7.2b correspond to a small vicinity at the leading edge for b = 0.05 and 0.1.
The dash line in Fig.5.7.2a shows how the difference of velocities Δv=V(0)–u(0) at middle
point x = 1 depend on thickness of ellipse b. The ordinates of the line must be divided by
100. This implies that the values calculated by nonlinear and linear expressions differ very
little for small thicknesses. Also, speed at a slender ellipse changes very slow along the
length except for a small vicinity at edges. Thus the ellipsoid nearly describes a cavity of
cavitation number σ = V(0)2–1. The dependence of cavitation number (multiplied by 5) on
thickness is shown in Fig.5.7.1a.
5.7.5. Axisymmetric Flow Past a Slender Body
A source distribution is used mostly to study a flow around a slender body of revolution. The
first asymptotic solutions have been obtained by Grigoryan [1959]. He considered an
unsteady problem and reduced it to a nonlinear differential equation, which looks by
marking (5.7.10) as Pu = 0 , (5.7.14)
2
1 (ut )
where differential operator Pu = utt ln u + − 2σ , (5.7.15)
2 u
u = f 2 ( x, t ) = S / π ; ut and utt are first and second order derivatives with respect to time t.
If a body moves steadily with unit speed then variable x of the moving coordinates is
equal to time t, and equation (5.7.14) becomes an ordinary differential equation for function
u ( x) . Differential operator (5.7.15) can be reduced to Pu = (u ′2 ln u )′ / 2u ′ , and the
differential equation may be integrated analytically u ′2 ln u = 4σ u + C1 . Whence,
u
ln u
x(u , C1 , C2 ) = ± ∫ du (5.7.16)
C2
4σ u + C1
Arbitrary constants C1 and C2 could be determined from conditions at the detachment point
on a body from continuity conditions of ordinates and their derivatives. But in that case, the
cavity cannot be closed. Yakimov [1968] added to equations (5.7.14) some terms which
201
account for the body as well as the trailing surface (Riabouchinsky model). If both bodies
are cones located at intervals ( x0 , x1 ) and ( x2 , x3 ) then the differential equation looks like
2u ( x1 ) x − x0 + ( x − x0 ) + u 2u ( x2 ) x3 − x + ( x3 − x) + u
2 2
P (u ) = ln + ln . (5.7.17)
x1 − x0 x − x1 + ( x − x1 ) 2 + u x3 − x2 x2 − x + ( x2 − x) 2 + u
Both constants can be obtained from the derivatives continuity conditions at points x1 and x2.
More accurate asymptotic expansions for axisymmetric cavities have been made by
Petrov [2005]. He considered also the source distribution but gave algorithms of taking into
account small terms of high order. Following Petrov’s presentation, let’s first rewrite integral
1 q (ξ ) − q( x) q ( x) d ξ
1 1
4π −∫1 4π −∫1 r
(5.7.8) in an identical form ϕ = − d ξ − . The first integral is
r
non-singular as y→0; the second one can be integrated analytically and asymptotic
expansion can be found, so the leading terms of the disturbed potential are
1 ⎛ 4(1 − x 2 ) ⎞
ϕ= ⎜ Iq ( x ) − q ( x ) ln ⎟ + O(ε ln ε ) , y → 0 ,
4 2
(5.7.18)
4π ⎝ y2 ⎠
q ( x) − q (ξ )
1
where ε = max | f ( x) | ; Iq( x) is integral operator Iq ( x) =
x∈( −1,1)
−1
∫ x −ξ
dξ (5.7.19)
It was shown in [Petrov, 1986; Fedoruk, 1987] that the Legendre polynomial is an eigen-
n
function of operator (5.7.19): IPn = λn Pn , where λn = 2∑1/ k , λ0 = 0 . (5.7.20)
k =1
Changing variables q(x ) = 4πs ′(x ), s(x ) = f 2 (x ) / 4 (5.7.21)

4(1 − x 2 )
makes it possible to simplify potential as ϕ = Is ′( x) − s′( x) ln + O (ε 4 ln 2 ε ) . Hence,
y2
the speed distribution and the pressure coefficient on the surface of thin body are as follow:
2
v 2 − v 2∞ (s ′) s 2s ′(1)
−cp (x ) = 2
= + 2s ′′ ln 2
+ 2I (s ′′) + −
v∞ s 1−x 1−x (5.7.22)
′
2s (−1)
− + O(ε 4 ln2 ε)
1+x
If s(x ) is polynomial then coefficient cp (x ) is simply calculated using (5.7.20).
E.g. for ellipsoid y = ε 1 − x 2 , s = ε 2 (1 − x 2 ) / 4, Is′′ = 0 one can obtain from (5.7.22)

the Van Dyke’s expansion (5.7.4). In the case of cavitating flow, the dynamic condition is
−c p = σ and so the integral-differential equation from (5.7.2) is created as
(s ′)
2
s ′′ ln s +
2s
+
d
dx
( (
I (s ′) − s ′ ln 1 − x 2 )) = σ2 . (5.7.23)
Eg. (5.7.23) was solved in the form of expansion by [Petrov, 1986]
202
y 2 = ε (1 − x 2 ) + εε1 ⎡⎣1 − (1 − 2ln 2) x 2 − (1 + x) ln(1 + x) − (1 − x)ln(1 − x) ⎤⎦ + O(εε12 ),

(5.7.24)
σ = ε ln(1 / ε ), ε1 = 1 / ln(1 / ε )
Eq. (5.7.23) is similar to Eq. (5.8.1) obtained earlier by [Serebryakov, 1973; Logvinovich &
Serebryakov, 1975]. Eq. (5.7.23) shows that, unlike planar cavity, the axisymmetric cavity’s
shape differs from ellipse though this difference is not great. The cavity’s aspect ratio is:
H 2 / L2с = ε + ε / ln(1/ ε ) + O(εε12 ) . (5.7.24)
Drag of cavitating flow can be found from the variation principle. Namely, the drag of the
cavitator is expressed in terms of an extreme value of energy functional [Petrov 1981]
D = πρ v∞2 L2cε 2 ( (1/ 2)ln(1/ ε ) + 1 − ln 2 + …) . (5.7.25)
The leading term of the expansion (5.7.25) was obtained in [Garabedian, 1956].
Sec. 5.8. Asymptotic Models of Axisymmetric Supercavitating Flows†

5.8.1. Brief Notes
Investigations of axisymmetric cavitating flows have been proceeding along three venues: 1)
heuristic semi-empirical approaches, 2) asymptotic theories, and 3) nonlinear numerical
approaches. The most important results of nonlinear numerical approaches have been
obtained in [Bloch, 1969a; Brennen, 1969; Guzevsky, 1979; Terentiev & Dimitrieva, 1998;
Terentiev, 2005; 2006]. Some numerical methods are presented below in Ch. 11. A wide
range of problems could be computed numerically but some problems such as flow around a
slender body should be analysed asymptotically or approximately only. Therefore, simpler
and universal approaches and dependencies are needed for practical applications.
As the minimal cavitation drag is achieved for the most slender cavities, these cases
are most important for applications. Slenderness is also an effective factor of simplification
allowing one to apply the known Slender Body Theory (SBT), see [Adams & Sears, 1953;
Frankel & Karpovich, 1948]. A number of studies [Nishiyma & Kobayshi, 1969; Vorus,
1986; Vargnezer, Kirschner & Uhlman, 2003] used steady integral-differential equations
(IDE) with the SBT for numerical calculations of cavity shape and drag. Also the known
linear theory of 2D supercavitation [Tulin, 1964a] stimulated the development of a similar
theory for an axisymmetric flow. Asymptotic and approximated statements of cavitating
flow of axisymmetric slender bodies have been given in [Serebryakov, 1973; 1986;
Yakimov, 1983a, Petrov, 1986]. The theory of slender axisymmetric cavities with gas
injection and their pulsation was developed in [Paryshev, 1978]. Asymptotic theory of 3D
perturbations of slender axisymmetric cavities was developed in [Zhuravlev, 1973, Voronin,
1983].
5.8.2 Integral-differential equations (IDE) of slender axisymmetric cavities
Let the cavity together with the cavitator is an axisymmetric slender body characterized by a
small parameter of order δ ∗ = O(2 R* / L* ) were R* is the maximal radius and L* is the
length of the body. Then the governing hydrodynamic equations may be, with an accuracy to
δ*2 ln δ* , transformed to the integral-differential equation with initial conditions at the
section of separation, x = xs , [Serebryakov, 1973]:
†
This section was prepared by Prof. Vladimir V. Serebryakov.
203
d 2 rn 2 d2R 2
⎛ d R2 ⎞
2 xs | −
1 d2 R 2 R2 dx 2
x=x1
dx 2 dx −
2 R2
⎜⎜ ⎟⎟ +
d x2
ln − ∫ |x1 − x|
1
⎝ dx ⎠ 4x(L - x) 0
(1)
(ln1/ δ*2 ) −1 (ln1/ δ*2 )−1
(5.8.1)
d2R 2 d2R 2 drn2 dR 2
L | x=x1 − | x=0 |x=L
dx 2 dx 2 dx − dx dx
− ∫ |x1 − x|
1
x
+
L-x
= 2σ ( x )
xo (1)
(ln1/ δ*2 )−1 (ln1/ δ*2 )−1
(ln1/ δ*2 ) −1
⎡ dR 2 drn2 ⎤
with initial conditions a) ⎡⎣ R = rn ( x) ⎤⎦ , ⎢ = ⎥ , b) ⎡⎣ R 2 = 0 ⎤⎦
⎣⎢ dx dx ⎦⎥ x = x
x = xo x = Lc
o
Here r, x are the cylindrical coordinates connected with the nose of the moving cavitator
r = rn ( x) . For simplicity, the location of separation, x = x0 , is assumed to be fixed. The
cavity is enclosed on some small trailing surface at x = Lc . Under terms of equation, the
order of small magnitude is indicated. The first order equation for δ* → 0 here is:
2
⎛ d R2 ⎞
1 d2 R 2
⎜ ⎟ + lnδ*2 = 2σ ( x) (5.8.2)
2 R 2 ⎜⎝ dx ⎟⎠ d x2
Similarly, the IDE for unsteady flow in the coordinate system connected with still
fluid was presented in [Serebryakov, 1977]:
∂ 2 rn 2 ∂2 R2
−
2 xs ( t ) ∂t 2 ∂t 2
1 ⎛ ∂R 2 ⎞ ∂ 2 R 2 R2 x = x1
4 [ xn (t ) − x ][ xc (t ) − x ] x ∫(t )
⎜ ⎟ + ln − dx1 −
2 R 2 ⎜⎝ ∂t ⎟⎠ ∂t 2 x1 − x
n
(5.8.3)
2 2
∂ R 2 2
∂ R ∂R 2
∂rn 2
−
∂t 2 x = x ∂t 2 dx (t ) ∂t x = xc (t ) dxn (t ) ∂t x = xn (t ) 4ΔP( x, t )
xc ( t )
− ∫ 1
dx1 − c + =
x (t )
x1 − x dt xe (t ) − x dt xn (t ) − x ρ
s
2
1 ⎛ ∂R 2 ⎞ ∂ 2 R 2 4 ΔP ( x , t )
2⎜
⎜ ⎟⎟ + lnδ*2 = (5.8.4)
2 R ⎝ ∂t ⎠ ∂t 2
ρ
Equation (5.8.3) has been obtained from (5.8.1) neglecting of infinitesimal terms of order
δ **2 ln | δ ** | , were δ ** is characteristic constant value of the small parameter δ * at some
typical time.
5.8.3. Foundations of the Asymptotic Theory
Due to low accuracy of the formal first order solutions the theory in general is developed on
the basis of second order solutions in asymptotic expansions of type:
⎡ R1 ⎤
R 2 = δ 2 ⎢ Ro + + ....⎥ (5.8.5)
⎣ ln(1 / δ 2 ) ⎦
204
The cavitator and cavity shape as a whole can generally be characterized by two
independent slenderness parameters for the cavitator, ε , and cavity, δ , and then three
typical approach have been developed:
1. Regular case for δ / ε = O(1). In this case the radiuses of the cavitator and the cavity and
their lengths are supposed to be of the same order: l / Lc = O(1) .
2. Singular case for δ / ε → 0 . In this case for cavity length Lc = O(1) the cavitator size for
σ → 0 can be very small, such as Rn = O(δ 2 ln1 / δ ) .
3. Practical approach for predicting the cavity form is developed on the base of first order
equations, such as (5.8.5) and (5.8.7), which are improved using the parameters based on
second order solutions.
In the two first cases a number of second order asymptotic solutions for different cases of
flow was developed by Serebryakov [1973; 2009].
5.8.4. Singular Steady Solutions
The most important for applications is the case of δ / ε → 0 . An approach for this case was
developed on the base of known Matched Asymptotic Expansions Method (see Sec. 5.6).
The asymptotic solutions are found separately in an inner region near the cavitator and an
outer solution at the middle part of cavity. The inner solution region is very small
O (δ 2 ln1 / δ ) as compared to the outer solution region. Matching here can be made
universally for both slender and blunted cavitator of a disc type. The first order solution is
found direct from equations (5.8.3) and (5.8.4) but for second order solution the additional
terms of asymptotic (5.7.1) should be used. This asymptotic solution with additional terms in
the Levinson & Gurevich form at Rn = 1 is found in [Serebryakov, 1986]:
R2 =
2 cdo x
[1 −
ln(4 / e)
+ ...] ≈ R 2
=
2 cdo x
[1 − +
(
1 ln ln x 1 ln e cdo 2)+ ...]
ln( x 2 / R 2 ) 2ln( x 2 / R 2 ) lnx 4 ln x 2 ln x
(5.8.6)
Integral-differential equation in the case of δ / ε → 0 , ignoring terms δ ln1 / δ 2 , is 2
simplified as follows:
d 2 R2 d 2 R2 dR 2 dR 2
2 +1 | − | |x =+1
1 ⎛d R ⎞ 2 2
d R 2
R 2
dx 2
x = x1
dx 2 dx − dx
x =− 1
dx
⎜ ⎟
2 R 2 ⎜⎝ dx ⎟⎠
+
d x2
ln − ∫
4(1 − x 2 ) -1 | x1 − x |
1
1+ x
+
1− x
=
= 2σ ( x)
(5.8.7)
The outer asymptotic solution of second order outer asymptotic solution at the given
cavity semi-length of Lk =1 was obtained by [Serebryakov, 1973]:
1 ⎡ x 2 ln 4 − ln(1 + x)(1+ x ) − ln(1 − x)(1− x ) ⎤ 2 λ
R 2 = 2 ⎢ (1 − x 2 ) + 2 ⎥, σ = 2 ln (5.8.8)
λ ⎣ ln λ ⎦ λ e
were λ=Lc/Rc is aspect ratio of the cavity, Rc - radius in the middle section. The first order
solution here defines an ellipsoidal cavity.
Eq. (5.8.8) for cavity aspect ratio λ improves the known first order dependence
(5.8.3). Figs.5.8.1 & 5.8.2 illustrate the accuracy of solution (5.8.8) as compared to nonlinear
205
numerical data. Matching the inner solution in the form of the asymptotic (5.8.6) and outer
(5.8.8) solutions made it possible to obtain the system of second order asymptotic
dependencies (5.8.9) for cavity dimensions, [Serebryakov, 1986]. Figs.5.8.1 and 5.8.2 as
well as Table 5.8.1 show a very good coinciding of asymptotic and nonlinear numerical
results.:
Fig.5.8.1. Second order dependence for Fig.5.8.2. Second order dependence for
aspect ratio λ=λ(σ): solid line — second cavity R ( x) = (1 / δ 2 ) R ( x) : solid line —
order by Eq. (5.8.8); dash line − second second order by Eq. (5.8.8); dash line −
order by Eq. (5.8.3); circles − nonlinear first order for ellipsoidal cavity; crosses −
numerical solution. nonlinear numerical solution.
2ln λ / m e cd ⎡ ln 2 / e ⎤ cd ln λ 2 / m 2 ⎡ ln e / 2 ⎤
σ= , Rk2 = Rn2⎢1 + 2 ⎥, Lk = Rn ⎢1− 2 ⎥
.(5.8.9)
λ 2
σ ⎣ ln λ 2 / m 2 ⎦ σ ⎣ ln λ / m ⎦
2
Conclusion: Solutions (5.8.12) are

applicable both for a cavity behind a Table 5.8.1. Сomparison asymptotical (5.8.2)
slender cavitator and for a cavity behind a and numerical results in case of disc cavitator,
disk-type cavitator. As can be found from σ 0.03 0.04 0.05
comparing the results in Fig.5.8.1 & 5.8.2, λ (asymptotic) 11.327 9.291 7.924
the first order solutions are very good for λ (numerical) 11.456 9.453 8.132
approximating the cavity shape, while the
Rk (asymptotic) 5.537 4.834 4.357
second order solutions are very good for Rk (numerical) 5.5544 4.8448 4.3624
calculating cavity dimensions. Lk (asymptotic) 63.497 45.583 36.109
(numerical) 63.5181 45.8 35.4731
5.8.5 The Simplest Physical Model of Axisymmetric Supercavitating Flows

The process of creating a slender axisymmetric cavity can be explained using a simple
model of radial flow expansion, Fig.5.8.3, expressed by equations (5.8.6).
For slender cavities with a small cavitator, the drag is practically independent of the
cavity shape and the cavity is independent of the cavitator shape; it is defined by the
cavitator’s drag only. A moving cavitator pushes motionless fluid to asides and its work is
transformed into kinetic energy of mainly radial near-cavity flow at each motionless section
206
which cavitator passed. The real zone of perturbations, containing the main part of energy
and pulse of flow, is concentrated in the finite
region limited by surface r=ψ(x,t), which is wider
than the cavitator’s half-length, and the cavity
surface (here x is the axis of coordinates
connected with the still fluid).
Fig.5.8.3.
Radial flow model.
Further expansion of the cavity section

together with the near-cavity radial flow is going
by inertia due to pressure gradient in the
undisturbed flow and cavity. The expansion depends mildly on surface r=ψ(x,t). The cavity
section reaches the max radius at the middle and then decreases due to the outer pressure.
Decreasing the cavity section is finished in the back unstable zone of the cavity with chaotic
flow, and all radial flow energy is transformed into the energy of the wake behind cavity, as
seen in Fig.5.8.3. Some elements of this simple model were proposed by G. Birkhoff along
with R. Isaacs [Birkhoff & Zarantonello, 1957]. An updated variant of this model has been
developed using SBT and is currently the base for practical estimates of cavitating flows.
The system of the simplest equations for predicting a slender steady axisymmetric cavity
based on first order equation (5.8.5) was given in [Serebryakov, 1976]:
d 2 R2 dR 2 D
μ 2
+ σ ( x ) = 0 , = 2 2
, R2 = 0 . (5.8.10)
dx dx k πμρU x=0
x=0 ∞
Here D is the cavitator drag, the sizes of cavitator and inner solutions region to order
O (δ 2 ln1/ δ ) are neglected. These equations for constant cavitation number define an
ellipsoidal cavity and the known typical dependencies as:
2cd σ 2 c c 2μ / k 2μ
R2 = x− x , Rk = Rn d , Lk = Rn d , λ2 = . (5.8.11)
kμ 2μ kσ σ σ
where Rk is maximal radius and Lk is semi-length of cavity.
Eq. (5.8.10) at σ = const can also be presented in the form of an equation for total kinetic
and potential energy conservation at each section of the cavity:
2
μρ ⎛ dR 2 ⎞
Ek x + E px = kπ U ∞2
⎜⎜
2
⎟⎟ + π R ΔP = kD
4 ⎝ dx ⎠
The key idea for obtaining system of equations (5.8.13) is as follows. We
approximate the cavity shape by the first order equation (5.8.5), and improve these equations
introducing coefficients μ, k based on more accurate second order dependencies. This can
be done using integral differential equations (5.8.10) for steady slender cavities. The
coefficients μ, k have clean physical meanings. Value of μ characterizes inertial properties
of the cavity sections like a kind of added mass. Coefficient k characterizes small
longitudinal energy transfer along the cavity sections.
For cavities, which are moderately different from a cavity at σ=const, these values
can be defined based on steady second order asymptotic dependencies (5.8.9):
207
2μ 2 λ 2 ln 2 / σ λ ln 2 / σ
σ= = ln ≈ ln → μ = ln ln , μ λ ~12 ~ 2 (5.8.12)
λ 2
λ 2
e λ 2
eσ e eσ
ln 2 / e 2 ln 2 / e
k =1− 2 2 2
≈1− , k λ ~7 −15 ~ 0.92 − 0.94 (5.8.13)
ln λ / β ln 4 / β 2σ
The results of calculating μ and k are shown in Fig.5.8.4 & Fig.5.8.5. A precise version of
the system of equations based on the known formula for cavitator drag D = cd π Rn2 ρU ∞2 / 2
d 2 R2 dR 2 2(cd − kσ )
is μc + σ = 0 , = Rn , R2 = Rn2 , (5.8.14)
dx 2 dx x =0
k μc x =0
where the value of μc = μ cd /(cd − kσ ) is a small correction for the cavitator’s finite-size.
The solution for a cavity with σ = const and dependencies for cavity maximal
radius Rm and length Lc are:
2(cd − kσ) σ 2 2μc
R 2 = R 2n + R n
kμ c
x−
2μc
c
kσ
R
x , R m = R n d , Lc = n
σ k
( )
cd − kσ + cd .(5.8.15)
These equations and solutions can be used both for slender cavitators and for cavities behind
Fig.5.8.4 Value of σ vs. λ for a Fig.5.8.5 Value of k vs. Rk for a

steady cavity at σ=const: solid line steady cavity at σ=const: dashed
- second order by Eq. (5.8.15); line – experimental data for k
dashed line - 10% of μ inclination; [Logvinovich, 1969]; solid line –
dots - μ=σλ2 by the nonlinear second order by Eq (5.8.16);
numerical solution. triangles - nonlinear numerical.
the disc-type cavitators. For predicting the disc-type cavitator, the known dependence
cd = cdo (1 + σ ) is applied, for disk cdo ~ 0.82 − 83 . For slender cavitator cd = cdo + σ . For
calculating cdo, e.g. for a slender cone with semi angle γ 0 and ε=tanγ, nonlinear asymptotic
⎧ 2⎫
equations cdo = 2 ⎨1 − (1 − ε 2 ) ⎡1 − (ε 2 / 4) ln(4 / ε 2 ) ⎤ ⎬ can be used [Taic, 1985].
⎩ ⎣ ⎦ ⎭
For slender cavitators Eqs. (5.8.15) and (5.8.16) are applicable up to the cone’s semi angle of
γ=4°−5°. Applicability and accuracy of Eqs (5.8.14) at σ = const behind the disc are
illustrated in Fig.5.8.6 and 5.8.7.
208
Fig.5.8.6. Solid line – accuracy of

predicting an ellipsoidal cavity at σ=0.04: Fig.5.8.7. Form of the cavity’s forward
triangles – numerical; dashed line – part: triangles – numerical at σ=0.04;
ellipsoidal cavity by Eq. (5.8.18). solid line – by Eq. (5.8.18) at σ=0.04;
dashed line –by Eq. (5.8.18) at σ=0.05.
5.8.6 Unsteady Cavities – A Practical Approach
One of the most important properties of axisymmetric cavitation flows in the case of slender
cavities is that the cavity section expansion is essentially independent of the expansion of
neighboring cavity sections, cavitator and the cavity shape as a whole. This property was
discovered long ago experimentally by [Reinhardt, 1946]. Based on numerous experimental
data [Logvinovich, 1959a] formulated the principle of independence of the cavity
expansion. More description and application of this principle are given in Ch.8. It should
noted that the principle gives sufficiently good coincidence with experiments for steady as
well as unsteady cavitating flows and also for the entry of body [Abelson, 1970; Zhuravlev,
1973]. Like for steady cavity, the simplest unsteady equation system based on first order
equation (5.8.4) was developed in [Logvinovich and Serebryakov, 1975]:
∂ 2 R 2 2ΔP ( x, t )
μc 2 + = 0,
∂t ρ
∂R 2 2 [ cd ( x) − kσ ( x) ]
= Rn ( x)U ( x) , (5.8.16)
∂t k μc
t =t n ( x )
R2 = Rn2 ( x).
t =t n ( x )
In the most general case the cavitator can have variable size Rn (t ) and its base moves
unsteadily due to low, x = xn (t ) , and so the drag coefficient is depends on the time, cd (t ) .
Let the base of cavitator intersects the fixed abscissa x at time tn which may by found from
equation x = xn (tn ) . At the time, tn , the vertical speed of the point of cavitator and its
ordinate may be obtained from given of cavitator boundary and its moving. Just these values
can be used as starting condition for different equation of second degree (5.8.16). Its solution
in general case Rn (t ) , cd (t ) , U (t ) , ΔP ( x, t ) is defined by expression:
209
t t
cd ( x) − kσ ( x) 2
2
R = Rn2 ( x) + 2 Rn ( x)U ( x) [ t − tn ( x ) ] − ∫ ∫ ΔP ( x, t )dtdt (5.8.17)
2k μ ρμ tn ( x ) tn ( x )
In the particular case of ΔP = ΔP( x ) the universal integral exists:
cd ( x) − kσ ( x) ΔP ( x )
R 2 = Rn2 ( x) + 2 Rn ( x)U ( x) [ t − tn ( x ) ] − [t − tn ( x)]2 (5.8.18)
2k μ ρμ ( x)
This integral for U , ΔP = const define ellipsoidal cavity and includes really all obtained
earlier by traditional way simplest solutions on the base of independence principle. With
account of week change of the values of μ , k simplest variant for estimation of cavities,
which are not very strongly different as compared ordinary steady cavity, can apply these
values in the range of σ ~ 0.05 ÷ 0.02 as universal constants μ ~ 2, k ~0.93-0.96 . More
accurate solutions can be obtained for using of these values as functions μ ( x ), k ( x ) .
Simplest case is when these values are calculated depend on σ ( x) with help of steady
dependencies (5.8.12), (5.8.13) at the moment when cavitator pass the section of motionless
fluid. Application of μ ( x ), k ( x ) is some averaging of this values along time for each
cavity section. More precisely construction of the values for μ ( x , t ), k ( x , t ) on the base of
linearized theory give the possibility to obtain more accurate approximation including
solutions in near disc zone. Elementary solutions (5.8.17)-(5.8.18) give the possibility to
analyze typical cavities form under different cases of flow. For calculations used for
Fig.5.8.8 and 5.8.9 on the base of solutions (5.8.17)-(5.8.18) the values of μ , k for first
approximation are accepted as for σ = σ o ~ 0.04 . Cavity form is presented finally in the
coordinate system connected with cavitator. Acceleration action is calculated for the case,
when accelerated and decelerated motion is started from the same speed under not changing
of other parameters.
Fig.5.8.8. Change of the cavity shape Fig.5.8.9. Effect of alternating pressure on

under acceleration (solid line, a>0) and shape of unsteady cavity: solid line – cavity
deceleration (solid line, a<0) of the shape due to harmonic pressure
cavitator speed; dashed line – undisturbed oscillations; dashed line – undisturbed
cavity at σ o = 0.04 . cavity at σ o = 0.04 .
Fig.5.8.8 illustrates the cavity shape at the instant when the cavitator reaches the
same size as the length of the undisturbed cavity at the initial moment. Similarly the cavity
under alternating pressure are estimated by equations (5.8.17). Fig.5.8.9 illustrates the effect
of harmonic pressure alternations in the cavity. Similar shapes have been observed, in
particular, in ventilated axisymmetric cavity oscillations [Paryshev, 1978]. The main limits
210
to apply these equations is conditions U *T* L* = O (1) for typical values of speed, cavity
length, time duration of this process. Nevertheless these equations have been working,
though roughly, even for the cases of sharp changes of speed, Fig.5.8.8, drag and cavitator’s
dimensions.
Fig.5.8.10 shows a more general case of water entry than the axisymmetric flow
presented above, in this case a round-nosed body entering the water at an oblique angle.
Modeling of such practical cases would involve more than the simple prediction of the
cavity shape, since the time-dependent forces on the full body govern its trajectory in six
degrees of freedom. The figure highlights the importance of afterbody interaction with the
cavity boundary, which produces the scar visible on the lower surface near the center of the
image. Such a problem is highly nonlinear, since the drag and lift on the cavitator and the
cavitator trajectory determine the cavity shape, which in turn has a strong influence on the
afterbody forces. The case in the figure is subject to yet another nonlinearity, namely the
physics of cavity detachment from the rounded nose, an effect that is discussed in Ch. 11.
This figure highlights the close relationship between cavitation and water entry. An excellent
resource for further information on that topic, including the complicated trajectories that are
often observed, is the well-known report by [May, 1975].
Fig.5.8.10.
Water entry of a round-
nosed body at an
oblique angle.(Image
courtesy of California
Institute of Technology
and C.E. Brennen).
In conclusion
one should note that
equations (5.8.19) – (5.8.22) have been many time compared with experiment data as well as
with numerical results. Of course, the principle of independence would not be considered as
exact law but approximate only.
211
Ch. 6. Unsteady Cavitating Flows

Sec. 6.1. Unsteady Movement of a Slender Hydrofoil
6.1.1. Introduction
In the theory of non-stationary movement of thin bodies there are applied mostly two
methods based on determination of a complex velocity and a function of pressure,
respectively. Application of the first method entails necessity of accounting free vortices past
a hydrofoil, which form a line of speed discontinuities [Sedov, 1950]. By determining a
pressure function one can avoid such necessity, since the pressure is continuous at the
transition of discontinuity line. The function of pressure, a.k.a. potential of acceleration, has
recently received wide application in connection with non-stationary problems of continuous
and cavitating flows [Panchenkov, 1975; Ivchenko, 1966].
However, it should be noted, that a majority of works is devoted to research of non-
stationary movement of a body in a stream having a constant speed at infinity. Acceleration
potential θ = − p / ρ v∞2 is connected with speed potential by the linearized equations:
1 ∂ ∂
∇θ = N ∇ϕ , N = − .
v∞ ∂t ∂x
In a planar flow the acceleration potential is determined as a real part of analitic
function by methods of the theory of analytic functions of complex variable. But for the full
solution of problem, it is necessary to find inverse operator N −1 , which connects
acceleration potential and speed potential. Determining the inverse operator is a
mathematical challenge. The kind of operator and the inverse operator essentially depend on
the choice of coordinate systems. As shown in [Terentiev, 1981b], a statement problem and
its solution become considerably simpler if the coordinate system is fixed (absolute system).
Plane problem statement of unsteady movement of thin bodies is considered below, as well
as solutions of some problems.
Problem statement can be illustrated using a
specific example of movement of a thin
body under a free surface above a solid
bottom, Fig.6.1.1a. As before, the fluid is
assumed to be incompressible and
weightless, and flow is potential. Fig.6.1.1. (a) - movement of foil;
Let's consider the moving ( x′, y ′ ) (b) - domain of solution.
and fixed (x, y) systems of coordinates
connected by relationship z ′ = z − s (t ) , where s (t ) is a running distance to the body. Let
function y = f ( x ', t ) be the equation of the moving body relative to the moving system of
coordinates x′, y ′ ; in the fixed system, the body movement is described by equation
y = f ( x − s, t ) . It is assumed that partial derivatives ∂f / ∂x′ and ∂f / ∂t are infinitesimal of
the same order. With these assumptions, partial derivatives of speed potential are small of
first order, and the pressure expressed from Couchi – Lagrange integral p = − ρ ∂ϕ / ∂t are
212
smalls of the same order. Here the pressure is assumed to be difference p − p∞ . Since the
pressure is determined only by partial derivative with respect to time t, it is desirable to
obtain harmonic function ϕt ( x, y, t ) or analitic function wt ( z, t ) = ϕt + iψ t on the domain,
Fig.6.1.1b, bounded with a split corresponding to a thin hydrofoil γ , and with a strait line
corresponding to the free boundary γ 1 and the given curvilinear bottom γ 2 .
Let's begin with proving the following theorem:
Let function F ( x, y , t ) in domain G with boundary γ is determined; its partial derivatives
are infinitesimals of the same order as ε . If boundary γ is determined by function
y = g ( x, t ) and their partial derivatives are infinitesimals also of the same order as ε , then
the boundary value of the derivatives is equal to derivatives of boundary value of function
F ( x, y , t ) , i.e. the following equalities are valid in the linear theory:
⎧ ∂F ⎫ ∂ { F }γ ⎧ ∂F ⎫ ∂ { F }γ
⎨ ⎬ = , ⎨ ⎬ = . (6.1.1)
⎩ ∂x ⎭γ ∂x ⎩ ∂t ⎭γ ∂t
Proof: Since the boundary value of the function is { F }γ = F ( x, g ( x, t ), t ) then
∂ { F }γ ⎧ ∂F ⎫ ⎧ ∂F ⎫ ∂g ∂ { F }γ ⎧ ∂F ⎫ ⎧ ∂F ⎫ ∂g
=⎨ ⎬ +⎨ ⎬ , =⎨ ⎬ +⎨ ⎬ .
∂x ⎩ ∂x ⎭γ ⎩ ∂y ⎭γ ∂x ∂t ⎩ ∂t ⎭γ ⎩ ∂y ⎭γ ∂t
The second term of both equalities is infinitesimal of second order, the same as ε 2 , so that
they should be ignored. The theorem is proven.
Due to this theorem it is easy to write boundary conditions for partial derivative
wt ( z, t ) . The normal derivative of speed potential, or curvilinear derivative of stream
function, vanishes on fixed solid boundary γ 2 , i.e. ∂ψ / ∂s = 0 . In fact the stream function is
determined as an arbitrary function of time. But if the fluid is undisturbed at infinity then
stream function should be constant, and the partial time derivative vanishes.
Now, the boundary conditions look as following:
• the kinematic condition at fixed boundary γ 2 , ψ t = 0, z ∈ γ 2 ; (6.1.2)
• the dynamic condition at free boundary γ 1 ,
ϕt = ( p0 − p) / ρ = 0, z ∈ γ 1 ; (6.1.3)
• the kinematic condition at the moving boundary γ ,
∂f ( x′, t ) ∂f ( x′, t )
ψ x = −ϕ y = − yt = − +s . (6.1.4)
∂t ∂x′
x−s
Whence the streams function is ψ = T (t ) − ∫
0
yt (x′, t )dx′ , (6.1.5)
where T(t) is an arbitrary function. Due to the second equality in (6.1.1), the kinematic
x−s
condition can be written as ψ t = T (t ) + syt ( x − s, t ) − ∫
0
ytt ( x′, t ) dx′ , (6.1.6)
where T = dT / dt , s = ds / dt = U (t ) is the speed of a hydrofoil.

If function T(t) is given, then conditions (6.1.4) – (6.1.6) determine the partial
derivative of complex function wt ( z , t ) in the same way as ω ( z , t ) . Generally, this function is
213
to be found. The value problem solution, as well as function ω ( z ) , have singularity at the
leading edge, and then the stream function can have discontinuity at that point. Therefore,
the continuity condition at the leading edge of a moving hydrofoil should be satisfied
lim ψ ( x, t ) = T (t ) . (6.1.7)
x→s+0
Imaginary part of function wz ( z , t ) on the x-axis beyond the split could be presented as
ψ t = F ( x, t ) + T (t )[1 + K ( x, t )] .
Integrating with respect to time in interval (0, t) and satisfying condition (6.1.7), one
can obtain an integral equation as Volterra form of first kind:
t t
∫ T (τ ) K (s(t ),τ )dτ = − ∫ F (s(t ),τ )dτ −ψ 0 (s(t ),0) ,

0 0
(6.1.8)
where ψ 0 ( s (t ),0) is the initial stream function in a flow domain if the fluid was disturbed at
the beginning.
Only some integral equations can be obtained analytically. Generally, it should be
solved numerically, in particular, iteratively [Terentiev, 1979].
6.1.3. Iteration Algorithm
t
Let the integral equation be as ∫ K (t ,τ )μ (τ )dτ = f (t ) .
0
(6.1.9)
Dividing interval (0, t) into 2n+1 pieces, we can calculate numerically on each segment of
length 2h = 2t /(2n + 1) by using average value of function μ . Then the average values can
be obtained consecutively by recurrence formula:
1 ⎛ f 2 m m −1 ⎞ f2
μ2 m −1 = ⎜ − ∑ μ 2 k −1 K 2 m ,2 k −1 ⎟ , μ1 = , m = 1, n + 1 (6.1.10)
K 2 m ,2 m −1 ⎝ 2h k =1 ⎠ 2hK 2,1
t
where μ2 m −1 = μ ((2m − 1)h), f 2 m = f (2mh), K 2 m,2 k −1 = K (2mh, (2k − 1)h), h = .
2n + 1
The nth iteration μ ( n ) = μ2 n +1 is calculated from (6.1.10) if m = n + 1 . The order as
n = 1, 3, 7, 15, … can be used to satisfy the required accuracy | ( μ ( n ) − μ ( n −1) ) / μ ( n ) |≤ δ .

Denoting the pressure on the lower and upper boundaries by subscripts 1 and 2, and
integrating the pressure along hydrofoil, one can obtain the lift and the torque as follows:
0
L = ∫ ( p1 − p2 )dx′ = − ρ Re ∫ wt ( z , t ) dz , (6.1.11)
−l γ
0
and M = ∫ ( p1 − p2 ) x′dx′ = − ρ Re ∫ wt ( z , t )( z − s )dz . (6.1.12)
−l γ
The drag is a sum of a sucking force applied at the leading edge due to singularity and a
longitudinal force applied at the hydrofoil boundary:
X = X1 + X 2 , (6.1.13)
214
iρ
where first term is calculated by Sedov’s formula [Sedov, 1950] X 1 = ∫
2 z =s
wz2 ( z , t ) dz .
Since the leading parts of derivatives wt ( z, t ) and wz ( z, t ) at singular edge differ from each
other by factor − s then this force is calculated as
iρ
X 1 = 2 ∫ wt2 ( z , t )dz . (6.1.14)
2s z = s
The second term of longitudinal force can be calculated from integral
0
⎡⎛ ∂y ⎞ ⎛ ∂y ⎞ ⎤
X 2 = ρ ∫ ⎢⎜ ϕt ⎟ − ⎜ ϕt ′ ⎟ ⎥ dx′ . (6.1.15)
−l ⎣ ⎝ ∂x ′ ⎠ 2 ⎝ ∂x ⎠1 ⎦
Statement and expressions (6.1.11) – (6.1.15) show that the value problem for a
partial derivative of complex potential wt(z,t) and its solution are similar to the value
problem for small perturbation complex velocity ω ( z ) of steady flow; moreover, it is not
necessity to account the trailing free vortices past hydrofoil. All these make function wt ( z , t )
more preferable then complex velocity ω ( z , t ) used in unsteady flows [Sedov 1950].
The value problem of unsteady movement of a hydrofoil with deformable boundary
in an unbounded domain is determined by condition (6.1.6). Function ψ t ( x) has generally
different values at each side of cut but the same function T(t). Following Sedov’s solution
(5.2.14) - (5.2.15), function wt(z,t) could be presented as a sum of two functions, one that
determines the longitudinal motion of a symmetrical deformable body, and the other
determining the motion of a curvilinear deformable arc. The first function is independent of
function T(t) and can be determined by integral (5.2,14). The unknown function T(t) enters
in the second function, which determines a motion of curvilinear arc only and could be
found from equation (5.2.15). As an example, a simple problem of pulsating movement of a
flat plate is considered below.
6.1.5. Wagner’s Problem
Let the speed of the moving flat plat be given as
⎧U1 = const , t < 0,
s (t ) = ⎨ (6.1.16)
⎩U 2 = const , t ≥ 0.
It is assumed that the plate inclination angle α is small, and the plate length and density are
equal to unity. According to equation (6.1.16) the flow, depending on the plate position, is
determined by two functions: w1 ( z ) at t < 0 and w2 ( z , t ) at t > 0 . The first function
determines a steady flow relative to moving coordinates ( x′, y′ ) and satisfies kinematic
condition ( ∂ψ 1 ( x′) / ∂y′ = −αU1 ), so that the complex velocity is expressed as
dw1 ( z ′) ⎛ z′ + 1 ⎞
= −iαU1 ⎜⎜ 1 − ⎟, t < 0 . (6.1.17)
dz ′ ⎝ z ′ ⎟⎠
whence, (
w1 ( z ′) = −iαU1 z ′ − z ′( z ′ + 1) − ln ( ))
z′ + z′ + 1 . (6.1.18)
Besides, the plate changes its speed suddenly on U 2 − U1 so that another potential w0 ( z )
should be considered satisfying condition ∂ϕ0 / ∂y = −α (U 2 − U1 ) . It would be expressed
215
similarly to (6.1.18) but without the logarithmic term because the flow is symmetrical about
the middle of the plate. Now, the initial stream function is presented as
ψ 2 ( x, 0) = ψ 1 ( x, 0) + ψ 0 ( x, 0) = αU 2 ( x − x( x + 1)) − αU1 ln( x + x + 1) . (6.1.19)
Kinematic condition (6.1.6) changes to (ψ 2t = T − αU 22 ), so that partial derivative is
⎛ z − s +1 ⎞
expressed similarly to (6.1.17) as w2t ( z , t ) = i (T − aU 22 ) ⎜⎜ 1 − ⎟, t > 0 . (6.1.20)
⎝ z − s ⎟⎠
Integrating with respect to time τ and assuming z = x > s in interval (0, t), we have:
x − s (τ ) + 1
t
ψ 2 ( x, t ) = T (t ) − αU 2 s(t ) − ∫ (T (τ ) − αU 22 ) dτ + ψ 2 ( x,0) .
0
x − s (τ )
Substituting x = s, s(τ ) = s′ and equating to the value of stream function at the plate
ψ 2 ( x) = T − αU 2 ( x − s′) , one can obtain integral equation in the form
s
1+ s − s' U
∫
0
s − s'
β ( s ') ds ' = s ( s + 1) + 1 ln( s + s + 1) ,
U2
(6.1.21)
where β ( s ) = 1 − T (t ) / αU 22 . The last term in (6.1.21) can be excluded and the integral
s
1+ s − s'
equation becomes ∫
0
s − s'
μ ( s ')ds ' = s( s + 1) , (6.1.22)
U1 ⎛ U1 ⎞
where β ( s ') =
+ ⎜1 − ⎟ μ ( s′) .
U2 ⎝ U2 ⎠
Coefficients of lift, torque, and longitudinal force are presented, respectively,
πα
CL = 2παβ ( s ), Cm = β ( s ), Cx = 2πα 2 β ( s )[ β ( s ) − 1] . (6.1.23)
2
All special cases can be obtained from (6.1.22) and (6.1.23): Wagner’s flow
( U1 = 0 ), a steady movement of a foil ( U1 = U 2 ), impact problem
( s → 0, β = (U1 + U 2 ) / U 2 ), and a steady movement again ( s → ∞, β → 1 ).
Mikhailov [Terentiev & Mikhailov, 1979] gave an asymptotic expansion in power
series on τ = s /(1 + s ) for a solution of equation (6.1.22) as
1 1 1 ⎛1 1 ⎞ ⎛ 1 5 ⎞ 4
μ (s) =
+ τ + τ 2 + ⎜ + ⎟τ 3 + ⎜ + ⎟τ + O(τ ) .
5
2 4 8 ⎝ 16 96 ⎠ ⎝ 32 384 ⎠
On other hand an expansion of the rational fraction is
1+ s 1 1 1 1 1
= + τ + τ 2 + τ 3 + τ 4 + O(τ 5 ) .
2+s 2 4 8 16 32
One may notice that coefficients of both power expansions differ slightly and have
the same limit, 1, as distance s approaches to infinity, and so the solution of equation
1+ s
(6.1.22) could be well approximated as the fraction μ (s) ≈ . (6.1.24)
2+s
The integral equation could also be solved numerically by iteration (6.1.10) but
because of singularity at s′ = s the iteration will convergence slow. The singularity of the
integrand could be eliminated multiplying by ( s1 − s ) −1/ 2 then integrating with respect to s in
216
interval (0, s1 ) . Integral equation (6.1.22) is transformed to

s
1
∫ 1 + s − xE ( k1 ) μ ( x)dx =1 + s [ (1 + 2 s ) E (k2 ) − K (k2 ) ] , (6.1.25)
0
3
where E and K are full elliptic integrals of first and second kinds accordingly, and
s−x s
k1 = , k2 = .
1+ s − x 1+ s
A comparison of results calculated using approximation formula (6.1.24) with
numerical solution calculated from (6.1.25) iteratively (6.1.10) shows a good agreement, for
instance the iteration from (6.1.25) gives μ = 0.500, 669, 0.857 0.989 for distances s = 0, 1,
5, 50, while approximated formula (6.1.24) for the same distances results in μ =0.500,
0.667, 0.857, 0.981, respectively. Numerical results are plotted in Fig.6.1.2. The curves in
Fig.6.1.2a show a dependence of lift coefficient CL = 2 L / ρU 22lCL0 on distance s at different
initial speeds U1 . The curves in Fig.6.1.2b characterize drag D = − X relative to D0 = L0α .
As expected, curves CL have the same asymptote CL (∞) = 1 , while curves CD have
asymptote CD (∞) = 0 .
Fig.6.1.2. Wagner’s problem: lift CL = L / παρU 22 l (a) and drag CD = D / πα 2 ρU 22 l (b)

coefficient as functions of distance s at different initial speeds U1 ( U = U1 / U 2 ).
6.1.6. Perturbation Velocity Method

For comparison let’s consider the integral equation of Wagner problem obtained using
vortex layer and complex velocity of perturbation flow [Sedov, 1950]:
S − s′ + 1
S
∫0 S − s′ γ (s′)ds′ = π U α , (6.1.26)
where γ=ϕx2 – ϕx1 is intensity of free vortex layer past a foil.

1
s
γ ( s′)ds′
The lift coefficient is CL ( s ) = 2πα − ∫ . (6.1.27)
U 0 (1 + s − s′)( s − s′)
217
Integral equation (6.1.26) can be transformed to equation (6.1.22) multiplying by
[( s − S )( s − S + 1)]−1/ 2 and integrating with respect to S in interval (0, s), and denoting,
U
s
γ ( s′)ds′
μ ( s ) = CL ( s ) / 2πα , T =
2π ∫
(1 + s − s′)( s − s′)
0
.
It is evident that the integral equations in both manners are of the same kind but the
partial time-derivative wt ( z , t ) has some advantages, namely, the problem statement and
solving the integral equation, as well as calculating hydrodynamic characteristics are simpler
than when using complex velocity wz - there is no need to calculate extra integrals like
(6.1.27). Particularly, an advantage of function wt ( z , t ) is in using it for a fluid with free
boundaries which will be considered in the next sections.
6.1.7. Motion of Plate with Arbitrary Speed
⎧U = const , t < 0,
Let a plate moves at given speed s (t ) = ⎨ 1 (6.1.28)
⎩U (t ), t ≥ 0.
Boundary conditions for a stream function and its time derivative are obtained from (6.1.5)
and (6.1.6) ψ ( x, y, t ) = T (t ) + αU (t )( x − s), x ∈ [ s − 1, s], y = 0, (6.1.29)
ψ t ( x, y, t ) = T (t ) − αU 2 (t ) + αU (t )[ x − s(t )], x ∈ [ s − 1, s ], y = 0 .
(6.1.30)
The value problem solution satisfying the last condition at both sides of cut
( x ∈ [ s − 1, s ], y = 0 ) is expressed by function
⎛ z − s +1 ⎞ ⎛ z − s +1 ⎞
wt ( z , t ) = i (T − αU 2 ) ⎜⎜1 − ⎟⎟ + iαU ⎜⎜ z − s − ( z − s − 1/ 2) ⎟. (6.1.31)
⎝ z−s ⎠ ⎝ z − s ⎟⎠
Integrating the imaginary part of this function at z = x > s with respect to time and
equating z = s = s (t ) to T (t ) , one obtains an integral equation
1 + s − s′
t
1
∫ [αU (τ ) + αU (τ ) − T (τ ) − αU1U (τ )] dτ =
2
0
2 s − s′
t
(6.1.32)
= α (U (0) − U1 ) s ( s + 1) + α ∫ U (τ ) (1 + s − s′)( s − s′)dτ .
0
U ⎛1 ⎞ T
Denoting β ( s ) = U − U1 +
⎜ + s⎟ − , (6.1.33)
U ⎝2 ⎠ αU
or β ( s ) = (U (0) − U1 ) β1 ( s) + β 2 ( s) , (6.1.34)
one obtains two integral equations, the one for β1 (s) coincides with equation (6.1.22),
1 + s − s′ U ( s′) 1 + s − s′
s s
another is expressed as ∫
0
s − s′
β 2 ( s′)ds′ = s ∫
0
U ( s′) s − s′
ds′ . (6.1.35)

Hydrodynamic characteristics are calculated from (6.1.11) – (6.1.15) as following:
⎛ 3 ⎞ ⎛ 1 + 4s ⎞
• lift L = πρ ⎜ α s 2 − T (t ) + α s ⎟ = πρα ⎜ U [ β ( s ) + U1 ] + U ⎟; (6.1.36)
⎝ 4 ⎠ ⎝ 4 ⎠
218
πρ ⎛ 1 ⎞ πρα
• torque M = ⎜α s − T + α s ⎟ =
4 ⎝
2
2 ⎠ 4
(U [β (s) + U1 ] − sU ) ; (6.1.37)
• drag
πρ ⎛ αs⎞
2 ⎧⎪ ⎡ U⎤
2
1 − 4s ⎫⎪
X = 2 ⎜α s − T + ⎟ − α L = πρα ⎨ ⎢ β ( s ) + U1 − s ⎥ − U [ β ( s ) + U1 ] − U
2 2
⎬ (6.1.38)
s ⎝ 2 ⎠ ⎪⎩ ⎣ U⎦ 4 ⎪
⎭
Let us consider a special case of moving with constant acceleration ( s = U = b ) from
initial static state ( U (0) = 0, U1 = 0 ). In this case speed s is equal to 2bs . Denoting
β = β 2 = 2bγ ( s) , the integral equation (6.1.35) is transformed to
s
1 + s − s′ ⎛ s ⎞
∫ s − s′
γ ( s′)ds′ = s 1 + sE ⎜⎜ ⎟⎟ . (6.1.39)
0 ⎝ 1+ s ⎠
The solution of the integral equation can be presented as asymptotic expansion
⎛ τ τ2 τ3 ⎞ s
γ ( s ) = s ⎜1 + + + + O(τ 4 ) ⎟ , τ = . (6.1.40)
⎝ 6 10 15 ⎠ 1+ s
For numerical calculation, integral equation (6.1.39) as above should be transformed to
s
⎛ s−x ⎞
∫0 1 + s − xE ⎜⎜ 1 + s − x ⎟⎟ γ ( x)dx = f (s) , (6.1.41)
⎝ ⎠
π⎧ ⎫
where f (s) =
4⎩
⎛1 3 ⎞ ⎛1 1⎞
(
⎨ s ( s + 1) ⎜ + s ⎟ + ⎜ s − ⎟ ln s + s + 1 ⎬ .
⎝8 4 ⎠ ⎝2 8⎠ ⎭
)
Hydrodynamic characteristics (6.1.36) – (6.1.38) are expressed as following:
⎛
⎝
1⎞
4⎠
1
(
L = πρα b ⎜ 2 sγ ( s ) − s + ⎟ , M = πρα b 2 sγ ( s ) − s ,
4
)
(6.1.42)
⎛ ⎡ 2 s 1 ⎞
X = πρα b ⎜ 2 ⎣γ ( s ) − s ⎤⎦ − − ⎟ .
2
⎝ 2 4⎠
As follows from (6.1.40), initial value of solution vanishes γ (0) = 0 while the initial
1
values of lift and drag do not vanish and are equal to L(0) = πρα b, X (0) = −α L(0) .
4
Some results of numerical calculation are plotted in Fig.6.1.3, where
L = L / πρα bl , M = M / πρα bl 2 , D = − X / πρα bl
CL = 2 L / παρ s 2 l , CM = 2 M / παρ s 2 l 2 , CD = −2 X / πα 2 ρ s 2l .
6.1.9. Harmonic Oscillation of a Thin Hydrofoil

Consider motion of an oscillating hydrofoil from an initial static state. It should be noted that
the oscillating foil has been usually considered assuming quasi-stationary movement
[Keldysh & Lavrent’ev, 1935; Nekrasov, 1947; Panchenkov, 1975; Gorelov, 1975]. Let the
oscillating of hydrofoil is given as y = Re ( F ( x′)e jωt ) , x′ ∈ [ − a, a ] , (6.1.43)
219
where j = −1 is imaginary unity untied with another imaginary unit I; ω is oscillation

frequency; 2a is the length of foil. Further, as usually, we shall drop symbol “Re” and take it
into account only in the final expressions.
Fig.6.1.3. Uniformly accelerated motion of plate.
Denote the Strouhal number as λ = ω a / U 0 , and dimensionless time as τ = U 0t / a ;

U 0 is the initial speed. Integral equation (6.1.8) is given in the form [Terentiev & Mikhailov,
τ τ
2 +τ −τ ′ (1 + τ − τ ′)e jλτ ′
1979] ∫
0
τ −τ ′
q (τ ′)dτ ′ = ∫
0 (τ − τ ′)(2 + τ − τ ′)
dτ ′ , (6.1.44)
where unknown function q(τ ) contains function T (t ) and some other expressions which
appear in this problem. As before, hydrodynamic characteristics can be obtained integrating
the pressure. Namely, for horizontal flat plate, coefficients of lift, torque relative to the plate
center, and drag are determined as
L ⎡λ2 ⎤ π
= 2πε ⎢ e jλτ − jλ q(τ ) ⎥ , CM = − ε jλ q(τ ), CD = 2πε 2 [ jλ q(τ )] . (6.1.45)
2
CL =
ρU 0 a
2
⎣2 ⎦ 2
Coefficient CD is obtained as a concentrated force at the leading edge so that its sign
is always positive and the longitudinal force is a pull manifested at foil oscillations.
At the beginning of moving, asymptotic solution of integral equation (6.1.44) is
⎛1 τ ⎞
q (τ ) ≈ ⎜ + ⎟ e jλτ , τ → 0 , (6.1.46)
⎝2 8⎠
and equation (6.1.45) yields the following asymptotic expressions:
⎡ ⎛ τ⎞ ⎤
CL = πε ⎢λ 2 cos λτ + λ ⎜1 + ⎟ sin λτ ⎥ , τ → 0,
⎣ ⎝ 4⎠ ⎦
2
(6.1.47)
πε ⎛ τ ⎞ πε 2 ⎡ ⎛ τ ⎞ ⎤
CM = λ ⎜1 + ⎟ sin λτ , τ → 0, C X = λ ⎜ 1 + ⎟ sin λτ ⎥ , τ → ∞ .
4 ⎝ 4⎠ 2 ⎢⎣ ⎝ 4 ⎠ ⎦
220
When τ → ∞ (quasi-stationary movement), the solution and all characteristics become
oscillating with the same frequency. Presenting the solution as q(τ ) = q0 e jλτ , one can obtain
from integral equation (6.1.44) the expression for the unknown function q0 (λ ) as
H1(2) (λ )
q0 = = C (λ ) , (6.1.48)
H1(2) (λ ) + jH 0(2) (λ )
where H n(2) (λ ) is the Hankel function of second order; C (λ ) is the Theodorsen function.
All well-known results for quasi-stationary movement are obtained from (6.1.45).
Namely, average values of coefficients for the period of oscillation ( 2π / λ ) are obtained as
CLav = 0, CMav = 0, CDav = πε 2 λ 2 | C (λ ) |2 . (6.1.49)
Whence, two kinds of asymptotic expressions are valid:
CDev = πε 2 λ 2 , λ → 0; CDav = πε 2 λ 2 / 4, λ → ∞ . (6.1.50)
Sec. 6.2. Unsteady Movement of a Plate with a Kirchhoff Cavity

6.2.1. Paradoxes of Unsteady Cavity
Unsteady cavitating flows have been considered by many authors [Parkin, 1957; Wu, 1957;
1958; Timman, 1958; Geurst, 1960]. It was found that studying a non-stationary cavitating
flow in a planar case is connected with the main paradoxes which essentially consist in the
following: in planar cavitating flow, changing the cavity’s area makes it impossible to
satisfy the condition of pressure limitation in infinity. This follows from the common
equations of the hydrodynamics. Indeed, variation of the cavity area is equivalent to a source
from variable intensity, i.e. the speed potential and its partial derivative on time grow
according to the logarithmic law with an increase of distance from cavity, and according to
Couchi – Lagrange integral, the pressure also grows as logarithm of distance. It is possible to
overcome this paradox considering open models of cavitating flow.
The simplest one is the Kirchhoff model with a semi-infinite cavity. Pressure inside
cavity is considered constant and equal to pressure at infinity. A non-stationary movement of
a plate inclined at a small angle to the velocity direction is considered below within the
frame of this model [Terentiev & Mikhailov, 1979]. The model cannot describe the full
cavitating flow, nevertheless, it allows finding some features of non-stationary cavities.
6.2.2. Pulsation of a Kirchhoff Cavity Downstream of a Plate
Consider a cavity attached to a plate at its trailing edge. At some point on its upper side, it is
moving from the leading edge to distance a, Fig.6.2.1a. The plate length is assumed to be
unity. The plate speed is given and is denoted as U1 before the pulse and U 2 after, so that
time derivative s is given as before by (6.1.6). The stream function and its time derivative
on the wetted part of the plate are determined as
ψ ( x, 0, t ) = T ( t ) + α s ( x − s ) , (6.2.1)
ψ t ( x, 0, t ) = T ( t ) − α s 2 . (6.2.2)
At the cavity boundary equation (6.1.3) is valid. This mixed boundary problem could be
solved parametrically by conformal mapping the z-plane with semi-infinite cut DB on the
upper half ζ -plane, Fig.6.2.1b. The transformation is ζ =i z−s . (6.2.3)
The problem consists of three consecutive problems: 1) stationary movement before pulse,
221
2) impact and 3) further movement. The first problem can
be considered directly before the impact. So long as the
problem in moving ( x′, y′ ) coordinates is steady, its
solution can be found from the value problem presented as
follows:
ψ 1t = −α U12 , ξ ∈ (− a ,1), η = 0,
(6.2.4)
ϕ1t = 0, ξ ∈ (−∞, − a ) ∪ (1, ∞), η = 0 .
Besides, function w1t ( z , t ) has a simple pole at point
Fig.6.2.1. (a) - plate with ζ = 0 , limited at points ζ = − a and ζ = 1 , and vanishes
Kirchhoff cavity; (b) the at infinity. The boundary problem solution is expressed by
parametric plane. function
⎛ 1 ⎞
w1t ( z , t ) = −iα U12 ⎜1 −
⎝ ζ
(
(ζ − 1) ζ + a ⎟ . (6.2.5)
⎠
)
The complex potential could be found from (6.2.5) by integrating with respect time
t, or as long as time is connected with variable ζ with function t = (ζ 2 + z ) / U1 , it can be
integrated with respect to ζ :
{
w1 ( z , t ) = iT1 + iα U1 −ζ 2 + (ζ − b ) (ζ ( )
− 1) ζ + a − 2c 2 ln ( ζ −1 + ζ + a )} , (6.2.6)
1− a 1+ a
where b= , c= .
2 2
Whence, the stream function before pulse t < 0 for z = x > 0 is expressed as
(
⎧x + i x − b
⎪ ) ( )(
i x − 1 i x + a −⎫
⎪ )
ψ 1 ( x, − 0 ) = T1 + α U1 Re ⎨ ⎬. (6.2.7)
⎩ (
⎪−2c 2 ln i x − 1 + i x + a ⎪
⎭ )
The second problem is to find potential w0 ( z ) for pulse movement from speed U1 to
U2, i.e. the following boundary condition should be satisfied:
ϕ0 = 0, ξ ∈ (−∞, − a ) ∪ (1, ∞), η = 0, ψ 0 = T0 + α (U 2 − U1 ), ξ ∈ (− a ,1). (6.2.8)
Besides it should be bounded at points A(− a , 0) and C (1, 0) . A solution of this mixed
⎝ ⎠
(
value problem is w0 ( z ) = iT0 + i α (U 2 − U1 ) ⎛⎜ −ζ 2 + (ζ + b ) (ζ − 1) ζ + a ⎞⎟ . ) (6.2.9)
Whence, the stream function on z = x > 0 is found by function
( )
ψ 0 ( x ) = T0 + α ( U 2 − U1 ) Re x + (i x + b) (i x − 1)(i x + a ) . (6.2.10)
Now, we can confirm that the initial stream function for unsteady movement of the plate is
the sum ψ ( x, 0) = ψ 1 ( x) + ψ 2 ( x) . (6.2.11)
The partial time derivative of complex potential for unsteady movement is obtained
from value problem (6.2.2) and is determined by the same function as (6.2.5), but the factor
222
should be changed to i (T (t ) − αU 22 ) . Besides, variable ζ is a function of z and t .

Substituting equation (6.2.3), one can express the potential and also stream function on the x-
axis at any moment. As before, distance s = U 2t should be used instead of time and
integrated variable s′ = U 2τ , s′ ∈ (0, s ) . Satisfying continuous condition (6.1.7), one obtains
an integral equation
( )
2
s
β1 ( s′) 1 1+ s + a + s
∫ (1 + s − s′ )( a + s − s′ ) + a + s − s′ds′ = s . (6.2.12)
0 s − s′ 2 (1 + s )( a + s ) + a+s
Here the following designations are entered
−1
T (t ) ⎛ U ⎞ ⎛ U ⎞
β (s) = 1 − , β1 ( s ) = ⎜1 − 1 ⎟ ⎜ β ( s ) − 1 ⎟ . (6.2.13)
αU 2 2
⎝ U2 ⎠ ⎝ U2 ⎠
Hydrodynamic coefficients are determined by function β(s) only:
1
the lift L = 2 ραU 22 β ( s ) ∫ (1 − x ) ( x + )
a dx = πραU 22 c 2 β ( s ) ; (6.2.14)
− a
the concentrated force at leading edge and longitudinal pressure force

X 1 = πρ a (α U 2 β ( s ) ) ,
2
X 2 = −α L ; (6.2.15)
the drag D = πρα 2U 22 β ( s ) ( )

a β ( s ) −c 2 . (6.2.16)
Equation (6.2.15) shows that the concentrated force vanishes if the attached point
coincides with the leading edge, a=0; if the attached point is at the trailing edge, all
expressions coincide with those for a smooth flow. Integral equation (6.2.12) in this case
transforms to equation (5.5.22). Consider integral equation for a=0:
s
1+ s − x + s − x 1
( )
3/ 2
0
∫ s−x 2
β1 ( x) dx = s1/ 4
s + s +1 (6.2.17)
The solution of the last equation permits an asymptotic expansion as

γ 1 9 − 4πγ 2 33 + 4πγ 2 1
β1 ( x) ≈ + +γ s = 0.20865 + 0.42572 + 0.91779 s , (6.2.18)
2 s 16 16 s
Γ(5 / 4)
where γ= , Γ( z ) is Gamma-function.
π Γ(3/ 4)
As follows from the equation, the lift becomes indefinitely big as distance approaches zero
( s → 0 ). Physically it can mean that at pulse movement the stream cannot come off from the
leading edge. The attached point apparently moves along the upper side of plate downwards
on a stream. This theoretical conclusion waits for further experimental proof.
For the separation point at any position a, for a small distance s<<a, the equation
solution can be presented as an asymptotic series of small parameter τ = s /(1 + s ) << 1 . The
first two terms of expansion are:
223
2
1 ⎛1+ a ⎞ ⎛ 1− 4 a + a ⎞
β1 ( x) = ⎜ ⎟ ⎜1 − τ ⎟⎟ + O (τ 2 ) , τ << a ≠ 0 . (6.2.19)
2 a ⎜⎝ 2 ⎟⎠ ⎜⎝ 4a ⎠
If the distance grows, the solution and all formulas approach the known formulas for a
steady flow of a plate with attached Kirchhoff cavity. The lift and drag coefficients are
expressed as follows:
2 2
⎛1+ a ⎞ ⎛
2 1− a⎞
CL = 2πα c = 2πα ⎜⎜
2
⎟⎟ , CD = −2πα b = −2πα ⎜⎜
2 2
⎟⎟ , s → ∞ . (6.2.20)
⎝ 2 ⎠ ⎝ 2 ⎠
For solving the integral equation numerically it is expedient to get a bounded
integrand. Applying a method similar to that used in the previous section, we can transform
s
equation (6.2.12) to ∫ K ( s − x ) β ( x)dx = f ( s )
0
1 1 1 (6.2.21)
∫( )
π /2 1/ 2
with bounded equation kernel K1 ( x ) = 1 + x sin 2 ϕ a + x sin 2 ϕ + a + x sin 2 ϕ dϕ ,
0
( )
2
1 1 + s (1 − x 2 ) + a + s (1 − x 2 ) 1 − x2
s
f (s) = ∫ dx .
( 1 + s (1 − x ) + ))
1/ 2
a + s (1 − x )+ a + s (1 − x
20 2 2 2
Dependence of solution β1 ( s ) on distance for several values

of parameter a, calculated numerically, is plotted in
Fig.6.2.2. The first value for initial distance was calculated
by formula (6.2.19).
6.2.3. Acceleration of a Plate with Cavitation Effects
The kinematical condition on the wetted part of a plate is
ψ t = T (t ) − αU 2 (t ) + αU (t )( x − s (t )) ,
while dynamic condition is determined by second equation
in (6.2.4). Besides, if the movement is given by constant
speed U1 before initial time ( t < 0 ) and then at any
acceleration, the initial stream function is determined by Fig.6.2.2. Curves of β1 at
equation (6.2.7) only. Partial derivative wt ( z , t ) can be found different distances a.
on the ζ plane as
⎛
⎜
w t ( z, t ) = i (T ( t ) − α U (t ) ) ⎜ 1−
(ζ (
− 1) ζ + a ⎞⎟ )
⎟+
2
⎜ ζ ⎟
⎝ ⎠
(6.2.22)
⎛
⎜ 2 ⎛ 2
+iαU (t ) ⎜ −ζ + ⎜ ζ + bζ + b + ⎟
2
⎞
c 2 ⎞ (ζ − 1) ζ + a ⎟ ( )
2⎠ ζ ⎟,
⎜ ⎝ ⎟
⎝ ⎠
where parameters b and c are determined as in equation (6.2.6).
224
Let the plate moves at constant acceleration g = U = const . Denoting
⎛ c2 ⎞
k = g / U12 , T (t ) − αU (t ) ⎜ b 2 + ⎟ − αU 2 (t ) + αU1U (t ) = −αU1U (t ) β ( s ), (6.2.23)
⎝ 2⎠
one can find speed U = U1 1 + 2ks and, satisfying continuity condition (6.1.7), obtain the
integral equation
s
(1 + s − x )( a + s − x ) + a +s−x
∫ β ( x)
0
s−x
dx =
(6.2.24)
( )
2
k
s 1+ x + a + x x
= ∫ dx .
20 (1 + x )( a + x ) + a + x 1 + 2k ( s − x)
If attached points are located at both edges of plate ( a = 0 ) the integral equation looks like
s s
1+ s − x + s − x k ( 1 + x + x )3 / 2 1/ 4
∫
0 s−x
β ( x)dx =
2 ∫0 1 + 2k ( s − x)
x dx . (6.2.25)
Whence the asymptotic behavior of the solution is

Γ(5 / 4)
β ( x) ≈ − k s, s → 0 . (6.2.26)
π Γ(3/ 4)
If both attached points are located at the trailing edge a = 1, integral equation (6.2.24) has
s s
1+ s − x x 1+ x
the form ∫0 β ( x) s − x dx = k ∫0 1 + 2k (s − x) dx . (6.2.27)
This equation differs from equation (6.1.35) for smooth flow, but its solutions are connected
β ( x) ks
with relationship β ( x) = 2 − . In the case of s → 0 (initial
U1 1 + 2ks
movement with acceleration), equation (6.2.27) has asymptotic solution as
k k (1 − 2k ) s 2
β ( x) ≈ s + , s →0. (6.2.28)
2 8
For large distance the solution doesn’t depend on attached point location a and is equal to
β ( x) ≈ 2ks , s → ∞, 0 ≤ a ≤ 1 . (6.2.29)
Hydrodynamic forces (lift and concentrated force at the leading edge) are, respectively,
⎛ ⎞
L = πρ c 2 ⎜ αU1U ( s ) (1 + β ( s ) ) + α U ( s ) ( 2b 2 + c 2 / 4 ) ⎟ ,
.
⎝ ⎠ (6.2.30)
X 1 = πρ c 2U12 a (1 + β ( s ) ) .
2
Drag is calculated as difference D=aL−X1. Lift from the start of movement varies step-wise:
2
ρU12πα k ⎛ 1 + a ⎞
ΔL =
16
⎜⎜
8
(
⎟⎟ 9 + 9a − 14 a . ) (6.2.31)
⎝ ⎠
225
Numerical results are plotted in Fig.6.2.3 with the following notation used:
L = 2 L / ρU 2lα , X 0 = 2 X 0 / ρU 2lα 2 , D = 2 D / ρU 2 lα 2 .
Frames a), b) and c) show the solution and forces at distance s for a = 0, 0.5 and 1,
respectively.
Fig.6.2.3. Solution and forces at distance s for different locations of the attached point.
Curve L ( s ) coincides with drag coefficient D( s ) in Fig.6.2.3a but concentrated force X 0

vanishes. Curves L ( s ), D ( s ) and X 0 ( s ) on frame b) have horizontal asymptotes
y = 2π c 2 , y = 2π b 2 and y = 2π a , respectively; on frame c) asymptotes for curves L ( s )
and X 0 ( s ) coincide while the drag vanishes as s → ∞ . Since solution β ( s ) vanishes as
s → 0 , forces L ( s ), D( s ) and X 0 ( s ) limit to:
⎛ k (9 + 9a − 14 a ) ⎞ ⎛ 2 kc 2 (9 + 9a − 14 a ) ⎞
lim L ( s ) = 2π c 2 ⎜⎜ 1 + ⎟ , lim
⎟ s →0 D ( s ) = 2π ⎜⎜ b + ⎟⎟ , lim X 0 ( s ) = 2π a .
s →0
⎝ 16 ⎠ ⎝ 16 ⎠
s →0
6.2.4. Harmonic Oscillations of a plate with Cavitation Effects

Let the plate be parallel to the x-axis and move with constant velocity U towards the positive
direction of x. At initial moment t=0 the plate starts to oscillate harmonically as
y = ε 0 e jω t − ε 0 , t > 0, j 2 = −1 . (6.2.32)
The horizontal plate, when moving, does not disturb fluid: though the kinematic and
dynamic conditions are satisfied, the fluid is still and the stream function vanishes. Thus
function ψ 1 ( x) is zero at any moment. The vertical speed of any point of plate is equal to
V = Re(ε 0 jω e jω t ) and is zero at the initial time. Thus, function ψ 0 ( x) is also zero. As
before, the kinematical condition could be written in complex form as
ψ t (ξ ) = T (t ) + ε 0Ujω e jω , x ∈ (− a,1) . (6.2.33)
The dynamic condition remains unchanged. The mixed value problem solution is as follows:
226
⎛ (ζ − 1)(ζ + a ) ⎞
wt ( z , t ) = i (T + ε 0Ujω ) ⎜1 − ⎟+
⎜ ζ ⎟
⎝ ⎠
(6.2.34)
⎡ ⎛ c 2 ⎞ (ζ − 1)(ζ + a ) ⎤
+iε 0ω 2 e jω t ⎢ −ζ 2 + ⎜ ζ 2 + bζ + b 2 + ⎟ ⎥.
⎢ ⎝ 2⎠ ζ ⎥
⎣ ⎦
Satisfying condition (6.1.7) one obtains an integral equation in the form:
e j 2λ ( s − x ) ⎡ 4b 2 ⎤
s s
∫0 β − = λ ∫0 (1 + x)(a + x) ⎢⎣ K ( x) + ( a − b ) K ( x) ⎥⎦dx ,
2
( x ) K ( s x ) dx j 6.2.35)
where K ( x) = (1 + x)( a + x) + a + x / 4 x , λ = ω / 2U is Strouchal number,

β ( x) is a new unknown function connected linearly with T (t ) .
Vertical force can be found integrating the pressure along the wetted part of plate
⎡ 9 + 9a − 14 a ⎤
Y = ρU 2πε 0 c 2 ⎢ β ( s ) − j 2λ e j 2 λ s + λ 2 e j 2 λ s ⎥. (6.2.36)
⎣ 4 ⎦
The longitudinal force is calculated as concentrated one at the leading edge only
X = ρU 2πε 02 a ( β ( s ) − j 2λ e j 2 λ s ) .
2
(6.2.37)
Force X is always positive so that it can be considered as draft of the oscillated plate but if
parameter a vanishes, the force vanishes as well.
Consider some special cases. Let both attached points are located at thetrailing edge:
a=1, then b = 0, c =1. The integral equation is transformed to
e j 2λ ( s− x)
s
1+ s − x
∫0 β ( x) s − x dx = jλ ∫ x(1 + x) dx . (6.2.38)
This equation differs from similar integral equation (6.1.44) for the plate without cavity but
their solutions are connected as β ( s ) = jλ (e jλ s − q( s)) . (6.2.39)
In two limiting cases at initial moment s → 0 , and for steady oscillation s → ∞ , the solution
β ( s ) → jλ e j 2 λ s , s → 0,
of equation (6.2.38) becomes, respectively,
β ( s ) → j 2λ e j 2 λ s [1 − C (λ )], s → ∞.
Here C (λ ) is the Theodorsen function.
For numerical calculations singularities of the integrands in (6.2.38) should be
removed, as used earlier, multiplying by function ( s1 − s ) −1/ 2 and integrating with respect to
variable s in interval (0,s1). As a result, integral equation (6.2.38) changes to
s
∫ β (s′) F (s − s′)ds′ = f (s) ,

0
(6.2.40)
⎛ x ⎞
( )
2
e− j 2λ t
s
∫0
j 2λ s
where F ( x) = 2 1 + x E ⎜⎜ ⎟⎟ , f ( s ) = 2πλ je λ − 2
erf j 2 ( s t ) dt .
⎝ 1+ x ⎠ 1+ t2
Let the attached point coincides with the leading edge of plate: a = 0, b = ½, c = ½.
The integrand in (6.2.35) becomes K ( x) = 1+ x + x / 4 x . (6.2.41)
227
For initial movement the integral equation (6.2.35) is transformed to asymptotic equation
s
β ( x) s
⎛ 7 1 ⎞
∫0 4 s − x ∫0
jλ ( s − x )
= j λ e ⎜ 4
− ⎟dx, s → ∞ .
⎝8 x 4 x ⎠
4 3
Asymptotic expressions of the solution look like

⎛7 Γ(1/ 4) ⎞
β ( s ) ≈ jλ e j 2 λ s ⎜ − ⎟, s → ∞ . (6.2.42)
⎝8 s 4 π Γ(3/ 4) ⎠
Consider now a steady oscillating movement of the plate with attachment points at
both edges (a = 0) of the plate at infinitely large distance ( s → ∞ ). As long as the frequency
at a limited large distance for a = 1 due to (6.2.39) is equal to λ / π , equation (6.2.35) can be
found as β ( s ) = 2λ jAe 2 λ j s . (6.2.43)
Substituting it in integral equation (6.2.35) and approaching s → ∞ , one can write the
M (λ ) + M 3 (λ )
unknown factor as A= 2 , (6.2.44)
M 1 (λ )
∞ ∞ 2
1 1 e − ju u
∫ M 2 (λ ) = ∫
2
here M 1 (λ ) = e − ju u (u + 2λ + u 2 ) du , du ,
2λ 0 2 0 2λ + u 2 u + 2λ + u 2
∞ 2
1 e − ju u + 2λ + u 2
M 3 (λ ) = − ∫ du .
8 0 2λ + u 2 u
These integrals for small Strouchal number as well as in unsteady theory could be
represented by Hankel functions. Multiplying and dividing the integrand in M 1 (λ ) by
function 2λ + u 2 and then expanding in power series on small parameter λ as
(
u (2λ + u 2 ) u + 2λ + u 2 = 2u 2 +
5
4
)
2λ + , λ → 0 ,
one can express the first integral as
∞ 2 ∞
1 e − ju ⎛ 5 ⎞ 2 jλ − jλ cosh x
2λ ∫0 2λ + u 2 ⎝ ∫0 e
M 1 (λ ) ≈ ⎜ 2u 2
+ 2λ ⎟ du = e (cosh x + 1/ 4)dx.
4 ⎠ 4
Using integral identity of Hankel function [Yanke et al., 1968]
∞
π
∫e
− j λ cosh x
coshν xdx = ( − j )ν +1 Hν(2) (λ ), λ → 0 ,
0
2
2 jλ ⎛ (2)π j ⎞
we present M 1 (λ ) ≈ −e ⎜ H1 (λ ) + H 0(2) (λ ) ⎟ , λ → 0 . (6.2.45)
2 4 ⎝ 4 ⎠
The other functions could be expressed similarly:
π 2 jλ (2)
M 2 (λ ) + M 3 (λ ) ≈ − e jH 0 (λ ) . (6.2.46)
2 16
So factor (6.2.44) is expressed by Hankel functions as
jH 0(2) (λ )
A≈ . (6.2.47)
4 H1(2) (λ ) + jH 0(2) (λ )
Whence, the amplitude of the solution for small Strouchal number is expressed as
228
λ (π j + 2C + 2ln(λ / 2)
A≈ , λ << 1 , (6.2.48)
8 j + λ[π j + 2C + 2ln(λ / 2)]
where C = 0.5772… is Euler constant. Now asymptotic relationship for perpendicular force
⎡ 1 ⎤
⎢ − j 2λ + λ ln 2λ + ⎥
π 4
can be written Y = ρε 0 e j 2 λ s ⎢ ⎥ + O(λ 3 ln λ ), λ → 0 . (6.2.49)
4 ⎢ 2⎛π C − ln 4 9 ⎞ ⎥
⎢⎣ + λ ⎜⎝ 4 j + 2
+ ⎟⎥
4 ⎠⎦
In another case for large Strouchal number, λ → ∞ , the integrals in (6.2.44) could
be asymptotically expressed by Gamma functions as
je jπ / 8 ⎛ Γ(5 / 4) − jπ / 4 ⎞
M 1 (λ ) ≈ − 3/ 4 ⎜
Γ(3/ 4) + e ⎟,
2(2λ ) ⎝ 2 2λ ⎠
je jπ / 8 ⎛ Γ(5 / 4) − jπ / 4 ⎞
M 2 (λ ) ≈ − 3/ 4 ⎜
Γ(3/ 4) − e ⎟,
4(2λ ) ⎝ 2 2λ ⎠
je jπ / 8 ⎛ Γ(3/ 4) − jπ / 4 3 jΓ(5 / 4) ⎞
M 3 (λ ) ≈ − 3/ 4 ⎜
Γ(1/ 4) − j e + .
16(2λ ) ⎝ 2 2λ 16λ ⎟⎠
1⎛ Γ 2 (5 / 4) ⎞ j 2λ
Hence, A≈ ⎜ 7 + 4 ⎟ − (1 + j ) , λ →∞. (6.2.50)
16 ⎝ Γ (3/ 4) ⎠
2
8
Asymptotic expansion of vertical force looks like
⎡9 (1 − j ) Γ(5 / 4) ⎤
⎢16 (2λ ) + (2λ )3/ 2 − ⎥
2
π 8 Γ(3/ 4)
Y = ρε 0 e j 2 λ s ⎢ ⎥ + O ( 2λ ) . (6.2.51)
4⎢ j ⎛ Γ 2 (5 / 4) ⎞ ⎥
⎢− ⎜ 7 − 4 2 ⎟ (2λ ) ⎥
⎢⎣ 16 ⎝ Γ (3/ 4) ⎠ ⎥⎦
Factors of the given expansion practically coincide with Kuznetsov's results [Kuznetsov,
1975] who solved harmonic oscillations of plate with detachment points at its both ends by
other methods. Some results for oscillations of a plate are plotted in Figs.6.2.4 and 6.2.5.
Fig.6.2.4. Oscillations of plate at a = 1 :

left - changing the amplitude by movement; right - amplitude vs. λ for steady oscillation.
229
Fig.6.2.5. Oscillations of plate at a = 0 : (a) - changing amplitude by movement;

(b) - amplitude of vertical force versus λ for steady oscillation;
doted line is calculated by (6.2.51), dashed line – by (6.2.49)
Sec. 6.3. Water Entry of a Thin Hydrofoil

Penetration of a body into a fluid is one of complicated problems in the hydrodynamics.
Only a few cases of this problem have been solved, limited mostly to symmetric entry
[Sagomonyan, 1974; Korobkin & Pukhnachov, 1988; Terentiev, 2002). A non-symmetric
but perpendicular entry of thin foils was studied by Yim [1971]. Two approaches to solving
the inclined penetration of thin foils have been considered by Terentiev [1977b, 1979] – one
is based on complex velocity, another on the partial time-derivative of complex potential.
These methods are discussed below.
6.3.1. Complex Velocity Approach
Consider the motion of a partially submerged slender body (Fig.6.3.1a). Let the abscissa axis
be parallel to principal velocity U of the body so that the y-component of speed v is small.
The linear value problem is formulated on the half plane with an inclined slit (the
definitional domain in Fig.6.3.1a is marked by the dashed line). A linear kinematic condition
on the wetted surface of the body is imposed as:
⎛ ∂w ⎞
Im ⎜ ⎟ = f ( x, t ) , x ∈ (l ,0) , y = ±0 , (6.3.1)
⎝ ∂z ⎠
where f(z,t) is described by the given motion of
the body. The dynamic condition on the free
boundary (inclined dashed line) is described by
Fig.6.3.1. Inclined entry in fluid: ⎛ ∂w − iπγ ⎞
Re ⎜ e ⎟ = 0 on Im ( zeπγ i ) = 0 . (6.3.2)
(a) - flow in the z-plane; ⎝ ∂z ⎠
(b) – the parametric ζ - plane. Besides, a solution of the mixed value problem
should possess at the leading edge a singularity
of the type ∂w ∂z ≈ ( z − l ) −1/ 2 , (6.3.3)
230
and a zero of the second kind at infinity ∂w ∂z ≈ z −2 , z → ∞ . (6.3.4)

The mixed value problem solution can be determined in a parametric form by
mapping conformably the definitional domain in the z-plane onto the upper half-plane of the
auxiliary ζ-plane, Fig.6.3.1b. The transformation is:
1−γ
⎛ ζ ⎞ 1− γ
z = l (1 − ζ )γ ⎜ 1 + ⎟ , a = . (6.3.5)
⎝ a ⎠ γ
Instead of function wz ( z, t ) , it is more convenient to solve another value problem of the
partial derivative ∂w ∂ζ = wζ (ζ , t ) , which satisfies the following conditions:
Re wζ = 0 on ξ ∈ (−∞, − a ) ∪ (1, ∞), η = 0 , (6.3.6)
Im wζ = g ( x(ξ , t ), t ) xξ (ξ , t ) on ξ ∈ [− a,1] . (6.3.7)
Multiplying by function g (ζ ) = (1 − ζ ) (a + ζ ) , the mixed value problem is reduced to
Schwartz problem with boundary condition,
⎧ f ( x(ξ , t ) xξ (ξ ) g (ξ ) ξ ∈ [− a,1], η = 0,
Im[ wζ g (ζ )] = ⎨ (6.3.8)
⎩0 ξ ∈ (−∞, − a) ∪ (1, ∞), η = 0 .
Thus a solution of the value problem is in the form
1− 2γ
laγ −1 ⎛ a + τ ⎞ 2 gτ dτ
1
wζ = ∫ ⎜
π (1 − ζ )(a + ζ ) − a ⎝ 1 − τ ⎠
⎟
τ −ζ
. (6.3.9)
Complex velocity wz = wζ (ζ , t ) zζ (ζ , t ) is a function of variable ζ and time t;

together with equation (6.3.4) they represent the value problem solution presented in
equations (6.3.1) to (6.3.4). Since variable ζ is a function of z and t, it is quite complicated
to obtain complex potential w( z , t ) and its partial time derivative wt ( z , t ) = ∂w( z , t ) ∂t . On
the surface of the body, potential ϕ (ξ , t ) can be determined integrating equation (6.3.9) over
interval (-a, ξ), and then pressure can be expressed in the form:
⎛ ∂ϕ (ξ , t ) ∂ξ ( x, t ) ∂ϕ (ξ , t ) ⎞
p = −ρ ⎜ + ⎟, (6.3.10)
⎝ ∂t ∂t ∂ξ ⎠
∂ξ ( x, t ) l (1 − ξ )(a + ξ )
where = . (6.3.11)
∂t lξ
The potential on [ − a, 1] is obtained from (6.3.9) by integral
ξ (1− 2 γ ) / γ
⎧ ∂w ⎫ l ⎛ a +τ ⎞
ϕ (ξ , t ) = ∫ Re ⎨ ⎬ d ξ = 1−γ ∫ f (τ ) I (τ , ξ ) ⎜ ⎟ τ dτ , (6.3.12)
−a ⎩ ∂ζ ⎭ζ =ξ πa ⎝ 1−τ ⎠
ξ
dξ ln | G (τ , ξ ) |
where I (τ , ξ ) = ∫
−a (a + ξ )(1 − ξ )(τ − ξ )
=−
( a + τ )(1 − τ )
,
( a + ξ )(1 − τ ) − (a + τ )(1 − ξ )
G (τ , ξ ) = ln .
(a + ξ )(1 − τ ) + (a + τ )(1 − ξ )
Integrating the pressure along the wetted part of the foil, one can calculated force
and torque. The transverse force is equal to
231
∂x(ξ , t )
0 1
Y = ∫ ( p1 − p2 )dx = − ∫ ( p − p0 ) dξ . (6.3.13)
−1 −a
∂ξ
∂ξ ( x, t ) ∂x(ξ , t ) l
Since = − x(ξ , t ) , integral (6.3.13) is expressed in integral form
∂t ∂ξ l
ρ 1 1
∂x(ξ , t ) ln | G (τ , ξ ) |
Y =−
πa 1−γ ∫ ∫ [2l g (τ ) + l g (τ )]
−a −a
∂ξ (a + τ )γ (1 − τ )1−γ
τ d ξ dτ , (6.3.14)
where l = dl / dt = −U is the speed of the moving body.

The integral with respect to variable ξ can be found analytically. For that we consider
∂z (ζ ,τ )
function F (ζ ,τ ) = eπγ i ln G (τ , ζ ) , which is pure imaginary on the real axis
∂ζ
1 ∞
1
outside interval (-a, 1), thus, Re ∫ F (ξ ,τ )d ξ = Re ∫ F (ξ ,τ )d ξ = − Re ∫ F (ζ ,τ )d ζ .
−a −∞
2 ζ =∞
1
πl
Taking Taylor series of the integrand, one obtains Re ∫ F (ξ ,τ )d ξ = (a + τ )(1 − τ ) .
−a
a1−γ
On the other hand the integral could be presented as
∂x(ξ , t )
1 1
Re ∫ F (ξ ,τ )d ξ = cos πγ ∫ ln | G (τ , ξ ) | d ξ − π x(τ , t )sin πγ .
−a −a
∂ξ
1
∂x(ξ , t ) π ⎛ l (a + τ )(1 − τ ) ⎞
Hence, ∫− a ∂ξ ln | G (τ ,ξ ) | dξ = cos πγ
⎜⎜ x(τ , t )sin πγ +
a1−γ
⎟⎟ .
⎝ ⎠
Taking into account this expression, the transverse force (6.3.14) is expressed as
ρl 1 ⎛ g (τ ) ⎞
Y= ∫ ⎜ l g (τ ) + l
2 −a ⎝ 2 ⎠
⎟ L(τ )dτ , (6.3.15)
4τ ⎡⎛ a + τ ⎞1− 2γ ⎛ a +τ ⎞
(1− 2 γ ) / 2
⎤
where L(τ ) = 2(1−γ ) ⎢⎜ ⎟ sin πγ − ⎜ ⎟ ⎥.
a cos πγ ⎣⎢⎝ 1 − τ ⎠ ⎝ 1−τ ⎠ ⎦⎥
Function L(τ) has uncertainty as the inclination angle γ→1/2; its limit is as
4τ 1 + τ 1
L(τ ) = ln , γ= . (6.3.16)
π 1−τ 2
Longitudinal force consists of concentrated force at the leading edge O ( x = −l ) and
longitudinal projection of pressure on the real axis. The force is determined as
(1− 2γ ) / 2 2
iρ wζ 2
ρl ⎡1 ⎛ a +τ ⎞ ⎤
Xc = − ∫
4 ζ = 0 zζ
dζ = − ⎢∫
2π a 2(1−γ )
⎢⎣ − a
g (τ ) ⎜
⎝ 1−τ ⎠
⎟ dτ ⎥ .
⎥⎦
(6.3.17)
Its projection on the x-axis is calculated by integrating pressure along the foil
1
Xp = ∫ ( p − p )β ( x + l ) zζ dξ .
−a
0 (6.3.18)
where β ( x + l ) = β ( x′) is the slope of tangent to the foil boundary.
232
For parallel movement of a foil, function g (ξ , t ) = −l β ( x + l ) and force (6.3.18) could be

ρl 1 1
expressed as Xp =
2 ∫ ∫ q(τ ,ξ ) R(τ ,ξ )dτ dξ ,
−a −a
(6.3.19)
2τξ ln | G (τ , ξ ) |
where q(τ , ξ ) = [l g (τ ) + l g (τ )]β (ξ ), R (τ , ξ ) = 2(1−γ )
.
πa ( a + τ )γ (1 − τ )1−γ (a + ξ )γ (1 − ξ )1−γ
Torque relative to the origin of the fixed coordinates is determined by integral
∂x(ξ , t )
1
M = − ∫ ( p − p0 ) x(ξ , t ) dξ .
−a
∂ξ
After integration by parts the integral can be transformed to
ρ 1 1 zζ2 (ξ , t ) ln | G (τ , ξ ) |
2π a1−γ −∫a −∫a
M =− [3l g (τ ) + l g (τ )] τ dτ d ξ . (6.3.20)
(a + τ )γ (1 − τ )1−γ
Expression (6.3.20) can be used for calculation of torque. But it can be transformed to single
integral similarly to transverse force (6.3.15). Here function e2πγ i zζ2 (ζ ,τ ) ln G (τ , ζ ) should
be considered instead of F(ζ,τ). After manipulations, the final expression of torque is:
ρl2 1 l
M =− ∫
2 −a
[l g (τ ) + g (τ )]K (τ )dτ ,
3
(6.3.21)
(1− 2γ ) / 2
3τ ⎛ 3(1 − 2γ ) ⎞ ⎛ a + τ ⎞ ⎤
where K (τ ) = 3(1−γ )
⎡ (a + τ ) 2 −3γ (1 − τ )3γ −1 sin 2πγ − ⎜τ +
⎣ ⎟⎜ ⎟ ⎥.
a cos 2πγ ⎝ 2γ ⎠⎝ 1−τ ⎠ ⎥⎦
For γ→1/2 the function approaches the limit
1/ 4
τ ⎡ 3 +τ ⎤⎛ 3 +τ ⎞ 1
K (τ ) = 5 / 4 ⎢ (3 + τ )ln − 4⎥ ⎜ ⎟ , γ= .
π3 ⎣ 1−τ ⎦ ⎝ 1−τ ⎠ 2
6.3.2. Entry of a Wedge with Straight Sides
Consider a parallel motion of the wedge with speed U = −l . Let angles α1 and α 2 stand for
inclination angle to the x-axis at the upper and lower sides, respectively, i.e.
g (τ ) = U α1 , τ ∈ (0,1) and g (τ ) = U α 2 , τ ∈ ( − a, 0) . Then the following expressions for
transverse force, longitudinal force, leading edge force and moment, respectively, are:
ρ l ⎛ lU 2⎞ ρl 2 ⎛ α 12 I1 + α 22 I 2 ⎞
Y= ⎜ − U ⎟ (α1 I1 + α 2 I 2 ), X = (U − lU ) ⎜ + (α1 − α 2 ) 2 I 3 ⎟ ,
2 ⎝ 2 ⎠ 2 ⎜ 2 ⎟
⎝ ⎠ (6.3.22)
ρU l 2
ρl ⎛ 2 U l ⎞
2
Xc = − (α1 I 4 + α 2 I 5 ) 2 , M = ⎜U − ⎟ (α1 I 6 + α 2 I 7 ),
2 2 ⎝ 3 ⎠
(1− 2γ ) / 2
⎛ a +τ ⎞
1 0 1 1 1
1
here I1 = ∫ L(τ )dτ , I 3 = ∫ ∫ R(τ , ξ )dτ d ξ , I 4 = 1−γ ∫ ⎜ ⎟ dτ , I 6 = ∫ K (τ ) dτ .
0 −a 0 π a 0 ⎝ 1−τ ⎠ 0
Factors I 2 , I 5 and I 7 are calculated by the same integrals as I1 , I 4 and I 6 , respectively, but
integrating should be on interval ( − a, 0 ). It can be shown using beta-functions that the sums
are expressed as:
233
2
2π ⎛ 1 − 2γ ⎞ π (1 − 2γ )(16γ 2 − 16γ + 3)
I1 + I 2 = 2( I 4 + I 5 ) = 2(1−γ )
2
⎜ ⎟ , I + I = . (6.3.23)
cos 2 πγ ⎝ 2γ ⎠ 8a 3(1−γ )γ 3 cos πγ cos 2πγ
6 7
a
All the integrals for coefficients I k can be calculated numerically. They can also be
expressed by hyper geometric functions F (α , β , γ , z ) [Gradstain & Ryzhik 1962]:
4 ⎧ 2γ (1− 2γ ) / 2 sin πγ ⎫
I1 = 2(1−γ ) ⎨ [(1 − γ ) F1 − F2 ] + [ F4 − (1 − γ ) F3 ]⎬ ,
(1 − γ ) cos πγ ⎩ 1 + 2γ 2γ ⎭
2 γ
I4 = F4 ,
π (1 + 2γ )(1 − γ )1−γ
3 ⎧ F6 − (1 − γ ) F5 ⎡ (1 − 4γ )(1 − γ ) 2 F7 ⎤ ⎫
I6 = ⎨ sin 2πγ + γ (1− 4γ ) / 2 ⎢ F1 + F2 − ⎬,
(1 − γ ) 3(1−γ )
cos 2πγ ⎩ 3γ ⎣ 1 + 2γ 1 + 2γ ⎦⎥ ⎭
where F1 = F (γ + 0.5, γ − 0.5, γ + 1.5, γ ), F2: = F (γ + 0.5, γ − 1.5, γ + 1.5, γ ),
F3 = F (2γ , 2γ − 1, 2γ + 1, γ ), F4 = F (2γ , 2γ − 2, 2γ + 1, γ ),
F5 = F (3γ , 3γ − 2, 3γ + 1, γ ), F5 = F (3γ , 3γ − 3, 3γ + 1, γ ),
F7 = F (γ + 0.5, γ − 2.5, γ + 1.5, γ ).
The hyper geometric function may be calculated by the power series expansion
∞ k −1
(α + m)( β + m)
F (α , β , γ , z ) = 1 + ∑∏ z. (6.3.24)
k =1 m = 0 (γ + m)(1 + m)
As inclination angle γ approaches zero, coefficients I1 , I 3 , I 4 , I 6 approach zero but other

coefficients approach as follows: I 2 → π / 2 , I 5 → π / 2 , I 7 → 3π / 8 . This case
corresponds to glissade of the plate with variable length; the leading edge moves with a
speed U but the other edge is fixed.
For perpendicular entry of a wedge as γ = π / 2 all integrals in (6.3.22) are expressed
in simple analytical forms and the hydrodynamic forces and moment are presented as
Y = ρ l (lU − 2U 2 )(α1 + α 2 ) / π ,
X = ρ l (U 2 − lU )[(α1 + α 2 ) 2 / 2 + (α1 − α 2 ) 2 ln 2]/ π ,
(6.3.25)
X 0 = − ρ lU 2 (α1 + α 2 ) 2 / 2π ,
M = ρ l 2 (3U 2 − lU )(α1 + α 2 ) / 6.
For example, consider an entry of a plate and a wedge with flat sides. Let the plate’s
slopes are α1 = α 2 = −α , and wedge’s α1 = −α 2 = α . The following coefficients are used:
CY = 2Y / ρU 2 lα , C X = 2 X / ρU 2lα 2 , C0 = −2 X c / ρU 2lα 2 , CM = −2M / ρU 2l 2α . (6.3.26)
The leading edge force and torque are negative in these examples, thus all the coefficients
are positive now. Parameter s = 1 − CM / CY is the distance of the pressure center from
leading edge related to length l. Dependences of the coefficients are plotted in Figs.6.3.2 and
6.3.3. Note: the scale for dependences of s (γ ) is shown on the right side on both figures.
One can find that at γ → 0 the flat plate and the wedge give the same results: CY → π / 2 ,
C X → π / 4 , CM → 3π / 8 , C0 → π / 4 and s → 1/ 4 . That is because only one side of
thewedge is contiguous to liquid. It should be noted that the lift coefficient is a half of that
234
for a planning plate of a constant length. In another limited case for normal entry, γ → 1/ 2 ,
we have for the flat plate CY → 8 / π , CM → 2 , C0 → 4 / π , s → 1 − π / 4 , and for the wedge
C X → 8ln 2 / π ; all other coefficients CY , CM and C0 vanish, so distance s is not meaningful
but not zero. It should be noted that the nonlinear problem of perpendicular entry of a wedge
with flat sides was considered by using Wagner function in [Dobrovolskaya, 1969].
Fig.6.3.2. Entry of flat plate. Fig.6.3.3. Entry of wedge.
6.3.3. Partial Time Derivative

The above-mentioned approach is quite difficult for studying many problems associated with
unsteady flow with free boundaries due to integration and partial differentiation of very
complex functions. It is preferable, as before, to formulate a value problem for partial
derivative with respect to time, wt ( z , t ) = ∂w( z , t ) ∂t . Below is an example of using this
approach to solve unsteady linear problems of an inclined entry with an attached cavity. The
boundary condition on the moving foil can be rewritten integrating along the x-axis
x
ψ ( x, t ) = − ∫ g ( x, t )dx + T (t ) . (6.3.27)
s (t )
x
Hence, ψ t = sg ( x, t ) − ∫ g ( x, t )dx + T (t ) , (6.3.28)
s
where T is to be found.
The dynamic condition on the free boundary is ϕt = 0 . (6.3.29)
Therefore, function wt ( z , t ) should be a solution of a mixed value problem with conditions
(6.3.28) and (6.3.29). The mixed value problem solution has a certain singularity at the
leading edge, but the stream function should be uninterrupted at that point. The condition of
continuity of stream function at the leading edge develops an integral equation for unknown
function T(t).
6.3.4. Entry of a Flat Plate
Consider an inclined entry of a flat plate with a ventilated cavity at a constant velocity
(Fig.6.3.4a). On the slit bank OB, which has a variable length l(t), the kinematic condition is
235
ψ x = −ϕ y = lα . Due to equation (6.3.27), the stream

function on the plane is in the form:
ψ = lα ( x − l ) + T (t ) (6.3.30)
Hence, the boundary conditions for partial time
derivative of complex potential on the ξ -axis are as
Fig.6.3.4. Entry of plate with follows:
ventilated cavity: (a) - the z-plane; ψ t = T − l 2α , ξ ∈ [0,1] , ϕt=0 outside (0, 1). (6.3.31)
(b) - the parametric ζ -plane.
Besides, function wt has a singularity in the form of
ζ −1 2 at point O, is limited at point B, and approaches zero in infinity. These requirements
⎛ ζ −1 ⎞
are met by the function wt = i (T − l 2α ) ⎜⎜ 1 − ⎟⎟ . (6.3.32)
⎝ ζ ⎠
The same function on the ξ -axis determines partial time derivative of the stream
function outside (0, 1), so that at any moment the stream function itself can be obtained by
integration with respect to time over interval (0, t). Equating with the function for T(t), one
obtains an integral equation in the form
t
ζ −1
Re ∫ (T − l 2α ) dτ = −l 2tα . (6.3.33)
0
ζ
Variable ζ is a function of times t and τ as a result of the equality
s (t ) = s (τ )(1 − ζ )γ (1 + ζ / a ) .
1−γ
(6.3.34)
For constant speed V = l = const , the solution of equation (6.3.33) and all the
hydrodynamic characteristics can be expressed analytically. Passing to new variable σ=τ /t,
one can write both last equations as
1
ξ −1
Re ∫ (T − U 2α ) dσ = −U 2α , (6.3.35)
0
ξ
σ = (1 − ζ ) − χ (1 − ζ / a ) .
χ −1
(6.3.36)
Since function ζ (t , τ ) depends only on variable σ , and equality (6.3.35) should be satisfied
for any value of time, unknown function T − U 2α should be constant. Hence,
. 1
ξ −1
T − U 2α = −U 2α I −1 , I = Re ∫ dσ . (6.3.37)
0
ξ
In view of equality (6.3.36) the integral in (6.3.37) can be submitted through hyper-
geometrical and beta-functions [Gradstain & Ryzhik, 1962] as
∞
I = a1− χ Re ∫ δ 1/ 2 ( eπ (1− χ ) i + δ ) (δ + ae )
− ( χ +1/ 2 ) − πχ i − ( 2 − χ )
dδ =
0
⎧ ⎛3 ⎞ ⎛1 3 5 1 ⎞⎫
= a1/ 2 Re ⎨e −π i / 2 B ⎜ ;1⎟ F ⎜ + γ , ; ; ⎟ ⎬ = (6.3.38)
⎩ ⎝2 ⎠ ⎝2 2 2 γ ⎠⎭
γγ ⎡ ⎛1 ⎞ 1 ⎛ 1⎞ ⎤
= 1/ 2 ⎢
F ⎜ + γ , γ − 1; γ ; γ ⎟ cos (πγ ) + B ⎜ γ ; ⎟ γ 1−γ ⎥ .
(1 − γ ) ⎣ ⎝ 2 ⎠ 2 ⎝ 2⎠ ⎦
236
Resulting force can be found by integrating pressure along the body’s surface:
s
ρU 2α 1 1 − ξ ∂x (ξ , t ) ρU 2 sαγ γ ⎛ 1 3⎞ ⎛ 1 ⎞
Y = ∫ pdx = − ∫ d ξ = 1−γ
B ⎜ γ + ; ⎟ F ⎜ γ , γ + ; γ + 2; γ ⎟ (6.3.39)
0
I 0
ξ ∂ ξ Ia ⎝ 2 2 ⎠ ⎝ 2 ⎠
⎛ 1 ⎞
π F ⎜ γ , γ + ; γ + 2; γ ⎟
2Y ⎝ 2 ⎠ .
Whence its coefficient is Cy = = (6.3.40)
ρU 2 sα ⎛ 1⎞
I (1 − γ ) (1 + γ ) B ⎜ γ ; ⎟
1−γ
⎝ 2⎠
The torque relative to the origin of the fixed coordinate is calculated similarly. Its
⎛ 1 ⎞
π F ⎜ 2γ − 1, 2γ + ; 2γ + 2; γ ⎟
2M ⎝ 2 ⎠ .
coefficient is CM = = (6.3.41)
ρU s α
2 2
( ) ⎛ 1 ⎞
2 (1 − γ ) (1 + 2γ ) B ⎜ 2γ ; ⎟ I
2 1− γ
⎝ 2⎠
Depending on values of arguments it is possible to use various representations of special
functions [Gradstain & Ryzhik, 1962]. In particular, at small values it is necessary to use
other formulas. Namely, equations (6.3.40) and (6.3.41) for γ → 0 yield the above obtained
results for coefficients C X and CM ( π / 2 and 3π / 8 ).
Fig.6.3.5
Dependence of lift coefficient C L and pressure center
s0 on entry angle γ .
Fig.6.3.5 shows dependencies of lift coefficient

CL and the pressure center x0 of a plate on inclination
angle. For γ = 1 , lift coefficient CL = π / 2 and pressure
center x0 = 3/ 4 . This result coincides with those in the
previous section for a wedge when the angle of
inclination approaches zero.
Consider the vertical entry of a wedge of finite
length l into a fluid as shown in Fig.6.3.6. The wedge
sides are equal and form angles α1 and α 2 with the x-
axis. The wedge moves at constant velocity U. The
boundary conditions are:
⎧⎪ −V 2α1 + T , ξ ∈ (−b,0),
Fig.6.3.6. Vertical entry of a ψt = ⎨ 2 (6.3.42)
⎪⎩ −V α 2 + T , ξ ∈ (0, b),
finite-length wedge: (a) the
z-plane ; (b) the ζ plane. ϕ t = 0 outer interval (-b, b).
The value problem can be solved analytically in the form
V 2β ζ 2 − b 2 + ib ⎛ ζ 2 − b2 ⎞
wt = − ln + iV 2α ⎜ 1 − ⎟, (6.3.43)
π ζ ⎜ ζ ⎟
⎝ ⎠
where α = (α1 + α 2 ) / 2, β = α 2 − α1 . (6.3.44)
237
Two stages of the wedge entry should be considered: without and with cavity ventilation.
The former case was considered above with the results in (6.3.25). Nevertheless we shall
consider this problem using the Volterra’s type integral. Then, function (6.3.43) could be
expressed directly in the z variable as
∂w ( z , t ) U 2β z+s ⎛ z ⎞
=− ln + iU 2αμ (t ) ⎜1 − ⎟. (6.3.45)
∂t π z−s ⎝ z −s ⎠
2 2
Whence the stream function of interval ( s, ∞ ) is found as

t ⎞ ⎛ x
ψ ( x, 0, t ) = U α ∫ μ (τ ) ⎜ 1 −
2 ⎟dτ , (6.3.46)
0
⎜
⎝ x 2
− s 2
(τ ) ⎟
⎠
where μ = T V 2α − 1 . Substituting function x = s (t ) and equating to unknown function T,
t
s (t )
we receive the required integral equation: ∫ μ (t )
0 s ( t ) − s 2 (τ )
2
dτ = −t . (6.3.47)
This equation, as Abel’s integral equation, can be solved analytically. If the wedge moves
with a constant speed, then s (t ) = Ut , s (τ ) = Uτ , and after substitution τ = t sin θ , integral
π /2
equation (6.3.47) becomes ∫ μ (t sin θ )dθ = −1
0
with the only solution as μ (t ) = −2 / π .
Consider now a motion of a fully immersed wedge ( s > l ). The stream function on
the x-axis outer wedge is determined from (6.3.43) by substituting ζ = s 2 − z 2 / s directly
⎛ x 2 − ( s (τ ) − l ) ⎞⎟
t
2
2⎜
in the x variable as ψ ( x, 0, t ) = U α ∫ μ (τ ) 1 − dτ + ψ 0 ( x ) . (6.3.48)
t0
⎜⎜ x 2 − s 2 (τ ) ⎟⎟
⎝ ⎠
2U 2α ⎛ x Ut ⎞
Initial stream function at t0 is found from (6.3.46) as ψ 0 ( x ) = − ⎜ t0 − arcsin 0 ⎟ .
π ⎝ U x ⎠
.
Equating function (6.3.48) for x = Ut , s (τ ) = Uτ , one obtains the integral equation in the
t 2 − (τ − t0 )
t 2
⎛ 2 t ⎞
form:
t0
∫ μ (τ )
t −τ
dτ = −t ⎜ 1 − arcsin 0 ⎟ .
⎝ π
2 2
t ⎠
(6.3.49)
This integral equation has an analytical solution only if time approaches to infinity, μ → −1 .
This case corresponds to a wedge flow with attached Kirchhoff’s cavity in boundless
domain. For any time, equation (6.3.49) can be solved numerically by iterations, as above.
Hydrodynamic forces and torque are calculated as usually by integration of pressure
along the wedge. Skipping manipulations, the resulting formulas are:
lift (transverse force)
x2 − ( s − l )
s 2
Y = −2 ρU 2αμ (t ) ∫ dx = −2 ρU 2αμ (t ) ⎡⎣ E (b) − (1 − b 2 ) K (b) ⎤⎦ ; (6.3.50)

s −l
x −s 2 2
concentrated force at the leading edge
238
2
iρ ⎛ ∂w( z , t ) ⎞ ρU 2 lα 2 s 2
2U 2 z∫= s ⎝ ∂z ⎠
X0 = ⎜ ⎟ dz = π μ (t )b 2 ; (6.3.51)
2 l
longitudinal force without concentrated force
s s
ρUl β 2
X 1 = −α1 ∫ p1dx + α 2 ∫ p2 dx = −α Y − R , (6.3.52)
s −l s −l
2
where R is the drag of the symmetric wedge
bs + x 2 − ( s − l )
s 2
2 2
R= ∫ ln dx = ⎡⎣(1 + b ) ln (1 + b ) + (1 − b ) ln (1 − b ) ⎤⎦ .
π l s −1 s −x
2 2 πl
Drag of an asymmetric wedge is equal to D = X 1 + X 0 .
⎛ π b2 ⎞
The pressure center is λ = s ⎜1 − ⎟. (6.3.53)
⎝ 4[ E (b) − (1 − b ) K (b) ⎠
2
When distance s between the wedge and the free water

surface approaches infinity as ( s → ∞ ), all hydrodynamic
parameters approach the limits corresponding to the
Kirchhoff’s model: i.e. lift L∞ = πρV 2l , drag
D∞ = − ρV 2 β 2 l π , and pressure center λ∞ = 1 4 . Fig.6.3.7
shows dependencies of ratios L / L∞ , D / D∞ and λ / λ∞ on
length-to-immersion ratio l/s.
Fig.6.3.7.
Dependence of hydrodynamic parameters of wedge on
submergence depth.
Sec. 6.4. Analytic Extension Method for a Free Surface between Two
Fluids
Analitic extension as a method of solving value problems is widely used in the
hydrodynamics. The method was already used above for solving both nonlinear and
linearized flow problems. Clearly, the method can be effectively used for domains with
boundaries as arches and straight lines where the required function should satisfy certain
conditions, i.e. its imaginary or real parts, or its module, or argument should be constants.
If a fluid consists of two or more components with different densities, then two
conditions on their common free boundary should be satisfied, namely, one is the equality of
normal speeds and the other is the equality of pressures. Thus it is impossible to apply the
analitic extension method directly to such problems. However, using an additional function,
one can apply the analitic extension for crossing over the boundary between two media
[Sheshukov, 1968; Nikitin & Terentiev, 1980, 1985]. Unsteady problems are considered
below using analitic extension only. For simplicity, consider first moving of a thin
symmetric body perpendicular to the free boundary, and then of a wedge with Kirchhoff’s
cavity across the free boundary of two fluids.
239
6.4.1. Movement of a Body Perpendicular to a Free Boundary

The movement is shown in Fig.6.4.1. Flow of both fluids at the left and right sides of the
free boundary is determined by two different
complex potentials w1 ( z, t ) and w2 ( z , t ) ,
respectively.
Fig.6.4.1.
Crossing the free boundary by a wedge with the
Kirchhoff cavity: (a) before crossing;
(b) crossing the free boundary;
(c) after crossing.
Due to the symmetry all real axes can be

considered as the body boundary. In fact, more
than one body can move along the x-axis. In the moving axes the body is given by function
z = f ( x) , then the kinematic condition on the whole x-axis can be written using partial time-
⎧ sf ( x − s ) − s 2 f ′( x − s ), x ∈ ( s − 1, s ), y = 0,
derivatives ψ 1t = ⎨ (6.4.1)
⎩0, x ∈ (0, s − 1) ∪ ( s, ∞), y = 0 ,
ψ 2t = 0, x ∈ (−∞, 0), y = 0 . (6.4.2)
Two conditions should also be satisfied on the y-axis corresponding to the free boundary:
• kinematic condition of equality of velocities normal components (the x-components)
written in the form ψ 1t = ψ 2t , y ∈ (0, ∞), x = 0 ; (6.4.3)
• dynamic condition of equality of pressures
ρ1ϕ1t = ρ 2ϕ2t , y ∈ (0, ∞), x = 0 . (6.4.4)
Consider auxiliary function
⎧⎪ w1t ( z , t ), Re z ≥ 0,
Wt ( z , t ) = ⎨ (6.4.5)
⎪⎩ Aw2t ( z , t ) + Bw1t ( − z , t ), Re z < 0 .
Satisfying conditions (6.4.3) and (6.4.4), one obtains unknown factors
2ρ2 ρ − ρ1
A= , B= 2 . (6.4.6)
ρ1 + ρ 2 ρ1 + ρ 2
Function (6.4.5) is analytical on the upper side of the z -plane including free boundary x=0.
On the real axis the boundary condition following from (6.4.5), (6.4.1) and (6.4.2) are to be
⎧ψ 1t ( x, t ), x ≥ 0,
satisfied Im Wt ( x, t ) = ⎨ (6.4.7)
⎩ Aψ 2t ( x, t ) − Bψ 1t (− x, t ), x < 0 .
This boundary value problem (6.4.7) can be solved by Schwartz’s operator (Appendix 1) as
∞
1 Im W ( x′, t )
Wt ( z , t ) = ∫ dx′. (6.4.8)
π −∞ x′ − z
The required functions are determined by formulas:
w1t = Wt ( z , t ), Re z ≥ 0 , (6.4.9)
1 B
w2t = Wt ( z , t ) − Wt (− z ), Re z < 0 . (6.4.10)
A A
240
Pressure in both fluids is obtained by linearized Cauchi-Lagrange integral as
p1 = − ρ1ϕ1t (Re z ≥ 0), p2 = − ρ 2ϕ 2t (Re z < 0) . (6.4.11)
Intersection of straight and free boundaries of a circular form by some foils are calculated in
[Nikitin & Terentiev, 1980].
6.4.2. Movement of a Foil with an Attached Kirchhoff Cavity
In this case the conditions at solid boundaries (6.4.1) and (6.4.2) are valid but in addition to
them the following dynamic conditions at the cavity boundary should be satisfied:
• cavity located in the first fluid ϕ1t = 0, 0 < s(t ) ≤ x < ∞, y = 0 ; (6.4.12a)
ϕ2t = 0, − ∞ < s(t ) ≤ x < 0, y = 0,
• cavity intersects the free boundary (6.4.12b)
ϕ1t = 0, 0 < x < ∞, y = 0 .
Let's consider again the movement of a body in the first fluid. Denoting the
imaginary part of function w1t ( z , t ) at the unknown boundary of cavity as β and at the solid
boundary of the foil as h, we can express solution (6.4.9) as integral (6.4.8) and then satisfy
the dynamic condition at the cavity boundary:
∞ ∞
β dτ β dτ s hdτ s
hdτ
π Re w1t = ∫ + B∫ + ∫ +B∫ =0. (6.4.13)
s
τ −x s
τ + x s −l τ − x s −l
τ+x
The first term of sum (6.4.13) is an integral of Couchi’s kind and concerning to it the
equation could be inversed, so that a regular integral equation for unknown function β ( x) is
B
∞
β (τ , s ) dτ
obtained as β ( x, s ) + x − s∫ = F ( x, s , B ) , (6.4.14)
π s τ + s (τ + x )
x−s ⎛ hdτ hdτ ⎞
s s
where F =− ⎜∫ −B∫ ⎟. Substituting the new
π ⎜⎝ s −l s − τ (τ − x ) ⎟
s − l τ + s (τ + x ) ⎠
variable x = s + tan 2 (u / 2), τ = s + tan 2 (v / 2) u , v ∈ (0, π ) , (6.4.15)

one can reduce expression (6.4.14) into β (u , s ) = F (u , s ) − ( B / π ) tan(u / 2) J , (6.4.16)
π
β (v, s )sin(v / 2)dv
where J =∫ .
0 2s + tan (v / 2)(2s + tan 2 (u / 2) + tan 2 (v / 2)) cos3 (v / 2)
2
This equation is convenient for calculating iteratively. The initial value of function β (v, s )
could be zero or –1. For comparing the accuracy of calculations, and for substantiating the
initial value of required functions, we will express, after [Terentiev, 1981b], analytical
solutions of separate value problems of wedge movement at constant speed and attached
Kirchhoff’s cavity in boundless fluid ( B = 0 ), and towards a vertical wall ( B = 1 ), and
toward a free surface ( B = −1 ), as follows:
α s2 i z − s + 1
1) boundless domain w0 t ( z , s ) = ln ; (6.4.17)
π i z − s −1
2) moving towards a free boundary
α s 2 ⎛ s 2 − z 2 − 2s − 1 ⎞
w1t ( z , s ) = ln ⎜ ⎟; (6.4.18)
π ⎜ s 2 − z 2 + 2s − 1 ⎟
⎝ ⎠
3) moving towards a solid wall
241
α s 2 ⎛ ( s − 1) s 2 − z 2 − z 2s − 1 ⎞
w2 t ( z , s ) = ln ⎜ ⎟. (6.4.19)
π ⎜ ⎟
⎝ ( s − 1) s − z + z 2 s − 1 ⎠
2 2
The imaginary parts of these functions on interval ( s, ∞) are the exact solutions for integral
equation (6.4.16). As follows from the last function by approaching y → +0 and s → 1 , the
solution is constant and equal to −1 . Fig.6.4.2 shows the solution of integral equation
obtained iteratively for different parameters B. The exact solutions coincide with the curves
for B = −1, 0, 1. The iterative process should be made
as follows: the integrals in (6.4.16) are calculated at N
nodal points un = π (n + 1) /( N + 2) , then function
β (u ) is obtained by interpolation and the next
iteration begins.
Fig.6.4.2. Solutions for s=1.
Drag can be found by integrating pressure (6.4.11)

along the wedge sides. More difficult is to calculate
the cavity boundary with ordinates satisfying
differential equation
∂ 2 y ∂β ( x, t )
= . (6.4.20)
∂t 2 ∂x
The right part of equation (6.4.20) can be calculated from obtained solution β ( x, t ) , so it
should be a double integral with respect to two times. Instead of time the integration can be
made along the s(t)-variable in interval (s, x). If solution β(x,t) is given analytically then the
cavity ordinates are easy to calculate, namely, the ordinates for functions (6.4.17) – (6.4.19)
x x
∂ Im w( x, s2 )
are calculated as: y ( x, s ) = − ∫ ∫ ds2 ds1 (6.4.21)
s s1
∂x
Partial derivative βx(x,s) can be calculated numerically as follows: first, partial derivative
βu(u) is calculated iteratively by determining the corresponding expression from (6.4.16),
then partial derivative β x ( x, s ) = β u (u ) / x − s (1 + x − s ) is obtained and, finally, cavity
boundary (6.4.21) can be found. The iteration allows achieving any given accuracy.
Intersection of the free boundary and the further movement in the second liquid are
considered similarly. Only boundary conditions in the second fluid are determined in view of
the boundary conditions at the moving foil. Integral equation (6.4.16) can also be solved
numerically using expansions in Fourier series [Nikitin & Terentiev 1985]. Nikitin obtained
some calculated data in that manner, which are plotted in Figs.6.4.3 and 6.4.4. Fig.6.4.3
shows dependence of drag coefficient CD = 2 D / ρ1V 2lα 2 on distance s / l for B = 1, 0.5, -
0.5, −1. The curves for B = −1 and B = 1 coincide with the exact lines calculated by analitic
expressions (6.4.18) and (6.4.19). Fig.6.4.4 shows the cavity shapes when crossing the free
boundary of two fluids. Fig.6.4.4 shows that the cavity boundary breaks up at the free
boundary. In a real process, vortices or some other phenomena such as jets and bubbles are
produced. As marked above, the curve for B=1 corresponds to the wedge movement to
242
the solid wall if the wedge is at the right of it, and goes from the wall if the wedge is at the
left. In the latter case in the beginning of movement there is no cavity and, hence, the cavity
area should be equal to zero.
Fig.6.4.3. Drag coefficient versus s/l. Fig.6.4.4. Cavity shapes at crossing free surface.
Sec. 6.5. Momentum and Energy of Impact Upon or Entrance Trough

a Free Surface
6.5.1. Problem Statements for Impact and Entrance
Fig.6.5.1 shows the schematics of impact and entry. Iimpact as a hydrodynamic phenomenon
is characterized by an instantaneous
change of speed in fluid. Formulations of
impact problems of solid bodies against
the free surface of an incompressible
fluid were given by Sedov [1934].
Mathematical statement of the impact
problem can be obtained by limiting
transition in the equations of movement.
Namely, if the speed changes during an
extremely short time interval, the
governing terms in the integral of
movement are local acceleration ∂ϕ/∂t
and pressure p, which become infinitely
great values. As a result, the pulse of
pressure is proportional to potential of
speed at any point: P = − ρϕ . (6.5.1)
Fig.6.5.1. Schematics of vertical impact (a, Now, the impact problem statement can
b) and entry (c, d) of a slender body (a, c)
be formulated for speed potential, which
and a blunt body (b, d). is a harmonic function and satisfies the
following conditions:
• at the free boundary ϕ =0; (6.5.2)
• at the wetted solid boundary ∂ϕ / ∂n = vn . (6.5.3)
Here we shall pay attention, first of all, to problems with free boundaries in the linear theory.
243
As shown above, in the linear theory the pressure is proportional to speed potential’s
partial derivative over time. Thus, potential of speed on the free surface does not depend on
time, and if at the initial moment the potential is equal to zero, it will be zero at any moment,
i.e. in linear theory, condition (6.5.2) at the free boundary is valid also for entry. Besides, the
same kinematical condition (6.5.3) should be satisfied for a moving solid body. Thus the
impact problem coincides with the entry problem of the same foil and, therefore, both
problems are characterized by the same function. With apparent similarities, there are certain
differences in the problems, which can sometimes be latent and difficult for finding,
resulting in mistakes. Some of the impact and entry problems of slender and blunt bodies are
considered below.
6.5.2. Impact of Floating Thin and Blunt Bodies
For simplicity, the bodies are assumed to be symmetric relative to the x-axis for the case in
Fig.6.5.1a and to the y-axis for the case in Fig.6.5.1b. As a result, the impact problems are
reduced to value problems for complex velocity wz = u − iv on a semi-plane, Im z > 0:
⎧ −Vf '( x − h), x ∈ (0, h)
• for Fig.6.5.1a case: v=⎨ , (6.5.4)
⎩Vf '(− x − h), x ∈ (− h, 0)
v = 0 with, (h < x < ∞) ,
• for Fig.6.5.1b case: v = V with ( − a ≤ x ≤ a ) and u = 0 with (a < x < ∞) . (6.5.5)
Solutions of both problems are defined, respectively, by
2V f ' (τ − h )
h
π ∫0 τ 2 − z 2
wz = τ dτ , for Fig.6.5.1a; (6.5.6)
⎛ z ⎞
wz = −iV ⎜1 − ⎟ , for Fig.6.5.1b. (6.5.7)
⎝ z2 − a2 ⎠
Next, one can find velocity potentials for both cases and calculate all the dynamic
characteristics. Namely, the kinetic energy and vertical pulse of fluid are determined by:
ρ
2 C∫
K0 = − ϕϕ n ds, I 0 = − ρ ∫ ϕ nV ds , (6.5.8)
C
where contour C consists of the wetted boundary of foil and the free boundary; nV is a
normal directed towards velocity V. Since the potential vanishes at the free boundary, and
ϕn = ϕ y , nV = − f ′( x − h) at the wetted boundary of thin foil (Fig.6.5.1a), then the kinetic
energy and momentum can be expressed as K 0 = μV 2 / 2 , I 0 = μV , (6.5.9)
where virtual mass coefficient μ, is calculated for the Fig.6.5.1a case as
2ρ x +τ
h h
μ= ∫
π 00 ∫ f ′( x − h) f ′(τ − h) ln
x −τ
dxdτ . (6.5.10)
Namely, for entry of a wedge with top angle 2α , the virtual mass is equal to
4ρ 2 2
μ= α h ln 2 . (6.5.11)
π
For the Fig.6.5.1b case the kinetic energy and vertical pulse are expressed as equations
(6.5.9) with μ = πρ a 2 / 2 . (6.5.12)
244
6.5.3. Entry of a Thin Foil

Value problems for entry of the same bodies (Fig.6.5.1c) have the same conditions as (6.5.4)
and (6.5.5) because of the linearity. Therefore, the complex velocities for entry and for
impact coincide and can be obtained by formulas (6.5.6) and (6.5.7). It would seem that the
kinetic energy and vertical pulse of fluid could be calculated by equations (6.5.9) – (6.5.12).
However, this is not correct. Although value problems of impact and entry are similar, they
differ from each other. Pressure at the free boundary is the same ( p = 0 ) both for impact and
entry, but the potential and its time derivative are equal exactly to zero for the impact, while
for an entry problem they are equal to zero with an accuracy up to a second-order small
value. Noticing that ϕ ∂ ϕ / ∂ n at the free boundary is a small value of a higher order than that
at the wetted boundary of the body, the kinetic energy has the same expressions as K0 in
(6.5.9) with (6.5.10) or (6.5.11). Momentum for the entry of slender bodies must be
calculated by integrating ϕ nV over both wetted and free boundaries so the momentum is
defined as I = I 0 + I1 , (6.5.13)
where I 0 is calculated by (6.5.9) – (6.5.12); I1 is the integral over the free boundary.
For finding the last term in (6.5.13), one should first obtain the time derivative
∞ ∞
[Terentiev, 1987] I1 = 2 ρ ∫ ϕt dy = − ρ ∫ ϕ x2 dy . Since differential dx and partial derivative
0 0
ϕ y are equal to zero on the imaginary axes, equality ϕ x2 dy = Im( wz2 dz ) is valid. Considering
the integral of analitic function wz2 on the closed contour and deforming the integration
h
contour up to imaginary and real axes yields: I1 = 2 ρ ∫ ϕ xϕ y dx . Substituting now the
0
expression for ϕ x from (6.5.6) and the boundary condition for ϕ y from (6.5.4), one can write
4 ρV 2 f ′( x − h) f ′( x′ − h) x′dx′dx
h h
I1 = −
π 00 ∫ ∫ x ′2 − x 2
. Swapping variables x and x’ and summing the
integrals, one can get an integral convenient for numerical calculations:

2 ρV 2 f ′( x − h) f ′( x′ − h)dx′dx
h h
I1 = −
π 00 ∫ ∫ x′ + x
. (6.5.14)
Thus, the additional pulse can be calculated by integration with respect to time at
given velocity. Namely, for a wedge with top angle 2 α the integral is expressed analytically
ρV 2α 2 h
I1 = − 4ln 2 . (6.5.15)
π
If the wedge penetrates with constant speed h=Vt the additional pulse is determined as
ρVh 2α 2
I1 = − 2ln 2 .
π
Hence, I = I 0 + I1 = T / V . (6.5.16)
As shown by Logvinovich [1969], the last identity is valid for a flat-side wedge with
an arbitrary apex angle. Factor (6.5.10) for penetration of foil is a function of time. Its time
derivative can be expressed by integral (6.5.14) [Terentiev, 1987] as
I1 = − μV / 2 . (6.5.17)
245
Thus, the vertical force is X = − μV − μV / 2 , (6.5.18)

where μ is calculated by (6.5.10). The last expression can be obtained from kinetic energy
(6.5.9) using relationship X = −T / V , and also integrating pressure distribution along the
wetted boundary of foil. For the wedge with top angle 2α, the vertical force due to equations
ρα 2
(6.5.11) and (6.5.18) is X =− (Vh 2 + V 2 h)4ln 2 . (6.5.19)
π
It should be noted that in some works (see [Gurevich, 1979]), integral (6.5.14) has
been incorrectly ignored what resulted in a vertical force twice as high as it is in reality.
6.5.4. Water Entry of a Blunt Body
Initial entry of a blunt body (Fig.6.5.1d) is given by equation y = f ( x) , (|ƒ′(x)|<<1), as a
result, products ϕ ϕn and ϕ nV at the free boundary are small values of an order higher than
that at the body wetted boundary. Thus, the kinetic energy and pulse are calculated by the
same equations (6.5.9) and (6.5.12). It seems that equality (6.5.16) for wedge is not fulfilled
again. Wagner [1932] clarified without analytical proof that a part of the energy is contained
in the splashes formed along the wedge sides. This can be precisely proven calculating the
work of the concentrated forces at points x = ± a . The concentrated forces acting on the
ρ ρπ V 2 a
splashes are calculated using Sedov’s formula F± a = i ∫ Wz 2 dz = ± , and so, only
4 z =± a 4
the sprays’ kinetic energy differential is calculated as a function of the forces on differential
ρV 2
da. Hence, dK1 = ( F+ a − F− a ) da = π ada . (6.5.20)
2
Thus, the total kinetic energy is given by K = K 0 + K1 .
Distance a at entry is unknown and must be obtained from integral equation which
satisfies the condition of free boundary attachment to the penetrated surface. The free
boundary ordinate from solution (6.5.7) is determined from kinematical condition as
xdτ
t t
y = ∫ V (τ )dτ − ∫ V (τ ) . On the other hand, the ordinate of moving solid
0 0 x − a 2 (τ )
2
t
boundary is given: yb = f ( x) + ∫ V (τ )dτ . Equating both ordinates at attached point x = a (t )
0
and using new variables a = a (t ) and a′ = a(τ ) yields an integral equation of Volterra’s first
a da′
a
dh
kind for unknown derivative dh/da: ∫ da′
a 2 − a′2
0
= − f (a) . (6.5.21)
This integral equation, as Abel’s equation, can be solved analytically; for that it should be
multiplied by 1/ b 2 − a 2 and integrated in interval (0, b), and then the limits of integration
b b b
f (a )da dh a da
should be changed in double integral: − ∫ =∫ da′∫ . The last
0 b2 − a 2 0
da′ a′ (b − a 2 )(a 2 − a′2 )
2
f (a′)da′
a
2
π∫
integral is equal to π / 2 . Hence, h( a ) = − . (6.5.22)
0 a 2 − a′2
246
If the entry velocity is constant, equation (6.5.22) for a wedge with flat sides yields
K1 = πρV 2 a 2 / 4 so that the total energy is K = πρV 2 a 2 / 2 = μ V 2 = V I , i.e. equality (6.5.16)
is also satisfied.
Using the momentum and kinetic energy laws, one can offer an approximate formula
for calculating the jet configuration. As the jet’s thickness is an infinitesimal as compared
with its length, it can be approximately assumed that speed of particles in a cross section is
the same and does not change along the jet. If U and δ(τ) stand for fluid velocity and splash
jet thickness at time τ and point a, then the following two conditions determine the rate of
U2
changing momentum and kinetic energy: ρδ (U − a )U = F+ a , ρδ (U − a ) = F+ a a .
2
Whence, the speed in splashes is twice as high as that of attached point ( U = 2a ) as obtained
by Wagner [1932]. The thickness is δ = 2 F+ a / ρU 2 at point a. We assume that fluid particles
at point a at time τ move along the solid boundary at constant velocity U(τ). Thus, the
configuration of the splash thickness at time t>τ is calculated by the following equations:
π a (τ )V 2 (τ )
δ= , x(t ) = a(τ ) + (t − τ )U . (6.5.23)
2U 2 (τ )
Namely, the angle of the top of splash can be obtained from the previous equations as
δ (0)
follows: γ= . (6.5.24)
2ta (0) − a (a)
Equations (6.5.23) and (6.5.24) coincide with the results obtained by Wagner [1932] and
Logvinovich [1969] for the entry of wedge only at a constant velocity.
Sec. 6.6. Green’s Functions in Linear Theory

The reader has already noticed that mathematical modeling has been carried based on
boundary value problems of the theory of analitic functions. It allows to derive analytical
solutions in many cases or to offer effective numerical algorithms based on the known
methods of the theory of functions of complex variables. In some cases, analytical methods
can be applied to axisymmetric and even to three-dimensional problems though the latter
requires additional research. Besides analytical methods in the hydrodynamics, other
methods, including those based on Green's integral identities, have been used as well. More
details on Green’s integral relationships are given in Ch. 8 and App. 2. For completeness,
applications of Green function in problems of thin bodies entering water at finite Froude
numbers is briefly discussed here.
Moran [1961] was the first one who considered the effects of finite Froude number
during the water entry and exit of thin bodies. He obtained a calculated data for biconvex flat
and axisymmetric bodies crossing the water surface vertically. Mackie [1962] also
considered an initial value problem of water entry at a finite Froude number. Later, Yim
[1973] obtained analytical and numerical results for many problems of water entry both with
and without gravity taken into account. He used time-depended Green function. Another
approach with application of Green function was offered by Porfiriev [1978]. Within the
framework of this approach a number of problems were considered about the gravity effects
on vertical movement of thin flat and axisymmetric bodies in boundless fluid [Porfiriev,
1981], in a final depth fluid [Galanin at al, 1982], during crossing the free boundary between
247
two liquids [Nikitin & Porfiriev, 1985]. Two of these approaches, slightly modified and
computed anew, are discussed below.
6.6.1. Problem Statement of Entry into Fluid Considering Gravitational Effects
Let the boundary of a moving body is given by equations y=±f(x−s(t)), Fig.6.6.1, which
satisfy all required conditions of smallness.
Fig.6.6.1.
Then the entry problem is determined by
Entry of a
initial-boundary value problem for the speed
wedge into
potential satisfies the following conditions:
gravity
on the y-axis, y = ±0 , x ∈ (0, h) ,
water of
finite depth. ∂ϕ
= ∓ sf ′( x − s )[θ ( x) − θ ( x − s )] ; (6.6.1)
∂y
at the bottom,
x = h, y ∈ (−∞, ∞) , ∂ϕ / ∂x = 0 ; (6.6.2)
at the free boundary, x=0; y ∈ ( −∞,0) ∪ (0, ∞) , ηt − ϕ x = 0, ϕt − gη = 0 (6.6.3)
where g is the gravity acceleration; η=(y) is the free height; θ(x) is the Heaviside function.
Besides, speed potential ϕ(x,y,t) and its time-derivative should vanish at the initial time and
in infinity at any time, ϕ ( x, y,0) = ϕt ( x, y,0) = 0 , ϕ ( x, y, t ) → 0, y → ±∞ . (6.6.4)
Deformation of the free boundary x = η ( y, t ) can be found from the condition at the free
boundary as η = ϕt / g , x = 0 . (6.6.5)
6.6.2. Time-Dependent Green Function

Deformation of the free boundary in flow problems is considered mostly using time-
dependent Green function [Stoker, 1957]. Yim [1973] applied it to solve the entry problem
for infinitely deep water, h = ∞ , in a longitudinal gravity field. Presented below is the Yim’s
solution, modified slightly. Using time-dependent Green function G (ξ , η , τ | x, y , t ) that
satisfies Laplace's equation in the flow domain and initial-boundary conditions on the real
axis as Gtt − gGx = 0, y = 0 , (6.6.6)
G = Gt = 0, t = τ , y = 0 (6.6.7)
and behaves like ln r for x → ξ , y → η with r = (ξ − x) 2 + (η − y ) 2 , one can express the
partial derivative of speed potential via Green’s formula as follows
1 ⎛ ⎞
ϕt ( x, y, t ,τ ) = ⎜ ∫ ϕt Gn ds − ∫ ϕtn Gds ⎟ . (6.6.8)
2π ⎝ C C ⎠
Here boundary C coincides with axes y and x that correspond to the free boundary and the
body’s wetted boundary. Unknown function G can be obtained as follows
G (ξ , η , τ | x, y , t ) = ln(r / R ) + F (ξ , η , τ | x, y , t ) , (6.6.9)
where R = (ξ + x) 2 + (η − y ) 2 , so that the first term in (6.6.9) determines the source
potential under the free boundary. Substituting (6.6.9) into (6.6.6), one obtains the partial
2 gξ
differential equation as Btt − gBx = − 2 , x = 0 . Using integral representation
ξ + (η − y ) 2
248
∞ ∞
ξ
∫0 e and setting F = ∫ T (t , s )e −[ξ + x + i (η − y )]k dk ,
−[ξ + i (η − y )] k
= Re dk (6.6.10)
ξ 2 + (η − y ) 2 0
one obtains an ordinary differential equation at the free boundary, x = 0 ,
Ttt + gkT = −2 g (6.6.11)
with initial conditions T = Tt = 0 for t = τ . A solution of equation (6.6.11) is expressed by
2
function: T (t , k ) = − {1 − cos[ gk (τ − t )]} . (6.6.12)
k
∞ − k (ξ + x )
r e
Thus G = ln − 2 ∫ {1 − cos[ gk (τ − t )]}cos[k (η − y )] dk . (6.6.13)
R 0
k
The speed potential obtained in [Yim, 1973] can be found from (6.6.8) using boundary
1 ⎛ ⎞
t ∞
conditions (6.6.3) and (6.6.4) ϕ ( x, y , t ) = ∫ ⎜ ∫ ([ϕτ ]Gη − [ϕτη ]G ) d ξ ⎟ dτ , (6.6.14)
2π 0 ⎝ 0 ⎠
where square parentheses denotes a jump of the value when crossing the x-axis.
The boundary value of partial derivative [ϕτη ] can be obtained from (6.6.1) but
partial derivative [ϕτ ] is unknown for an arbitrary foil. For a symmetrical wedge, however,
since ϕt and Gη are both symmetrical with respect to the x-axis, the first integral of equation
(6.6.14) vanishes, and the speed potential and all dynamic parameters could be calculated
completely. From the boundary condition (6.6.1), one obtains
[ϕτη ] = −2 sf ′( x − s ) + 2s 2 f ′′( x − s ) − 2 s 2 f ′( x − s)δ ( x − s ), x ∈ (0, s), y = ±0 , (6.6.15)
where δ is the Dirac delta function, which behaves like
∞
∫ F ( x)δ ( x − s)dx = F (s) .

0
(6.6.16)
The potential can be calculated by double integrating with respect to ξ∈(0,x) for η=0 and to
time τ∈(0,t).
Let a flat-sided wedge with open angle 2α moves at a constant unit speed, so that
the submergence depth is s = t . Taking into account (6.6.15) and (6.6.16), potential (6.6.14)
α t
π ∫0
can be written as ϕ= G (τ ,0,τ | x, y, t )dτ , (6.6.17)
α⎛ t
⎞
whence ϕt = ⎜ G (t ,0, t | x, y, t ) + ∫ Gt (τ ,0,τ | x, y, t )dτ ⎟ . (6.6.18)
π⎝ 0 ⎠
Now, the dynamic pressure on the wedge side can be calculated by
α⎛ t+x t ∞
g − k ( x +τ ) ⎞
p ( x, t ) = − ρϕt = ρ ⎜⎜ ln + 2 ∫ dτ ∫ e sin[ gk (t − τ )]dk ⎟⎟ . (6.6.19)
π ⎝ t−x 0 0
k ⎠
Integrating with respect to time yields the final expression for pressure,
α ⎛ t + x ∞ e− kx ⎛ ge + kg sin( kg t ) − ⎞ ⎞⎟
− kt
p ( x, t ) = ρ ⎜ ln + 2∫ ⎜ ⎟ dk . (6.6.19a)
π ⎜ t−x k (k + g ) ⎜⎝ − g cos( kg t ) ⎟ ⎟
⎠ ⎠
⎝ 0
249
Results of computations using two last equations are plotted in Figs.6.6.2 and 6.6.3.They
show pressure coefficient C p = 2 p / α ρ s 2 at the wedge and a ratio of drags D = D / D0 for
Fig.6.6.2. Pressure distribution at a wedge Fig.6.6.3. Drag ratio versus Froude

entering into water. number.
different Froude numbers Fr = s 2 / g s .Drag D is calculated by integrating the pressure
distribution along the wedge side; drag D0 = ρ ss 2α 2 4ln 2 / π corresponds to weightless
fluid. High accuracy is obtained even when the upper limit is equal to 50 instead of infinity.
As seen in Fig.6.6.2, pressure is not equal to zero at point x = 0 due to rising water level at
the wedge. The water level can be calculated using equation (6.6.5), or water level at the
wedge using equation (6.6.19a): η (0, t ) = − p (0, t ) / g . Fig.6.6.3 shows that the dynamic drag
converges fast to the drag at infinite Froude number.
6.6.3. Fourier Transformation
Speed potential is expressed as a sum [Galanin et al, 1982] ϕ = ϕ0 + ϕ1 , (6.6.20)
where first term is the speed potential of entry into weightless liquid satisfying conditions
(6.6.1) and (6.6.2) and condition ϕ0 = 0 at the free boundary
s
s (t )
2π ∫0
ϕ0 = f ′(ξ − s )G ( x, y, ξ )d ξ , (6.6.21)
cosh(π y / 2h) + cos(π u / 2h)

with G ( x, y, ξ ) = G0 ( x − ξ , y ) − G0 ( x + ξ , y ), G0 (u , y ) = ln .
cosh(π y / 2h) − cos(π u / 2h)
The second term ϕ1 ( x, y, t ) satisfies Laplace’s equation in the flow domain and the boundary
conditions as follows: ϕ1 y ( x,0, t ) = 0, x ∈ (0, h) , (6.6.22)
ϕ1x (h, y, t ) = 0, y ∈ (−∞, ∞) , (6.6.23)
ϕ1tt (0, y, t ) − gϕ1x (0, y, t ) = gϕ0 (0, y, t ), y ∈ (−∞, ∞) (6.6.24)
ϕ1 → 0, y → ±∞,
. (6.6.25)
ϕ1 (0, y,0) = 0, ϕ1t (0, y,0) = 0
Using Fourier even transformation
250
∞
2
ϕ1 =
π ∫ [ A(k , t ) cosh kx + B(k , t )sinh kx]cos ky dk ,
0
(6.6.26)
one can satisfy condition (6.6.22). Satisfying the second condition of (6.6.23) B=−Atanhkh.
Substituting functions (6.6.21) and (6.6.26) into equation (6.6.24), an ordinary differential
equation is obtained as Att + ω 2 A = F (t ), ω = kg tanh kh , (6.6.27)
with initial conditions A(k ,0) = At (k ,0) = 0 obtained from equation (6.6.25) and with the
s
gs
cosh kh ∫0
function F (k , t ) = f ′(ξ − s ) cosh k ( h − ξ )d ξ . The required solution of equation
t
1
ω ∫0
(6.6.27) is expressed in integral form A(k , t ) = F (k ,τ )sin ω (t − τ )dτ . (6.6.28)
Now, the second function in (6.6.20) is known. After manipulations, pressure

coefficient with respect to potential (6.6.26) can be expressed as follows
4 g cosh[k (h − x)] ( cosh[k (h − s )] − cosh(hk )cos(ω s ) )
∞
π ∫0
C ′′p = dk +
k cosh(hk )[k cosh(hk ) + g sinh(hk )
∞
(6.6.29)
4 cosh[k (h − x)]ω sin(ω s )
+ ∫ dk .
π 0 k[k cosh(hk ) + g sinh(hk )
The drag coefficient due to gravity is represented as
8 g ( cosh[ k (h − s )] − cosh(hk ) ) cos(ω s )
∞
π h ∫0
′′
CD = Fdk +
cosh(hk )
∞
(6.6.30)
8
( sinh[k (h − s)] − sinh(hk ) ) ω sin(ω s) Fdk ,
π h ∫0
+
sinh[k (h − s )] − sinh kh
where F= .
k 2 [k cosh kh + g sinh kh]
Using another approach Kuznetsov and Manevich [1979] obtained results similar to
equations (6.6.29) and (6.6.30), which can be transformed asymptotically to Yim’s results.
The water entry without gravity force is determined by speed potential (6.6.21). Coefficients
of pressure and drag in this case are calculated by
s
1 2
π s ∫0
C ′p = G ( x,0, s ), C 'D = G ( x,0, s )dx . (6.6.31)
π
The drag coefficients of a flat-sided wedge entering water are computed using equations
(6.6.31) and (6.6.30). Fig.6.6.4 shows the effect of immersion depth on drag of the wedge
entering weightless water. Coefficient CD′o = CD′ (h → ∞) = 8ln 2 / π corresponds to infinite
depth; coefficient CD1 = CD′ ( s = h) = 2.970 corresponds to the case of the wedge top touching
the bottom. The effect of Froude numbers on drag coefficient CD′′ is shown in Fig.6.6.5.
In conclusion, it should be noted, that the full drag of the wedge consists of dynamic
drag, which is proportional to the angle squared, α 2 , and hydraulic pressure ph = ρ gx ,
which is proportional to gravity acceleration g and angle α . Thus, the dynamic pressure
251
and dynamic drag should be taken into account if Fr ≈ 1/ α , i.e. Froude number should be
quite large. Otherwise it suffices to take into account the hydraulic pressure only.
Fig.6.6.4. Effect of depth on drag for a Fig.6.6.5. Gravity effect for a wedge
wedge entering a weightless fluid. entering finite depth water.
Sec.6.7. Unsteady Movement of a Wing of Finite Span

Green functions and equations are available for investigating a finite span foil as well. In
spite of numerous works on finite span foils neither a theory nor experiments have led us to
full understanding of the flow problem involved. Considered below is a finite span wing in
the water, which is one of great challenges of the hydrodynamics.
6.7.1. Problem Statement of Unsteady Movement of a Foil of Finite Span
Let a thin wing of any form in the plan view moves in
incompressible fluid at an angle of attack β, with speed
V(t)=ds(t)/dt. The speed depends on time and form of foil. The
sketch of foil is shown in Fig.6.7.1. The coordinates of foil, both
fixed and moving, are marked as ( x′, y ′, z ′ ) and ( x, y, z ),
respectively. Let the foil is described by function z = f ( x, y , t ) in
domain Σ bounded by leading edge xa=a(y,t) and trailing edge
xb=b(y,t). The wing is assumed to be thin, thus the necessary
conditions of mathematical problems can be satisfied at z=0. Here
non-linear terms with respect to small partial coordinates and time Fig.6.7.1.
derivatives may be ignored. The flow around a foil can be Sketch of a finite
determined by potential φ ( x′, y ′, z ′, t ) in the fixed, or span wing moving.
Φ = Vx + ϕ ( x, y , z , t ) in the moving coordinates. The coordinates
are: x’=s+x; y’=y; z’=z, thus φ ( x′, y′, z ′, t ) = ϕ ( x − s, y, z , t ) and time derivative is
∂ ∂ ∂
φ ( x′, y ′, z ′, t ) = Nϕ , N = + V (t ) (6.7.1)
∂t ∂x ∂t
Time derivative φt , or function θ ( x, y, z , t ) = Nϕ ( x, y, z , t ) , is called acceleration
potential; the pressure of fluid is linearly related with acceleration potential,
252
p = ρθ / V 2 + const . (6.7.2)
The flow problem statement has the following conditions:
• the continuity condition is satisfied on the wetted part of wing: ϕ z = Nf ( x, y, t ) , (6.7.3)
or θz = N 2 f ; (6.7.4)
• on the free surface, θ = const , x, y, z ∈ C . (6.7.5)
The free surface may be a water surface, or a cavity surface, etc. Due to boundary
conditions (6.7.4) and (6.7.5) the integral equation for the acceleration potential can be
written in a form convenient for obtaining function θ . As mentioned above in in this section
the speed potential is obtained to any function T ( y , t ) , which could be calculated from an
additional integral equation derivable from condition (6.7.3). The speed potential is related
to acceleration potential by inverse operator N −1 as follows
t
ϕ ( x, y, z, t ) = N −1θ = ∫ θ ( x − s + s′, y, z , t ′) dt ′ , (6.7.6)
t0
where s and s′ are given functions of t and t ′ , respectively. If a foil moves with a constant
velocity then Eq. (6.7.6) can be transformed to [Panchenkov, 1975]
1 x ⎛ ξ −x⎞
ϕ = ∫ θ ⎜ ξ , y, z, t + ⎟ dξ . (6.7.7)
V −∞ ⎝ V ⎠
6.7.2. Integral Equations for Unsteady Movement of a Wing

An integral equation is created from Green equation
∂
θ = − ∫∫ [θ ] G1 ( x, y, z; ξ ,η , ζ , s )d ξ dη + ∫∫ [θζ ]G1 ( x, y, z; ξ ,η , ζ , s )d ξ dη , (6.7.8)
Σ ∂ζ Σ
1 ⎡1 ⎤
where G1 = − ⎢ + G0 ( x, y , z; ξ ,η , ζ , s ) ⎥ , R = ( x − ξ ) 2 + ( y − η ) 2 + ( z − ζ ) 2 .
4π ⎣ R ⎦
The regular part of Green function G0 is obtained taking into account other
conditions, for instance, free surfaces of fluid or fixed straight boundary, etc. The second
integral in (6.7.8) appears due to thickness of wing. The kinematical condition (6.7.3) should
be satisfied, which can be written in integral form [Mikhailov, 1986],
x
M (∗) = N ∫ (∗)dx′ . (6.7.9)
a( y)
Using this operator to both terms of equation (6.7.3) one may obtain an additional equation
M ϕ z = MNf ( x, y, t ) . (6.7.10)
Equations (6.7.8) and (6.7.10) allow one considering a wide range of linear problems of
unsteady motion of foil though the analyses and calculations are associated with complicated
expressions. Many problems have been considered in [Mihkailov, 1986 -1988]. Some of
them are represented below.
6.7.3. Motion of a Wing with Attached Cavity in an Unbounded Liquid
Let a wing with attached cavity moves with unsteady velocity V (t ) . The cavity is attached
along the leading edge, a ( y ) , and trailing edge, b( y ) , and is also closed along unknown line
253
c( y ) . The cavity surface and the wetted surface of wing are denoted C and Σ , respectively.
The boundary conditions are
θ ( x, y, ± 0, t ) = V 2σ / 2 on C and ϕ z ( x, y, −0, t ) = Nf ( x, y, z, t )| z →−0 on Σ . (6.7.11)
The above-mentioned integral equations in this case look as
1 ∂ γ ⎡ ( x − ξ )2 + ( y − η )2 ⎤
∫∫ ⎢1 + ⎥ + Q ( x, y , t ) = − g ( x, y , t ) − T ( y , t ) , (6.7.12)
4π ∂y Σ y − η ⎢ x −ξ ⎥
⎣ ⎦
1 ∂ y −η Q (ξ ,η , t )
γ ( x, y , t ) + ∫∫ = V 2σ , (6.7.13)
2π ∂y C x − ξ ( x − ξ ) 2 + ( y − η ) 2
1 ∂ t ⎡∂ γ ⎛⎜ ( x − s + s′ − ξ )2 + ( y − η )2 ⎞ ⎤
where T ( y , t ) = − ∫ dt ′ ⎢ ∫∫ 1 + ⎟d ξ dη ⎥ ,
4π ∂t −∞ ⎢ ∂y Σ y − η ⎜ x − s + s′ − ξ ⎟ ⎥
⎣ ⎝ ⎠ ⎦ x=a( y )
Q( x, y , t ) = Mq( x, y.t ) ; γ = [θ ] ; q = [θ z ] .
The source intensity, q, should also satisfy the integral condition
c ( y ,t )
∫ q(ξ , y, t )d ξ = 0 . (6.7.14)
a ( y ,t )
Similar equations but in other forms have been obtained by many authors [Efremov
& Roman, 1973; Efremov, 1974; Panchenkov, 1975]. Theoretical and experimental results
of cavitating finite-span wings have been obtained in [Kermeen, 1961; Schiebe & Wetzel,
1961; Widnall, 1966; Nishiyama, 1970; Leehey, 1973]. The latter author used the Matched
Asymptotic Expansion. The numerical results for rectangular foil with unit chord plotted in
Figs.6.7.2 and 6.7.3, are reproduced from [Roman &Makaseev, 1984] and [Efremov &
Roman, 1973]. Both have been computed by the method of discrete singulars which was
proposed by I.I. Efremov. A foil under free water surface at submersion depth h was also
considered in [Efremov & Roman, 1973]. Comparisons between numerical and experimental
investigation for rectangle flat wing show that the results are distinctly different, and so the
problem should be studied in more detail.
Fig.6.7.2. Rectangular wing for aspect Fig.6.7.3. Comparison of numerical and

ratios λ = 2c / d : λ = 2,3,8, ∞ experimental result; h is dimensionless depth
of immersion.
254
6.7.4. Vertical Entry of a Rectangular Plate into a Fluid

The problem is similar to the wing without cavity. In this case, the Green function is
1
G= (1/ R1 − 1/ R ) , (6.7.15)
4π
where R 2 = ( x − ξ ) 2 + ( y − η ) 2 + ( z − ζ ) 2 , R12 = ( x + ξ − 2 s ) 2 + ( y − η ) 2 + ( z − ζ ) 2 .
Asymptotic formulas have been obtained in [Mihkailov, 1986] for a rectangular wing with
high aspect ratio of wetted part, λ = 2c / s >> 1 , where 2c is the wing length, s = Vt is the
immersion depth. E.g. the lift coefficient is
⎡ ⎛ 16 ⎞ ⎛ 4 ⎞⎤
CL = L / ρV 2 cs = πλα ⎢1 − exp ⎜ 3 3 ⎟ erfc ⎜ ⎟⎥ , (6.7.16)
⎣ ⎝π λ ⎠ ⎝ π π ⎠⎦
x
2 −t 2
π ∫
where erfc( x) = 1 − Φ ( x) , Φ ( x) = e dt is Frenel integral [Gradstein&Ryzhik, 1962].
0
For high aspect ratios expanding in power series yields
8⎡ 2 32 ⎤
CL / α = ⎢1 − + 3 2 ⎥ + O(λ −3 ) . (6.7.17)
π ⎣ πλ 3π λ ⎦
The center of pressure is the same as in two-dimensional problem of entry of a flat plate,
x0 = − M y / L = (1 − π / 4) s .
6.7.5. Inclined Entry of a Foil

The sketch of entry of a finite span foil into water is
shown in Fig.6.7.4. The problem at each section is
divided into two stages: the first is the entry of a foil
section only and the second is the entry of the fully
submerged foil, with attached cavity past them. The
acceleration potential is as in Eq.(6.7.8) while the
governing integral equation (6.7.10) for a jump of
acceleration potential, γ = [θ ] , is represented as
∂
∂y ∫∫
γ (ξ ,η , t )G ( x, y, ξ , s )d ξ dη =
Σ (6.7.18)
Fig.6.7.4. Sketch of entrance. = − g ( x, y , t ) − T ( y, t ),
x2
∂
where g ( x, y , t ) = V (t )ϕ z ( x, y, t ) +
∂t ∫ ϕ z (ξ , y, t )dξ ,
x1
∂
t ⎡∂ ⎤
∂t ∫ ⎢ ∂y ∫∫′
T ( y, t ) = − dt ′ ⎢ γ (ξ ,η , t ′)G (α − s + s ′, y , ξ ,η , s ′) d ξ dη ⎥
0 ⎣ Σ ⎥⎦
α = x1
Here, Σ′ = Σ(t ) shows that the foil can change its form; x1 = a( y, t ) is the form of the wing
leading edge in plan view; x2 = b( y, t ) is the form of the trailing edge (or an intersection
point with free boundary). The Green function is expressed as
G ( x, y, ξ ,η , s ) = G1 + G2 , (6.7.5)
255
⎡ ( x − ξ )2 + ( y − η )2 ⎤
1 ⎢1 + ⎥,
where G1 =
4π ( y − η ) ⎢ x −ξ ⎥
⎣ ⎦
⎧ ⎡ ( s − x) cos 2β − ( s − ξ ) ⎤⎫
⎪ ⎢ + ⎥⎪
y − η ⎪ cos 2 β X0
G2 = −
1 ⎢ ⎥⎪ ,
⎨ ⎢ ( s − x) cos 2β − ( s − y ) cos 2 2 β ⎥ ⎬
4π ⎪ Y0 ( y − η )2 + X 0 ⎢+ ⎥⎪
⎪ Y0 ⎪
⎩ ⎣⎢ ⎦⎥ ⎭
X 0 = ( s − x) 2 + ( s − ξ )2 − 2( s − x)( s − ξ ) cos 2 β , Y0 = ( y − η ) 2 + ( s − ξ )2 sin 2 2β .
Function G2 accounts for the free surface.
A number of asymptotic expressions which are rather complicated have been
obtained by [Mihkailov, 1987; 1988]. Plotted in Figs.6.7.5 and 6.7.6 are some of his results.
The curves in Fig.6.7.5 show the lift coefficient of a rectangular plate in the first stage of
entry. Aspect ratio λ = ∞ corresponds to two-dimensional problem (see Sec.6.3). Dashed
curves in Fig.6.7.5 are calculated for aspect ratio
λ = 6 . Fig.6.7.6 shows the dimensionless lift,
ψ = L( s,τ ) / L(∞, ∞) , of a rectangular flat wing
entering at angle β = π / 2 as a function of
penetration depth τ = s / l ; τ ≤ 1 corresponds to
the first stage, τ > 1 – to second stage when the
wing in fully submerged. A ventilated cavity is
attached at the leading edge as well as at trailing
edge.
Fig.6.7.5.
Lift coefficient of rectangular wing entering at
angle β .
Fig.6.7.6.
Dependences of dimensionless lift
ψ = L( s,τ ) / L(∞, ∞) and dominant function ζ on
dimensionless penetration depth τ = s / l ;
τ ≤ 1 corresponds to first stage,
τ > 1 – to second stage of fully submerged wing.
6.7.6 Pitch Oscillation of a Wing of Finite Span

This section poses, without mathematical formulation, experimental results for a wing of
finite span oscillating in pitch. First, the basic features of steady flow past a supercavitating
wing of finite span are shown in Fig.6.7.7 [Wosnik & Arndt, 2009].
256
Fig.6.7.7.
The basic flow features of a supercavitating wing of
finite span penetrating a nominally flat free surface.
(Image courtesy of R.Arndt.)
Specifically, Fig.6.7.7 shows a wedge-type

wing in a partially cavitating condition. A partial
sheet cavity originates at the sharp leading edge of
the wing on the suction surface. For this
configuration at the low angle of attack shown, a base cavity also forms at the downstream
end of the wedge. (If conditions produced a longer leading edge cavity, these two features
would merge into a single supercavity extending past the suction surface of the foil.)
In Fig.6.7.7, three-dimensional effects produce the strong tip vortex system that is
apparent in the figure. It can be seen that, along the tip of the pressure surface of the foil
itself, flow driven from the pressure to the suction side of the wing generates a small
vaporous cavity, which merges with the far more dominant cavitating tip vortex that
originates near the corner where the wing pressure surface, the squared-off tip surface, and
the base cavity surface that detaches from the trailing edge of the pressure face all meet.
In the case presented, the wing penetrates a free
surface, which here is nominally flat, but would be nominally
axisymmetric if the wing represented the fin of a
supercavitating body, or the strut supporting an axisymmetric
body in a water tunnel or tow basin. Due to the pressure
difference across the lifting surface representing the wing, the
flat free surface that it penetrates undergoes a deformation:
streamlines are deflected toward the tip on the suction surface
(as can be seen in Fig.6.7.7) and away from the tip on the a
pressure surface. It should be noted that, if the wing under
consideration featured a rounded leading edge, the flow would
be even more complicated, since the locus of cavity
detachment would be established by a balance of several
effects, including surface tension, as is discussed in detail in
Ch. 11.
Fig.6.7.8.
Three frames from a water tunnel experiment with a wing of
finite span that is oscillating in pitch. (Images courtesy of E.A. b
McKenney and C.E. Brennen.)
A series of frames from an unsteady water tunnel

experiment is presented in Fig.6.7.8 [Hart, et al. 1990;
McKenney & Brennen, 1994; Reisman et al., 1994; Reisman
et al., 1998]. In this case, the wing features a rounded leading
edge, and it does not penetrate a free surface; rather, it is
mounted to the wall of a water tunnel in an arrangement that
allows it to pivot about a point along its chord. The three c
257
frames shown are from an experiment in which the wing was forced to oscillate in pitch. It is
immediately clear that the tip vortex is present in each frame, and the direction of its
circulation does not change from frame to frame, indicating that the lift on the wing, while
varying, was in the same direction over the course of the period in which the three frames
were acquired.
The behavior of the flow along the outboard edge of the sheet cavity is particularly
intriguing: in the first frame it can be seen that a strong re-entrant flow system is produced
that – while very three-dimensional in nature – is clearly visible through the cavity boundary
down along the span, to the point where the re-entrant jet is dominated by flow in the
upstream direction and significant jet-cavity boundary interaction clouds the image. The
effects of the tunnel wall at the base of the wing are likely to play a role in this region,
although the same effect would be expected near mid-span on a symmetrical wing in free
flow. In the following two frames, the angle of attack decreases, the cavity shortens toward
the leading edge, and the strength of the tip vortex decreases, as is evident from the visible
decrease in the tip vortex core size. The cavity closure region evolves into a cloud cavity,
which dissipates as the angle of attack decreases and the sheet cavity that feeds the closure
region diminishes in strength.
Modeling of such an unsteady, three-dimensional flow is particularly challenging,
involving the interacting dynamics of the sheet cavity (which detaches in a nonlinear fashion
from the rounded leading edge, as is discussed in Ch. 11), the three-dimensional re-entrant
flow under the cavity, the tip vortex and any associated cavitation, and the evolution of the
sheet cavity into a cloud cavity that is advected downstream in a transient manner. Finally,
there are local effects that are clearly apparent in this sequence of images, notably the V-
shaped features originating at the leading edge of the wing that are most prominent in the
second frame. Signficant progress in the modeling of unsteady cloud cavitation from fixed
hydrofoils has been made recently using advanced methods of grid-based computational
fluid dynamics. (See, for example, [Schnerr et al., 2008; Schmidt et al., 2009].) A laboratory
photograph of cloud cavitation in the flow past a fixed hydrofoil is shown in Fig.6.7.9
[Wosnik et al., 2006; Arndt, 2010];
note that, although the body is fixed,
cloud cavitation is an inherently –
even violently – unsteady process.
Fig.6.7.9.
Cloud cavitation in the flow past a
fixed hydrofoil. (Image courtesy of R.
Arndt.)
258
Ch. 7. Approximation Method
Ch. 7. Approximation Methods

Asymptotic methods used in the previous two chapters give exact or accurate solutions for
nonlinear value problems. Exact or accurate are understood here in the sense that integrated
values can be obtained with any accuracy by decreasing the small parameters. However,
there are many hydrodynamic problems, such as air effects, evaporation from cavity
boundary, turbulent wake past cavity, and so on, which are too complicated even for
formulating a problem and more for solving it. Approximation methods are used only to
obtain some results in that case. Some problems using different approximations are
considered below. Inasmuch as unsteady moving of a body will be considered further, it is
desirable first to solve a time-depended value problem in a nonlinear statement. These
methods could be helpful when investigating unsteady analytically the movement of a body.
Sec. 7.1. Dropping of a Plate on a Flat Bottom - Exact Solution

7.1.1. Separation of Variables in Boundary and Initial Value Problems
Pressure in fluid is determined by Cauchy – Lagrange integral
p = − Re wt − 0.5wz wz + C (t ) , (7.1.1)
where partial derivative with respect to either time wt , or complex variable z, wz , should be
found. Even just finding these functions is a challenge. The fact is that complex potential or
complex velocity is determined analytically as a function of an auxiliary variable and other
parameters, which depend on z and t by complicated relationships that makes computing
quite difficult. A reader could become acquainted with the former problem in Sec. 6.3. It is
therefore expedient to determine functions wz and wt separately as the solutions of
appropriate boundary value problems [Terentiev, 1981b].
Suppose the deformation of an impermeable contour γ is given by equations
x = xγ ( s, t ) , and y = yγ ( s, t ) ; then Neumann condition for the complex velocity on contour
∂ϕ ∂yγ ∂xγ
γ takes the form = xγ − yγ (7.1.2)
∂n ∂s ∂s
whence the stream function on contour γ is obtained integrating the right-hand term with
⎛ ∂yγ ∂xγ ⎞
respect to curvilinear abscissa s ψ = f ( s, t ) = ∫ ⎜ xγ − yγ ⎟ ds . (7.1.3)
s⎝ ∂s ∂s ⎠
Differentiating the latter with respect to time, we find the total derivative:
ψ ( xγ , yγ , t ) = ψ x xγ + ψ y yγ + ψ t .
From this we obtain the following boundary condition for the partial time-derivative
ψ t = ϕ y xγ − ϕ x yγ + f ( s, t ) . (7.1.4)
Thus the solution of a unsteady flow problem can be split in two separate value
problems: 1) Neumann problem (7.1.2) for complex velocity wz and 2) Schwartz problem
(7.1.4) for partial time-derivative wt = ϕt + iψ t , thus the speed problem should be solved
first, then condition (7.1.4) found and the second value problem solved. These functions may
have singularities in the flow domain and at the boundary which must be consistent with one
another. Some singularities of complex velocity have been considered above. Suppose at
259
point a the complex velocity has a power singularity as wz ≈ A( z − a ) μ . From this, after
integration over z and differentiation with respect to t we have wt → − aA( z − a ) μ as z → a .
Hence, lim( wt / wz ) = − a . (7.1.5)
z →a
It follows from this that for a mobile singular point a ≠ 0 , functions wz and wt have the
same singularities with coefficients differing by factor − a ; if the singular point is stationary,
then function wt at this point is bounded while the complex velocity may have a singularity
if μ < 0 . Based on boundary conditions (7.1.2) and (7.1.4), derivatives wz and wt can be
determined for arbitrary configuration bodies. Exact analytical solutions can be found for
some bodies (mostly shaped with straight lines) [Terentiev, 1981b, 1989]. One such solution
is presented in the next sub-section below.
7.1.2. Vertical Motion of a Plate near a Wall
Suppose a flat plate parallel to the bottom moves vertically with variable velocity V(t),
Fig.7.1.1a. The distance from the plate to the bottom is denoted by h, the plate length is 2l. It
is assumed that the flow around the plate edges is continuous. Due to (7.1.2) – (7.1.4) the
boundary conditions at the plate and bottom take the forms:
• at the plate ∂ϕ / ∂y = V , ψ = −Vx, ∂ψ / ∂t = −Vx − V ∂ϕ / ∂x , (7.1.6)
• at the bottom ∂ϕ / ∂y = 0, ψ = 0, ∂ψ / ∂t = 0 . (7.1.7)
Moreover, derivatives wz and wt at endpoints A and B have singularities of the form,
( z ± l ) −1/ 2 . The thirst equation in (7.1.6) shows
that function wt can be found from the value
problem if the velocity function is known.
To solve the boundary value problems
(7.1.6) and (7.1.7), the upper half z plane with a
horizontal cut is mapped onto the rectangle’s
interior on the parametric ζ -plane, Fig.7.1.1b. Fig.7.1.1.
The horizontal side of the rectangle is equal to π , Plate above the
while its vertical side is equal to πτ i / 2 . Points A bottom (a) and
and B are mapped with ζ = a and ζ = π − a . the parametric
The mapping function is periodic with rectangle (b).
period π . Its imaginary part is equal to h on the
rectangular bottom and vanishes at the upper side.
Besides, the mapping function has a simple pole at C (ζ = π 2 + iπτ 2) which corresponds
to infinity of the z-plane. Transformation from the z-plane to the ζ -plane can be expressed
in terms of theta function with parameter τ i : z = h ( i + ln ′ϑ3 (ζ ) ) . (7.1.8)
Conformality is violated at points A and B so that conditions z ′(a) = 0 , z ′(π − a ) = 0 must be
satisfied. The half length of plate is l = z(a) − z(0). These conditions constitute a nonlinear
system of equations for a and τ ln ′ϑ3 (a ) = l / h , ln ′′ϑ3 (a ) = 0 . (7.1.9)
According to conditions (7.1.6) and (7.1.7), at rectangle’s horizontal sides, the
complex velocity has values similar to those of the mapping function, but unlike it has two
simple poles at points A and B.
260
⎧ 1 ⎫
Thus, wz = −V ⎨i + [ ln ′ϑ1 (ζ − a ) + ln ′ϑ1 (ζ + a )]⎬ . (7.1.10)
⎩ 2 ⎭
Expression (7.1.10) is very difficult to integrate with respect to z and then differentiate with
respect to t because parameters a and τ and even variable ζ depend on z and t. Therefore, it
is much easier to solve the value problems (7.1.6) and (7.1.7) separately. Taking into account
(7.1.8), boundary conditions (7.1.6) and (7.1.7) are applied on the rectangle’s horizontal
sizes as follows: ψ = 0 for η = πτ / 2 and ψ = −Vh ln ′ϑ3 (ξ ) for η = 0 . (7.1.11)
The complex potential may be solved using Schwartz integral operator. In this case,
the problem can be solved analytically without integration. Consider the auxiliary function:
W (ζ ) = w(ζ ) + Vh { i ln ′ϑ3 (ζ ) + [ln ′ϑ3 (ζ )]2 } .
Its imaginary part vanishes at both horizontal sides of the rectangle and it has a pole of
second kind at point C. Continuing analytically through the horizontal sides, we obtain an
elliptic function of periods π and πτ i . It can be expressed via theta function due to
singularities [Whittaker & Watson, 1927] as W (ζ ) = A + B ln ′ϑ3 (ζ ) + C ln ′′ϑ3 (ζ ) .
Unknown coefficients can be determined using the equality of leading parts of both sides.
⎪⎧i ln ′ϑ3 (ζ ) + 0.5[ln ′′ϑ3 (ζ ) + (ln ′ϑ3 (ζ )) − ⎪⎫
2
Thus w = − hV ⎨ ⎬. (7.1.12)
⎩⎪ − ln ′′ϑ2 (ζ ) − ln ′′ϑ3 (ζ ) − ln ′′ϑ4 (ζ ) − 1] ⎭⎪
Due to boundary conditions (7.1.6) and (7.1.7), the partial derivative of time is
presented as a sum wt = w1t + w2t . (7.1.13)
The first term satisfies condition ψ 1t = −Vx at the plate and ψ 1t = 0 at the wall, i.e., the
boundary conditions differ from ψ by factor V / V ; therefore,
w1t = (V / V ) w . (7.1.14)
Function w2t satisfies condition ψ 2t = −V ϕ x at the plate, ψ 2t = 0 at the bottom, and it has
simple poles at points a and π − a . One may see that at the plate and bottom
ψ 2t = Im( wz 2 / 2) . (7.1.15)
Thus, auxiliary function G = w2t − wz 2 / 2 is real only at both horizontal sides and it can be
continued analytically over the ζ -plane. Again we have an elliptical function with periods
π and πτ i and with singularities as doublets at points A( a) and B (π − a ) , so that it can be
expressed in terms of theta function as
G = A + B1 ln ′ϑ1 (ζ − a) + B2 ln ′ϑ1 (ζ + a ) + C1 ln ′′ϑ1 (ζ − a ) + C2 ln ′′ϑ1 (ζ + a ) .
All unknown coefficients could be determined satisfying the equality of leading parts of both
sides. Hence,
w 2 V2
w2t = z + [ln′′ϑ1 (ζ − a) + ln′′ϑ1 (ζ + a) − 2ln′′ϑ3 (a)] −
2 2 (7.1.16)
V2
− ln ′ϑ1 (2a ) [ ln ′ϑ1 (ζ − a ) − ln ′ϑ1 (ζ + a ) + 2ln ′ϑ3 (a ) ].
4
Functions (7.1.10) - (7.1.16) together with (7.1.8) constitute an exact analytical solution of
the boundary value problems (7.1.6), (7.1.7). Using these functions it is possible to
determine pressure distribution on the plate and bottom, and the resultant force on the plate.
Pressure is calculated by Cauchi-Lagrange integral (7.1.1) and can be presented in the form
261
p = ρ (Vlp1 + V 2 p2 ) , (7.1.17)
where p1 and p2 are dimensionless functions
p1 = − h Re w / lV , p2 = − Re w2t / V 2 − wz wz / 2V 2 . (7.1.18)

Distribution of p1 and p2 over the bottom for h / l = 0.5 is shown in Fig.7.1.2a.
Fig.7.1.2.
(a) Pressure distribution
on the bottom;
(b) vertical force as
function of distance;
dotted lines show
dependence on l/h.
The vertical force can be calculated by integrating pressure along both sides of the
plate. Since the complex velocity has a constant imaginary part, the real part of expression
wz 2 can be integrated instead of wz wz , then the pressure is determined by an analytic
function that can be integrated over any line enveloping the plate. It is important due to the
singularities at points A, B and C. According to equation (7.1.15) the resulting force is
presented in the form X = ρ l ( −VlC1 + V 2 C2 ) . (7.1.19)
Coefficients C1 and C2 are plotted in Fig.7.1.2b as functions of ratios h / l and l / h .
Coefficient C1 approaches the limit π / 4 as distance h approaches infinity (doted line). The
second term of equation (7.1.17) shows that coefficient C2 keeps the same sign regardless of
the motion direction. On the same token, calculations show very interesting but unexpected
results: dependence of coefficient C2 is non-monotonic; with distance h decreasing the
coefficient increases and then decreases so that the resulted force can be directed to the
bottom, i.e. the flat bottom or the flat wall attract the flat plate if the distance is sufficiently
small. This is a theoretical conclusion only and requires experimental verification.
Sec. 7.2. Gravity and Air Effects during Water Entry

7.2.1. Notes on Approximation of Problem Statement
The problem of a body moving in air or water and its interaction with water and air is quite
difficult to study both in non-linear theory and asymptotically. However, some results can be
obtained approximately under simplified assumptions. The air density is less than hat of
water by a factor of approximately 1/800. Therefore, air does not have much effect on water
movement in general but can affect the free surface near the moving body. On the other
hand, deformation of the free water surface is small and has little effect on velocity field of
air due to a body moving in the air. Thus the free water surface can be assumed as a steady
boundary, usually flat. Then the general problem can be broken into separate problems:
determination of air velocity field due to body movement; and deformation of the free water
262
surface under air pressure. The former problem is unsteady and nonlinear (see Sec. 7.1.),
while the latter can be solved by a linear theory.
In this section the simplified approximation is applied to obtain deformations of free
water surface due to a plate movement and also due to a thin body entrance in water.
7.2.2. Flat Plate Dropping on a Free Surface
Schematics of the flat plate dropping on a free water boundary is shown in Fig.7.2.1.
Consider first a drop on a solid surface. Pressure distribution on the horizontal line (free
water boundary) can be calculated by equation (7.1.17) in the parametric rectangle:
p = ρ a (Vlp1 + V 2 p2 ) , (7.2.1)
where ρ a is the air density, functions p1 and p2 are
calculated by (7.1.18) for ζ = ξ + πτ i / 2 . Variables ξ and x
on the bottom are interconnected by the equation obtained
from (7.1.8): x = h ln ′ϑ2 (ξ ) . (7.2.2)
Fig.7.2.1. Schematics of Since pressure on the free water boundary due to
flat plate dropping. the moving plate, is proportional to air density pa, the air
makes a small perturbation on the water with density
pw>>pa. Thus, the deformation of the free boundary can be calculated by the linear theory. If
all partial derivatives of potential for water are negligible relative to pressure and gravity, the
free boundary can be determined by y = − p / ρw g . (7.2.3)
The free surfaces calculated by equation (7.2.3) for dropping the flat plate
[Terentiev, 2002] are shown in
Fig.7.2.2. Ordinate (7.2.3) corresponds
to a hydrostatic case when inertial forces
are not ignored, thus the ordinates could
be overestimated. As pressure (7.2.1) is
proportional to speed squared, it varies
quickly with a change of drop height.
The partial time-derivative of speed
should be of the same order of
magnitude as pressure while term g y
cannot be of a noticeably higher order, Fig.7.2.2. Deformation of the free water
and thus it can also be ignored. With this boundary due to a plate (its length 2l=2m)
assumption the problem of movement of dropping from height H=2m.
a semi-infinite rectangle at a constant
speed was considered in [Terentiev & Nikitina, 1983]. It is necessary to remind, that the
movement of a semi-infinite body is possible only at a constant speed. Below, using the
drop-plate case as an example, movement of a finite-size body at variable speed is discussed.
If the pressure on the real axis is given as a function of time and coordinates, the
time derivative of potential is equal to ϕ = −p(x,t)/ρw. The function of partial derivative on
1 ϕ t ( x′, t )
∞
time is determined by Scharz’s integral as wt ( z , t ) = ∫ x′ − z dx′ .
π i −∞
(7.2.4)
Thus, partial derivative of stream function at the free boundary is expressed as Cauchy’s
263
1
∞
ϕ t ( x′, t )
integral ψ t ( x, t ) = −
π ∫
x′ − x
−∞
dx′ . Since the kinematical condition at the free boundary
is presented as equality ∂y / ∂t = −∂ψ / ∂x , it is expedient to find the derivative with respect

1 ϕ t ( x′, t )
∞
to coordinate x: ψ tx ( x, t ) = − ∫ dx′ . As function ϕ t ( x, t ) has regular continuous
π −∞ ( x′ − x) 2
second derivative over the entire x-axis and vanishes in infinity, the singular integral can be
integrated by parts [Krikunov, 1966]. Thus the free boundary ordinates could be found from
∞
∂ 2 y 1 ϕtx ( x ', t )
a differential equation of second order = ∫ x′ − x dx′ .
∂t 2 π −∞
(7.2.5)
The problem becomes complicated since the integral is singular and should be
calculated as Cauchy’s principal value. Similar integral was calculated in [Terentiev &
Nikitina, 1983] presenting it, as usually, as a sum of two integrals: one with a continuous
integrand calculated numerically, and the other with a Cauchy’s integrand calculated
analytically. As shown before, integral (7.2.4) can be calculated easily in complex form at
point ( x + δ i ) instead of point (x), and the interval of integrating can also be finite ( −Δ, Δ )
instead of infinite. Besides, function ϕtx ( x, t ) in integral (7.2.5) is equal to ϕtξ / xξ so that
integral (7.2.5) can be transformed to one with respect to variable ξ on interval (0, π). Partial
differential equation (7.2.5) can be considered as ordinary one for fixed coordinate x. Instead
of time t, it is expedient to integrate with respect to the distance of plate from the beginning
of falling, s=s(t). Speed of falling plate as a function of distance is equal to
V = s (t ) = 2 gs . Deformation of the free boundary for different distances from the
unperturbed water level plotted in Fig.7.2.3. The plate is falling from the height of one
meter; the half length of the plate is also one meter.
Fig.7.2.3.
Deformation of free
boundary due to plate of
length l falling from
initial height H = l.
The calculations
show that the plate in this
case touches the wave
crest surface at a distance
about h=0.015. Air
cushion formed under the plate reduce the hydrodynamic force of the plate impact on water.
It should be noted that the numerical data received reflect the real process only qualitatively
as at a close distance the air pressure can grow sharply and the air discharge rate increases
sharply as well. Therefore compressibility of air and failure of stream from sharp edges of
plate can significantly affect the geometrical and hydrodynamic characteristics of stream.
Nevertheless, our results agree well with numerical calculations obtained within the
framework of one-dimensional model of compressible media flow from the air cushion with
264
viscosity and friction against water taken into account [Verhagen, 1967; Buford &Koehler,
1977].
7.2.3. Entry of a Thin Wedge into Water
The cavity following a body upon entry into water can collapse at a depth or at the free
surface. The cavity closure at a depth is caused by the gravity, but its collapse at the free
surface closure is due to air movement behind the body. This phenomenon for axisymmetric
flow was studied in [Logvinovich & Yakimov, 1973; Yakimov, 1973], for a slender body in
[Buivol et al, 1973], and for a 2-D flow in [Gusev & Terentiev, 1981. Cavity closure at entry
into the water has been discussed by Tulin [2001].
The solution by Dobrovolskaya [1969] for the non-linear value problem of entry of a
wedge shows that the splashing of the free surface is negligible for a thin wedge, and the free
surface can be assumed to be flat. The problem of slender body entry into gravity-affected
water can be divided into three separate linear problems: a) a slender body entry into water
in non-gravity field, b) moving of the vertical free boundary due to gravity, and c)
deformation of the vertical free boundary by the pressure of moving air. The time-dependent
pressure of moving air must be calculated using a non-linear problem only. Schematics of
these flows are shown in Fig.7.2.4a-c.
Fig.7.2.4.
Components of a slender body
entry problem: (a) - entry of a
wedge in weightless fluid;
(b) - vertical plate moving in a
gravity liquid,
(c) - a slot of suction at a wall.
The value problem of the

a-case is a mixed one for a partial
time-derivative of complex potential w1t ( z, t ) = ϕ1t ( x, y ) + iψ 1t ( x, y ) . (7.2.6)
The value problem for the function in the first quadrant has to satisfy the following boundary
conditions: ϕ1t = 0, x = 0, y ≥ 0, x ∈ (0, s ), y = 0;
ψ 1t = β ( x, t ) = Vf ( x − s − a ) − V 2 f ′( x − s − a ), x ∈ ( s, s + a ), y = 0; (7.2.7)
ψ 1t = 0, x > s + a, y = 0,
where V is the body velocity; s is the length of cavity; a is the body length; and y = f(x-s-a)
is the equation of the moving body.
The value problem can be solved by analytic continuation through the y-axis and
s+a
2 2 2 x′β ( x′, t )
using the integral operator w1t = z −s ∫ dx′ . (7.2.8)
π s x ′ 2 − s 2 ( x ′2 − z 2 )
For a flat-sides wedge with top angle 2α , integral (7.2.8) can be presented analytically as
w1t ( z , t ) = α [V 2W1 ( z , t ) + VW2 ( z , t )] , (7.2.9)
1 z 2 − s 2 − ( s + a)2 − s 2
where w1t ( z , t ) = ln , (7.2.10)
π z 2 − s 2 + (s + a)2 − s 2
265
w2t ( z , t ) = ( s + a ) w1t ( z , t ) +
1⎛ 2 s + a − ( s + a)2 − s 2 ( s + a ) z 2 − s 2 + z ( s + a ) 2 − s 2 ⎞ (7.2.10)
+ ⎜ z − s 2 ln + z ln ⎟.
π ⎜⎝ s + a + ( s + a)2 − s 2 ( s + a ) z 2 − s 2 − z ( s + a ) 2 − s 2 ⎟⎠
Now, the partial derivative with respect to coordinate x, and then the double integral of its
imaginary part with respect to time or to distance s on interval (0, s ) should be calculated.
Namely, the cavity boundary for a wedge with unit length a = 1 and top angle 2α, moving at
constant velocity V = 1, is obtained by:
s s′
2α x ( s′′ + 1) 2 − s′′2 ds′′ds′
π ∫x ∫x ⎡ x 2 − ( s′′ + 1) 2 ⎤ s′′2 − x 2
y1 ( x, s ) = . (7.2.11)
⎣ ⎦
Partial time derivative w2t for the (b) problem satisfies the boundary conditions as
ϕ2t = 0 with x = 0, y > 0,
ψ 2t = 0 with y = 0, x > s, (7.2.12)
ϕ2t = gx with y = 0, 0 < x < s .
The value problem solution is w2t = g ( z − z 2 − s 2 ) . (7.2.13)
Thus, the vertical free boundary moving at the constant unit speed can be calculated as
t ν
2α xd μ dν
Fr ∫x ∫x μ 2 − x 2
y2 = , (7.2.14)
where Fr = V 2 / gb is Froude number, and b is the length of the wedge base.

The (c) problem of the air effect can be solved after obtaining the pressure due to air
movement in the cavity. For simplicity, this problem will be considered as a flow due to slot
suction in a wall. The slot is assumed to be infinitely long of width b. The flow is two-
dimensional, steady, ideal and non-rotational. The flow speed in the infinite slot is assumed
to be equal to velocity V of the moving body. This problem can be solved easily by mapping
onto the upper half-plane. Transformation of an air domain onto the upper half plane can be
b ⎛ i ζ −1 +1 ⎞
written in analytical form: z = ⎜⎜ 2i ζ − 1 − ln −πi⎟ .
⎟
(7.2.15)
2π ⎝ i ζ −1 −1 ⎠
The function of complex velocity of airflow is expressed as
dW iV
= . (7.2.16)
dz ζ −1
Pressure on the cavity boundary can be found in parametric form by two equations:
p = − ρ aV 2 /[2(1 − ξ )], ξ ∈ (0, 1) , (7.2.17)
⎛ b 1− 1− ξ ⎞
⎜⎜ 2 1 − ξ + ln
x=− ⎟⎟ , ξ ∈ (0,1) . (7.2.18)
⎝ 2π 1 + 1 − ξ ⎠
After determining the pressure distribution at the cavity boundary, the value problem of w3t
will be the same as w2t except the last condition of (7.2.12), which is
ϕ3t = − p( x) / ρ w , x ∈ (0, s ), y = 0 . (7.2.19)
The solution of the last value problem is expressed as
266
s2 − z 2
s
ϕ3t ( x′) x′dx′
w3t ( z , t ) =
πi ∫ ( x′
0
2
− z 2 ) s 2 − x ′2
. (7.2.20)
The ordinate cavity boundary is calculated by

ρ 1 ⎡s 2 2 s + s2 − x2 ⎤
y3 ( x, s ) = a ⎢ s − x − x ln ⎥. (7.2.21)
ρ w 2π ⎢⎣ x x ⎥⎦
The ventilated cavity boundary is equal to a sum of ordinates (7.2.11), (7.2.14) and (7.2.21).
Fig.7.2.5 shows cavities past the penetrating wedge. Cavities at different distances
s = 1, 2, 3, 4 and Froude number Fr=2 are shown in Fig.7.2.5a (air is ignored). The ordinate
is divided by α. The cavity jointing point and the free boundary moves at a constant unit
velocity in the vertical direction. The depth of
collapse in this case is approximately equal to
11
sd = (2α 2 Fr ) 2 / 3 . (7.2.22)
4
Fig.7.2.5b shows a cavity in a gravity field, which
takes air into account. In this case two horizontal
jets at the cavity end are formed because of a
singularity angle at the domain boundary. These
jets are very thin at first and can be ignored. Then
they increase and we can imagine that the surface
closing the cavity is as thick as the jets, or a half
of the body thickness. Calculations show that for
the wedge moving at a constant speed the surface
closing can be approximated by:
Fig.7.2.5. Deformation of cavity past a 21 ρw
ss = α . (7.2.23)
wedge by entry into water: 8 ρa
(a) - α = 1 , Fr=2, s=1 to 4; Equations (7.2.11), (7.2.14) and (7.2.21)
(b) - α=10°, Fr=2, ρa/ρ=0.01, s=4. allow calculating the moments of depth or surface
collapse. For example, at α=10°, b=1 cm,
ρa/ρ=0.00125 (air-to-water) and V=100 cm/s the depth collapse is at sd = 123.5 cm and the
surface collapse is at ss = 37 cm, i.e. it occurs before the depth collapse. However, at V=25
cm/s the depth collapse at sd =19.4cm occurs before the surface collapse at ss =37cm. The
air after the depth collapse does not move, thus the cavity cannot collapse at the free surface.
However, ventilated cavity always collapses upon entry.
It should be noted that the described approach reflects the real process rather
approximately; and that the free water surface deformation, impact of both jets, and other
effects here are not taken into account. Nevertheless, this approach allows estimating
influence of such factors as weightiness of fluid or air above water surface. Hydrodynamic
forces can be calculated by integrating the pressure distribution along the body surface.
7.2.4. Non-Symmetric Entry into Water
Gravity effect on an inclined entry into water can br assessed similarly. All boundary
conditions are the same as in Sec. 6.3 except the dynamic condition at cavity boundaries,
267
which looks like ϕt = gx sin πγ . The gravity effect on free water surface is, as before,
ignored. The problem is divided into two problems: one similar to the problems considered
in Sec. 6.3, and the other is a problem of a plate moving with zero angle of attack in a
gravity field. Both problems require solving additionally integral equations of Volterra’s first
type. Inclined entry of a plate with ventilated cavity and its collapse time were calculated by
Mikhailov & Nikitina [1980]. The case of a partial entrance of a plate with a ventilated
cavity moving at a constant speed was solved analytically. For other cases, however, only
numerical solutions could be found. Other methods have been applied for studying an
entrance of semi-infinite plate with a ventilated cavity [Kuznetsov, 1977], and a ventilated
foil [Yim, 1973; Wang, 1977, 1979].
Let's present some numerical results for a plate. According to the problem division,
hydrodynamic force and moment can be separately analyzed with or without taking into
account weightiness, i.e. lift coefficient CL = 2 L / αρV 2 s and moment relative to the point of
plate/free surface contact CM = 2 M / αρV 2 s 2 can be presented as a sum of two terms:
CL = C1L + Fr −1 C2 L , CM = C1M + Fr −1 C2 M , (7.2.24)
where Fr = V / gs is the Froude number; s is the distance between the leading edge and the
2
point of intersection with the free surface. Coefficients C1L and C1M are expressed by
equations (6.3.40) and (6.3.41), see Fig.6.3.5. Coefficients C2 L and C2M correspond to
Archimedes force and for inclination angle πγ are approximated with a high accuracy (less
2
than 10−2 ) as C2 L = sin πγ , C2 M = sin πγ . (7.2.25)
3
In the specific case of perpendicular entry ( γ = 1/ 2 ), exact values of coefficients are:
C1L =
4 2
π
( )
2 − ln(1 + 2) ≈ 0.959,
⎛1 2 − ln(1 + 2) ⎞
(
C2 L = 3ln(1 + 2) − 2 ⎜⎜ + 2 ) ⎟ ≈ 1.005,
3 2 − ln(1 + 2) ⎟⎠
(7.2.26)
⎝2
1 π ⎛ 3ln(1 + 2) − 2
1⎞
C1M = ≈ 0.707, C2 M = ⎜⎜ + ⎟⎟ ≈ 0.680 .
2 4 ⎝ 3 2 − ln(1 + 2) 2 ⎠
Equation (7.2.25) gives the exact hydrostatic force and moment of an inclined plate, thus
weightiness can be ignored for slender body entry and then hydrostatic pressure
phs = ρ g x sin πγ can be added in the final expression.
The second case of entry of a finite-chord foil with fully ventilated cavity will solve
the flow value problems with and without gravity taken into account. The integral equations
for that case could not be solved analytically, but numerically only. Denote the
dimensionless distance as s/a=1+τ (a is the plate length), lift coefficients as
CL = 2 L / αρV 2 a = C1L + ν 0C2 L , and coefficient of torque relative to the plate center as
CM = 2M 0 / αρV 2 a 2 = C1M + ν 0C2 M ; (ν 0 = ga / V 2 is the Froude number inversed).
Dependence of the coefficients for perpendicular entry into liquid is shown in
Fig.7.2.6. The dash lines are asymptotes of curves for lift coefficient C1L=π and torque
coefficient C1M=3π/32π corresponding to Kirhoff’s model in unbounded domain. The
268
coefficients correspond to gravity increase as C2 L ≈ π τ / 2, C2 M ≈ (π / 8) τ / 2, τ >> 1 .
Fig.7.2.6.
Vertical entry of
fully ventilated
plate:
(a) – lift
coefficients;
(b) - torque
coefficients.
Sec. 7.3. Evaporation from a Cavity Boundary

In a real cavitating flow a part of fluid evaporates at the free boundary, and is then evacuated
as gas-liquid mixture into the main stream generating cavity cloud or hole vortexes behind
the cavity. Diffusion through the cavity boundary for supercavitating flow is insignificant
and can be ignored. If however, the fluid is gazing or overheated, diffusion becomes
intensive and can affect hydrodynamic and geometric characteristics of the flow. Various
applications of evaporation using cavitating processes [Fedotkin & Machinsky, 1983], such
as capturing blebs by cavity [Savchenko et al., 1985], confirms the interest to cavity flow
with evaporation. Below a simple model of flow with evaporation is proposed.
7.3.1. Physical and Mathematical Problem Statement
As known from experiments, evaporation rate depends linearly on surface and difference
between sated steam pressure and pressure above fluid surface (in the case of cavitation –
pressure in cavity) [Fedotkin & Machinsky, 1983]. This pressure difference can be created
by removal of the steam-water mix from cavity. Assuming that the difference remains
constant, evaporation rate depends only on cavity surface, i.e. it is possible to assume that
the normal fluid speed at the cavity surface is constant. This simplified physical model is
used in this section. A planar flow of a flat-sided wedge is considered, Fig.7.3.1a.
Fig.7.3.1. (a) - Flow on the z-plane (b) - potential domain on the w-plane; (c); the ζ -plane.
Let the complex potential be zero at detachment points A and B of the wedge/cavity
boundary. Since normal velocity ∂ϕ /∂n=vn is assumed to be constant as well as that at the
cavity boundary, then tangent velocity ∂ϕ /∂s=vs should be constant as well. Also assumed
269
that ratio vn / vs = tan πκ is given, then ratio ϕ /ψ = tan πκ at the cavity boundary. The
potential domain is shown in Fig.7.3.1b while the hodograph of the velocity, Fig.7.3.1c, is
the same as without evaporation.
7.3.2. Solution for a Wedge with Flat Sides
The problem can be solved by conformal mapping. If the parametric domain is in the first
quadrant of the ζ -plane (Fig.7.3.1c), then complex velocity wz (ζ ) and derivative of
potential wζ (ζ ) can be obtained, as above, by its singularities and zeros:
2μ
⎛ ζ − 1 ⎞ − bζ /(ζ 2 +1)
wz = e − iπμ ⎜ ⎟ e , (7.3.1)
⎝ ζ +1⎠
(1 − ζ 2 )ζ 1− 2κ (ζ 2 + 1)1+ 2κ
wζ = . (7.3.2)
(ζ 2 − e2α i ) (ζ 2 − e−2α i )
2 2
Transformation from the ζ -plane into the z-plane is calculated, as usually, by integrating
function zζ (ζ ) = wζ (ζ ) / wz (ζ ) . Namely, the length of wedge side is equal to
(1 − ξ )1− 2 μ (1 + ξ )1+ 2 μ (1 + ξ 2 )1+ 2κ ξ 1− 2κ bξ /(ξ 2 +1)
1
l=∫ e dξ . (7.3.3)
0
(1 − 2ξ 2 cos 2α + ξ 4 ) 2
Two unknown parameters α and b can be obtained from conditions for velocity in infinity
wz (eiα ) = v∞ = 1/ 1 + σ and for “closing” of cavity limiα d ln[(ζ − eiα ) 2 zζ (ζ )]/ d ζ = 0 .
ζ →e
Taking into account functions (7.3.1) and (7.3.2), one obtains two equations as
μ ln tan(α / 2) − (b / 4)cos α + с = 0 , (7.3.4)
b = 4 μ cot 2 α + 4κ cos α , (7.3.5)
where с = 0.25ln(1 + σ ) . Substituting the second equation into the first one we get the only
equation for unknown angle α : μ cos α + [κ − c − μ ln tan(/ 2) ] sin 2 α = 0 . (7.3.6)
The left part of equation (7.3.6) for two limited cases, α → 0 and α → π / 2 , is equal to μ
and κ − c , respectively. Thus, equation (7.3.6) can be solved only if κ ≤ c . A solution of
equation for κ = 0 is α = π / 2 , i.e. point D coincides with point C and the cavity has an
infinite length as long as the cavitation number differs from zero, σ≠0.
Integrating the pressure along the wedge and dividing by d = 2l sin πμ , one obtains
the drag coefficient as CD = (1 − I / l )(1 + σ ) . (7.3.7)
where expression I is obtained from length l, or equation (7.3.3) by changing the sign of
parameters μ and b. Numerical calculations for cavitation numbers σ=0.5 and σ=1 and
wedge angles 2πμ = 30 , 90 ,180 show [Terentiev, 1985] that drag coefficient mildly
depends on diffusion parameter κ . For instance, the drag coefficient of a perpendicular plate
at σ=0.5 and parameter κ varying from 0 up to a decreases from 1.323 to 1.301 only, and
for σ=1 decreases from 1.774 to 1.714. Accordingly, for a wedge with angle 90 the drag
coefficient decreases from 0.977 to 0.928 and from 1.354 to 1.207. With the wedge apex
angle decreasing, the effect of parameter κ becomes more noticeable. For a wedge angle of
30° it is possible to specify the following approximated formulas:
270
CD = 0.539 − 0.685κ − 5.972κ 2 for σ = 0.5,

(7.3.8)
CD = 0.921 − 0.624κ − 9.598k 2 for σ = 1.
Approximated functions for cavity length L(κ ) and width H (κ ) can also be given:
3/ 2
L(κ ) H (κ ) ⎛ ln(1 + σ ) ⎞
≈ ≈⎜ ⎟ , (7.3.9)
L(0) H (0) ⎝ ln(1 + σ ) − κ ⎠
they are in a good agreement with theoretical results for the considered range of σ and μ.
Cavity shapes for various ratios κ0
=κ/c are plotted in Fig.7.3.2.
Fig.7.3.2.
Cavity shapes for various diffusion
numbers: (a) – plate; (b)- wedge.
It is seen that the diffusion parameter

depends significantly on the cavity size.
An increase of cavity by evaporation has
been shown experimentally in [Fedotkin
& Machinsky, 1983].
7.3.3. Kirchhoff Model

For detail analysis, consider a flow of a perpendicular plate with zero cavitation number,
σ=0, or b=0. Solution of the first equation (7.3.4) for b=0 is α=π/2, and from equation
(7.3.5) it follows that c=0. In this special case for the plate at μ=0 expressions (7.3.1) and
ζ −1 (ζ 2 − 1)ζ 1− 2κ
(7.3.2) are reduced to wz = e − iπμ , wζ = . (7.3.10)
ζ +1 (ζ 2 + 1)3− 2κ
Cavity boundary AC is determined by integrals
η η
x = ∫ r cos θ dη , y = ∫ r sin θ dη , (7.3.11)
0 0
(1 + η 2 )η 1− 2κ π
where r= 2 3− 2κ
, θ = − − πκ + 2arctanη . From the second equation (7.3.11) it
(1 − η ) 2
follows that the angle between the cavity tangent line and the plate is πκ at cavity boundary
detachment point A ( η = 0 ), and asymptotic angular coefficient ± tan πκ in infinity at
η → ∞ . Integrals (7.3.3) and (7.3.11) can be calculated analytically in special cases when:
π η (1 + η 2 ) 1 1 − η η2
• κ =0 l =1+ , x= + ln , y = − ; (7.3.12)
4 4(1 − η 2 ) 8 1 + η 2(1 − η 2 )
π 1 1 −η η2
• κ = 1/ 2 l = 1 + , x = ln , y=− ; (7.3.13)
2 2 1+η 1 −η 2
π 1 η (1 + η 2 ) 1 1 − η 1⎛ 1 − 3η 2 ⎞
• κ = −1/ 2 l= + , x= + ln , y = ⎜1 − ⎟. (7.3.14)
16 6 8(1 − η ) 16 1 + η
2 2
6 ⎝ (1 − η 2 )3 ⎠
The first case (7.3.12) corresponds to classical Kirchhoff flow; the cavity boundary is a
streamline. In the second case all liquid evaporates in the cavity, and in the third case the
271
liquid flows from cavity to main stream, so that the potential in cavity is constant.
The drag coefficient takes the form as
⎧2π /(π + 4), κ = 0;
⎪
CD = ⎨4 /(π + 2), κ = 1/ 2; (7.3.15)
⎪16 /(3π + 8), κ = −1/ 2 .
⎩
Fig.7.3.3.
Shapes of Kirchhoff cavity near the upper edge of a
plate.
From here it is seen that despite of significant difference

in flow modes, the drag coefficient changes only within
the range from 0.7780 to 0.9182. Fig.7.3.3 shows cavity
shapes near the attachment point for some values of evaporation parameter κ .
7.3.4. Linearized Approach for Cavitating Flow with Evaporation
Let’s assume the derivative of foil ordinate y ′( x) , as well as cavitation number σ and
evaporation parameter κ , are infinitesimals of the same order. Then function
ω = u − iv ≈ wz − 1 is an infinitesimal of the same order and satisfies the following boundary
conditions: v = y ′( x) on a thin foil and u = σ /2 on the cavity boundary (7.3.16)
These conditions coincide with conditions of the linearized problem of cavitating flow
without evaporation and its solution can be found using the integral operators. But the
kinematical condition at the cavity boundary differs from the condition at a non-penetrated
boundary. Since speed at the cavity boundary is directed at angle πκ to the boundary, the
derivative of boundary ordinate is yc′ = v ± πκ , (7.3.17)
where singes “+” and “–” correspond to the upper and lower cavity boundaries, respectively.
Integrating the last equation, one can obtain the cavity shape. Integrating
additionally the first equation of (7.3.16), one obtains the “closing” condition as
L
Im ∫ ω dz = −πδ , δ = 2∫ κ dx . (7.3.18)
C 0
Conditions (7.3.16) – (7.3.18) allow to solve a more general problem when
cavitation number and evaporation parameter remain small as functions of variable x. E.g. if
the second integral in (7.3.18) vanishes, δ = 0 , but κ ( x) is a known function, then all
conditions and the problem solution coincide with those for the above-considered cavitating
flow without evaporation. Thus the cavity length and drag should be the same, as for flow
without evaporation but cavity shapes are different. For δ ≠ 0 the solution and all
characteristics depend on κ(x).
For example let’s consider the dependence of cavity on evaporation parameter κ(x)
at a constant cavitation number. A cavitator assumed to be infinitesimal so that it is modeled
as a singular point; the same singularity is at the cavity end. Consider the upper half-plane of
the flow domain. Then we have a mixed value problem on the half-plane with the following
u = σ / 2, x ∈ [0, 1], y = 0,
boundary conditions: (7.3.19)
v = 0, x ∈ (−∞, 0) ∪ ( L, ∞ ), y = 0 .
272
σ z ⎛A ⎞
The solution can be found as ω= + ⎜ + B⎟ . The function should
2 z−L⎝ z ⎠
vanish in infinity so B = −σ / 2 . Constant A is obtained using condition (7.3.18):
σ⎛ L − 2z ⎞ δ
A = −δ / 2 + σ L / 4 . Hence, ω = ⎜1 + ⎟ − . (7.3.20)
2 ⎜⎝ 2 z ( z − L) ⎟⎠ 2 z ( z − L)
dyc σ ( L − 2 x) δ
Equation (7.3.17) due to last function looks like = − + πν ( x) .
dx 4 x( L − x) 2 x( L − x)
If the initial value is yc (0) = 0 , the cavity boundary is determined as
σ δ ⎛ 2 x ⎞ πδ
x
yc = x( L − x) − arcsin ⎜ − 1⎟ − + π ∫ν ( x′)dx′ . (7.3.21)
2 2 ⎝ L ⎠ 4 0
The first term represents the equation of ellipse and corresponds to a cavity with diffusion
ignored. Other terms appear due to the diffusion of steam.
It should be noted that the cavity should be single valued. This imposes additional
conditions on function κ(x) and cavitation number σ . Let function κ(x) is everywhere
positive, then the magnitude of δ is positive as well. From asymptotic expansion of the right
part near point x = 0 it follows that under condition δ / L ≤ σ / 2 the cavity ordinate
derivative is positive, yc′ (0) > 0 , and the cavity is single valued locally.
If δ / L = σ / 2 then the slope is yc′ (0) = πκ (0) . In this case function (7.3.20) is
bounded at point z = 0 , and the tangent line to cavity at that point has a slope of πκ (0) .
Physically such a flow can be realized for horizontal flow of a highly heated plate.
The cavity boundary in a
neighborhood of the trailing point is always
single valued because the derivative for
positive δ approaches negative infinity
y ′( x) → −∞ , as x → L and yc ( L) = 0 . The
cavity shapes are shown in Fig.7.3.4. Fig.7.3.4. Shapes of cavity due to
longitudinal flow of a heated plate at
7.3.5. Conclusion various diffusion functions.
The cavity shapes in Figs.7.3.2 and 7.3.4 can
be considered as the simplest model of longitudinal flow of a heated plate of unit length.
Diffusion function κ(x) is not known exactly, but it should be less than the values shown on
these figures. Function κ(x) could probably be obtained experimentally or from other
theoretical models using also thermo-dynamic relationships.
Sec. 7.4. Perturbed Flow around a Foil

7.4.1. Perturbed Flow
When two bodies of essentially different sizes are in a flow, the larger one disturbs the fluid
much more than the other, and so the small body can be considered of being in a perturbed
flow. Thus the total flow problem is divided into two separate problems: an undisturbed flow
around a large body only, and a perturbed flow around a small body only. In this statement,
three-dimensional problems have been considered in [Voinov et al., 1973] and two
273
dimensional problems in [Petrov, 2007; Terentiev, 2000]. Choosing some large bodies one
can simulate different situations of flow around a small foil, e.g. an intensification or
diminution of the gravity [Bolotin et al., 1981; Dimitrieva & Terentiev, 1988].
Let’s denote characteristic dimensions of the large and small body as L and l.
Suppose that ratio l / L = ε is infinitesimal ε << 1 , so that the effect of small body located at
point Z 0 is negligible. If W ( Z ) is the complex potential of flow in the Z-plane perturbed by
large body, then velocity dW / dZ can be expanded into Taylor series about point Z 0 :
WZ ( Z ) = WZ ( Z 0 ) + WZZ ( Z 0 )( Z − Z 0 ) + … . (7.4.1)
Keeping the needed number of terms in (7.4.1) and using new variable z = Z − Z 0 , one can
consider a perturbed flow with velocity in infinity z → ∞ given as
( wz )∞ = a0 + a1 z + a2 z 2 + … + an z n , z → ∞ (7.4.2)
1 ⎛ d nWZ ( Z ) ⎞
with coefficients a0 = WZ ( Z 0 ), a1 = WZZ ( Z 0 ), … , an = ⎜ ⎟ .
n! ⎝ dZ n ⎠ Z = Z
0
1
The pressure field ph = P∞ + ⎡⎣V∞2 − ( wz )∞ ( wz )∞ ⎤⎦ (7.4.3)
2
can be considered as hydrostatic pressure of any perturbed fluid in the z-plane. In particular,
first two terms in (7.4.2) neglecting term | a1 z |2 as an infinitesimal of second order produces
the hydrostatic pressure of a gravity field
1
ph = P∞ + [V∞2 − (a0 a0 + a0 a1 z + a0 a1 z )] . (7.4.4)
2
Constants P∞ and V∞ are the pressure and velocity at infinity of mainstream with the small
body ignored; pressure ph determines the pressure at infinity of flow domain with only the
small body taken into account.
7.4.2. Analytical Solution for Perturbed Flow
Consider a foil in the z-plane in a perturbed flow. Let functions ζ=f(z) or z=g(ζ) be
transformations of foil exterior in the z-plane onto exterior of unit circle on the ζ-plane.
Consider first the perturbed flow around the unit circle on the parametric ζ-plane.
Derivative wζ (ζ ) at infinity should have the same singularities as complex velocity
n
(7.4.2), or the function Fζ (ζ ) = ∑ a k [ z (ζ )]k (7.4.5)
k =0
n
with leading part at infinity, ∑b ζ
k =0
k
k
. Factor bk can be found analytically by coefficients ak
and coefficients of power series expansions of [ z (ζ )]k . But it is much easier to calculated by
2π
1
2π r k ∫0
integral bk = Fζ (reit )e − ikt dt , (7.4.6)
where r is any radius grater than 1. Potential w = ϕ + iψ should satisfy the ordinary
kinematical condition ψ = 0 on the foil as well as on the circle so that it can be extended
analytically through the circle. Then complex potential w(ζ) have the leading part at infinity
274
n
1
and at the origin due to equation (7.4.2) as G (ζ ) = ∑ ( )
bk ζ k +1 + bk ζ − ( k +1) . (7.4.7)
k =0 k + 1
Adding logarithmic singularity, one can write the complex potential as
Γ
w(ζ ) = G (ζ ) + ln ζ + const . (7.4.8)
2π i
If trailing point is given ζ = 1 , then velocity wζ (1) = 0 . Whence the unknown circulation is
n
Γ = −2π i ∑ (bk − bk ) . (7.4.9)
k =0
The complex potential for an arbitrary foil in the z-plane is determined from (7.4.8) by
substitution ζ=f(z).
7.4.3. The Joukowski Foil near a Circular Cylinder
Consider a non-circulating flow around a circular cylinder of radius L. Let the trailing edge
of the hydrofoil is located at point Z 0 = L0 eiβ ,
Fig.7.4.1. Parameter ε = 1/ L , as well as ratio
δ = 1/( L0 − L) , are supposed to be small.
Fig.7.4.1.
Flow around a cylinder and the Joukowski foil.
Complex velocity for non-circulation flow

around the cylinder on complex Z-plane is prescribed
by WZ ( Z ) = (1 − L2 / Z 2 ) . Its power series expansion at point Z 0 has the following
L2 k +1 L2
coefficients: a0 = 1 − , ak = ( − 1) ( k + 1) , k = 1, 2, 3, … (7.4.10)
Z 02 Z 02 + k
The exterior of Joukowski hydrofoil is mapped onto the exterior of unit circle on the ζ-plane
by z (ζ ) = ( u (ζ ) + 1/ u (ζ ) − 2 ) / 2 , (7.4.11)
where u (ζ ) = ( Rζ − d )eib + ih, R = d + 1 + h 2 + c, b = − arctan h . (7.4.12)
Here the trailing edge z=0 corresponds to point ζ =1; parameters d, h and c stand, as above,
for the thickness, flexure and curvature at the trailing edge, respectively. The foil’s
t
curvilinear abscissa is calculated by s (t ) = ∫ | dz (eiτ ) / dτ | dτ , τ ∈ [0, 2π ] . (7.4.13)
0
The foil velocity is calculated by Vex = iwζ (eit )eit / s′(t ), s′(t ) = ds / dt . (7.4.14)
Some results of calculations will be shown compared with numerical computing
below. It should be noted here that satisfying any conditions at a free boundary in perturbed
flow is very difficult so the proposed method should be useful for simulating any flow by
special perturbation of fluid. One of such simulations is considered below.
7.4.4. Cavitating Flow in a Tunnel with Variable Cross Section
A simulation of the gravity field by shaping a working section of a tunnel in a special way
has been discussed in [Acosta, 1961]. A comparison of results calculated in [Dimitrieva &
Terentiev, 1988] shows a good agreement between flows in a tunnel and in a gravity field. It
275
should also be noted that it is possible to approximate the viscosity effect on a cavitation
flow by designing the tunnel wall as in [Karlikov & Sholomovich, 1966]. Cavitation
modeling with gravity effect compensation was
considered in [Bolotin et al, 1981].
As an example, cavitating flows around
a wedge inside another wedge generated by a
source or sink are considered below. The source
or sink is located at the apex of the wedge.
There is another wedge on the inside of the
wedge with distance s from the source or sink.
Fig.7.4.2. (a) Flow The flow in the z-plane and the auxiliary
inside a wedge: domain are shown in Fig.7.4.2.
(b) - auxiliary
The derivative of complex potential wζ
domain on the ζ -
and function ω (ζ ) can be found by
plane.
recognizing their singularities as
ζ
wζ = , (7.4.15)
2
(ζ − a )(ζ 2 − b 2 )
2
ζ −1 (ζ − a )(ζ + b) c
ω = μ ln + κ ln − − πμ i . (7.4.16)
ζ +1 (ζ + a )(ζ − b) ζ
Three essentially unknown parameters a, b, and c are determined from three
conditions for ratio R/l (R is the distance | BO | ) with cavitation number σ and assuming
non-disturbance of the infinite stream. The third condition can be written as
Im ∫ wζ (ζ )e −ω (ζ ) d ζ = 0 , (7.4.17)
L
where L is the curve connecting two arbitrary points on the x-axis at both sides of the cavity.
There is a problem of determining the cavitation number σ = ( v0 / vE ) − 1 . As usually, v0 is
2
the velocity at the cavity boundary, but velocity vE is undetermined by the flow inside of the
wedge while it is the inlet velocity at infinity for undisturbed flow. It is supposed that
velocity vE is calculated at point E on the wedge side BA, length | BE |=| BO |= R , Fig.7.4.2.
The drag of the wedge can be calculated by integrating the pressure along the wedge side.
Fig.7.4.3. Cavity shapes in the source flow. Fig.7.4.4. Cavity shapes in the sink flow.
Figs.7.4.3 and 7.4.4 show the cavity shapes for the source and the sink flow inside
the wedge with angle 2πκ = π / 6 which where obtained by Dimitrieva. The cavities were
276
calculated for the flat plate and the wedge of angle 2πμ = π / 6 situated away from the apex
of length R/l = 10. The cavity in Fig.7.4.3 transforms to Kirchhoff’s one with infinite length
for certain cavitation numbers σ 0 , e.g. σ 0 = 0.75 for the flat plate and σ 0 = –0.12 for the
wedge with angle π / 6 . The cavitation number cannot be less than σ0. The cavity in
Fig.7.4.4 transforms to Colsher’s model with a closed cavity. For the flat plate σ0 = 1.3 and
for the wedge σ0=1.07. In that case the cavitation number must also be grater than σ 0 . The
flow is similar to cavitating flow in the longitudinal gravity field, and for greater R/l ratios
the calculated data are compared with the results in Sec. 4.3.
Sec. 7.5. The Approximation Method for a Compressible Fluid Flow

The Prandtl law, Eqs. (5.1.12) - (5.1.14), which follows from an asymptotic solution, allows
to consider compressibility of fluid, but the theoretical results should be valid for small
Mach number and small perturbation of flow only. In that case the approximation extends
well the validity range of Mach numbers. Approximation for a compressible gas was first
suggested by Chaplygin [1902]. His studies are remarkable in a number of aspects. They
constitute the first rational and systematic studies of gas flow. He also inspired many other
works. His approximation is used here for studying cavitating flow of compressible fluid.
7.5.1. Chaplygin Approximation
Assuming an adiabatic process p = ρ n , the governing equations for potential and stream
∂ψ ∂ϕ ∂ψ ∂ϕ
functions look like = (1 − τ ) β , = −(1 − τ ) β , (7.5.1)
∂y ∂x ∂x ∂y
where β = 1/(n − 1), τ = v 2 / vmax
2
= (n − 1) M 02 ; M 0 is Mach number corresponding to the
sonic speed in fluid at v = 0 .
For the angle of velocity θ and using variable τ instead of x and y , equations
∂ϕ 2τ ∂ψ ∂ϕ 1 − (2 β + 1)τ ∂ψ
(7.5.1) are transformed to = β
, =− . (7.5.2)
∂θ (1 − τ ) ∂τ ∂τ 2τ (1 − τ ) β +1 ∂θ
τ0
(1 − τ ) β
Substituting function λ=∫ dτ (7.5.3)
τ 2τ
Chaplygin transformed equations (7.5.2) into
∂ϕ ∂ψ ∂ϕ ∂ψ
=− , =K , (7.5.4)
∂θ ∂λ ∂λ ∂θ
1 − (2 β + 1)τ
where K= .
(1 − τ ) 2 β +1
Variable τ is connected to local Mach number by τ = M 2 /(2β + M 2 ) so that coefficient K
is a function of Mach number M. Its power series looks like K ≈ 1 + aM 4 + ... , so that for
subsonic cavitating flows it can be assumed equal to unity (K≈1). Then differential equations
(7.5.4) look like Couchy conditions of analyticity for complex potential ϕ + iψ = w(ω ) ,
where ω = −λ − iθ . If parameter τ 0 is a constant magnitude of variable τ at the free
boundary, analytic function ω ( w) looks like a logarithmic function of complex velocity for
277
incompressible fluid. The above obtained formulas are valid for the water with isentropic
dependence [Cole, 1950] p = k ρ n − B, n = 7.1, B = 3000atm . (7.5.5)
dp p+B
The local sonic speed of water is a 2 = =n . If
dρ ρ
p = 1atm = 9.81 ⋅ 104 kg / m ⋅ sec 2 , ρ = 103 kg / m3 then the
local sonic speed is equal to a = 1445.8m / sec , that
corresponds to the sonic speed of water at temperature
T = 10 C [Batchelor, 1970]. The dependence of function
K on Much number is plotted in Fig.7.5.1. It shows that
function K depends slightly on n, and its magnitude is
grater than 0.95 if M < 0.4 for water and if M < 0.5 for
air. Differentiating function 1/ τ twice with respect to
variable λ yields a second-order differential equation Fig.7.5.1. Function K(M):
whose solution is λ −λ
1/ τ = (C1e + C2 e ) / 2 . The solid line - n = 1.4,
integration constants are obtained satisfying conditions for dotted line - n = 7.1.
that function and its derivative at the free boundary by σ = 0 :
(1 − τ 1 ) β + 1 (1 − τ 1 ) β − 1
C1 = , C 2 = . (7.5.6)
τ 1 (1 − τ 1 ) β τ 1 (1 − τ 1 ) β
Transformation from the w-plane to the z-plane can be written in differential form
[Chaplygin, 1902] as dz =
1
2vmax
(
C1e −ω ( w) dw + C2 eω ( w) dw . (7.5.7) )
Analitic function w(ω ) can be found in the same way as potential w(ω ) for
incompressible fluid. Chaplygin considered an inclined plate with Kirchhof’s cavity in a
compressible fluid.
7.5.2. Cavitating Flow of a Compressible Fluid around a Wedge
If cavitation number σ and Mach number M at infinity are given, then the value of
parameter τ both in infinity and at the cavity boundary can be calculated as
(n − 1) M 2
τ∞ = , (7.5.8)
2 + (n − 1) M 2
τ 0 = 1 − (1 − τ ∞ ) (1 − nσ M 2 / 2 )
( n −1) / n
. (7.5.9)
Thus, variable λ due to (7.5.3) vanishes at the cavity boundary; at infinity it is equal to
τ0
(1 − τ ) β
λ∞ = ∫ 2τ dτ .
τ∞
(7.5.10)
Now, we have to obtain analitic function ω ( w) satisfying the following conditions:

• at the part of stream function related to the wedge sides Im ω (ϕ ) = ∓πμ ;
• at the cavity boundary Re ω (ϕ ) = 0 ;
• in infinity ω (∞) = −λ∞ .
Besides, it should have an integrated singularity like ω ≈ ( w − wc )1/ 2 at the cavity end.
278
Hence, the value problem looks exactly like for incompressible fluid and could be
solved similarly. Only the “closed” condition differs, which could be obtained from (7.5.7)
as C1 ∫ e −ω ( w) dw + C2 ∫ eω ( w) dw = 0 . (7.5.11)
∞ ∞
Some calculation results from [Terentiev, 2002] are represented on Figs.7.5.2 - 7.5.3.
Fig. 7.5.2 shows the cavities past a flat plate and the wedge with apex angle 2πμ = 30o at
Mach numbers M = 0 and M = 0.5 and cavitation numbers σ = 0.5 and 0.75.
Fig.7.5.2. Cavity
shapes past:
(a) - the plate
(b) - the wedge with
angle π / 6 .
A comparison of cavity boundaries calculated for water and air by using Chaplygin’s
approximation, and based on linear theory is plotted in Fig.7.5.3. As was shown in Sec. 5.1,
linear problems of a compressible cavitating flow around a thin wedge can be converted
using the Prandtl approach for incompressible flow around a wedge of angle 2πμ / 1 − M 2 ;
it gives the same results both
for water and air.
Fig.7.5.3.
Comparison of approximation
and linear theory.
Calculations show that the drag coefficient depend not much on compressibility of liquid.
For instance, changing the Mach number from 0 to 0.5 results in changing the drag
coefficient of plate from 1.323 to 1.353 for σ = 0.5 and from 1.547 to 1.556 for σ =0.75.
The drag coefficient of a wedge of angle 2πμ = 30o changes from 0.538 to 0.571 and from
0.721 to 0.764 for the same σ and M. Also, compressibility slightly affects cavitating flow,
except for the cavity length which increases with an increase in Mach number.
At the present time, very high sub- and supersonic velocities can be reached
experimentally by moving a body in water [Kirschner, 2001; Savchenko, 2001]. Much
attention is paid to numerical approaches for axisymmetric supercavitation in compressible
fluids [Vasin, 1998; Serebryakov, 1998] in the theory of cavitating flow around
axisymmetric bodies. There are some asymptotic relations for axisymmetric flow obtained in
[Yakimov, 1983].
279
Ch. 8. Supercavity as a Dynamical System
Ch. 8. The Supercavity as a Dynamical System

It is well known that a ventilated cavity may be subject to a more complex dynamics than
steady flow past a vaporous cavity. Under many conditions, vaporous supercavitation is a
fairly simple process: low pressure at a point on the cavitator causes a phase change in the
ambient liquid, the resulting gas advects downstream toward a region of higher pressure,
where it returns to the liquid phase. As has been discussed elsewhere in this book, closure is
an inherently unsteady process (except under the most theoretical of conditions), and is
complicated on a small scale. However, in the time-averaged sense, the basic process –
vaporization, advection, and collapse – is dynamically straightforward, absent complicating
effects associated with body interaction (for example, the quarter-chord instability of
partially cavitating foils discussed in Ch.2).
Once gas is introduced into the cavity, however, the situation becomes more
complicated. Under the low-to-moderate pressure conditions of supercavitating flow, the
gas does not change to a liquid phase in the high-pressure region, and must escape from the
rear of the cavity via some sort of mixing process. Whereas the vapor in a vaporous cavity
acts as a discontinuous spring (one that collapses when compressed back to ambient
conditions), the gas in a ventilated cavity is continuously and significantly elastic until it
breaks up into the bubbly wake downstream of the cavity. (Even then, the multiphase flow
in the bubbly wake is far more elastic than the ambient liquid. High-speed motion through
such a medium is discussed in a later chapter.) This more continuous compliance allows for
a more persistent exchange of potential and kinetic energy, and the resulting system may
display some interesting oscillatory effects. Therefore, it is worth to simplify flow problems
by introducing additional assumptions about some physical parameters for solving
complicated equations.
This chapter addresses the transient problem of cavity initiation by a pressurized gas
canister. The approximate method is intended to assess the importance of added mass to the
dynamics of cavity expansion, and it is shown that, under most conditions of interest to the
start-up problem, inertial effects are of little importance. In this approximation, the cavity is
treated as a simple sphere, and the Rayleigh equation of bubble dynamics is solved
simultaneously with the so-called blow-down equations that govern the discharge of gas
from a pressurized canister through an orifice or nozzle. If the pressure difference across the
nozzle is sufficiently high, the nozzle flow is choked, and the discharge rate from the
canister is unaffected by the dynamics of the cavity. Once the canister pressure drops below
a critical value, the flow through the nozzle is no longer choked, and the pressure in the
canister (and therefore the discharge rate) is coupled to the cavity dynamics.
It also presents a higher fidelity – albeit still approximate – solution to the cavity
dynamics problem. This method highlights the importance of gas loss from the downstream
end of the cavity. In order to account for the changing conditions as a cavity section evolves
from its birth at the cavitator to collapse in the closure region, the governing equation is
transformed to a system of delay differential equations, which class of problems possesses
some very interesting mathematical properties. This approach is sufficient to capture the
effect known as cavity auto-oscillation, which can lead to violent dynamical behavior. Since
the methods discussed below are based mostly on the approximated principle of
independence of cavity cross section given by Logvinovich it makes sense first to present
that principle adding some new results and its formalization.
280
Sec. 8.1. Logvinovich’ Principle and its Formalization

8.1.1. Logvinovich’ Principle of Independence of Cavity Cross Section
Numerous experiments carried out in CAGI under guidance by Logvinovich show that the
cavity cross sections deform almost normally to the trajectory of the cavitator and nearly the
same as it does in a straight movement. Fig.8.1.1 from [Logvinovich, 1959] confirms this;
the corresponding sections of two different cavities merge. It shows a good coincidence of
sections of two cavities located at the same distance from the cavitator. Based on
experiments and the conservation theorems in hydrodynamics, Logvinovich called this
phenomenon as “principle of independence of cavity sections”: each section deforms
independently on posterior moving of a cavitator and other sections but is determined by the
pressures at infinity and in cavity, and on initial velocity of deformation. This principle can
∂S 2 (h, t ) k
be written mathematically as = − Δp (h, t ) , (8.1.1)
∂t 2
ρ
where S is the cross-sectional area, Δp = p∞ − pc , k = k (σ ) is a coefficient depending
slightly on cavitation number, but can be taken mostly as a constant; h is a coordinate of a
cross-section on the cavitator trajectory. The
principle, as written in [Logvinovich, 1969], is
approximate in essence and could not be proved
analytically.
Fig.8.1.1.
Experiments showing independence of deformation
of a cavity section by moving with the same speed
but different curved trajectory.
The value of factor is determined mostly

experimentally, but it can be approximately found
from speed distribution on ellipsoid (5.7.3) and
(5.7.4). Consider a moving ellipsoid of axes 1 and
ε . A time dependence of cross-sectional area of
ellipsoid in case of uniform motion can be written as S (t , x) = πε 2 (t − x)(2 − (t − x)) , and so
the first and second time-derivatives of the area are St = −2πε 2 (t − x − 1) and Stt = −2πε 2 ,
respectively. It is shown that the second derivative, Stt, depends on minor semi axis only
which may be calculated at given cavitation number from equation (5.7.4) as V02 (ε ) − 1 = σ .
The difference of pressure is p∞ − pc = σ / 2 , and so the factor can be presented as
k = 4π b(σ ) 2 / σ . Using Eq. (5.7.26) one can calculate a cavity past any body by double
kσ
integration, S (t , x) = − (t − τ ) 2 + π y ′(t − τ ) + π , here t − τ = x . If the body moves with
4
t
time-depended speed v(t ) , then x = ∫ v(t ) dt . The cavity past any body moving steadily with
τ
unit speed is described by equation y ( x) = S (t , x) / π , i.e. defined as a part of ellipsoid; the

initial conditions at τ are y(τ)=1, yt (τ ) = tan θ . The length of cavity is calculated as a
281
solution of quadratic equation ε 2 x 2 − 2 tan θ x + 1 = 0 ; the abscissa of the largest radius of

cavity is x0 = ε −2 tan θ so the maximum radius is Y=y(x0). Calculations show that the results
obtained by the above formulas can describe a cavitating flow for the middle part of cavity
only and also at very small cavitation numbers. Therefore, the Logvinovich independence
principle can be used for describing a cavitating flow very approximately.
Sec. 8.1.2. Formalization of Logvinovich’ Principle

Logvinovich’ principle of independece of cavity cross sections as stated in the previous sub-
section has been posed as experimental observations. However, via application of the
mathematical method of matched asymptotic expansions, as described above, Logvinovich’
principle may be formalized. In order to do so, it is necessary to apply matched asymptotic
expansions to the more general problem of three-dimensional flow past a slender cavity. As
usual, the fully three-dimensional problem is governed by Laplace’s equation in the field, a
no-flux condition on the surfaces of any wetted portion of the body, and both no-flux and
constant-pressure conditions on the cavity boundary, see Fig.8.1.2 indicating horizontal
motion in a gravity field.)
∇ 2ϕ = 0
u(x) ⋅ n(x) = 0 p ( x ) = pc
0.5
W
y0 g
z
x
Fig.8.1.2. The problem of supercavitating flow past a vehicle moving horizontally in a

gravity field.
The full three-dimensional problem may be stated as follows. Under the

assumptions of ideal flow, the total velocity u ( x) = (W + w ( x ) ) k + ur e r + uθ eθ may be
computed as the gradient of the total velocity potential, u ( x ) = ∇ϕ ( x ) . Here W is the free
stream velocity, and the stream-wise coordinate is in the z -direction (not to be confused
with the general position and velocity vectors, x and u ).
Considering that for many problems of interest the cavity sections are approximately
circular just downstream of the cavitator, it is natural to solve the problem in cylindrical-
polar coordinates. For incompressible flow, the total potential satisfies Laplace’s equation in
∂ 2ϕ 1 ∂ϕ 1 ∂ 2ϕ ∂ 2ϕ
the field: ∇ 2ϕ = 2 + + + = 0.
∂r r ∂r r 2 ∂θ 2 ∂z 2
On the entire hydrodynamic envelope consisting of the cavitator surface Sn and the cavity
surface Sc, the no-flux condition applies:
282
1 ∂R ∂R
ur − uθ − (W + w) = 0 on S n ∪ Sc , (8.1.2)
R ∂θ ∂z
where r = R (θ , z ) is the set of points defining the surface of the cavitator-cavity system.
The surface of the cavity is unknown a priori, and an additional condition is required to
obtain a unique solution of the problem. Specifically, the pressure on the cavity surface
must equal the cavity pressure, p = pc on Sc , which for the current simple case is
considered to be constant in both space and time. Application of Bernoulli’s equation
provides the following dynamic condition on the cavity surface:
pc 1 2 p 1
+ ⎡⎣ur + uθ2 + (W + w) 2 ⎤⎦ + gR sin θ = ∞ + W 2 on Sc , (8.1.3)
ρ 2 ρ 2
where p∞ is the pressure at upstream infinity and ρ is the density of the ambient liquid.
The circumferential coordinate θ is measured from the horizontal plane passing through the
origin.
For the current problem, it will be convenient to make quantities dimensionless with
respect to the total length of the combined cavitator-cavity system, = n + c , (where n is
the cavitator length and c is the cavity length) and the free stream velocity, W . Then the
field equation, equation (5.7.21), is left unchanged, and the boundary conditions become
1 ∂R ∂R
ur − uθ − (1 + w) = 0 on S n ∪ Sc (8.1.4)
R ∂θ ∂z
2
and ur2 + uθ2 + 2 w + w2 + 2 R sin θ = σ on Sc , (8.1.5)
F
where the cavitation number is defined as σ = 2( p∞ − pc ) / ρW 2 (8.1.6)
and the Froude number based on total system length is F =W / g . (8.1.7)
The boundary conditions may be written in terms of the dimensionless total potential
∂ϕ 1 ∂ϕ ∂R ∂ϕ ∂R
as − − = 0 on S n ∪ Sc (8.1.8)
∂r R 2 ∂θ ∂θ ∂z ∂z
2 2 2
⎛ ∂ϕ ⎞ 1 ⎛ ∂ϕ ⎞ ⎛ ∂ϕ ⎞ 2
and ⎜ ⎟ + 2 ⎜ ⎟ ⎜+ ⎟ + 2 R sin θ = σ + 1 on Sc . (8.1.9)
⎝ ∂r ⎠ R ⎝ ∂θ ⎠ ⎝ ∂z ⎠ F
8.1.3. Simplifying Assumptions
The current model thus falls under the category of slender-body theory, which allows the
solution to be treated separately as those for longitudinal and transverse flow problems
according to the classical theory presented in [Van Dyke, 1964; Ashley & Landahl, 1965;
Cole, 1968; Newman, 1977; Kevorkian & Cole, 1996] to name a few. The longitudinal flow
problem is solved as the classical matched asymptotic expansion for slender bodies, in which
an inner solution valid very close to the body is matched to an outer solution valid in the far
field. The variant of this problem applicable to axisymmetric supercavitating flows has been
addressed analytically, to various levels of rigor, by many researchers, for example
[Reichardt, 1946; Münzer & Reichardt, 1950; Garabedian, 1956; Cuthbert & Street, 1964;
Brennen, 1969; Logvinovich, 1969; Chou, 1974; Logvinovich & Serebryakov, 1975; Vorus,
1991; Varghese et al., 1997; Kuria et al., 1997; Serebryakov, 1998; Tulin, 1998; 1999] and
others, and has been discussed above in Secs. 5.7.1-4.
283
To fully formalize Logvinovich’ principle, it remains to address the distortion of this
axisymmetric solution due to lift acting on the cavitator or other surfaces penetrating the
cavity boundary, and for gravitational effects. We address only the latter here. As in Sec.
5.7.4, a simple surrogate has been applied to serve as the axisymmetric solution about which
the distortions occur. The surrogate axisymmetric solution has been selected for
convenience in illustrating the current approach. Any other valid axisymmetric solution may
be substituted, a step that may, in fact, improve accuracy. In fact, if additional fidelity is
desired, more numerically intensive and (presumably) more accurate versions of the
axisymmetric problem have been presented in many articles and texts, for example
Guzevsky (in work dating back to the 1970s, exemplified in English in [Guzevsky, 1979;
2002; Kirschner et al., 1995; Uhlman et al., 1998; Krasnov, 2002; Varghese et al., 2005;
Terentiev, 2005]) which addresses the partially and fully cavitating axisymmetric problem of
one or more bodies. The relative merits of each of the various approximate and high-fidelity
approaches remains the subject of ongoing consideration, especially the validity of the
corresponding results via comparison with data.
It should be noted that the transverse flow problem may be solved using one of
several methods, including a technique involving a multipole expansion in each plane,
wherein the lowest term in the series corresponds to the inner solution of the longitudinal
flow problem, which is therefore treated as known, namely the surrogate axisymmetric
solution mentioned above. The coefficients of the remaining terms in the multipole
expansion are then determined such that the solution satisfies the dynamic and kinematic
condition on the cavity boundary. The resulting method is cast as a marching problem, and,
via application of the Galilean transformation discussed below, the solution of the transverse
flow problem may be thought of as the evolution of a two-dimensional cavity under the
influence of gravity. An extension of this approach is theoretically applicable to cases
involving lift on the cavitator or on other appendages such as canards, and could also be
extended to time-dependent problems.
The resulting method is similar to the Fourier decomposition described in
[Logvinovich, 1969; Buivol, 1980; Logvinovich et al., 1985]. However, whereas those
references state the problem as a geometric perturbation of the cavity geometry about the
axisymmetric geometry, the current approach attempts to formalize and justify the basic
technique in the context of asymptotic and multipole expansions. In this sense, the current
approach is an attempt to link Logvinovich’ principle of independent expansion of cavity
sections [Logvinovich, 1969; Logvinovich et al., 1985; Paryshev, 2006; Savchenko et al.,,
2000; Semenenko, 2001; Kirschner & Arzoumanian, 2008; Paryshev, 1978; Kirshner, 2006;
Buivol, 1980; Logvinovich & Serebryakov, 1975; Serebryakov, 1976] to the classical
asymptotic expansions as originally developed for fully-wetted bodies, and to overlay a
physically realistic asymmetric distortion in a simultaneous Fourier-type expansion for both
the cavity geometry and the potential, resulting in a multipole representation of the latter.
Rather than a geometric perturbation of the cavity geometry, the result is a more
straightforward Fourier expansion of the in-plane potential, amenable to a simpler
specification of classical mixed boundary conditions which are, in general of the Robin type.
However, the two methods should be essentially equivalent, and it is expected that they
should provide similar if not identical solutions.
One cost of the current approach is the appearance of the multipole singularity
within the cavity contour in each plane. As will be illustrated below, if not treated carefully,
284
this is associated with unrealistic distortion at the end of the cavity. In contrast, the
Logvinovich approach appears to lack a singularity within the cavity contour. Since both
methods hint at the twin-vortex regime as the computational plane approaches cavity closure
for low values of the cavity Froude number, a more complex solution topology is suggested,
further development of which will be deferred.
The slender system formed by the cavitator-cavity system allows for a simplification
of the type presented for flows past fully-wetted bodies by the authors described above.
With the total system length defined above, and the maximum diameter d m , then the
slenderness parameter, ε = d m , is small for the cases of interest. Although the formulation
may be extended to more general cases, it will be assumed that the body is moving at
constant forward speed. Cavity dynamics effects will be ignored for the present, such that
the entire flow field is steady.
Roughly following Newman [1977], the solution of the longitudinal flow problem
proceeds as a matched asymptotic expansion, wherein an inner solution for the disturbance
velocity potential Φ ( r ,θ , z ) is sought, valid near the surface of the cavitator-cavity system,
satisfying the boundary conditions on that surface and the two-dimensional Laplace
equation. This solution does not satisfy the condition at infinity in the outer field, and it
contains an arbitrary additive constant. A corresponding outer solution for the disturbance
velocity potential φ ( r ,θ , z ) with the proper behavior at infinity has the form of a three-
dimensional source distribution along the axis of the cavitator-cavity system. The inner
expansion of the outer solution is required to match the inner solution in a matching region
ε r . After matching, resulting expressions for the inner and outer solutions of the
1 r
longitudinal flow problem are given by Φ = − S ' ( z ) log + f ( z ) , r = ε , (8.7.10)
2π
1 r
and φ =− S ' ( z ) log + f ( z ) , r= , (8.1.11)
2π
where S ( z ) is the sectional area of the combined cavitator-cavity system, and
1
z
2 ( z − z ') 1 2 ( z '− z )
f ( z) = ∫ S '' ( z ') log dz ' − ∫ S '' ( z ') log dz '
4π 0 4π z
1 2 z − z' (8.1.12)
∫ S '' ( z ') sgn ( z − z ') log
= dz '.
4π 0
The cavitator-cavity system need not be axisymmetric, but all of the asymmetry is
reflected in the inner solution; its outer limit and the outer solution are axisymmetric. It is
important that the argument of the logarithm appearing in the solution is made dimensionless
with a quantity on the order of the length of the combined cavitator-cavity system, in order
that proper matching is achieved in the overlap region.
For the supercavitation problem, S '( z ) is unknown in way of the cavity, and must be
determined as part of the solution process. For purposes of simplicity in the current analysis
(more than any consideration of accuracy), the solution of the axisymmetric longitudinal
flow problem will be estimated as a Garabedian cavity, that is, a cavity profile for which the
maximum cavity diameter d m and the length of the combined cavitator-cavity system are
approximated by Garabedian’s formulae [Garabedian, 1956]:
285
d m / d n = CD / σ and / d n = (CD / σ 2 ) log(1/ σ . (8.1.13)

Here d n is the cavitator diameter at cavity detachment (or a characteristic diameter over the
locus of detachment if it is not axisymmetric) and CD is the cavity drag coefficient of the
cavitator based on its projected area. It should be noted that the second of Garabedian’s
formulae (8.1.13) is generally written for the length of the cavity, rather than the length of
the combined system. However, since it is an expression for the asymptotic behavior as the
cavity length increases, the distinction is negligible.
For the current illustration, the Garabedian cavity dimensions were applied to
specify a cavity of ellipsoidal form. For other applications, the authors have often applied
the Münzer-Reichardt [1950] profile to estimate the cavity shape, but, since it violates the
slender-body assumptions at the cavity endpoints and would thus require complicating
corrections, the simpler shape was selected. Moreover, in general the Münzer-Reichardt
profile does not match commonly applied cavitator profiles at the cavity detachment point,
and modification of the cavity profile is required to improve the estimate of the geometry in
that region.
For the present, this surrogate solution comprised of the Garabedian ellipsoid will
serve as the known leading axisymmetric term in a multipole expansion, the shape of which
is distorted asymmetrically by the remaining terms in the multipole expansion in order to
satisfy the boundary conditions for the case of a cavitating body traveling horizontally in a
gravitational field. Although reasonably straightforward in theory, the extension of this
model to the case of a lifting body will be deferred to a future publication.
8.1.4. The Inner Problem
For the inner problem, the solution must satisfy the field equations very close to the
surface of the combined cavitator-cavity system, along with the consistent boundary
condition. Since a prerequisite to this assumption is the slenderness of the body, stream-
wise gradients are dominated by gradients in the transverse plane, and the governing
equation of the inner problem reduces to the two-dimensional Laplace equation:
∂ 2 Φ 1 ∂Φ 1 ∂ 2 Φ
∇ 22D Φ = 2 + + = 0.
∂r r ∂r r 2 ∂θ 2
Consistent with the slenderness approximation, the boundary conditions become
∂R ∂Φ 1 ∂Φ ∂R
= − on S n ∪ Sc , (8.1.14)
∂z ∂r R 2 ∂θ ∂θ
2 2
∂Φ σ 1 1 ⎛ ∂Φ ⎞ 1 ⎛ ∂Φ ⎞
and = − 2 R sin θ − ⎜ ⎟ − ⎜ ⎟ on Sc . (8.1.15)
∂z 2 F 2 ⎝ ∂r ⎠ 2 R 2 ⎝ ∂θ ⎠
These boundary conditions have been rearranged to highlight the distinction between the
axial and transverse aspects of the problem: from a numerical standpoint it can be seen that
the solution can be treated as a nonlinear system of differential equations in the stream-wise
direction in the form of successive transverse flow problems, although the nonlinearities
inherent in the system must be addressed.
8.1.5. Galilean Transformation of the Inner Solution
It is conceptually convenient to convert the boundary-value problem for the inner
solution to the problem of the approximate Lagrangian “evolution” of a cavity section as it
286
traverses the system from the cavity detachment locus on the cavitator to the point of cavity
closure by applying the Galilean transformation: z = Wt. (8.1.16)
Under this transformation, the first partial derivative in the stream-wise direction is given by
∂ 1 ∂ ∂ ∂
= , or in dimensionless form ( z = t ), = . (8.1.17)
∂z W ∂t ∂z ∂t
Substituting these expressions into the time-independent form of the boundary
conditions, equations (8.1.14) and (8.1.15), produces the following kinematic and dynamic
conditions governing the “evolution” of a cavity section:
∂R ∂Φ 1 ∂Φ ∂R 1 ∂R
R= = − 2 = ur − uθ on Sn ∪ Sc (8.1.18)
∂t ∂R R ∂θ ∂θ R ∂θ
2 2
∂Φ σ R sin θ 1 ⎛ ∂Φ ⎞ 1 ⎛ ∂Φ ⎞ σ R sin θ ur2 uθ2
and Φ = − − ⎜ ⎟ − ⎜ ⎟ = − − − on Sc . (8.1.19)
∂t 2 F2 2 ⎝ ∂r ⎠ 2 R 2 ⎝ ∂θ ⎠ 2 F2 2 2
It can be seen that the transformed dynamic condition, equation (8.1.19), has the same form
as the dynamic condition – derived from the unsteady Bernoulli equation – governing the
evolution of an unsteady two-dimensional cavity.
Sec. 8.2. A Simple Approach to Estimating 3D Supercavitating Flow

8.2.1. Multipole Distribution for Slender Cavities
Consider again the problem presented in Sec. 8.1, namely, ideal flow past a long, slender
cavity. As was decribed therein, the fully three-dimensional problem can be addressed via
the method of matched asymptotic expansions, resulting in the following form for the inner
and outer problems (after matching):
1 r
Φ=− S ' ( z ) log + f ( z ) , r =ε , (8.2.1)
2π
1 r
and φ = − S ' ( z ) log + f ( z ) , r= , (8.2.2)
2π
where S ( z ) is the sectional area of the combined cavitator-cavity system, and
1 2 z − z'
f ( z) = ∫ S '' ( z ') sgn ( z − z ') log dz ' . (8.2.3)
4π 0
The total cavitator-cavity system length is l defined above, and the maximum cavity
diameter is d m . The slenderness parameter, ε = d m , is small for the cases of interest.
All of the asymmetry is reflected in the inner solution; its outer limit and the outer
solution are axisymmetric. Various approaches to predicting the axisymmetric outer
solution are presented elsewhere in this text. (See, especially, [Loyzyansky, 1959.) The
remainder of this sub-section presents one approach to predicting the asymmetric part of the
inner solution. The method is closely related – but not identical – to the solution presented
in [Logvinovich, 1969; ] and the work of Buivol (see [Logvinovich et al, 1985]), which
involves a somewhat different Fourier expansion, specifically for the geometric distortion of
the nominally axisymmetric geometry.
As is presented in [Kirschner et al, 2009], the method employs the Galilean-
transformed forms of the kinematic and dynamic conditions governing the “evolution” of a
cavity section as presented in previous section:
287
∂R ∂Φ 1 ∂Φ ∂R 1 ∂R
R= = − 2 = ur − uθ on Sn ∪ Sc , (8.2.4)
∂t ∂R R ∂θ ∂θ R ∂θ
2 2
σ 1 1 ⎛ ∂Φ ⎞ 1 ⎛ ∂Φ ⎞ σ 1 1 2 1 2
and Φ = − 2 R sin θ − ⎜ ⎟ − 2⎜ ⎟ = − 2 R sin θ − ur − uθ on Sc .(8.2.5)
2 F 2 ⎝ ∂r ⎠ 2 R ⎝ ∂θ ⎠ 2 F 2 2
Thus, the challenge for the numerical approximation is to predict the evolutionary solution to
this system of equations starting with some initial condition representing the flow at the
upstream end of the cavity, at or near the point of detachment from the cavitator.
Separating the two-dimensional Laplace equation in cylindrical-polar coordinates
leads to the classical two-dimensional multipole expansion for the velocity potential (see, for
example, [Newman, 1977]):
∞
1
Φ ( r ,θ , t ) = A0 ( t ) ln r + ∑ m ⎡⎣ Am ( t ) cos ( mθ ) + Bm ( t ) sin ( mθ ) ⎤⎦ , (8.2.6)
m =1 r
where the “time” dependence of the coefficients has been retained in consideration of the
results of the Galilean transformation discussed above. The cavity radius may also be
expressed in terms of a Fourier series:
∞
R (θ , t ) = R0 ( t ) + ∑ ⎡⎣ Rcm ( t ) cos ( mθ ) + Rsm ( t ) sin ( mθ ) ⎤⎦ . (8.2.7)
m =1
The partial time derivatives of the cavity surface function and the potential are thus given by
∞
R (θ , t ) = R0 ( t ) + ∑ ⎡⎣ Rcm ( t ) cos ( mθ ) + Rsm ( t ) sin ( mθ ) ⎤⎦ (8.2.8)
m =1
∞
1
and Φ ( r ,θ , t ) = A0 ( t ) ln r + ∑ m ⎣ m( )
⎡ A t cos ( mθ ) + Bm ( t ) sin ( mθ ) ⎤⎦ . (8.2.9)
m =1 r
Evaluating the potential on the cavity surface Sc , at r = R (θ ) , and substituting the resulting
expression and equation (8.2.7) into the dynamic and kinematic conditions, equations (8.2.4)
and (8.2.5), produces the following nonlinear system of dynamical equations governing the
evolution of the Fourier coefficients of index greater than zero:
∞
u R'
∑m =1
⎡⎣ Rcm cos ( mθ ) + Rsm sin ( mθ ) ⎤⎦ = − R0 + ur − θ
R
on Sc , (8.2.10)
∞
1 σ 1 1 1
and ∑ m ⎣⎡ Am cos ( mθ ) + Bm sin ( mθ ) ⎦⎤ = − A0 ln R + − 2 R sin θ − ur2 − uθ2 , (8.2.11)
m =1 R 2 F 2 2
where primes denote differentiation with respect to θ , and where the leading Fourier
coefficients for the cavity geometry and velocity potential are treated as known from the
axisymmetric base solution. The local in-plane velocity components on the cavity surface
are the appropriate in-plane derivatives of the velocity potential:
A (t ) ∞ 1
ur ( R ,θ , t ) = 0 − ∑ m m +1 ⎡⎣ Am ( t ) cos ( mθ ) + Bm ( t ) sin ( mθ ) ⎤⎦ , (8.2.12)
R m =1 R
∞
m
and uθ ( R,θ , t ) = −∑ m +1 ⎡⎣ Am ( t ) sin ( mθ ) − Bm ( t ) cos ( mθ ) ⎤⎦ . (8.2.13)
m =1 R
The angular derivative of the cavity surface function is computed as
∞
R ' (θ , t ) = −∑ m ⎡⎣ Rcm ( t ) sin ( mθ ) − Rsm ( t ) cos ( mθ ) ⎤⎦ . (8.2.14)
m =1
288
The “time”-dependent area of the cavity is given by
π
1 π ∞
A ( t ) = ∫ R 2 (θ , t ) dθ = π R02 ( t ) + ∑ ⎡⎣ Rc2m ( t ) + Rs2m ( t ) ⎤⎦ . (8.2.15)
2 −π 2 m =1
It can be seen that the area of the base solution is therefore a lower bound on the area of the
distorted solution, a fact that might serve as the basis for a improved approximation via
iteration on the base inner solution.
The solution of the problem of three-dimensional flow past a slender cavitating body
in a gravity field is thus reduced to determining the evolution of the coefficients describing
the velocity potential and the geometry of each section as it evolves downstream of its
generation at the cavitator. The leading coefficients of each are treated as known from the
selected axisymmetric base solution. The initial values of the remaining coefficients for the
cavity geometry and velocity potential are specified to properly represent the cavitator
geometry and flow at cavity detachment. The solution then proceeds numerically by
computing the evolution of the remaining coefficients in truncated forms of the two series.
Truncating the Fourier expansion at M terms, the kinematic and dynamic
conditions, equations (8.2.10) and (8.2.11), may be approximately satisfied at K points as
M R km Rm = U k ; k = 1: K , m = 1: 2M , (8.2.16)
and M Φ km Φ m = Pk ; k = 1: K , m = 1: 2 M , (8.2.17)
⎧⎪ Rc m ( t ) , m = 1: M
where Rm = Rm ( t ) = ⎨ ; (8.2.18)
⎪⎩ Rs m − M ( t ) , m = M + 1: 2M
⎧⎪ Am ( t ) , m = 1: M
Φm = Φm (t ) = ⎨ ; (8.2.19)
⎪⎩ Bm − M ( t ) , m = M + 1: 2M
uθ ( R (θ k , t ) ) ∂R (θ k , t )
U k = U k ( Rm , Φ m ) = ur ( R (θ k , t ) ) − − R0 (θ k , t ) ; (8.2.20)
R (θ k , t ) ∂θ
Pk = Pk ( Rm , Φ m ) =
and 1 2 R (θ , t )sin θ 1 (8.2.21)
=−
2
( ur ( R (θ , t )) + uθ2 ( R (θ , t )) ) −
Fr 2
+ σ − A0 ( R (θ , t ) ) ln R(θ , t ).
2
The so-called mass matrices are defined as
⎧⎪cos ( mθ k ) , m = 1: M
M R km = ⎨ ; (8.2.22)
⎪⎩sin ( ( m − M )θ k ) , m = M + 1: 2M
⎧ cos ( mθ k )
⎪ m , m = 1: M
⎪ R (θ k , t )
and M Φ km = M Φ km ( Rm ) = ⎨ (8.2.23)
⎪ sin ( ( m − M )θ k )
⎪ R m − M (θ , t ) , m = M + 1: 2M .
⎩ k
(These are not related to the actual mass of the fluid nor the added mass of the cavitator-
cavity system, but are so denoted using the parlance of the theory of numerical solution of
nonlinear ordinary differential equations.)
289
It is emphasized that the geometry mass matrix defined in equation (8.2.22) does not
depend on the solution, and therefore is independent of time, whereas the potential mass
matrix defined in equation (8.2.23) does depend on time via the geometric solution but not
via the potential.
The system can be concatenated to form the following single partitioned matrix equation
comprising the approximate system of linearly implicit, nonlinear ordinary differential
equations governing the time-dependent behavior of a cavity in a gravity field:
M αβ xβ = bβ ; α = 1: 2, β = 1: 2 , (8.2.24)
⎧ M R km , α = 1, β = 1
⎧ Rm , β = 1 ⎧U m , β = 1 ⎪
where xβ = ⎨ , bβ = ⎨ , M αβ ( x, t ) = ⎨ M Φ km , α = 2, β = 2
⎩Φ m , β = 2 ⎩ Pm , β = 2 ⎪0, otherwise
⎩
The structure of the concatenated solution system may be written in matrix form as
⎛ M 11 = M Rkm M 12 = 0 ⎞ ⎛ x1 = Rm ⎞ ⎛ b1 = U m ⎞
⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ ⎟⎜ ⎟=⎜ ⎟
⎜ M =0 M = M ⎟⎜ x = Φ ⎟ ⎜ b = P ⎟
⎝ 21 22 Φkm ⎠ ⎝ 2 m⎠ ⎝ 2 m ⎠
Verification and preliminary validation of the model for supercavitating flow in a

gravity field has proceeded by exploring the cavity shape at a fixed cavitation number as the
cavity Froude number varies. The predicted behavior is illustrated in the following set of
examples, after which some issues associated with implementation of the code are discussed.
Logvinovich [1969] provides several formulae for the distortion of the line of
centers of the cavity as a function of cavity Froude number. The most reliable seems to be
the following (here given in the “time”-dependent form):
h ( t; F ) =
(1 + σ ) t 2
. (8.2.26)
6F 2
Using this expression to distort a base solution comprised of a parabolic cavity contour
defined by Paryshev (2006) with minor corrections made to the cavity dimensions as
described in Kirschner and Arzoumanian [2008], it is possible to plot an estimated profile of
the cavity as it intersects its plane of bilateral symmetry. The results of the current model
may then be overlaid on that profile for purposes of comparison, as presented for a cavitation
number of σ = 0.030 and for several values of the cavity Froude number. The results are
presented in Fig.8.2.1. On the left-hand plots, the predicted profiles are compared with the
distorted base solution. Bear in mind that the comparison is for the profile intersecting the
plane of symmetry in both cases. The right-hand plots show transverse slices through the
cavity, some of which have been labeled to facilitate correspondence with the associated
location in the profile view. It can be seen that the profile comparisons are quite good. The
model also predicts increasing distortion of the planar cavity sections as the downstream end
of the cavity is approached. These effects are more pronounced at lower cavity Froude
numbers, where the effect of gravity is more important.
290
Fig.8.2.1. Results of the current model (dashed lines in plots on left; all contours on right)
compared with low-order estimates of cavity distortion based on the Logvinovich formula
(solid lines in plots on left) for flow past a non-lifting supercavitating disk for a cavitation
number of σ=0.030 and for several values of the cavity Froude number based on total
system length: a – F=2.5; b – F=5.0; c – F=7.5; d – F=10.0. Note that the horizontal and
vertical scales do not match in the profile views shown on the left-hand plots.
The method is computationally efficient: the results shown in Fig.8.2.1, which were
computed at 300 stations along the cavity length, required only a few seconds to compute on
a modern laptop computer. Provided the multipole is allowed to migrate with the contour of
the cavity sections as they evolve, the method appears to be robust for the first few terms of
291
the multipole expansion. If too many terms are included, however, the computation
becomes ill-conditioned. At a certain point the geometry predicted by the method becomes
physically unrealistic. Beyond some number of terms, rank deficiencies are reported in the
matrix solver. These issues are illustrated in Fig.8.2.2, which compares the physically
realistic solution of Fig.8.2.2a with the problematic solutions shown in Fig.8.2.2b.
Fig.8.2.2. Some issues associated with numerical implementation of the current model
(example computations performed for cavitation number σ=0.030 and cavity Froude
number Fr=2.5): a – the physically reasonable result previously presented in Fig.8.2.1a;
b – illustration of numerical problems that arise if too many terms of the multipole
expansion are retained; c – illustration of the need to allow the multipole to migrate with the
cavity sectional contour. Note that the horizontal and vertical scales do not match in the
profile views shown on the left-hand plots.
The results of Fig.8.2.2a (which are the same as those shown in Fig.8.2.1a) were
computed with the series truncated at M=6 and the multipole specified to migrate upward at
292
a rate proportional to the square of the time-like variable. The results of Fig.8.2.2b were
computed with the series truncated at M=9. It can be seen that the downstream end of the
cavity becomes unrealistically distorted. It has been shown that such numerical issues may
be avoided if the boundary-value problem at each computational plane is re-scaled to make
the average cavity radius around the local circumference take on a value of unity.
Fig.8.2.2c shows the results if the multipole is not allowed to migrate as the cavity
rises. Although the upstream portion of the solution is remarkably robust to the choice of
multipole location, as the solution proceeds the predicted cavity boundary comes so close to
the multipole that the solution distorts very rapidly in a physically unrealistic way until, a
short time later, the solution can no longer proceed numerically.
Thus, based on the presentation above, a simple multipole method may be
formulated and implemented to estimate the effects of gravity on a slender supercavity. By
way of review, the method extends classical slender-body theory to allow computation of the
terms of a multipole expansion for the inner solution of the potential flow problem. For
convenience in illustrating the method, the lowest-order term for the axisymmetric flow was
assumed known from some other solution method. The method allows for numerical
marching of the non-axisymmetric distortion the cavity sectional geometry and concurrent
evolution of the velocity potential in the stream-wise coordinate, which is treated as a time-
like variable. Preliminary validation against an independent formula for the tail-up of the
line of cavity centers shows promising results. Work remains to improve the numerical
method to allow computation for higher terms of the multipole expansion, and to provide for
a rational approach to moving the multipole as the cavity contour deforms. Most
importantly, future efforts should address the effects of lift, maneuvering, and afterbody
planing interaction on the cavity geometry. In principle, the method could be extended to
unsteady flow.
Sec. 8.3. Mathematical Model of an Unsteady Supercavity*

8.3.1. Equation of Cavity Cross-Section Expansion
The process of cross-section expansion for an axisymmetric cavity is approximately
described by differential equation (8.1.1) derived by Logvinovich in 1959. If k is constant,
t u
k
the integral of equation (8.1.1) is S (h, t ) = S0 + S0' V (th ) (t − th ) − ∫ ∫ Δp (h, v) dv du , (8.3.1)
ρt h th
Here S0 = π R and S = S (σ ) is initial rate of cavity expansion (a weak function of

2
n
'
0
'
0
cavitation number), th = th (h) is the instant when the cavitator

passes through coordinate h, Fig.8.3.1.
Fig.8.3.1
Sketch of a cavity at arbiter moving of a cone.
The coefficients of equation (8.1.2) were calculated for

various cavitators as S0' / 2π Rn = f1 (σ ), k / 4π = f 2 (σ ) (the
approximations by Guzevsky [1979] for Ryaboushinsky’s
*
This section was prepared by Emil Paryshev
293
model were used). For example, at σ=0.02 we have the following values for a disk: f1=0.439,
and f2=0.219. Actually, the dependence of functions f1 and f2 on σ have moderate effect on
the size of cavity mid-section. It should be noted that coefficient f2(σ), which governs the
narrowing of cross-section, is virtually the same for all types of cavitators.
8.3.2. Closed Problem Formulation
A cavity is regarded as a variable-volume gas-filled space with gas injection and losses.
Equation (8.3.1) has to be completed with the following:
h3 ( t )
• expression of the gas volume: Q (t ) = ∫ [ S ( h, t ) − S

H (t )
T (h, t ) ] dh , (8.3.2)
where ST - cross-sectional area of the inside cavity space; H - cavity inception coordinate;
h3 - coordinate of the section of cavity closure (Fig.8.3.1);
• equation of gas mass balance inside the cavity
d
(Q ρ k ) = mn − m y , ρ k = ρ k ( pk ) , (8.3.3)
dt
where ρk - gas density in the cavity (constant throughout the whole volume; no wave
processes in the gas are considered); mn , m y - mass rates per second for gas injection and
losses;
dV
• equation for cavitator motion = F (V , H , h3 , p∞ , pk ,…) . (8.3.4)
dt
8.3.3. Delay Differential Equations
The integral-differential system (8.3.1–8.3.4), which does not have any regular methods of
solution apart from numerical analysis at discrete sections, is transformed into the system
(8.3.5) of non-linear time-delay differential equations, the time lag being variable. The main
idea of the transformation is to exclude integral relations by means of differentiating them
with respect to a parameter (time in this particular case).
∂S ( h3 , t )
h
k 3 k
Q = − ∫ p∞ (h, t ) dh + pk (t )l + h3 + S0' V 2 − V [ ST (l ) − S0 ] − VST' (l )l ;
ρH ρ ∂t
∂S (h3 , t ) / ∂t − VST' (l ) h3
h3 = − ; l = h3 + V ; l = h3 − H ; H = −V ; τ = 1 + ;
∂S (h3 , t ) / ∂h − ST (l )
'
V (t − τ )
d ∂S (h3 , t ) ⎧⎪ S ' V (t − τ ) k ⎡ Δp (h3 , t − τ ) t ∂p∞ (h3 , v) ⎤ ⎪⎫ k
= −⎨ 0 + ⎢ + ∫ dv ⎥ ⎬ h3 − Δp (h3 , t ); (8.3.5)
dt ∂t ⎩⎪ V (t − τ ) ρ ⎣ V (t − τ ) t −τ
∂h ⎦ ⎭⎪ ρ
∂S (h3 , t ) ⎡ V (t − τ )τ ⎤ k ⎡ Δp (h3 , t − τ )τ t u
∂p∞ (h3 , v) ⎤
= S0' ⎢1 − ⎥− ⎢ + ∫ ∫τ dv du ⎥ ;
∂h ⎣ V (t − τ ) ⎦ ρ ⎣ V (t − τ ) t −τ t− ∂h ⎦
t
∂S (h3 , t ) k
Δp (h, t ) = p∞ (h, t ) − pk (t ), ST = ST ( x); = S0' V (t − τ ) − ∫ Δp(h3 , v) dv;
∂t ρ t −τ
where l is the cavity length, x = h − H is the distance between the cavitator and the given
cross-section; t-τ is the instant when the cavitator passes through coordinate h3 ; τ is a time
294
lag. It is assumed that the gas pressure inside the cavity is almost constant lengthwise and
depends on time only [Michel, 1973].
Equations (8.3.5) are completed with the differential equation for gas pressure
mn − m y − CQpkn
pk = , (8.3.6)
nCQpkn −1
derived from (8.3.3) assuming that gas density and pressure conform to the polytropic
equation ρ k = Cpkn , where 1/ n is a polytropic factor. Used here are the data on gas
withdrawal along the trailing vortices [Cox & Clayden, 1955; Epshteyn, 1961], and
under conditions close to vapour cavitation [Logvinovich, 1969, p. 146], as well as a model
of withdrawal due to gas jet ejection. It is modeled numerically for cavity pulsations
[Paryshev, 1985] due to periodic pinching and detachment of cavity tail part, using non-
linear equations (8.3.5) and (8.3.6), see Fig.8.3.1. The gas withdraws in the form of periodic
detachment of bubbles, as was observed in experiments conducted by Michel [1973] and
others.
Fig.8.3.2.
The wavy cavity shapes.
By selecting special scales, e.g. different

length-wise and in radial direction, the system of
equations (8.3.5) – (8.3.6) can be transformed to a dimensionless form (the bars above all
dimensionless parameters in the equations are omitted herein):
∂S (h3 , t ) (1 − Δp )Q′ − qn ∂S (h3 , t ) / ∂t
Q′′ = −τ Δp + h3′ + 1, Δp′ = , h3′ = − ,
∂t Q ∂S (h3 , t ) / ∂h
t
(8.3.7)
∂S (h3 , t ) ∂S (h3 , t )
l ′ = h3′ + 1, τ = l , = 1 − ∫ Δp (v) dv, = 1 − τ Δp(t − τ ).
∂t t −τ
∂h
2σ v Qn
Dimensionless value qn = is the only governing parameter. Equation (8.3.7)
π cxo Vd 2
controls partial detachments as well.
Sec. 8.4. Linear Theory of Cavity Stability and Oscillations

8.4.1. Problem Formulation
Nonlinear equations (8.3.5) – (8.3.7) are still too complicated for theoretical analyses and
computing. But they can be simplified for some certain conditions. Considered below is the
simplest cavitation flow under steady input conditions, namely an axisymmetric cavity in a
weightless liquid ( p∞ = const ) at constant cavitator velocity V and constant gas mass in the
cavity ( Qpkn = const ). After linearization, equations of cavity dynamics (8.3.5) – (8.3.7)
yield the equation for small pressure oscillations:
2 2
pk′′′( t ) + pk′′ ( t ) + pk′ ( t − τ 0 ) − pk ( t ) + pk ( t − τ 0 ) = 0 , (8.4.1)
τ0 τ0
where T = n ρ Q0 /(kl0 pk 0 ) is the time scale, t = t / T is the dimensionless time; Pk0, Q0,
295
l0, τ0=l0/V are steady parameters. Derivative sign ′ denotes derivatives with respect to t .
Equation (8.3.7) and the total problem have only one governing parameter, the
dimensionless time lag τ 0 = τ 0 /T . Approximating the cavity shape by an ellipsoid of
12 pk 0 12 ⎛ Eu ⎞
revolution yields: τ0 = ⋅ = ⎜ − 1⎟ . (8.4.2)
n p∞ − pk 0 n⎝σ ⎠
8.4.2. Characteristic Equation and Eigenvalues

Time-delay equation (8.3.7) belongs to the class of differential equations with a divergent
argument, whose mathematical analysis is usually very difficult. Equation (8.3.7) leads to
the following transcendental characteristic equation
2 2
λ 3 + λ + λ e− λτ 0 − + e − λτ 0 = 0 , (8.4.3)
τ0 τ0
with an infinite set of eigenvalues (roots). A method for solving such equations has been
developed based on numerical analysis of complex eigenvalues trajectories depending on
one or several parameters, the method that makes it possible to determine a root of any given
number [Paryshev, 1978].
The essence of the method is as follows. When τ 0 = π j 2 ( j = 1, 2,…) , equation
(8.4.3) has pure imaginary eigenvalues λ j = iω j , ϖ j = 2 . By differentiating (8.4.2) with
dλ 1 2 − exp(−λτ 0 ) (λ 2τ 02 + 2λτ 0 + 2)
respect to τ 0 , yields =− 2 , which is equivalent to
dτ 0 τ 0 3λ 2 + 1 − exp(−λτ 0 ) (λτ 0 + 1)
dμ 1 AC + BD d ω 1 BC − AD
two equations in terms of real variables: =− 2 2 , =− 2 2 ,
dτ 0 τ0 C + D 2
dτ 0 τ 0 C + D2
where A, B, C and D are the appropriate functions of τ 0 , μ и ω .
These equations are integrated numerically with initial conditions
τ 0 j = π j 2, μ (τ 0 j ) = 0, ω (τ 0 j ) = 2 .
Thus, the total system of eigenvalues of equation (8.4.2) is determined, which includes a
zero root of the third order and an infinite set of complex eigenvalues
λ j (τ 0 ) = μ j (τ 0 ) ± iω j (τ 0 ), j = 1, 2, …, 0 ≤ τ 0 < ∞ . (8.4.4)
Note that the third order zero root may cause extraneous solutions like c1 + c2 t + c3t 2
if the accuracy of the numerical integration of (8.3.7) is not sufficient.
8.4.3. Examination of the System of Eigenvalues
The obtained system of eigenvalues has a very well-defined structure, which makes it
possible to find out the main cavity dynamics properties. The trajectories of the first three
roots of equation (8.4.3) are plotted in Fig.8.4.1. Dimensionless circular frequency ω j (τ 0 )
forms a family of monotonically decreasing functions. The real parts of eigenvalues μ j (τ 0 )
are mutually similar intersecting functions that intersect the horizontal axis at points
τ 0 = π j 2 . To the left of such a point, μ j < 0 , i.e. jth eigenvalue is a stable one. To the right
of this point μ j > 0 , i.e. the eigenvalue is an oscillatory unstable one. The same is for j.
296
Curves μ j (τ 0 ) show that a cavity is asymptotically stable at τ 0 < π 2 , and

oscillatory unstable for all τ 0 > π 2 with the stability
boundary laying at the value of τ 0 = π 2 ( Eu / σ = 2.64 ).
The instability of a linear system transforms to real nonlinear
systems (which a cavity can be referred to) into self-excited
oscillations (cavity pulsations) with finite amplitude and with
a frequency equal or close to that of the linear system. The
main nonlinear factor in the case of a cavity is the detachment
of gas bubbles caused by the pinching of the cavity tail part
by traveling waves.
Fig.8.4.1.
Dependences of dimensionless circular frequency ω j (τ 0 )
and eigenvalues μ j (τ 0 ) on dimensionless time τ 0 .
The cavity oscillation frequency is determined by the

physically meaningful root which has the greatest real part
among all the other unstable roots at given τ 0 . Thus, the dimensionless frequency of
asymptotic oscillations (actually the frequency of self-oscillations) is a saw-tooth function of
τ 0 , Fig.8.4.2, consisting of portions of consequent curves ω j (τ 0 ) . The greater the value of
τ 0 , the smaller is the tooth height; also ω ≈ 1 at increasing values of τ 0 . The tooth spacing
seems to be 2π. The dimensional circular frequency of oscillations is
ω = ω / T = ω kpk 0l0 /(nQ0 ρ ) . (8.4.5)
Fig.8.4.2. The stepwise dependences of ω j and μ j on τ 0 for self-excited oscillations of

cavity pulsations.
297
If τ 0 is great enough, then ω ≈ 1 and ω is proportional to pk 0 . Pressure

oscillations inside cavity are accompanied by waves on the cavity surface with wavelength
lb = 2π V / ω . The number of waves lengthwise the cavity is N = l0 lb = ωτ 0 (2π ) . Function
N (τ 0 ) calculated using the “saw” in Fig.8.4.2 is shown in Fig.8.4.3. This is also a
discontinuous function. The number of waves along the cavity is close to an integer value
equal to j of the dominating root. In the discontinuity points, N increases stepwise by ∼1. A
simple estimate at ω ≈ 1 is:
τ 0 1 3 ⎛ Eu ⎞
N≈ = ⎜ − 1⎟ .
2π π n ⎝ σ ⎠
Fig.8.4.3.
The stepwise pulsation mode;
experimental points are obtained in
[Epshteyn & Lapin, 1985].
8.4.4. Comparison with Experiments

The present theory clearly explains the most
typical features of cavity oscillations, both in axisymmetric and planar cases. The results of
numerous experiments and their detailed examination for oscillations of planar cavities past
a wedge are presented in [Michel, 1973]. The detailed comparison [Paryshev, 1978] shows
that, generally, the features described in that paper are quite close to those given by the
present theory.
The theory and the experiment for axisymmetric cavities past disks have been
compared on the basis of multiple experimental data by different authors, the data varying
greatly in terms of test conditions (10≤Fr≤120, 0.02≤σ, 0.04≤Eu≤10). Some experimental
data reduced to the form of ω (τ 0 ) are plotted as points over the theoretical line in Fig.8.4.2.
The experiments at Eu=0.05 and Eu=0.5 ( τ 0 ≈ 20 ) were conducted in an open natural
reservoir and were not intended specifically for investigating oscillations, that explains some
scattering. Experiments at Eu=2–6 were made in a water tank especially for pulsation
studies. Positions of points on the theoretical plot depend on the value of 1/n. Solid points
are related to 1/n =1 (isothermal process in gas), hollow points are for 1/n =1.4 (adiabatic
process). Comparing positions of solid and hollow points of the same shape, one can see that
better matching between test and theory is for 1/n =1. Most typical, from this viewpoint, are
the experimental cavities with 12 waves. The appropriate point in Fig.8.4.2 lies within the
12-wave zone at 1/n =1 and within the 15-wave zone at 1/n =1.4. Stepwise pulsation mode
change is a subject of comparison between the theory and the Epshteyn’s experiments on
oscillations of horizontal disk-induced cavities, Fig.8.4.3. The number of waves along the
cavity are plotted by points as a function of τ 0 together with theoretical stepwise
dependence N (τ 0 ) . Using different polytropic factors 1/n shows that the best coincidence
between the theory and experiment is for 1/n =1.1, i.e. an almost isothermal process.
298
8.4.5. Major Physical Conclusions

12 ⎛ Eu ⎞
• The parameter of similarity for unsteady ventilated cavities is ⎜ − 1⎟ . τ0 =
n⎝σ ⎠
• The generic property of ventilated cavities under various conditions is their oscillatory
instability, which is behind the phenomenon of pulsation (self-oscillations) of cavities.
• The number of waves along an oscillating cavity changes in steps by unity (change of
oscillation mode) under continuous variation of flow parameters (particularly τ 0 ).
Sec. 8.5. Development of the Linear Theory

Using the developed method one can solve some other problems of unsteady cavitation in a
linear formulation.
8.5.1. Vertical Cavities in a Gravity Field
Dynamic properties of vertical cavities in a liquid have been investigated for the case where
gravity cannot be ignored. The peculiarity of such cavities is the vertical gradient of external
pressure. The vertical cavities have the same dynamic properties as those in a weightless
liquid. Parameter τ 0 for vertical cavities is determined based on external pressure p∞ (h) in
the middle of cavity length.
8.5.2. Cavities with Loss of Gas
Another issue of great interest is gas loss from the cavity and its effect on the cavity itself.
The cavity model with constant gas mass inside leads to the conclusion that a cavity is
always unstable if τ 0 ≥ π 2 . Meanwhile experiments in a test tank ( τ 0 ≈ 20–60) show that
horizontal cavities van be both stable and oscillating, the latter can usually be stopped by an
increase of gas injection. Let’s consider a cavity model with of gas loss taken into account,
the loss amount being a monotonically increasing function of gas pressure, i.e. a
monotonically decreasing function of cavitation number σ : m y = m y ( pk ), ∂m y / ∂ pk > 0 .
In this case, the equation of small cavity oscillations is:
2 2
pk′′′( t ) + α pk′′ ( t ) + pk′ ( t ) + pk′ ( t − τ 0 ) − pk ( t ) + pk ( t − τ 0 ) = 0. (8.5.1)
τ0 τ0
dm y dpk ρ pk 0
The equation features an additional damping term α pk′′ , with α = . (8.5.2)
Cpkn0 nkl0Q0
Considering the eigenvalues of the
characteristic equation, the general stability
boundary α кр (τ 0 ) for the cavity is plotted
as a solid line in Fig.8.5.1.
Fig.8.5.1.
Stability boundaries at dimensionless time:
solid lines stand for critical value αcr;
dashed lines stand for stable value αj.
299
The stability boundaries for different eigenvalues α j (τ 0 ) are plotted by dashed

lines. The cavity is stable at α > α cr . The appearance of α pk′′ does not significantly affect
the frequency values. In the case of gas withdrawal from horizontal cavities along trailing
Qy 0.42cx20
vortices, the Epshteyn’s formula can be used: Qy = 2 = , where Qy
Vd σ (σ 3 Fr 4 − 2.5cx 0 )
is volume per second rate of gas withdrawal from the cavity. The rate of gas loss is
m y = CpknVd 2Qy . By differentiating this expression with respect to pk and substituting the
derivative into (8.5.2), one can obtain coefficient α as
6 Qyσ ⎛ τ 02 Q yσ ⎞
α= ⎜⎜1 + + 1.5τ 02 ⎟. (8.5.3)
π (1 + σ )τ 0 cx 0 ⎝ 3 cx 0 ⎟⎠
There is a complicated Qyσ cx 0 considered as a stability criterion for weight-
affected cavities. Damping coefficient α grows as Qyσ cx 0 increases. Substituting
α = α cr (τ 0 ) into equivalence (8.5.3) yields the value of Qyσ cx 0 corresponding to stability
boundary:
2
⎛ Q yσ ⎞ 3 + τ 02 ⎛ 3 + τ 02 ⎞ π (1 + σ ) α cr (τ 0 )
⎜⎜ ⎟⎟ = − + ⎜ 2 ⎟
+ .
⎝ cx 0 ⎠ кр 9τ 02 ⎝ 9τ 0 ⎠ 9τ 0
Fig.8.5.2.
Theoretical stability threshold and gray area; the
Epshteyn’s & Lapin’s [1985] experimental points
(marked by × for stable cavities, and by fore
pulsating cavities).
A part of the theoretical stability boundary

in coordinates τ 0 , Qyσ / cx 0 is plotted on Fig.8.5.2 by
a solid line. Experimental and analytical results are in close agreement.
8.5.3. Jet-Induced Ejection of Gas from a Cavity
According to well-known ejection properties of turbulent jets, ejection rate is proportional to
the length of the part of jet located inside the cavity. Survey this model shows that the jet has
a little effect on cavity stability. The trail jet can damp cavity oscillations within the
κ R0V0 ρ 0
following range of parameters 0.8–0.9 < < 1.1–1.2. Here κ = 0.155 is the jet
cx 0 RnV ρ k
ejection coefficient; R0 ,V0 , ρ 0 are radius, velocity and density of the jet; cx 0 , Rn ,V are
cavitator’s drag coefficient, radius and velocity; ρ k is the gas density inside cavity.
300
8.5.4. Forced Oscillations of a Cavity

It is interesting to study the reaction of cavity as a dynamic system to an external
disturbance. Considered here is a method to excite cavity oscillations by means of periodic
variation of the volume q connected with the internal cavity volume Q . Wave processes in
the gas are ignored, and it is assumed that the total mass in volume Q + q is constant. After
linearization, equations (8.3.5) & (8.3.6) yield the equation for forced cavity oscillations:
2 2
pk''' ( t ) + pk' ( t ) + pk' ( t − τ 0 ) − pk ( t ) + pk ( t − τ 0 ) = − q ''' . (8.5.4)
τ0 τ0
iω t
Let’s denote disturbance as q = Ae and pressure oscillations as pk = Beiω t .
Equation (8.5.4) gives amplitude and phase frequency response functions b = B / A ,
arg B = π + ϕ , χ = ωτ 0 / 2 . The frequency response functions of a stable cavity at τ 0 = 3 are
shown in Fig.8.5.3. There is a resonance ( b > 1 ) at χ ≈ π . While χ → ∞ , ϕ → 0 , b → 1 ,
the cavity volume does not keep pace with pressure disturbance at high frequencies, i.e. the
cavity becomes “rigid”.
For unstable cavities, theoretically, there can be a method of external influence
which would dampen self-oscillations of the cavity. For the volume control as q = α ∫ pk dt ,
equasion (8.5.4) will be reduced to equation (8.5.1) considered above. A cavity is stable if
α > α кр , Fig.8.5.1. For an unstable cavity with τ 0 = 5 , the value α кр = 0.282 . The frequency
response functions of a stable cavity in terms of amplitude at different values of α > α кр are
plotted in Fig.8.5.4. The amplitude at resonance decreases as α increases. As can be seen
from Fig.8.5.1, there is an estimate: α кр max ≈ 1.2 .
Fig.8.5.3. The frequency response Fig.8.5.4. The frequency response

functions of unstable cavity. functions of stable cavity ( α > α cr )
8.5.5. Nonlinear Cavity Oscillations with Periodic Detachment of Gas Bubbles

Along with linear analysis, numerical integration of the full nonlinear system of time-delay
equations (8.3.5) & (8.3.6) has been performed with periodic detachment of gas bubbles
taken into account. Let’s use the following conditions: V=const, p∞=const, with gas
301
injection rate mn (t ) and periodic bubbles detachment (ignoring other components of gas
loss). Theoretically, waves on a cavity surface cause instant pinches of cavity profile and
detachment of some volume (this do occur in reality though not so explicitly, e.g. [Michel,
1973]).
The qualitative peculiarities of non-linear oscillations are in stepwise changes at the
detachment instant of not only cavity volume and length, but derivatives Q and pk as well.
Thus, this instant requires re-defining of the initial conditions. At the instant of bubble
detachment, the cavity length features an infinite (though integrable) variation rate l → −∞ .
By selecting specific scales (e.g. different in lengthwise and radial directions) the
nonlinear system of equations (8.3.5)–(8.3.6) transforms into a dimensionless form in which
all the coefficients are equal to 1, and the only governing parameter, the dimensionless
2 Eu Qn
injection rate, is qn = , where d is the cavitator diameter.
π cx 0 Vd 2
Comprehensive numerical analysis has been performed, covering a cavity pulsation
range from single- to ten-wave mode under different types of injection. The analysis results
are in good qualitative agreement with experimental data and the above linear theory. An
example of multi-wave (N≈6) analysis results is presented in Fig.8.5.5, qn = 10 . The
oscillations are not strictly regular and feature variations in the pressure pulse and number of
waves (also observed in experiments) due to the excitation of unstable lower eigenvalues. A
phase trajectory ( Δp ' , Δp ) is of the converging (stable) spiral type, with instant energy
supply at the
bubble
detachment
instant as in a
pendulum clock.
Fig.8.5.5.
The main
parameters for
multi-wave
(N≈6) cavity.
The
most important
result is that the accepted theoretical model reflects deep inherent properties of real cavities.
Thus the analysis shows that an increase in injection rate leads to switches between pulsation
modes towards a greater number of waves along the cavity; this is also observed in
experiments. This is especially pronounced for the following analysis conditions: under
continuous or very slow change in the injection rate and in the absence of external
disturbances. There is a critical injection rate of qn = 3.26 at which single-wave oscillations
R1 spontaneously turn into two-wave ones R2 (see. Fig.8.5.6). Here N is the number of
waves along the cavity, and RN denotes the corresponding mode.
302
Fig.8.5.6.
Dependence of
number of waves on
injection rate.
Multi-wave
transitions due to
changes of injection
rate are not so
explicit. The multi-
wave modes
themselves are
unstable. Both their
parameters (cavity
length, pulsation
frequency, bubble volume), and number of waves N are unstable. Along with the main
mode, there are cavities with a lower or higher number of waves. Fig.8.5.6 shows
distribution of pulsation modes RN as a function of dimensionless injection rate qn ≤ 13.75 .
Cavities with different numbers of waves can occur at the same injection rate kept for a
rather long time (~20–30 periods of pulsation). The occurrence rate of a cavity mode in
percentage points is displayed by length of horizontal lines (100% relates to Δ qn = 0.5 ). One
can see that, generally, mode number N increases with an increase of injection rate.
Approximate boundaries between modes can be indicated: transition from R2 to R3 occurs at
qn ≈ 5.5 ; from R3 to R4 at qn ≈ 8.5 ; from R4 to R5 at qn ≈ 13.3 .
Sec. 8.6. Approximating Cavity Initiation as a Rayleigh Bubble

For very short-term operation of a vehicle involving a ventilated cavity, it is practical to
provide the gas from a pressurized air canister. This section presents the combined thermo-
hydrodynamic formulation for a first-principles (albeit approximate) model of this process,
along with simulation results, as well as the effects of varying several key system parameters
on the time required to fully envelop the vehicle. For a small part of the space of cases
studied, it is shown that the dynamics of the expanding cavity can have a significant effect
on the time required to achieve supercavitation. The contents of this section were derived
from the work by [Kirschner & Rosenthal, 2006].
In general, initiation of the supercavity with such a ventilation system involves the
coupled dynamics of gas discharge from the pressurized canister through a choke into the
expanding cavity. The question arises as to the effect of added mass of the cavity in the
expansion mode of motion on the time required to fully envelop the body, a subject that may
be addressed via time-domain simulations of cavity initiation using a pressurized air canister.
An upper bound on the added mass of the cavity in the expansion mode is approximated by
applying the Rayleigh equation, which is the governing equation of spherical bubble
dynamics ignoring the effects of viscosity and surface tension. This approximation for the
303
cavity dynamics is coupled to the release of gas from a pressurized canister via assumed
isentropic discharge through a converging nozzle, as is given schematically in Fig.8.6.1b.
For the cases of interest, the flow is at first supersonic at the choke, and added mass
has no effect on the gas discharge rate until the critical time; that is, until the pressure in the
expanding cavity reaches the nozzle critical pressure, at which point the canister “feels” the
presence of the expanding bubble. The one-dimensional results for the expansion of a
spherical bubble are then applied to approximate the expansion of an axisymmetric cavity,
the shape of which at any instant is defined by the Münzer-Reichardt profile. It is shown
that, for the cases of interest, the effects of added mass can be neglected until the critical
time, and are very small thereafter for much of the operational space studied. Results of key
quantities are presented as functions of time for several different flow cases, and
parametrically as functions of design parameters.
State in Mass flow State in Ambient
canister through bubble conditions
nozzle
Inflow
PT ( t ) mE ( t ) PB ( t ) P∞ ( t )
a
ρT ( t ) ρB ( t ) T∞
TT ( t ) TB = T∞
Fig.8.6.1. A supercavitating vehicle and its
ventilation system:
(a) the vehicle showing the cavitator and the
body, partially enveloped in a cavity;
(b) a simple ventilation system consisting of
a pressurized tank or canister coupled to a
cavity (idealized as a spherical bubble)
through a converging nozzle. b Bubble RB ( t ) , RB ( t ) , RB ( t )
growth
In the formulation that follows, the thermodynamics of blow-down (that is, discharge
of pressurized gas from a canister through a nozzle) is developed based on the choked-flow
theory presented in [Otta, 2003], and extended to non-choked flow based on information
presented in [Van Wylen & Sonntag, 1973]. The model for the dynamics of a spherical
bubble is based on Rayleigh’s theory as presented in [Brennen, 1995]. In order to
approximate the geometry of a supercavity once the coupled problem has been solved, the
spherical bubble volume is apportioned to the dimensions of an axisymmetric supercavity
using the cavity profile estimated from first principles by [Münzer & Reichardt, 1950] as
presented in [May, 1975]. The length and diameter of the supercavity are selected using the
formulae of [Garabedian, 1956], also presented in [May, 1975], for the cavitator geometry
and operational conditions of interest, here selected based on a notional model-scale system.
8.6.1. Nozzle Analysis.
Consider a converging nozzle, where the throat is coincident with the nozzle exit plane,
denoted by subscript E. Let the exit area be AE. The mass flow through throat is given by
mE = ρ EVE AE , (8.6.1)
where ρE is the density and VE. is the velocity of the gas, both evaluated at the exit plane.
The exit density is related to the exit pressure, pE, and temperature, TE, by the ideal gas law:
304
ρ E = pE / RTE , (8.6.2)
where R is the gas constant for the inflation gas being used, with a value of R=287.0 J/kg-K
for air. The velocity at the throat can be written in terms of the exit Mach
number and sonic velocity: VE = M E cE . (8.6.3)
AE pE M E cE
Thus, the mass flow is given by mE = (8.6.4)
R TE
Considering the flow through the nozzle to be an isentropic process, the critical
pressure at the nozzle throat, p*, is that at which the exit Mach number is unity: M E p = p* = 1
E
An important result of thermodynamics is the fact that, for a converging nozzle, the
exit Mach number cannot exceed unity, and the exit pressure cannot fall below the critical
pressure. Thus, even if the back pressure on a converging nozzle drops below the critical
pressure, the exit Mach number is unity: M E p ≤ p* = 1 , (8.6.5)
E
the exit pressure does not fall below the critical pressure, and the flow is choked. The ratio
of the critical pressure to the total stagnation pressure is given by the following equation:
γ (γ −1)
p* ⎛ 2 ⎞
=⎜ ⎟ = Γ +γ Γ− 2
= Γ +γ − (8.6.6)
p0 ⎝ γ + 1 ⎠
where γ is the polytropic constant of the gas for air, γ=1.4, and the following abbreviations
Δ 2 Δ γ Γ
have been defined: Γ± = and γ ± = ±
(8.6.7)
γ ±1 2
The corresponding critical temperature and density ratios of the gas at the nozzle exit
1 ( γ −1)
T* 2 ρ* ⎛ 2 ⎞
are, respectively: = = Γ + and = = Γ + Γ− 2 . (8.6.8)
T0 γ + 1 ρ0 ⎜⎝ γ + 1 ⎟⎠
If the back pressure on the converging nozzle is greater than or equal to the critical
pressure, the exit Mach will be less than unity, and the exit pressure is related to the total
stagnation pressure and the exit Mach number by the following expression:
γ (γ −1)
p0 ⎡ γ − 1 2 ⎤ γ−
= ⎢1 + ME ⎥ = ⎡⎣1 + M E2 Γ − ⎤⎦ . (8.6.9)
pE ⎣ 2 ⎦
The corresponding exit temperature and density are given by
(γ −1) γ 1 γ− 1γ
T0 ⎛ p0 ⎞ ⎛p ⎞ ρ0 ⎛ p0 ⎞
=⎜ ⎟ =⎜ 0 ⎟ and =⎜ ⎟ . (8.6.10)
TE ⎝ pE ⎠ ⎝ pE ⎠ ρ E ⎝ pE ⎠
Equation (8.6.9) can be solved for the exit Mach number:
− ( γ −1) γ
2 ⎡⎛ pE ⎞ ⎤ γ−
⎡⎛ p ⎞−1 ⎤
ME = ⎢ ⎜ ⎟ ⎥
− 1 = Γ − ⎢⎜ E ⎟ − 1⎥ . (8.6.11)
pB ≥ p*
γ − 1 ⎢⎝ p0 ⎠ ⎥ ⎢⎣⎝ p0 ⎠ ⎥⎦
⎣ ⎦
The sonic velocity is given by cE = γ g c RTE , (8.6.12)
where g c is a constant relating the units of force, mass, length, and time, with a value of
unity for the SI system. Thus, the mass flow through the nozzle is given by the expressions
presented in Table 8.6.1, where the new constants appearing in the expressions are defined
Δ
as follows: ℜ = gcγ R (8.6.13)
305
(γ +1)
Δ⎛ 2 ⎞ 2(γ −1)
and Γ =⎜ ⎟ = Γ + Γ− 2Γ+
. (8.6.13)
⎝ γ +1⎠
Table 8.6.1. Mass flow through a converging nozzle.

Back Pressure Mass Flow
pE ≤ p * mE = ℜΓAE ρ0 T0 (8.6.14)
⎡⎛ ρ ⎞ −2 Γ− ⎤
pE ≥ p * mE = ℜ Γ − AE ρ E ⎢⎜ E ⎟ − 1⎥ TE (8.6.15)
⎢⎣⎝ ρ0 ⎠ ⎥⎦
8.6.2. Coupled Tank-Nozzle-Bubble System

For quasi-static blow-down from a canister into a bubble through the isentropic nozzle, the
time-dependent total stagnation pressure is the pressure in the canister, p0 = pT ( t ) , which is
assumed to be spatially uniform. The corresponding stagnation temperature and density are
T0 = TT ( t ) and ρ 0 = ρ T ( t ) , respectively, and relate to one another and to the stagnation
pressure via the ideal gas law: p0 = R ρ0T0 = R ρT ( t ) TT ( t ) = pT ( t ) . (8.6.17)
If the pressure within the bubble, also assumed to be spatially uniform, is greater
than or equal to the critical pressure, the exit pressure is the back pressure, which is the time-
dependent pressure in the bubble, pE ( t ) = pB ( t ) ; otherwise, pE ( t ) = p * ( t ) , which is time-
dependent by virtue of the fact that the canister pressure and therefore the total stagnation
pressure are falling with time. Neglecting the effects of any vapor in the bubble, the back
pressure in the bubble is related to the temperature, TB ( t ) , and density, ρ B ( t ) , of the gas in
the bubble via the ideal gas law: pB ( t ) = R ρ B ( t ) TB ( t ) . (8.6.18)
8.6.3. Isentropic Canister Blow-Down

The classical control-volume analysis of blow-down of a constant-volume canister results in
the following ordinary differential equation for the density of the air in the canister:
VT ρT ( t ) + mE ( t ) = 0, (8.6.19)
where VT is the canister volume; ρT is the air density in the canister; and mE is the mass
flow rate through the discharge nozzle, given by equation (8.6.15) or (8.6.16), depending on
the back pressure. The initial condition is a specified density at some time t = 0 :
ρT ( t = 0 ) = ρT0 . (8.6.20)
If the initial temperature in the canister is also given: TT ( t = 0 ) = TT0 , (8.6.21)
the initial pressure in the canister is given as pT ( t = 0 ) = pT0 = R ρT0TT0 , (8.6.22)
For rapid discharge of air via the nozzle into the bubble, heat transfer through the
canister walls will be negligible, and the blow-down process may be considered isentropic:
γ −1 2Γ
⎛ ρ (t ) ⎞ ⎛ ρT ( t ) ⎞ −
TT ( t ) = TT0 ⎜ T ⎟ = TT0 ⎜ ⎟ . (8.6.23)
⎝ ρT0 ⎠ ⎝ ρT0 ⎠
For choked flow, where the bubble pressure is less than or equal to the critical pressure,
equation (8.6.23) allows elimination of the stagnation temperature, T0 = TT (t ) , in equation
306
(8.6.15). Making this substitution and applying the initial conditions, equations (8.6.20) and
(8.6.21), results in the following solution for isentropic blow-down through a choked nozzle:
2
−Γ −
⎡⎛ ( γ − 1) ⎞ ⎤ 1−γ ⎡ ℜΓτ 0 ⎤
ρT ( t ) = ρT0 ⎢⎜ ℜΓτ 0 ⎟ t + 1⎥ = ρ T0 ⎢ t + 1⎥ , (8.6.24)
⎢⎣⎝ 2 ⎠ ⎥⎦ ⎣ Γ− ⎦
Δ A
where τ 0 = E TT0 .
VT
The corresponding mass flow through the nozzle may now be evaluated by
substituting this solution for density into equation (8.6.19):
Γ− Γ+
τ0 ⎡ ℜΓτ 0 ⎤
mE = − VT ρT ( t ) = ℜΓVT ρ 1 Γ+
T ( t ) = ℜΓVTτ 0 ρT0 ⎢ t + 1⎥ . (8.6.25)
ρT01 Γ− ⎣ Γ− ⎦
The absolute pressure in the canister is expressed as
2γ
−2 γ −
⎡⎛ ( γ − 1) ⎞ ⎤ 1−γ ⎡ ℜΓτ 0 ⎤
pT ( t ) = pT0 ⎢⎜ ℜΓτ 0 ⎟ t + 1⎥ = pT0 ⎢ t + 1⎥ . (8.6.26)
⎢⎣⎝ 2 ⎠ ⎥⎦ ⎣ Γ− ⎦
8.6.4. Bubble Dynamics

Neglecting the compressibility of the liquid outside the bubble, as well as surface tension
and the liquid viscosity, the expansion of a spherical bubble is governed by the Rayleigh
pB ( t ) − p∞ ( t ) 3 2
equation: = RB RB + RB , (8.6.27)
ρL 2
where ρ L is the liquid density; RB = RB ( t ) is the bubble radius; and the pressure at an
infinite horizontal distance from the bubble center, p∞ ( t ) , is taken as time-dependent to
account for far-field disturbances or quasi-static vertical motion of the frame of reference.
The initial conditions for this system are the initial bubble radius, and time-rate-of-change of
the bubble radius at some time t = 0 : RB ( t = 0 ) = RB0 , (8.6.28)
and RB ( t = 0 ) = RB0 . (8.6.29)
8.6.5. The Coupled System

The canister and bubble are coupled through the nozzle. Note that, for choked flow, with the
bubble pressure less than the nozzle critical pressure, the canister blow-down is independent
of the bubble pressure, but that the situation is much more complicated otherwise.
The Rayleigh equation, (8.6.27), can be rewritten in terms of the bubble density and
temperature via the ideal gas law, equation (8.6.18):
R p∞ ( t ) 3 2
ρ B (t )TB (t ) − = RB RB +(8.6.30)
RB .
ρL ρL 2
The average gas density in the bubble is the total mass in the bubble divided by its volume:
mB mB
ρB = = . (8.6.31)
VB 4 3π RB3
Assuming that the temperature of the gas in the bubble is constant, TB ( t ) = TB ,
simplifies the problem considerably. Such an assumption is physically realistic if mixing
inside the bubble is rapid and the liquid outside the bubble is an efficient heat sink. Making
307
this substitution in equation (8.6.30) and taking the time derivative of the result produces the
following equation for the bubble dynamics in terms of the time rate of change of the mass
⎛p 3 ⎞ p
in the bubble: RB4 RB + 7 RB3 RB RB + 3RB2 RB ⎜ ∞ + RB2 ⎟ + ∞ RB3 − λ mB = 0, (8.6.32)
⎝ ρL 2 ⎠ ρL
Δ
where the constant in the last term on the left-hand side is λ = 3RTB 4πρL .
Neglecting any transport of gas across the bubble surface, the time rate of change of
the mass in the bubble is simply the mass flow into the bubble via the nozzle: mB = mE .
Thus, the governing equation relating the bubble radius, the nozzle exit mass flow, and the
⎛p 3 ⎞ p
far-field conditions is RB4 RB + 7 RB3 RB RB + 3RB2 RB ⎜ ∞ + RB2 ⎟ + ∞ RB3 − λ mE = 0, (8.6.33)
⎝ ρL 2 ⎠ ρL
In the special case where the far field pressure is constant, the fourth term on the left-hand
side is zero.
In the case of choked flow through the nozzle, the nozzle exit mass flow does not
depend on the response of the bubble, and the last term on the left-hand side becomes a
⎛p 3 ⎞ p
forcing function: RB4 RB + 7 RB3 RB RB + 3RB2 RB ⎜ ∞ + RB2 ⎟ + ∞ RB3 = λ mE . (8.6.34)
⎝ ρL 2 ⎠ ρL
This 3rd-order nonlinear ordinary differential equation for the bubble radius is subject to
initial conditions on the initial bubble diameter and its 1st and 2nd rates of change:
RB ( 0 ) = RB0 , RB ( 0 ) = RB0 , RB ( 0 ) = RB0 . (8.6.35)
If, on the other hand, the flow is not choked, the nozzle exit mass flow is a function
of the pressure in the bubble.
8.6.6. The Critical Bubble Pressure
If a solution to equation (8.6.39) is determined, the pressure in the bubble can be determined
⎛ 3 ⎞
from the Rayleigh equation: pB ( t ) = ρ L ⎜ RB RB + RB2 ⎟ + p∞ ( t ) , (8.6.36)
⎝ 2 ⎠
For the case of monotonic increase in the bubble pressure during canister blow-
down, the flow will be choked until the pressure in the bubble reaches the nozzle critical
pressure as given by equation (8.6.35), with the nozzle stagnation pressure given by the
quasi-static canister stagnation pressure: pB = p*, or
2γ −
pB p 1 ⎡ ⎛ 3 2⎞ ⎤ ⎡ ℜΓτ 0 ⎤
= γB = γ
p * Γ + pT Γ + pT0 ⎢ ρ L ⎜ RB RB + 2 RB ⎟ + p∞ ( t ) ⎥ ⎢ Γ t + 1⎥ = 1. (8.6.37)
− −
⎣ ⎝ ⎠ ⎦⎣ − ⎦
8.6.7. The Governing Equation for Choked Flow

Substituting equation (8.6.25) into equation (8.6.36) yields the following governing equation
for the bubble radius:
Γ − Γ+
⎛ p∞ 3 2 ⎞ p∞ 3 ⎡ ℜΓτ 0 ⎤
RB4 RB + 7 RB3 RB RB + 3RB2 RB ⎜ + RB ⎟ + RB = ℜΓVT λτ 0 ρ T0 ⎢ t + 1⎥ , (8.6.38)
⎝ ρL 2 ⎠ ρL ⎣ Γ− ⎦
subject to the initial conditions on the bubble, equations (8.6.35). This equation is valid until
the canister pressure decreases and the back pressure on the nozzle increases to the
308
corresponding critical pressure, at which time the bubble dynamics begin to have an effect
on the canister blow-down.
8.6.8. Non-Dimensional Form for Choked Flow
A dimensionless form of equation (8.6.38) can be developed based on appropriate
characteristic quantities. It is convenient to select these based on the initial conditions in the
canister: p = pT0 , ρ = ρT0 , T = TT0 , (8.6.39)
where the over-bar indicates the characteristic quantity used to make the system
dimensionless. The natural length scale associated with the canister-nozzle system is the
VT
canister volume divided by the nozzle exit area: = . (8.6.40)
AE
A characteristic time scale can be formed by combining the length, pressure, and density
ρ VT ρT0
characteristic quantities: t = = . (8.6.41)
p AE pT0
Similarly, characteristic scales of force and mass can be formed by appropriate combination
of the other characteristic quantities, respectively:
2 3
⎛V ⎞ ⎛ VT ⎞
f = 2
p = ⎜ T ⎟ pT0 . and m= 3
ρ =⎜ ⎟ ρT0 . (8.6.42)
⎝ AE ⎠ ⎝ AE ⎠
Using a tilde to denote dimensionless quantities, the dimensional quantities
appearing in equation (8.6.39) can be expressed, for example, as
ρ L = ρT0 ρ L , p∞ = pT0 p∞ , t = ( VT / AE ) ρT0 / pT0 (8.6.43)
3
pT0 pT0 ⎛V ⎞
etc. Note, especially: λ= λ, ℜ = ℜ, VT = ⎜ T ⎟ VT , gC = gC (8.6.44)
ρT0
2
ρT0TT0 ⎝ AE ⎠
n
d n ⎛ AE pT0 ⎞ d n
Also, time derivatives of order n are expressed as =⎜ ⎟ .
dt n ⎜⎝ VT ρT0 ⎟⎠ dt n
Finally, it should be noted, that with these selections of characteristic quantities, several
dimensionless quantities take on specific values:
ρ T ρ T pT0 ρ2 3 TB
R = T0 T0 R = T0 T0 = 1, λ = T0 λ = ,
pT0 pT0 ρ T0TT0 pT0 4π ρ L
(8.6.45)
VT VT AE TT0 VT
τ0 = τ0 = = 1, = 1.
AE TT0 AE TT0 VT AE
Making these substitutions into equation (8.6.38), dividing through by common
factors, and dropping the tildes, the dimensionless governing equation becomes
Γ− Γ+
⎛p 3 ⎞ p ⎡Γ ⎤
RB4 RB + 7 RB3 RB RB + 3RB2 RB ⎜ ∞ + RB2 ⎟ + ∞ RB3 = ℜΓVT λ ⎢ ℜ t + 1⎥ . (8.6.46)
⎝ ρL 2 ⎠ ρL ⎣ Γ− ⎦
A similar procedure is used to make the initial conditions, equations (8.6.39), dimensionless.
Equation (8.6.46) is amenable to numerical solution. The 3rd-order ordinary
differential equation can be rewritten as a system of three 1st-order equations. Denote
Δ d n −1
yn ( t ) = RB (t ), n = 1, 2,3. (8.6.47)
dt n −1
309
Then the system of three 1st-order ordinary differential equations equivalent to equation
⎡ y1 ⎤ ⎡ y2 ⎤
(8.6.46) is ⎢y ⎥ = ⎢y ⎥, (8.6.48)
⎢ 2⎥ ⎢ 3⎥
⎢⎣ y3 ⎥⎦ ⎢⎣ f ⎥⎦
{ }
Δ 1 ⎛ 3 ⎞
where f (t ) = φ (t ) − ⎣⎡7 y13 (t ) y2 (t ) y3 (t ) + 3 y12 (t ) y2 (t ) ⎜ P(t ) + y22 (t ) ⎟ + P(t ) y13 (t ) ⎦⎤ ,
y14 (t ) ⎝ 2 ⎠
Δ p∞ (t ) Δ p∞ (t ) Δ
φ (t ) = C [ A t + 1] .
B
P (t ) = , P(t ) = ,
ρL ρL
The constants in the last of the equations are defined as
Δ Γ Δ Γ Δ
A= , B = − , and C = ℜΓVT λ . (8.6.49
Γ− Γ+
The initial conditions are specified as a vector comprised of the initial bubble radius and its
⎡ y1 ( 0 ) ⎤ ⎡ y10 ⎤ ⎡ RB0 ⎤
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
first two time rates of change: ⎢ y2 ( 0 ) ⎥ = ⎢ y20 ⎥ = ⎢ RB0 ⎥ . (8.6.50)
⎢⎣ y3 ( 0 ) ⎥⎦ ⎢⎣ y30 ⎥⎦ ⎢⎣ RB0 ⎥⎦
8.6.9. System Behavior for Non-Choked Flow

In the case of non-choked flow, it is necessary to solve the fully-coupled problem for which
the bubble pressure affects the discharge rate from the canister through the nozzle.
Denote the critical time (that is, the time when the bubble pressure first reaches the
nozzle critical pressure) as t* . At any time after this critical time, the mass in the bubble,
t
denoted as u ( t ) , is given by u ( t ) = mE ( t* ) + ∫ mE (τ )dτ , (8.6.51)
t*
where t* is the time when the bubble pressure first reaches the nozzle critical pressure.
Differentiating this expression with respect to time yields u ( t ) = mE ( t ) , (8.6.52)
which is the equation of mass continuity for the control volume representing the bubble.
Noting that the density of the bubble can be written as
ρ B ( t ) = κ u ( t ) / RB3 ( t ) , (8.6.53)
where κ 4π 3 , equation (8.6.16) can be rewritten as
3 −γ 2
⎡ u (t ) ⎤ ⎡ u (t ) ⎤
mE ( t ) = ℜ Γ − AE TB ρT γ −1
( t ) ⎢κ 3 ⎥ − ⎢κ 3 ⎥ . (8.6.54)
⎢⎣ RB ( t ) ⎥⎦ ⎢⎣ RB ( t ) ⎥⎦
The equivalent dimensionless expression is
3−γ 2
⎡ u (t ) ⎤ ⎡ u (t ) ⎤
mE ( t ) = μ AE ρT γ −1
( t ) ⎢κ 3 ⎥ − ⎢κ 3 ⎥ . (8.6.55)
⎣⎢ RB ( t ) ⎥⎦ ⎣⎢ RB ( t ) ⎥⎦
where μ ℜ Γ − TB .
The dimensionless form of equation (8.6.19), a statement of mass continuity for the
control volume representing the canister, is VT ρT ( t ) + mE ( t ) = 0, (8.6.56)
310
Combining equations (8.6.56) and (8.6.55), squaring the result, and dropping the
tildes produces the following expression relating the density of gas in the canister, its time
rate of change, and the radius of and the mass of the gas within the bubble:
3− γ 2
⎡ u (t ) ⎤ ⎡ u (t ) ⎤
ρ − μ ρT
2 2 γ −1
( t ) ⎢κ 3 ⎥ + μ ⎢κ 3 ⎥ = 0.
2
(8.6.57)
⎣⎢ RB ( t ) ⎥⎦ ⎣⎢ RB ( t ) ⎥⎦
T
Differentiating with respect to time results in the following dimensionless second-order

ordinary differential equation for the density of gas in the canister in terms of the remaining
ρ γ −1 ( t ) 1
quantities: ρT ( t ) = f1 ( t ) ρTγ − 2 ( t ) + f 2 ( t ) T + f3 ( t ) . (8.6.58)
ρT ( t ) ρT ( t )
3−γ 2 −γ
⎧ (γ − 1) μ 2 ⎡ u (t ) ⎤ (3 − γ ) μ 2 ⎡ RB ( t ) u ( t ) − 3u ( t ) RB ( t ) ⎤ ⎡ u ( t ) ⎤
⎪ f1 ( t ) ⎢κ 3 ⎥ ; f2 (t ) ⎢κ ⎥ ⎢κ 3 ⎥ ;
where ⎪⎪ 2 ⎢⎣ B ( t ) ⎦⎥
R 2 ⎣⎢ RB4 ( t ) ⎦⎥ ⎣⎢ RB ( t ) ⎥⎦
⎨
⎪ ⎡ R ( t ) u ( t ) − 3u ( t ) RB ( t ) ⎤ u ( t )
⎪ f3 ( t ) − μ 2 ⎢κ B ⎥κ 3 .
⎩⎪ ⎣⎢ RB4 ( t ) ⎦⎥ RB ( t )
Similarly, a second-order equation for the mass of gas in the bubble can be found by
combining equations (8.6.52) and (8.6.54), squaring the result, and differentiating with
respect to time. In dimensionless form, the result is
ρ γ −1 ρ γ −2 ρ 1
u ( t ) = g1 ( t ) T + g 2 ( t ) T T + g3 ( t ) , (8.6.59)
u (t ) u (t ) u (t )
where g1 ( t ) Θ2
(3 − γ ) (1 − γ )
ρ B2 −γ ρ B ; g 2 ( t ) −Θ2 ρ B3−γ ; g3 ( t ) −Θ 2 ρ B ρ B ; Θ μ AE .
2 2
Thus, for the case of non-choked flow, seven state variables must be simulated: the
bubble radius, y1 = RB ( t ) , and its first and second time rates of change, y2 = RB ( t ) and
y3 = RB ( t ) , respectively; the mass of gas in the bubble, y4 = u ( t ) , and its time rate of
change, y5 = u ( t ) , (which is equal to the mass flow rate through the nozzle); and the density
of the gas in the canister, y6 = ρT ( t ) , and its time rate of change, y7 = ρT ( t ) . The governing
system of dimensionless equations can be stated as follows:
1 ⎛ p∞ ⎞ y ⎛p 3 ⎞ y
y1 = y2 , y2 = y3 , y3 = ⎜ + 7 y2 y3 ⎟ + 3 22 ⎜ ∞ + y22 ⎟ − λ 54 ,
y1 ⎝ ρ L ⎠ ρ
y1 ⎝ L 2 ⎠ y1
1
y4 = y5 , y5 = ⎡ g1 ( t ) y6γ −1 + g 2 ( t ) y6γ − 2 y7 + g3 ( t ) ⎤ , y6 = y7 , (8.6.60)
y5 ⎣ ⎦
y6γ −1 1
y7 = f1 ( t ) y6γ − 2 + f 2 ( t ) + f3 ( t ) ,
y7 y7
The time-dependent functions on the right-hand side can be expressed in terms of the state
3− γ 2 −γ
(γ − 1) μ 2 ⎡ y4 ⎤ ( 3 − γ ) μ 2 ⎡ y1 y5 − 3 y4 y2 ⎤ ⎡ y4 ⎤
f1 ( t ) = ⎢κ 3 ⎥ , f 2 ( t ) = ⎢κ ⎥ ⎢κ 3 ⎥ ,
2 ⎣ y1 ⎦ 2 ⎣ y14 ⎦ ⎣ y1 ⎦
variables as: (8.6.61)
⎡ y y − 3y y ⎤ y
f3 ( t ) = − μ 2 ⎢κ 1 5 4 4 2 ⎥ κ 43 ,
⎣ y1 ⎦ y1
311
2 −γ 3−γ
(3 − γ ) ⎛ y4 ⎞
⎡ y1 y5 − 3 y4 y2 ⎤ 2 (1 − γ ) ⎛ y4 ⎞
g1 ( t ) = Θ2 ⎜⎜ κ ⎢κ ⎟ ⎥ , g 2 ( t ) = −Θ ⎜⎜ κ 3 ⎟⎟ ,
2 ⎝ y13 ⎟⎠
⎣
4
y1 ⎦ 2 ⎝ y1 ⎠
and (8.6.62)
⎛ y ⎞ ⎡ y y − 3y y ⎤
g3 ( t ) = −Θ2 ⎜⎜ κ 43 ⎟⎟ ⎢κ 1 5 4 4 2 ⎥ .
⎝ y1 ⎠ ⎣ y1 ⎦
Now using a numerical ordinary differential equation solver and specifying as initial
conditions the results of the choked flow case at time t* , the subsequent time history of the
bubble can be found. Note that it is conceivable that conditions may exist for which the
system is in an oscillatory mode at the critical time, such that the canister-nozzle-bubble
system endures multiple cycles of transition between the choked and non-choked conditions.
In the current implementation of this formulation, such an event has been ignored for
simplicity.
8.6.10. Approximating the Cavity Geometry
Once the governing system of equations has been solved, the instantaneous volume of the
spherical bubble is used to estimate the cavity geometry, as follows.
If the instantaneous cavitation number, σ ( t ) , were known, Garabedian’s formulae
(which were developed assuming steady flow conditions) could be applied to estimate the
effective quasi-steady length, , and maximum diameter, dmax, of the cavity:
⎛ d max ( t ) ⎞ CD ( t )
2
⎜ ⎟ ≅ , (8.6.63)
⎝ d ⎠ σ (t )
⎛ (t ) ⎞ CD ( t ) ⎛ 1 ⎞
2
⎜ ⎟ ≅ 2 ln ⎜ ⎟. (8.6.64)
⎝ d ⎠ σ ( t ) ⎜⎝ σ ( t ) ⎟⎠
Here, d is the axisymmetric cavitator diameter and CD ( t ) is the instantaneous cavitator
D (t )
drag coefficient: CD ( t ) , (8.6.65)
1 2 ρV 2 ( t ) A
where A is the cavitator projected area. The value of the drag coefficient depends on both
the geometry of the cavitator and the cavitation number. For a disk (used as an example in
the simulations below), its value may be computed as CD ( t ) ≅ CD 0 (1 + σ ( t ) ) , (8.6.66)
where CD 0 ≅ 0.815 is the limiting value of the cavitator drag coefficient at a cavitation
number of zero.
Again assuming that the instantaneous value of the cavitation number were known,
the axisymmetric cavity profile, r = Rc ( x; t ) , could be estimated assuming quasi-steady
cavity growth using the Münzer-Reichardt formula:
2 n
⎛ x ⎞ ⎛ Rc ( x; t ) ⎞
⎜⎜ ⎟ +⎜ ⎟ = 1, (8.6.88)
⎝ ( t ) 2 ⎟⎠ ⎜⎝ d max ( t ) 2 ⎟⎠
where n = 2.4 . If d max and were known via equations (8.6.63) and (8.6.64), this profile
could be ignored (most easily using a numerical scheme) to estimate a cavity volume.
In the current formulation, the cavitation number driving the cavity geometry via
equations (8.6.63) through (8.6.67) must account for dynamical effects, primarily the added
mass due to cavity expansion. Thus, an approximation is made, consistent with the current
312
approach, that the cavity volume as determined via integration of the Münzer-Reichardt
profile, equation (8.6.67), is equal to that of the spherical bubble predicted via solution of the
equations governing the coupled canister-nozzle-bubble system, equations (8.6.48) for
choked flow, and equations (8.6.60) for non-choked flow.
In the current implementation, a numerical root-finding algorithm is employed to
determine that value of the instantaneous cavitation number, σ ( t ) in equations (8.6.63),
(8.6.64), and (8.6.66), for which the cavity volume as found via numerical integration of the
Münzer-Reichardt profile, equation (8.6.48), matches that of the spherical bubble as
predicted via solution of the coupled system.
8.6.11. Cavity Behavior at Start-up
Applying standard solution methods (the standard MATLAB solver ODE45 was used to
solve system (8.6.48) numerically for choked flow through the critical time, then system
(8.6.60) thereafter), it may be verified that setting the initial spherical bubble radius to some
very large value, such that the canister volume is effectively negligible in comparison,
recovers the standard results for isentropic blow-down through a converging nozzle. The
results are left as an exercise for the reader.
More interesting simulations employ the parameters and initial conditions presented
in Table 8.6.2 to explore the dynamics of the canister-nozzle-bubble system without regard
to the implications for supercavitating vehicle envelopment. The dimensionless results are
presented in Fig.8.6.2 and the associated dimensional quantities appear in Fig.8.6.3. The
black vertical line on each plot indicates the critical time – that instant when the pressure in
the bubble equals the nozzle critical pressure, and the flow switches from the choked to the
non-choked condition. That this is the crossing point can be seen clearly in Fig.8.6.2d. In
this example, although the bubble radius oscillates about its short-time-averaged non-
stationary mean value, the oscillations are so slight that their effect on both the bubble
pressure and the canister pressure (after the critical time) are negligible.
Table 8.42. Input to simulation for results presented in Figs.8.6.2 & 8.6.3.
Input Dimensional Value Units Dimensionless Value
γ 1.4 1.4
ρL 1025 kg/m 3
111.77
3
VT 0.1 m 4.845E-11
2
AE 7.854E-5 m 4.845E-11
T∞ 288 k 1.0
P∞ 1.516E5 pa 0.2
RB0 0.05 m 3.927E-5
TT0 288 k 1
PT0 7.580E5 Pa 1
ρT0 9.1706 kg/m 3
1
313
a b
c d
Fig.8.6.2. Dimensionless results for blow-down from a pressurized canister into a spherical
bubble for the case defined in Table 8.6.2: (a) bubble radius; (b) bubble radius time rate of
change; (c) bubble radius second time rate of change; (d) tank, bubble, and critical
pressures, and ratio of tank pressure to critical pressure.
The effects of canister volume and orifice diameter on supercavitating vehicle

envelopment are of special interest under this class of problems. To illustrate these effects,
the results of independently varying canister volume from a baseline of 500 cc to 300 cc and
the orifice diameter from a baseline of 2.4 mm to 4.0 mm are presented below. The other
simulation parameters are presented in Table 8.6.3, and the notional vehicle geometry is
shown in Fig.8.6.4.
Table 8.6.3. Approximate simulation parameters for a notional

supercavitating vehicle for results presented in Figs.8.6.4 thru 6.
vehicle length 1.9 m
vehicle maximum diameter 0.1 m
cavitator diameter 0.05 m
cavitator type disk
operating depth 2m
speed 37 m/s
canister initial pressure 10.34 MPa
canister initial temperature 291 K
314
a b
c d
Fig.8.6.3. Dimensional results for blow-down from a pressurized canister into a spherical
bubble for the case defined in Table 8.6.2: (a) bubble radius; (b) bubble radius time rate of
change; (c) bubble radius second time rate of change; (d) tank and bubble pressures.
Fig.8.6.4. 2
The notional 0.1
supercavitating test vehicle
Inflow
geometry used in the
simulations. The insert Radial 0.0
shows the geometry with the Coordinate 1 -1 0 1 2
radial scale expanded for (m)
clarity.
Inflow Cavitator
The results of three 0
simulation cases are
-1 0 1 2
presented in Figs.8.6.5 thru
Axial Coordinate (m)
7, where the bubble radius
and its first time rate of
change, the pressure in the canister and in the bubble, and the mass flow rate through the
nozzle are each plotted as functions of time. Vertical lines indicate the vehicle envelopment
time as predicted by the fully-coupled simulation, and as predicted if the blow-down
equation for discharge into an infinite volume is used. The critical time is also indicated. It
can be seen that, for the cases tested, the time to envelop the vehicle as predicted by the fully
coupled model can be significantly greater than that predicted by the blow-down equation
315
alone: the blow-down equation under-predicts the actual envelopment time by approximately
12-15% for the cases presented in Fig.8.6.5 and Fig.8.6.7. It can also be seen that decreasing
the canister volume from 500 cc to 300 cc eliminates the ability of the ventilation system to
generate a cavity large enough to envelop the vehicle, and that independently increasing the
orifice diameter from 2.4 mm to 4.0 mm reduces the envelopment time from approximately
2.1 s to 0.8 s.
However, consider the total effective vehicle endurance. This is taken as the time
required to equalize the pressure across the nozzle, at which point the ventilation system
ceases supplying gas to the cavity, which then begins to deflate as the hydrodynamic
processes near the cavity closure point entrain gas out of the cavity into the ambient fluid
(processes which are not modeled herein). It can be seen from the Fig. that this total run time
also decreases approximately proportionately. The inset curves in Figs. 8.6.5 and 8.6.7 show
the difference in the nozzle mass flow rate as predicted by the fully-coupled model and by
the blow-down equation once the critical time has passed, and the canister feels the
dynamical effects of the bubble.
A qualitative understanding of the evolution of the cavity geometry during this
process can be derived from Fig.8.6.8, which shows the cavity condition at various times
during a simulation based on the fully-coupled model. The parameters are the same as those
presented in Table 8.6.3, except that the cavitator diameter has been increased to a value of
0.07 m. As described above, the geometry at each instant is estimated by deforming the
spherical bubble into the equivalent Münzer-Reichardt cavity geometry using Garabedian’s
formulae to predict the instantaneous length and maximum diameter of the cavity based on
the instantaneous cavitation number.
vehicle envelopment time:

blow-down equation
coupled system
critical time
Fig.8.6.5. Simulation results for the notional supercavitating vehicle and operating
conditions in Table 8.6.3 for a canister volume of 500 cm3 and an orifice diameter of 2.4mm.
316
no vehicle envelopment:
blow-down equation
coupled system
critical time
conditions in Table 8.6.3 for a canister volume of 300 cm3 and an orifice diameter of 2.4mm.
vehicle envelopment time:

blow-down equation
coupled system
critical time
conditions in Table 8.6.3 for a canister volume of 500cm3 and an orifice diameter of 4.0mm.
317
e
Fig.8.6.8. Animation frames for notional vehicle at the following times: (a) 0.000s; (b)
0.558s; (c) 1.176s; (d) 1.794s; (e) 2.370s. The vehicle geometry and operating conditions
are defined in Table 8.6.3, except that the cavitator diameter is 0.07m.
A more global understanding of the effects of cavity growth dynamics on ventilation

system performance may be developed via parametric analysis in which the key elements of
the set of baseline parameters presented in Table 8.6.4 are varied independently.
Table 8.6.4. Approximate simulation parameters for a notional supercavitating

vehicle for results of parametric variation presented in Figs.8.6.9 and Fig.8.6.10.
Constant Parameters for All Cases
vehicle maximum diameter 0.1 m
cavitator diameter 0.05 m
cavitator type Disk
speed 37 m/s
canister initial temperature 291 K
Baseline Parameters for Variation
operating depth 2m
canister volume 500 cm3
orifice diameter 2.4 mm
canister initial pressure 10 MPa
Using such an approach, the effects of canister initial pressure, canister volume,
orifice diameter, and absolute depth pressure (a function of vehicle operating depth) may be
explored. Results of such an analysis are presented in Figs.8.6.9 & 10. Several conclusions
can be drawn from these graphs. Firstly, increasing the orifice diameter has a very strong
effect on reducing the vehicle envelopment time, the critical time, and the total run time.
The linear increase in critical time (and run time) with canister volume, which is obvious
from the formulation presented above, is also clear from the results presented in Fig.8.6.9.
Conversely, the vehicle envelopment time decreases monotonically with this parameter. The
318
effects of varying the canister initial pressure are also important and are qualitatively similar
to the variation in canister volume, although not as pronounced (and the relationship is not
linear in the case of the variation of critical time). The back pressure imposed by a change in
the operating depth of the vehicle has a rather weak effect. What is perhaps of greatest
interest to the would-be designer is the fact that, in that part of the design space for which the
run time is greatest, and therefore of most practical value to testing, the error introduced by
employing the blow-down equation, rather than the fully-coupled model, is negligible; it is
only in cases for which the ventilation system has marginal performance to begin with that
the effects of cavity dynamics are predicted to be significant.
8.6.12. Recapitulation of the Simplified Cavity Dynamics Problem
It is thus seen that a first-principles (albeit approximate) fully-coupled model may be
formulated, implemented, and used to predict the dynamical effects of cavity expansion on
the performance of a pressurized gas system for supercavity ventilation. The underlying
physical representations incorporated in the model are believed to provide a conservative
estimate of the effects of the added mass due to cavity expansion.
• Small range for which bubble dynamics are important

• Occurs at low tank pressures or volumes
• Notional system falls in this range
Notional system
Fig.8.6.9. Time required to acheive key simulation events as a parametric variation of

canister initial pressure (left pair of graphs) and canister volume (right pair of graphs).
However, based on the results of the model, it is concluded that added mass due to
cavity expansion is important only over a small portion of the design space that is of interest
to designers of supercavitating test vehicles employing pressurized gas systems as a means
of cavity ventilation. Specifically, it is only when the design parameters are such that the
system would display merely marginally acceptable performance that the fully coupled
model predicts significantly different behavior than would be predicted using the blow-down
319
equation (i.e. ignoring the effects of cavity expansion dynamics on the gas discharge
process).
absolute depth pressure [Pa]
Cavity oscillations result in jagged parametric variation for coupled system
Notional system
absolute depth pressure [Pa]
Fig.8.6.10. Time required to acheive key simulation events as a parametric variation of

orifice diameter (left pair of graphs) and absolute depth pressure (right pair of graphs).
The parametric analysis presented above shows that the nozzle orifice diameter must
be selected small enough that test run times are not reduced below a useful duration.
Assuming that there is sufficient room on board a test vehicle, increasing the canister volume
has the greatest effect on improving ventilation system performance. This very simplified
example illustrates certain elementary relationships between the thermodynamics of the
ventilation gas and the dynamics of cavity expansion. It also helps bound the design and
operational space over which the dynamical effects of cavity expansion are important.
This model is not, however, sufficient to address another problem of interest, namely
the sustained ventilation of a cavity by means of a gas generator. Although part of the
analysis presented above may be useful in such an analysis, that problem involves
considerations of chemical reaction and heat transfer that are beyond the scope of this book.
Moreover, with respect to the dynamics of fully-evolved supercavities, the model
presented above is deficient in the sense that it ignores the importance of forward speed to
the supercavitation problem. Among other shortcomings, the spherical bubble geometry
selected to model the expanding cavity dynamics is so compact that it loses an essential
element of supercavity dynamics, namely the advective delay that can lead to such important
effects as cavity auto-oscillation. That effect and the important concept of gas loss are
discussed in the following section.
Figs.8.6.11 [Wosnik et al., 2003] and 8.6.12 [Kawakami et al., 2009] show two
cavity start-up sequences, one a supercavity downstream of a disk cavitator mounted on an
upstream sting, and the other a supercavity enveloping a body. In both cases, the cavities
320
were ventilated with air injected just behind the cavitator, such that the cavitation number
decreased as the sequence progressed. The end result of each test was a well-formed, glassy
cavity, suggesting that the final cavities were both operating in the twin vortex regime of
flow. Initially, when the cavitation number was still rather high, the flow traversed the re-
entrant jet regime. The presence of the strong ring vortex structure in the third frame of
Fig.8.6.11 suggests that the flow also traversed the toroidal shedding regime. In general, the
ventilation coefficient required to maintain a cavity in the twin vortex regime is greater than
that in the re-entrant jet regime, because the twin vortex system provides a low-impedance
path for the loss of gas from the closure region.
The toroidal shedding regime typically features a local maximum in the curve of ventilation
coefficient versus cavitation number (at a fixed cavity Froude number), since the intermittent
shedding of the gas-filled toroidal vortex structures is
a moderately efficient mechanism for transferring gas
out of the cavity. [Logvinovich, 1969] presents an
elegant formulation proving that the cavitation number
in a cavity in a transverse gravity field cannot be
reduced below a fixed limit, although the practical
lower limit of the cavitation number for a nominally
axisymmetric cavity is typically higher than that value.
Fig.8.6.11.
Start-up of a supercavity downstream of a disk
cavitator mounted on an upstream sting. (All images
courtesy of R Arndt.)
Fig.8.6.12.
Start-up of a supercavity enveloping a body, including
the partial cavitation phase of flow. (All images
courtesy of R. Arndt.)
Between this so-called “left wall” and the toroidal

shedding maximum of the ventilation curve is a
portion of the curve in which a local minimum in the
ventilation requirements exists. The presence of a
body inside the cavity can complicate these
relationships.
321
Ch. 9. Numerical Methods

The above-considered analytical methods allow to obtain results with the desired accuracy
and, more importantly, to establish typical behaviors of flows with free boundaries. In spite
of numerous works and a rapid spread of analytical methods, there is a wide range of applied
problems which cannot be solved analytically. Some numerical methods are considered in
this chapter, most of them have been developed by the authors. The methods provide great
simplicity and efficiency in computing applied problems including those that can be solved
analytically. Considered first is the well-known boundary method connected with cavitating
flows, and then other methods applied to different problems of flow with free boundaries.
Sec. 9.1. Boundary Element Method

9.1.1. Remarks on the BEM
The history of boundary element method (BEM) began probably in the 1960’s. The concept
was first presented and used in [Jaswon, 1963; Symm, 1963], and another concept was
presented in [Zienkiewicz & Cheung, 1967]. Since then both methods have been widely used
to investigate many problems of continuum mechanics. More information on both methods
can be found in [Mackerle, 1989]. The finite and boundary element techniques and their
various applications are presented in [Brebbia, 1978; Brebbia et al., 1984; Banerjee &
Butterfield, 1981; Terentiev & Afanasiev, 1987].
The BEM is applied mostly to boundary-value problems including the development
of governing integral equations, approximation of the integration domain of a boundary
elements system, and approximation of continuous functions of those boundary elements.
Primary consideration is given to ensuring that the sought solution satisfies certain
conditions (e.g. governing differential equation) inside the domain. The boundary integral
equations can be obtained from the governing differential equations using Green’s identities,
App. 3, and methods of weighted residual functions [Brebbia et al., 1984]. Then the integral
equations are reduced by the approximation methods to systems of linear equations.
The BEM consists of two operations: approximations of boundary and unknown
functions. The simplest approximation for a two-dimensional domain is a polygon with
straight sides, so-called, “elements”, and for three-dimensional domain a polyhedron with
triangular sides. At each element the function can be expanded in Taylor series with one,
two, or more terms kept, in which, according to approximations, the elements are called
constant, linear, or higher-order elements. Generally, the use of high-order elements
increases the accuracy of calculations but the algorithms become much more complicated.
Constant and linear elements will mainly be used below.
For a two-dimensional domain, one can estimate the accuracy of approximations.
Based on variation principles of conformal mapping [Lavrent’ev & Shabat, 1965],
transformations of two close domains onto the same circle differ from each other by
difference of their areas. For small elements the area of segment is a small of the order of
O (ln3 / Rn ) , where ln and Rn are the length of segment and the radius of curvature at a node.
The length of element is inversely proportional to the number of elements, N−1, and so the
total error is about O(N−2). Besides, approximation of a function of an element yields also an
error, which can be estimated roughly as O (ln2 ) . Indeed, if a node is a middle point, then
322
integral of expansion f ( s ) = a0 + a1 ( s ) + a2 s 2 + ... along the n-th element is

ln / 2
∫− ln / 2
f ( s )ds = a0ln + a2ln3 / 3 + O(ln4 ) . Thus, the relative error for both approximations,
constant or linear, is O (ln2 ) or O ( N −2 ) . This justifies using constant elements in the BEM.
9.1.2. Green’s Identity

Let the function u be double-differentiable and single-valued; domain D is generally multi-
connected with non-self-intersecting and closed boundary C. Then the Green’s integral
identity valid Lu = − ∫ ∇ 2u G (r )dV , (9.1.1)
D
∂G (r )
where Lu = ε u ( z ) + ∫ u (τ ) ds − ∫ q (τ )G (r )ds . (9.1.2)
C ∂n C
Here q = ∂u / ∂n is a derivative directed along an external normal to contour C at integration
point τ ; G (r ) is Green's function ( G (r ) = − ln r / 2π for 2D domain; G (r ) = 1/ 4π r for 3D
domain); r is a distance between boundary points z and τ . The value of constant ε depends
on the location of that point at which Green’s identity is applied: ε = 1 for an internal point,
or ε = 1/ 2 for a point on the smooth boundary. In general, the normal n at some points on
the boundary can be discontinuous with increment Δθ; at these points, factor ε is equal to
1 − Δθ / π but the appropriate choice of elements can provide no discontinuities of the
normal to the sides of a polygon.
Equation (9.1.1) is an identity and doesn’t give a solution of any problem.
Appropriate conditions in domain D and on boundary C should also be satisfied. Further, we
shall consider mostly harmonic functions, which satisfy Laplace equation ∇ 2 u = 0 , thus
equation (9.1.1) is reduced to Green’s expression Lu = 0 . (9.1.3)
The latter together with certain conditions on the boundary determine the solution of inner
value problem of potential flows. The flow domain is usually outer, thus the function’s
behavior at infinity should be given. For meromorphic functions, Green’s equation (9.1.3)
could be generalized [Terentiev, 1994] to equations (A3.12) or (A3.13) which is widely used
below.
9.1.3. Discretization of Green’s Equation
Consider first an inner value problem bounded by
closed curve C. Boundary element method consists of
two interpolations: a boundary of flow domain and
unknown functions on each element. There are many
modes of interpolation for both boundary and
functions. The simplest interpolation curve is a
Fig.9.1.1.
polygon with straight sides, the so-called elements,
Boundary of internal domain and
Fig.9.1.1 (the number of elements is N = 7.
approximating polygon.
Denote joint points (knocks) of two neighboring
elements ck −1 and ck by zk (k = 1, N ) with z0 ≡ z N and the middle point (nodal) of the kth
boundary element by Z k = 0.5( zk + zk −1 ) . The length of elements
lk =| zk − zk −1 | (k = 1, N ) , (9.1.4)
323
is supposedly small and elements ck form a polygon which approximates boundary C so that
∂G (r )
operator (9.1.2) is approximated by Lu = ε u ( z ) + ∑ ∫ u (τ ) ds − ∑ ∫ q(τ )G (r )ds .
k ck ∂n k ck
Expanding function u (τ ) and its derivative q(τ ) into Taylor series at the middle of
each element and keeping the first terms only, one can write the value of operator at other
middle points Z m as Lum = ∑ Akm uk − ∑ Bkm qk , or in matrix form
k k
L = AU − BQ . (9.1.5)
The components of matrix-vectors U and Q are values of functions u and q at the middle
points, the array cells of matrices A and B are the integrals:
∂G (rm )
Akm = εδ km + ∫ ds , Bkm = ∫ G (rm )ds , (9.1.6)
ck
∂n ck
where rm = ( X m − τ x ) 2 + (Ym − τ y ) 2 is a distance between integrated point ( τ x , τ y ) and

nodal points ( X m , Ym ). Analitic expressions of the latter integrals are given in App. 5. This
approximation will mainly be applied for computing flow problems. Another simple
approximation is linear, which is also given in App. 5.
9.1.4. Governing Equations for Foils without Cavities
Flow problems of hydrofoils without cavitation are well known and determined by potential
function ϕ which satisfies Laplace’s equation ∇ 2ϕ = 0 inside the fluid region, and
Neumann’s condition ∂ϕ / ∂n = 0 at the foil boundaries. Additionally, the Joukowski-Kutta
condition of limited velocities at the trailing edge of the foil must be satisfied
limV ( s ) + limV ( s) = 0 , (9.1.7)
s →0 s→L
where s is the arc-length along the foil boundary and L is the total length of the foil
perimeter. Instead of condition (9.1.7), the circulation around a foil may be given
∫ V (s)ds = −Γ .
C
(9.1.8)
Besides, if inflow speed at infinity is v∞ eiβ then the complex potential at infinity has a
Γ
leading part as w( z ) ≈ v∞ e − iβ z + ln z . (9.1.9)
2π i
Whence the behavior of speed potential and stream function are
Γ
ϕ ≈ −v∞ ( x cos β + y sin β ) + θ , θ = arg z , (9.1.10)
2π
Γ
ψ ≈ v∞ ( y cos β − x sin β ) − ln | z |, | z |→ ∞ . (9.1.11)
2π
Thus the speed potential is a multi-value function because of angle θ , and so the Green’s
equation (9.1.3) cannot be used directly on contour C. Therefore it is necessary to apply
additional mathematical techniques to Green's equation, though the speed potential is very
useful for numerical investigation of many problems of steady and unsteady flow.
In the case of steady problems of planar and axisymmetric flows, a stream function is much
more preferable. The stream function is single-valued and so the Green’s equation (A.3.13)
324
is valid for it. Contrary to Fig.9.1.1, the flow domain is outside the foil; and the function has
singularity at infinity. The logarithmic singularity of the stream function and the Green’s
function cancel each other at infinity, thus it ignored. It should be taken into account the next
singularity coinciding with F ( z ) = v∞ e − iβ z , z → ∞ . (9.1.12)
The stream function is constant at the foil boundary:
ψ = c = const . (9.1.13)
Constant c is unknown and is different at each foil. If c1 and c2 are constants on two foils,
then the amount of liquid flowing between these foils is equal exactly to
q =| c2 − c1 | . (9.1.14)
9.1.5. Single Foil

Taking into account condition (9.1.12) and identity (A3.16), equation (A3.13) for the stream
function is transformed to ∫ v( s′)G ( s, s′)ds′ + c = y ( s )cos β − x( s )sin β , (9.1.15)
C
where v( s′) = −∂ψ / ∂n is an unknown distribution of velocity at the foil boundary; x( s ) and
y ( s ) are coordinates of foil, s and s′ are curvilinear abscissas; the velocity at infinity is
assumed to be unity.
Dividing the foil boundary C into N elements, Fig.9.1.2, and satisfying integral
equation (9.1.15) at N middle points of elements, one obtains the set of N linear equations
N
∑B
k =1
V + c = Ym cos β − X m sin β ,
km k m = 1, N , (9.1.16)
sk +1
where Bkm = ∫ G(Z

sk
m , s′)ds′, Vk = v( Z k ) . (9.1.17)
Analitic expression of integral Bkm is given in App. 5 by equation (A5.6).
Fig.9.1.2.
Foil and its approximation
polygon of 10 elements; the
joint and middle points are
marked by o and x.
Point s1 = 0 of the first element coincides with point sN +1 of the last element. Let
these points locate at the foil trailing edge, where the Joukowski - Kutta condition should be
satisfied as follows: V1 + VN = 0 . (9.1.18)
Denoting constant c as VN +1 , equations (9.1.16) and (9.1.18) can be combined in an equation
set and written in a matrix form as: BV = Z , (9.1.19)
where B is the ( N + 1) × ( N + 1) - matrix with elements
325
⎛ B11 B12 … B1, N −1 B1N 1 ⎞

⎜ ⎟
⎜ B21 B22 … B2, N −1 B2 N 1 ⎟
Bmk =⎜ 1⎟, (9.1.20)
⎜ ⎟
⎜ BN 1 BN 2 BN , N −1 BNN 1 ⎟
⎜ ⎟
⎜1 0 0 1 0 ⎟⎠
⎝
Z is the (N+1)-vector with elements
⎛ Y1 cos β − X 1 sin β ⎞
⎜ ⎟
⎜ Y2 cos β − X 2 sin β ⎟
Zm = ⎜ ⎟, (9.1.21)
⎜ ⎟
⎜ YN cos β − X N sin β ⎟
⎜ 0 ⎟
⎝ ⎠
V is the (N+1)-vector with unknown elements, V1 , V2 , , VN , c .
The solution of equation (9.1.19) V = B −1 Z gives the velocity at the middle of
elements (the first N elements) and the value of stream function, c = VN +1 .

The hydrodynamic force and torque of the k-th element are calculated by
Fkx = −Vk2 ( yk +1 − yk ) / 2,
Fky = Vk2 ( xk +1 − xk ) / 2, (9.1.22)
M k = Fky X k − FkxYk .
Summing equations (9.1.22) one obtains the total force and the torque as
N N
Fx = ∑ Fkx , Fy = ∑ Fky , (9.1.23)
k =1 k =1
N
M = ∑ Mk . (9.1.24)
k =1
The lift and drag are found from (9.1.23) L = Fy cos β − Fx sin β , (9.1.25)
D = Fy sin β + Fx cos β . (9.1.26)
N
The circulation around a foil is Γ = −∑Vk lk . (9.1.27)
k
The lift can also be found by circulation L = −Γ . Besides, the drag should vanish, D = 0 .
These two equalities could be used to examine the numerical accuracy.
9.1.7. Neutral Direction
Let the vector e − iγ be inlet velocity for which the lift or circulation vanishes. Equation
(9.1.16) gilts in this case too but angle β should be changed to γ . Dividing that equation by
cos γ and moving the item X m tan γ from the right-hand side to the left, we obtain an
equation set of N + 2 unknown variables. Adding equation (9.1.27) by Γ = 0 to equations
(9.1.19), the linear equation set can be written also in a matrix form as (9.1.19) of N + 2
326
dimensions. The ( N + 2) th column in matrix B consists of Bk , N + 2 = X k if k ≤ N ; Bk , N + 2 = 0

if k = N + 1 and k = N + 2 ; the ( N + 2) th line consists of BN + 2, m = lm if m ≤ N ; BN + 2, m = 0 if
m = N + 1 and m = N + 2 ; the first N components of vector Z are Z k = Yk , the last two
components are equal to zero. The required angle γ is calculated from the ( N + 2) th
component of matrix-vector V and then the speed at the middle points of elements are found
from the first N components: tan γ = VN + 2, vk = Vk cos γ . (9.1.28)
As above, the ( N + 1) th component determines the value of stream function VN +1 = c .
9.1.8. Joukowski’s Hydrofoil

Let’s consider as an example a flow with unit stream speed around the Joukowski’s
hydrofoil. Though its equation is well known, it is given here because it will be often used
for demonstration of various methods. The Joukowski’s foil is determined by equations
x(t ) = 1 2 (1 + 1/ g (t ) )[a cos(γ − t ) − d cos(γ )],
(9.1.29)
y (t ) = 1/ 2 (1 − 1/ g (t ) ) [a sin(γ − t ) − d sin(γ ) + h],
where g (t ) = a 2 + d 2 + h 2 − 2ad cos t + 2ah sin(γ − t ) − 2d h sin γ ,
a = d + 1 + h 2 + c, γ = −arctg h .
Parametric variable t ∈ [0, 2π ] , parameters d, h and c characterize the thickness, flexure and
roundness of the trailing edge. These parameters can be found by introducing any three
conditions of a foil profile, e.g. radius of the leading edge curvature, maximum thickness and
deflection in the mid-chord. The chord of foil is obtained as l = max x (t ) − min x (t ) . Angle
γ determines the neutral axis, and the angle between the inflow speed direction and neutral
axis is angle of attack α = β − γ , where β is the inflow speed angle.
Exact solution of flow problem is given by the speed potential at foil boundary
Γσ
ϕ (σ ) = a cos(σ + α ) − 0 , Γ 0 = −2π R sin α . (9.1.30)
2π
dϕ ϕ ′(t )
Velocity at the foil is v(σ ) = , or v(t ) = . (9.1.31)
ds x′(t ) 2 + y ′(t ) 2
Now, we can obtain all dynamic characteristics; the lift is exactly equal to the
2π
1
circulation ( L = Γ ) while the torque is M = − ∫ v 2 ( xx′ + yy′)dt . (9.1.32)
20
Dividing the unit circle ζ = e − iσ uniformly into N elements as σ k = 2π k / N ( k = 0, N ), one
obtains N elements of different lengths lk on the foil.

Numerical results have been obtained for a foil
with h=0.1, d=0.05, c=0.025, and angle of attack
α=15°, inlet angle β=0.1621. The contour of
Joukowski’s hydrofoil is shown in Fig.9.1.3.
Fig.9.1.3. Joukowski’s foil.

327
The dotted straight line is the neutral line. Directions of the stream speed and the lift action
line are plotted as well. Velocity distribution over the contour is shown in Fig.9.1.4. The
curve has a peak trajectory, but the numerical approach of BEM calculates it very well even
for this case.
Fig.9.1.4.
Speed distribution
on Joukowski foil:
solid lines
correspond to
numerical data,
points – to
analytical results.
Dependence of numerical data on element number is given in Table 9.1.1. Taking

about hundred elements is enough for practical purposes. But the calculation time increases
little with an increase in the number of elements. The computer time is spent mostly for
calculating matrix (9.1.20) only. For the same foil and a fixed number of elements, changing
the angles take very short computer time.
Table 9.1.1. Dependence of numerical data on number of elements.

N 50 100 200 300 400 Exact
γ -0.0990 -0.0995 -0.0995 -0.0996 -0.0996 -0.0997
c 6.7579 6.7787 6.7865 6.7874 6.7880 6.7858
Γ 1.7460 1.7538 1.7557 1.7560 1.7561 1.7563
CM 0.2618 0.2622 0.2624 0.2624 0.2624 0.2624
9.1.10. A System of Hydrofoils
The numerical method can be easy applied to calculating hydrofoil systems. Curvilinear
abscissas are selected on each foil, and each contour is divided onto finite elements. Since
the stream function is an unknown constant at each boundary, there are as many constants as
the number of hydrofoils. To clarify the numerical procedure, consider two foils with
boundaries C1 and C2 . Integral equation at both boundaries due to integral identity (9.1.15)
can be written as ∫ v(s′)G(s, s′)ds′ + c(C , C ) = y(s) cos β − x(s)sin(β ) ,
C
1 2 (9.1.33)
where C = C1 ∪ C2 , c(C1 , C2 ) = c1 if s ∈ C1 , and c(C1 , C2 ) = c2 if s ∈ C2 . Functions x( s )

and y ( s ) determine both hydrofoils C1 and C2 .
Let N1 and N2 be the numbers of elements at these contours, respectively. Number
now the elements from 1 to N1 for the first foil and from N1 +1 to N= N1 +N1 for the other
foil, and add two unknown constants c1 and c2, and also two additional conditions similar to
(9.1.18) at trailing edges, then an equation set can be written in a matrix form like equation
(9.1.19). The (N+2)x(N+2) -matrix B is arranged by NxN elements Bmk, which are calculated
by equation (9.1.17). The first N1 elements of (N+1)th column are unity, the rest are zero; the
elements from N1 + 1 to N of (N+2)-th column are unity, the rest are zero, i.e.
Bm , N +1 = θ ( N1 − m), Bm, N + 2 = θ ( N1 + N 2 − m) − θ ( N1 − m) , (9.1.34)
328
where θ ( x) is Heviside-function. Elements of two last matrix rows are zero except two
elements of each row: BN +1,1 = 1, BN +1, N1 = 1, BN + 2, N1 +1 = 1, BN + 2, N1 + N2 = 1 . (9.1.35)
Elements of (N+2)-th matrix-vector Z are expressed exactly as (9.1.21) but two last elements
should be zero; the first N1 elements are determined by the C1 -foil, the next N 2 elements
are determined by C2 . The first N1 elements of (N+2)-th matrix-vector V determine speed at
contour C1 , the next N 2 elements determine speed at C2 ; the N + 1 -th element gives
constant c1 and the last element is equal to c2 .
In that way, it is possible to consider several hydrofoils or obstacles. Circulation
could be given instead of condition (9.1.18), and then all corresponding elements should be
equal to the length of elements of the contour.
9.1.11. Airfoil near a Bottom
An airfoil near a flat bottom could be considered as
two airfoils symmetrical to the bottom, Fig.9.1.5.
Denote the distance from the trailing edge to the
bottom as H. If the x-axis coincides with the bottom,
the stream speed is parallel to the x-axis. Thus the
components of matrix Z in equation (9.1.19) are
equal to ordinates Ym only. Some results of
numerical calculations for Joukowski’s airfoil are
plotted in Figs.9.1.6 and 9.1.7. Each boundary of a
foil was divided by 250 elements. The hydrofoil
corresponds to h = 0.1, d = 0.05, c = 0. Fig.9.1.5. A foil near a bottom.
Fig.9.1.6. Speed distribution over a foil Fig.9.1.7. Effect of distance of foil on lift.
near a bottom.
Fig.9.1.6 shows velocity distribution over the foil at distance H = 0.1 and angle β
=10°. The dotted line corresponds to an infinite distance. It worth noting that in this case the
velocity at both low and upper surfaces is less than that at an isolated foil. Fig.9.1.7 shows
the dependence of lift on distance. The lift is expressed as a ratio to the lift for single foil:
329
L = L( H ) / L(∞) . It is seen that near the bottom the lift generally decreases but increases at
small angles β and for very small distance H,. Using BEM one can study the flow of several
foils in more detail [Terentiev, 1994; Terentiev & Kartuzova, 1995, 1996].
9.1.12. Symmetric Flow
The flow is symmetric about the real axis x. At that case, a half symmetrical part of the flow
domain can be considered, thus reducing the dimension of equation set to a half. For
symmetric flow along the x-axis, angle α = 0 and equation (9.1.15) becomes
c + ∫ V ( s )G ( z ,τ )ds = P, P = y ∈ C , (9.1.36)
C
1 ( x − ξ )2 + ( y + η )2
where G ( z ,τ ) =
ln .
4π ( x − ξ ) 2 + ( y − η ) 2
Eq. (9.1.36) may be discretized similar to (9.1.16).
9.1.13. Axisymmetric Flow
Axisymmetric flow is similar to symmetric planar flow. The similarity is more striking if the
stream function is considered. The stream function ψ ( x, y ) in cylindrical coordinates x, y is
also constant on an axisymmetric surface like a ring wing, and its partial derivatives
determine the velocity component, Sec. 4.5. The stream function for a body of revolution
satisfies the integral equation [Sidorov, 1958] as
y2
∫
C1
v ( s ) H ( z ,τ ) ds = P , P =
2
, (9.1.37)
yη ⎡ 2 ⎛2 ⎞ ⎤
where H (ξ ,η ; x, y ) = ⎢ E (k ) − ⎜ − k ⎟ K (k ) ⎥ , k = 2 yη / ( x − ξ ) 2 + ( y + η ) 2 .
2π ⎣ k ⎝k ⎠ ⎦
Here K (k ) and E (k ) are complete elliptic integrals of the first and second kind,
respectively. This equation is fundamental in many numerical methods. It is noteworthy that
equation (9.1.37) is valid for axisymmetric bodies without any opening. For bodies such as
ring wings, unknown constant c should be added to equation (9.1.37)
y2
c + ∫ v( s ) H ( z ,τ )ds = P, P = . (9.1.38)
C1
2
The constant determines the volumetric flux of liquid through the ring surface:
q = 2π c . (9.1.39)
Integral equation (9.1.38) looks like equation (9.1.36) and so the problems of both planar
and axisymmetric flow can be solved by the same numerical method, BEM, but the
components of matrix B should be calculated numerically. A comparison with analytical
solution of the ellipsoid of revolution shows a very high efficiency of the method.
9.1.14. The Ring Wing
Consider a ring wing with a generatrix of the Joukowski’s foil. Some results of calculations
are plotted in Fig.9.1.8; the points correspond to analytical results for the same foil in a
planar flow. Coincidence of the theoretical and numerical results in the planar case are good
even for the sharp top of the speed distribution curve.
330
Fig.9.1.8. Ring wing:

(a) - the Joukowski’s foil,
(b) - distribution of velocity along
foil boundary. The points
correspond to exact results.
A ring wing is used for increasing the efficiency of screw propellers, but any
conclusive explanations are lacking. Compare two positions of the same generatrix foils and
the same radius determined as a distance from the trailing edge to the axis (Fig.9.1.9).
Calculations show that the amount of liquid
which flows through first ring (a) is q=0.969
while second one (b) –q=0.617 only, i.e. the
screw propeller in the first ring accelerates liquid
much more than the second one, and more than
without rings ( q0 = π R02 = 0.439 ). Therefore, the
first ring nozzle increases efficiency of the screw
propeller. At the same time, calculations show its
Fig.9.1.9. Two shapes of the rings. shortcomings. Since the least radius of the ring
wing (a) is R0=0.374, the mean speed at that
section is V0=2.202, and moreover, the speed on the shape is higher than V0. Hence, the
cavity can arise earlier than without a nozzle and this can make the use of a ring nozzle
problematical for fast ships [Rusetsky et al., 1971].
9.1.15. Axisymmetric Body with a Ring-Wing
A lot of engineering problems can be
calculated just by using systems of
two axisymmetric shapes: intake
diffusers, ring stabilizers, a body in a
body, in a wind tunnel and so on. For
example, a system of the ellipsoid of
revolution and a ring wing is shown
in Fig.9.1.10a.
Fig.9.1.10.
(a) - An ellipsoid and the Joukowski
hydrofoil; (b) speed distributions on
the foil (c) speed distribution on the
ellipsoid.
Velocity distributions along the foil and the ellipsoid are shown in Fig.9.1.10b & c.
Analytical results for single foil and ellipsoid are plotted for comparison. As seen in
Fig.9.1.10 the interference can be quite considerable. Namely, both curves at the left and
331
right plots have horizontal straight-line segments (two-sided arrows) of almost the same
values corresponding to a ring channel between two parts of the foil and ellipsoid shapes.
Sec. 9.2. Penetrable Hydrofoils

An increase of lift/drag ratio can be realized by sucking fluid via the permeable part of the
profile boundary or by squeezing the fluid from the profile to the main flow. Some problems
of penetrable hydrofoils have been studied analytically in [Golubev, 1935; Nekrasov, 1959;
Woods, 1961; Elizarov et al., 1994]. Generally, the stream function and the potential are
multi-value functions so the Green’s identities are invalid. In this case, flow problems
should be separated into two problems: numerical conformal mapping of a flow domain onto
some simple one of an auxiliary plane and then solving analytically a value problem to the
given boundary conditions on the auxiliary plane [Terentiev & Smirnova, 1998, 2000]. The
numerical-analytical method is demonstrated below on examples of a single and two
penetrable hydrofoils.
9.2.1. A Single Hydrofoil
The problem statement is shown in Fig.9.2.1. As
before, let the origin of curvilinear abscissa be at
the trailing edge of hydrofoil and its positive Fig.9.2.1. Flow around penetrable foil.
direction be clockwise. The normal speed
component is zero at the boundary except the
part (sm, sn), where it is given a value of q (q>0 for outflow from the foil). It is assumed that
the main stream and the outflow or inflow have the same Bernoulli constant. Practically
there can be different Bernoulli constants but the flow problem in that case should be solved
differently. The complex potential in infinity behaves as
sn
Γ + iQ
w( z ) = z + ln z + … , z → ∞, Q = ∫ q ( s )ds . (9.2.1)
2π i sm
Hydrodynamic resultant force is equal to −(Γ+ iQ), thus the lift and drag are expressed as
L = −Γ, D = −Q . (9.2.2)
The stated problem could be solved analytically if the hydrofoil would be a circular cylinder.
Hence, the flow domain in Fig.9.2.1 must be mapped onto exterior of the cylinder.
9.2.2. Numerical Mapping onto the Exterior of the Unit Circle
Consider first a non-circulating flow problem of non-penetrable hydrofoil. Solving
numerically equation (9.1.19) with (N+2)-matrix B including both equations (9.1.7) and
(9.1.8) for Γ = 0 like the problem of the neutral direction, (9.1.7):
⎛ B11 B12 … B1, N −1 B1N 1 X 1 ⎞
⎜ ⎟
⎜ B21 B22 … B2, N −1 B2 N 1 X 2 ⎟
⎜ 1 ⎟
Bmk = ⎜ ⎟. (9.2.3)
⎜ BN 1 BN 2 BN , N −1 BNN 1 X N ⎟
⎜ ⎟
⎜1 0 0 1 0 0⎟
⎜l l lN −1 lN 0 0 ⎟⎠
⎝ 1 2
332
The right-hand term is ( N + 2 )-vector Z with elements Z m = Ym ( m = 1, N ) and

Z N +1 = 0, Z N + 2 = 0 . The ( N + 2 )-th element of a solution of the matrix equation (9.1.19),
V = B −1Z , determines the neutral direction, and N first elements determine velocity at the
middle of finite elements as follows: tan γ = VN + 2 , Vk = Vk cos γ , k = 1, N . (9.2.4)
Instead of matrix (9.2.3) one can use another matrix with the same array of cells but
the (N+1)-th column; N cells of which are –Ym and the last two cells are zero. The N cells of
right-hand vector are unities and the last two cells are zeros. In that case all velocity
components are determined to the constant factor 1/c, which can be calculated by the given
length of the cut using the constant [Terentiev & Smirnova, 1998].
The speed potential at the joint and nodal points can be presented as N-vectors:
k
1
Φ k = ∑Vm lm , φk = (Φ k + Φ k −1 ), k = 1, N . (9.2.5)
m =1 2
Since equation (9.2.5) represents discrete data, one can replace continuous function Φ ( s )
with some interpolating functions or splines, and then find the length of the cut. However,
the left edge of the cut for a large number of N could be calculated as a = min(Φ ) and the
right one b=0, so that the cut length L = −a. Equation (9.2.5) together with
k
sk = ∑ lk , k = 1, N (9.2.6)
m =1
determine interrelations of velocity potential and the length of curvilinear abscissa.
Consider now a W-plane with a cut of length L. The transformation from the W-
plane to the ζ -plane with clipping unit circle is realized by function
W = L(ζ − 1) 2 / 4ζ . (9.2.7)
Whence, the dependence of the central angle of circle ζ = e iσ
on speed potential at nodal
⎧⎪2π − arccos ( 2φk / L + 1) if φk > Φ k ,
points can be found by σk = ⎨
⎪⎩arccos ( 2φk / L + 1) if φk ≤ Φ k .
(9.2.8)
9.2.3. Analytical Solution for Flow around a Circle
Consider a flow around a penetrable circular cylinder on the ζ -plane. The stream speed of
⎛ dw ⎞ ⎛ dw ⎞ ⎛ dz ⎞ ⎛ dW ⎞ L
flow is determined as ⎜ ⎟ =⎜ ⎟ ⎜ ⎟ ⎜ ⎟ = eiγ . (9.2.9)
⎝ d ζ ⎠ζ →∞ ⎝ dz ⎠ z →∞ ⎝ dW ⎠ w→∞ ⎝ d ζ ⎠ζ →∞ 4
Let the penetrable part of the circle be an arc δ mδ n , which is determined by penetrable part
of the hydrofoil boundary sm sn using equations (9.2.8). For sufficiently small elements, a
quantity of liquid leaking through each penetrable element can be written as qk lk . Exactly
the same quantity of liquid should leak through corresponding arc of the circle. Therefore,
the complex potential of the flow around penetrable cylinder on the ζ -plane is expressed as
L ⎛ iγ e − iγ ⎞ Γ 1 n ζ −ζ
w(ζ ) =
4⎝
⎜ ζ e + ⎟ +
ζ ⎠ 2π i
ln ζ + ∑ qk lk ln ζ k .
π k =m
(9.2.10)
Thus the speed potential and velocity can be found. Namely, the speed on the circle ζ = eiσ
333
L Γ 1 n σ −σk
is obtained by v (σ ) = − sin(σ + γ ) +
2
+ ∑
2π 2π k = m
qk lk cot
2
. (9.2.11)
Satisfying the Joukowski – Kutta condition at stagnation point, σ = 0 , one obtains the
n
σk
circulation Γ = π L sin γ + ∑ qk lk cot . (9.2.12)
k =m 2
For continuous distribution of the penetration of intensity q ( s ) , the sum in equations
(9.2.11) and (9.2.12) can be presented in integral form,
σn
n
σk ds σ
∑
k =m
q l
k k cot
2
→ ∫ q(s)
dσ
cot dσ .
2
(9.2.13)
σm
The integral for velocity distribution (9.2.11) in that case should be calculated as the
principal value. However, for numerical sense it is preferably to use equations with the sums.
The Cauchy singularity at point σ k can be eliminated excluding that term from the sum
L Γ 1 n
σi −σk
vi = −
2
sin(σ i + γ ) + +
2π 2π
∑q l
k =m
k k cot
2
. (9.2.14)
k ≠i
Returning to the hydrofoil on a physical plane, velocity at the foil boundary is found by
dϕ dϕ dσ d Φ
v( s ) = = , Hence, the velocity at nodal points is calculated by
ds dσ d Φ ds
2
Vi = −vi Vi . (9.2.15)
L sin σ i
9.2.4. Penetrable Hydrofoils near a Bottom

The problem, as above, can be solved using the following algorithm:
1) solving numerically a conformal mapping of the exterior of two non-penetrable
hydrofoils in Fig.9.1.5 onto the exterior of two parallel open cuts in Fig.9.2.2;
2) analytical conformal mapping of the exterior of two parallel open cuts onto the interior
of a rectangle;
3) analytical solution of a value problem in the rectangle.
The first step is the same as a numerical solution given in the previous section. The
horizontal cuts can be obtained integrating speed along the foil boundary or by summation
(9.2.5) with the only difference that the cuts are open, Fig.9.2.2a. Magnitudes of the stream
function on the sides of cuts are equal to c1 = VN +1 and c2 = VN + 2 so that the distance between
cuts is equal to d =| c1 − c2 | .
Now, the multi-sheet domain should be
a conformal map onto a rectangle with sides π
and π ti on the ζ -plane, Fig.9.2.2. Let the sides
of the cuts correspond to the rectangle’s
horizontal sides and its edges comply with
points A(a + π ti / 2) , B (π − a + π ti / 2) ,
A '(a − π ti / 2) , and B '(π − a − π ti / 2) ,
respectively; and infinite point D of the w-plane
Fig.9.2.2. w-plane (a); ζ -plane (b).
be in keeping with ζ=0 and ζ=π. Consider
334
derivative Wζ(ζ), which is pure real at the horizontal sides and has singularity as pole of
second order at points ζ=0and ζ=π. Extending through sides of the rectangle, one can be
sure that the derivative is an elliptic function so that it can be expressed by theta-functions as
Wζ = C[ln ′′ϑ4 (a) − ln ′′ϑ1 (ζ )] . (9.2.16)
Thus, the transformation is W = C[ζ ln ′′ϑ4 (a) − ln ′ϑ1 (ζ )] . (9.2.17)
Unknown parameters a, t, C are calculated from equations for length of the lower and upper
sides of the cut, L1 and L2 , and from the distance between two cuts (d):
L1 = C[2ln ′ϑ4 (a ) + (π − 2a) ln ′′ϑ4 (a )],
(9.2.18)
L2 = 2C[ln ′ϑ4 (a ) − a ln ′′ϑ4 (a )], d = C[π t ln ′′ϑ4 (a) + 2].
Function (9.2.17) together with the numerical approach for arbitrary foil presents the
desirable conformal map.
Considering an analytical solution of a value problem in the rectangle, one should
establish that the stream function on horizontal sides is constant outside the penetrable parts
and its tangent derivative is given on the penetrable parts Gj. Thus the derivative of complex
potential as singular function can be written as
1
2π G∫j
wζ = − Q(ξ )ln ′ϑ4 (ξ − ζ )d ξ + C0 + C1 ln ′ϑ1 (ζ ) + C2 ln ′′ϑ1 (ζ ) , (9.2.19)
where Q(ξ ) = q( s )ds / d ξ . Unknown coefficients C0 , C1 , C2 are obtained satisfying the

• function (9.2.19) should be double periodic with periods π and π ti , thus
1
C1 = ∫ Q(ξ )d ξ ; (9.2.20)
π Gj
• coefficient on the pole of second order can be calculated as

dw dz dW
C2 = − lim ζ 2 wζ (ζ ) = − lim lim lim ζ 2 =C; (9.2.21)
ζ →∞ z →∞ dz w→∞ dW ζ →∞ dζ
• function (9.2.19) should vanish at point B (condition at the trailing edge)
C0 = − ∫ Q (ξ ) ln ′ϑ1 (ξ + a)d ξ − C1 ln ′ϑ4 (a) − C2 ln ′′ϑ4 ( a) . (9.2.22)
Gj
Derivative (9.2.19) together with transformation (9.2.17) and the numerical solution
of flow problem of non-penetrable hydrofoil allows to calculate the penetrable foils both
near bottom and isolated. The tangent speed at nodal points at the boundary of hydrofoil are
wζ (ζ k )
calculated as before Vk = Vk . (9.2.23)
Wζ (ζ k )
The lift is calculated summing the pressure on finite elements. Namely, the lift coefficient is
2 L 1 N1 2
CL = = ∑Vk ( xk − xk −1 ) , (9.2.24)
l l k =1
where N1 is the number of finite elements of the hydrofoil; l is the chord of foil.
Numerical results are plotted in Figs.9.2.3 and 9.2.4. The former plot was calculated for a
single Joukowski’s foil, the second one for a foil near a bottom. It should be noted that Eq.
335
(9.2.15) for a proximity to the leading edge, σk ≈π, gives a small splash but it does not
virtually affect integral results if a number of elements is sufficiently large. E.g, the number
N = 500 gives for the circulation four exact digits after comma. The curves in Fig.9.2.3 have
logarithmic singularity at the ends of penetrated parts because intensity q ( s ) has a break of
first kind.
Fig.9.2.3.
Flow around a
Joukowski’s foil:
(a) - configuration
of hydrofoil;
distribution of the
tangent speed for
constant source
(b) and for a sink
(c). Solid lines -
penetrated part at
the upper
boundary (0.8L,
0.85L), dashed
curves – for
(0.15L, 0.2L).
Fig.9.2.4 shows the influence of penetrable part of hydrofoil near the bottom. The
hydrofoil is the same as that shown in Fig.9.2.3a but is turned to angle α = 10°. The distance
between the bottom and the trailing edge h =0.1. Solid lines show calculations for a non-
penetrable foil. Coefficients
CL∞ , CLh and CLq correspond
to a single foil, to a non-
penetrable and penetrable
foils near a bottom
respectively. Distributions
of source and sink on
interval ( sm , sn ) are given
parabolic with a maximum a) b)
unit value. As shown in
Fig. 8.2.4. The
Fig.9.2.4. TheJouckovski’s
Joukowski’shydrofoil
hydrofoil near bottom:
near bottom:
Fig.9.2.3 that the positive a source (a) and a(b)
sinka (b).
effect of penetrable part (a) a source; sink.
exists only if it is located near the trailing edge, but a source should be at the lower side and
a sink at the upper side. The same results were obtained for a foil near a bottom, Fig.9.2.4. A
bottom for this hydrofoil decreases the lift but it can be increased by sucking or blowing-out
fluid into mainstream.
336
Sec. 9.3. Numerical Modeling of Cavitating Flow

9.3.1. Preliminary Notes
Numerical modeling of cavitating flows is a complicated problem in the hydrodynamics
since the free boundary and cavity termination are a priori unknown, and they can hardly be
simulated numerically due to single spirals at the ends of cavity boundaries. This is a result
of the multi-sheet flow at the cavity end, which is difficult to calculate directly in a physical
plane It is possible to apply analytical methods including conformal map, but analytical
formulas and systems of nonlinear equations are not only too complex for any body, but also
too difficult for numerical calculations. Therefore, many scientists have focused their
attention on direct numerical modeling of a physical plane, or an axial plane, for
axisymmetric flows. Riabouchinsky model has mostly been used as it gives almost the same
results as theoretical models with single spirals for a wide range of cavitation numbers.
Instead of a flow domain with unknown cavity boundary it is possible to consider a
domain with known boundaries for function y ( x,ψ ) that is determined in the plane ( x,ψ )
with a horizontal cut [Kuznetsov et al., 1967; Zuykov & Shepelenko, 1967; Shepelenko,
1968]. Indeed, the desirable function should satisfy the most complicated differential
equation, which can be calculated numerically using the finite difference method. The above
authors made computations for some plane and axisymmetric problems of cavitating flows.
More details of this numerical analysis can be found in [Shepelenko, 1968]. At almost the
same time, Bloch [1969] used variation principles and the method of steepest descent to
calculate a cavitating flow around a disk in the auxiliary w -plane. However, the method
converges rather slowly. Snider [1971] proposed a faster procedure of an iterative scheme to
handle nonlinear boundary conditions from Lewy’s fundamental reflection rules. Finally,
Garabedian [1971] in collaboration with Bloch, Frieman and Snider developed a finite
difference scheme for solving numerically the stream function expressed in a conformal
invariant form mapping some rectangle in a w -plane onto one quadrant of the flow domain.
In this way all singularities were taken into account and the numerical results for a disk were
obtained with high accuracy. Some numerical results for a cavitating sphere in an
axisymmetric water tunnel are calculated in [Brennen, 1969]. It should be noted that all these
solutions overcome the difficulty of unknown cavity shape though the equations and their
numerical computing are rather complicated.
Another way of numerical study of plane and axisymmetric cavitating flows is
boundary integral equations, which can be obtained from continuously distributed ring-
shaped sources or vortexes, or from generalized Green’s identities. The boundary sources
layer was used to calculate axisymmetric cavities in [Krylov, 1963; Terentiev, 1994] while
the ring-shaped vortexes in [Ivanov, 1980; Kojouro, 1980; Todorashko 1977]. Boundary
element methods based on Green’s integral identity for potential were used in [Krasnov &
Kuznetsov, 1989; Yas’ko, 1993; Kinnas, 1998, Kirschner et al., 2001; Vaz, 2005]. Generally
it is not difficult to obtain the appropriate boundary integral equations but solving those
equations is rather complicated, particularly, to satisfy the conditions at cavity boundary.
One of the methods is satisfying simultaneously both kinematic and dynamic
conditions. This method reduces a flow problem to a set of non-linear equations, Sec. 4.5.
Because of complicated equations the number of collocation points should be not too large.
This method has been successfully used to investigate cavitating flow of a cone [Guzevsky,
1975, 1979]. He divided the wetted and cavity boundaries into finite numbers of nodes and
337
approximated unknown functions by cubic splines. Satisfying boundary conditions and
behaviors of unknown functions near the stagnation point and detached points where the free
surface of cavity separates from the cone, he obtained a set of non-linear equations, which
was solved numerically. Despite small numbers of nodes (five points on a wedge and six
points on a cavity) the numerical results, obtained for the cavitated cone are close to the
exact values. He used the Riabouchinsky model with a mirror trailing cone so that the results
might differ from the expected theoretical results for a small cone angle. Anyway, the
results received later have been coordinated to Guzevsky’s results. Below, the direct iterative
method is considered as the simplest way of calculation [Terentiev, 2005].
9.3.2. Direct Iterative Method
Numerical methods can be based on several techniques, namely, on auxiliary
transformations, additional functions, and so on. Here all mathematical operations are made
on physical space only, thus calling the method “direct” is correct.
Let the inlet velocity be unity and directed towards the x-axis. Turning back to
equations (8.1.16), (8.1.36) and (8.1.38), all these equations and constant c can be combined
to a single matrix equation Q(Y) = B( X, Y)V . (9.3.1)
Here matrix B is written as a function of nodal coordinates X k , Yk . The latter equation puts
some idea to find vector Y in iterative manner for given speed distribution V,
Q( Y ( n ) ) = B( X( n −1) , Y ( n −1) )V . (9.3.2)
Designing a body by the given speed distribution constitutes inverse problems of the
hydrodynamics. The theory of inverse value problems has been widely disscussed in
[Tumashev & Nuzhin, 1965; Elisarov et al., 1994]. The main problem in the theory is still
giving a speed distribution; contour C can generally be self-intersecting. No condition is
available for a non self-intersecting line. Nor there is a mathematical proof of convergence
of iteration (9.3.2), though numerous examples show a good convergence if the speed
distribution corresponds to a body without self-intersecting boundary.
9.3.3. Design of Ellipsoid of Revolution
Analytical distribution of speed on an ellipsoid
of revolution x + y / b = 1 is given by Eq.
2 2 2
(5.7.3). For example, the design of an ellipsoid

with aspect ratio ε = 0.5 is considered,
Fig.9.3.1.
Fig.9.3.1.
Iterations for recovering an ellipsoid design
from its specified analytical velocity
distribution.
The initial shape was selected as a

cylinder of unit radius (shown as the chain
line). The number of elements was N = 30. Convergence is good for any initial shape, but it
becomes slower for a slender ellipsoid.
Prima facie, the Green’s identity seems to be invalid for the initial and following
shapes because the boundary is not closed. This is because the open domain on a sheet plane
may be represented as a closed domain on a multi-sheet Riemann plane, or by conformal
338
map on a parametric plane. Since the Green’s identity can be represented as Couchi integral,
which is valid on a parametric plane too, then using Green’s equation is valid. The velocity
on edges should be infinite; this behavior of velocity is shown by fine lines in Fig.9.3.1.
The direct iterative method is preferable because the interim shape and speed
distribution over it are calculated at each iteration. It is particularly effective when a given
velocity distribution corresponds to a self-intersecting line. An iterative procedure is possible
for positive ordinates only and, if one of them takes a negative value, the calculation stops.
Then the preceding ordinates and the speed distribution can be considered as final results,
although the body will differ from the desirable one, i.e. the iterative method makes it
possible to control the speed distribution during computing. The speed distribution control
problem was discussed analytically in [Elisarov et al., 1994] for 2D flows. The body can be
also non-closed in a physical space. One of them is presented below.
9.3.4. Cavity Determination
The cavity can be considered as a body of
constant velocity distribution ( V = 1 + σ ).
0
Abscissas xk (and then X k ) are fixed on

segment [-1,1] while ordinates Yk are calculated
from (9.3.2) iteratively, and then function y(x) is
approximated and ordinates yk are found at
each iterative step and, simultaneously, the
speed distribution is computed. Iterations for the
axisymmetric cavity at cavitation number Fig.9.3.2. Determination of the
σ = 0.5 using N=100 boundary elements are axisymmetric cavity shape.
shown in Fig.9.3.2. The points on the horizontal
segment are shown over each five nodes. Theoretically, the cavity boundary should be ended
by a single spiral, but it is impossible by this procedure. This feature is a cause of speed
fluctuation at the cavity ends, but the fluctuation has not much influence upon overall flow
characteristics and can be ignored. As expected, the cavity boundary is not closed.
A comparison of the present results with numerical data obtained in [Guzevsky,
1979] and experimental data [Egorov et al., 1971] is shown in Fig.9.3.3. The good
comparison of the numerical results with both of these data sets can be seen, even though
Guzevsky calculated the cavity
with two disks (that is, with the
Riabouchinsky cavity closure
model).
Fig.9.3.3.
Comparison of cavity slenderness
ratios calculated using the current
method (solid line) with
Guzevsky’s results (dashed line)
and experimental data (points).
339
9.3.5. Direct Iteration for Cavitating Flow around a Symmetric Body

In the case of cavitating flow around a foil, there are two types of boundaries: one is the
given boundary of a body Cb , another one is cavity boundary C0 . The finite elements can be
chosen by a fixed x-abscissa, or by a fixed curvilinear s-abscissa. The latter is preferable
particularly for the foil boundary if the boundary has a vertical tangent. Further, the cavity
length Lc will be fixed instead of cavitation number. The numbers of elements on the fixed
boundary and cavity can be different. Besides, the cavity can be simulated by Riabouchinsky
model, then a third boundary appears whose shape is given but the length is unknown and it
must be attached to the cavity boundary. Usually, the trailing shape is chosen as a straight
line for planar, and a disk for axisymmetric, flow. Since the trailing flow has a little effect on
total characteristics, the number of elements on the trailing shapes can be small. Denote the
number of elements on a body and on a cavity boundary as N b , N c , respectively. The initial
shape of cavity can be given as a horizontal straight line ykc = yNb b = const.
The iterative procedure for cavitating flow is divided into two stages. The first stage,
associated with speed distribution, is arranged as follows:
1) calculate the speed for a given boundary C ∈ C ∪ C from (9.3.2);
( n −1)
b c
( n −1)
(n) (n)
2) obtain the average speed Vc = ∑ Vk lk ∑ lk on the free boundary;
k k
3) correct the speed on the cavity boundary C by V ( n − 1)
c
=V = constant , while on the
( n −1)
0
∗ ( n −1)
wetted boundary C the calculated speed is retained.

b
The second stage, associated with the boundary, consists of the following steps:
1) update the ordinates of the control points Y using equation (9.3.2);
∗( n )
2) update the vector of boundary element endpoints y using any appropriate

∗( n )
interpolating functions;
3) correct the wetted boundary by given C and the cavity boundary by moving
b
ordinates yk∗( n ) , in order to fulfill a continuity of boundaries C and Сс at the attached

b
th
point, the n iteration is determined by the ordinates of the cavity nocks, y = y . (n) ∗( n )
The iterative procedure can be introduced in the following sequence:

Y ( n −1) ⎯⎯⎯(8.3.1)
→ V *( n ) ⎯⎯⎯⎯
correcting
→ V ( n ) ⎯⎯⎯
(8.3.2)
→ Y*( n ) ⎯⎯⎯⎯
correcting
→ Y(n) . (9.3.4)
The iteration is terminated when a desired measure of convergence is achieved, for example
|1− σ /σ |< ε . The initial boundary of a cavity may be given by any curve, for example
(n) ( n −1)
by horizontal straight lines.

9.3.6. Comparison of the Current Method with Others
Consider a cavity past a wedge of half vertex angle β = π / 6 shown in Fig.9.3.4. The initial
free boundary is given as a horizontal straight line. The wedge side l = 1 and cavity length
Lc = 5 are divided into a free boundary of 40 elements, and the aft line is divided into 200
elements. The finite elements are given as a geometric series from the attached point to the
middle of the free boundary and as a converged series for another half of boundary. The
initial trailing line is equal to l sin β and is divided into 20 elements. Calculations show a
relatively high convergence of iterations. The points correspond to exact analytical solution
that agree very well with numerical curve. Cavitation number is calculated at each stage of
340
iteration due to the average speed, therefore the analytical results (points) differ at each
iteration. The final relative tolerance is 1 − σ (14) / σ (15) = 0.00009 ; the tolerance for lift
coefficient CL relative to
analytical lift coefficient
CL ,analyt . is equal to 0.0027.
Fig.9.3.4.
Iteration convergence for the
initial cavity (horizontal line);
points correspond to analytical
calculations using interim
values of cavitation number.
Figs.9.3.5 - 9.3.7 are shown mainly for comparing the presented method with others
for cavitating flow of a disk. Convergence is shown in Fig.9.3.5. The initial free boundary is
plotted as a straight cylinder of length Lc = 8 . Calculated speed distributions for the disk are
plotted in Fig.9.3.5a by dashed line (1st stage), and solid lines correspond to the corrected
velocity (2nd stage). As expected, the calculated velocity increases to infinity at the
cylinder’s edges and approaches iteratively a finite value. The numbers of elements on the
front disk, on the cavity surface and on the trailing disk are 50, 100 and 20, respectively. For
comparison it was also calculated for the number of elements at the cavity boundary,
N с = 200 . The results for tolerance ε = 10−4 are:
N b = 50, N c = 100, Ntr = 20 , Lc = 8 : σ = 0.355 , CD = 1.145 , H c = 1.963
N b = 50, N c = 200, N tr = 20 , Lc = 8 : σ = 0.335 , CD = 1.121 , H c = 1.977
Fig.9.3.6 shows a comparison of current results (solid curves) with experimental
data and with the approximate Reichardt formula from [Knapp et al., 1970]. Here H is the c
maximum diameter of the cavity; d is the diameter of disk or base of cone cavitator, and L c
is the cavity length. Approximated formula gives results slightly lower than the iterative
manner gives but acceptable for practical
purposes. The experimental results for the
cavity maximum diameter are higher than
the theoretical and approximated curves.
The difference can be due to measurement
errors, but can also mean that the cavitation
models used might not necessarily simulate
the real process accurately.
Fig.9.3.5.
Cavitating flow of a disk: calculated (dashed
lines) and corrected (solid lines) distribution
of a speed (a); convergence of cavity shapes
by initial straight cylinder (b).
341
Fig.9.3.6 Comparison
of the calculation
results with
experimental
data [Knapp et al.,
1970] for the disk (a)
and the cone (b);
dashed line
corresponds to
Reichardt’s
approximation.
A comparison of computations by the direct iterative method and by other methods

used in [Guzevsky, 1979; Ivanov, 1980] is presented in Fig.9.3.7. The left plot shows the
velocity distribution over the cone generatrix; the right plot – dependence of the maximum
diameter of cavity on cavitation number.
Fig.9.3.7.
Comparison of
present method
(solid lines) with
Guzevsky (points)
and Ivanov (dotted
lines) results:
(a) - velocity
distribution over a
cone
(b) - cavity
thickness vs.
cavitation number.
9.3.7. Conclusion
The direct iterative procedure can be used for computing with high accuracy a cavitating
flow of arbitrary foils. The results agree well with all known data on cavitating flow
[Brennen, 1969; Garabedian, 1971; Guzevsky, 1979; Wrobel, 1991; and others]. But
iterative procedure converges rather slowly for a large cavity (or for small cavitation
number). In this case the Serebryakov asymptotic theory of axisymmetric body can be used.
Sec. 9.4. Non-Symmetric Cavitating Flow

9.4.1. Partial Cavitation
A partial cavity has only one free boundary, thus the above-mentioned numerical algorithm
is valid but the Joukowski – Kutta condition (8.1.18) must be satisfied, i.e. matrix equation
(8.1.19) and its inverse should be used. If inlet velocity is unity and directed towards the x-
axis, then matrix Z = Y and computing of each cycle should be made as follows:
342
∗( n ) −1 ( n −1) (n) ∗( n ) ∗( n ) (n) Y ∗( n ) =Y
Vk =Vc
Y ( n −1) ⎯⎯⎯⎯⎯
V =B Y
→ V ∗ ⎯⎯⎯⎯
⎯ = BV
→ V ( n ) ⎯⎯⎯⎯⎯
Y
→ Y∗( n ) ⎯⎯⎯⎯
j j
→ Y(n) (9.4.1)
where subscripts k and j correspond to the cavity and wetted foil boundary, respectively. The
initial cavity boundary can be considered a part of the foil or any curve above it. As before,
the cavity length and the detachment point are supposed to be given. Generally, the first
detachment point should be found from the Brillouin condition. Some notes on finding the
detachment point will be discussed below. The partial cavitating flow can be computed both
as a value problem statement (Sec. 2.1) or using the Riabouchinski model; the latter is
preferable for a simple and closed cavity boundary. As shown above, both models give
virtually the same results.
9.4.2. Convergence and Accuracy
Convergence and accuracy of iterations can be shown for a cavitating flow of plate using the
Riabouchinski model’s analytical solution (Sec. 2.4). It is shown in Fig.9.4.1 for the test
problem of flow around the plate with
partial cavity. The plate was simulated as
two close plates with spacing between them
0.001. The both sides of the plate, [0, 1],
were divided into 100 elements and only the
40th element corresponded to the trailing
plate of the cavity. Fig.9.4.2 shows a good
agreement of results obtained by the current
method and using speed potential [Dang &
Kuiper, 1998].
Fig.9.4.1.
Comparison of numerical results with an
analytical solution for flat plate with partial
cavity: points are obtained analytically.
Fig.9.4.2.
Comparison of the present method (dots) with
numerical calculations by Dang & Kuiper (solid
line).
9.4.3. Using the Single Spiral Model for Partially Cavitating Flow
Plotted in Fig.9.4.3 are the calculations results for a nonlinear problem of flow around a
plate at angle of attack α = 5 and cavity length Lc=0.4L=0. The cavity configuration is
shown in Fig.9.4.3a, and in a larger scale in Fig.9.4.3b near the leading edge, and in
Fig.9.4.3c – at the cavity end. A flow at the cavity end takes place on infinite-sheet planes,
but speeds on these sheets are virtually zeros, therefore one of the elements can represent a
curve connecting the spiral center with a point on the plate inside the cavity. As the speed on
this curve is practically equal to zero (Fig.9.4.3d) it cannot be taken into account in
numerical calculations. Fig.9.4.3b shows that the cavity boundary’s ambiguity near the fore
343
edge (b) and at the cavity end (c) takes place only in very small vicinities which can be
ignored in calculations. The closure condition (2.1.6d) is satisfied automatically because the
Green’s identity is valid only for closed boundaries. Here the cavity closure takes place on a
multi-sheet plane though at the sheet in question the cavity is not closed.
Fig.9.4.3. Flow around a plate: (a) configuration of cavity; (b) cavity boundary near the
leading edge; (c) spiral at the cavity end; (d) speed distribution over the plate and cavity;
solid lines – the exact solution, points – numerical solution at N=400.
Fig.9.4.3d shows speed distribution over the plate and cavity; at the lower side of
plate the mark of speed is changed. Almost perfect coincidence of analytical and numerical
results is observed. The cavitation number and lift coefficient calculated analytically and
numerically are σ = 1.2388 , CL = 0.6561 and σ = 1.2390 , CL = 0.6546 , respectively.
9.4.4. Point of Smooth Cavity Attachment

If there is a corner point on a foil where the flow speed attains infinity and, hence, pressure
reaches a negative infinite value, the separation point has to coincide with the given corner
point. If the body has a smooth configuration, the attachment point is unknown a priori and
should be defined from additional conditions. Usually satisfied in the theory of cavitating
flow is the Brillouin condition about equality of curvatures of foil boundary and cavity
boundary. However, this condition has a local character and doesn't guarantee univalence of
flow, i.e. the cavity boundary can cross the foil boundary with the Brillouin condition
satisfied. Besides satisfying the Brillouin condition, it is necessary to calculate the first and
second derivatives, what is difficult to do numerically with high accuracy. Therefore the
Brillouin condition is replaced with a principle of positivity of difference of ordinates, which
is written as follows: min( yc ( x) − y ( x)) ≥ 0, min( yc ( x) − y ( x)) < 0, (9.4.2)
x∈γ c x∈δγ c
where γ c and δγ c are the cavity boundary and its variation. The second condition in (9.4.2)
means that when the cavity border varies, the first inequality is not satisfied at least at one
point of the interval.
This principle can easily be realized by a method of consecutive approach to the
attachment point at the left. The method is similar to the method of dividing a piece in halves
344
that can be easily realized by computer modeling. If cavity length is zero, Lc = 0 , the
attachment point should coincide with point am of the maximum speed of continuous flow.
For real foils this point is near the maximum curvature of structure, and the attachment point
should be between it and the critical point. Choosing between them the initial separation
point a (0) = (am + a0 ) / 2 , at which the first condition (9.4.2) is not satisfied, one can find the
following approximation a (1) = ( am + a (0) ) / 2 . If only first inequality of (9.4.2) at point a (1)
is satisfied then a (2) = ( a (1) + a (0) ) / 2 ; if both conditions (9.4.2) are satisfied, then
a (2) = (am + a (1) ) / 2 . This process continues until condition | a ( j ) − a ( j −1) |< δ is satisfied.
Results of calculations for the Joukowski foil are shown in Fig.9.4.4. The left figure
shows the open-end cavity with a smooth attachment point marked by dashed line, while the
right plot shows basic characteristics of cavitated foil. Shown in Fig.9.4.4b is the speed
distribution on the foil; the origin of curvilinear coordinate coincides with the trailing edge,
perimeter of foil p = 2.059. The horizontal line in Fig.9.4.4b corresponds to the cavity. At
the right-hand side, a decrease of speed is observed at approaching the cavity.
Fig.9.4.4. (a) The Joukowski foil and attached cavity; (b) speed distribution on foil and
cavity.
Plotted in Fig.9.4.5a is abscissa b of the smooth attachment point vs. cavity length Lc ,
while cavitation number σ and lift coefficient CL are plotted in Fig.9.4.5b.
Fig.9.4.5.
(a) Abscissa of the
smooth attachment
point versus cavity
length Lc ;
(b) cavitation number
and lift coefficient
versus cavity length.
The plot was calculated for the attachment point fixed at the point corresponding to the
maximum speed of continuous flow. These points closely approximate smooth attachment,
so the attachment point can be fixed at different cavity lengths.
345
9.4.5. Two Cavities on a Foil

The direct iteration method is easy to
apply when calculating complicated
cases. Two cavities on a foil are
shown in Fig.9.4.6. The cavities in
Fig.9.4.6a have the same cavitation
number of σ=0.680, while in
Fig.9.4.6b they are different (σ =
0.987, σ1 =−0.316) for the upper and
lower cavities, respectively). The
detachment and trailing points are
marked by circles.
Fig.9.4.6.
Two cavities on a foil with the same (a) and different (b) cavitation numbers.
9.4.6. Two Foils
The direct iteration can be applied for a numerical study of cavitating flow around a system
of a finite number of arbitrary bodies and foils. Some numerical results are plotted in
Figs.9.4.7 and 9.4.8. Two equal foils with partial cavities, one above another, are shown in
Fig.9.4.7. The lengths of cavities on both foils are the same but the cavitation numbers are
different. The hydrodynamic characteristics of both foils are presented in Table 9.4.1.
Fig.9.4.7 shows two foils located one after another. Their attack angles are α1 = 10
Table 9.4.1. Hydrodynamic values.
Upper foil
Lc σ CL CD ψ
0.191 3.454 0.957 0.033 0.110
0.449 2.642 0.756 0.058 0.112
0.722 2.370 0.681 0.108 0.104
Lower foil
Lc σ CL - CD -ψ
0.191 2.206 1.243 0.030 0.072
Fig.9.4.7. One above another foils.
0.449 2.028 1.525 0.055 0.069
0.722 1.891 1.872 0.108 0.007
and α 2 = 20 at the fore and rear foils, respectively. Calculations show that the partial cavity
on the rear foil cannot grow if the angle of attack α 2 ≤ α1 . Interaction of two or more foils is
illustrated in Fig.9.4.8. Many foils could be calculated similarly by direct iteration.
Fig.9.4.8.
Foils: one
behind the
other.
346

In case of full cavitation (cavity length Lc is greater than foil cord l), there are two cavity
boundaries and both should be attached to the plate edges – both upper and lower boundaries
must be adjusted at each iteration cycle. Therefore, to be found at each cycle is either the
length of both boundaries for the cavitating model with a single spiral, or the trailing plate
length for the Riabouchinsky model. As stated above, the Riabouchinsky model is more
suitable for fully cavitating flow. The end points of cavity boundaries must be on the same
perpendicular plate; since its length is unknown a’ priori, it has to be calculated iteratively.
For simplicity, the termination plate can be approximate by one element with a stagnation
point at the middle. The last condition should be considered as the additional one instead of
the Joukowski-Kutta condition. It simulates approximately both models: the one with a
termination plate and the other with a single spiral cavity. Comparison between numerical
calculations and analytical solution is shown in Fig.9.4.9. The full cavity past a curvilinear
(Joukowski’s) foil is shown in Fig.9.4.10.
Fig.9.4.9.
Comparison of
numerical
calculations (solid
lines) and
analytical results
(points).
Fig.9.4.10.
Fully cavitating
curvilinear foil.
9.4.8. Ring Wing with Cavitation

The numerical algorithm for computing cavitating flow of a
ring wing is similar to that for the above-considered planar
flow. Some numerical results are plotted in Figs.9.4.11 -
9.4.12. Fig.9.4.11 shows the partial cavities for three cavity
lengths Lc = 0.251, 0.526, 0.791 . Dependence of cavitation
number on length is similar to that for a single foil: with
length increasing, cavitation number decreases and then
increases. An unexpected and very interesting effect is
observed: fluid consumption increases with cavity extension.
Fig.9.4.11.
Ring wing with partial cavities
347
The cavity shape past a ring wing obtained by an inclined plate is shown in
Fig.9.4.12. The trailing ring was divided into
two rings (elements). The ratio of its width
should be correlated with the cavitation
number on both cavities. In fact, the cavity
with different speeds on the boundaries may be
calculated by variation of that ratio. It is
remarkable, the inside cavity surface is similar
to a jet stream.
Fig.9.4.12.
Full cavity past a ring wing.
9.4.9. Cavitating Flow around a Foil with an Interceptor

An interceptor can increase lift. The theory of interceptor was developed by Rozhdestvensky
and his followers [Rozdestvensky & Fridman, 2001; Fridman & Uryadov, 2004]. Some
results of numerical calculations of using the direct iterative method for cavitating flow
around a foil with a spoiler are shown in Figs.9.4.13 – 9.4.15. Fig.9.4.13 shows dependence
of cavity size and
cavitation number on
stagnation point
location on the
trailing edge and the
spoiler plate.
Fig.9.4.13. Dependences of cavity size on stagnation point location at cavity length Lc=0.4:
(1) without spoiler ( σ = 1.37, CL = 1.20 ); (2) stagnation point at the middle of spoiler plate
( σ = 2.04, CL = 1.88 ); (3) stagnation point at the spoiler edge ( σ = 3.95, CL = 3.27 ).
Different cavity lengths Lc=0.1, 0.2, 0.4 are considered in Fig.9.4.14. The same foil
without a spoiler has lift coefficient CL = 1.15 . It is shown that the lift decreases while the
length increases. The same foil with two kinds of full cavities is shown in Fig.9.4.15.
Fig.9.4.14. Different cavities past the spoiler plate.
348
Fig.9.4.15. Two sorts of cavitating flow of the same foil.
Sec. 9.5. Numerical Modeling of Perturbed Flows

The domain of a bounded flow usually contains two or more points at infinity that
complicate a numerical model. These points at infinity are often excluded by considering a
truncated domain: new boundaries where the conditions are satisfied are introduced, that
conforms the known or assumed behavior of the solution at infinity. The above-considered
numerical method is based on generalized Green’s identity (see App. 3), which includes the
behavior of a stream at infinity, though the method is universal for planar and axisymmetric
problems of steady flow. Numerous examples presented in previous sections verify this. The
direct iteration method considerably expands the BEM capability and allows computing the
cavitating flows past a system of foils and cavities [Terentiev, 2000; Pavlova &, Terentiev
2010]. Approximate modeling of bounded flows perturbed by a system of bodies is
presented here. A majority of problems discussed below were computed by Pavlova.
9.5.1. Perturbed Flow around a Foil
Governing equations have been presented in Sec. 7.4. Namely, the complex speed at infinity
can have asymptotic expansion (7.4.2), thus the generated Green’s equation for stream
function is transformed to ∫ v(s)G (s, sz )ds + c = Im F ( z ) ,
C
(9.5.1)
where F ( z ) = a0 + a1 z + a2 z 2 + … + an z n , z ∈ C . (9.5.2)
Further, the current numerical method can be applied to
computing this equation. E.g. Joukowski hydrofoil near a
circular cylinder is considered in Fig.7.4.1. Analytical
solution is given in Sec. 7.4. Fig.9.5.1 shows a comparison
of speed distribution obtained analytically (solid curve) with
Fig.9.5.1.
Speed distribution on a Joukowski foil near a circular
cylinder.
349
that given by equations (7.4.10) – (7.4.14) and numerically by equation (9.5.1) together with
the Joukowski – Kutta condition (points). The arc length in the x-axis is made dimensionless
by relating it to perimeter L=4.141. The chord of foil in this example is equal to l=2.017.
The problem was calculated without simplifications using generated Greens identity to the
system of a circle and a small foil. The calculations coincide with analytical results very
well. Comparisons show that the approximate method gives close results even for R > 10
and R0 > 15 , i.e. ε ≈ 0.2 and δ ≈ 0.4 . The coincidence of curves in Fig.9.5.1 also highlights
the efficiency of the numerical method (BEM) and the generalized Green’s identity. The last
conclusion and numerous other examples signify that the current numerical method provides
high accuracy for both small and large bodies of a system. This allows simulating various
flows, including channels and jets, free boundaries and bottom, gravity, and others. All these
problems can be computed using the same numerical algorithm.
9.5.2. Perturbation Flow between Two Bodies
To study the flow between two bodies, let’s consider two bodies C2 and C3 of different
configurations. The one in Fig.9.5.2 simulates a flow in a channel with converging straight
boundaries (shaded area); the other in Fig.9.5.2b simulates a flow between two concentric
circles. Considering the streamline patterns, these flows can be represented as those due to a
point sink and a point vortex, respectively.
Fig.9.5.2.
Flow around two bodies which simulate:
(a) slow converging channel;
(b) channel with circular sides;
(c) relative errors between the flows predicted
using the perturbation method and the
corresponding analytical solution for a point sink
(a) and a point vortex (b), respectively.
For the first case, the mass flow rate through any section of a channel with slowly
converging boundaries is equal to VH where H is the channel width and V is the average
flow speed normal to the section. Alternatively, the amount can be found as the difference
between two values of the stream function, q =| c2 − c3 | . Hence, agreement between the flow
350
computed using the perturbation method and the actual flow in a channel is characterized by
ε = (VH − q) / q . Relative error ε for the shaded area in Fig.9.5.2a is plotted for different
values of channel width h, as the family of curves labeled (a) in Fig.9.5.2c. The flow
between the arcs of two concentric circles, Fig.9.5.2b, corresponds to that of a point vortex,
with agreement between the computed and actual flows characterized by ε = V3 / V2 − R2 / R3 .
The relative error for the shaded area in Fig.9.5.2b is plotted for different values of channel
width h = R3 − R2 , as the family of curves labeled (b) in Fig.9.5.2c. The agreement between
the perturbed and exact analytical flows is good for both a converging channel (point sink)
and a channel formed by concentric arcs (point vertex).
Now, any relatively small body in the shaded area in the plots simulates a flow
around such a body in a channel. Hence, the flow in a channel can be reduced to a flow
problem of three bodies in unbounded region which can be calculated using BEM.
9.5.3. Hydrofoil in a Channel
Consider the flow around a foil between two large bodies featuring inner flat boundaries,
Fig.9.5.3a. The channel width is denoted by H and the distance between the trailing edge and
the lower channel boundary by h. The length of the surrounding bodies is 10 times greater
than the unit chord length of foil. The circulation around large bodies is assumed to be zero.
Results for the Joukowski foil in a channel of width H=1, and offset h=0.2 are plotted in
Fig.9.5.3b and c. The inflow speed in the channel is obtained via the flux of liquid between
the walls, V∞ = q / H . In this case V∞ = 1.075 . Dividing all resulting speeds by V∞
normalizes all results to a unit inflow speed through the channel.
Fig.9.5.3.
Simulation of flow around the
Joukowski’s foil in a channel:
(a) three bodies shown to
scale ;
(b) normalized velocity
distribution on the foil; (c)
normalized velocity
distribution on the straight
boundaries of the channel (the
dashed line corresponds to the
lower boundary).
Lift coefficient CL , coefficient of hydrodynamic moment relative to trailing edge

CM and abscissa of reaction point xc are plotted in Fig.9.5.4 as functions of trailing edge
offset from lower boundary h. Since both CM and xc = CM / CL are negative, their absolute
values are plotted. For clarity, the abscissa of the reaction point is multiplied by 5.
One can see that the lift coefficient first slightly decreases then increases very
rapidly. This implies that the channel’s upper boundary has a greater effect than the lower
351
one. On the whole, the lift coefficient in a channel is
always greater than that without it: the lift coefficient of
this foil in an unbounded domain is equal to CL = 1.148 .
Fig.9.5.4.
Influence of channel on hydrodynamic characteristics of
foil
9.5.4. Foil between a Free Surface and a Wall

Let us consider three bodies in Fig.9.5.3a, but now
assume that the lower segment of the upper body is a free
surface where speed is constant though unknown. The initial step is: speed distribution is
calculated for the given boundaries exactly as was described in the preceding section. Next,
average speed is calculated at the free boundary. Then, retaining the average speed at the
free boundary and the computed speed distribution at the given boundaries, new boundary
coordinates are calculated at the free surface. The iterative process is then repeated. When
speed at the free boundary converges to a constant, this value corresponds to the inlet
velocity, which is an estimate of the free stream velocity far ahead of the foil. By treating the
foil-bottom distance and free surface as parameters, the ideal flow past a foil near either a
free surface or a wall alone can be estimated to any desired accuracy by specifying a
sufficiently large value for the other distance. An analytical solution of the problem of flow
past a flat plate near a bottom has been presented in Sec. 3.7. Numerical results obtained by
Pavlova are plotted in Figs.9.5.5 – 9.5.7. A comparison of numerical results with the
analytical solution is presented in Fig.9.5.5. The foil’s angle of attack is α = 10 . Fig.9.5.6
shows a comparison of numerical and analytical results for free boundary configurations for
different plate–to–free boundary distances.
Fig.95.5. Comparison of numerical ( ) Fig.9.5.6. Comparison of numerical ( ) and

and analytical ( – ) results for a plate analytical ( – ) results for a plate near a free
near a wall. boundary.
Analytical results are presented in Sec. 3.5 above. Velocity V1 at the free boundary is
constant but different for each distance h, so this distance is defined, as before, as
h = (ψ 1 − ψ 0 ) / V1 . Free boundary configurations and a sheet cavity originating at the leading
edge of a hydrofoil as a function of depth below free surface are plotted in Fig.9.5.7.
Calculations were performed for distance d between the trailing edge and the horizontal line
352
of the upper body changing as follows: d=1, 0.8, 0.6, 0.5, 0.4, 0.3 corresponding to h =
0.86, 0.68, 0.48, 0.37, 0.26, 0.13, respectively, at angle of attack α = 15 . The cavity length
was lc = 0.4 . Dashed lines correspond to the free surface configuration for the same foil
without a cavity. These numerical calculations show that partial cavity affects free boundary
only locally. The cavity shapes shown in Fig.9.5.7 correspond to h=0.13.
Fig.9.5.7.
Joukowski foil near
free boundary, with
and without a
cavity.
9.5.5. Simulation of Axisymmetric Flow in a Pipe or in a Jet

The flow around the ellipsoid of revolution in a pipe, or in a jet can be calculated by the
same algorithms, Fig.9.5.8a; the area of flow in a pipe is shadowed. The speed distributions
on inner pipe and on ellipsoid are plotted in Fig.9.5.8b and c, respectively. Three diameters
of the pipe have been calculated: D = 1.5, 2 and 100. The ellipsoid of revolution is
determined by
equation of axial
section,
x 2 + y 2 / 0.52 = 1 .
Fig.9.5.8.
Simulation of flow
around ellipsoid in a
pipe. Points –
analytical solution by
eq. (9.3.3).
Jet flow can be simulated iteratively considering a ring with free surface inside a
cylinder. Instead of the Joukowski–Kutta condition at the trailing edge, circulation around
foil is assumed to be zero. The jet flow around the same ellipsoid is plotted in Fig.9.5.8 by
broken line. The effect of jet is moderate, so the results are givens for jet diameter D = 1.5
only.
353
9.5.6. Conclusions
The presented examples demonstrate the great potentials for predicting perturbed flow by the
developed numerical methods, including direct iteration and the BEM approximation. It is
noteworthy that the same numerical algorithm has been used for all of the problems
presented. It should also be noted that there are some mathematical problems that should be
addressed involving analytical error estimates.
Sec. 9.6. Numerical-Analytical Method for Arbitrarily-Shaped Foils

9.6.1. Introductory Notes
Although the numerical methods presented in previous Sections can be applied to numerous
problems of cavitating flows, it appears important to develop other methods, including
analytical. The method of power series expansion presented in Chapter 4 (Levi-Civita
method) is not effective foils accommodating highly curved surface areas near the leading
edge. Solving problems for high-curvature foils requires finding a great number of
coefficients due to a poor convergence of Fourier series. Errors due to highly curved surface
obstacles are analyzed in [Birkhoff et al., 1953]. Maklakov [1997] offered an alternative
approach for numerical studies of those problems. He reduced the problem to a Villat-type
equation which is solved by means of discretization and matrix approximation of linear
integral operators. A similar discretization technique was used for computing a free-surface
flow obstructed by arbitrary bed topography [King & Bloor, 1990].
Cavity flows over hydrofoils using the Wu cavity model and applying the Levi-
Civita method were computed in [Konhauser, 1984]. To investigate the symmetric profile
NACA 0012, Konhauser retained 64 coefficients in the corresponding Fourier series. But the
Wu cavity model did not allow him to define the cavity length; therefore the Konhauser
computation did not give a full picture of the cavitation flow. A summary of the method
proposed by Maklakov [1997] and some numerical results are discussed below.
Consider a two-dimensional steady cavitating flow of an ideal incompressible fluid over an
arbitrary-shaped profile (Fig.9.6.1a). Assume that its trailing edge coincides with
detachment point B of the cavity. The other
detachment point A can be either fixed (by
specifying the wetted arc length) or can be
determined from the Brillouin condition of
smooth separation. Velocity at the cavity
boundary is unity, and the velocity vector at
infinity is directed along the x-axis. The foil
shape is given by function
β = F ( s) , (9.6.1)
Fig.9.6.1. The sketch of a cavitating
flow (a), parametric plane (b). where β is the inclination angle of the tangent to
the profile surface, and s is the arc coordinate
reckoned from trailing edge B; the inclination angle at this point to the lower boundary is
denoted as βB..
Conformal mapping of the flow domain in the z-plane onto a unit-radius semi-circle
on the ζ -plane (Fig.9.6.1b), can yield the derivative of complex potential due to its simple
354
zeros at points A, B, C, D and poles of second order at points ζ ∞ , ζ ∞ ,1/ ζ ∞ ,1/ ζ ∞ :

ζ (ζ 2 − 1)(ζ 2 − 2ζ cos t D + 1)
wζ = k . (9.6.2)
(ζ − ζ ∞ ) 2 (ζ − ζ ∞ ) 2 (ζ − 1/ ζ ∞ ) 2 (ζ − 1/ ζ ∞ ) 2
As before the logarithmic function of complex conjugated velocity is expressed as a sum
ω (ζ ) = ln( dw / dz ) = ω0 (ζ ) + ω1 (ζ ) + ω2 (ζ ) , (9.6.3)
ζ − eit D
where ω0 = (π − β B )i + ln , (9.6.4)
1 − ζ eit D
ω1 = iM (ζ − 1/ ζ ) . (9.6.5)
Functions ω0 correspond to the flow around a flat plate with angle of attack π − β B ;
function ω1 takes into account the simple pole at the cavity end ζ = 0 . Both these functions
are pure imaginary on diameter AB. Besides, the imaginary part of the second function is
zero on circle arc BDA, Im ω1 (eit ) = 0 . Function ω2 (ζ ) is analytical in the domain and
should be pure imaginary on diameter AB, hence it could be expressed as power series with
imaginary coefficients. But as noted above the series converges very slow.
9.6.3. Integral Equation
Maklakov [1997] reduced the problem to a Villat-type integral equation solved it
numerically by discretization on a parametric semi-circumference using spline
representations. Continuing analytically the function ω2 (ς ) from the upper semicircle to the
entire circle and then using the Shcwarz formula, one can find the following expression in terms of
π
1−ζ 2 λ (t ′)dt ′
real function λ (t ) = − Im ω2 (eit ) : ω2 (ζ ) = ∫
π i 0 1 − 2ζ cos t + ζ 2
. (9.6.6)
Whence its real part on circumference ζ = eit can be found as

π
1 ⎛ t′ + t t′ − t ⎞
μ (λ , t ) = ∫
2π 0
[λ (t ′) − λ (t )] ⎜ cot
⎝ 2
− cot
2 ⎠
⎟ dt ′ . (9.6.7)
Using identity ds / dt =| wζ (ζ )e −ω (ζ ) | and the above functions can yield a relationship between
curvilinear abscissas on the wetted boundary of foil and circle ζ = eit in integral form as
t
s (λ , t ) = k ∫ g (τ )e − μ ( λ ,τ ) − 2 M sinτ dτ , (9.6.8)
0
4b 4 sin τ [1 − cos(τ + t D )]
where g (τ ) = , ζ ∞ = beiγ .
[1 − 2b cos(τ − γ ) + b 2 ]2 [1 − 2b cos(τ + γ ) + b 2 ]2
Taking into account equation λ=β−βB and function (9.6.1), one can write a nonlinear integral
equation for unknown function λ (t ) : λ (t ) = − β B + F [ s (λ , t )] . (9.6.9)
This equation differs from the Villat [1911] equation, which presents the sought arc
coordinate s as a as a function of polar angle t, though both equations are equivalent.
9.6.4. Complementary Equations for Finding Mathematical Parameters
The obtained functions contain unknown parameters: k, M, t B , b and γ , so it is necessary to
satisfy five additional conditions:
355
• the given velocity at infinity: ω (ζ ∞ ) = ln σ + 1 ;

d
• the closure condition at infinity: lim [ln(ζ − ζ ∞ ) 2 wζ (ζ ) − ω (ζ )] = 0 .
ζ →ζ ∞ d ζ
These two conditions in complex form yield four real equations. The fifth condition could be
a length of arc BA, s (π ) = l , or the Brillouin condition of smooth free boundary detachment
at point A, which is expressed in terms of function λ (t ) as
π
tD dλ t
π tan +∫ tan dt = 2π M . (9.6.10)
2 0 dt 2
9.6.5. Numerical Method of Discretization

Locating a mesh of nodes ti , i = 1, N ; t1 = 0 , t N = π on arc [0, π ] we seek the values of
function λi = λ (ti ) . Let functions Ai (t ) , i = 1, N be a system of fundamental cubic splines
[Ahlberg et al., 1967] on mesh ti with natural boundary conditions at the ends of segment
[ 0, π ] (the second derivatives equal to zero). Then any function R (t ) twice continuously
differentiable within this segment can be approximated as
N
R (t ) ≈ ∑ Ri Ai (t ) , (9.6.11)
i =1
where Ri = R (ti ) , Ai (t j ) = δ ij , j = 1, N ; Ai′′(0) = Ai′′(π ) = 0 ; δ ij are the Kronecker symbols.

Now, all integrals of equations (9.6.7) – (9.6.10) are presented as following:
ti N ti π N
∫ λ (t )dt ≈ ∑ Sij λ j , Sij = ∫ Aj (t )dt ;

0 j =1 0
∫ λ (t )dt ≈ ∑ a λ ,
0 j =1
j j a j = S Nj ;
N N a j (1 − ζ 2 )
ω2 (ζ ∞ ) ≈ ∑ ω2 j (ζ ∞ )λ j , ω2′ (ζ ∞ ) ≈ ∑ ω2′ j (ζ ∞ )λ j ; ω2 j (ζ ) = i .
j =1 j =1 π (1 − 2ζ cos t j + ζ 2 )
Further we need to calculate derivatives of function λ (t ) at the mesh nodes. Due to
equation (9.6.9), function λ (t ) satisfies conditions λ ′(0) = 0 and λ ′(π ) = 0 , hence the
derivative can be approximated by a system of fundamental cubic splines B j (t ) , j = 1, N
with zero values of derivative at the ends: Bi (t j ) = δ ij ; Bi′(0) = Bi′(π ) = 0 . Thus,
π N −1
N
⎛t⎞ N
t
μi ≈ ∑ Cij λ j , ∫ λ ′(t ) tan ⎜ 2 ⎟ dt ≈ ∑ J j λ j ; J j = ∑ ai B′j (ti ) tan 2 − 2aN B′′j (π )
i
j =1 0 ⎝ ⎠ j =1 i =1
⎧ aj ⎛ t j + ti t j − ti ⎞
⎪ ⎜ cot − cot ⎟ , i ≠ j;
ai ⎪ 2π ⎝ 2 2 ⎠
Cij = − B′j (ti ) + ⎨ N
t +t t −t
⎪−∑ j ⎛⎜ cot j i − cot j i ⎞⎟ , i = j;
π a
⎪ j =1 2π ⎝ 2 2 ⎠
⎩ j ≠i
Substituting these expressions into the above-obtained equations, one obtains a system of
N + 5 nonlinear equations with N + 5 unknowns λi (i = 1, N ), k , M , t D , b, γ :
356
⎛ N ⎞
N
− ∑ Cik λk − 2 M sin t j
• integral equation (9.6.9) λi = F ⎜ k ∑ Sij g j e k =1 ⎟ − β B , i = 1, N ; (9.6.12)
⎜ j =1 ⎟
⎝ ⎠
• velocity at infinity (two equations)
N
⎛ 1−ζ ∞2 ⎞
− ∑ ω2 j ( ζ ∞ ) λ j ζ − eit iM ⎜⎜⎝ D
ζ ∞ ⎟⎠
⎟ + i (π − β B )
e j =1
− σ +1 ∞ e =0; (9.6.13)
1 − ζ ∞ eit D
• closure condition at infinity (two equations)

N
⎛ 1 ⎞ 1 2 i 2e − iγ
∑ ω2′ j (ζ ∞ )λ j − iM ⎜1 + 2 ⎟ − − − + = 0 ; (9.6.14)
⎝ ζ∞ ⎠ ζ∞ ζ∞ − e b sin γ b − 1/ b
− it D
j =1
N
N − ∑ C jk λk − 2 M sin t j
• the length of arc BA l = k ∑ a j g je k =1
. (9.6.15)
j =1
Equations (9.6.12) – (9.6.15) are sufficient if detachment point A is given. For a smooth
detachment, equation (9.6.15) should be replaced with the Brillouin condition
t N
π tan D + ∑ J j λ j = 2π M . (9.6.16)
2 j =1
This equation set was solved by [Maklakov, 1997] using Newton’s numerical
method. Calculating analytically Jakobian of the system makes the iterative procedure faster
and more accurate.
To check effectiveness of the method, a sample computation of the flow over a circular
cylinder was carried out. In the sample computation the lower point of detachment was
chosen according to the data from [Terentiev, 1981a; Terentiev & Dimitrieva, 1998] while
the upper one was determined by the Brillouin condition. As a result a symmetric flow was
obtained for the Helmholtz model, as well as for the cavitating flow (see Sec. 4.2). Drag
coefficients have coincided with accuracy of 10−4 with the data obtained earlier by
Terentiev, and cavity lengths determined with accuracy of 10−2 .
Some of numerical results for cavitating flow over a Joukowski hydrofoil are
depicted in Fig.9.6.2. Taking the profile chord as unity, its geometrical parameters are as
follows: width k = 11.82%, radius of curvature at the leading edge r = 0.016 , perimeter P =
2.053; angle of attack α = 5 . Fig.9.6.2 shows dependencies of drag coefficient CD (a) and
lift coefficient CL (b) on detachment point location la for fixed cavitation numbers σ .
Similar dependencies are plotted in Fig.9.6.2c for cavity length. Points on the curves
correspond to the Brillouin condition. These are the points where all curves have extremes,
namely, cavity drag and length reach maximums while lift has a minimum. Similar results
were obtained in Sec. 4.2 for a circular cylinder, but in that case we observed two maximums
for resistance. It should be noted that the numerical method presented in Secs. 9.3 and 9.4
gives the same results as obtained in [Maklakov, 1997].
357
a) b)
Fig.9.6.2. Numerical results for

Joukowski foil: effects of cavitation
number σ , and detachment point
location la on drag coefficient (a), lift
coefficient (b) and cavity length (c).
c)
Sec. 9.7. Complex Variable BEM for Plane Stationary Problems

9.7.1. Brief Notes
In planar flows, discretization of continuous functions can be realized directly in the plane of
complex variable using Cauchy integral formula instead of Green’s identity. This approach
was described in [Hromadka & Lai, 1987]. Using Cauchy integral expression, Mokry [1990]
wrote integral relationships for the tangent and normal velocity components and provided
solutions either for a single airfoil or for a system of airfoils in infinite flow. This method
was further evolved conformably to a planar flow in a free-boundary gravity field by
Afanasiev and his colleagues [Afanasiev & Goudov, 2001; Afanasiev & Samoilova, 1995;
Afanasiev & Stukolov, 1999, 2002]. Some of their results are presented below.
Free boundary deformations in a gravity field have been considered mostly for the
sub-critical range of Froude numbers. As for the supercritical range, little, if any, attention
has been paid to problem solutions for that Froude number range in the bulk of works
devoted to restrained flows over obstacles. Presented below are the results of numerical
studies of steady flows over obstacles. The numerical method based on complex variables
can be used effectively for studying flows with free boundary including cavitating flows.
9.7.2. Cauchy Integral Formula
If domain D is bounded by piecewise-smooth curve C, then any analitic function w is
358
determined by Cauchy integral via its boundary value
1 w( z ′)
w( z ) = ∫
2π i C z ′ − z
dz ′ . (9.7.1)
As above, wk = w( zk ) is a value of function w( z ) at points zk ; zk ∈ C , k = 1, N are corner

points of the polygon and C is an approximating contour. Since linear approximation is used,
the corner point coincides with a node. Consider linear analitic functions
z − zk −1 z − zk +1
N1k ( z ) = , N2k ( z) = , k = 1, N ,
zk − zk −1 zk − zk +1 (9.7.2)
( z0 ≡ z N , z N +1 ≡ z1 ),
which could be integrated analytically over any curve connecting two points
zk zk +1
N1k ( z ′) z − zk −1 zk − z N 2 k ( z ′) z − zk +1 zk +1 − z
∫z z′ − z dz′ = 1 + zk − zk −1 ln zk −1 − z , z∫ z′ − z dz′ = −1 + zk − zk +1 ln zk − z .
k −1 k
Due to the Schwartz principle of analitic extension, function w( z ) can be replaced

with functions (9.7.2) near boundary C and integrated over any curve connecting the nodal
points, and then the Cauchy integral (9.7.1) is presented as
1 N ⎛ z − zk −1 z −z z − zk +1 zk +1 − z ⎞
w( z ) = ∑ wk ⎜ ln k +
2π i k =1 ⎝ zk − zk −1 zk −1 − z zk − zk +1
ln ⎟
zk − z ⎠
(9.7.3)
at the nodal points, wn = w( zn ) . (9.7.4)

Eq. (9.7.4) can be written in matrix form as W = ZW , (9.7.5)
where W is an N-dimensional vector with complex array cells wk = w( zk ) ; Z is the N x N
⎧ Z n′,k , n ≠ k,
⎪
matrix with complex array cells Z n k = ⎨ 1 zn +1 − zn (9.7.6)
⎪ 2π i ln z − z , n = k ,
⎩ n n −1
where
⎧ 1 ⎛ zn − zk −1 z − zn z −z z −z ⎞
⎪ ⎜ ln k + n k +1 ln k +1 n ⎟ , n ≠ k − 1, n ≠ k + 1,
⎪ 2π i ⎝ zk − zk −1 zk −1 − zn zk − zk +1 zk − zn ⎠
⎪⎪ 1 z − z z −z
Z n′,k = ⎨ n k +1
ln k +1 n , n = k − 1,
⎪ 2π i zk − zk +1 zk − z n
⎪ 1 zn − zk −1 z − zn
⎪ ln k , n = k + 1.
⎪⎩ 2π i zk − zk −1 zk −1 − zn
As seen from equations (9.7.4) and (9.7.5), the Cauchy singularity of equation (9.7.1) at
point zn vanishes. Besides, contrary to the boundary Cauchy integral, equations (9.7.4) and
(9.7.5) contain an additional item, which appears while approaching the boundary.
Designating the real and imaginary parts of matrix and vector as
Z = A′ + iB, W = Φ + iΨ (9.7.7)
equation (9.7.4) yields two scalar identities:
AΦ = BΨ, AΨ = −BΦ , (9.7.8)
where A = A′ − E . Since equation (9.7.5) is valid for any analitic functions including
359
constant, then the determinant of matrix W – E and matrix A should be zero. It means that
the potential and the stream function are determined to within a constant.
For example, consider a flow induced by dipole w( z ) = m( z − zm ) −2 with m = 1 + 2i
and zm = 1 + 0.5i . The potential and the stream function are calculated exactly at any points
as ϕ = Re w( z ) and ψ = Im w( z ) . Let the domain be an ellipse of half-axes a = 1 and b = 0.5
with a cutout sector of angle γ = π / 4 , Fig.9.7.1, so that its boundary has three angular
points with angles 2π − γ , π / 2 and less than π / 2 . The
dipole located out of the domain, thus function w( z ) is
analytical in the domain including the boundary. The
ellipse is given in parametric form as:
x = a cos t , y = b sin t , t ∈ (α , 2π ) , α = arctan(b −1 tan γ ) .
Dividing interval ( α , 2π ) into N1 equal pieces,
we set N1 nodal points on the ellipse arc; the nodal
points on the horizontal and inclined section are given Fig.9.7.1. Calculation domain.
by dividing them into equal pieces of numbers N 2 and
N 3 . A number of total points are equal to N = N1 + N 2 + N 3 . Accuracy of numerical
calculations can be estimated by value ε = max(| Wk − w( zk ) |) / max(| w( zk |)) .
k ∈N k ∈N
9.7.3. Problem Statement of Flow over Obstacles

For the low shown in Fig.9.7.2, its domain has free boundary C1 and the bottom with
cylindrical bulge C3 of radius R, and also two
infinite points at the left and right sides. But the
domain should be closed for using equations (9.7.5)
and (9.7.8). It can closed by any additional straight
lines C2 and C4 drawn at a far distance. It is
assumed that the inflow speed, as well as the depth
Fig.9.7.2. Flow over an obstacle. and density, are unities, v∞ = 1, H = 1, ρ = 1 ; while
the fluid flows in a transverse gravity field.
The complex potential w( z ) = ϕ + iψ satisfies the following conditions:
• both dynamic and kinematic conditions at free boundary C1 ,
ψ = 1, | wz |2 = 1 − 2( y − 1) / Fr , where Fr = v∞2 / gH ; (9.7.9)
• kinematic conditions on the bottom and bulge, and at both perpendicular straight
ψ = 0, z ∈ C3 ,
lines (9.7.10)
ψ = y , z ∈ C2 , C4 .
9.7.4. Calculation Algorithm for a Free Boundary
The free boundary can be calculated iteratively to determine the potential and then the free
boundary. Let boundary C1 be known in the k -th approximation. To begin an iterative
process, it is necessary to find potential ϕ ( x, y ) at boundary C1 . Due to second condition
(9.7.9) the velocity is expressed as v =| wz |= 1 + 2(1 − y ) / Fr . (9.7.11)
360
Let velocity v0 at a free point with ordinate y0 is known. Then the Froude number can be
⎛ v2 ⎞ y − 1
eliminated v = 1 − ⎜1 − 02 ⎟ . (9.7.12)
⎝ v∞ ⎠ y0 − 1
Since boundary C1 is a streamline, the velocity vector is a tangent line to it. Thus
v = ± ∂ϕ ∂s (the plus sign corresponds to curvilinear abscissa s directed toward the stream).
Since the potential is determined with the accuracy to an additive constant, let’s set Φ1 = 0 .
The potential at other nodal points of free boundary can be calculated step by step as
Φ k +1 = Φ k + 0.5(Vk + Vk +1 )lk , lk =| zk +1 − zk | . (9.7.13)
Now, the free boundary calculation algorithm is as follows:
1) let a certain position of free boundary C1( n ) be known (the initial position of C1(0) is given
as horizontal straight line Yk(0) = 1 );
2) the values of Φ (kn ) at nodes, Z k( n ) on C1( k ) are determined from (9.7.12) and (9.7.13);
3) the system of linear equations (9.7.8) is solved;
⎛ dw( n ) ⎞
4) the velocity vector components at free boundary points C1( k ) are U k( n ) − iVk( n ) = ⎜ ⎟ ;
⎝ dz ⎠ k
5) the collinearity condition of velocity vector and tangent to the boundary, dy /dx = V /U ,
yields the new position of free boundary C1( k +1) as follows: Yk( n +1) = Yk( n ) + ΔYk( n ) .
v( x)
Increment ΔYk( n ) is determined by expansion of function y ′( x) = in the Taylor series at
u ( x)
each nodal point. Here, we kept four terms of the series
V (n) 1 d 3 ⎛ Vk( n ) ⎞
ΔYk( n ) = k( n ) ( xn +1 − xn ) + … + ⎜ ⎟ ( xn +1 − xn ) .
4
Uk 4! dx 3 ⎝ U k( n ) ⎠
The iterative loop is repeated until achieving the required accuracy, max | yk( n ) − yk( n −1) |≤ ε .
k
Then, the Froude number is evaluated using formula Fr = 2( y0 − 1)(1 − V ) , except a vicinity
2
of point y0 where the initial value y0(0) = 1 + 0.001 is assigned.

The derivatives at the boundary points were calculated by high-order accuracy
schemes [Terentiev & Afanasiev, 1987; Afanasiev & Samoilova, 1995]. Earlier, Guzevsky
[1982] used a similar algorithm for calculating a free boundary with high accuracy and good
convergence. The difference of the present approach is that increment Δy is found
expanding the function obtained from (9.7.3) in the Taylor series. This reduces the number
of iterations and increases accuracy of the method, especially for limiting modes. Some
examples are considered below.
9.7.5. Solitary Waves over Obstructions
Consider a flow without waves past an obstacle. If the problem of fluid flow over an obstacle
is solved using the Bernoulli integral, it can provide only a trivial solution that describes a
uniform flow ignoring the obstacle. In this case, the solution is valid for Froude numbers
Fr ≥ 1 , depending on ratio R / H ; below these values no steady solution exists. Another
solution is also possible for this problem; it was obtained in [Guzevsky, 1982]. Vanden-
361
Broeck also sought two solutions and calculated the wave amplitude and Froude number for
R/H=0.2 and 0.5 but he didn’t completed calculations for amplitudes close to the limiting
value. Guzevsky [1982] made detailed calculations only for R/H=0.1 and found no multi-
valued dependence a=a(Fr) for amplitudes a close to the limiting values. Numerical results
calculated by Afanasiev & Stukolov [2002] are plotted in Fig.9.7.3.
Fig.9.7.3. Solitary wave over circular

obstacle: left - amplitude vs. Froude number
and radius; right – free-surface shapes.
It is shown, that the nonlinear problem of a flow of an ideal heavy fluid over an
obstacle has also a third solution in the range of wave amplitude limiting values. The
ambiguity of the solution for a solitary wave is noted in [Longuet-Higgins, 1980; Vanden-
Broeck, 1989]. Fig.9.7.3 shows the calculated values of amplitude a versus Froude number
for different values of R/H. Curve 1 corresponds to R/H=0 and describes the solitary wave,
and curves 2-9 correspond to ratios R/H=0.1, 0.2, 0.3, 0.5, 0.7, 0.9, 1.0 and 1.1. The
rectangular insert shows the range where the solution is ambiguous in a limiting wave zone.
This region is scaled up separately in the left-bottom plot of Fig.9.7.3a, where the
calculations for R /H = 0 (solitary waves) completely agree with the calculations by
Maklakov [1997] using his theory (the calculations for R/H = 0.1 are taken from [Guzevsky,
1982], and the calculations for R/H=0.2 are taken from [Vanden-Broeck, 1989]).
The right plots of Fig.9.7.3 show the free-surface shapes for R/H=0.1 corresponding
to three solutions for the same Froude number (Fr)0.5=1.2910 with amplitudes equal to
a=0.8332, 0.7531, 0 (curves 1-3). The right-bottom plot of Fig.9.7.3 shows solutions for
R/H=0.2 and (Fr)0.5=1.3322 at a=0.8875, 0.8220, 0.1106 , 0.1106 (curves 1, 2 and 4). For
the latter, the wave shape for the Froude number which is maximum for the first solution and
the beginning of the second solution is also given, (Fr)0.5=1.1704 and a=0.2355 (curve 3).
Further study of the problem is given in [Afanasiev & Stukolov, 1999].
9.7.6. Conclusion: Limiting Waves
The problem of the limiting amplitude of a solitary wave has been studied extensively, but
the results obtained by many authors differ from each other. The most accurate calculations
of solitary waves have been obtained by Williams [1981], a = 0.833197, (Fr)0.5=1.290888,
362
Evans & Ford [1996], a = 0.833199179 , (Fr)0.5=1.29089053. Later investigations [Zhitnikov
et al, 2000] confirm this conclusion: a = 0.8331990845196 , (Fr)0.5=1.2908904558634. The
values of all characteristics obtained at the total number of elements N = 290 differ between
each other by the fourth digit after decimal.
9.7.7. Hydrofoil in a Stream of Finite Depth
Unlike the preceding problem, the flow domain in
this case is double-connected (Fig.9.7.4). Conditions
(9.7.9) at the free boundary and at the bottom and on
perpendicular lines (9.7.10) are valid, but a constant
value of the hydrofoil’s stream function C5 is
unknown. To find it, as before, the Joukowski-Kutta Fig.9.7.4. Hydrofoil in a stream.
condition should be satisfied at the foil trailing edge.
Instead of complex potential w( z ) , complex velocity χ ( z ) = dw( z ) / dz is used for
solving this problem. Function χ ( z ) = u ( x, y ) − iv( x, y ) is analytical and can be interpolated
through its values at nodal points similar to (9.7.4). According to equations (9.7.9) and
(9.7.10), the complex velocity should satisfy the same dynamic condition (9.7.11) at the free
V ⋅ n = 0 z ∈ C1 , C3 , C5 ,
boundary and following kinematic conditions: (9.7.14)
V ⋅ n = ±1 z ∈ C2 , C4 .
The Joukowski-Kutta condition at the trailing edge is given by
lim[(V ⋅ s ) + + (V ⋅ s ) − ] = 0 . (9.7.15)
z → zT
The velocity normal and tangent components on a contour are expressed by

coordinate components of velocity and tangent angle as
V ⋅ n = −u cosθ + v sin θ , V ⋅ s = u cosθ + v sin θ . (9.7.16)
The tangent angle at nodal points can be calculated separately from given contour, but can
also be calculated directly at the given nodal points. The trigonometric functions look like
y − yk −1 x −x y − yk −1
tan θ k = k +1 , cos θ k = k +1 k −1 , sin = k +1 . (9.7.17)
xk +1 − xk −1 | zk +1 − zk −1 | | zk +1 − zk −1 |
Conditions (9.7.14) and (9.7.15) at nodal points are expressed as
vk = 0, k ∈ N (C2 ),
−uk cos θ k + vk sin θ k = 0, k ∈ N (C1 ), N (C5 ),
. uk = 1, k ∈ N (C2 ), (9.7.18)
uk = −1, k ∈ N (C4 ),
u1 cos θ1 + v1 sin θ1 + u N −1 cosθ N −1 + vN −1 sin θ N −1 = 0, k ∈ N (C5 ).
Here, a number of elements on each boundary part is denoted as N (Cr ) ; the N (C5 ) -th nodal
point coincides with the trailing edge. N-vectors U{ uk } and V{ vk } are interrelated by
matrix equations (9.7.8) with U and -V instead of Φ and Ψ . Because of linear dependence
of the relationship equations only N − 1 equations are independent, which together with
equality of (9.7.18) gives a set of N linear equations with respect to uk for given vk and vice
versa. Afanasiev & Stukolov [2002] calculated directly the tangent components of velocity
363
instead of values uk and vk , and then obtained speed components. The free boundary can be
obtained as above iteratively.
For example, consider two problems: a circulation flow around a circular contour of
diameter d=0.4H and around the Joukowski airfoil (d=0.2H, h=0) designed and reduced
proportionally to l=0.4H, where l is the airfoil chord length. The domain boundary is
approximated by 500 elements (100 elements for the object, 200 elements for the free
boundary, 150 elements the flat bottom, and 50 elements for the inflow and outflow sectors).
The convergence point of streamlines is located at the sharp edge for the Joukowski
foil, but at an unknown point for the circular contour. Therefore, to obtain the unique
solution, one has to specify either velocity circulation on the circular contour or the
convergence point. To specify the convergence point, one should use the Joukowski-Kutta
condition at the trailing edge. It follows from the numerical analysis that the flow problem
around Joukowski foil has also a non-unique solution for the Froude numbers close to unity.
Fig.9.7.5. Hydrofoils in a stream.

Free-surface shapes above a circular
cylinder (a) and Joukowski foil (b);
stream around Joukowski foil under
the free boundary of fluid (c).
x
Fig.9.7.5a&b shows the free surface of flow around two bodies: (a) the circular
contour with given stagnation points (-0.2, 0.5) and (0.2, 0.5) for Fr = 1.351 and a=0.904
(curve 1), Fr = 1.346 and a =0.818 (curve 2), Fr = 1.235 and a=0.356 (curve 3), and
Fr = 1.346 and a=0.220 (curve 4); (b) the Joukowski airfoil for β=0°, Fr = 1.264 and
a=0.634 (curve 1), Fr = 1.173 and a=0.383 (curve 2), Fr = 1.064 and a = 0.087 (curve
3), and Fr = 1.173 and a = 0.035 (curve 4). Curves 1 correspond to the maximum
amplitude wave for which a steady solution was obtained, curves 2 and 4 – to the identical
Froude number but different amplitudes (these curves demonstrate that the solution is not
unique), and curves 3 to the Froude number below which there are no steady solutions.
Fig.9.7.5c shows the streamlines near the Joukowski airfoil for β=15°, Fr = 1.237
and a=0.355. Good agreement of the flow pattern and the Joukowski condition implicitly
confirm accuracy of the calculations. The examples show that the boundary element method
in complex variables is highly accurate for studying the planar flow problems.
364
Sec.9.8. Unsteady Flow with Free Boundaries

9.8.1. Brief Literature Review
A non-stationary flow with free boundaries is quite difficult for analytical study due to a
non-linearity of dynamic conditions at the free surface. It can be studied numerically only.
Various approaches have been used for solving non-linear problems of unsteady deformation
of the free surface [Longuet-Higgins & Cokelet, 1976; Kedrinskij & Lavrent’ev, 1983;
Afanasiev & Terentiev, 1984; Petrov & Smolyanin, 1993; Tuck, 2000; Kim et al., 2002].
Preceding sections show high efficiency of the boundary element method for steady
flow problems. Now we can conclude with confidence that almost all steady problems can
be solved numerically. Particularly, the planar and axisymmetric problems can easily be
computed using stream functions. But for unsteady flows, the stream function cannot
simplify problems due to continuous changing. In that case, the speed potential is much
preferable than the stream function though it can be useful for some special problems. The
boundary element method remains also an effective tool of numerical calculations of
unsteady flow with free boundaries. Many problems have been solved using this method
[Liu & Ligget, 1986; Afanasiev et al., 1986; Terentiev & Afanasiev, 1987; Terentiev et al.,
1987; Afanasiev et al., 1989; Terentiev, 1989, 1997; Petrov, 1996; Afanasiev, 2002].
Numerical solution of an unsteady flow with free boundaries, in spite of its
unsteadiness, has an advantage over that with stationary free surfaces. The steady free
surface is a’priori unknown whereas the initial position of an unsteady free boundary is
usually given. Thus the further movement of free surface can be calculated step by step by
alteration of time. Computing an unsteady problem is divided into two stages: first is the
steady flow with given boundary conditions as Dirichlet or Neumann’s or another one at
fixed time t; second is a transfer of each point of the free boundary in a small time interval
Δt .
9.8.2. Numerical Algorithm for Calculating Free Boundaries in the First Stage
The first stage of numerical algorithm concerns to solving a steady problem numerically by
BEM with complex variable as in the previous section for planar flow, or Green’s identities
(9.1.3) for both planar and 3D flows. Digitizing finite elements by the Green’s identity can
be written in matrix form AΦ = B Q , (9.8.1)
where A and B are square matrixes of dimension N × N with array cells Φ and Q are N
dimensional vectors with values of potential and its normal derivative at the nodes. The array
cells could be calculated analytically for two-dimensional flow or numerically for three-
dimensional problem, App. 5. Using equation (9.8.1), one can calculate normal velocity Q if
potential Φ is known or vice verse, and then the axial speed components are determined.
Consider first a 2D problem including both planar and axisymmetric flows. Let the
speed potential be given at nodal points, i.e. matrix-vector Φ is known, then using any
approximations the potential may be presented as a differentiable function at the boundary
and the tangent projection of velocity can be calculated W = ∂ϕ / ∂s . For numerical
procedure, velocity components are preferable to find in Cartesian coordinates. The velocity
vector is V = Qn + Wm , whence the Cartesian components of the velocity are equal to
u = Vx = Qnx + Wmx , v = Vy = Qn y + Wm y . (9.8.2)
365
Components of normal vector are connected with components of unit tangent vector by
equations nx = m y = sin θ , n y = − mx = − cosθ , (9.8.3)
where θ is the tangent angle to the boundary. It can be calculated by finite-difference
approximation. Considering small time increment Δt , one can find fluid particle
displacement Δr = V Δt. If a fluid particle at time t is located at point r , then at time t + Δt
the particle moves to point r = r + V Δt , (9.8.4)
or in Cartesian coordinates the new position of the fluid particle is
x = x + u Δt , y = y + vΔt . (9.8.5)
The second stage of the algorithm includes calculating the new potential distribution
on deformed free boundary С2 using equations (9.8.5). If the speed potential at time t is
given for a fluid particle at point r , then the increment of potential for the same particle
with respect to time increment Δt can be found from equation (2.1.2) regarding to identity
dϕ / dt = ∂ϕ / ∂t + u 2 + v 2 as Δϕ = [(u 2 + v 2 ) / 2 − gy − p / ρ + C ]Δt , (9.8.6)
and the new potential distribution on the boundary due to (9.8.5) is calculated by
ϕ = ϕ + Δϕ . (9.8.7)
As a rule, the pressure on the free boundary is constant, and constant C is determined by
given parameters in undisturbed fluid. Now, the algorithm looks as follows:
(n)
Φ ( n −1) ⎯⎯⎯
(8.8.1)
→ Q ( n −1) ⎯⎯⎯
(8.8.5)
→r ⎯⎯⎯
(8.8.7)
→Φ ( n ) . (9.8.8)
9.8.3. Vertical Motion of Cylinders and Spheres in a Fluid with a Free Surface
For example, consider vertical motion of a circular cylinder and a sphere in a fluid with free
surface and in the gravity field. Some numerical results computed by MM Afanasieva and
elaborated in [Terentiev et al., 1987; Terentiev, 1989 ] are plotted in Figs.9.8.1 and 9.8.2.
Linear approximations at boundary elements have been used. Fig.9.8.1 shows the free
surface for a circular cylinder (solid lines) and a sphere with unit diameter (dashed lines)
rising at a constant unit speed to a free surface. Froude number Fr = V 2 / gd = 1 ; the initial
Fig.9.8.1 Exit of the cylinder (solid Fig.9.8.2 Evolution of the free

lines) from fluid in a gravity field. surface for immersion of a sphere.
Curves 1-4 are for the times τ = 2.5, Curves 1-6 are for the times τ =
2.7, 3.1, 3.25. 0.15, 0.3, 0.375, 0.4, 0.5, 0.575.
366
immersion of the cylinder and the sphere is h(0)=3D. The value of dimensionless time
τ = tD / V is shown in Figs.9.8.1 and 9.8.2. It is seen that some fluid emerges with the body.
As some fluid emerges with the body, wave formation is observed on the fluid surface in the
plane case. Similar results were obtained in [Telste, 1986] for a circular cylinder rising to a
free surface but without any waves. This dysfunction can be explained by different initial
conditions, time-dependent velocity, and also by calculation errors. The free surface
evolution due to immersion of a sphere at unit velocity for Fr = 0.0625 is plotted in
Fig.9.8.2. Initially, the sphere was immobile at depth h(0)=0.6 under the free surface. It is
curious to note that during submerging the sphere, small waves are generated on the risen
free water surface, which propagates out from the center (curves 5 and 6).
9.8.4. Horizontal Movement of a Half Cylinder
Let a semi-cylinder moves in a ponderable fluid of finite depth H from its quiescent state
along the bottom, Fig.9.8.3. Cartesian coordinates x, y are used for considering a non-
stationary motion of ideal incompressible fluid with a free boundary. The flow domain is
bounded by semi-circle C1, flat bottom C2, free boundary C3, and two truncated
perpendicular dash-lines C4 and C5. The
semi-cylinder moves horizontally with
constant speed V, within fluid bounded by
free surface S(t), moving semi-cylinder Q(t),
fixed bottom W. It is also bounded by two
vertical lines, drawn sufficiently far from the
cylinder by distances –12H and 6H. As
Fig.9.8.3. Half cylinder moving in a before, the initial free boundary is horizontal.
ponderable fluid. The distance traveled by the cylinder center
from its initial position corresponds to the
dimensionless time, i.e. the cylinder speed and diameter are assume to be unities.
A system of linear algebraic equations for determining the unknown potential and its
normal derivative in the nodes at each time-step was solved by the Gauss method with
selecting the pivotal element. The solution procedure for varying boundary conditions is
described in [Terentiev et al., 1987, Afanasiev & Samoilova, 1995].
Besides geometric characteristics, physical peculiarities of the arising wave are of
interest in this problem. A breaking wave is the most difficult object for studying the wave
phenomena due to high rates and considerable accelerations of processes, and large
deformations of the free surface. Typical for this problem is that before the wave breaks, an
element of water surface becomes vertical and then the crest breaks forming a jet directed
towards the wave motion line. This follows from calculations presented in Fig.9.8.4. The
number of nodes on the free
boundary is 68, and 190 on the whole
boundary. In the initial time the
points on free boundary were
concentrated about the cylinder top in
a special way.
Fig.9.8.4.
Formation of a breaking wave behind
the moving cylinder.
367
Many authors have considered an overturning wave [Grilli et al., 1989; Haack et al.,
1981; Petrov, 1996; Afanasiev & Stukolov, 1998]. Namely three regions in a flow domain
near an overturning solitary wave were distinguished in [Petrov & Smolyanin, 1993;
Afanasiev & Stukolov, 1998] (it was earlier predicted in [Peregrin, 1987]):
1) velocity of separate fluid particles exceeds the wave velocity;
2) a thin layer is formed in the leading zone of the wave where acceleration of particles is
noticeably higher than accelerations in the rest of the wave;
3) a large zone appears at the back surface of the wave
where acceleration of particles is very small.
The waves considered in the present work, though different
from solitary waves, have the above-mentioned peculiarities
proper for breaking solitary waves. This is well seen in
Fig.9.8.5 showing the calculated fields of velocities and
accelerations.
As known the problem of steady flow of a finite
depth ponderable fluid around an obstacle has a steady
solution for some conditions associated with sizes and
Froude number [Sturova, 1990; Vinje & Brevig, 1981]. For
a subcritical flow, a stationary wave train can exist only at
Froude numbers lesser than a certain critical number
Fr ∗ ( R /H ) . A steady supercritical flow is possible only if
the Froude number is greater than the second critical value
Fr ∗∗ ( R / H ) . Thus in the range Fr* < Fr < Fr ** the flow
could be unsteady only. This effect was verified Fig.9.8.5. Water overturning:
experimentally [Forbes & Schwartz, 1982]. More detailed (a) velocity; (b) acceleration.
study of a semi-cylinder moving in a ponderable fluid was
performed by KE Afanasiev with his colleagues. Some of their numerical results are plotted
in a slightly modified form in Fig.5.8.4 thru Fig.5.8.7 for ratio R/H=0.2. Based on
calculations for Froude numbers ranging from 0.01 up to 100, three flow modes are
distinguished: the first is a steady mode in supercritical flows (1.44 < Fr < 100); the second
is the steady subcritical mode (0.01 < Fr < 0.36); and the third is near-critical flows with no
steady solutions (0.36 < Fr < 1.44).
Fig.9.8.6.
Wave above cylinder in steady flow (solid line) and
due to moving body (dotted line).
The first mode has been studied theoretically

and experimentally in detail [Peregrin, 1987; Petrov
& Smolyanin, 1993; Terentiev & Afanasiev, 1987]. In
the case of steady flow there is a hump above the body, and no disturbances spread up and
down the flow. There is regularity, common for this mode, of transition to the stationary
mode: at the initial moment of motion a deepening appears behind the body, which stays
virtually unchanged at the same place while the body moves. A hump is formed above the
solid and moves together with it. The humps above the moving cylinder (doted curve) and
also above the fixed cylinder in steady flow (solid line) are shown in Fig.9.8.6. The hump
formed above the moving body transits quite fast to the stationary mode.
368
The first flow mode is shown in
Fig.9.8.7a. In this mode the resistance of a
cylinder moving at a constant speed drops to zero
while a steady wave is formed above the cylinder.
The free boundary deformation behind the
cylinder diminish with time and don’t affect flow
near the body. a)
Fig.9.8.7.
Typical waves at free boundary: (a) first mode;
(b) second mode;(c) third mode.
The second mode flows, Fig.9.8.7b, are

characterized by quiescent state harmonic waves
that are formed and run in front of the moving
b)
solid while harmonic waves of smaller amplitude
are generated behind the solid. Resistance force in
this flow mode has a damped wavelike nature.
The third mode flows, Fig.9.8.7c, have
unambiguous unsteady character. Waves of
increasing amplitude are formed behind the solid
c)
and the resistance does not tend to zero.
Sec. 9.9. Generation of Waves on Free Boundaries
9.9.1. Brief Analogy with a Tsunami
The above-mentioned numerical algorithm can be applied for studying the generation
and propagation of waves on free boundary due to a fast movement of bottom. Such
problems are of special significance in connection with tsunami generated by
underwater earthquakes. Suffice to remind about the underwater earthquake off
Sumatra Island in 2004 with tsunami waves that caused the loss of about 300,000
lives, and the incident in 2011 that had devastating effect on Japan. Visual detection
of the first waves is very difficult, even from space, because of insignificant initial
disturbance of the free water surface. For instance, surface waves 60 cm high were
first detected by space radar only two hours after the earthquake near the Sumatra
shore (Wikipedia). The tsunami wave size and propagation pattern depend greatly on
the seabed displacement rate and large-scale relief. In the Sumatra earthquake, the
earth's crust broke over a length of about 1200 km with the break edges displacing,
one upward and the other downward, at a rate of 2 km/s, i.e. instantly. When all these
parameters (seabed geography, break line orientation and length, displacement rate
and orientation, etc.) are known, the problem could, in principle, be solved
numerically if not analytically. There are many physical and mathematical models
considering a number of those parameters. The simplest model of wave generation
and propagation on a free boundary is considered here, as well as the approaches for
solving it numerically.
Let the initial domain of fluid be a strip of width h ; the initial free boundary and
369
bottom are horizontal, y = 0 . The above-mentioned numerical algorithm (9.8.8)
allows computing a value problem for the given deformation rate (which is usually
unknown prior to the earthquake though it would be measured by underwater sensors).
For simplicity we shall consider the bottom displacement starting instantly with initial
rate V ( x) and then instantly standing still again [Terentiev, 1996]. The initial value
problem is similar to the impact of a rigid bottom against water. Stricktly saying, an
abrupt stop of a rigid surface constitutes an impact with speed −V ( x) , and so the
water should stop. However, the water cannot resist high depression generated by
impact, and so we can assume that the water continues moving though the bottom
remains still. Also, the problem can be divided to three consecutive problems: 1)
impact of a rigid bottom against water with a free boundary and obtaining a normal
velocity on the free boundary 2) determining a speed function by given vertical speed
of the free boundary and the fixed bottom; 3) a time-dependent value problem.
9.9.3. Bottom or Water Impact
In the first problem the following boundary conditions should be satisfied:
• at the free boundary: Φ = 0 on y = 0, x ∈ (−∞, ∞) ;
• at the bottom: d Φ / dy = f ( x) on y = − h, x ∈ (−∞, ∞) .
Function f ( x) should vanish at infinity and the equilibrium condition should be satisfied
∞
∫
−∞
f ( x)dx = 0 . (9.9.1)
This equation has a hydrodynamic interpretation as the equality of flow permeated through
the lower horizontal boundary of the strip. This equation should be a necessary condition for
the tsunami generation. Using boundary elements technique, a rectangle with aspect ratio
L = l / h is introduced to truncate the computational domain instead of the infinite strip,
Ratio L is chosen to be large enough so that during the time period of computation, waves
which propagate near the origin have not reached these fictitious vertical boundaries. One
can assume approximately that Φ =0 at the fictitious boundaries. Using the above-mentioned
collocation technique and Green’s identity (9.8.1), one can calculate normal velocity Q at
the free boundary and potential Φ at the bottom. The problem can be solved analytically for
complex velocity Ω = U − iV . Extending analytically through the free boundary, where
U = 0 , one obtains a Schwartz problem in a strip of width 2h . Its solution is obtained by
∞
1 f ( x′)dx′
(A1.16): Ω= ∫
2hi −∞ cosh[π ( x′ − z ) / 2h]
. (9.9.2)
Whence the vertical velocity at the free boundary is

∞
1 f ( x′)dx′
V ( x) = ∫
2h −∞ cosh[π ( x′ − x) / 2h]
. (9.9.3)
9.9.4. Initial Conditions

Having the liquid disturbed by a moving pulse, the bottom itself remains unmoved,
and the free surface is being deformed by inertia and gravity actions. It is assumed
that vertical speed v0 at the free surface at the initial time is equal to obtained vertical
velocity (9.9.3); the vertical speed on the bottom vanishes. Thus,
370
v0 = V ( x) on y = 0; v0 = 0 on y = − h . (9.9.4)
This problem has an analytical solution for function ω0 = u0 − iv0 as
∞
1 V ( x′)dx′
ω0 = ∫
2h −∞ tanh[π ( x′ − z ) / 2h]
. (9.9.5)
Whence horizontal speed u0 at the free boundary is calculated from (9.9.5) substituting
z = x , and then the initial speed potential may be found as
x
ϕ0 = ∫ u0 ( x)dx + c , (9.9.6)
0
where constant c can be chosen as convenient for calculations.
In spite of simplicity of Eqs. (9.9.3)–(9.9.6) calculating takes much more computer time.
Sometimes it is preferable to compute using the above-mentioned numerical method.
9.9.5. Numerical Simulation of Free Boundary Deformation
The above-obtained initial distributions of velocity and speed potential at truncated rectangle
are sufficient for computing the free boundary evolution by time steps (9.8.8). To verify the
performance of the present numerical and analytical formulations, an example has been
considered. Let a part [ a, b ] of the bottom rises with rate (speed) V0 , while the other part
[ − a, −b ] sinks with the same speed, and the rest of the bottom stays unmoved so that
condition (9.9.1) is satisfied. Integral (9.9.3) can be calculated analytically
⎡ ⎛ π ( x − a) ⎞ ⎛ π ( x − b) ⎞ ⎤
⎢ arctan ⎜ tanh 4h ⎟ − arctan ⎜ tanh 4h ⎟ + ⎥
2V ⎝ ⎠ ⎝ ⎠
V ( x) = 0 ⎢ ⎥, (9.9.7)
π ⎢ ⎛ π ( x + a ) ⎞ ⎛ π ( x + a ) ⎞⎥
⎢ + arctan ⎜ tanh ⎟ − arctan ⎜ tanh ⎟⎥
⎣ ⎝ 4h ⎠ ⎝ 4h ⎠ ⎦
while integral (9.9.5) at u0 = ω ( x) can be calculated numerically only as Cauchy principal
value. It should be noted that computing all these problems using Green identity is usually
straightforward.
The water depth as well as speed and density are assumed to be unity, i.e. all
dimensionless parameters are obtained with respect to h, V0 and ρ . The Froude number is
Fr = V0 / gh . The initial truncated domain is taken as a rectangle of sides 20 × 1 . Fig.9.9.1
shows time evolution of the free boundary for a = 1, h = 1, Fr = 0.1 , where dimensionless
time is τ = V0 t / h . The deformation of free boundary is anti-symmetric; the first waves at
both sides have the largest amplitudes but in the opposite direction; then other waves arise
with smaller amplitudes and finally waves vanish. But the waves at both sides make different
effect. The first wave at the right has the amplitude a half of that at the left, and so the right
coasts can be destroyed much less than the left one. One can see in Fig.9.9.1. that a hollow is
formed in front of the first wave, thus the water recedes first from a coast and then a huge
wave crushes at the coast with much destruction. The above-mentioned Sumatra tsunami
made most destruction just at the left side because the left part of bottom sunk. Some
destructions, though considerably lesser, can also be expected in the opposite
direction from the epicenter, especially by the second and third waves.
371
a) b)
Fig.9.9.1. Time history of free surface: (a) nonlinear theory for Fr=0.1;
(b) linear theory for Fr=1.
On advice by Terentiev, the problem was solved in a linear statement [Porfiriev &
Troeshestova, 1996]. The results of calculations for the same initial conditions but at Froude
number Fr = 1 are presented in Fig.9.9.1b. Both methods give similar results. The tsunami
model presented here is very simple but simulates the real tsunami sufficiently well. The
method allows computing much more complicated conditions at an extra computer time.
This section was to demonstrate an application of the above-mentioned numerical method to
predict such an important hydrodynamic phenomenon as tsunami. There are more
complicated models, e.g. [Murty, 1981; Marchuk et al., 1983; Voit, 1987].
Sec. 9.10. Deformations of Gas Bubbles

Deformation of bubbles is of interest in connection with cavity collapse and erosion. A
review on bubble problems can be found in [Pernick, 1966, Knapp et al., 1970; Voinov &
Petrov, 1976; Brennen, 1995]. It has been studied mostly for a single bubble in unbounded
flow or near rigid boundaries [Voinov & Voinov, 1976; Plesset & Chapman, 1971; Blake &
Gibson, 1981; Dommermuth & Yue, 1987; Takahira & Yasuda, 1998]. Interaction of two
bubbles was studied experimentally and numerically in [Tomita et al., 1994; Sato & Tomita,
1998]. All these problems were symmetrical or axisymmetrical and could be solved
analytically or numerically in two-dimensional plane. A non-symmetrical interaction
between bubbles and a vertical wall was considered in [Chahine & Duraiswami, 1998].
Bubble deformations, both symmetrical and non-symmetrical, were considered in [Terentiev
et al., 1987; Afanasiev et al., 1989; Afanasiev & Stukolov, 2002]. Some of them are
discussed below.
372

Ignoring gas diffusion through the bubble boundary, the pressure at the bubble boundary
γ
⎛ V ⎞
С (t ) is defined as pΓ = pv + p0 ⎜ 0 ⎟ . (9.10.1)
⎜ V (t ) ⎟
⎝ ⎠
Here pv is a vapor pressure in the bubble; p0 is an initial gas pressure in the bubble; V and
V0 are the current and initial bubbles, respectively, γ is an adiabatic index. The total
differential of speed potential due to Bernoulli equation is
γ
⎛ 1 2 ⎛ V0 ⎞ ⎞
dϕ = ⎜ 1 + ∇ϕ − α z − β ⎜ ⎟ ⎟ dt , (9.10.2)
⎜ 2 ⎝ V ⎠ ⎟⎠
⎝
where α = Rm ρ g Δp is buoyancy coefficient; β = p0 Δp ; Δp = p∞ − pv ; ρ is mass
density; p∞ is the pressure at infinite point of fluid; Rm is the maximum radius of bubble at
its explosion in infinite imponderable. Kinematical conditions on the bubble boundary and a
free surface are satisfied by Eqs. (9.8.5).
9.10.2. Deformation of a Single Bubble near a Free Boundary
The numerical algorithm (9.8.8) is used for axisymmetric problems of bubble deformation
near a free boundary in a gravity field taking into account the surface tension as well as a gas
in bubble [Terentiev et al., 1987; Afanasiev et al., 1989]. The initial bubble is assumed to be
a sphere of radius R0, so the initial speed is zero. The bubble center is far from the free
surface on h = 13R0 . The initial pressure in the bubble is p0 = pa + k ρ gh , and so the inner
pressure coincides with hydrostatic pressure at k=1 and greater or less than hydrostatic
pressure if k>1 or k<1,espectively. Calculations were carried out for k=0.94 unless time is
τ = t g / 2 R0 .
Results of numerical Fig.9.10.1.
computing are plotted in Fig.9.10.1. Deformation of a
The left configurations correspond to bubble near free
adiabatic changes of the inner gas, surface; the times due
p = p0 (V0 / V )κ , κ = 1.4 . Curves 2 and to numbers:
3 correspond to instances τ=0.3 and 1 – τ = 0;
0.575 , respectively. It is shown that 2 – τ = 0.3 ;
the bubble at first shrinks then 2 – τ = 0.575 ;
enlarges and surfaces. The right 4 and 5 – τ = 0.375 ;
curves are obtained for constant inner 6 and 7 – τ = 0.4 .
pressure p = p0 . This assumption can
simulate an abrupt formation of cavities and, then, disappearance. The solid and broken lines
correspond to deformation without and with surface tension, respectively. Curves 4 and 6 are
obtained for instance τ=0.375 while curves 5 and 7 – for τ=0.4. It is seen that the bubble
deformation disappears and the surface tension relaxes.
373
9.10.3. Interaction of Two Bubbles

Experimental observation show that bubbles interacting deform variously, namely, one of
bubble can break through another one [Tomita et al., 1994; Sato & Tomita, 1998; Arn et al.,
1998]. Fig.9.10.2a shows the interaction of two bubbles near a wall at two consecutive
instances. At first the bubble was generated near the wall, and then, after the time difference,
another bubble was generated between the first one and the wall; the first bubble was just in
a collapse process and was much smaller than the second one. Plotted in Fig.9.10.2b are the
numerical results using the source distribution obtained in [Tomita et al., 1994].
The procedure was successfully applied also for studying the interactions of two or
more bubbles. The numerical results obtained in [Afanasiev & Gudov, 2001] are plotted in
Fig.9.10.0c. They agree with experimental pictures very well. It should be noted that the
theoretical models and numerical methods allow studying much more complicated problems
of interaction of bubbles and can help developing a model of cavity clouds.
a) b) c)
Fig.9.10.2. Interaction of two bubbles hear a wall: (a) experimental pictures by
Tomita et al. [1994]; (b) numerical results by Sato & Tomita [1998];
(c) numerical results by Afanasiev & Gudov [2002].
Sec. 9.11. Three Dimensional Problem of Bubble Deformation

Fig.9.11.1
In the previous section the problems were Initial
axisymmetric and so the Green’s integral position of
equation could be reduced to the one- the bubble
dimensional equation and constant or linear and the wall.
elements were used. Below the Green’s integral
equation is used to 3D surfaces. Designations of
parameters are given in Fig.9.11.1. At initial
time, a spherical bubble is located near an inclined wall at distance h. This problem was
numerically studied in [Chahine & Duraiswami, 1998] where deformations of bubbles were
considered taking vorticity into account. Without vorticity but in a gravity field, the problem
was considered in [Afanasiev & Grigorieva, 2002]. Results of their investigation, presented
374
below, can be viewed as a demonstration of using BEM with triangular elements. Numerical
algorithm (9.8.8) is valid in this case too, but the Green function should be preferred
1 ⎛ 1 1 ⎞
G( z) = ⎜ + ⎟, (9.11.1)
4π ⎝ | z − ζ | | z − ζ ∗ | ⎠
where ζ and ζ ∗ are symmetrical points relative to the inclined straight line. Thus the
kinematic condition on this line is satisfied identically, and so only a bubble boundary is to
be considered. Other boundary conditions are the same as (9.8.4), (9.10.1), (9.10.2).
9.11.3. Triangular Elements
The initial sphere of unit radius is approximated by a polyhedron with triangular sides
(elements). Let r1 , r2 , r3 be the radius-vectors of vertices of a triangular element; l1 and l2
are the lengths of edges of a polyhedron side; ξ1 and ξ 2 are the local coordinates of points
on the edges, Fig.9.11.2, so the unit vectors e1 = (r1 − r3 ) / l1 , e 2 = (r2 − r3 ) / l2 , and the
radius-vector of a point on element sides is
r = r3 + l1ξ1e1 + l2ξ 2e 2 , or
r = ξ1r1 + ξ 2r2 + ξ 3r3 , ξ 3 = 1 − ξ1 − ξ 2 . (9.11.2)
Since the elementary of area is ds = 2 Sdξ1dξ 2 , hence
integrals of Green’s function and its normal derivative
with respect to the element can be calculated as
1 1−ξ1
B = 2S ∫ ∫ G (r, R )dξ1dξ 2 , (9.11.3)
0 0
1 1−ξ1
∂G (r, R )
Fig.9.11.2. Triangular element A = 2S ∫ ∫ dξ1dξ 2 . (9.11.4)
0 0 ∂n
and local coordinates.
If nodal point R coincides with a corner of element, for
instance, with r3 , then the integrands have singularities at this point. For a smooth boundary
the second integral is equal to 1/2, i.e. all diagonal elements of matrix A are Ai ,i = 1 / 2 ,
while the integral in (9.11.3) may be expressed analytically using local polar coordinates.
The area of triangle, S, height h, and angles θ1 and θ 2 between the height and adjoining
sides can be expressed in vector terms as S =| (r1 − r3 ) × (r2 − r3 ) | / 2 , h = 2S / | r2 − r1 | ,
cosθ1 = h / l1 and cosθ 2 = h / l2 . Integral (9.11.3) for R = r3 is transformed to
θ 2 h / cos θ
1 S ⎡ ⎛ π θ2 ⎞ ⎛ π θ ⎞⎤
B = 2S ∫ ∫ ρdρdθ =
ln ⎢ tan ⎜ + ⎟ tan ⎜ − 1 ⎟⎥ . (9.11.5)
−θ1 0 4πρ 2π ⎣ ⎝ 4 2 ⎠ ⎝ 4 2 ⎠⎦
These formulas determine diagonal components of matrices A and B, and could be applied if
unknown functions are found via some approximated functions. In that case a nodal point
coincides with vertex of triangle and at each node some adjoining triangular elements should
be considered. For approximation by a constant equal to the value at the nodal point, for
instance, at point of intersection of median R = r1 + r2 + r3 , a triangle should be divided into
three triangles and then calculated by the above-mentioned technique. Other non-diagonal
375
components of matrices can be calculated by any cubature formula. If the integrated domain
is triangular, one may use the Hammer’s cubature formula (A.5.28).
The technique of containing a set of linear equations is presented very well in
[Banerjee & Butterfield, 1981] where some other finite elements are also considered.
Description and application of triangular, as well as quadrilateral, elements can be found also
in [Terentiev & Afanasiev, 1987; Vaz, 2005]. The quadrilateral element has its own
distinctive feature; it should be a curved surface for the given four corner points and it can be
flat only with a specific arrangement of corner points. Anyway, for unsteady flow problems
with free boundary, quadrilateral elements should be curved while triangular elements can
always be flat. Thus the latter is preferable for investigation of flow problems with free
surfaces. Though, the use of quadrilateral elements can be standardized by approximation of
4
any function: u = ∑ ui N i (ξ ,η ) , (9.11.6)
i =1
(1 − ξ )(1 − η ) (1 + ξ )(1 − η ) (1 + ξ )(1 + η ) (1 − ξ )(1 + η )
where N1 = , N2 = , N3 = , N4 = , i.e.
4 4 4 4
the element is mapped onto square and all functions including coordinates of corner points
are transformed by (9.11.6). Quadrilateral elements have been widely used for investigation
of cavitating flow of finite-span wings and marine propellers in [Vaz, 2005].
Triangular elements and BEM-techniques were successfully used for numerical
studies of 3D deformation of bubble surface in [Afanasiev & Grigorieva, 2002] where linear
approximations of unknown functions was considered. The problem is unsteady, so it is
important to keep accurate calculation and stable solution. For that, time step Δt was chosen
to restrict maximal travel of the nodes on the bubble surface for one time step
Δt = θ ∗ lmin max Δϕ ( xi , t ) , (9.11.7)
i
where θ - mesh type coefficient; lmin - minimal rib height. Tangent components of velocity
∂ϕ / ∂ξ1 and ∂ϕ / ∂ξ 2 are calculated as finite differences. The normal vector at the node is
calculated by averaging the normal vectors of surrounding elements using Green’s integral
equation, or from a linear equation set. Then the velocity components are rearranged to
Cartesian coordinates and the new positions of nodal points are found by Eq. (9.8.4).
9.11.3. Test Example of a Single Bubble
The numerical convergence was tested and confirmed using various mesh sizes and time
steps [Afanasiev & Grigirieva, 2002]. The results were found in a good agreement with
Rayleigh solution for a spherical cavitation bubble and with experiment results. Calculating
the bubble dynamics in infinite imponderable fluid with different coefficients β provides an
excellent opportunity examining the numerical algorithm. In this case a spherical bubble of
initial radius Rm decreases to small radius Rmin. Minimum radius Rmin is obtained via
coefficient β as follows [Levkovsky, 1973]: ( β > 0.3 )
Rmin ≈ 3β (1 + 3β − β 3 2 ) . (9.11.8)
After that the bubble increases again to maximum radius Rm. Theoretically, the bubble can
pulsate unlimitedly, in our numerical modeling the bubble performs from one to four full
pulsating cycles before breaking its spherical symmetry. In case of axisymmetric modeling
we can successfully contend with this numerical instability by smoothing. There are
376
analytical and numerical quantities of
the bubble minimum radiuses for two of Table 9.11.1. Analytical and numerical
the first collapse phases for different minimal radiuses for two first pulsating and
quantities of β and scaled lifetime of the scaled lifetime of the buoyant gas bubble in
one in Table 9.11.1. infinite imponderable fluid
The bubble radiuses as function β Rmin Rmin1 Rmin2 tend
of time for different β are shown in 0.4 0.6163 0.6056 0.6031 3.9129
Fig.9.11.2. 0.5 0.6988 0.6898 0.6870 4.3220
0.6 0.7707 0.7638 0.7607 5.4710
0.7 0.8352 0.8304 0.8274 7.2098
0.8 0.8940 0.8912 0.8885 8.4509
0.9 0.9486 0.9473 0.9456 9.890
Fig.9.11.2: 1
Bubble radius as a function of
time at different β 0.9
5
6
(1 - β =0.4, 2 - β =0.5, 3
4
3 - β =0.6, 4 - β =0.7,
0.8
5 - β =0.8, 6 - β =0.9). 0.7

2
0.6
0 2 4 6 8 10 t
9.11.4. Computational Results

For studying the mutual influence of the wall and gravity forces, some results of bubble
deformations are presented. All calculations were made for β = 100 and γ = 1.4 . The
gravity acceleration is directed downward. Some results of a buoyant gas bubble near a wall
are plotted in Figs.9.11.3-9.11.5. Fig.9.11.3 shows a collapse of the gas bubble near an
inclined wall. This problem using another numerical method was solved in [Wang, 1998].
Both results are in a good agreement.
The most unusual case is the bubble above a horizontal wall ε = 0 , Fig.9.11.4. If the
buoyancy coefficient is small, α = 0.05 , the wall influence prevails. The bubble is flattened
Fig.9.11.3.
Collapse of a
buoyant gas
bubble near
inclined wall:
ε = π 4,
h=Rm=1, α=0.2;
β=100
against the wall during the growth phase and acute jet directed toward the wall is formed
during the collapse phase. Unfortunately the thin and acute jet promotes numerical instability
377
and a premature distraction of computing. The greater the buoyancy coefficient increases,
the smaller is the wall influences on the flow pattern. Namely at α = 0.2 , one can see an
approach of small jet directed opposite to gravity vector in spite of the influence of close
disposed wall. At a vertical wall the jet is opposite to gravity vector and is deflected by the
wall, Fig.9.11.5. If α = 0.2 the jet is formed in the described direction but has a greater
volume and a smaller maximum velocity. It is possible that jet velocity is smaller because
the computing breaks before the jet touches the opposite boundary of the bubble.
α = 0.05 α = 0.5 α = 0.05 α = 0.2

Fig.9.11.4. Collapse of a buoyant Fig.9.11.5. Collapse of a buoyant
gas bubble near the horizontal wall: gas bubble near the vertical wall:
ε = 0 , h = Rm = 1 , β = 100 ε = 3π / 2 , h = Rm = 1 , β = 100 .
Sec. 9.12. Finite Particle Method in Entry Problems

9.12.1. Preliminary Remarks
The compressibility of fluid has been considered primarily for an initial entry of a blunt body
into water or by an impact of a rigid body against water, under certain assumptions
compatible with the linear theory [Galin, 1947; Sagomonyan, 1959; Poruchikov, 1964;
Skalak & Feit, 1966; Galanin et al, 1979]. Non-linear problems of penetration into
compressible fluids have been solved numerically in [Gridneva et al, 1979; Bazhenov et al,
1984; Shahverdy, 1984; Terentiev & Chechnev, 1985, 1986]. Different calculation
algorithms with step-by-step accounting of time have been used. It should be noted that the
numerical method of finite particles developed by Belotserkovskii & Davydov [1971, 1973]
is convenient for development of stable computational procedures. The finite particles
378
method is similar to the method of particles in a cell [Harlow & Welsh, 1967], which
simulates continuous fluid in a cell by a finite number of particles. The method allows to
easily calculate a free boundary but it is prone to numerical instability due to discontinuity of
fluid in a cell. In this section two methods are combined: the fluid is assumed to be
continuous in a cell as in the finite particles method, and to be replaced with discrete
particles on a free boundary due to Harlow’s approach [Harlow & Welch 1966].
9.12.2. Description of the Method and Calculation Algorithm
The flow domain is broken into cells, in each cell the equations of gas dynamics in integral
d
dt ∫∫
form are: ρVdτ + ∫ ρV (V ⋅ n )ds = − ∫ pnds , (9.12.1)
τ s s
d
dt ∫∫
ρ dτ + ∫ ρ (V ⋅ n )ds = 0 , (9.12.2)
τ s
p = f (ρ ) . (9.12.3)
It is assumed that function (9.12.3) is given. The submitted equations allow calculating flow
parameters at subsequent time instances t ( k +1) = t ( k ) + Δt if their values at present time t ( k )
are known. Computing process can be broken into four steps as follows:
1. It is assumed, that x and y are Lagrangian variables, and a cell is a finite particle of fluid
as a solid body which moves under action of external pressure. In this case, particle’s
speed at the next instance of time t ( k +1) can be determined from the vector equation of
Δt
movement: V ′ = V ( k ) − ( k ) ∫ p ( k ) nds , (9.12.4)
ρ τ s
where τ is the sell area; n is the outer normal on the cell boundary.
At the following steps it is supposed that x and y are Euler variables; the cell is a
motionless relative to Euler coordinate, the fluid penetrates through its boundary. Then it
is possible to find a change in quantity of fluid and a pulse in the cell.
2. Calculating the change of mass and a stream of a pulse through the cell boundary:
Δm = −Δt ∫ ρ ( k ) (V ′ ⋅ n )ds , (9.12.5)
s
ΔP = −Δt ∫ ρ ( k )V ′(V ′ ⋅ n )ds . (9.12.6)

s
3. Calculating mass and pulse of fluid in a cell at the following time t ( k +1) :
m( k +1) = m ( k ) + Δm, P ( k +1) = m ( k )V ′ + ΔP . (9.12.7)
( k +1)
4. Calculating all parameters for the subsequent moment of time t :
ρ ( k +1)
=m ( k +1)
/τ , V ( k +1)
=P ( k +1) ( k +1)
/m , p ( k +1)
= f ( ρ k +1 ) . (9.12.8)
A moving free boundary is calculated following the Harlow’s approach using marks, which
are calculated as rm( k +1) = rm( k ) + Vm( k ) Δt . (9.12.9)
A few of marks can be in each cell, and they are given codes (0, 1, 2) accordingly to types of
cell (empty, or full of fluid, or with free boundary,). The next step is a conversion of codes.
Integrals (9.12.4) – (9.12.6) can be linearly approximated using neighboring cells [Terentiev
& Chechnev 1985, 1989].
379
9.12.3. Water Entry of a Plate

Let a plate or a disk touch the free surface of undisturbed compressible water and begin
immersion with constant velocity U. The stream breaks away from the plate edges so that a
ventilated cavity forms past the plate, Fig.9.12.1. The liquid is assumed to be inviscid but
compressible. Thus the normal component of the fluid on the moving plate is to be equal to
U. Pressure on the free surface including the water boundary past the plate is zero. Without
losing generality one can assume that the half-length of plate or the disk radius as well as the
sound speed and density of the undisturbed fluid are unity.
Then all parameters (time t, density ρ , pressure p , fluid
velocity V ) are dimensionless. The sink velocity
corresponds to the Mach number: U = M . Function (9.12.3)
for water looks like
p = ( ρ γ − 1) / γ , γ =7. (9.12.10)
Fig.9.12.1. Water entry of The starting pressure on the plate can be found from the
a plate or disk piston problem in a compressible fluid
p′ = M 1 + 4 M 2 + 2M 2 . (9.12.11)
In all other points the starting pressure should be zero. However, when pressure distribution
on the free surface is step-wise, computation of hydrodynamic parameters at the cell centers
in time have unstable and oscillating character. To avoid the initial oscillations, the pressure
jump (9.12.11) and density are to be distributed linearly at the four adjacent cells.
Numerical calculation can be realized in the orthogonal coordinates. Consider a cell
of number ij and sides Δx and Δy, so that the cell area is equal to τij=ΔxΔx. Equation
(9.12.4) with respect to the x-axis can be written as
Δt Δt
ui′ j = ui( kj ) − ( k ) ( pi(+k1,) j − pi(−k1,) j ) , vi′ j = vi( kj ) − ( k ) ( pi(,kj)+1 − pi(,kj)−1 ) . (9.12.12)
2 ρi j Δx 2 ρi j Δy
Changes of mass and pulse due to a stream through the lower side of cell are calculated by
Δt Δy
Δmi + , j = − (ui′ j + ui′+1, j ) ρ , ΔPi + , j = Δmi(+k,+j1)V ′ , (9.12.13)
2
⎧⎪ ρi( kj ) , if ui′,(jk ) + ui′+( 1,k )j ≥ 0, ⎪⎧Vi ′j , if ui′, j + ui′+1, j ≥ 0,
(k ) (k ) (k )
where ρ = ⎨ ( k ) V ′ = ⎨ (k )
⎪⎩ ρi +1, j , if ui′, j + ui′+1, j < 0, ⎪⎩Vi ′+1, j , if ui′, j + ui′+1, j < 0.
(k ) (k ) (k ) (k )
The stream through other sides is calculated ditto. Calculation procedure and the above
formulas can be applied to an axisymmetric problem as well. The (i,j)-cell for axisymmetric
flow is presented as a ring with radiuses yj- and yj+ ; its cross section is a rectangle with
sides Δx and Δy = y j + − y j − . All the formulas are valid except the one for mass changes via
sides of cells y j + or y j − , where the right-hand term is to be multiplied by 2π y j + or 2π y j − .

Calculations were carried out for a grid with cell sizes Δx = Δy = 0.05 . The time step got out
⎧ ⎫
⎪ Δx Δy ⎪
of Courant’s condition Δt = μ min ⎨ , (k ) ⎬
, (9.12.14)
⎪⎩ max(| ui j | + ci j ) max(| vi j | + ci j ) ⎪
(k ) (k ) (k )
i, j i, j ⎭
380
where ci( kj ) is the speed of sound in cell (i, j) for time t ( k ) ; ui( kj ) and vi( kj ) are components of
speed with respect to the x and y-axes. Courant number μ was taken equal to 0.4. Examples
with a smaller value of μ were calculated as well; the results were virtually the same.
For calculating the free boundaries, the cells attached to the undisturbed free
boundaries were marked initially by 0. At the mth marker, velocity at the cell center xm, ym
were calculated by linear approximation with respect to coordinates x and y. For example for
the (i, j)-cell in the fist quadrant velocities Vi (jk ) , Vi +( k1)j , Vi +( k1)j +1 , Vi (jk+)1 were:
( xm − xi +1 )( ym − y j +1 ) ( xm − xi )( ym − y j +1 )
Vm( k ) = Vij( k ) + Vi +( k1)j +
( xi − xi +1 )( y j − y j +1 ) ( xi +1 − xi )( y j − y j +1 )
(9.12.15)
( xm − xi )( ym − y j ) ( xm − xi +1 )( ym − y j )
+V (k )
i +1 j +1 +V (k )
ij +1 .
( xi +1 − xi )( y j +1 − y j ) ( xi − xi +1 )( y j +1 − y j )
Some of the results obtained in [Terentiev & Chechnev, 1985, 1989] are presented
below. Plotted in Fig.9.12.2 is drag coefficient CD = D / ρ 0U 2 l = D / M 2 (for a plate) and
CD = 2 D / ρ 0U 2π l 2 = 2 D / π M 2 (for a disk) as a function of dimensionless time. The
horizontal line corresponds to resistance coefficient of disk for Kirchhoff model (Cv =0.82).
Lines 1 and 2 correspond to a disk for M = 0.2 and M = 0.4, respectively; line 4 corresponds
to a plate for M = 0.2 . Curves 3 and 5 for disk and plate were calculated for M = 0.2 using
the linear approach [Sagomonyan, 1974]. The linear theory gives understated results for
resistance. The resistance coefficient approaches the asymptote very fast, e.g. for a disk of
radius r = 1 m the coefficient approaches its asymptote in t = 2.5/1445= 0.002 s. It is very
difficult to measure such a short time experimentally. Therefore, Logvinovich [1959]
concluded about independence of resistance coefficient on time of disk entry into water.
Fig. 8.11.2.
Fig.9.12.2. Drag coefficient
Drag coefficient of the
of the plate The pressure
Pressure
Fig. 8.11.3.
Fig.9.12.3. coefficient in
(1, 2) and disc (3-5) as a function ofa
plate (1,2) and disk (3-5) as coefficient
front of the disc atofMach
in front the disk at
number
function
time of
t. the time, t. Mach number, M=0.1, and times,
M=0.1 and times t = 0.5, 0.8, 1.0,
t=0.5, 0.8,
1.5.1, 1.5
Fig.9.12.2a shows also free surfaces during the disk entry at two depths h = 0.4 and
0.6. The dashed curve corresponds to Kirchhoff’s cavity. As seen from Fig.9.12.2,
compressibility of fluid decreases slightly size of cavity
381
The pressure distribution in time over the x-axis of the moving disk at a constant
velocity is plotted in Fig.9.12.3. It is seen that the shock wave is going away and pressure at
the disk center can be negative. Thus, a cavity can arise in front of the disk. At initial time
both for plate and disk, according to pressure (9.12.11), the drag coefficient is given by
CD = 4 + 2 1 + 4M 2 / M . (9.12.16)
A linear approach gives CD = 2 / M . Empirical formula CD = 1.87 + 2.13/ M was
obtained in [Eroshin et al., 1980]. Difference between the calculated and experimental data
can be due to an air cushion formed in the fluid
between the disk and free surface, which lowers the
impact force.
Fig.9.12.4 shows a time dependence of
velocity of a disk falling into water. The initial Mach
number is M=0.1; m=ρb/ρw is the density ratio ρb −
disk density; ρw − initial water density). As seen,
acceleration of the penetrating light body can be
both negative and positive. This is due to a pressure
reduction in front of the body directly after the shock
wave goes away from the solid surface.
Fig.9.12.4.
Velocity of the disk drop as a function of time for
initial Mach number M=0.1 and different m.
Sec.9.13. Smoothed Particle Hydrodynamics1

In spite of the great capability of BEM one should understand that it cannot be used for large
deformationa of free boundaries. More preferable in this case would be the method of
Smoothed Particle Hydrodynamics (SPH). This method was offered for problems of the
astrophysics in [Lucy, 1977] and independently in [Gingold & Monaghan, 1977]. Later the
method was applied for numerical simulations in the hydrodynamics, solid mechanics, etc.
[Monaghan, 1992; Liu et al., 1997; Liu GR & Liu MB, 2003; Morris et al., 1997; Afanasiev
et al., 2006; Cueto-Felgueroso et al., 2009]. The Smoothed Particle Hydrodynamics is a
fully meshless method; it doesn’t require a mesh (set of connected nodes) at any stage of
numerical simulation. This makes it attractively different from the hybrid methods which can
also be applied for solving problems with highly nonlinear deformations of free surfaces, but
requiring time-consuming algorithms for constructing a mesh and defining free surfaces. The
SPH method is a powerful tool for numerical simulation with low time resource
requirements. 2
9.13.1. Governing Equations
Motion of a Newtonian viscous fluid in a certain domain Ω is described by the Navier-
Stokes equations, the continuity equation and some equation of state p = p( ρ ) :
1
This section was prepared using results by K.E.Afanasiev and S.V. Stukolov.
2
The method was developed at Kemerovo State University by K.E. Afanasiev and his colleagues who
applied it to free boundary flow problems. Some results of their studies are presented below.
382
dv n ∂p ∂
= ρ F n − n + μ k (T nk ) ,
ρ (9.13.1)
dt ∂x ∂x
dρ ∂v k
= −ρ k , (9.13.2)
dt ∂x
where n, k = 1, 2, 3 are numerical indices of the coordinates; v n and F n are components of
velocity vector and gravity force, respectively; the viscose stress tensor component is
∂v n ∂v k 2∇ ⋅ v nk
T nk = k + n − δ , δ nk are Kronecker symbols; p and ρ are pressure and density
∂x ∂x 3
of fluid, respectively.
9.13.2. SPH Method
The major idea of the SPH method is a discretization of problem domain Ω by a set of
Lagrangian particles, which can be considered as small liquid volumes and do not require
any connections between each other. Integral approximation is used for the functions,
included in the equations of motion: f ( r ) = ∫ f ( r ′)δ ( r − r ′)dr ′ (9.13.3)
Ω
where δ is the Dirac delta function.
For numerical simulation the δ -function is substituted with a certain function
W ( r − r ′, h ) referred to as the kernel function and having a compact support domain, while
the integral is approximated by the summation [Monaghan, 1992]:
n
m
f s ( r ) = ∑ f i i W ( r − ri , h ), (9.13.4)
i =1 ρi
where ri , mi , ρi are the position vector, mass and density of the i -th particle, respectively,
and n is the number of particles neighboring the i -th one. Two particles i and j are called
neighboring or interacting, if the distance between them doesn’t exceed (hi + hj). The value
of (hi + hj) is a support domain of the kernel function W, and hi is called the smoothing
length of the i -th particle and determines the radius of its interaction with the neighbors.
Usually polynomial splines are used for W. It follows from (9.13.4 that the function gradient
n
m
is expressed as ∇f s ( r ) = ∑ f i i ∇Wr ( r − ri , h ). (9.13.5)
i =1 ρi
Using the above assumptions, the following equations can be derived from (9.13.1)-(9.13.2):
n ⎛ pj ⎞ mj
dvia p v n
= ∑ ⎜ 2i + 2 + Π ij ⎟m j ∇Wra (ri − rj , hi )∑∑ ⎡⎣ ( μiTi da + μ jT jda )∇Wra (ri − rj , hi ) ⎤⎦ (9.13.6)
dt ⎜
j =1 ⎝ ρ i ρj ⎟ ρ ρ
⎠ d = 1 j = 1 i j
d ρi v n
= −∑∑ m j (v dj − vid )∇Wra (ri − rj , hi ) (9.13.7)
dt d =1 j =1
where pi , vi , ui , μi are the pressure, velocity and dynamic viscosity of the i-th particle,
respectively;ν is the dimension of the problem; and Ti is the tensor of viscose stress, the
normal and tangent components of which are found by the formulae:
383
n mj
Ti aa = ∑ ⎡ 2 ( v aj − via ) ∇Wra ( ri − rj , hi ) + ( vbj − vib ) ∇Wrb ( ri − rj , hi ) + ( v cj − vic ) ∇Wrc ( ri − rj , hi ) ⎤ ;
j =1 ρj ⎣ i i i ⎦
(9.13.8)
n mj
Ti ab = ∑ ⎡( v aj − via ) ∇Wrb ( ri − rj , hi ) + ( v bj − vib ) ∇Wra ( ri − rj , hi ) ⎤ , (9.13.9)
j =1 ρj ⎣ i i ⎦
where a ≠ b ≠ c .
For stabilizing the method, especially in simulations with dynamic viscosity μ = 0 , an
additional term Π ij , called artificial viscosity, is included to the right part of equation
(9.13.6) [Brebbia, 1989.]: Π ij = ⎨

(
⎧⎪ −α cijσ ij + βσ ij2 / ρij , vij ⋅ rij ≤ 0 ; )
(9.13.10)
⎪⎩0, vij ⋅ rij > 0,
where σ ij = hvij rij / ( rij2 + ηi2 ) , cij = ( ci + c j ) / 2 ; is average sound speed at the particles i and
j, ηi2 = 0.001hi2 , vij = vi − v j and rij = ri − rj .
To enclose the system of equations (9.13.6)-(9.13.7), one should add an equation of
state, which is mostly used in the SPH method in the form of Theta [Batchelor, 1970]:
⎡ ⎛ ρ ⎞γ ⎤
p = B ⎢ ⎜ ⎟ − 1⎥ . (9.13.11)
⎢⎣ ⎝ ρ 0 ⎠ ⎥⎦
Here B is volumetric expansion; ρ 0 is initial density; γ = 7 is adiabatic exponent. For dam
breaking problems with gravity, coefficient B is selected by the Monaghan’s formula:
200ρ 0 gH
B= (9.13.12)
γ
where g is the gravity acceleration; and H is the initial height of the fluid column.
9.13.3. Kernel Function
Kernel function is very important in the SPH method. Both consistency and stability of the
method depend on the function selection. For numerical simulation using the SPH many
different kinds of the kernel function are applied, beginning with the Gaussian function, and
finishing with splines of different orders. Besides the already known kernel functions one
can develop other ones, but should follow the minimum requirements as
W (r , h) = 0, r > h , ∫ W (r , h) dr = 1 , limW ( r , h ) = δ ( r ) .
h→∞
Ω
where Ω - flow domain; r =| r − ri |, q = r / h .
Besides the above-mentioned requirements, some additional conditions can be
imposed on the kernel function for better stability of the method and higher order
consistency. Such additional conditions and ensuing ways of constructing the kernel
functions result, e.g. in appearance of the RKPM (Reproducing Kernel Particle Method)
[Liu, 1997]. In the problems below the classic Monaghan cubic spline was applied
384
⎧2 / 3 − q 2 + q 3 / 2 at 0 ≤ q ≤ 1;
15 ⎪
[Monaghan, 1992]: W (r , h) = ⎨(2 − q) / 6
3
at 1 ≤ q ≤ 2; (9.13.13)
7π h 2 ⎪
0 at q > 2.
⎩
9.13.4. Conditions of Solid Boundaries
There are many different ways of imposing conditions on solid boundaries, the most
common of which is the virtual particle method. Monaghan used virtual particles, locating
them along the solid boundary in one layer. These particles carry no physical characteristics
unlike inner fluid particles, however, they interact with the latter by means of a certain
interaction potential. The selected potential is included as an additional body force into the
equations of motion. The most common potential used in the Smoothed Particle
Hydrodynamics is the Lennard-Jones potential [Monaghan 1999.13.]:
D ⎡⎛ r0 ⎞ ⎛ r0 ⎞ ⎤
12 6
U ( r ) = ⎢⎜ ⎟ + ⎜ ⎟ ⎥ , (9.13.14)
r ⎣⎢⎝ r ⎠ ⎝ r ⎠ ⎦⎥
where D is the depth of potential well, r0 is interaction radius. The potential is two-
parameter, what allows independently setting any two properties of the fluid. This method of
imposing boundary conditions is applied in the paper for simulation of model problems.
9.13.5. Time Integration
For time integration of the obtained ordinary differential equations (9.13.6)-(9.13.7) the leap-
frog scheme is used [Liu & Liu 2003].
ρin = ρin −1/ 2 + (Δt / 2)(d ρin −1 / dt );
"Prediction": (9.13.15)
vin = vin −1/ 2 + (Δt / 2)(dvin −1 / dt ).
ρin +1/ 2 = ρin −1/ 2 + Δt ( d ρin / dt );
"Correction": vin +1/ 2 = vin −1/ 2 + Δt (dvin / dt ); (9.13.16)
n +1 n +1/ 2
x i = v + Δt (dv
n
i i / dt ).
ρ 1/ 2
i = ρ + (Δt / 2)(d ρ / dt );
i
0
i
0
The first time step: v1/ 2

i = vi0 + (Δt / 2)(dvi0 / dt ); (9.13.17)
xi1 = v 0 + Δt ( dvi1/ 2 / dt ).
The time step is selected based upon the Courant-Friedrichs-Lewy condition [Morris et al ,
lim(hi )
1997]: Δt ≤ C i
, (9.13.18)
lim(ci + vi )
i
where hi , ci , vi – smoothing length, sound speed and velocity of the i -th particle,
respectively. The constant C ∈ (0,1) , the stable calculation is observed if C = 0.3 .
9.13.6. Testing the Method: the Droplet Problem

Consider the droplet problem [Ovsyannikov 1976] for testing the SPH-method. The problem
is formulated as follows: at the initial moment the calculation domain is a circle of the radius
R = 1 m containing the incompressible fluid. Deformation of the circle into an ellipse starts
385
under the influence of initial velocity distribution in the absence of external forces. To
provide incompressibility it is required that the area of the ellipse remains constant, i.e.
ab = 1 throughout calculation, where a and b are semi-axes of the ellipse.
x2
At any moment t the equation of ellipse takes the form: 2
+ a 2 (t ) y 2 = 1 . (9.13.19)
a (t )
A dependence of the semi-axis on the time, a(t ) , can be found by means of integrating the
system of ordinary differential equations: a′ = c; c′ = 2c 2 /(a 5 + a ) (9.13.20)
with the initial conditions: a(0)=1; c(0)=100 (9.13.21)
The initial velocity field for solving the problem by the SPH method is
v1 (0) = xc(0) / a (0); v2 (0) = − yc(0) / a (0). (9.13.22)
The set of equations (9.13.20)-(9.13.21) is integrated using the fourth-order Runge-
Kutta method to obtain the free surface of the ellipse. Comparisons between the results
obtained by the SPH and the Runge-Kutta methods are presented in Fig.19.13. For
numerical simulation of the problem by the SPH method the following values are used:
ρ0 = 1000 kg/m3 (initial density of particles); c0 = 1440 m/s (sound speed); α = 0.1 , β = 0
(constants in the formula of artificial viscosity (9.13.10)). Irregular distribution was used
with equal number of particles neighboring the given one in different directions.
Fig. 9.13.1. shows that the results of calculations by the SPH are in good accordance with
the results obtained by the Runge-Kutta method. The difference between the numerical and
analytical solutions doesn’t
exceed 0.5%. More detailed
Initial location
testing of the method and solution t = 0.01s
of some model problems are
provided in [Afanasiev et al.,
2006].
Fig.9.13.1.
Droplet problem
9.13.5. Examples of Numerical Computations

For numerical simulation of model problems the following values of parameters are used:
c0=1440 m/s (sound speed); α=0.1 α = 0.1 , β=0 (constants for artificial viscosity); r0=dx
(interaction radius for the Lennard-Jones forces), where dx - initial distance between the
particles; D=0.049 (depth of the potential well). The time step is selected using the CFL
condition (9.13.18). For imposing the solid boundaries conditions the Monaghan virtual
particles are used.
Dam breaking. In this problem the fluid column of density ρ 0 = 1000 kg/m3 , viscosity
μ = 10−3 kg/(m ⋅ s) and the zero initial velocity field starts breaking under gravity at t = 0 .
Fig.9.13.2 presents flow patterns for different moments with 900 particles, used for
simulation of liquid column.
Fluid-fluid interaction. Two fluid columns located at opposite sides of a tank start breaking
under gravity at the moment t = 0 . In a certain moment the fluid flows come into collision
386
and start interacting. The dynamic viscosities of the both fluids are the same and equal to
kg/(m⋅s). The initial densities of the fluids are ρ01 =1000 kg/m3 and ρ02 =2000 kg/m3. In
the calculation the fluid columns are simulated with 900 particles each, the solid boundary
consists of 661 Monaghan virtual particles.
Fig.9.13.2. Initial t = 0.3s t = 1.3s

Dam location
breaking
(N = 900).
Fig.9.13.3 presents the flow patterns at different times. The obtained results show that the
SPH method is an efficient tool for numerical simulation of multiphase fluid flows.
Initial location t = 0.1s t = 0.25s

t=0
Fig.9.13.3. Fluid-fluid interaction.
Cavity collapse. The problem of a cavity flapping on a free surface is of interest because
during the flapping process a cumulative jet of fluid with high velocity and complex
geometry is formed. This phenomenon becomes difficult for mathematical modeling. The
problem was first formulated by Lavrent’ev and investigated by Kedrinskij with the EGDA
method [Kedrinskij & Lavrent’ev, 1983]. Cumulating effects arising during underwater
explosions were discussed in the review [Kedrinskij, 2000]. The applicability of the
boundary element method on the basis of third Green formula for solving these problems
was considered in [Terentiev & Afanasiev 1987]. Despite the long history, this problem is
still of interest to many researchers worldwide [Kedrinskij 2000; Tuck 2000].
Below, the three-dimensional problem is considered. At the initial moment a
semicircular cavity is located on the free surface of liquid in a tank with initial density
ρ=1000 kg/m3 and dynamic viscosity μ=0 kg/(m⋅s). The cavity starts collapsing under
gravity at initial time t = 0 . The results of numerical simulation are provided for 9625
particles. To verify reliability of the obtained results we compared them with those obtained
using Boundary Element Method. Fig.9.13.4 presents a comparison of the numerical
simulation results: the particles whose location is found by the SPH method (grey dots); the
boundary nodes found by the Boundary Element Method – black dots.
Drop falling. The problem is formulated as follows: a circular drop 0.01 m in diameter is
falling into the tank with a fluid having the initial velocity 2 m/s. The initial density of the
drop and the fluid in the tank is ρ01 =1000 kg/m3 , the viscosity μ=10–3 kg/(m⋅s).. A series of
387
calculations was performed for different number of particles and different time steps. Fig.
9.13.5 presents the flow patterns at different moments.
Initial location
t = 0.6s t = 1s
Fig.9.13.4. Cavity collapse.
Dam breaking with inclined boundary. The given problem is one of classic model of free
surface problems, which is used for verification of Lagrangian particle methods. The solid
boundary geometry and the initial position of the fluid column are presented in Fig.9.13.6.
The calculations are performed for 2500 particles. The same problem has been considered
using another method in [Cueto-Felgueroso et al., 2009]. For comparison, both results are
depicted on Fig.9.13.6. One can see from Fig. 19 a good coincidence between both methods.
t=0.0018c t=0 0297s
t=0.1017s
Fig.9.13.5.
Drop falling.
Fig.9.13.6. Fluid spreading over an inclined plane, as in the left-hand sketch, and the flow
simulation by the above calculations at t=0.208 s and using the Cueto-Felgueroson’s
method.
388
Ch. 10. Method of Singularity in Cavitating Flow
Ch. 10. Method of Singularities in Cavitating Flow

In the previous chapter were presented main numerical methods used to computing
cavitating flows in plain as well as in axisymmetric statements of value problems. The
numerous examples have been evidence of their effectiveness. In spite of the fact that the
above-mentioned methods allow computing of wide frame of cavitation theory some
problems as three-dimensional or unsteady flow have been investigated not so completely. In
this chapter using method of singularity some new problems of cavitating flow are
considered. The method of singularity has been widely used just at the beginning of last
century, but it may be efficiency applied at present too, see [Belotserkovskii & Nisht, 1978,
Efremov, 1974]. Some new results have been obtained by many others authors.
Sec. 10.1. Velocity-Based Method for Cavitating Flow past Thin Foils
10.1.1. Integral Equations
The numerical approach to two-dimensional linear theory of partially and super cavitating
hydrofoils is generally based on the representation of the disturbance velocity on the foil,
(u,v), as a distribution of sources and vortices along the chord of the foil and the extent of the
1 q (ξ ) 1 γ (ξ )
l c
cavity where u ± ( x ) = ∓ 12 γ ( x ) + ( ) ( )
2π ∫0 x − ξ 2π ∫0 x − ξ
d ξ , v ±
x = ± 1
2 q x + dξ (10.1.1)
and γ = u+−u−, q = ν+−ν−, are the vortex and source distributions, respectively, Fig.10.1.1.
For a hydrofoil with thickness t(x), camber η(x) and angle-of-attack α, the linearized
kinematic boundary condition on the wetted part of the boundary is
v ± = Uf ′± (10.1.2)
where f ± ( x ) = ± 12 t ( x ) + η ( x ) − α x and the linearized dynamic boundary condition on the
cavity is u ± = 12 U σ , (10.1.3)
where σ is the cavitation number and U is the free stream speed. Employing these
boundary conditions in equations (10.1.1) yields the coupled singular integral equations
1 q (ξ ) 1 γ (ξ )
l c
( ) ( )
2π ∫0 x − ξ 2π ∫0 x − ξ
1
2
U σ = ∓ 1
2
γ x + d ξ , v ±
= ± 1
2
q x + dξ . (10.1.4)
l
Fig.10.1.1.
Discretized partially
c
cavitating hydrofoil. xγ xq
An auxiliary condition must be enforced that the net source strength for the hydrofoil-cavity
l
system must be zero or ∫ q (ξ ) dξ = 0 .
0
(10.1.5)
This condition guarantees closure of the hydrofoil-cavity system. Eqs. (10.1.4) and (10.1.5)
uniquely determine the flow for either partial cavitation (l<c) or supercavitation (l>c).
389
10. Method of Singularity in Cavitating Flow
10.1.2. Discretization of Integral Equations

Note that in the analytical solution of the above equations, a Kutta condition that γ(c)=0
must be imposed to ensure uniqueness. However, the numerical method described below
does not require this. The positioning of the singularities manages to enforce this condition
by itself. Note also that the kinematic boundary condition is not enforced on the cavity. This
is the usual approximation for linear theory where the kinematic boundary condition is
employed subsequent to the solution to obtain the cavity shape.
For sharp edged foils these equations may be solved using a variety of techniques,
the simplest of which is the point source-vortex method.
In the point source-vortex method when applied to a partially cavitating hydrofoil,
the regions 0 ≤ x ≤ l and l ≤ x ≤ c are divided up into elements of roughly the same size. In
each of these elements a point vortex is placed on quarter of the way back from the upstream
end and a point source is placed three quarters of the way back*, Fig.10.1.1. Equations
(10.1.4) and (10.15) may then be expressed in discreet form where the kinematic boundary
conditions are satisfied at the source points and the dynamic boundary conditions are
satisfied at the vortex points to yield
1 qj
1
2
U σ = − 12 γ i + ∑ γ
2π j xi − x j
q
Δx j on y = 0+ , 0 ≤ xiγ ≤ l
1 γj
v + = + 12 qi +
2π
∑x j
q
− xj γ
Δx j on y = 0+ , l ≤ xiq ≤ c
i
γj (10.1.6)
1
v − = − 12 qi +
2π
∑x j
q
− xj γ
Δx j on y = 0− , 0 ≤ xiq ≤ c
i
⎡ q j ( 12 Δx j + 14 Δx j −1 ) + q j −1 ( 14 Δx j ) ⎤
0 = ( 43 ) 4 q1Δx1 + ∑ ⎣ ⎦ Δx
3
j =2 ( 4 Δx j + 4 Δx j −1 )
3 1 j
where Δxj is the length of the jth element and use has been made of the fact that q(x) ~ x–1/2
as x→0. As stated, the cavitation number, σ , is given and the cavity length, l, is to be found.
However, it is easier to assume that the cavity length is given and the cavitation number is to
be found. In that case the above equations become
1 qj
0 = − 12 U σ − 12 γ i + ∑ γ
2π j xi − x j
q
Δx j on y = 0+ , 0 ≤ xiγ ≤ l
1 γj
v + = + 12 qi +
2π
∑x
j
q
− xγj
Δx j on y = 0+ , l ≤ xiq ≤ c
i
γj (10.1.7)
1
v − = − 12 qi +
2π
∑x
j
q
− xj γ
Δx j on y = 0− , 0 ≤ xiq ≤ c
i
⎡ q j ( 12 Δx j + 14 Δx j −1 ) + q j −1 ( 14 Δx j ) ⎤
0 = ( 43 ) 4 q1Δx1 + ∑ ⎣ ⎦ Δx
3
j =2 ( 4 j 4 j −1 )
3
Δ x + 1
Δ x
j
*
I.I. Efremov (1974) offered preliminarily change of variables, ξ =τ, x = z .
390
If there are N elements on the foil-cavity boundary, then the above equations yield a
(2N+1)x(2N+1) matrix equation for the quantities γ i , qi i = 1,…, N and σ .
10.1.3. Hydrodynamic and Geometric Characteristics

Once the solution has been obtained the various quantities of interest, i.e. lift, drag, moment,
cavity volume and cavity shape may be obtained. Since the linearized pressure is given by
p = p∞ − ρUu we find that the lift and moment are
c c
L = − ∫ ( p + − p − ) dx = ρU ∫ γ dx = ρU ∑ γ j Δx j ,
o o j
c c
(10.1.8)
M = − ∫ x ( p − p ) dx = ρU ∫ xγ dx = ρU ∑ x j γ j Δx j .
+ − γ
o o j
The cavity drag may be obtained by taking the component of the pressure in the direction of
the flow yielding D = 12 ρ ∑ ⎡⎣ q j ( u +j + u −j ) + γ j ( f '+j + f '−j ) ⎤⎦ . (10.1.9)
j∈0 < x < c
The cavity shape may be obtained by employing the kinematic boundary condition on the
cavity surface which states that h′± ( x ) = v ± ( x ) . Then for the partially cavitating case
x
h + ( x ) = ∫ v (ξ , y = 0+ ) d ξ =
0
. (10.1.10)
⎡ v +j ( 12 Δx j + 14 Δx j −1 ) + v +j −1 ( 14 Δx j ) ⎤
J
= ( 43 ) v Δx1 + ∑ ⎣ ⎦ Δx , 0 ≤ x ≤ l
3
4 +
1
j =2 ( 4 Δx j + 4 Δx j −1 )
3 1 j
where the same integration scheme has been employed as that used for the closure condition.
For the supercavitating case we have similarly
x
h + ( x ) = ∫ v (ξ , y = 0+ ) d ξ =
0
⎡ v +j ( 12 Δx j + 14 Δx j −1 ) + v +j −1 ( 14 Δx j ) ⎤
J
=( ) v Δx1 + ∑ ⎣ ⎦ Δx , 0 ≤ x ≤ l
3
4 4 +
3 1
j =2 ( 4 Δx j + 4 Δx j −1 )
3 1 j
x
(10.1.11)
h −
( x ) = h ( c ) + ∫ v (ξ , y = 0
− −
) dξ =
c
⎡ v −j ( 12 Δx j + 14 Δx j −1 ) + v −j −1 ( 14 Δx j ) ⎤
J
= h (c) + ∑ ⎣
− ⎦ Δx , c ≤ x ≤ l
j = Jc ( 34 Δx j + 14 Δx j −1 ) j
Similarly, the cavity volume for the partially cavitating case may be calculated from
l
V = ∫ ( h + (ξ ) − f + (ξ ) ) d ξ . (10.1.12)
0
c l
and for the supercavitating case V = ∫ ( h + (ξ ) − f + (ξ ) ) d ξ + ∫ ( h + (ξ ) − h − (ξ ) ) d ξ (10.1.13)
0 c
Figs.10.1.2–10.1.7 present representative results for the solutions obtainable through
391
the above computational approach for both partially cavitating and supercavitating cases.
Fig.10.1.2 shows the solution quantities q and γ versus x for a supercavitating flat plate at 4°
angle-of-attack and a cavity length of 2.0 for which the cavitation number is 0.133 showing
the singularities in both q and γ at the leading edge and in q at the end of the cavity as well as
the approach of γ to zero 0.60
at the trailing edge of the
hydrofoil. 0.40 q
ℵ
Fig.10.1.2. 0.20
Solution for a
q, ℵ
0.00
supercavitating flat plate
at 4° angle-of-attack and ‐0.20
cavity length 2.0.
‐0.40
Fig.10.1.3 shows the
cavity shape for this case ‐0.60
(note the exaggerated 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00
vertical scale) and x
Fig.10.1.4 shows the negative of the pressure coefficient with the stagnation point at the
leading edge and the 0.04
smooth transition to the 0.02
cavity pressure at the
0.00
trailing edge of the y+
hydrofoil. ‐0.02 y-
y
yfoil
Fig.10.1.3. ‐0.04
Cavity shape for a ‐0.06
supercavitating flat plate ‐0.08
at 4° angle-of-attack and ‐0.10
cavity 0.00 0.50 1.00 1.50 2.00
length 20. x
0.20
0.00
Fig.10.1.4.
Pressure distribution for ‐0.20
a supercavitating flat
-cp
‐cp+
plate at 4° angle-of- ‐0.40
‐cp‐
attack and cavity length ‐0.60
2.0.
‐0.80
Fig.10.1.5 shows ‐1.00

the solution quantities q 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00
and γ versus x for a x
partially cavitating 10%
thick biconvex hydrofoil at 4° angle-of-attack and a cavity length of 0.5 for which the
cavitation number is 0.695 showing the singularities in q and γ at the leading edge and the
singularities and discontinuities in q and γ at the end of the cavity.
392
Fig.10.1.5. 2 .5 0
Solution for a partially 2 .0 0

q
q
cavitating biconvex foil 1 .5 0

γ
of 10% thickness at 4° 1 .0 0
angle-of-attack and q 0 .5 0
q,
cavity length 0.5. γ 0 .0 0
-0 .5 0
-1 .0 0
Figs.10.1.6 and 10.1.7 -1 .5 0
show the shape of the 0 .0 0 0 .2 0 0 .4 0 0 .6 0 0 .8 0 1 .0 0
cavity on the biconvex x
hydrofoil and the negative 0.12

of the pressure coefficient 0.08
for the same case. 0.04
Fig.10.1.6. 0.00
Cavity shape for a partially y
-0.04
cavitating biconvex foil of -0.08
10% thickness at 4° angle-
-0.12
of-attack and cavity length
-0.16
0.5. 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
x
Fig.10.1.7. 1.00
Pressure distribution for a 0.75
partially cavitating 0.50
biconvex foil of 10% 0.25
thickness at 4° angle-of- 0.00
-cp
attack and cavity length 0.5 -0.25

-0.50 -cp+
The pressure -0.75 -cp-
coefficient shows stagnation -1.00
points at the leading edge -1.25
on the pressure side, the 0.00 0.20 0.40 0.60 0.80 1.00
trailing edge of the cavity x
on the suction side and at
the trailing edge of the hydrofoil as expected.
10.1.4. Alternate Approach
An alternate approach to the solution of equations (10.1.6) is to forego the requirement of
zero net source strength and take the cavitation number as given, but the cavity length to be
determined. In this approach, the cavity closure may be defined as
l
δ = h + ( l ) − f + ( l ) = ∫ v (ξ ,0+ ) d ξ (10.1.14)
0
for the partially cavitating case, where f + ( x ) denotes the upper surface of the foil, and
l
δ = h + ( l ) − h − ( l ) = ∫ ⎡⎣ v (ξ ,0+ ) − v (ξ ,0− ) ⎤⎦ d ξ (10.1.15)
0
393
for the supercavitating case. The solution approach is then to assume a cavity length for the
specified cavitation number, solve the problem and compute δ. If δ is not zero then the
cavity length is altered and the problem solved again. Iterating in this way the solution
procedure becomes a root finding problem for δ (l)=0, where δ (l) is determined by the
solution of the problem. This approach becomes essential in three dimensional problems.
10.1.5. Leading Edge Corrections for Round-Nosed Hydrofoils
The approaches discussed above apply strictly to sharp edged hydrofoils. Linear theory fails
in the vicinity of the leading edge of a round nosed foil due to the failure of the assumption
that u, v << U in that region. For fully wetted foils a leading edge correction due to
Lighthill [1951] yields the development of a uniformly valid approximation to the velocities.
This approach expresses the uniformly valid surface velocity in terms of the streamwise
x
disturbance velocity as q ± = (U + u ± ) , (10.1.16)
x + 2 ρ LE
1
where ρ LE is the radius of the leading edge of the foil. Kinnas [1991] applied this to the
cavitating problem and noted that, over the cavity, q = U 1 + σ we find that we can write
⎡ x + 12 ρ LE ⎤ ⎡ x + 12 ρ LE ⎤
u+ ( x) = U ⎢ 1 + σ − 1⎥ ≈ U ⎢(1 + 12 σ ) − 1⎥ . (10.1.17)
⎣⎢ x ⎦⎥ ⎣⎢ x ⎥⎦
The dynamic boundary condition may then be written as
⎡ x + 12 ρ LE ⎤ 1
c
q (ξ )
U ⎢(1 + 12 σ ) − 1⎥ = − 12 γ ( x ) + ∫ x − ξ dξ on y = 0+ , 0 ≤ x ≤ l , or (10.1.18)
⎣⎢ x ⎥⎦ 2π 0
1
c
q (ξ ) x + 12 ρ LE ⎡ x + 12 ρ LE ⎤
− 12 γ ( x ) + ∫0 x − ξ d ξ − 12 U σ
= U ⎢ − 1⎥ on y = 0+ , 0 ≤ x ≤ l (10.1.19)
2π x ⎢⎣ x ⎥⎦
which may be solved as before for the unknowns γ ( x ) , q ( x ) and σ . Note that this
approach assumes that the leading edge radius of the cavity foil system is the same as that of
the foil alone. This assumption may be justified theoretically. The detachment point of the
cavity from the foil must be specified as part of the solution. A suitable approximation is to
select the cavity detachment point as the location of the pressure minimum in the fully
wetted flow. The cavity closure condition that the net source strength is zero must be
modified in this case from that employed in the case of a sharp edged hydrofoil to take
−1
account of the fact that the source strength now behaves as q ( x ) ∼ x 2
at x → 0 yielding as
⎡ q j ( 12 Δx j + 14 Δx j −1 ) + q j −1 ( 14 Δx j ) ⎤
a closure condition 0 = 3 q1Δx1 + ∑ ⎣ ⎦ Δx . (10.1.20)
j =2 ( 4 Δx j + 4 Δx j −1 )
3 1 j
Similar integration considerations apply to the computation of the cavity shape.

Note, as shown in Fig.10.1.8, that without the leading-edge correction the cavity
length increases with increasing section thickness, while with the inclusion of the leading-
edge correction the cavity length decreases with increasing section thickness, in addition to
the cavities being considerably smaller with the leading edge correction. This trend is
consistent with results found in fully non-linear partial cavitation models.
394
Fig.10.1.8.
Comparison of cavity extents
and shapes with and without
the leading edge correction at
an angle-of-attack of 4° and a
cavitation number of 0.90 for
NACA 16 series sections of 6,
9 and 12% thickness ratios.
Solid lines are with the
leading edge correction and
dashed lines are without the
leading edge correction.
Note, as shown in Fig.10.1.8,

that without the leading-edge
correction the cavity length
increases with increasing
section thickness, while with
the inclusion of the leading-
edge correction the cavity
length decreases with
increasing section thickness,
in addition to the cavity being
considerably smaller. This
trend is consistent with results
found in fully non-linear
partial cavitation models.
Sec. 10.2. Nonlinear Velocity-Based Method for Cavitating Hydrofoils

In order to go beyond the limitations of the 2D linear theory, a nonlinear cavitation model is
required wherein the cavity shape is determined as part of the solution process. This was
first done by Uhlman [1983; 1987; 1989] using a velocity based approach. In this approach
the velocity field in the fluid is represented by a distribution of vortices on the hydrofoil-
cavity boundary such that the disturbance velocity is given by
1 ⎧⎪
γ ( s ) ⎨i
( y −η ) (x −ξ ) ⎫⎪
v=
2π C∫ ( x − ξ )
2
+ ( y − η )
2
−j
( x − ξ )
2
+ ( y − η )
2 ⎬
ds , (10.2.1)
⎩⎪ ⎭⎪
where the integral is over contour C given by (ξ ( s ) ,η ( s ) ) , see App. 6 and Fig.10.2.1.
The kinematic boundary condition in this case is n ⋅ v = −n ⋅ U on C, (10.2.2)
395
where n is the unit surface normal vector and U is the freestream velocity so that the total
velocity is given by U + v . For reasons that will be discussed shortly we note that the
kinematic condition may also be expressed as t ⋅ v − = −t ⋅ U , (10.2.3)
−
where v denotes the disturbance
velocity on the interior of the foil-
cavity boundary, after [Uhlman,
1983; 1987].
Fig.10.2.1.
Partially cavitating hydrofoil, after
[Uhlman, 1983; 1987].
The dynamic boundary condition

may be derived from Bernoulli’s
equation applied to the cavity boundary to obtain p∞ + 12 ρU 2 = pc 12 ρ qc2 , (10.2.4)
where U = U and qc is the speed of the fluid on the cavity boundary. Employing the usual
definition of the cavitation number, σ = ( p∞ − pc ) 1
2 ρU 2 , then yields
qc2 = U 2 (1 + σ ) , on Cc. (10.2.5)
This equation is still nonlinear however. In order to obtain a linear expression of the
dynamic boundary condition we note that qc = t ⋅ ( U + v ) on the cavity and take the
appropriate root of equation (10.2.5) to obtain the condition
t ⋅ v = U 1 + σ − t ⋅ U , on Cc. (10.2.6)
This expression is now exact and linear in the quantities t ⋅ v and 1 + σ . These boundary
conditions may now be combined with equation (10.2.1) to obtain
−
1 ⎧⎪ t x ( x, y )( y − η ) − t y ( x, y )( x − ξ ) ⎫⎪
∫ ⎨
γ ( s ) ⎬ ds = −t ( x, y ) ⋅ U (10.2.7)
2π ( x − ξ ) + ( y −η )
2 2
C ⎩⎪ ⎭⎪
+
1 ⎧⎪ t x ( x, y )( y − η ) − t y ( x, y )( x − ξ ) ⎫⎪
and a
2π ∫C γ ( s ) ⎨ ( x − ξ ) + ( y −η )
2 2 ⎬ ds − U 1 + σ = −t ( x, y ) ⋅ U , (10.2.8)
⎪⎩ ⎪⎭
where we have employed the alternate kinematic boundary condition and the ± superscripts
denote evaluation on the exterior or interior of the boundary respectively. The reason for
employing the alternate kinematic boundary condition is that the standard form results in a
poorly conditioned matrix, while the alternate form yields a well conditioned matrix when
discretized. In principle, the kinematic boundary condition, equation (10.27), should be
applied on all portions of the foil-cavity boundary. However, since the cavity shape is
initially unknown we again employ the strategy of applying only the dynamic boundary
condition on the cavity boundary.
These equations, combined with a Kutta condition and a cavity termination model
may be solved to obtain an update of the surface vorticity distribution and cavity shape. The
kinematic boundary condition over the cavity is then employed to update the cavity shape
and the problem is again solved with this new geometry.
396
10.2.2.Discretization and Solution of the Integral Equations

These equation may be solved by discretizing the surface vorticity distribution into flat
elements or panels of piecewise constant surface vorticity. Denoting the Green’s function by
G=i
( y −η ) −j
(x −ξ ) (10.2.9)
( x − ξ ) + ( y −η ) ( x − ξ ) + ( y −η )
2 2 2 2
N
1
∑γ t ⋅ ∫ G
−
the discretized equations take the form ds = −t ⋅ U (10.2.10)
2π
j
j Cj
N
1
∑γ t ⋅ ∫ G
+
and ds − U 1 + σ = −t ⋅ U , (10.2.11)
2π
j
j Cj
where Cj denotes that portion of the contour corresponding to the local panel and N is the
number of panels and the equations are imposed at the mid-points of the panels. The
integrals in the above equation may be evaluated analytically.
Since the problem has been posed in terms of a distribution of surface vorticity only,
there is no need for a closure condition of zero net source strength. In its place is a cavity
termination model. The model employed originally by Uhlman 1983 was a modified
Riabouchinsky wall model wherein an impermeable wall is dropped from the aft end of the
cavity to the surface of the hydrofoil. This wall is also broken up into panels on which the
kinematic boundary condition is imposed (see Fig.10.2.2). The panels on this wall may then
be considered to be part of the wetted portion of the foil-cavity boundary.
Fig.10.2.2.
Modified Riabouchinsky wall cavity
termination model, aftr [Uhlman, 1983;
1987].
The Kutta condition supplies the extra

condition necessary to yield N+1 equations for
the N + 1 unknowns, γ j , j = 1,… , N and 1 + σ . This condition may be expressed as
stating that the velocities at the trailing edge (TE) as approached from the upper and lower
surfaces must be equal or lim t A ⋅ ( U + v A ) = lim t B ⋅ ( U + v B ) , (10.2.12)
A→TE B →TE
where A and B denote points on the upper and lower surfaces of the foil respectively. This
condition is imposed numerically by extrapolating the trailing edge velocities on the upper
and lower surfaces from the two points adjacent to the trailing edge and setting the values
equal.
These equations are solved starting from an assumed cavity shape to yield a surface
vorticity distribution from which the velocities may be obtained fro the discreet form of
1 N
equation (10.2.1) as v= ∑ γ j Gds
2π j C∫j
(10.2.13)
These velocities are then used with the kinematic boundary condition in the form
dh V + v
= , (10.2.14)
dx U + u
397
where U = (U ,V ) and v = ( u , v ) , to integrate the cavity shape from its detachment point to
its end at the specified cavity length. A new Riabouchinsky wall is then erected connecting
the end of the cavity with the foil and the problem is reformulated with this new geometry
and solved again. This process is repeated until convergence is achieved, typically on the
order of 10-20 iterations. At convergence the cavity shape is known and both the kinematic
and dynamic boundary conditions are satisfied on the cavity.
Note also that no mention has been made of the cavity detachment point. If the
hydrofoil is sharp edged, this salient edge is taken as the point of detachment of the cavity.
If the leading edge of the hydrofoil is rounded, then additional considerations come into
play. In the case of ideal flow the detachment point would be chosen to satisfy the Brillouin-
Villat condition of tangential detachment. However, it has been shown (see e.g. Arakeri
1975) that viscous effects are important in the location of the detachment point. As a
reasonable approximation the point of pressure minimum in the fully-wetted condition may
be employed, as has been done in the figures that follow.
A similar approach may be taken for supercavitating hydrofoils. The main difference
there is in the nature of the Kutta condition, since only half the trailing edge is wetted and in
the Riabouchinsky wall, which must now span the gap between the end of the upper cavity
boundary and the end of the lower cavity boundary
The forces acting on the body are determined by computing the pressure acting at each panel
using Bernoulli’s equation and performing integration time the local unit normal as
F = ∫ npds (10.2.15)
C
the only caveat being that the integration must be over the original hydrofoil surface. Thus
any contribution from the Riabouchinsky wall is neglected and the cavity drag may be
obtained. An alternate way to obtain the forces is to evaluate the integral
F = ∫ n ( p − pc ) ds (10.2.16)
C
where the integral need only be computed over the wetted portion of the hydrofoil since the
portions of the boundary under the cavity will contribute nothing. The cavity shape is
determined as part of the solution process and the cavity volume is readily found as the area
between the cavity boundary and the foil.
Results of this approach to
the computation of both partially
cavitating and supercavitating
hydrofoils are presented in
Fig.10.2.3-10.2.10.
Fig.10.2.3.
Partially cavitating NACA 16-006,
α =4 degrees, l c =0.50, final
cavity shape after 15 iterations,
after [Uhlman 1983; 1987].
398
Fig.10.2.4.
Partially cavitating NACA 16-006,
α =4 degrees, l c =0.50, negative
pressure coefficient, after [Uhlman,
1983; 1987].
Fig.10.2.5. Cavity length l/c versus α/σ Fig.10.2.6. Meijer’s 1959 experimental data
NACA 16 series sections with varying for a 4% thick bi-convex foil at 2, 4 and 6
thickness at 4 degrees angle-of-attack degrees angle-of-attack versus the linear
[Uhlman, 1983; 1987]. theory of Geurst [1959] and the nonlinear
theory of Uhlman [1983, 1987].
Fig.10.2.7.
Converged cavity shape
for a supercavitating flat-
plate hydrofoil at 4
degrees angle-of-attack
with a cavity length of
1.40 after 25 iterations,
after [Uhlman, 1983;
1989].
399
Fig.10.2.8.
Pressure distribution over a
supercavitating flat-plate
hydrofoil at 4 degrees angle-
of-attack with a cavity length
of 1.40 after 25 iterations,
after [Uhlman, 1983; 1989].
Fig.10.2.9. Cavity length l/c (left-hand plot) and normalized lift coefficient Cl (πα 2 )
(right-hand plot) versus α/σ for a supercavitating flat-plate hydrofoil with the theoretical
predictions of Geurst [1960] and Uhlman [1983; 1989] and the experimental data of Wade
and Acosta [1966], after [Uhlman, 1983; 1989].
Sec. 10.3. Velocity-Based Method for 3D Cavitating Flow past Thin Foils
In three dimensional linear theory the disturbance velocity field u may be represented as
u = ∫∫ γ × GdS + ∫∫ qGdS (10.3.1)
S S
where γ is the surface vorticity vector tangent to the surface of the hydrofoil S , q is the
source strength and G ( x, ξ ) denotes the velocity induced at the field point x by a unit point
source located at ξ . The kinematic boundary condition takes the usual form
n ⋅ u = −n ⋅ U (10.3.2)
where U is the freestream velocity with magnitude U and n is the local unit normal vector.
400
( U + u ) = U (1 + σ )
2
The dynamic boundary condition becomes (10.3.3)
Denoting by s and t the unit vectors tangent to the surface in the chordwise and
spanwise directions respectively (and assuming orthogonality of s and t for simplicity of
exposition) the dynamic boundary condition may be rewritten as
s ⋅ u = U (1 + σ ) − ⎡⎣t ⋅ ( U + u ) ⎤⎦ − s ⋅ U
2
(10.3.4)
and the equations to be solved become −n ⋅ U = n ⋅ ∫∫ γ × GdS + n ⋅ ∫∫ qGdS on Swetted (10.3.5)
S S
U (1 + σ ) − ⎡⎣ t ⋅ ( U + u ) ⎤⎦ − s ⋅ U = s ⋅ ∫∫ γ × GdS + s ⋅ ∫∫ qGdS on Scavity

2
and (10.3.6)
S S
where on Swetted and Scavity denote the wetted and cavity portions of hydrofoil-cavity
boundary respectively. Note that in equation (10.3.6) the quantity t ⋅ u is not known a priori
and hence an iterative solution scheme will be required. Note also that the cavitation number
is a specified quantity and the span-wise distribution of cavity extent is to be determined.
10.3.2. Discretization and Solution of the Integral Equations
To solve these equations an extension of the two-dimensional point source-vortex method to
three dimensions may be employed called the Vortex Lattice Method (VLM). In this
approach the mean surface of the hydrofoil is discretized into quadrilateral panels in both the
chordwise and spanwise directions. In each of these panels a line source segment is placed
at three-quarters of the panel chord and a line vortex segment is placed at one-quarter of the
panel chord. The vortex segment is augmented with streamwise vortex segments at each end
which progress downstream and together constitute a single horseshoe vortex as shown in
figure 10.3.1. These horseshoe vortices, consisting of the cross-stream or bound portion and
the streamwise or trailing portions, are introduced to automatically satisfy the requirement
that vorticity can neither begin nor end in the fluid. In a manner similar to the two
dimensional case, the kinematic boundary conditions are satisfied at the mid-span of each
source segment and the dynamic boundary conditions are satisfied at the mid-span of each
bound vortex segment. The kinematic and dynamic boundary conditions then become
± 12 qi + ∑ γ j n i ⋅ G γij + ∑ q j ni ⋅ G ijq = −n i ⋅ U on Swetted (10.3.7)
j j ≠i
and ∓ 12 γ i + ∑ γ j ni ⋅ G γij + ∑ q j ni ⋅ G ijq = U (1 + σ ) − ⎡⎣ t ⋅ ( U + u ) ⎤⎦ − s ⋅ U on Scavity (10.3.8)

2
j ≠i j
applied at the appropriate control points where γi is now the circulation about the i-th
horseshoe vortex, qi is the source strength of the i-th source segment, G γij is the velocity
induced at control point i by the j -th horseshoe vortex and G ijq is the velocity induced at
control point i by the j-th source segment.
The major new hurtle to be overcome in the extension of linear methods to three
dimensions is the determination of the cavity extent. The cavity extent is no longer described
by one number, but rather by a distribution over the span of the hydrofoil. As in the two
dimensional case the solution of the problem and the updating of the cavity shape yields a
cavity closure quantity δj where j is the span-wise index. The problem then becomes to
determine the cavity lengths lj over the span such that δj=0, ∀j. In principle the cavity
401
closure at any spanwise location will depend on the cavity lengths at that spanwise location
and all others. In practice, it is found that assuming that δ j only depends on l j (essentially
employing only the diagonal of the Jacobian matrix) yields a scheme that converges rapidly.
Another feature that has been found to be useful in the three dimensional problem is
the use of “split panels” wherein the kinematic and dynamic boundary conditions are
applied over portions of a given panel,
thereby obviating the need to regrid at
each iteration.
Fig.10.3.1. Horseshoe
Elements of a vortex lattice model, vortices
showing mesh, horseshoe vortex
segments, source segments, kinematic dbc control point
t
boundary condition (kbc) control points,
dynamic boundary condition (dbc) mesh s
control points and surface tangent
vectors s and t . Source segments
10.3.3. Hydrodynamic and

Geometric Characteristics
The forces acting on the foil are usually
computed by determining the total
velocities at each panel and employing kbc control point
Bernoulli’s equation to yield pressures
which are then integrated over the foil.
Alternate approaches using the Kutta-Joukowski and Lagally theorems may also be
employed. Computation of the cavity volume follows readily from the method discussed in
two dimensions.
Sec. 10.4. Nonlinear Potential-Based Method for Cavitating 2D

Hydrofoils
10.4.1. Integral Equations for Cavitating Hydrofoils with Reentrant Jet Cavity
Closure
Potential based methods employ Green's third identity which, in two dimensions, has the
⎧ ∂φ ∂G ⎫
form βφ = ∫ ⎨ G − φ ⎬ ds (10.4.1)
∂S ⎩
∂n ∂n ⎭
where φ is the disturbance potential such that the total potential is given by Φ=Ux+φ, ∂S is
the integration contour enclosing the domain S which is entirely within the fluid (see
Fig.10.4.1), and G ( x, y; ξ ,η ) = − ln ( R ) where R = (x −ξ ) + ( y − η ) is the 2D free space
2 2
⎧2π , ( x, y ) ∈ S
⎪
Green’s function and β = ⎨π , ( x, y ) ∈ ∂S (10.4.2)
⎪ 0, ( x, y ) ∈ S c
⎩
402
n
C∞
Fig.10.4.1.
Geometry for a
supercavitating
hydrofoil with a
S
reentrant jet cavity
closure. Ccav+
n Cwak
Cjet
To apply this Cfoil
identity to the inflow
Cca v-
problem of a
supercavitating
hydrofoil with a
reentrant jet cavity
termination model it
is necessary to deform
the contour in such a
manner that a portion
of the contour coincides with the surface of the foil, C foil , a portion of the contour coincides
with the cavity boundary, Ccav , a portion is the cross section of the reentrant jet, C jet , a
portion forms a, "contour at infinity”, C∞, and a portion that covers both sides of the wake,
Cwak. The identity then takes the form
⎧ ∂φ ∂G ⎫ ∂G
βφ = ∫ ⎨ G −φ
C foil + Ccav + C jet ⎩
∂n
⎬ ds + ∫ Δφ
∂n ⎭ Cwak
∂n
ds ,. (10.4.3)
where the integral over C∞ has vanished and Δφ = φ + − φ − is the jump in potential across
the wake surface Cwak.
To solve the nonlinear problem of a supercavitating hydrofoil section wherein the
cavity shape is to be determined as part of the solution we must satisfy the kinematic
boundary condition over the wetted portions of the foil and both the kinematic and dynamic
boundary conditions over the cavity at convergence. In addition, there must be conditions
specified on the jet cross section since that constitutes part of the computational boundary.
Over the foil and cavity boundaries the kinematic boundary condition is the usual
∂φ
= − nxU , (10.4.4)
∂n
dynamic condition on the cavity may be derived from Bernoulli’s equation to yield
2
⎛ ∂φ ⎞
⎜ sxU + ⎟ = U (1 + σ )
2
(10.4.5)
⎝ ∂s ⎠
Choosing the appropriate branch of the square root, this equation may be re-expressed as
∂φ / ∂s = U 1 + σ − sxU , (10.4.6)
which may be integrated to obtain φ = φ0 + U 1 + σ ( s − s0 ) − U ( x − x0 ) , (10.4.7)
where φ0 may be taken to be the potential at the upstream end of the cavity located at s0 or
403
( x0 , y0 ) .
Note that for a supercavitating foil this initial point is different for the upper and
lower portions of the cavity.
To determine the conditions to be satisfied on the reentrant jet cross section we must
first assume that the jet has asymptoted to a constant diameter and constant speed. In this
∂Φ
case the speed is just that of the fluid at the cavity wall, thus = U 1 + σ on Cjet (10.4.8)
∂n
∂φ
and = U 1 + σ − nxU on Cjet (10.4.9)
∂n
The Kutta condition for this situation may be implemented by having the wake
where there is a jump in potential emanate from the center of the jet cross section. Due to
our assumptions on the jet, the potential on the upper (lower) portion of the wake and the
upper (lower) portion of the jet cross section will equal that found where the upper (lower)
cavity wall intersects the jet. This is the so-called Marino Kutta condition.
Employing these boundary conditions in equation (10.4.3) then leads to coupled
integrals equations for the lifting flow past a supercavitating hydrofoil of the form
∂G ∂φ ∂G ∂G
πφ + ∫ φ ds − ∫ Gds + φ0+ ∫ ds + φ0− ∫ ds
C foil ∂n Ccav ∂n C+
∂ n C−
∂n
cav cav
∂G ∂G ∂G
+φ + ∫ ds + φ − ∫ ds − ∫ (φ + − φ − ) ds
C +jet
∂n C−
∂n Cwak ∂n
jet
⎡ (10.4.10)
∂G ∂G ⎤
∫C Gds + U 1 + σ ⎢⎢ ∫+ ( s − s0 ) ∂n ds + ∫− ( s − s0 ) ∂n ds ⎥⎥
−U 1 + σ
jet ⎣ Ccav Ccav ⎦
⎡ ∂G ∂G ⎤
= −U ∫ nx Gds + U ⎢ ∫ ( x − x0 ) ds + ∫ ( x − x0 ) ds ⎥ + U ∫ nx Gds
⎢⎣ Ccav ∂n ∂n ⎥
C foil + −
Ccav ⎦ C jet
on the wetted portion of the foil and

∂G ∂φ ∂G ∂G
π ⎡⎣φ0± + U 1 + σ ( s − s0± ) ⎤⎦ + ∫ φ ds − ∫ Gds + φ0+ ∫ ds + φ0− ∫ ds
C foil ∂n Ccav ∂n C+
∂n C−
∂n
cav cav
∂G ∂G ∂G
+φ + ∫ ds + φ − ∫ ds − ∫ (φ + − φ − ) ds
C +jet
∂n C−
∂n Cwak
∂n
jet
⎡ (10.4.11)
∂G ∂G ⎤
−U 1 + σ ∫ Gds + U 1 + σ ⎢ ∫ ( s − s0+ ) ds + ∫ ( s − s0− ) ds ⎥
⎢⎣ Ccav ∂n ∂n ⎥
C jet + −
Ccav ⎦
⎡ ∂G ∂G ⎤
= π U ( x − x0± ) − U nx Gds + U ⎢ ∫ ( x − x0+ )
∫ ds + ∫ ( x − x0− ) ds ⎥ + U ∫ nx Gds
C foil ⎢⎣ Ccav
+ ∂n −
Ccav
∂n ⎥⎦ C jet
on the cavity boundaries. These equations must be augmented with a closure condition
which guarantees zero net source strength including the flux through the jet or
∂φ
∫
C foil + Cavc + C jet ∂n
ds = U 1 + σ ∫ ds
C jet
(10.4.12)
404
∂φ
or equivalently ∫
Ccav ∂n
ds = U ∫ nx ds
C foil + C jet
(10.4.13)
Again, in these equations we have enforced the dynamic boundary condition over
the cavity and not the kinematic boundary condition. The kinematic boundary condition will
be used to update the cavity shape until the kinematic boundary condition is satisfied over
the cavity as well.
The simplest manner in which these equations may be solved is to employ a low-order
boundary element method (BEM) wherein the boundary of the hydrofoil and cavity is
discretized into a number of elements or panels in each of which the potential and its normal
derivative are assumed to be constant and the resulting integrals over the panels may be
evaluated analytically. Applying the above integral equations at the centers of the elements
then yields a discreet system of equations of the form.
∂G ∂φ
πφi + ∑ φ j ∫ ds − ∑ ∫ Gds + φ0+ ∑ ∫ Gds + φ0− ∑ ∫ Gds +
j ∋ C j ∈C foil Cj
∂n j ∋ C j ∈Ccav
∂n j Cj +
j ∋ C j ∈Ccav Cj
−
∂G ∂G ∂G
+φw+ ∑ ∫ ∂n ds + φ ∑ ∫ ∂n ds − (φ −
w
+
w − φw− ) ∫ ∂n
ds − U 1 + σ ∫ Gds +
j ∋ C j ∈C +jet C j j ∋ C j ∈C −jet C j Cwak C jet
⎡ (10.4.14)
∂G ∂G ⎤
+U 1 + σ ⎢ ∑ ∫ ( s − s0 ) ds + ∑ ∫ ( s − s0 ) ds ⎥ = −U ∑ ∫ nx Gds +
⎢⎣ j ∋C j ∈Ccav ∂n ∂n ⎥
⎦
+ − j ∋ C j ∈C foil C j
Cj j ∋ C j ∈Ccav Cj
⎡ ∂G ∂G ⎤
+U ⎢ ∑ ∫ ( x − x0 ) ds + ∑ ∫ ( x − x0 ) ds ⎥ + U ∫ nx Gds
⎢⎣ j ∋C j ∈Ccav ∂n ∂n ⎥
⎦
+ −
Cj j ∋ C j ∈Ccav Cj C jet
on the wetted portion of the foil and

∂G ∂φ
π ⎡⎣φ0± + U 1 + σ ( si − s0± ) ⎤⎦ + ∑ φ j ∫ ds − ∑ ∫ Gds + φ0 ∑ ∫ Gds +
+
j ∋ C j ∈C foil Cj ∂n j ∋ C j ∈Ccav ∂n j Cj +
∂G ∂G ∂G
+φ0− ∑ ∫ Gds + φ ∑ ∫ ∂n ds + φ ∑ ∫ ∂n ds − (φ
−
+
w
−
w
+
w − φw− ) ∫ ∂n
ds
j ∋ C j ∈Ccav Cj j ∋ C j ∈C +jet Cj j ∋ C j ∈C −jet Cj Cwak
(10.4.15)
⎡ ∂G ∂G ⎤
−U 1 + σ ∫ Gds + U 1 + σ ⎢ ∑ ∫ ( s − s0 )
⎢⎣ j ∋C j ∈Ccav ∂n
ds + ∑ − ∫ ( s − s0 ) ∂n ds ⎥⎥ = π U ( xi − x0± ) −
⎦
+
C jet Cj j ∋ C j ∈Ccav C j
⎡ ∂G ∂G ⎤
−U ∑ ∫ n Gds + U ⎢⎢ ∑ ∫ ( x − x ) ∂n ds + ∑ ∫ ( x − x ) ∂n ds ⎥⎥ + U ∫ n Gds
x 0 0 x
⎣ j ∋C j ∈Ccav C j ⎦
j ∋ C j ∈C foil C j + −
j ∋ C j ∈Ccav Cj C jet
on the cavity boundaries. The closure condition becomes

∂φ
∑ ∫ ds = U j ∋C j ∈∑
j ∋ C j ∈Ccav ∂n j C j
∫ nx ds
C foil + C jet C j
(10.4.16)
Once a solution has been obtained for the unknowns, φ j , j ∋ C j ∈ C foil ,

∂φ ∂n j , j ∋ C j ∈ Ccav and 1 + σ , the kinematic boundary condition may be used to update
405
the cavity shape. The kinematic boundary condition may be written in the form
dy v
= tan (θ ) = . (10.4.17)
dx U +u
The cavity shape is updated between iterations by computing the velocities induced at the
center of each panel and rotating the panels such that they are parallel to that velocity
starting at the point of detachment of the cavity from the cavitator. Specifically, if the old
position of a panel endpoint relative to its upstream end is given by ( Δx, Δy ) , then the new
position of the panel endpoint is assumed to be given by ( Δx + δ x, Δy + δ y ) . The discrete
form of the kinematic boundary condition then requires that
Δy + δ y v
= tan (θ + δθ ) = (10.4.18)
Δx + δ x u
where the geometric quantities are illustrated in Fig.10.4.2. Assuming rotation of the panel,
the new displacements of its endpoints are readily shown to be given by
δ x = −Δy δθ
(10.4.19)
δ y = Δx δθ
v Δx − u Δy
for small rotations, where δθ = (10.4.20)
u Δx + vΔy
Fig.10.4.2.
δθ (Δx+δx, Δy+δy)
Geometry for panel alignment, after [Uhlman, 2006].
(Δx, Δy)
These displacements, (δx,δy), are computed at
each panel endpoint starting at the upstream θ
detachment point of the cavity and are then added to
all panel endpoints downstream of the current panel.
These displacements have the effect of altering the length of the cavity. To counter this, the
cavity is then stretched in such a manner that the cavity length and jet length are returned to
their original specified values. The problem is then reformulated using this new geometry
and the entire problem resolved. This process is continued until the cavitation number has
converged to some specified level of accuracy, see Fig.10.4.3.
5.00
Fig.10.4.3.
Solution for a 4.00
supercavitating 3.00
phi, dphidn
flat plate at 60 2.00

phi
degrees angle- dphidn
1.00
of-attack and
l/c=2.0. 0.00
‐1.00
‐2.00
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00
s
406

The forces acting on the body may again be determined by computing the pressure acting at
each panel using Bernoulli’s equation and performing integration time the local unit normal
as F = ∫ n ( p − pc ) ds (10.4.21)
C
where the integral need only be computed over the wetted portion of the hydrofoil. The
cavity shape is
determined as part of 2.00
the solution process 1.50
and the cavity volume
1.00
is readily found. The
cp
results are presented 0.50
cp, q q
in Figs.10.4.4 & 0.00
10.4.5. ‐0.50
Fig.10.4.4. ‐1.00
Pressure coefficient
‐1.50
and dimensionless
surface speed for a ‐2.00
supercavitating flat ‐0.50 ‐0.40 ‐0.30 ‐0.20

x
‐0.10 0.00 0.10 0.20
plate at 60 degrees 1.00

angle-of-attack and 0.80
l/c=2.0. 0.60
0.40
Fig.10.4.5. 0.20
y
Cavity shape for a 0.00

supercavitating flat ‐0.20
plate at 60 degrees ‐0.40
angle-of-attack and ‐0.60
l/c=2.0. ‐0.80
‐0.50 ‐0.25 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00
x
Sec. 10.5. Nonlinear Potential-Based Method for Supercavitating Flow

past Axisymmetric Bodies
10.5.1. Integral Equations for Reentrant Jet Cavity Closure
The problem of axisymmetric supercavitation with a reentrant jet is simpler than the 2D
problem in that there is no lift involved. The only complication is the requirement to
compute axisymmetric Green’s functions. Again Green’s third identity is employed in the
form
⎧⎪ ∂φ ( ρ ( s ) ,ϕ ) ∂G ( r ,θ ; ρ ( s ) ,ϕ ) ⎫⎪
βφ ( r ,θ ) = ∫∫ ⎨ G ( r ,θ ; ρ ( s ) , ϕ ) − φ ( ρ ( s ) , ϕ ) ⎬ ρ ( s ) dϕ ds (10.5.1)
S ⎪⎩ ∂n ∂ n ⎪⎭
where the normal is directed out of the fluid, s is arclength along a meridian,
407
⎧4π , in V
⎪
β = ⎨ 2π , on S . (10.5.2)
⎪ 0, in V c
⎩
Derivation and evaluation of the axisymmetric Green’s functions is given in Apps. C and D.
The boundary conditions are the same as found in the 2D case, leading to coupled
integral equations of the form
∂G ∂φ ∂G ⎡ ∂G ⎤
2πφ + ∫∫ φ dS − ∫∫ GdS + φ0 ∫∫ dS + 1 + σ ⎢ ∫∫ ( s − s0 ) dS − ∫∫ GdS ⎥
Sb
∂n Sc
∂n Sc + S j
∂n ⎢⎣ Sc + S j ∂n Sj ⎥⎦
(10.5.3)
∂G
= − ∫∫ nx GdS − ∫∫ nx GdS + ∫∫ ( x − x0 ) dS
Sb Sj Sc + S j ∂n
∂G ∂φ ⎡ ∂G ⎤
∫∫ φ ∂n
dS − ∫∫ GdS + φ0 ⎢ 2π + ∫∫
∂n ⎢⎣
dS ⎥
∂n ⎥
Sb Sc Sc + S j ⎦
⎡ ∂G ⎤
on the body and + 1 + σ ⎢ 2π ( s − s0 ) + ∫∫ ( s − s0 ) dS − ∫∫ GdS ⎥ (10.5.4)
⎢⎣ Sc + S j
∂n Sj ⎥⎦
⎡ ∂G ⎤
= − ∫∫ nx GdS − ∫∫ nx GdS + ⎢ 2π ( x − x0 ) + ∫∫ ( x − x0 ) dS ⎥
⎢⎣ ∂n ⎥
Sb Sj Sc + S j ⎦
∂φ
on the cavity, with the closure condition ∫∫ dS = ∫∫ nx dS (10.5.5)
Sc ∂n Sb + S j

These equation may be discretized in a manner very similar to that employed in the two
dimensional problem and the solution found using a similar iteration scheme, Discretization
of the integral equations then yields a discreet system of equations of the form.
∂G ∂φ ∂G
2πφi + ∑ φ j ∫ ds − ∑ ∫ Gds + φ0 ∑ ∫ ds
j ∋ C j ∈C foil Cj
∂n j ∋ C j ∈Ccav ∂n j C j j ∋ C j ∈Ccav + C jet C j ∂n
⎡ ∂G ⎤
+U 1 + σ ⎢ ∑ ∫ ( s − s0 ) ds − ∫ Gds ⎥ = −U ∑ ∫ nx Gds − U ∫ nx Gds (10.5.6)
⎢⎣ j ∋C j ∈Ccav C j ∂n C jet ⎥⎦ j ∋ C j ∈C foil C j C jet
∂G
+U ∑ ∫ ( x − x0 ) ds
j ∋ C j ∈Ccav C j ∂n
along arclength on the wetted portion of the foil and
408
∂G ∂φ ⎡ ∂G ⎤
∑ φj ∫
∂n
ds − ∑
j ∋ C j ∈Ccav ∂n
∫ Gds + φ 0 ⎢ 2π +
⎢⎣
∑ ∫ ∂n ds ⎥⎥
j ∋ C j ∈C foil Cj j Cj j ∋ C j ∈Ccav + C jet C j
⎦
⎡ ∂G ⎤ (10.5.7)
+U 1 + σ ⎢ 2π ( si − s0 ) +
⎢⎣
∑ ∫ ( s − s0 ) ∂n ds − ∫ Gds ⎥⎥ = −U j ∋C∑∈C ∫ nxGds − U ∫ nxGds
j ∋ C j ∈Ccav C j C jet ⎦ j foil C j C jet
⎡ ∂G ⎤
+U ⎢ 2π ( xi − x0 ) +
⎢⎣
∑ ∫ ( x − x ) ∂n ds ⎥⎥ 0
j ∋ C j ∈Ccav C j
⎦
along arclength on the cavity boundaries. The closure condition becomes
∂φ
∑ ∫ rds = U j ∋C j ∈∑
∫ nx rds
C foil + C jet C j
(10.5.8)
The cavity updating scheme at each

iteration follows exactly that
employed in sub-Sec. 10.4.2, see
Fig.10.5.1.
Fig.10.5.1.
Cut-away view of an axisymmetric
cavity with a reentrant jet behind a
disk cavitator with l d = 5 . Shades
represent the disturbance potential.
Solution has been revolved about the
axis of symmetry to obtain cut-away
view, after [Uhlman, 2006].
Fig.10.5.2 presents the initial assumed

cavity geometry, the resultant cavity geometry after one iteration and the final cavity
geometry to illustrate
1.2
the rapid convergence
of the solution 1.0
procedure. 0 iter ations
0.8
1 iter ation
Fig.10.5.2. 0.6 final
y
Initial cavity shape, 0.4

cavity shape after first
iteration and final 0.2
cavity shape, after 0.0
[Uhlman, 2006]. 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Fig.10.5.3 shows the independence of the solution with assumed jet length.
Fig.10.5.4 compares the reentrant jet cavity solution with those employing a modified
Riabouchinsky wall at the same cavitation number and cavity length.
409
1.2
Fig.10.5.3. 1.0
Comparison of
solutions for l d = 5 for
0.8
Reenax (lcav/d=5.0, sigma=0.269436)
various values of the jet 0.6
y
length, after [Uhlman, Scax (lcav/d=5.0, sigma=0.262471)
2006]. 0.4
Scax(lcav/d=4.84475,
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
1.2 x
Fig.10.5.4.
Comparison of cavity 1.0
shapes from the ljet /lcav= 0.1
0.8 ljet /lcav= 0.2
reentrant jet and ljet /lcav= 0.4
Riabouchinsky wall 0.6
y
ljet /lcav=0.6
formulations at equal 0.4 ljet /lcav= 0.8
cavity lengths and ljet /lcav= 0.9
0.2
cavitation numbers,
after [Uhlman, 2006]. 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
x

The forces acting on the cavitator may again be found by computing the pressure on a panel
through Bernoulli's equation and integrating the quantity F = ∫∫ n ( p − pc ) dS (10.5.9)
S
over the wetted portion of the cavitator. For this axisymmetric case there is no moment and
the only force is the drag which becomes D = 2π ∫ nx ( p − pc ) r ( s ) ds (10.5.10)
C
where the integral is in arclength along the wetted portion of the cavitator. Figs.10.5.5 to
10.5.7 present a comparison of the predicted drag with experimental results for a disk
cavitator over a range of
2.00
cavitation numbers
1.80
(Figs.10.5.5 & 10.5.6) and for 1.60
a conical cavitator
1.40
(Fig.10.5.7). 1.20 Brennan (1969)
Fig.10.5.5. Dawson-Seigel (1970)
Cd
1.00 Kermeen
Comparison of predicted and 0.80 σ Eisenberg-Pond (1948)
experimental drag coefficients 0.60 Klose-Acosta (1965)
for a disk cavitator for 0.40 Oversmith (1959)
1 ≤ lcav d ≤ 10 (data from 0.20 Prediction
[May, 1975]), after [Uhlman, 0.00

2006] 0.00 0.20 0.40 0.60 0.80 1.00 σ
1.20 1.40
410
1.15
Fig.10.5.6. 1.10
Comparison of 1.05
predicted and 1.00
experimental drag Oversm ith (1959)
0.95
coefficients for a Reic hardt (1946)
Cd
0.90 Brennan (1 969)
disk cavitator for O'N eill (19 54)
4≤ l/d ≤40 (data 0.85 Kerm een
from [May, 1975]), 0.80 Kice niuk (19 54)
after [Uhlman, 0.75

A.R.D.E .
Pre dicti on
2006].
0.70
0.00 0.05 0.10 0.15 0.20 0.25 σ
0.30 0.35
The comparisons for the disk cavitator (which has no viscous drag component) are well
within the range of experimental spread, while adding a viscous component to the drag (via
Thwaites method) of the conical cavitator is seen to improve the agreement.
0.40
Fig.10.5.7. 0.35
Comparison of 0.30
predicted and
experimental drag 0.25 O'Neill (1954)
coefficients for a 0.20 3/4 Inch (May 1975)
Cd
15 degree half Kiceniuk (1954)

angle cone 0.15 1/2 Inch (May 1975)
cavitator, 0.10 Kermeen (May 1975)
2 ≤ lcav d ≤ 10 Prediction (inviscid)
0.05 Prediction (Re=10^5)
(data from [May,
1975]), after 0.00
[Uhlman, 2006]. 0.00 0.05 0.10 0.15 0.20 0.25 σ 0.30 0.35
Fig.10.5.8 shows nominally axisymmetric supercavitating flows past circular disk

cavitators [Brennen, 1970]. Note the similarity of the experimental results to the
computational predictions presented in this section. The cloudy character of the cavity – at
least over the downstream portion – indicates that the cavities are operating in the re-entrant
jet regime: interference of the upstream flowing jet with the downstream flowing liquid
outside the cavity boundary disrupts the otherwise smooth flow. In a transverse gravity field,
the re-entrant jet coexists with the twin vortex structure that becomes prominent when
gravity is relatively more important, although one feature tends to dominate the other
depending on the relative values of the cavitation number and the cavity Froude number,
e.g., [Brennen, 1969b]. In the cases presented in Fig.10.5.8, the cavity Froude number is
high enough that the re-entrant jet dominates the flow, and the twin vortex structure is
relatively unimportant. The interaction of the flow with the downstream sting can also have
an influence over the detailed structure of the flow near closure, although the effects are
considered negligible in the images of Fig.10.5.8.
411
a c
Fig.10.5.8. Nominally axisymmetric cavities downstream of circular disk cavitators:
a – A close-up view of the flow near the cavitator; note the similarity of the cavity shape
near detachment to the predictions presented in Figs. 10.5.2-10.5.4;
b and c – The full cavities under two different conditions; although both cavities are
operating in the re-entrant jet regime of flow, the effects of interference between the re-
reentrant jet and the cavity persist upstream to the cavitator in image b, whereas they are
confined farther downstream in image c. (The images courtesy of C.E. Brennen and the
California Institute of Technology.
Sec. 10.6. Nonlinear Potential-Based Method for Partially Cavitating

Flow past Axisymmetric Bodies
10.6.1. Integral Equations for Partially Cavitating Axisymmetric Bodies with a
Riabouchinsky Wall
Boundary element methods may also be used to examine partial cavitation over bodies.
Varghese, Uhlman and Kirschner [2005] employed such a method using the modified
Riabouchinsky cavity termination model. The boundary value problem remains largely the
same as that employed above with the reentrant jet cavity termination replaced by the
Riabouchinsky model where a no-flux kinematic boundary condition is imposed on the wall
and the cavity shape update is performed in the same manner as described for the two-
dimensional, nonlinear velocity based hydrofoil problem described earlier, see Fig.10.6.1.
The required integral equations then take the form
∂G ∂φ ∂G ∂G
2πφ + ∫∫ φ dS − ∫∫ GdS + φ0 ∫∫ dS + 1 + σ ∫∫ ( s − s0 ) dS
Sb + S wall ∂n Sc ∂n Sc ∂n Sc ∂n
(10.6.1)
∂G
= − ∫∫ nx GdS + ∫∫ ( x − x0 ) dS
Sb + S wall Sc
∂n
on the body and
∂G ∂φ ⎡ ∂G ⎤ ⎡ ∂G ⎤
∫∫ φ dS − ∫∫ GdS + φ0 ⎢ 2π + ∫∫ dS ⎥ + 1 + σ ⎢ 2π ( s − s0 ) + ∫∫ ( s − s0 ) dS ⎥ =
Sb + S wall
∂n Sc
∂n ⎢⎣ Sc
∂n ⎥⎦ ⎢⎣ Sc
∂n ⎥⎦
⎡ ∂G ⎤
=− ∫∫ nx GdS + ⎢ 2π ( x − x0 ) + ∫∫ ( x − x0 ) dS ⎥ (10.6.2)
Sb + S wall ⎢⎣ Sc
∂n ⎥⎦
412
∂φ
on the cavity, with the closure condition ∫∫ ∂n dS = ∫∫
Sc Sb + S wall
nx dS . (10.6.3)
Note the differences between this formulation and that with a reentrant jet.
10.6.2. Discretization of the Integral Equations for Partially Cavitating
Axisymmetric Bodies with a Riabouchinsky Wall
These equation may be discretized in a manner very similar to that employed in the two
dimensional problem and the solution found using a similar iteration scheme. Discretization
of the integral equations then yields a discreet system of equations of the form.
∂G ∂φ ∂G
2πφi + ∑ φj ∫
∂n
ds − ∑ ∫
Gds + φ0 ∑ ∫ ds
j ∋ C j ∈C foil + Cwall Cj j ∋ C j ∈Ccav C j ∂n
∂G
+U 1 + σ ∑ ∫ ( s − s ) ∂n ds = −U 0 ∑ ∫ n Gds
j ∋ C j ∈C foil + Cwall C j
x (10.6.4)
∂G
+U ∑ ∫ ( x − x ) ∂n ds 0
along arclength on the wetted portion of the foil and

∂G ∂φ ⎡ ∂G ⎤
∑ φj ∫
∂n
ds − ∑
j ∋ C j ∈Ccav ∂n
∫ Gds + φ0 ⎢ 2π +
⎢⎣
∑ ∫ ∂n ds ⎥⎥
j ∋ C j ∈C foil + Cwall Cj j Cj j ∋ C j ∈Ccav C j
⎦
⎡ ∂G ⎤ (10.6.5)
+U 1 + σ ⎢ 2π ( si − s0 ) +
⎢⎣
∑ ∫ ( s − s ) ∂n ds ⎥⎥ = −U
0 ∑ ∫ n Gds x
⎦ j ∋ C j ∈C foil + Cwall C j
⎡ ∂G ⎤
+U ⎢ 2π ( xi − x0 ) +
⎢⎣
∑ ∫ ( x − x ) ∂n ds ⎥⎥ 0
⎦
along arclength on the cavity boundaries. The closure condition becomes
∂φ
∑ ∫ rds = U j ∋C j ∈C∑foil +Cwall C∫ nx rds
(10.6.6)
j
The cavity updating scheme at each iteration follows exactly that employed in sub-Sec.
10.2.2 with a new Riabouchinsky wall being dropped from the end of the cavity to the body
at the end of each iteration.
Fig.10.6.1.
Partial
cavitation
problem,
after
[Varghese,
Uhlman &
Kirschner,
2005].
413
Fig.10.6.2.
Top plot:- surface pressure
distribution,
Bottom plot: cavity shapes for
different cavity lengths.
Dimensionless body length 5,
dimensionless body radius 0.8, after
[Varghese, Uhlman & Kirschner,
2005].
One of the interesting features of

partially cavitating flow for bodies of
this shape is that cavities do not
appear to exist for all cavity lengths.
Specifically, cavity lengths wherein
the converged cavity shape would
intersect the body must be excluded
from consideration and are non-
realizable. Cavities of this type are
illustrated in Fig.10.6.3.
Fig10.6.3.
Illustration of non-
realizable cavity cases.
A frame from the

initiation of a
supercavity is presented
in Fig.10.6.4. The flow
is still in the partially
cavitating phase. Note
the similarity with the
predicted results presented in Fig.10.6.2.
Fig.10.6.4.
Nominally axisymmetric
partial cavitation during
the initiation of a
supercavity, the first
frame of the sequence
presented in Fig.8.6.12. (Image courtesy of R. Arndt.)

The force acting on the body may again be obtained from the integration of the pressure, as
found from Bernoulli's equation, integrated over the surface of the body as
414
F= ∫∫ npdS
S
(10.6.7)
Alternately, the force may be found from the integral

F = ∫∫ n ( p − pc ) dS (10.6.7)
Sb
over the wetted portion of the body, excluding the cavity boundary and the Riabouchinsky
wall. For this axisymmetric case the only force is the drag which becomes
D = 2π ∫ nx ( p − pc ) r ( s ) ds (10.6.8)
C
where the integral is in arclength along the wetted portion of the body. Note that for this
drag value to compare to experiment a base drag component and a viscous drag contribution
may be required.
Sec. 10.7. Nonlinear Potential-Based Method for 3D Hydrofoils

10.7.1. Integral Equations for 3D Cavitating Hydrofoils
Boundary element methods may also be used to examine cavitating flow over three
dimensional hydrofoils, see e.g. [Fine & Kinnas, 1993; Kinnas & Fine, 1993]. Again
Green's third identity for the disturbance potential takes the form
∂G ∂φ ∂φ ∂G ∂G
2πφ + ∫∫ φ dS − ∫∫ GdS = ∫∫ GdS − ∫∫ φ dS + ∫∫ Δφ dS (10.7.1)
S wetted ∂n Scavity ∂n S wetted ∂n Scavity ∂n S wake ∂n
where G = 1 R is the usual freespace Green's function and S wetted , Scavity and S wake denote
the wetted portion of the foil-cavity boundary, the cavity boundary and the wake surface
respectively. On the S wetted this takes the form
∂G ∂φ ∂φ ∂G ∂G
2πφ + ∫∫ φ dS − ∫∫ GdS = ∫∫ GdS − ∫∫ φ dS + ∫∫ Δφ dS (10.7.2)
S wetted ∂n Scavity ∂n S wetted ∂n Scavity ∂n S wake ∂n
and on the cavity
∂G ∂φ ∂φ ∂G ∂G
∫∫
S wetted
φ
∂n
dS − ∫∫
Scavity ∂n
GdS = −2πφ + ∫∫
S wetted ∂n
GdS − ∫∫ φ
Scavity ∂n
dS + ∫∫ Δφ
S wake ∂n
dS (10.7.3)
where the known quantities are on the right-hand-side and the unknowns on the left-hand-
side. The Morino Kutta condition Δφ = φte+ − φte− may again be used to absorb the wake
integral into the equations for the unknowns on the foil-cavity boundary, keeping in mind
that the trailing edge denoted in this expression is the trailing edge of the foil or cavity,
whichever is further downstream. The kinematic boundary condition takes the usual form
∂φ
= −n ⋅ U, on S wetted (10.7.4)
∂n
and if s denotes a unit tangent vector in the streamwise direction and b the unit tangent
vector locally orthogonal to s , the dynamic boundary condition becomes
⎡⎛ 2
∂φ ⎞ ⎛ ∂φ ⎞ ⎤
2
⎢⎜ s ⋅ U + ⎟ ⎜+ b ⋅ U + ⎟ ⎥ = (1 + σ ) on Scavity (10.7.5)
⎢⎣⎝ ∂s ⎠ ⎝ ∂b ⎠ ⎥⎦
which may be integrated to yield
415
s ⎧⎪ ⎛ ∂φ ⎞
2 ⎫⎪
φ = φ0 + ∫ ⎨ (1 + σ ) − ⎜ b ⋅ U + ⎟ − s ⋅ U ⎬ ds on Scavity , (10.7.6)
s0 ⎪ ⎝ ∂b ⎠ ⎪
⎩ ⎭
where s0 is the location of cavity detachment and φ0 the value of the disturbance potential
there.
For a given assumed cavity extent (i.e. spanwise distribution of cavity length) the
above problem may be formulated and solved with iterations being performed to determine
∂φ ∂b . Once a solution is achieved the cavity height must be determined. This may be
done by employing the kinematic boundary condition expressed as
D ⎛∂ ⎞
⎡⎣ n − h ( s, b ) ⎤⎦ = ⎜ + q ⋅ ∇ ⎟ ⎡⎣ n − h ( s, b ) ⎤⎦ = 0 (10.7.7)
Dt ⎝ ∂t ⎠
where n is the coordinate normal to the foil boundary (or other reference surface) and
h ( s, b ) is the height of the cavity above that boundary. For steady problems this becomes
∂h ∂h
qs+ qb = qn (10.7.8)
∂s ∂b
where qs , qb and qn are the total velocity components in the orthogonal s , b and n
directions respectfully. Assuming that the cross flow term is small allows this expression to
be readily integrated to obtain the cavity height. In principle this expression may be solved
iteratively to find the contribution of the crossflow term, in practice its contribution is
usually neglected. These integration then yields a cavity boundary which, in general, does
not satisfy cavity closure in that it either the cavity end fails to reach the foil or the cavity
end penetrates the foil. The displacement of the cavity closure from the foil then may be
( ) (
expressed in terms of the quantity δ ( b ) = h s ( l ( b ) ) , b − f + s ( l ( b ) ) , b ) where f + ( s, b )
denotes the upper surface of the hydrofoil. In order to achieve cavity closure, that is to
obtain δ ( b ) = 0 ∀b , the cavity length versus spanwise location l ( b ) must be solved for
iteratively. Again, in principle, the cavity length at any spanwise location should depend on
the cavity lengths at all other spanwise locations. In practice it has been found that treating
each spanwise location independently and including numerical damping yields a solution
algorithm which is stable and converges rapidly.
Note that these expressions readily generalize to the situation where the coordinate
directions employed are not orthogonal. Orthogonal coordinates have been assumed here for
ease of exposition. Note also, that while an exact solution would regrid the new foil-cavity
boundary and apply a further iterative step to the solution, it has been found, for thin
hydrofoils, that the initial cavity estimate is sufficiently good that additional iterations are
generally not required.
The integral equations for this problem are discretized in the usual way by employing panels
wherein the disturbance potential φ and its normal derivative ∂φ ∂n are assumed to be
constant. Such a discretization yields the coupled equations
416
N cavity
N wetted
∂G ∂φ
2πφi + ∑ j
φj ∫∫ ∂n
dS − ∑ j ∂n ∫∫ GdS =
S wetted , j j Scavity , j
(10.7.9)
∂φ
N cavity
N wetted
∂G ∂G
= ∑ ∫∫ GdS − ∑ φ j ∫∫ dS + (φte+ − φte− ) ∫∫ dS
j ∂n j S wetted , j j Scavity , j ∂n S wake ∂n
on the wetted portions of the foil-cavity boundary and

N cavity
N wetted
∂G ∂φ
∑j j ∫∫ ∂n
φ dS − ∑j ∂n ∫∫ GdS =
S wetted , j j Scavity , j
(10.7.10)
∂φ
N cavity
N wetted
∂G ∂G
= −2πφi + ∑ ∫∫ GdS − ∑ φ j ∫∫ dS + (φte+ − φte− ) ∫∫ dS
j ∂n j S wetted , j j Scavity , j ∂n S wake ∂n
on the assumed cavity boundary where the expression (10.7.6) is substituted for the value of
the potential on the cavity. This yields
N cavity
∂φ
N cavity
N wetted
∂G ∂G
2πφi + ∑ φ j ∫∫ dS − ∑ ∫∫ GdS + φ0 ∑ ∫∫ dS −
j S wetted , j
∂n j ∂n j Scavity , j j Scavity , j ∂n
(10.7.1)
∂φ
N cavity
∂G N wetted
∂G
− (φte − φte ) ∫∫
+ −
dS = ∑ ∫∫ GdS − ∑j S ∫∫ F ∂n dS
S wake ∂n j ∂n j Swetted ,j cavity , j
N cavity
∂φ
N cavity
N wetted
∂G ∂G
∑ φ ∫∫
j
j
∂n
dS − ∑j ∂n ∫∫ GdS + φ0 ∑ ∫∫
j ∂n
dS −
S wetted , j j Scavity , j Scavity , j
and (10.7.11)
∂φ
N cavity
∂G N wetted
∂G
− (φ − φ ) ∫∫
+
te
−
tedS = −2πφi + ∑ ∫∫ GdS − ∑ ∫∫ F dS
S wake ∂n j ∂n j S wetted , j j Scavity , j ∂n
⎧⎪
s
⎛ ∂φ ⎞
2 ⎫⎪
where F ( s ) = ∫ ⎨ (1 + σ ) − ⎜ b ⋅ U + ⎟ − s ⋅ U ⎬ ds . (10.7.12)
s0 ⎪ ⎝ ∂b ⎠
⎩ ⎭⎪
The integrals appearing in the above expressions are readily integrated for flat panels (see
for example Newman 1986).
Among the techniques employed by Fine & Kinnas to simplify and stabilize the
solution process are the introduction of a smooth pressure recovery function f ( s ) which is
zero on the cavity and selected so that the surface speed varies smoothly from the cavity
value to the wetted value and is such that the velocity at the cavity termination is given by
U ∞ 1 + σ ⎡⎣1 − f ( s ) ⎤⎦ , see [Fine, 1993] for more details. Note that this alters equation
⎧⎪ s
⎛ ∂φ ⎞
2 ⎫⎪
F ( s ) = ∫ ⎨ (1 + σ ) ⎡⎣1 − f ( s ) ⎤⎦ − ⎜ b ⋅ U +
2
(10.7.12) to yield ⎟ − s ⋅ U ⎬ ds (10.7.13)
s0 ⎪ ⎝ ∂b ⎠
⎩ ⎭⎪
In addition, Fine & Kinnas introduced a "split panel" technique which allowed the
solution process to proceed without the need to regrid at every iteration. This technique also
helped stabilize the cavity planform determination. In this technique the panel wherein the
cavity ends in each spanwise swath is split into two parts, the part under the cavity and the
part not under the cavity. The disturbance potential φ and its normal derivative ∂φ ∂n in
417
each portion of the split panel are represented by extrapolation from the quantities on its side
of the split, be they knowns or unknowns. The values of these quantities in the split panels
∂φ ∂φ
φ l + φd ld ∂φ ∂n u lu + ∂n d ld
are defined by weighted averages as φ = u u , = (10.7.14)
lu + ld ∂n lu + ld
where the subscripts u an d denote upstream and downstream values respectively. and lu and
ld denote the streamwise lengths of the split portions of the panel and φu , φd , ∂φ ∂nu and
∂φ ∂nd the extrapolated values.. The known portions of these averages are placed on the
right-hand-side of the equation and the unknown portions on the left-hand-side.
The cavity height is obtained by recursively integrating equation (10.7.8) in the
spanwise direction with updated estimates of the spanwise derivative ∂h ∂b . The cavity
will, in general, not be closed at the cavity trailing edge and an iterative root finding
algorithm must be employed to update the cavity shape between each iteration. The
solution will have converged when δ ( b ) = 0 to within the specified tolerance.
The algorithm described so far works fine for a partially cavitating hydrofoil. To
extend the technique to supercavitating hydrofoils wherein no a priori initial guess for the
cavity boundary is available, Fine & Kinnas constructed a hybrid model where the flow over
the hydrofoil is modeled as discussed above and the cavity downstream of the foil is
modeled as a vortex-source lattice in a manner similar to that described in section 10.3. The
hybrid method then combines the potential and velocity based methods to achieve its
solution using potential methods over the hydrofoil and velocity methods downstream of the
hydrofoil. Problem solution then yields the potential and its normal derivative over the foil
and the cavity source strengths downstream of the foil. The iterative scheme for the cavity
planform it largely the same and a split plane technique is employed for the source panels in
way of the supercavity. Note, however, that to obtain the cavity shape downstream of the
hydrofoil, the cavity "camber" must first be determined by integrating the expression
∂c
= w ( s ) + U ∞ sin (α ) (10.7.15)
∂s
where c is the cavity camber height, w is the velocity normal to the wake (excluding the
local contribution) and U ∞ sin (α ) is just the angle-of-attack contribution. Once this camber
line is know the cavity heights above and below it may be determined by integrating the
local contribution along the top and bottom of the camber line to obtain the cavity thickness.
Once a solution has been obtained the potential and the coordinates may be differentiated
along chordwise and spanwise directions s and t respectively to obtain ∂φ ∂s , ∂φ ∂b and
the tangent vectors s and b . The disturbance velocity components may then be found by
⎛ nx n y nz ⎞ ⎛ u ⎞ ⎛ ∂φ / ∂n ⎞
⎜ ⎟⎜ ⎟ ⎜ ⎟
solving the equations ⎜ sx s y sz ⎟ ⎜ v ⎟ = ⎜ ∂φ / ∂s ⎟ (10.7.15)
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ bx by bz ⎠ ⎝ w ⎠ ⎝ ∂φ / ∂b ⎠
even in the case when the s and b directions are not orthogonal. The freestream contribution
may then be added to these to obtain the total velocities and Bernoulli's equation employed
to yield the surface pressures. The forces and moments can then be obtained from the
418
integrals F = ∫∫ n ( p − pc ) dS (10.7.16)
S
and M = ∫∫ r × n ( p − pc ) dS (10.7.17)
S
where the integration is only over the wetted portions of the foil.
Sec. 10.8. Computation of Added Mass and Damping of

Supercavitating Axisymmetric Bodies†
10.8.1. Formulation
Consider a supercavitating body, such as the one shown in Fig.10.8.1, undergoing nominally
steady forward motion with velocity U0 in the negative x direction. We seek expressions for
the added mass and damping forces experienced by the body in response to its unsteady
velocity, the magnitudes of which will be assumed to be small relative to the forward speed.
To solve for these quantities, we will formulate a boundary value problem using a Cartesian
coordinate system x = ( x, y, z ) moving through the fluid with a constant velocity U. The
fluid is assumed to be inviscid and incompressible and the flow irrotational, so that the
velocity field may be written as the gradient of a potential, Φ(x, t ) , which satisfies Laplace’s
equation ∇ 2Φ = 0 . (10.8.1)
The total fluid velocity is assumed to be an unsteady perturbation superposed on a steady
mean flow. Accordingly, the velocity potential is decomposed into a steady mean velocity
G G
potential and a time-dependent perturbation potential Φ( x , t ) = Φ0 ( x ) + φ(x, t ) . (10.8.2)
10.8.2. Kinematic Boundary Condition

If the body and cavity surface is represented by the equation F (x, t ) = 0 , then the
exact potential flow kinematic boundary condition to be satisfied on the surface is
DF ∂F
= + ∇(Φ0 + φ) ⋅ ∇F = 0 . (10.8.3)
Dt ∂t
y U0 Cavity surface
x
Caviator
Fig.10.8.1 Notional sketch of a supercavitating vehicle.
In addition to the coordinate system x = ( x, y , z ) , we will also employ a coordinate system

relative to which the wetted part of the body is stationary. This coordinate system,
x0 = ( x0 , y0 , z0 ) , will satisfy x = x0 + ξ (x0 , t ) , (10.8.4)
†
This section has been adapted from material presented, in part, in [Uhlman et al, 2001].
419
ξ ∂ξ / ∂t
where ξ is a function that will be defined later, other than to note that 1. ,
d U0
Using these definitions, the time derivative and gradient of F may be written as
∂F ∂ξ ∂ξ ∂ξ ∂ξ
= − ⋅ ∇F ; ∇F = ∇ x0 F − iˆ ⋅ ∇ x0 F − ˆj ⋅ ∇ x0 F − kˆ ⋅ ∇ x0 F
∂t ∂t ∂x ∂y ∂z
where the subscript “x0” denotes that the gradient operator is to be evaluated with respect to
the x0 coordinate system and i, j and k are the unit vectors in the x0 , y0 and z0 coordinate
directions, respectively. Substituting these expressions in equation (10.8.3), the kinematic
boundary condition becomes
∂ξ ⎛ ∂ξ ∂ξ ∂ξ ⎞
− ⋅∇ x0 F + ∇(Φ0 + φ) ⋅ ⎜⎜∇ x0 F − i ⋅∇ x0 F − j ⋅ ∇ x0 F − k ⋅ ∇ x0 F ⎟⎟⎟ = 0 (10.8.5)
∂t ⎝⎜ ∂x ∂y ∂z ⎠⎟
Eq. (10.8.5) applies on the exact body surface F (x, t ) = 0 . Following [Timman & Newman,
1962], this expression may be linearized under the assumption that the unsteady
displacement amplitudes, the perturbation potential and its spatial derivatives are all small.
Using a Taylor expansion for the velocity field
∇(Φ0 + φ ) = ∇ x0 Φ0 + ξ ⋅∇ x0 (∇ x0 Φ0 ) + ∇ x0 φ + O(ξ 2 ) , (10.8.6)
and assuming that φ = O(ξ ) , we may recast equation (10.8.5) as follows
∂ξ
− ⋅∇ x0 F + (∇ x0 Φ0 + ξ ⋅ ∇ x0 (∇ x0 Φ 0 ) + ∇ x0 φ) ⋅
∂t
⎛ ⎞ (10.8.7)
∂ξ ∂ξ ∂ξ
⋅⎜⎜⎜∇ x0 F − iˆ ⋅ ∇ x0 F − ˆj ⋅ ∇ x0 F − kˆ ⋅ ∇ x0 F ⎟⎟⎟ + O(ξ 2 ) = 0.
⎜⎝ ∂x 0 ∂y 0 ∂z 0 ⎠⎟
Rearranging terms, the unsteady components of equation (10.8.7) become
∂ξ
∇ x0 φ ⋅ ∇ x0 F = ⋅ ∇ x0 F + (∇ x0 Φ 0 ⋅ ∇ x0 ξ − ξ ⋅ ∇ x0 (∇ x0 Φ 0 )) ⋅ ∇ x0 F + O(ξ 2 ) . (10.8.8)
∂t
Noting that ∇ x0 F is normal to the body/cavity surface, equation (10.8.8) reduces to
∂φ ∂ ξ
= ⋅ nˆ + (∇ x0 Φ 0 ⋅ ∇ x0 ξ − ξ ⋅ ∇ x0 (∇ x0 Φ 0 )) ⋅ n . (10.8.9)
∂n ∂ t
Equation (10.8.9) is the linearized kinematic boundary condition to be applied on the mean
body-cavity surface.
We now consider only the wetted portion of the boundary and assume that its motion
is associated with rigid-body motion of the vehicle. The displacement amplitude may be
expressed as the sum of a rigid body translation, ξT , and rotation, ξ R as
ξ = ξT + ξ R × x0 . (10.8.10)
The perturbation potential may be decomposed into a sum of components proportional to
G
complex harmonic displacement amplitudes, φ = ζ j ϕ j ( x )eiωt , 0
(10.8.11)
where summation over repeated indices is implied and where the magnitudes of the complex
(ζ1 , ζ 2 , ζ3 )eiωt = ξT
amplitudes are (10.8.12)
(ζ 4 , ζ 5 , ζ 6 )eiωt = ξ R .
420
∂ϕ j ( x 0 )
Substituting Eq. (10.8.11) in Eq. (10.8.9) then yields = iω n j + m j (10.8.13)
∂n
where the generalized normal and so-called “m-terms” are given in [Ogilvie & Tuck, 1969].
⎧⎪ nˆ j = 1, 2,3
nj = ⎪⎨G
⎪⎪⎩ x0 × nˆ j = 4,5,6
⎧⎪ −nˆ ⋅∇ (∇ Φ ( x )) G (10.8.14)
⎪⎪ x0 x0 0 0 j = 1, 2,3
mj = ⎨
⎪⎪−nˆ ⋅∇ ( xG ×∇ Φ ( xG )) j = 4,5,6
⎪⎩ x0 0 x0 0 0
Eq. (10.8.13) is the linearized kinematic boundary condition to be applied on the wetted
portion of the body for each of the six modes of oscillation (surge, sway, heave, pitch, yaw,
and roll).
10.8.3. Dynamic Boundary Condition
To derive the linearized dynamic boundary condition, we start Bernoulli’s equation
G
∂Φ( x , t ) 1 G G
ρ0 + 2 ρ0∇Φ( x , t ) ⋅ ∇Φ( x , t ) = ( p∞ − pc ) + 12 ρ0U 02 , (10.8.15)
∂t
where p∞ and pc are the pressures at infinity and in the cavity, respectively, and where the
effects of buoyancy are ignored. Inserting the decomposition of the total potential given in
equation (10.8.2) and the Taylor expansion of its gradient from equation (10.8.6) then yields
∂φ ( x 0 , t )
2 + (∇ x0 Φ0 (x0 ) + ξ ⋅∇ x0 (∇ x0 Φ0 (x0 )) + ∇ x0 φ( x0 , t ))
∂t (10.8.16)
⋅ (∇ x0 Φ0 ( x0 ) + ξ ⋅ ∇ x0 (∇ x0 Φ0 (x0 )) + ∇ x0 φ(x 0 , t )) = U 02 (1 + σ )
where σ is the cavitation number.
Keeping only terms linear in ξ, equation (10.8.16) reduces to
∂φ ( x 0 , t )
2 + ∇ x Φ0 (x 0 ) ⋅ ∇ x Φ0 (x0 ) + 2∇ x Φ0 (x0 ) ⋅ ∇ x φ (x 0 , t )
∂t 0 0 0 0
(10.8.17)
+ 2∇ x Φ0 (x 0 ) ⋅ (ξ ⋅ ∇ x (∇ x Φ0 (x 0 ))) = U 02 (1 + σ )
0 0 0
The steady potential, Φ0(x0), is assumed to satisfy the steady dynamic boundary condition
∇ x Φ0 (x0 ) ⋅ ∇ x Φ0 ( x0 ) = U 02 (1 + σ )
0 0
on the cavity (10.8.18)
so equation (10.8.17) becomes
∂φ ( x 0 , t )
∂t 0 0 0
(
+ ∇ x Φ0 (x0 ) ⋅ ∇ x φ (x0 , t ) + ∇ x Φ0 ⋅ ξ ⋅∇ x (∇ x Φ 0 ( x0 )) = 0 (10.8.19)
0 0
)
The last term on the left-hand-side of equation (10.8.19) may be evaluated as follows
∇ x Φ0 ⋅ (ξ ⋅ ∇ x (∇ x Φ0 )) = ξ ⋅ (∇ x (∇ x Φ 0 ) ⋅ ∇ x Φ 0 ) = 12 ξ ⋅∇ x (∇ x Φ0 ⋅ ∇ x Φ 0 ) = 0 . (10.8.20)
0 0 0 0 0 0 0 0 0
This term vanishes on SC because the last bracketed term of (10.8.20) is constant, according
to Eq. (10.8.18).
Substituting the decomposition of the perturbation potential, Eq. (10.8.11), in Eq.
(10.8.19) and making use of equation results in the linearized dynamic boundary condition
iωϕ j (x 0 ) + ∇Φ0 (x0 ) ⋅ ∇ϕ j (x0 ) = 0 on the cavity, j = 1,...,6 (10.8.21)
where equation (10.8.21) will be applied on the steady mean cavity surface, SC.
421
10.8.4. Cavity Termination

The need for a termination model arises from the inconsistency inherent in forcing a
constant-pressure streamline to end at a stagnation point. In this effort, a re-entrant jet cavity
termination model is employed, as shown in Fig.10.8.2. This termination model was
originally devised by Efros [1946] and Kreisel [1946]. Details of the numerical
implementation in the boundary element method may be found in [Uhlman, 2006].
Reentrant
3 jet face, SJ
Fig.10.8.2. 6
Cut-away sketch of a 4 1
2 y
supercavitating cone x
5 z
showing the various Cavity, SC
surface definitions.
Cavitator, SBW
10.8.5. Boundary Integral Equation

The perturbation potential satisfies Green’s third identity:
⎡ ∂ ∂ ⎤
2πφ(x, t ) + ∫∫
w ⎢φ (ξ, t ) G ( x, ξ ) − G (x, ξ ) φ (ξ , t )⎥ dS = 0
⎢⎣ ∂n ∂n ⎥⎦
(10.8.22)
S
1
where the Green’s function, G, is given by G ( x, ξ ) = (10.8.23)
x −ξ
and where the field point, x , lies on the boundary S. Substituting the decomposition
(10.8.11) and noting that the complex amplitudes, ζ j , have no spatial dependence, we find
that each complex potential, ϕ j , satisfies
⎡ ∂G ( x, ξ ) ∂ϕ j (ξ ) ⎤
2πϕ j (x) + w ⎢⎢⎢ϕ (ξ)
∫∫ j
∂n
− G ( x, ξ )
∂n ⎥⎥⎦
⎥dS = 0 ∀j . (10.8.24)
S ⎣
The boundary conditions to be satisfied are the kinematic boundary condition (equation
(10.8.13)), which defines the source distribution ∂ϕ j ∂n on the wetted portion of the body,
and the dynamic boundary condition (equation (10.8.21)) which may be used to define the
dipole distribution ϕ j on the cavity. Inserting the kinematic boundary condition in (10.8.24)
and rearranging the terms so that the known quantities are on the right-hand-side, we arrive
at the following expression:
∂G ∂ϕ j ∂G
2πϕ j + ∫∫ ϕ j dS − ∫∫ GdS + ∫∫ ϕ j dS = ∫∫ {iω n j + m j } GdS (10.8.25)
S BW
∂n SC + S J
∂n SC + S J
∂n S BW
Here, the wetted body surface is SBW, the cavity surface is SC and the jet face is SJ, as shown
in Fig.10.8.2. Note that the dipole strength on the cavity and jet may be determined by
numerically integrating equation (10.8.21). This will be described in more detail in the
context of an axisymmetric basis flow below.
422
10.8.6. Hydrodynamic Coefficients

Once the source and dipole distributions have been computed, the forces and moments may
be found by integrating the pressure over the wetted surface
Fi (t ) = ∫∫ p(x, t )n dS ,
i (10.8.26)
SB (t )
where S B (t ) is the exact surface of the body. The unsteady pressure on S B (t ) may be
∂φ ( x , t ) 1
written p (x, t ) = −ρ0 − 2 ρ∇(Φ0 (x) + φ (x, t ))⋅ ∇(Φ0 (x) + φ(x, t )) (10.8.27)
∂t
If we again introduce the intermediate coordinate system, x0 , as in equation (10.8.4),
equation (10.8.27) may be linearized for small ξ to yield
∂
p (x0 , t ) = −ρ (φ(x0 , t )) − ρ∇ x0 Φ0 ( x0 ) ⋅ ∇ x0 φ (x0 , t ) −
∂t (10.8.28)
− 12 ρξ ⋅∇ x0 (∇ x0 Φ0 (x0 ) ⋅ ∇ x0 Φ0 (x0 ))
For the remainder of this paper, there will be no need to distinguish between the
coordinates x and x0 . It will be implicitly assumed that all quantities are defined relative
to x0 . Inserting the decomposition in Eq. (10.8.11) into Eq. (10.8.28) and the resulting
expression into Eq. (10.8.26) then yields the following equation for the force as a function of
{ }
frequency [Nakos & Sclavounos, 1990]: Fi (ω ) = Re eiωt ⎡⎢ξ j (ω 2 aij − iωbij − cij )⎤⎥ , (10.8.29)
⎣ ⎦
where the force has been split into three components: the component in phase with the
body’s acceleration, aij, the component in phase with its velocity, bij, and the component in
phase with its displacement, cij. The three coefficients are the added mass, damping and
restoring coefficients, respectively, where the expressions for the three coefficients are given
ρ ⎪⎧⎪ ⎡ ∂Φ0 ∂ϕ j ⎤⎥
⎫
⎪
⎪
by aij = − 02 Re ⎪⎨∫∫ ⎢iωϕ j + ni dS ⎪⎬ (10.8.30)
ω ⎪⎪ S ⎢⎣ ∂xk ∂xk ⎥⎦ ⎪
⎪
⎩⎪ B0 ⎪
⎭
ρ ⎪⎧⎪ ⎡ ∂Φ0 ∂ϕ j ⎤⎥ ⎪⎫⎪
bij = 0 Im ⎪ ⎨∫∫ ⎢⎢iωϕ j + ni dS ⎪ ⎬ (10.8.31)
ω ⎪⎪ S ⎣ ∂xk ∂xk ⎥⎦ ⎪⎪
⎩⎪ B0 ⎭⎪
⎪⎧⎪ ∂ ⎜⎛ 1 ∂Φ0 ∂Φ0 ⎞⎟
⎫
⎪
⎪
cij = ρ0 Re ⎪⎨∫∫ ⎜⎜ ⎟⎟ ni dS ⎪⎬ (10.8.32)
⎪⎪ S ∂ξ j ⎝ 2 ∂xk ∂xk ⎠⎟ ⎪
⎪
⎩⎪ B0 ⎪
⎭
The surface S B is the steady mean body surface.
0
10.8.7. Boundary Conditions for an Axisymmetric Base Flow

To this point, we have formulated the problem in terms of a general steady basis flow. To
demonstrate the method, we now define our steady basis flow to be that of an axisymmetric
cavitator with a reentrant jet cavity termination, as shown in Fig.10.8.2. The steady
axisymmetric flow is computed numerically via a low-order boundary element method
(Uhlman 2006). In the numerical solution of the basis flow, the cavity length is assumed to
be known, and the cavitation number is determined as part of the solution. An iterative
423
method is used to determine the cavity shape that satisfies the kinematic and dynamic
boundary conditions.
Noting that for an axisymmetric base problem we have
∂Φ0
∇Φ0 = e s = U 0 1 + σ e s on SC (10.8.33)
∂s
where s and eˆs are the arclength and unit vector along a meridian respectively, we find that
the dynamic boundary condition on SC (equation (10.8.21)) becomes:
∂ϕ j
iωϕ j + U 0 1 + σ = 0 on SC . (10.8.34)
∂s
Equation (10.8.34) is a first order ordinary differential equation for ϕ j and may be integrated
to yield ϕ j = ϕ j0 e−iγ ( s ) (10.8.35)
ω
where ( s − s0 )
γ (s) = (10.36)
U0 1+ σ
and where ϕ j0 is the value of ϕ j at s = s0 . The dynamic boundary condition on the jet cross-
∂Φ0 ∂ϕ j
section is iωϕ j +
= 0 on S J . (10.8.37)
∂n ∂ n
Noting that, on the jet cross section, ∂Φ0 ∂n = U 0 1 + σ , equation (10.8.37) becomes
∂ϕ j iω −i γ ( s )
=− ϕ j 0 e J0 , (10.8.38)
∂n U0 1+ σ
where sJ 0 is the arclength along the cavity to the edge of the jet cross-section and where we
have assumed that ∂ϕ j ∂s is zero on the jet face.
10.8.8. Boundary Integral Equations for an Axisymmetric Base Flow

On the wetted part of the body, SBW, Green’s third identity becomes:
∂ϕ j ⎧⎪ − iγ ( s )
ike J0 ⎫⎪
2πϕ j + ∫ ϕ j Ψ n ds − ∫ Ψds + ϕ j0 ⎨ ∫ e − iγ ( s ) Ψ n ds + ∫ ⎬⎪
Ψ ds
∂n ⎩⎪ sC + sJ d 1+σ ⎭
sBW sC sJ (10.8.39)
= ∫ {iω n
sBW
j + m j } Ψds
and on the cavity

∂ϕ j ⎧⎪ − iγ ( s )
ike J0 ⎫⎪
2πϕ j0 e − iγ ( sx ) + ∫ ϕ j Ψ n ds − ∫
∂n
Ψds + ϕ j0 ⎨ ∫ e − iγ ( s ) Ψ n ds +
d 1+σ
∫ Ψds ⎬⎪
sBW sC ⎩⎪ sC + sJ sJ ⎭ (10.8.40)
= ∫ {iω n
sBW
j + m j } Ψds
where sx is the arclength from the leading edge of the cavity to the field point, and sBW, sC
and sJ are the arclength domains on the wetted body, cavity and reentrant jet face. sJ0 is the
arclength at the edge of the jet. Ψ and Ψ n are the potentials induced by a ring source and a
ring normal dipole, respectively (see below for further discussion of these functions). d is
the body diameter at the cavity detachment location, and k is the dimensionless frequency,
424
k = ω d U 0 . The integrals in arclength along the meridian are computed by the boundary
element method, wherein the source and dipole distributions ( ϕ j and ∂ϕ j ∂n , respectively)
are assumed to be constant over each arclength segment (or panel). The induced potentials,
Ψ and Ψn , are then integrated over each panel. Following [Hess & Smith, 1964] and
Uhlman [2006], the induced potential at x due to a source ring with radius ρ and passing
π
ρ dθ
through the point ξ , Ψ ( x,ξ ) = ∫ , and that due to a ring normal dipole,
−π
x−ξ
∂ ⎛ ⎞
π
1
Ψ n ( x,ξ ) = ∫π ∂n ⎜⎜ x − ξ ⎟⎟ ρ dθ ,
may be written in terms of complete elliptic integrals of the
−⎝ ⎠
first and second kinds. A cylindrical coordinate system is used in which ρ is the radius of
the axisymmetric body at the plane of integration. For further details, see Appendix ?.
Modified source and dipole influence functions may also be defined for the case when there
is a crossflow component to the inflow, as is the case when the motion includes pitch, yaw,
heave or sway.
425
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
Ch.11. Selected Effects in the Modeling of Cavitating Flows

in Real Fluids
Ch. 1 provided an overview of the broad array of physical parameters that can have some
influence on the behavior of cavitating flows. It was also indicated that, for many if not most
flows of engineering interest, only a few of these parameters are important to the dominant
behavior of the system, especially (depending on the system and conditions under
consideration) the cavitation number, the cavity Froude number, the ventilation coefficient,
the Mach number, the Silberman-Song-Logvinovich number, the Campbell-Hilbert (or
alternatively the Buivol) number, and the volumetric void fraction
The Reynolds and Weber numbers are always important, insofar as they place many
cavitating systems of engineering interest in a flow regime wherein viscous and surface
tension effects are unimportant except in very localized regions of the flow system, or at the
operational margins of cavitation inception or dessinence. However, this is only strictly true
for cases of cavity detachment from a salient point. For cavitation on the smooth portion of a
body – such as the suction side of a hydrofoil that was designed for fully-wetted operation
the Reynolds and Weber number play very important roles in the location of cavity
detachment, which in turn has an effect on both the steady and unsteady hydrodynamic
performance of the system.
Studies of localized effects or the operational margins of cavitation inception or
dessinence open yet another rich treasure trove of physical concepts. For example, careful
studies by Keller (see, for example, [Keller, 1998]) on the scaling of cavity inception reveals
that the onset of cavitation is extremely sensitive to many influences, including not only the
very important level of air content in the ambient liquid, but also the distribution of the
population of nuclei in terms of their size. Other influences are also important, such as the
presence of surfactants, and the material comprising the outer coating of the body over
which cavitation may occur. Some particularly beautiful experimental results comparing the
effects of hydrophobic and hydrophilic coatings on the water entry of sphere have been
presented by [Truscott & Techet, 1998]. A sample of these is presented in Fig.11.1. The
time between images was 0.018s between the first and second frames and 0.020s between
subsequent frames for both image sequences shown. Each of the two cases was filmed from
the top and side with synced cameras. The images in the side views correspond in time with
those in the top view. The spheres were dropped into free fall from the same height of 18cm
in both cases. The effect of surface condition and the associated wetting angle on cavity
detachment is discussed in more detail in Sec.11.1.
In attempting to apply cavitation as a drag-reducing mechanism, real fluid effects are
often extremely important. For example, interaction of the cavity boundary with the
afterbody of a high-speed undersea vehicle tends to produce complicated interactions,
especially for flows that are even slightly three-dimensional, as the influences of viscosity
and surface tension compete with the inertial effects that tend to dominate the major portion
of the cavity upstream. Some drag reduction concepts for surface ships, such as so-called air
cavity craft, also involve complicated interaction among the ship, the air cavity, and the
ocean free surface. (See, for example, [Amromin, et al., 2010] or the many papers by
Matveev as exemplified by [Matveev, et al., 2009].)
426
Fig.11.1 The effect of surface properties on the behavior of flow during water entry of a
sphere, showing a time sequence of images: a and b – side views; b and c – views from
above. In a and c, the surface of the sphere was hydrophilic with a wetting angle of 68◦,
whereas in b and d, the surface of the sphere was hydrophobic, with a wetting angle of 120◦.
In the case of the hydrophilic surface, no significant water entry cavity forms, and the so-
called Worthington jet is the dominant feature of the flow once the sphere clears the free
surface. In the case of the hydrophobic surface, a very significant water entry cavity forms,
and the flow features a crown above the free surface around its circumference. A
continuation of the sequence in b and d would feature formation of a Worthington jet after
closure or “pinch-off” of the water entry cavity. (All images courtesy of T. Truscott.)
Even the typically reliable case of flow past a sharp-edged cavitator can display
some very interesting effects of the physics of real fluids when examined in close detail, as is
represented in Fig.11.2, which shows cavitating flow past two different cavitators, one
featuring a parabolic profile (in the first two images), and the other a profile that is a sector
of a circle [Brennen, 1969a; Brennen, 1970]. Re-entrant jet effects in the vaporous cavity
427
system of Fig.11.2a drive fluid upstream against the back of the cavitator, forming a high-
pressure region just downstream of the base that suppresses cavitation just behind the
cavitator; the liquid layer is clearly visible between the base of the cavitator and the
upstream extent of the vaporous cavity. In the ventilated cavitation cases presented in
Figs.11.2b&c, the introduction of gas at the base of each cavitator ensures that the cavity
boundary departs immediately and cleanly from the salient locus of detachment. However,
these latter two cases emphasize the capillary effects along the cavity boundary. Surface
tension clearly plays a role in all three flow
systems, as is evident from both the mottled cavity
surface and the periodic behavior of the cavity
flow in Fig.11.2a; in Figs.11.2b&c, surface
tension is responsible for the prominent capillary
waves along the clean cavity boundary. The flow
in Fig.11.2b appears to represent an intermediate
situation wherein the effects of surface tension on a
the clean portion of the cavity boundary compete
with the disruptive effects of the re-entrant jet
propagating back upstream toward the base of the
cavitator.
Fig.11.2.
Real fluid effects on cavity detachment from
sharp-edged cavitators. a – Vaporous cavitation; b
b and c – Ventilated cavitation. The cavitator with
a parabolic profile in a is the same as that in b;
the profile of the cavitator in c is a sector of a
circle. The interesting combination of effects in
these images is discussed in the text. (Images
courtesy of C.E. Brennen..) c
A comprehensive presentation of such a broad range of physical effects exceeds the

scope of this book. Instead the focus of this chapter is on two specific areas that are of
increasing interest. The first involves a model for real effects on cavity detachment from
smooth surfaces, and is intended to extend the classical Brillouin-Villat model (discussed in
greater detail below) to account for physical parameters that have been found to play an
important role in determining the cavity detachment location [Amromin, 1986; 2007]. Such
a prediction is important, for example, to reliable estimation of the cavitation performance of
hydrofoil sections.
The second area presented below involves high-speed motion in bubbly liquids, and
thus falls within the category of multiphase flow physics. This topic is important from at
least two perspectives. Firstly, air exists in the form of bubbles in many environments of
engineering interest, and complete models of cavitation at high speeds in such flows should
begin to account for that fact. Secondly, because even miniscule amounts of air content in
water lower the sound speed considerably (as will be discussed in more detail below), the
introduction of air into a laboratory test tank may ultimately allow for the investigation of
high-Mach-number supercavitating flow at much lower speeds, with an associated reduction
in cost and an improvement in safety. The genesis of the model presented in this chapter was
428
the recognition of a similarity between axisymmetric flows at high speeds in bubble-infested
liquids and the Kelvin wake of a surface ship: Both involve dispersive wave systems, and
both result in the appearance of a wave drag component for objects traveling at steady speed
in their respective media [Grant & Kirschner, 2003; 2004; 2005; 2006]. The theoretical
results of the model of high-speed motion in a bubbly liquid even generate wave patterns
with striking analogies to the Kelvin wake: although the topologies of the two systems are
somewhat different, both feature transverse and diverging wave systems (at least in certain
regimes). However, due to the existence of three important sound speeds in a bubbly liquid
(the sound speeds in the liquid, in the gas, and in the mixture), the physics of such high-
speed flows is rich, and several fundamental flow regimes may be identified, across the
boundaries of which the system behavior is predicted to display remarkable changes. The
model discussed herein has progressed to the prediction of the motion of a point source or a
slender axisymmetric body at high speed through a bubbly liquid; extension of the model to
account for the proper pressure condition on the cavity boundary is the subject of ongoing
research.
Sec. 11.1. First-Principle Model of Real Effects on Cavity Detachmentx

Note that the notation used in this section is independent from that of Sec. 11.2 (and, in
general, from the reminder of this book).
11.1.1. Introduction to Real Effects on Cavity Detachment
As was presented in Ch. 1 and discussed extensively throughout this book, many cavitating
flows of engineering interest are overwhelmingly dominated by inertial effects and include
only very limited regions of rotational flow. They are therefore well-described by the
techniques of potential flow. However, certain important details of such systems require
precise attention to more complicated physics, especially the effects of viscosity and surface
tension. One such problem of interest to hydrodynamicists and engineers is that of
determination of the point of cavity detachment from a smooth surface.
An example of this problem is the extent and behavior of back-bubble cavitation on
two-dimensional hydrofoil sections. A typical example involves a foil section that has been
selected for non-cavitating operation, but, due to off-design operational conditions, is subject
to reduced pressure at its design angle of attack. Under these conditions, the flow may
cavitate over the back or suction surface of the foil, with cavity detachment occurring
downstream of the leading edge, along a very smooth part of the foil contour where the
curvature is relatively small.
As is discussed in more detail below, the classical theory of potential flow requires a
uniqueness condition that allows for determination of the unknown cavity detachment point.
Historically, without consideration of viscous or surface tension effects, the appropriate
condition is the Brillouin-Villat condition [Birkhoff & Zarantonello, 1957], which presumes
that the pressure is continuous along the contour defined by the wetted foil surface and the
cavity boundary. Imposing this condition results in a cavity that detaches tangentially from
the foil surface at the point that satisfies continuity of pressure along the combined foil-
cavity boundary. However, this is a rather simplified model of the actual physical problem of
cavity detachment from a smooth boundary. This section is devoted to a more
x
This section was jointly developed with Dr. Edward Amromin based on [Amromin, 2007].
429
comprehensive consideration of the physics of cavity detachment, resulting in a model that
incorporates the effects of viscosity and surface tension.
Principal time-averaged characteristics of cavitating flows include lift and drag,
cavity length and volume, as well as the cavitation number at cavity inception. As follows
from numerous studies on cavitation of hydrofoils and bodies with smooth shapes, these
characteristics substantially depend on the location of cavity detachment points. In the
traditional Brillouin-Villat concept (here referred to as ideal cavitation), the cavitating flows
significantly depend on a single parameter, namely the cavitation number. Within the bounds
of applicability of this concept, the theoretical analysis could be performed on the ideal fluid
with a preliminary selected cavity closure scheme and a cavity detachment point or locus. As
has been shown, e.g. [Rowe & Blottiaux, 1993], using the cavity detachment point at its
observed location and tuning the closure scheme allows one to obtain satisfactory agreement
of ideal cavitation with some sets of experimental data, e.g. in [Yamaguchi & Kato, 1983]
and in Fig.11.1.1, though other sets may
require another scheme.
Fig.11.1.1.
Cavity length on the EN-hydrofoil at
Re=1.5×106, α=4.2°: Arrows show
experimental data spread. Curve A shows a
result obtained with CCVL (defined below).
Curves R&B-1 and R&B-2 show the results
for different cavity closure schemes in ideal
fluid [Rowe & Blottiaux, 1993].
As is shown in [Brennen, 1969;

Arakeri, 1975] and elsewhere, the difference between using the classical Brillouin-Villat
condition [Birkhoff & Zarantonello, 1957] for detachment predictions using this condition
and experimental data can be very significant. Analysis of experimental data shows that this
difference depends on hydrofoil (body) size C and free-stream speed U∞,. These two
quantities also affect the incipient cavitation number, σ i , as shown in [Keller, 2001] Thus,
the effects of two parameters which depend both on C and U∞, should be analyzed. There
are reasons to select as the primary parameter the Reynolds number, Re = U ∞ c ν (where ν
is the kinematic viscosity of the ambient liquid), which characterizes the importance of
viscosity to the flow. (Specifically, as described in Ch. 1, the Reynolds number is the ratio of
inertial to viscous forces in a flow; viscosity therefore becomes more important as the
Reynolds number decreases.) In addition to the Reynolds number, another parameter
depending on both c and U∞ is the Weber number, We = ρU ∞2 c γ (where ρ is the density
of the ambient fluid and γ is the local value of the surface tension with units of force per
unit length), which characterizes the importance of surface tension to the flow. It will be
shown below that surface tension in the vicinity of cavity detachment points is an important
factor. In the case of nominally axisymmetric bodies, the Reynolds number and Weber
number are important, as well, although in that case they are based on the cavitator or
maximum body diameter d, as an appropriate length scale.
It should be noted that, for both two-dimensional and axisymmetric flow, two forms
of the Reynolds number are important: (1) a global Reynolds number based on the
430
characteristic length scale, as defined above; and, (2) a local Reynolds number based on the
arc-length traversed by a fluid particle along a streamline from the point where it first
encounters the perturbation associated with the no-slip condition on the body boundary. On
the other hand, the Weber number is always a local quantity since, even if the length scale is
selected on the basis of global dimensions, the parameter is inherently dependent on the local
curvature of the interface between two phases of a fluid, or between two different fluids.
Therefore the gross character of the flow – laminar or turbulent – over the majority of a body
boundary may frequently be assessed by considering the global Reynolds number (although
the location of transition from laminar to turbulent flow will be characterized by the local
Reynolds number). The effects of surface tension, though, must be considered from a local
perspective, in general. In the following formulation, the importance of the local effects of
surface tension at the upstream and downstream extents of the cavity will become clear.
Currently, there are two differing concepts for analyzing viscous cavitating flows,
which will here be referred to as the uniform turbulent cavitation concept (UTCC) and the
concept of cavitation within viscous layers (CCVL).
The first concept widely used for analyzing viscous cavitating flows as described in
recent papers on cavitation – UTCC – is based on grid-based computational fluid dynamics
modeling of fully turbulent flows [Ahuja et al., 2001; Lindau et al., 2002, 2007; Iga et al.,
2003; Singhal et al., 2002; Vaidyanathan et al., 2003; Qin et al., 2003; Kim, 2009]. All of
these versions of UTCC are very similar. Most of the versions [Ahuja et al., 2001; Lindau et
al., 2002; Singhal et al., 2002; Vaidyanathan et al., 2003; Qin et al., 2003] are based on the
k − ε turbulence model, although different model constants are employed in different
papers. More recently, large eddy simulation (LES) has also been used, e.g. in [Kim, 2009].
These studies do not consider any boundary between the cavity and the surrounding liquid:
cavitation is represented using two-phase mixture models. The main computational results
are the pressure distributions over the body hydrofoil surface and the void fraction in the
field.
Interpretation of computational results and its validation are the main issues in
evaluating the UTCC method. Although experimental data on void fraction distributions is
already available [Gopalan & Katz, 2000], agreement between such distributions and the
cavity shapes and lengths (traditionally recorded in experiments) is rather uncertain so far.
Furthermore, validation of a majority of UTCC versions has been carried out only at a single
Reynolds number less than 2×106. These versions ignore the existence of laminar boundary
layers over hydrofoils or bodies and consider the entire flow field as a completely turbulent
flow. Such simplification is made despite experimental proof of the existence of a laminar
boundary layer over a broad range of conditions that includes the validation point. For
example, experimental results show that laminar boundary layers exist along streamlines
with uniform velocities (that is, with neutral pressure gradients) up to Reynolds numbers
(based on distance along the body surface, the local Reynolds number discussed above) of
Rex = 2 × 106 and for background turbulence levels up to 0.2% [Schubauer & Skramstad,
1948]. Furthermore, along streamlines with increasing velocities (that is, with favorable
pressure gradients), laminar boundary layers are observed up to much higher Reynolds
number, Rex 2 × 106 [Bourgoyne et al., 2003]. Laminar separation has also been observed
upstream of cavities on bodies of revolution for Rex 1.36 × 105 by [Arakeri, 1975] and
other researchers. It has also been shown in [Arakeri & Acosta, 1976] that the character of
431
the boundary layer – that is laminar versus turbulent – is certainly important for cavity
detachment.
Fig.11.1.2.
Pressure distributions on a body
with a hemispherical head-form computed
using CCVL for Re = 1.36 × 105 . Stars
show experimental data (after [Lindau et
al., 2002]). One can see that a prediction
that misses the pressure minimum can also
be in good accordance with the measured
data, if the data are sparse.
For the UTCC models,

comparisons of the computed
dimensionless pressure coefficient with its
values measured at several points have been frequently used as a major validation. Good
agreement of computed and measured values of the pressure coefficient, CP, has been
reported in several papers. However, as seen in Fig.11.1.2, theoretical results based on an
entirely different model may also agree very well with such experimental data, especially in
the vicinity of cavity detachment. Note, however, that a model that does not predict the
pressure minimum observed just upstream of cavity detachment in more detailed
experiments might be wrongly considered valid in that region, if the data used for
comparison are sparse, as are those in Fig.11.1.2.
Finally, no UTCC computations have been published for a range of Reynolds
number. Thus, there is not yet sufficient information to validate the ability of UTCC to
analyze the effects of Reynolds number on cavitation; more comparisons are required.
The second concept used for analyzing viscous cavitating flows – CCVL – involves
matching models of cavitation in an ideal fluid with boundary layer models [Amromin,
1985; 2002; Brewer & Kinnas, 1997; Chahine & Hsiao, 2000]. The terms used in this
concept are the traditional ones for cavitation experiments (for example, cavity length).
CCVL employs the traditional division of the whole flow into viscous layers (boundary
layers and a wake) surrounded by irrotational flow of an ideal fluid. The viscous-inviscid
interaction is mainly predetermined by two effects: the effect of the pressure gradient on the
thickness of the viscous layers, and the inverse effect of this thickness on the boundary shape
between the viscous and inviscid parts of the flow. The ability to correctly compute laminar
boundary layers upstream of the cavity has been proven for CCVL.
The following analysis of cavity detachment has been carried out using the approach
based on CCVL. In this model, both the Reynolds and Weber numbers are included as
potentially important factors, and it is found that the Weber number is indeed important in
the vicinity of cavity detachment points.
11.1.2. CCVL Model of Sheet Cavitation
CCVL can be looked at as a modification of the ideal cavitation model and, compared to this
traditional concept, can be nearly as easily explained. The flow schemes and pressure
distributions in the meridian semi-plane for the so-called axisymmetric ITTC body (a
standard form used in cavitation experiments) are presented in Fig.11.1.3 for both models.
432
Fig.11.1.3.
Comparison of ideal cavitation and CCVL: The top
plot shows cavity shapes. A Riabouchinsky wall is
the typical closure model for the ideal cavity,
whereas a separation-reattachment zone exists just
downstream of cavity closure in the CCVL model. A
separation-reattachment zone also exists upstream
of the cavity detachment point in the CCVL model
(too small to be visible in the upper figure). The
arrows show the direction of the time-averaged flux
of liquid through the cavity in the CCVL model. The
middle plot compares the corresponding pressure
distribution over the body in the vicinity of cavity.
The bottom plot shows the cavity detachment zone
in more detail; here β is the angle between the outer
normal to liquid on the liquid-cavity and liquid-
body boundaries.
There are three main regions in the

meridian sections shown: the body, the cavity, and the liquid. Although the boundary
between the second and third regions is initially unknown, in both models only the flow in
the liquid region is considered: as is typical of many of the cavitation models presented
throughout this book, the effect of the cavity on the flow is captured via accurate modeling
of the flow outside the cavity bounding streamline. A velocity potential Φ is used in both
concepts, but the potential flow does not make contact with the surfaces of the body and the
cavity in the CCVL model: it is separated from these surfaces by the system of boundary
layers.
As observed by [Arakeri, 1975], boundary layer separation takes place just upstream
of the cavity detachment point. If the same body is free of cavitation, viscous separation
would not occur at the same location for the same Reynolds number: this separation is
caused by the cavity itself. Because of surface tension, the cavity detachment region cannot
be infinitely thin, and the presence of the cavity induces separation similar to the case of
fully wetted flow over a surface with an irregularity or obstacle that is submerged within the
boundary layer. Viscous separation just downstream of cavity closure is caused by the re-
entrant jet. Thus, there are two viscous separation zones between the detachment of the
boundary layer and its reattachment to the body, with the cavity located in between.
The comparison of the two computational results in Fig.11.1.3 is made at the same
value of σ. The difference in the cavity lengths and pressure distributions is caused mainly
by the difference in the cavity detachment. For ideal cavitation, the pressure in the cavity is
the lowest pressure that occurs on the body surface and the detachment should be determined
from the above-mentioned Brillouin-Villat condition. Contrary to that result, the cavity
detachment in the CCVL (and in reality) is coupled with the boundary layer separation;
therefore, the lowest pressure points are located upstream of the cavity in this model.
Similar to boundary layer theory for non-cavitating flow systems, the boundary
layers are assumed thin, such that the pressure variation across them is negligible to lowest
433
order. The full formulation would involve asymptotic matching of an outer potential flow to
an inner boundary layer flow, the details of which are not presented here. (For further details
of such an approach as applied to non-cavitating flows, see, e.g., [Van Dyke, 1975, p121ff].)
The outer flow cavitation problem in CCVL is the free-boundary potential flow
problem defined by ΔΦ=0 (11.1.1)
∂Φ
=0 (11.1.2)
∂N S
2
grad (Φ ) = 1 − CP* ( x) (11.1.3)
S*
where S is the entire boundary of the inviscid flow field. These equations are written in
dimensionless form, with all lengths normalized by the appropriate length scale, either d or
c , as discussed above. The pressure coefficient is denoted as CP ≡ ( p − p∞ ) (1 2 ρU ∞2 ) ,
where p is the static pressure, p∞ is its value far upstream of the foil system, ρ is the
density of the ambient liquid, and U ∞ is the inflow velocity far upstream. The value of the
pressure coefficient on the free surface S * defining the boundary of the inviscid region
extending from the upstream boundary layer separation point to the downstream boundary
layer reattachment point, is denoted CP* ( x ) . As Fig.11.1.3 shows, the right-hand side of Eq.
(11.1.3) is constant over the major part of S * . This takes place in ideal cavitation as well,
but unlike that case, there are variable pressure regions near the ends of the cavity due to the
small-scale physics of surface tension and viscosity that are included in the CCVL model via
matching with the inner flow. The inner flow along the body-cavity system is presumed on
the basis of experimental observations to traverse five distinct zones:
• a wall boundary layer upstream of the cavity, in the region x ≤ X 0 ;
• an upstream separation-reattachment zone straddling cavity detachment, X 0 ≤ x ≤ x * ;
• a shear layer in way of the cavity along its boundary, x* ≤ x ≤ X 2 ;
• a downstream separation-reattachment zone near cavity closure, X 2 ≤ x ≤ X 3 ; and,
• a wall boundary layer downstream of the cavity, x ≥ X 3 .
As in boundary layer theory for non-cavitating flows past bodies of finite length, the
downstream wall boundary layer leaves the end of the body to form the viscous wake. In the
above descriptions and in the remainder of this section, the following key abscissae have
been defined: X 0 , the abscissa of the boundary layer separation point upstream of the cavity;
X 1 and X 2 , the abscissae of the upstream and downstream cavity endpoints; x * , the
downstream end of the upstream separation zone (which is downstream of the upstream end
of the cavity: x*>X1); and X1, the abscissa of the boundary layer reattachment point
downstream of the cavity. A quantity ξ also appears in the following formulation; it is
x − X2
defined as ξ ≡ , where X 4 ≡ ( 5 X 3 − 2 X 2 ) 3 , also unknown at the start of the
X4 − X2
computation.
The pressure coefficient, CP* = 1 − U 2 ( x ) , appearing in Eq. 3 can be expressed in
terms of the local velocity at the edge of the boundary layer, U ( x ) , along with two
434
undetermined coefficients C1 and C2 , and certain of the key points along the body-cavity
system, x * , X 0 , X 2 , and X 3 , defined above, each of which is also unknown at the start of
any CCVL computation, with the exception of X 2 , which is specified as a parameter of the
problem. With this approach, the edge velocity is given by
⎧C1 ( x − x *)2 + 1 + σ , X0 ≤ x ≤ x *
⎪⎪
U ( x) = ⎨ 1 + σ , x* ≤ x ≤ X 2 . (11.1.4)
⎪
⎪⎩C2ξ ( 3 − 2ξ ) + 1 + σ , X 2 ≤ x ≤ X 3
2
Eq. (11.1.4) is not a theoretical solution but it smoothly approximates experimental CP

in cavitating flows. The computed pressure distribution in Figs.11.1.2 & 11.1.3 was obtained
with Eq. (11.1.4). The above coefficients and abscissas should be found with matching
conditions for the five different flow zones listed above, which require four matching
conditions where each of two zones meet.
As for many of the other cavity flow models discussed in this book, the cavity
closure location, x = X 2 , is specified as a parameter of the computation. Since the objective
is to determine the point of cavity detachment from a smooth surface, in the CCVL model
the location of the cavity detachment point, x = X 1 , is also specified as a parameter. The
Reynolds number of the flow, Re , is also specified as a parameter. The upstream boundary
layer separation point, x = X 0 , the upstream boundary layer reattachment point, x = x * , the
downstream boundary layer reattachment point, x = X 3 , and the cavitation number, σ , are
all determined as part of the solution of the matched problem.
The boundary layer separation and reattachment points are determined using semi-
empirical conditions of boundary layer separation from a wall as presented in [Castillo et al.,
2004]. [Roberts, 1980] validated the application of such criteria for studying reattachment to
∂C
a wall. The general form of such conditions is as follows: δ * P = a* (11.1.5)
∂s
Here δ* is the boundary layer displacement thickness; a* is an empirical coefficient that
has different values for laminar and turbulent boundary layers, for separation and
reattachment. In producing the results presented herein, the value a* = 0.015 was used as the
reattachment criterion. Two criteria Eqs. (11.1.5) can be generally used for integral and
differential methods. There is also a second reattachment condition at X 3 that is the
consequence of Prandtl’s definition of the so-called body of displacement representing the
displacement of the outer flow streamlines by the boundary layer in the lowest-order
matching of the inner and outer flow: δ*(X3)=h(X3) (11.1.6)
Here the function h ( x ) is the distance between the body surface and the boundary
of the inviscid region, S . The value h(X3) is determined by solving Eqs.(11.1.1)-(11.1.4).
The displacement thickness δ * is determined as a result of the boundary layer prediction of
the inner problem, accounting for jumps at the cavity closure point, x = X 2 , in the values of
both the displacement thickness, δ * , and the boundary layer momentum thickness, δ ** ,
caused by the presence of the re-entrant jet. The jump in the value of δ * is calculated in
terms of the Riabouchinsky wall height, Yc ( X 2 ) , as Δδ * = τ Yc ( X 2 ) with τ = 0.77 . For the
laminar velocity profile used here, the corresponding jump in the value of δ ** is related to
435
the jump in the value of δ * by the expression Δδ ** Δδ * = 141 845 .Value h(X3) should be
obtained by solving Eqs. (11.1.1)-(11.1.4), whereas δ* is the result of computing the
boundary layer with jumps of δ* and the boundary layer momentum thickness δ** at x=X2
taken into account. These jumps are caused by the presence of the re-entrant jet. The jump of
δ* is calculated as τYC(X2) at τ=0.77. For the velocity profile used here, the corresponding
δ** jump is 141/845 of the δ* jump. It is assumed that dU/ds is continuous at x=X1 and
this derivative is used for determining C1 with Eq. (11.1.4).
To determine x*, there is the evident condition that the total displacement of the
streamlines from the body surface must account for the cavity and the boundary layer
enveloping it: h ( x *) = δ * ( x *) + Yc ( x *) . By definition, the cavity meets the body boundary
at the cavity detachment point: Yc ( X 1 ) = 0 with a slope dYc dx X that depends on the angle,
1
β , between the cavity and body surface normal vectors at that point, as defined in
Fig.11.1.3. Downstream of this point, computation of the cavity geometry accounts for
surface tension, which is small over most of the cavity where the curvature is gentle, but is
substantial in any region of cavity-liquid-body contact because there the cavity surface
curvature is very high. For example, [Arakeri, 1975] found that the curvature at cavity
detachment is greater than1 δ , where δ is the basic boundary layer thickness. The
expression of equilibrium among surface tension and the pressure internal and external to the
cavity at the cavity detachment point has the following form:
−3/ 2
2 d2 ⎡ ⎛ d ⎞ ⎤
2
C ( X1 ) + σ =
*
P Yc ( X 1 ) ⎢1 + ⎜ Yc ( X 1 ) ⎟ ⎥ . (11.1.7)
We dx 2 ⎢⎣ ⎝ dx ⎠ ⎥⎦
The β-dependent value of dYC /dx on the body surface at x=X1 should be used for integrating
Eq. (11.1.7).
With the values of the Reynolds number, Re , and the cavity detachment and closure
points, X 1 and X 2 , specified, and after imposing the matching conditions described above,
the general flow parameters remain to be determined: the cavitation number, σ , and the
Weber number, We . The latter is determined from Eq. (11.1.7) at the cavity detachment
point, x = X 1 . The value of the cavitation number must be determined by specification of a
cavity closure condition. Continuity of the surface normal, N , at the smooth boundary layer
reattachment point, x = X 3 , can be employed as the closure condition. As noted in
[Amromin, 2002; 2007], determination of h ( x ) is based on iterative solution of the quasi-
linearized Eqs. (11.1.2) and (11.1.3): ∂ϕ / ∂s = U − U * and q / 2 = ∂ ( h * U *) / ∂s , where the
distributions U * and h * are the distributions of U and h as determined in the previous
iteration of the computation. Here q is the source strength producing a perturbation
potential, ϕ . As follows from the second of these equations, continuity at the boundary layer
reattachment point is possible for only a unique solution of q , since the derivative in the
second equation can be computed from quantities known from the previous iteration.
However, since U is computed in the current iteration, the first equation must be solved as a
436
singular integral equation that has a unique solution for q provided the additional
s3
U −U *
σ − dependent existence condition is imposed: ∫
s0 ( s3 − s )( s − s0 )
ds = 0 . (11.1.8)
Here s0 and s3 are arc-length coordinates corresponding to x=X0 and x=X3. The edge
velocity, U, is defined by Eq. (11.1.4).
Thus, the complete problem of sheet cavitation in the CCVL model is a
superposition of nonlinear problems. The procedure to match the solution of the inviscid
flow outside the boundary layer and the solution of the boundary layer itself is based on
tuning the boundary between viscous and inviscid flow to the pressure distribution along this
boundary by the methods of potential theory. The procedure accounts for the boundary layer
criteria that result from this distribution. The iterative solution procedure includes the
following steps.
1. For a fixed Reynolds number, Re, and specified cavity endpoints, {X1, X2}, a preliminary
cavity shape is estimated by solving the outer flow problem subject to the assumed form
of the edge velocity, Eqs. (11.1.1)-(11.1.4). The pressure coefficient, CP, is computed
over the entire body-cavity surface, S, using BEM. (Thus, the numerical accuracy of the
entire solution is predetermined by the accuracy of this latter computation.).
2. The boundary layer solution is marched downstream from the foil or body leading edge
to the upstream boundary layer separation point, x=X0, which is determined from Eq.
(11.1.5). The Weber number, We, is determined from Eq. (11.1.7). The coefficient C1
appearing in Eq. (11.1.4) is determined by imposing continuity of the arc-length
derivative of the edge velocity, dU/ds at the cavity detachment point, x=X1.
3. The cavitation number σ and coefficient C2 appearing in Eq. (11.1.4) are found from
Eq. (11.1.5) at x=X0 and Eq. (11.1.8). The combination of these equations gives a
quadratic equation for C2. The value of the downstream boundary layer reattachment
point, x=X3 is adjusted iteratively to satisfy Eq. (11.1.6), each of these iterations involves
computing the boundary layer parameters downstream of the upstream boundary layer
separation point x=X0.
Of special importance in attached cavitation is to determine how derivative ∂CP/∂s in
Eq. (11.1.5) depends on cavity shape. Since the cavity detachment point is submerged in the
boundary layer, it affects velocities as a surface irregularity. The degree of this effect can be
estimated by the analogy between the meniscus of radius r* separating the cavity detachment
point and a source. The latter induces the normal velocity proportional to δ*λ-2 r*(1+cosβ)
and this also should be taken into account in calculating CP. Therefore, the following form
of Eq. (11.1.5) describes the separation of laminar boundary layer upstream of the cavity:
CL D δ * ∂U [1 + cos( β )]2
+ < (11.1.9)
δ * Re U ∂x (CC λ / δ *)3
Here angle β depends on the body surface wettability. Employing the Falkner-Skan laminar
boundary-layer profile [e.g., Schlichting, 1979; White, 1974], results in a value of the
parameter in the first term of Cβ = 1.1 , whereas the value of Cc is the result of tuning to
empirical data obtained from experiments with cavitating flows.
The boundary-element method for determining derivatives of the total potential Φ
and integral methods for predicting the behavior of the wall boundary layer are commonly
known. The correction procedure for the two-dimensional free boundary S * employing
437
auxiliary functions ϕ and h * is more fully described in [Amromin, 2002; 2007]. Additional
detail is required to fully understand the computation of the cavity boundary layer over
cavities. The following dimensionless velocity profile is employed:
u (η ) = u / U + (1 − u / U )η 2 (3 − 2η ) (11.1.10)
Here η=(y−Yc)/δ; δ is the basic boundary layer thickness; u is the velocity at the cavity
boundary η=0. The unknown parameters u(x) and δ(x) can be determined from two
equations. The first is the Karman - Poulhausen differential form of the momentum integral
equation [see, e.g., Schlichting, 1979], which takes the following form when applied to flow
dU d δ **
along the cavity boundary: U (δ * +2δ **) + U 2 + v(U − u ) = 0 , (11.1.11)
dx dx
where v is the velocity component of the liquid normal to the cavity surface (Fig.11.1.3). In
Eq. (11.1.11) friction at the cavity surface is ignored. The other equation is the momentum
differential equation for the point of zero friction on the body surface (that is the point of
separation within the thin boundary layer) in the Clauser’s form, see [Gogish, et al., 1979]:
du (U − u )3 dU
u = 0.015 −U (11.1.12)
dx Uδ * dx
11.1.3. Results of Predictions of Partially Cavitating Flows of Real Fluids Using
the CCVL Model
Different grids are used for different parts of the problem. When computing the edge
velocity, U , using a boundary-element method in two-dimensional and axisymmetric flows,
from 60 to 100 boundary elements are typically used, with element length varying widely
from 0.0003 chord lengths at the key abscissae to 0.05 chord lengths elsewhere. High-order
Gaussian quadrature is used within each element. In order to correct the unknown cavity
surface, S * , another non-uniform two-dimensional grid with 30 to 40 elements is generated.
To integrate the boundary layer differential equations, the grid density is increased by a
factor ranging from 400 to 8000; variations of boundary layer parameters between neighbor
nodes should be less than 1%.
Fig.11.1.4.
Computed cavitation number, σ , as a
function of foil chord length, c , for various
values of the abscissae of the upstream and
downstream extents of the cavity, X1 and , X2
(specified as parameters of the CCVL
problem), for the EN-hydrofoil at a specified
Reynolds number, Re = 1.5 × 106 .
As noted above, for given X1 and X2

at fixed Re, the corresponding pairs {σ, We}
can be determined using the CCVL approach. Alternatively, this pair of results can be
replaced with cavitation number and characteristic length, for example, d as defined above.
Determination of the cavity for a given cavitation number-characteristic length pair requires
mapping the specified values of the cavity endpoints in a form similar to that of Fig.11.1.4,
which in this case is for the EN-hydrofoil with an elliptic nose tested by Yamaguchi and
438
Kato. Such fitting leads to numerical results that are in good agreement with the measured
values of cavity length, as justified by a comparison with Fig.11.1.1.
As seen from Fig.11.1.4 the body characteristic length (for example, the foil chord
length) has a very small effect on the cavity closure point, X 2 (σ , c ) ≈ X 2 (σ ) , whereas the
cavity detachment point, X 1 (σ , c ) depends significantly on both the cavitation number
and the body characteristic length. The influence of the tuning coefficient, CC – defined in
the context of Eq. (11.2.9) – on the results depends on ∂U ∂x . For example, in the case of
an axisymmetric body with a hemispherical head operating at a Reynolds number of
Re=2×106 with a cavity length of L = 0.45d , a 30% increase of CC in Eq. (11.1.9) from
its tuned value of 0.0164 results in a 30% decrease in the predicted value of d , with σ
remaining virtually constant. For CC → ∞, d would be reduced threefold while σ would
be reduced by only 2.5%.
11.1.3. Cavity Detachment on Axisymmetric Bodies
Figs.11.1.5 & 11.1.6 present the predicted experimental positions of cavity detachment in
axisymmetric flows. Axisymmetric flows as the primary test cases are usually preferable,
since the absence of lift eliminates certain additional effects which may cause difficulties in
comparison. The cavity detachment location along a slender axisymmetric body with a
hemispherical head-form is presented in Fig.11.1.5 as the curvilinear coordinate of the
detachment point along the meridian section of the body. As seen in the figure, the CCVL
model provides a reasonable description of the effect of flow speed
Fig.11.1.5. 1.6
Comparison of theoretical (lines) and measured

Cavity detachment
(symbols, [Arakeri, 1975]) positions of cavity

detachment on an axisymmetric body of d = 0.045m 1.4
with a hemispherical head-form. The arc-length

coordinate of the detachment point is normalized by
the head-form radius, d 2 . The number in the legend 1.2
0.24 0.36 0.48 0.6
indicates the free stream speed, U ∞ in feet per second Cavitation number
55-exp 30-exp 55-th 30-th

The comparison of theoretical results with
experimental data [Amromin & Ivanov, 1982] for the ellipsoidal body in Fig.11.1.6 has a
double significance: it confirms the theory for another body shape and emphasizes the roles
of surface tension and body wettability: the angle β introduced in Fig.11.1.3 is close to zero
for metals, but is approximately π/2 for Teflon®. The axisymmetric body profile is defined
by the equation
2 2 2
⎛ x ⎞ ⎛ 2 y ⎞ ⎛ 2z ⎞
⎜ ⎟ + ⎜ ⎟ + ⎜ ⎟ = 1.
⎝d ⎠ ⎝ d ⎠ ⎝ d ⎠
439
Fig.11.1.6.
Computed and measured [Amromin & Ivanov,
1982] abscissas of cavity detachment Xo on
ellipsoids (x/D)2+(2y/D)2+(2z/D)2=1 made
from different materials. D=0.05m, U∞=18m/s.
Curve B-V was predicted with the ideal ideal
cavitation model based on the Brillouin-Villat
cavity detachment condition described above.
The differences in cavity detachment

for smooth bodies lead to differences in the cavity length. For cavities with the detachment
points shown in Figs.11.1.5 & 11.1.6, the detachment effect on the cavity length is shown in
Fig.11.1.7. The observed cavity length for the Teflon® ellipsoid [Amromin & Ivanov, 1982]
had variations from a value of c = 1.75d at a cavitation number of σ = 0.2 to a value as
low as c = 0.65d at cavitation numbers approaching σ ≈ 0.3 ; thus, the advantages of the
CCVL model relative to the ideal cavitation model increase with increasing values of the
cavitation number. For a body with a
hemispherical head-form, the situation is similar.
Fig.11.1.7.
Effect of cavity detachment on cavity
length: Solid curves give results of CCVL for D
and U∞ used in Figs.11.1.5 & 11.1.6, dashed
curves give the results of ideal cavitation with
The following consideration of 3D cavitating flows around axisymmetric bodies is

limited here by 3D flows at small angles of attack α. For these non-axisymmetric flows, the
inviscid flow was computed with the 3D approach, whereas the boundary layer along the
ensemble of 3D streamlines of inviscid flows around these bodies was computed with the 2D
equations. Further, the 3D cavity pattern is computed as a combination of 2D cavities in the
artificial non-uniform 2D incoming flows, with the initial pressure distributions determined
from the above 3D inviscid problem. The mutual influence of these cavities is supposed to
be negligible.
The application of this approach to a 12” model is compared in Fig.11.1.8 with
observations [Gorshkov, A., as cited without reference in Amromin, 2007] made in the KSRI
re-circulating water tunnel, which features a 1.3m×1.3m test section with rounded corners.
The wall effect was taken into account in reporting the values of the cavitation number, σ .
The Reynolds number in this experiment was quite high, but the boundary layer upstream of
the cavity remained laminar. The Michel criterion [Michel, 1951] was used for the prediction
of natural transition to turbulence as was done in [Bourgoyne et al., 2003]. According to
computations, there is no significant scale effect on cavity detachment on the upper part of
the body, but the observed asymmetry of the cavity disappears when the Reynolds number is
increased by a factor of ~11; the corresponding increase in the Weber number is by a factor
of ~44. A similar matching of solutions for three-dimensional cavitation-free flow with
quasi-two-dimensional solutions for cavitating flows was also successfully employed for
440
predicting cavitation inception and growth on marine propeller
blades [Amromin et al., 1995].
Fig.11.1.8.
Cavitation patterns on the body of revolution at α=3°. Body head
similar to the ITTC body head, but a little bit blunter (its block
coefficient was 2.5% greater). Solid lines show the body contour
(abscissa is normalized by D). Dashed and dotted curves limit
the computed cavity patterns. Observed cavity patterns are
shown by double-side arrows.
11.1.4. Cavitation Detachment and Cavitation Inception

A challenging problem associated with determining the cavity location and size is the
prediction of incipient cavitation number (or cavitation inception number), σI. The incipient
cavitation number is defined here as a maximum possible value of σ in the steady flow at
fixed Re and characteristic length (in this case, the diameter of the axisymmetric body, d )
above which cavitation does not occur. This definition is used in both contemporary
experimental practice and theoretical studies.
Cavitation inception can be significantly affected by water quality [Arndt, 2002].
Though the current theory does not take water quality effects into account, it is possible to
compare this theory with experimental data obtained in facilities with low gas content. For
such facilities, the gas pressure within incipient cavities differs insignificantly from the
vapor pressure of the ambient fluid, and the actual cavitation number is very close to the
vaporous cavitation number. It is important to emphasize that in CCVL there are two
solutions for the cavity length, c (σ ) , within a given range of the cavitation number, and
the cavity occurring at the highest cavitation number is not the cavity of minimum size.
Fig.11.1.9 presents the computed dependence between cavitation number and cavity length
for an ITTC body at a fixed combination of the Reynolds number, Re , and Weber number,
We . The values of cavitation number do increase with cavity length along the lower branch
of the curve, but this branch is only possible in the mathematical sense. In reality, the
cavitation number is a decreasing function of cavity length, and the decreasing branch of it
in Fig.11.1.9 is in good agreement with the experimental data presented in [Ceccio &
Brennen, 1992], provided the pulsation of the cavity length is taken into account.
Fig.11.1.9.
Cavitation number as a function of cavity
length over an ITTC body. The trianglesr and
diamonds show the observed maxima and
minima of the cavity length, respectively; the
curve approximates the result computed using
the CCVL model. The curve’s branch to the
left of its maximum is a non-physical
mathematical artifact..
Fig.11.1.10 presents the incipient cavitation number as a function of Reynolds

number and axisymmetric body diameter for bodies with a hemispherical head-form in
441
facilities with low air content, comparing results predicted by CCVL computations with
experimental data. The data used for comparison was acquired by [Billet & Holl, 1981] and
[Arakeri & Acosta, 1976].
There are two different trends there. The first trend includes results for a body with
turbulence stimulator only. The second trend accumulates all other results. The first trend is
related to the flow with the forced transition at abnormally low Re values. This trend was
obtained with a simplified method of simulating the stimulated transition (by 200% jumps of
δ* at the point of stimulator location) and with the oversimplification of the turbulent
boundary layer separation criterion upstream of the cavity, using it in the form λ=2δ*, after
[Amromin, 1985]. These simplifications resulted in a quantitative disagreement between our
prediction and the measurements by [Arakeri & Acosta, 1976], but a qualitative explanation
of the effect of stimulated transition on cavitation can be done with CCVL.
0.75
Cavitatio n in cep tio n n u m b er

Fig.11.1.10.
2
Cavitation inception number for bodies with
5
hemispherical head vs. Re for different D.
5-T
Symbols 2 show measurements with low air 0.5
2
content for D=0.02m, 5-T show measurements
5
for D=0.05m with a turbulence stimulator, 5
5-T
show measurements for this body without the
stimulator. Indexes of our computational 0.25
curves correspond to symbol indexes. 5.4 5.6 5.8 6
Log Re
The data of the second trend do not fit
in one dependency: there is a variation of the incipient cavitation number, σ i ( Re ) , with
body diameter, d . The larger bodies are associated with lower incipient cavitation numbers,
a behavior that is captured by both the CCVL model and experiments (also noted by Keller
[2001]). Thus, the ability of the CCVL model to correctly predict the incipient cavitation
number for sheet cavitation depends on the ability to accurately predict cavity detachment.
However, there are only two conservation laws in the integral form for two-dimensional and
axisymmetric incompressible boundary layers. Therefore, an integral method can work with
only two unknown functions (for instance, δ * and δ ** ). As a result, the velocity profiles
can depend on two parameters only. This limits the accuracy of predictions with the CCVL
model, as is manifest in the trend for stimulated turbulent flow in Fig.11.1.10. However, the
predictions are qualitatively correct.
Turbulent boundary layers can evolve without separation in a much more adverse
pressure gradient, ∂CP ∂s , than can laminar boundary layers. Since the cavity detachment
point is preceded upstream by a separation zone, as described above, this condition tends to
drive the equilibrium position of cavity detachment downstream, especially at low values of
the Reynolds number, which generally correspond to low values of the Weber number.
Fig.11.1.11 presents the predictions using the CCVL model for three different conditions:
two at moderate values of the Reynolds number, Re = 1 × 105 – one for a fully turbulent
boundary layer and one for laminar flow – and a third for a fully turbulent boundary layer,
but at a much higher Reynolds number, Re = 1 × 107 . For the case of the moderate value of
the Reynolds number with the fully turbulent boundary layer, separation is suppressed
442
upstream of the cavity detachment point, and equilibrium is not achieved for cavity
detachment in the low pressure region just forward of the intersection of the hemispherical
head-form and the cylindrical body: cavity detachment does not occur until downstream of
the intersection point, even at the relatively low value of the cavitation number specified,
σ = 0.3 , and the resulting cavity is quite limited in extent. Conversely, for the laminar
boundary layer at the same Reynolds number, the flow tends to separate more easily, the
equilibrium cavity detachment point can migrate forward into the low pressure region, and
the resulting cavity is much more extensive. For the case of a significant increase in
Reynolds number, the equilibrium cavity detachment point can migrate forward into the low
pressure region even for the fully turbulent boundary layer, as shown, and the resulting
cavity size approaches that of the
case in laminar flow.
Fig.11.1.11.
Boundary layer effect on cavity
shapes for σ=0.3 on a body with
hemispherical head. Here:
1 - body contour; 2 – pressure
coefficient Cp for cavitation-free
flow. The other curves are the
predicted cavity boundaries for the
3 - for Re=5×105, body diameter
d=0.05m and a fully turbulent
boundary layer; 3a – for the same
Re and d and a laminar boundary layer; 3b - for Re=107, d=1m and fully turbulent
boundary layer.
Sec. 11.2. High-Speed Motion of a Body through a Bubbly Liquidxi

As has been discussed in many sections of this book, supercavitating bodies can achieve
very high speeds under water by virtue of reduced drag: with proper design, a cavitation
bubble is generated at the nose and skin friction drag is drastically reduced. Depending on
the type of supercavitating vehicle under consideration, the overall drag coefficient can be an
order of magnitude less than that of a fully-wetted vehicle.
It is well known that the presence of bubbles in a liquid reduces the sound speed of
the medium. (See, for example, [Brennen, 1995].) Even at small void fractions, this effect
can be dramatic. For example, for air in water at standard atmospheric conditions, even a
concentration of 1% reduces the sound speed to only approximately 120m/s from its value of
approximately 1500m/s in the pure liquid. Moreover, if the finite radii of the bubbles is taken
into account, it can also be shown that the reduced sound speed is frequency dependent.
(See, for example, [Drew & Passman, 1999].) Thus, a body that moves at high-speed
through a bubbly mixture – whether or not it is supercavitating – is expected to operate at a
higher effective Mach number than a body moving through a pure liquid. This suggests that
xi
Dr. John R. Grant made significant contributions to the mathematical formulation of the model
discussed in this section.
443
a wave system will develop at much lower speeds in a bubbly flow than would be observed
in essentially pure liquids. Moreover, because the sound speed is frequency dependent, it is
anticipated that the wave pattern associated with a high-speed body moving through a
bubbly mixture will be more complicated than the rather simple shock wave structure
observed in essentially pure liquids [Kirschner, 1997; Hrubes, 2001] presented in Fig.1.1.2
and discussed in greater detail below.
The physics discussed in this section is intimately rooted in wave theory. As is
discussed in Chs. 1 and 7, a body moving at high speeds relative to the sound speed of the
ambient liquid or bubbly mixture is subject to the effects of compressibility. As is well
known, for single phase flow under conditions of Mach numbers approaching or exceeding
unity, the disturbance represented by the body moving through the medium generates a wave
field that radiates energy away from the disturbance, and therefore is associated with
increased drag on the body. Thus, it is anticipated that a projectile moving through a bubbly
liquid will be subject to wave drag at lower speeds than one traveling through a pure liquid.
This may have two important implications for research into supercavitation: (1) a
supercavitating projectile traveling through a real fluid that happens to be populated with
bubbles may experience a significant wave drag component in addition to cavity drag and
any other drag contributions; and, (2) Introducing bubbles into a laboratory research facility
may allow experimentation at high effective Mach numbers without the need for achieving
sonic speed with respect to essentially pure water, with attendant reductions in cost and
hazard.
[Grant & Kirschner, 2003] recognized the similarity of the problem of high-speed
motion of a point source in a bubbly mixture to Kelvin’s problem for a ship wake, where the
wave speed is also frequency dependent. They further presented a formulation for predicting
the wave patterns that can be expected in axisymmetric flow and developed Green’s function
for the potential due to a source in the bubbly mixture. In [Grant and Kirschner, 2004], this
line of development was continued to consider the wave drag acting on an object moving at
high speed in a bubbly mixture. In [Grant & Kirschner, 2005; 2006], the formulation was
extended to the case of a slender body moving at high speed through a bubbly liquid, and
estimates of the drag were presented. Ongoing research is extending these results to the case
of high-speed motion of an axisymmetric supercavitating body moving through a bubbly
liquid.
This section presents the analyses formulated in those references, which, although
linear, are suitable for predicting the wave patterns associated with a high-speed body
moving through a bubbly mixture and for drawing conclusions about the effects on drag.
First the results for a point disturbance are presented, and the computed field values of the
associated Green’s functions are presented and discussed. Once this groundwork is
developed, the results are extended to a slender body. This line of development is then
continued to consider the wave drag acting on an object moving at high speed in a bubbly
flow. The importance of these results is that wave drag due to supersonic effects is predicted
to occur at significantly lower speeds in a bubbly mixture than would be the case in the pure
liquid. Although the wave drag also decreases somewhat as the void fraction increases, it is
still comparable with wave drag for the case of very low void fraction (which may therefore
be thought of as representative of the single phase case), and therefore should not be
neglected.
444
Of course, any body moving through a bubbly liquid at the speeds being considered
in this section will display a strong tendency to cavitate, so the ultimate objective of this
thread of mathematical physics is a model that accounts for the occurrence of cavitation, a
topic that is the subject of ongoing research. However, the results presented herein are
limited to the motion of a solid, non-cavitating body, as an illustration of the basic physical
effects involved. It should also be noted that the predicted behavior has not been established
experimentally; although the well-established theory of the wave pattern associated with the
Kelvin wake of a surface ship provides a strong physical analogy that provides confidence in
the results presented (as is discussed in detail below); it will be the task of future laboratory
research to determine if theory matches reality.
The results of the model presented in this section predict that the character of the
flow changes depending on the mixture Mach number: the topology of the wave pattern is
predicted to undergo fundamental changes depending on the speed of the motion with
respect to both the mixture and liquid sound speeds. For the case of a point disturbance,
three regimes are of interest, characterized by propagation via a bubble mode, an acoustic
mode, or a combination of the two. Also, in certain regimes, the solution has two branches
which are associated with transverse and diverging waves, respectively. As will be
presented, the physics of a slender body is even more complicated, with additional regimes
of behavior appearing in the theoretical predictions.
11.2.1. General Formulation for Irrotational Flow in a Bubbly Liquid
This section presents a derivation of the equation for the potential flow of a body moving
through a bubbly liquid. The next section develops the linear approximation to this equation;
this result will be applicable, for example, to a slender body. Note that the notation used in
this section is independent from that of Sec. 11.1 (and, in general, from the other chapters of
this book).
Let Ψ denote the bubble number density and R the bubble radius, so that the gas
4
void fraction is given by α = π R 3 Ψ. (11.2.1)
3
The equations of conservation of mass in the gas and liquid (see [Drew & Passman, 1999,
p274ff]) are expressed in terms of the gas density ρ g and the liquid density ρ as
∂ (αρ g )
+ ∇ ⋅ (αρ g u g ) = 0 (11.2.2)
∂t
∂ [ (1 − α ) ρ ] + ∇ ⋅ ⎡ 1−α ρ u
and
∂t ⎣( )
⎤⎦ = 0 (11.2.3)
where u g is the bubble velocity, and u is the liquid velocity. The equations of motion for
the bubble and liquid phases are, respectively:
⎛ ∂u ⎞
αρ g ⎜ g + u g ⋅ ∇u g ⎟ = −α∇pg − f − ∇ ⎡⎢αρ Ci u g − u ⎤
2
(11.2.4)
⎝ ∂t ⎠ ⎣ ⎦⎥
⎛ ∂u ⎞
(1 − α ) ρ ⎜ + u ⋅ ∇u ⎟ =
and ⎝ ∂t ⎠ (11.2.5)
− (1 − α ) ∇p − ( pg − p ) ∇α + f − ∇ ⎡⎢(1 − α ) ρ Cr u g − u ⎤.
2
⎣ ⎦⎥
445
Here pg ( p ) is the pressure in the gas (liquid) and f is the interfacial force density. The
constants have the values Ci = 3 / 10; Cr = 1 / 5 (see [Drew & Passman, 1999, p274]).
The physical features of interest, namely, the effects of compressibility of the liquid-
bubble mixture, will not depend primarily on the difference in velocity of the two phases.
Hence for this formulation, take u g = u = u and add (11.2.4) and (11.2.5) to obtain
du 1 du p +αρ P dp
=− ∇ ( p + αρ P ) = = −∇ ∫ (11.2.6)
dt ρm dt p ∞ + α ∞ ∞ ∞ ρ
ρ P
m
d ∂
where the material derivative is ≡ + u⋅∇ (11.2.7)
dt ∂t
the mixture density is ρ m ≡ (1 − α ) ρ + αρ g (11.2.8)
P is associated with the difference in pressure inside and outside a bubble:
pg ≡ p + ρ P (11.2.9)
and the subscript ∞ indicates the undisturbed value.
Adding Eqs. (11.2.2) & (11.2.3) produces a conservation equation for the mixture
∂ρ m
density: + ∇ ⋅ ( ρmu ) = 0 (11.2.10)
∂t
Also, if the interfacial force density and the slip velocity, u g − u , are neglected (both
reasonable approximations for the cases of interest here), a mixture momentum equation in
terms of the mixture pressure, pm ≡ (1 − α ) p + α p , is obtained by adding Eqs. (11.2.4) &
du
(11.2.5): ρm = −∇pm , (11.2.11)
dt
where d dt = ∂ ∂t + u ⋅ ∇ is the total or material derivative. The similarity of the forms
of these equations to their counterparts in single-phase flow is notable, and will be exploited
in a later section in order to estimate the drag on a slender body moving at high-speed
through a bubbly liquid.
Consider the case of irrotational flow, ∇ × u = 0 , and introduce the velocity potential
du ⎛ ∂Φ 1 2⎞
Φ where ∇Φ = u , so that = ∇⎜ + ∇Φ ⎟ (11.2.12)
dt ⎝ ∂t 2 ⎠
Using this result in (11.2.5) leads to the equation
∂Φ 1
( ) dp
p +αρ P
∇Φ − u ∞ ⋅ u ∞ + ∫
2
+ =0 (11.2.13)
∂t 2 p ∞ + α ∞ ρ ∞ P∞ ρm
The fundamental equation for Φ is produced by taking the material derivative of
d ⎛ ∂Φ 1 2⎞ 1 d ( p + αρ P )
(11.2.13): ⎜ + ∇Φ ⎟ + =0 (11.2.14)
dt ⎝ ∂t 2 ⎠ ρm dt
The last term on the left-hand side of this equation may be written in terms of Φ by using
the conservation of mass, Eqs. (11.2.2) and (11.2.3). To do this, introduce the sound speeds
in the liquid and in the gas, denoted respectively as c and cg , which satisfy
dρ 1 dp d ρg 1 dpg
= 2 and = (11.2.15)
dt c dt dt cg2 dt
446
First, eliminate dα / dt from (11.2.2) and (11.2.3) to obtain an equation for dp / dt in terms
1 dp αρ dP
of ∇ 2 Φ and dP / dt + = −∇ 2 Φ (11.2.16)
ρ m cm dt ρ g cg dt
2 2
where Eq. (11.2.9) has been used to relate pg to p , and the mixture sound velocity,
1 1−α α cg2
denoted cm , satisfies = + with c 2
≡ (11.2.17)
ρ m cm2 ρ c 2 ρ g cg2
g
1 + P c2
Eliminate ∇ 2 Φ from (11.2.2) and (11.2.3) to obtain
dα ⎡⎛ 1 1 ⎞ dp ρ dP ⎤
= α (1 − α ) ⎢⎜ − ⎟ − ⎥ (11.2.18)
⎢⎣⎜⎝ ρ c ρ g cg ⎟⎠ dt ρ g cg2 dt ⎥⎦
2 2
dt
Using this result for dα / dt and Eq. (11.2.16) for dp / dt in Eq. (11.2.14) yields an
equation for Φ in terms of dP / dt , the rate of change following the flow of the difference in
pressure inside and outside the bubble (divided by the liquid density):
⎛ ∂ 2Φ ∂u ∂u ⎞
ρ m cm2 [1 + αρ PC ] ∇ 2 Φ − ρ m ⎜ 2 + 2ui i + ui u j i ⎟
⎜ ∂t ∂t ∂x j ⎟⎠
⎝
(11.2.19)
⎧⎪ (1 − α ) ρ P ρ m cm2 ⎫⎪ dP
= αρ ⎨1 − − [1 + αρ PC ]⎬ ,
⎩⎪ ρg cg2 ρ g cg2 ⎭⎪ dt
1 1 1
where indicial notation has been used and where C ≡ + − .
ρ m cm ρ c
2 2
ρg cg2
When α = 0 this equation collapses to the classical single phase equation for the
compressible flow velocity potential (see, for example, [Ashley & Landahl, 1965, Eq. (1-
74)]). This result (11.2.19) shows that irrotational motion in a bubbly flow is modified by the
rate of change of the difference between the pressure interior to the bubbles and the exterior
pressure. Importantly, the single phase (liquid) sound speed in the classical equation is
replaced in (11.2.19) by a term proportional to the sound speed of the mixture, which
decreases rapidly from the liquid value as α increases slightly from zero.
A second equation involving Φ and dP / dt is obtained by using the equation for
∂Ψ
conservation of bubble number density: + ∇ ⋅ ( Ψu ) = 0 (11.2.20)
∂t
1 dα 3 dR
which, with (11.2.1), implies = −∇ 2 Φ + (11.2.21)
α dt R dt
Here it has been assumed that no bubble coalescence or splitting occurs. Using (11.2.18) and
(11.2.16) to eliminate dα / dt and dp / dt , respectively, produces the desired relation
1 1 3 dR 1 − α ρ dP
∇2Φ − = . (11.2.22)
ρ g cg
2
ρ m cm2 R dt ρ c 2 ρ g cg2 dt
This description, Eq. (11.2.19) and (11.2.22), of irrotational flow in a bubbly liquid
is closed by specifying the relationship between the pressure differential P and the velocity
of the bubble surface dR / dt . The Rayleigh equation of bubble dynamics, having an error of
order ( dR / dt ) / c (see [Prosperetti, 1987]), provides the necessary equation:
447
2
d 2 R 3 ⎛ dR ⎞
P = ( pg − p
) ρ =R + ⎜ ⎟ . (11.2.23)
dt 2 2 ⎝ dt ⎠
Here surface tension and the viscosity of the fluids have been ignored.
11.2.2 The Linear Equations
The situation of interest is a body in constant translation with speed U in a fluid, with the
variables α , R, ρ , ρ g , c , cg uniform in the absence of the body. This section presents a
derivation of the linear approximations to the above equations. Note that the results of this
formulation apply to the case of small-amplitude waves: a higher-order formulation would
be required to extend the results to account for the presence of true shock waves.
Let ε be a small parameter: ε 1 . For the case of a body, ε would be the ratio of
maximum body diameter to body length, and the resulting equations would be the slender
body approximation to the above formulation. Expand the field variables in a series of
powers of ε ; for the x − direction defined by the direction of travel of the body, through
second order in ε , these expressions are
Φ = Ux + εΦ1 + ε 2 Φ 2 , R = R0 + ε R1 + ε 2 R2 , and P = Po + ε P1 + ε 2 P2 , (11.2.24)
plus similar expansions for the variables α , R, ρ , ρ g , c , cg .
dR0
The zeroth order terms describe the case of no body, so = 0 and P0 = 0 (11.2.25)
dt
must hold. To first order, from Eq. (11.2.23),
d 2R dP1 d 2⎛d R ⎞
P1 = R0 0 2 1 and = R0 0 2 ⎜ 0 1 ⎟ , (11.2.26)
dt dt dt ⎝ dt ⎠
where d 0 / dt is the material derivative following the free stream:
d0 ∂ ∂ d 02 ∂2 ∂2 2 ∂
2
≡ +U ; = + 2U + U . (11.2.27)
dt ∂t ∂x dt 2 ∂t 2 ∂t ∂x ∂x 2
Substituting the expansions (11.2.24) into (11.2.19) & (11.2.22), and equating terms of order
1 d 2Φ ⎛ 1 1 ⎞ d 0 P1
ε yields ∇ 2 Φ1 − 2 0 2 1 − α 0 ρ 0 ⎜ − 2 ⎟
=0 (11.2.28)
cm0 dt ⎜ ρ m0 cm0 ρ g0 cg0
2 ⎟ dt
⎝ ⎠
ρ g0 cg0 3 dR1 1 − α 0 d 0 P1
2
and ∇ 2 Φ1 − − 2 =0, (11.2.29)

ρ m0 c02 R0 dt c 0 dt
where the over-bar on cg has been dropped, since, to linear order, it does not depend on the
pressure differential, P . Similarly, the over-bar on the mixture sound speed cm will be
dropped in the following formulation, in consideration of Eq. (11.2.16). Each of these sound
speeds will also be subscripted with zero – that is cg0 and cm0 – for consistency with the
order of the formulation.
The quantity d 0 P1 / dt can be eliminated from these two equations to obtain an
equation for d 0 R1 / dt in terms of Φ1 . The result can be substituted into the equation for
d 0 P1 / dt in (11.2.26). Replacing d 0 P1 / dt in (11.2.28) with this latter expression leads to the
equation for the first-order perturbation potential for flight of a slender body in a bubbly
448
liquid: 2
Φ1 +
(1 − α 0 ) R02 ρ 0 ⎡ 2 ⎛ d 0 2 Φ1 ⎞ ρ m0 d 0 4 Φ1 ⎤
∇ ⎜ − = 0, (11.2.30)
2 ⎢ 2 ⎟ 4 ⎥
ρg0 cg0 ⎣ ⎝ dt ⎠ ρ 0 c 0 dt ⎦
m0 2
3
⎡
1 d n2 ⎤
where the D’Alembertian operator is defined as 2
qn { } ≡ ⎢∇ 2 − ⎥{ } .
⎣⎢ cq2n dt 2 ⎦⎥
If this equation is compared to the equivalent governing equation for single phase
compressible flow (see, for example, [Ashley & Landahl, 1965, Eq. (1-74)]):
0 Φ1 = 0
2
(11.2.31)
it may be seen first that, when the void fraction is zero, Eq. (11.2.30) reduces to
R02 ρ 0 d 0 2 2
2
Φ + ⎡ 0 Φ1 ⎤⎦ = 0, (11.2.32)
3 ρ g 0 cg2 0 dt 2 ⎣
0 1
which has as a solution the single phase solution. When the void fraction is nonzero (that is,
the case of a bubbly liquid) there are two qualitative differences to the single phase relation.
First, the single phase sound speed is replaced by the mixture sound speed. Secondly, higher
derivatives of the potential appear, indicating that regions of high variation in velocity (such
as near a leading edge, a cavitator, or cavity closure) will differ the most from the single
phase case. Thus it can be concluded that a main effect of bubbly flow, at least in the case of
steady flight, will be in the modification of the sound speed. Since the sound speed drops
radically even for small values of the void fraction, this analysis indicates that the flight of
underwater projectiles in bubbly flows will be supersonic. This result is to be expected on
the basis of previous formulations, such as those presented in [Brennen, 1995] and [Drew &
Passman, 1999]. Moreover, the formulation presented above lays the groundwork for further
analysis of the physics of high-speed motion in bubbly flows, as is presented in the
following sections.
11.2.3 Admissible Time-Independent Waves
Following a development similar to that presented in [Newman, 1977] for the Kelvin
wake of a surface ship, the dispersion relation for linear wave motion of the bubbly fluid is
obtained by Fourier transformation of (11.2.30) or, alternatively, by considering solutions of
the form Φ1 (r , t ) = exp ⎡⎣i ( K ⋅ r − ωt ) ⎤⎦ for wave number vector K and frequency ω . The
Ω2 2⎛Ω 2 ⎞
2
result is 2
− c 2
m0 − a2 Ω ⎜ 2 − bcm0 ⎟ = 0 (11.2.33)
K ⎝K ⎠
where Ω is the Doppler-shifted frequency and K is the magnitude of the wave number
1
vector: Ω ≡ ω − KU cos θ and K ≡ ( K ⋅ K ) 2 . (11.2.34)
The quantity θ is the angle between this vector and the x − axis (defined by the direction of
the mean flow U ). It is emphasized that the relevant sound speed is the mixture sound
speed, which varies as a function of the void fraction according to Eq. (11.2.17). The
remaining quantities are defined as follows:
ρ c2 1 ρ0 1
b ≡ 0 20 , a1 ≡ (1 − α 0 ) R02 , and a2 ≡ a1 . (11.2.35)
ρ m0 cm0 3 ρg0 cg0
2
b
For the case U = 0 ( Ω = ω ) this relation (11.2.33) reduces to the expression given by [Drew
& Passman, 1999, p281]. For α > 0 , Eq. (11.2.17) shows that b > 1 , a relation that will be
449
useful in understanding the properties of the dispersion relation.
As discussed above, the current interest is in the behavior of steady or time-
independent disturbances. For this case, take ω = 0 and re-arrange Eq. (11.2.33) slightly to
M 2 cos 2 θ − 1
obtain a2U 2 K 2 cos 2 θ = 2m , (11.2.36)
M m cos 2 θ − b
where M m is the Mach number of the mean flow relative to the mixture sound speed,
M m ≡ U / cm0 .
In the remainder of this development, it is convenient to define the following
M2 −b
quantity: γ ≡ m2 . (11.2.37)
M m −1
As seen in the following section, Green’s function for (11.2.30) is more conveniently
developed in terms of the components of the wave number vector K . Thus let k be the
component of K along the axis of motion, x , and, for axisymmetric disturbances about this
axis, let be the component of K perpendicular to this axis. Hence, cosθ = k / K , so that
the wave number components for admissible time-independent waves are related by
k 2 − 2 / ( M m2 − 1)
a2U 2 k 2 = 2 (11.2.38)
γ k − b 2 / ( M m2 − 1)
Note that for single-phase flow, cos 2 θ = 1 / M m2 or k 2 = 2 / ( M m2 − 1) is obtained by setting
a2 = 0 in these relations, achieved in the limit as the bubble radii approach zero according to
definitions (11.2.35). Eq. (11.2.38) has the solutions
2
1
k2 = c
⎡− (1 + bq ) ± s ⎤⎦ ≡ k±2 , (11.2.39)
2 b − Μ 2m ⎣
M m2 − 1
where 2
c ≡ , q ≡ 2 / 2c , s ≡ b 2 q 2 + 2 χ q + 1, and χ ≡ b − 2γ . (11.2.40)
a2U 2
The relation for two-phase flow, Eq. (11.2.38), reveals important features associated
with high-speed disturbances in a bubbly liquid. The following discussion provides a brief
overview of the properties of this relation; further ramifications will be seen in the following
sections, where, respectively, Green’s function for (11.2.30) is derived and wave phase plots
for various cases are presented. The features of relationships (11.2.38) or (11.2.39) in three
regimes of the mean flow speed, U are worth detailed consideration in order to understand
the nature of the resulting wave motion in the bubbly liquid. In order of increasing speed in
terms of M m , the three regimes are: I – 1 > M m2 ; II – b > M m2 > 1 ; and, III – M m2 > b .
As will be shown, two wave modes may be present in a bubbly liquid. It will also be shown
that, for M m2 > 1 , that is in speed regimes II and III, the wave parameters behave similarly to
the familiar acoustic wave of compressible single phase fluids. This wave mode is termed
the modified acoustic mode. In speed regimes I and III, a wave mode occurs which is
dependent on the presence of the bubbles. This additional mode is referred to as the bubble
mode. Note that both modes of propagation occur in speed regime III.
In order to clarify the behavior of k ( ) as specified by Eq. (11.2.39), the left-hand
and right-hand sides of Eq. (11.2.38) have been plotted versus k 2 in Figs.11.2.1a-c.
450
Intersections of the two sides are described by Eq. (11.2.39). In each of these figures the
units are arbitrary, and the arbitrary values = 2 , b = 2 , and a2U 2 = 1 have been used to
produce the figure. (Chosen for the sake of clarity in these figures, these values have no
particular physical significance, and the conclusions are independent of the values selected.)
In each speed regime, the right-hand side of Eq. (11.2.38) intercepts the axis defined by
k 2 = 0 at a value of 1 / b , and asymptotes for large k 2 to a value of
1 / γ = ( M m2 − 1) / ( M m2 − b ) . The intercept is shown, but the plots are not carried out to great
enough values of k 2 to make the asymptote discernible. In each speed regime, the left-hand
side plotted versus k 2 consists simply of a straight line of slope a2U 2 . For the special case
a2 = 0 , representing the single-phase relation, the slope of the left-hand side is zero.
The two sides of Eq. (11.2.38) for the highest speed regime, M m2 > b , are shown in
Fig.11.2.1a. The right-hand side changes sign at k 2 = ka2 and has a singularity at k 2 = ks2 ,
where ka2 ≡ 2
/ ( M m2 − 1) and ks2 ≡ b 2
/ ( M m2 − b ) (11.2.41)
With b > 1 , the following inequality applies: b / ( M m2 − b ) > 1 / ( M m2 − 1) . Therefore, ks2 > ka2 .
The right-hand side is positive for values of k 2 > ks2 and asymptotes for large k 2 to a value
of 1 / γ > 1 / b . As seen in Fig.11.2.1a, for a2 > 0 there are two points where the right- and
left-hand sides intersect, one in the interval 0 < k 2 < ka2 and the second in the range k 2 > ks2 .
The former point of intersection corresponds to the upper (+) sign in Eq. (11.2.39), and the
latter corresponds to the lower (-) sign. For the special case a2 = 0 , the slope of the left-hand
side of (11.2.38) is zero and there is only a single point of intersection of the two sides
plotted in Fig.11.2.1a, namely at k 2 = ka2 , which is the single-phase form of relation
(11.2.38). Thus the mode associated with the upper (+) sign in Eq. (11.2.39) is a
modification of the familiar acoustic wave mode in single-phase compressible fluids. The
mode associated with the lower (-) sign is an additional wave mode supported by the
presence of the bubbles. As is shown in Fig.11.2.1a, this mode is characterized by wave
number vectors having greater magnitude K = k 2 + 2 and making a smaller angle with the
x -axis θ = tan −1 ( / k ) than the modified acoustic mode.
Now consider relation (11.2.38) for intermediate mean flow speeds U such that
1 < M m2 < b . The two sides of (11.2.38) for this case are plotted in Fig.11.2.1b. The right-
hand side changes sign at k 2 = ka2 . However in this speed regime the right-hand side is not
singular; rather it approaches an asymptote for large values of k 2 to the value 1 / γ < 0 .
Consequently the two sides only intersect in the interval 0 < k 2 < ka2 and only the modified
acoustic wave mode is supported. In terms of Eq. (11.2.39), the lower (−) sign yields
negative values of k 2 in this speed regime, corresponding physically to evanescent rather
than propagating waves.
Finally, for low mean flow speeds U such that M m2 < 1 , the two sides of Eq.
(11.2.38) for this case are plotted in Fig.11.2.1c. In this speed regime the right-hand side
451
asymptotes for large k 2 to 1 / γ > 0 . As a result, the two sides can intersect only once, as for
the intermediate speed regime. In terms of Eq. (11.2.39), the upper (+) sign yields negative
values of k 2 in this low speed regime, corresponding physically to evanescent rather than
propagating waves, which is a result familiar for acoustic waves in single-phase
compressible fluids. When bubbles are present, the mixture supports propagating waves even
when M m2 < 1 , with wave parameters corresponding to the lower (-) sign.
4.0
0.9
3.0
0.7
Right and Left

Right and Left
2.0
0.5
1.0
0.3
0.0
0.1
-1.0
-0.1
-2.0
-0.3
-3.0
-0.5
-4.0
0.0 0.5 1.0 1.5 2.0
0.0 0.5 1.0 1.5 2.0
0.6
Fig.11.2.1. Right-hand (solid) and left-
hand (dashed) side of Eq. (11.2.38) for: 0.5
Right and Left

a (top left plot) – M m2 > b (in this case 0.4
M = 9 );
2
m
0.3
b (top right) – 1 < M m2 < b ( M m2 = 1.96 ); 0.2
and c (bottom) – M m2 < 1 ( M m2 = 0.9 ). 0.1
0.0
0.0 0.5 1.0 1.5 2.0
k squared
11.2.4 Green’s Function for the Steady Axisymmetric Linear Equation

The most straightforward case to consider is that of steady, axisymmetric flow. For steady
flow, all partial time derivatives are zero, ∂ / ∂t = 0 , so that the material derivative becomes
d0 ∂
=U . (11.2.42)
dt ∂x
For axisymmetric fields in cylindrical-polar coordinates, ( x, r ) , the gradient and Laplacian
⎡ ∂2 ⎤ 1 ∂ ⎛ ∂ ⎞
may be written as ∇ 2 { }=⎢ + ∇ r2 ⎥ { } ; ∇ r2 { }≡ ⎜ r ⎟{ }; r ≡ y 2 + z 2 , (11.2.43)
⎣ ∂x r ∂r ⎝ ∂r ⎠
2
⎦
respectively. Using these expressions, the linear Eq. (11.2.30) for Φ1 becomes
∂2 2 ∂ Φ1
4
δ (r )
∇ r2 Φ1 +
∂x
⎡
2 ⎣
− ( M 2
m − 1) Φ1 + a1U ∇
2 2
r Φ 1
⎤
⎦ + U 2
( a1 − a2 M m ) ∂x 4
= Aδ ( x )
r
(11.2.44)
for the case of a source of strength A located at the origin.
Introduce the Fourier-Hankel transform
∞ ∞
FH ( k , ) = ∫−∞ dx e−ikx ∫0 Φ1 ( x, r )H 0(1) ( r ) rdr , (11.2.45)
452
1 ∞ ∞
such that Φ1 ( x, r ) =
∫ dk eikx ∫ FH ( k , )H 0(1) ( r ) d . (11.2.46)
2π −∞ 0
Here H n(1) is the Hankel function of the first kind of order n . At this point the Bessel
function J 0 ( r ) might be considered as the basis of the transform, instead of the Hankel
function. However, the Bessel function leads to standing waves in the far field, while, as will
be seen below, the Hankel function leads to propagating waves, which is the focus the
present investigation. The Fourier-Hankel transform, here denoted FH , is appropriate to the
case of a point source, whereas the Fourier-Bessel transform will be appropriate to the
extension of this formulation to slender body theory, where it will be denoted FJ .
∞ ∞
Operating on Eq. (11.2.44) with ∫ dx e − ikx ∫ dr rH 0(1) ( r ) yields the equation for
−∞ 0
F (k , ) : {− 2
}
+ k 2 ⎡⎣ M m2 − 1 + a1 − U 2 2 ⎤⎦ +U 2 ( a1 − a2 M m2 ) k 4 FH ( k , ) = A. (11.2.47)
A
Grouping terms produces FH ( k , ) = . (11.2.48)
a2U ( b − M m )( k 2 − k+2 )( k 2 − k−2 )
2 2
Here k± are given by relation (11.2.39).

The corresponding potential – that is, Green’s function for Eq. (11.2.30) – is found
by substituting Eq. (11.2.48) in Eq. (11.2.46). A slight re-arrangement produces
A ∞ H 0(1) ( r )
Φ ( x, r ) =
4π a2U 2 ( b − M m2 ) ∫0
d ×
k+2 − k−2
(11.2.49)
⎧⎪ 1 ∞ ⎛ 1 1 ⎞ 1 ∞ ⎛ 1 1 ⎞ ⎫
⎪
⎨ ∫−∞ ⎜ − ⎟ e dk − ∫−∞ ⎜ −
ikx ikx
⎟ e dk ⎬ .
k
⎩⎪ + ⎝ k − k + k + k + ⎠ k − ⎝ k − k − k + k − ⎠ ⎭⎪
The integration over k , the x -component of the wave vector, may be performed
∞ eikx ∞ 1
∫−∞ k − k+ dk = e + ⋅ 2i ∫0 ξ sin (ξ x ) dξ = iπ e + .
ik x ik x
using (11.2.50)
For causality, only the outward propagating waves are admissible in Eq. (11.2.49), therefore
the terms involving 1 / ( k + k± ) are omitted. Also, only waves that propagate to the far field
are of interest. In view of the discussion of relation (11.2.38) in the preceding section, define
the potentials associated with an acoustic and a bubble propagation mode as, respectively:
ik x
iA 1 ∞e +
Φ a ( x, r ) = ∫ H 0(1) ( r ) d and (11.2.51)
4 M m − 1 k+ s
2 0
ik x
iA 1 ∞e −
Φ b ( x, r ) = − ∫ H 0(1) ( r ) d . (11.2.52)
4 M m − 1 k− s
2 0
In the high speed regime, where M m2 > b as discussed in the preceding section, the
potential is given by Φ1 ( x, r ) = Φ a ( x, r ) + Φ b ( x, r ). (11.2.53)
In the intermediate speed regime, 1 < M m2 < b , the following form is applicable:
Φ1 ( x, r ) = Φ a ( x, r ). (11.2.54)
Finally, in the low speed regime, M < 1 , the potential is given by
2
m
453
Φ1 ( x, r ) = Φ b ( x, r ) (11.2.55)
A far-field approximation for Φ1 for a point source can be developed using the
method of stationary phase. To begin with, use the property of the Hankel function
2 i ( z −π / 4 )
[Abramowitz & Stegun, 1972, p364] that H 0(1) ( z ) e , z >> 1. (11.2.56)
πz
Also introduce x G± ( ) = k± ( ) x + r , (11.2.57)
where x = x 2 + r 2 is the distance from the point source to the field point. With these
relationships, the following expression results in the far field:
iA 2 e − iπ / 4 ∞
Φ a ( x, r )
4 π r M m2 − 1 ∫0 k+ s
i x G (l )
e + d , r 1. (11.2.58)
A similar expression for Φ b may be written by replacing G+ with G− . Strictly speaking, the
condition r 1 is not met for values of in the neighborhood of the lower limit of
integration in (11.2.58). However, for r large enough, this portion of the integral becomes
vanishingly small. Approximating the integral in (11.2.58) by the method of stationary phase
(see, for example, [Newman, 1977, pages 275ff]) results in the far-field approximation
A 1 1 ⎡ 1 ⎛ iπ ⎞⎤
Φ a ( x, r ) ⎢
2 M m2 − 1 x r ⎣⎢ k+ s G+′′ ( )
exp ⎜ i x G + ( ) +
4
sgn ( G+ ( ) ) ⎟ ⎥ . (11.2.59)
⎝ ⎠ ⎦⎥ =
+
Here the primes on G indicates differentiation with respect to , and + is the value of
dG+ dk r
that satisfies =0⇒ + =− (11.2.60)
d d x
from Eq. (11.2.57); that is, + is a function of r / x . Also from Eq. (11.2.57),
d 2 G+ d 2k
G+′′ ≡ 2
= x 2+ (11.2.61)
d d
dk± 12 ⎡ 1 ⎤
Relation (11.2.39) yields = ⎢ −b ± ( b 2 q + χ ) ⎥ (11.2.62)
d b − M m k± ⎣
2
s ⎦
d 2 k± 1 dk± ⎛ dk± ⎞ 4 ( b − 1) M m q
2
and = ⎜1 − ⎟∓ (11.2.63)
d 2 d ⎝ k± d ⎠ ( M 2 − 1)2 k± s 3
Again, a similar expression is obtained for Φ b by replacing G+ with G− and k+ with k− in
Eq. (11.2.59) through (11.2.61). Using those results with Eq. (11.2.59) in Eq. (11.2.53)
through (11.2.55) provides the far-field approximation to the potential for radiation from a
point source in a bubbly liquid.
11.2.5 Wave Patterns and Field Computations for a Point Source
In this section, the relationships derived above are used to investigate the wave patterns
generated by a point source moving through a bubbly liquid.
For the steady motion of interest here, waves are considered whose phase speed
matches the speed of the body (in this case, the point source) generating the waves. Thus, the
locus of constant phase of waves that are stationary with respect to the body must be found.
While k± ( ) obtained from Eq. (11.2.39) and the corresponding statement of stationary
454
phase, Eq. (11.2.60), may be used, a more convenient approach is to use the polar
coordinates ( K ,θ ) of the wave number vector, because θ ranges only over the interval from
0 to π . For this approach, the form of the relation for K (θ ) given by the solution of Eq.
(11.2.36) and the statement of stationary phase may be employed:
d d
( K ⋅ x ) = ⎡⎣ K (θ ) ( x cosθ + σ sin θ ) ⎤⎦ = 0 (11.2.64)
dθ dθ
Accordingly, the desired wave pattern is given by solutions for the stream-wise and radial
coordinates, ( x, r ) , of the coupled equations
⎛ K' ⎞ ⎛ K' ⎞
x ( K cos θ ) + r ( K sin θ ) = κ and x ⎜ − sin θ + cos θ ⎟ + r ⎜ cosθ + sin θ ⎟ = 0, (11.2.65)
⎝ K ⎠ ⎝ K ⎠
where here the prime indicates differentiation with respect to θ and κ is a constant. To
display the locus of wave phase for successive waves, κ is increased by 2π .
Since both flows involve frequency-dependent wave speeds, it will be enlightening
to compare the wave patterns predicted by the current formulation with the Kelvin pattern
for gravity waves generated by a point source moving just below an otherwise undisturbed
free surface. This classical result is presented in Fig.11.2.2. The salient features of the
classical Kelvin wave pattern are the presence of transverse and diverging waves,
confinement of the significant waves to a well-defined sector that is independent of ship
speed (in deep water), the increasing curvature of the diverging wave crests with distance
from the axis of travel, and a phase offset of π / 2 between the transverse and diverging
waves. 20.0
Fig.11.2.2.
The classical Kelvin wave pattern.
10.0
radial distance
This classical result for dispersive waves due to

a point source moving just under the ocean surface may 0.0
be compared with results of the current model for high-

speed motion of a point source through a bubbly liquid, -10.0
as presented in Figs.11.2.3-5, and discussed in detail

with the aid of Fig.11.2.6. These figures display the -20.0
-40.0 -30.0 -20.0 -10.0 0.0
wave patterns generated under various conditions of axial distance
speed and void fraction. For these plots, relevant to the
case of a projectile moving at a speed of 1000m/s (Figs.11.2.3 and 11.2.4) or 1600m/s
(Fig.11.2.5), the following values are used:
c = 1500 m/s, cg = 350 m/s, ρ = 1000 kg/m3 , ρ g = ρ / 850, and R = 0.0001m. (11.2.66)
It is again emphasized that the relevant sound speed is the mixture sound speed, which varies
as a function of the void fraction according to Eq. (11.2.17). The curves plotted represent
successive crests of the linear waves. It should be noted that, when carefully applied,
stationary phase theory predicts the existence of phase shifts, which have not been plotted in
Figs.11.2.3-6, but which will be seen below in the full field computations presented in
Fig.11.2.7-8. The wave patterns display two distinct sets of curves. The transverse waves are
associated with the bubble propagation mode, whereas the diverging waves are associated
with the modified acoustic propagation mode.
455
Figs.11.2.3a through 11.2.3d show the wave patterns for a speed of 1000m/s and
void fractions of α = 10−5 , 10−4 , 10−3 , and 10−2 , respectively. Recall that the flow is
assumed to be axisymmetric, so these loci are actually meridional cuts through axisymmetric
surfaces. The Mach number with respect to the single-phase liquid sound speed is
M = 0.67 , so that, in the absence of the bubbles, no propagating waves will be generated.
For the case of nearly pure water (Fig.11.2.3a, showing results for a void fraction of
α = 10−5 ), the projectile is subsonic with respect to both the liquid and the mixture, with
M m = 0.72 . This is in the low speed regime M m2 < 1 identified above, and, as seen there,
these propagating waves occur only because of the presence of the bubbles. As will be
discussed in greater detail below, both diverging and transverse waves are present,
representing the existence of two branches in the solution. It is noteworthy that, whereas the
curvature of the loci of constant wave phase for the Kelvin wake increases downstream,
those for the projectile in a bubbly flow show higher curvature closer to the disturbance. As
the void fraction increases, it can be seen that the shapes of the loci change dramatically. In
Fig.11.2.3b, the void fraction is α = 10−4 , the flow is already supersonic with respect to the
mixture ( M m = 1.17 ), and no transverse wave system is present. The difference can be
understood using the discussion of relation (11.2.38) presented above: now the conditions
are in the intermediate speed regime 1 < M m2 < b . Hence the wave mode is the modified
acoustic mode, instead 0.4 0.4
of the bubble mode. As 0.3 0.3
the void fraction 0.2 0.2
continues to increase,
radial distance
radial distance
0.1 0.1
the diverging wave 0.0 0.0
pattern narrows and the -0.1 -0.1
curvature decreases. -0.2 -0.2
-0.3 -0.3
Fig.11.2.3. -0.4 -0.4

a (top left) a=0.00001, -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0
cm=1383m/s, axial distance axial distance

0.4 0.40
Mm=0.723.
b (top right) a=0.0001,
0.3 0.30
cm=904m/s, Mm=1.17. 0.2 0.20

radial distance
radial distance
c (bottom left) a=0.001, 0.1 0.10
cm=348m/s, Mm=2.873 0.0 0.00
d (bottom left) a=0.01, -0.1 -0.10
cm=113m/s, Mm=8.85. -0.2 -0.20
-0.3 -0.30
-0.4 -0.40
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 -0.80 -0.70 -0.60 -0.50 -0.40 -0.30 -0.20 -0.10 0.00
axial distance axial distance
In each of the
examples shown in Figs.11.2.4a through 11.2.4d, the projectile speed is again 1000m/s,
representing a Mach number with respect to the liquid of approximately M = 0.67 , and the
flow field is again assumed to be populated with air bubbles of radius R = 0.0001m . Here
the behavior at projectile speeds very close to the mixture sound speed is explored. These
456
plots show the nature of the transition from the bubble wave mode to the modified acoustic
wave mode as the mixture Mach number increases from slightly below unity to slightly
above unity due to increases in the void fraction. The true first image of the sequence shown
in Fig.11.2.4 is that of Fig.11.2.3a. The shape of the loci are difficult to deduce at that scale,
but are qualitatively similar to those shown in Fig.11.2.4a, where the void fraction has been
increased to α = 6x10−5 . Under these conditions the mixture sound speed decreases to
cm0 = 1047m/s , and the mixture Mach number is M m = 0.955 . Note that the flows for
Figs.11.2.3a and 11.2.4a show both transverse and diverging waves, so that the topology of
the shock wave system bears some resemblance to the Kelvin wake on a free surface.
However, the curvature of the diverging wave pattern again increases near the disturbance,
as opposed to the Kelvin wake. Finally, it can be observed that the loci begin to overlap
somewhere between the cases shown in Figs.11.2.3a & 11.2.4a. In Fig.11.2.4b, the mixture
Mach number is just less than unity, and the overlap has increased considerably. The
bounding sector angle has also increased. These plots display the nature of the bubble wave
mode. In Fig.11.2.4c, the projectile is just supersonic with respect to the mixture, so that the
wave form has switched to the modified acoustic mode. The transverse wave system no
longer exists, and the topology of the flow is homeomorphic to those of Figs.11.2.3b through
11.2.3d. This behavior becomes more dramatic at the lower void fraction and higher mixture
Mach number of 0.4 0.4
Fig.11.2.4d. 0.3 0.3
0.2 0.2
radial distance
radial distance
Fig.11.2.4. 0.1 0.1
a (top left) 0.0 0.0
a=0.000060, -0.1 -0.1
cm=1047m/s, -0.2 -0.2
Mm=0.955. -0.3 -0.3

b (top right) -0.4 -0.4
a=0.000071, -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0
cm=1001m/s, axial distance axial distance
Mm=0.999. 0.4 0.4
c (bottom left) 0.3 0.3
a=0.000072, 0.2 0.2
cm=997m/s,
radial distance
radial distance
0.1 0.1
Mm=1.003 0.0 0.0

d (bottom left) -0.1 -0.1
a=0.000080, -0.2 -0.2
cm=967m/s, -0.3 -0.3
Mm=1.034. -0.4 -0.4
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0
In each of the examples shown in Figs.11.2.5a through 11.2.5d, the projectile speed
is 1600m/s, representing a Mach number with respect to the liquid of approximately
M = 1.07 , and the flow field is again assumed to be populated with air bubbles of radius
R = 0.0001m . For these values, the mixture Mach number satisfies M m2 > b so that the wave
457
pattern is in the high speed regime identified above. Thus both wave modes are present.
Numerous features associated with relation (11.2.38) are visible, particularly the greater
magnitude of the wave number vector (smaller wavelength) for the bubble mode. In
addition, remembering that the vector lies perpendicular to the wave phase lines shown in
the figure, its smaller angle with the x − axis for the bubble mode is also observed.
Importantly, as can be seen in the sequence of increasing void fractions shown in
Figs. 11.2.5a through 11.2.5d, the modified acoustic waves tend to narrow toward the axis as
the Mach number increases, while the bubble waves tend to stack up toward a bounding
wave system, becoming sharper in curvature toward the limiting apex of the system.
0.4 0.4
0.3 0.3
0.2 0.2
radial distance
radial distance
0.1 0.1
0.0 0.0
-0.1 -0.1
-0.2 -0.2
-0.3 -0.3
-0.4 -0.4
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0
Fig.11.2.5a Fig.11.2.5b.
a=0.00001, cm=1383m/s, Mm=1.2. a=0.0001, cm=904m/s, Mm=1.8.
0.4 0.4
0.3 0.3
0.2 0.2
radial distance
radial distance
0.1 0.1
0.0 0.0
-0.1 -0.1
-0.2 -0.2
-0.3 -0.3
-0.4 -0.4
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0
Fig.11.2.5c. Fig.11.2.5d.
a=0.001, cm=348m/s, Mm=4.6 a=0.01, cm=113m/s, Mm=14.1.
The basic features of the wave patterns presented in Figs.11.2.3-11.2.5 are called out
in Figs.11.2.6a & b, and compared with the results of a supersonic supercavitating projectile
experiment [Kirschner, 1997; Hrubes, 2001] as presented in Fig.11.2.6c. The change in the
topology of the predicted wave field is apparent as the flow conditions change from a
mixture Mach number less than unity and a speed less than the liquid sonic velocity to a
mixture Mach number (much) greater than unity and a speed greater than the liquid sonic
velocity. Specifically, whereas in the former case the transverse waves are distributed well
downstream of the point source (located at the origin in the figures), in the latter case, they
are focused just downstream of the disturbance. Moreover, the diverging waves asymptote to
458
a wider angle with the line of flight in the former case, as may be expected for a mixture
Mach number slightly less than unity. Furthermore, if the wave pattern of Fig.11.2.6b is
compared with the results of supersonic supercavitation experiments, Fig.11.2.6c, it is easy
to mistake the bow-shaped bubble mode near the forward part of the system as the linear
precursor to the bow-shaped bow shock presented in [Kirschner, 1997] and [Hrubes, 2001].
This is not the case: the signature of the acoustic mode in either an essentially pure liquid or
a bubbly liquid is the diverging wave pattern, and in the supersonic supercavitation
experiments, the curved bow shock is due to the finite cavitator radius used in the body
design, the bluntness of which causes an offset of the shock wave from the nose and an
associated finite curvature. In contrast, the wave patterns presented in Fig.11.2.6a and b are
due to a point disturbance, and thus the acoustic mode forms an apex at the axis of travel.
The bowed shape of the bubble mode is a direct result of the solution as captured via Eq.
(11.2.65) and the method of stationary phase, and is unrelated to the bluntness of any body,
since the disturbance is a point source.
distributed focused
Mmm = 0.955 transverse waves transverse waves
0.4
Mm = 1.8
0.4
0.3
0.3
0.2
0.2
radial distance
0.1
radial distance
0.1
0.0
0.0
-0.1
-0.1
-0.2
-0.2
-0.3
-0.3
-0.4
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 -0.4
α= 1e-4
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0
α= 6.0e-5
axial distance
axial distance
cmix = 1047m/s cmix = 904m/s
a b
c
Fig.11.2.6. Examples of predicted wave patterns for a source moving through a
bubbly liquid compared with the shock wave generated in an essentially pure liquid:
a –mixture Mach number less than unity and speed less than the liquid sonic
velocity;
b – mixture Mach number greater than unity and speed greater than the liquid sonic
velocity; and,
c – high-speed image of a supercavitating body traveling at supersonic speed in
water [Kirschner, 1997; Hrubes, 2001]. Note that the bow shock visible in Fig.11.2.6b is not
due to the same physical effect as the similar feature apparent in Fig.11.2.6b. (See text.)
Confidence in the method of stationary phase as applied to this problem is bolstered

by direct numerical computation using the Green function solution for a point source, Eqs..
(11.2.53)-(11.2.55), of the various field quantities of interest. Various quantities, including
the potential and velocity fields have been explored. Computational results of the potential
and pressure fields are presented below. The challenging integrals in Eq. (11.2.44) have been
evaluated numerically.
459
Fig.11.2.7 shows contours of Eq. (11.2.101) for a void fraction of α = 0.0001 for
M m = 1 / 2 , M m = 2 , and, M m = 2 , and for the single
a
phase case [Ashley and Landahl, 1965] for M = 2 . Since
the acoustic wave speed for this void fraction as given by
Eq. (11.2.17) is about 904 m/s, the three bubbly flow cases
imply speeds of 639, 1278, and 1808 m/s, respectively. The
features of the wave motion discussed with respect to
Figs.11.2.3-5 can be seen in the plots shown in Fig.11.2.7.
For example, note in Fig.11.2.7c the presence of both the
modified acoustic wave mode and the bubble wave mode. In
Fig.11.2.7b only the modified acoustic mode is present, and
in Fig.11.2.7a, the case of M m < 1 , only the bubble wave b
mode is present.
Fig.11.2.7.
Contours in the xr − plane of Green’s function for a bubbly
liquid potential flow, of void fraction α = 0.0001 at three
values of the Mach number a – M m = 1 / 2 ; b – M m = 2 ;
and c – M m = 2 ; and d – supersonic single-phase flow with
M = 2.
The rough correspondence between the Mach angles
of Figs.11.2.7b&c is also noteworthy (and expected, since c
the Mach numbers are identical); however, the more
complicated structure of the wave field – including the
curved Mach cone – is apparent in the case of the bubbly
flow. In these plots, the point source producing Green’s
function is located at the origin. Interestingly, the latter two
plots can indicate an influence of the source ahead
(upstream) of the source, in marked contrast to the familiar
single phase case, where effects are felt only downstream of
a source. This difference reflects the different spectral
content produced by motion in a bubbly flow, compared to
motion in the pure liquid.
The field pressure coefficient for several cases,
including the case of supersonic motion in pure water, is
plotted in Fig.11.2.8. Both of these figures display the same
basic topology as the wave patterns as predicted via the
application of stationary phase, as exemplified in Fig.11.2.3-
6. Note, however, that a phase shift is predicted in one of the
flow cases that was not accounted for in the presentation of
the stationary phase results, as was noted above.
11.2.6 Slender Body Theory in a Bubbly Liquid
In the background presented above, the problem of motion in a bubbly liquid was linearized
on some small, unspecified parameter, ε . (See Eq. (11.2.24).) For the case of a body, ε
460
would be the ratio of maximum body thickness to body length, and the resulting equations
would be the slender body approximation to the above formulation.
Adapting the development presented in [Ashley and Landahl, 1965], the expansions
employed above are used to match an inner solution, Φ ( x , r ) = Φ i ( x, r ) , (11.2.67)
α = 0.0001 to an outer solution, Φ ( x, r ) = Φ o ( x, r ) , (11.2.68)
U = 630m/s Δ r
Mm = 0.71 as r → 0 , where r= (11.2.69)
M = 0.42
ε
is the inner radial coordinate.
Fig.11.2.8.
bubble Pressure coefficient from direct numerical integration for
mode
Green’s function: a – subsonic with respect to mixture and
liquid; b – supersonic with respect to mixture, subsonic with
phase shift
a respect to liquid; c – supersonic with respect to mixture and
α = 0.0001 liquid; and, d – supersonic, single-phase flow.
U = 1270m/s
Mm = 1.41
M = 0.85 In inner coordinates, substituting the perturbation
expansion into the steady equations, solving for the velocity
acoustic
mode
potential, and equating terms of second order yields the
following governing equation for the inner problem:
1
Φ i2 rr + Φ i2 r = 0, (11.2.70)
σ
b which represents the same simple source flow that results for
α = 0.0001 the single phase case.
U = 1800m/s Applying the kinematic condition that the flow is
Mm = 2.00
M = 1.20 tangent to the body:
acoustic Φ ir d d
mode = a ( x ) = ε a ( x ) = ε a ', (11.2.71)
bubble
Φ x dx
i
dx
mode yields the inner solution:
U
Φ i2 ( x, r ) = Uaa 'ln ( r ) = A 'ln ( r ) + g ( x ) , (11.2.72)
2π
where a ( x ) is the body radius and A ( x ) = π a 2 ( x ) is the
c
α = 0.0
U = 2100m/s body sectional area. The function g ( x ) is determined by
Mm = 1.41
M = 1.41 matching with the outer flow, in a procedure similar to that
presented in Sec. 8.1.3, but with the effects of
acoustic compressibility included. (See also Chapter 5, especially
mode
Sec. 5.7, as well as [Ashley & Landahl, 1965].)
Eq. (11.2.72) shows that the lowest order
contribution to the velocity potential in the inner region is a
d flow due to a simple source distribution of strength
Ua da dx . This is also the situation found in single-phase
flow. The corresponding radial velocity component is
Uaa′ Uaa′
uσ = ε uσ = ε =ε2 . (11.2.73)
σ σ
461
In [Grant & Kirschner, 2006], it is noted that the lowest order perturbation to the flow
velocity by a slender body moving through a bubbly liquid is of order ε2 , just as in the case
of single-phase compressible flow. These results show that the flow very near a slender body
moving in a bubbly liquid is unaffected to lowest order by the presence of the bubbles.
However, the bubbles modify the outer flow, whose lowest order contribution is also
presented in [Grant & Kirschner, 2006], where it is shown that the perturbation of a slender
body to a bubbly flow is zero through first order in the slenderness parameter. The expansion
in outer coordinates recovers the equation for linear waves, here rearranged:
(1 − M m2 ) Φ o2 xx + 1r Φ o2 r + Φ o2 rr +
(11.2.74)
1 ρ U 2 ∂ 2 ⎡⎛ M m2 ⎞ o 1 o ⎤
+ (1 − α 0 ) R02 0 2 ⎢⎜ 1 − Φ
⎟ 2 xx + Φ + Φ o
2 rr ⎥ = 0.
ρ g 0 cg 0 ∂x 2 ⎣⎝
2 r
3 b ⎠ r ⎦
The inner boundary condition is set by matching the solution as r → 0 to the inner
solution, Eq. (11.2.73). The radial component of velocity thus behaves as
Uada / dx
v =ε2 , r → 0, (11.2.75)
r
Uada / dx
so that Φ 02 r = , r → 0. (11.2.76)
r
The outer boundary condition is that the velocity should vanish at infinity. With the solution
Φ o2 , Eq. (11.2.74), the function g ( x) in Eq. (11.2.72) is set by matching the inner and outer
da
velocity potential for r → 0 : g ( x ) = Φ 02 ( x, r ) − a ln ( r ) , r → 0. (11.2.77)
dx
Consideration of the perturbation velocity potential, φ ( x, r ) ≡ ε 2 Φ 2 ( x, r ) , (11.2.78)
enables the inner boundary condition to be expressed in terms of the physical radius a ( x)
rather than the expanded radius a ( x) . From Eq. (11.2.74), matching the outer solution at
small radius to the inner solution yields the inner boundary condition
1 dA
φ ( x, r ) = ln ( r ) , r →0 (11.2.79)
2π dx
where A( x) ≡ π a( x) 2 is the cross-sectional area of the slender body at an axial location x.
The boundary value problem for the outer flow comprised of the governing Eq.
(11.2.74) and the boundary conditions is solved for the perturbation potential, φ ( x, r ) , using
a strategy similar to that presented above at Eq. (11.2.44):
1 L dA(ξ )
φ ( x, r ) ≡ ε 2 Φ o2 = G ( x − ξ , r ) dξ
2π ∫0 d ξ
(11.2.80)
where now Green’s function, G, is the solution to Eq. (11.2.74) with the right-hand side
replaced with δ ( x ) δ ( r ) / r .
Before this solution is developed, however, some physical discussion of Eq.
(11.2.74) is of interest. As noted in [Grant & Kirschner, 2004], a low-frequency
462
approximation is valid when the last term on the left-hand side of the governing Eq.
2 2
ρ ⎛ U ⎞ ⎛ R0 ⎞
2
1 ⎛U / h ⎞
(11.2.74) is negligible: β ( h ) ≡ (1 − α 0 ) 0 ⎜ ⎟ ⎜ ⎟ = (1 − α 0 ) ⎜ ⎟ 1 , (11.2.81)
3 ρg0 ⎜⎝ cg0 ⎟⎠ ⎝ h ⎠ ⎝ ωres ⎠
where h is a length scale that characterizes gradients in the axial direction and ωres is the
cg0 3ρ g0
bubble resonance frequency, given by ωres = . (11.2.82)
R0 ρ 0
Here the bubble resonance frequency is defined as the so-called peak bubble response
frequency when surface tension and viscous effects are ignored, assuming isothermal
polytropic behavior of the gas in the bubble. (For an extensive discussion of bubble
dynamics, including various important physical effects, see [Brennen, 1995, p37ff and
p99ff]. The resonance frequency presented above can be derived from [Brennen, 1995, eq.
4.5] by setting the surface tension and fluid kinematic viscosity to zero, and the polytropic
constant to unity, representing isothermal behavior. See also [Drew & Passman, 1999,
p280].) For air in fresh water in typical laboratory conditions, Eq. (11.2.81) can be
rearranged to provide the following approximate condition for the validity of the low-
U ρ g0 cg0 h 2
h
frequency approximation: Mm = << 3 ≅ 0.012 . (11.2.83)
cm ρ 0 cm R0
2
R0
Under the conditions for which the low-frequency approximation is valid, the
1
governing Eq. (11.2.74) reduces to (1 − M m2 ) Φ 2 xx + Φ 2 r + Φ 2 rr ≅ 0 (11.2.84)
r
which has the same form as the equation governing single-phase compressible flow, and it
can be seen that the solution will be dominated by the acoustic propagation mode, with little
contribution from the bubble mode. Thus, when either of Eqs. (11.2.81) or (11.2.83) holds,
the behavior can be predicted using the same equation as for the single-phase flow, except
that the Mach number must be replaced with the
mixture Mach number, M m . 1E+1
h
Fig.11.2.9. 0.012 t
1E+0 R0
Approximate limit of mixture Mach number
above which the acoustic propagation mode 1E-1
dominates the solution for air bubbles of radius Mm p
h
−4 1E-2 b 0.0012
R = 1 × 10 m (marked “b”) in fresh water. The R0
other marked points indicate order-of-magnitude 1E-3 (practical limit for bubble
length scales for: “p,”the cavitator size of a small mode contribution)
1E-4
supercavitating projectile cavitator; and, “t,” the
1E-4 1E-3 1E-2 1E-1 1E+0
cavitator size of a larger supercavitating underwater
vehicle. h
It should be noted that these two equations will not, in general, be valid in all regions
of the flow, since the characteristic length scale, h , is not necessarily uniform throughout the
domain. So, for example, near the cavitator of a supercavitating body, where conditions
change in a distance on the order of a cavitator radius, the limiting Mach number will be
much less than in most of the domain, where the conditions change more slowly. Eq.
(11.2.83) is plotted in Fig.11.2.9 for bubbles of radius R = 1 × 10−4 m . If a mixture Mach
463
number equal to 10% of the value given by this equation is considered small, the bold line in
Fig.6 represents an approximate practical limit for contribution of the bubble mode to the
solution. The figure also indicates length scales considered typical of two high-speed bodies
of interest.
Returning to the full form of Eq. (11.2.74), the mixture Mach number, M m , and the
quantity b defined as the first of Eqs. (11.2.35) provide two dimensionless parameters that
determine the nature of the solution to Eq. (11.2.74). To develop the third and final such
parameter, non-dimensionalize the coordinates ( x, r ) with the characteristic length scale L,
the length of the slender body: x → x L , and r → r L. (11.2.85)
In the remainder of this section, ( x, r ) refer to the non-dimensional coordinates. With this
non-dimensionalization, the governing Eq. (11.2.74) becomes
⎡⎛ ⎞ ⎤
(1 − M m2 ) Φ o2 xx + 1r Φ o2 r + Φ o2 rr + β ∂∂x 2 ⎢⎜1 − Mbm ⎟ Φ o2 xx + 1r Φ o2 r + Φ o2 rr ⎥ = 0, (11.2.86)
2 2
⎣⎝ ⎠ ⎦
where β ≡ β ( L ) .
With the inner solution providing boundary condition (11.2.76) for the outer flow, a
Green function strategy may be employed to obtain a formal solution to Eq. (11.2.74)
satisfying this boundary condition. With the Green function G ( x, r ) and boundary condition
(11.2.79), the formal solution to governing Eq. (11.2.74), which is the second order
1 Ξ dA(ξ )
perturbation velocity potential, is φ ( x, r ) = G ( x − ξ , r ) dξ ,
2π ∫0 d ξ
(11.2.87)
⎧1, Mm <1
⎪
with Ξ=⎨
⎪⎩ min x − r M (m
2
− 1,1 , M m > 1
,
)
(11.2.88)
in view of non-dimensionalization (11.2.85).

For the case of a slender body in bubbly flow, the equation for the outer potential
flow of order ε 2 has the same form as the homogeneous version of the linear steady
equation given in Eq. (11.2.44). Consequently, Green’s function for this case is
G ( x, σ ) = Re [ Φ a + Φ b ] , (11.2.89)
where Φ a and Φ b are specified in Eqs. (11.2.51) & (11.2.52), with the constant A set to
unity.
In determining Green’s function for a point source, the Fourier-Hankel transform
was the appropriate choice for solving the problem, since the solution must be singular at the
source location. For the slender body, the appropriate Green function solution for the outer
problem will be matched to a finite condition from the inner problem as r → 0 . Hence, a
Fourier-Bessel transform is the appropriate choice for this problem:
∞ ∞
FJ ( k , ) = ∫ dx e−ikx ∫ r dr J 0 ( r ) G ( x, r ) (11.2.90)
−∞ 0
∞ ∞
1
where G ( x, r ) = ∫ dk e ∫
ikx
d J 0 ( r ) FJ ( k , ) (11.2.91)
2π −∞ 0
464
1
∞
F ( x, )
The resulting Green function is G ( x, r ) = 2 ∫
d J 0 ( r ) J2 ,(11.2.92)
4π a2U ( b − M m ) 0
2
k+ − k−2
Δ ∞
where the form of the partial transform, FJ ( x, ) = ∫ dx e−ikx FJ ( k , ) (11.2.93)
−∞
depends on the solutions, k± , of dispersion relation (11.2.38), and thus on the speed regime,
which is further delineated than the three regimes previously identified for the point source.
For a slender body, the behavior can be described via Eqs. (11.2.106-109), which apply,
respectively, in each of the flow regimes indicated in Fig.11.2.10:
⎧1 1 ⎫
FJ ( x, ) = 2π i ⎨ e − − k+ x − eik− x ⎬ ,
2
(11.2.94)
⎩ k+ k− ⎭
⎧1 ⎫⎪
) = −4π ⎪⎨
1
FJ ( x, sin ( k+ x ) + e− − k−2 x
⎬, (11.2.95)
k 2 k-2
⎩⎪ + ⎭⎪
−4π − ki x
FJ ( x, )=
k + ki2
2
e {
kr sin ( kr x ) − ki cos ( kr x ) , } (11.2.96)
r
⎧1 1 ⎫
and FJ ( x, ) = −4π ⎨ sin ( k+ x ) − sin ( k− x ) ⎬ . (11.2.97)
⎩ k+ k− ⎭
(109) b
b =1 M2 = b=2 M2 = b
2−b Fig.11.2.10.
The flow regimes for a slender body moving in a
(50) bubbly liquid. The regions labeled 106, 107, 108, and
(50)
(109) 109, correspond to the applicability of Eq. (11.2.94),
(11.2.95), (11.2.96), and (11.2.97), respectively. Note
the more complicated structure compared with the
(48)
(107) case of a point source as represented in Fig.11.2.1.
(49)
(108)
In each of these equations, the first term represents
M2 = 1
(106)
(47) the contribution of the modified acoustic mode,
whereas the second term is due to the bubble mode.
For Eq. (11.2.96), the quantities are
determined as follows. With
k±2 = A ± B 2 = A ± i B 2 , (11.2.98)
or k± = kr ± iki , (11.2.99)
the real and imaginary parts can be evaluated as kr = A + A2 + B 2 2, (11.2.100)
and ki = ± B 2 (
2 A + A2 + B 2 , ) (11.2.101)
M m2 − 1 + ba2U 2 2
where A= , (11.2.102)
2a2U 2 ( M m2 − b )
465
(M 2
m − 1 + ba2U 2 ) − 4a U ( M
2 2
2
2 2
m − b) 2
and ± B =∓2
(11.2.103)
2a2U 2 ( M − b)
2
m
For comparison, Green’s function for the single-phase case is the well-known
expression [Ashley & Landahl, 1965, Eq. (11.2.5)-(11.2.26)]:
−1
G ( x, r ) = , (11.2.104)
2π x 2 − ( M 2 − 1) r 2
where M =U / c is the Mach number based on the sound speed in the liquid. The single-
phase results discussed below can be computed using this function in the governing
equation, which is represented by Eq. (11.2.74) with the last term on the left-hand side set
equal to zero.
Substituting the result for the case of the bubbly liquid, Eq. (11.2.89), with the form
of Green’s function as developed above, into the formal solution (11.2.87) yields the
following expression for the lowest order perturbation to the velocity potential for the outer
flow around a slender body in a bubbly liquid:
1 L dA(ξ ) ∞
φ ( x, r ) = 2 ∫ d ξ FJ ( x − ξ , ) J 0 ( r ) ,
d ξ ∫0
d (11.2.105)
8π 0 s( )
with FJ ( x, ) given by (11.2.94)-(11.2.97), a quantity that depends on the mixture Mach
number M m2 and the ratio b defined in Eqs. (11.2.35).
11.2.6. Drag on a Slender Body in a Bubbly Liquid

The velocity potential is governed by the following expression, which is the form of Eq.
∂Φ 1
( )
pm dp
+ ∇Φ − u ∞ ⋅ u ∞ + ∫
2
(11.2.13) applicable to the mixture: = 0. (11.2.106)
∂t 2 pm∞ ρ
m
The subscript ∞ denotes values at distances sufficiently far from the body so as to be
unperturbed. The second-order pressure perturbation, needed to compute the pressure drag on
the slender body, can be found from this governing equation. In the integral in that equation,
ρ m is replaced along with the integration limits with their expansions through second order
pm dp pm 0 + ε 2 pm 2 dp p 1 ε 2 pm 2 dp
to obtain ∫ =∫ ≅ ε 2 m2 − ε 2 2 ∫ . (11.2.107)
pm 0 ρ pm0 ρ m0 + ε ρ m2
2
ρ m0 ρ m0 0 ρ m2
m
Since the last term on the right is of higher order than second, to this order it is neglected. To
provide an expression for the pressure near the body in (11.2.106), the gradient is written in
terms of the inner radial coordinate. Thus, substituting (11.2.107) in Eq. (11.2.106), results
⎛ 1 ⎞
in pm2 = − ρ m0 ⎜ U Φ 2 x + Φ 22 r ⎟ . (11.2.108)
⎝ 2 ⎠
p − pm∞ ε 2 pm2
The pressure coefficient C p is defined as CP = 1 m = (11.2.109)
2 ρ m∞U 2 ρ m0U
2 1 2
to second order. With Eqs. (11.2.78) and (11.2.106), the pressure coefficient becomes
1
CP = 2 ( 2U φx + φr2 ) . (11.2.110)
U
466
Comparing 11.2.121) or (11.2.110) with their single phase analogs e.g. [Adams &
Sears, 1953] Eq. (11.2.11), [Ashley & Landahl, 1965] Eq. (11.2.6)-(11.2.23), it may be seen
that the leading (second order) relation between pressure and velocity for a slender body is
the same in single- and two-phase flow. Since Eq. (11.2.74) governs the velocity potential
(and therefore the velocity) itself, this result, with those obtained above for the inner flow,
show that, to leading order, the difference between single- and two-phase flow about a
slender body has its origin in those terms in that governing equation which appear for non-
zero void fraction, namely: (1) the mixture sound speed, cm , defining the applicable Mach
number, rather than the liquid sound speed, c , which would be applicable for single-phase
flow; and, (2) the entire last term on the left-hand side.
Since the relation (11.2.110) between the velocity potential perturbation φ and the
pressure coefficient is the same as for the single-phase case (with the single-phase pressure
and density replaced by their mixture counterparts), the pressure integral over a closed
bounding surface around the slender body is the same as for the single-phase case.
Consequently, the results of [Ashley & Landahl, 1965, p113-115] may be used to write an
expression for the pressure drag D and drag coefficient Cd in terms of φ :
2
D − DB ∂φ ∂φ ⎛ ∂φ ⎞
Cd − CdB ≡ = − ∫∫ dS 2 + 1
∫∫ ⎜ ⎟ dS3 . (11.2.90)
ρ m 0U 2
S2
∂x ∂r 2
S3 − S B ⎝
∂r ⎠
Here DB is the base drag acting on area S B , S 2 is the area of the right circular cylinder
having the same axis as the slender body, extending from some plane upstream of the body
to the base of the body, and S3 is the area of the disk forming the base of this cylinder
external to S B located at the truncated downstream end of the body, as shown in Fig.11.2.11.
When M m < 1 , a third integral, with the same integrand as that in the second integral on the
right in Eq. (11.2.111) is included to account for the elliptic nature of the flow.
S3 S base
S1
S2
Fig.11.2.11. A typical body used for the example forebody wave drag predictions presented
below, indicating the surfaces bounding the control volume. Surfaces S 2 and S3 comprised
the surfaces of integration for Eq. (11.2.111). Following the classical approach for
compressible flows of blunt-based bodies, base drag is not reported in the results for wave
drag presented below; base drag exceeds the scope of the current investigation of interest.
11.2.7. Drag Computations for an Example Slender Body in a Bubbly Liquid

The behavior of the forebody wave drag may be further understood by examining the results
of numerical computations presented in this section. These were generated via numerical
integration of Eq. (11.2.105) and (11.2.111) over ranges of interest of the key parameters or
467
the problem: the void fraction, the mixture Mach number, the speed, and the dimensionless
undisturbed bubble radius. The body selected for computation is presented schematically in
Fig.11.2.11, and is similar to those applied in studying the drag of bodies in classical single-
phase compressible flow. (See, for example, [Ashley & Landahl, 1965].) Specifically, the
apex at the nose allows for direct application of the theory presented above, without the need
of a bluff or blunt nose correction. Moreover, in addition to representing an afterbody shape
of practical interest, truncation of the body on a cylindrical portion downstream of the faired
nose avoids the complications associated with more streamlined afterbody shapes, provided
the base drag is treated as a separate component of the body drag (which, again, is common
in the classical literature). Thus, the forebody wave drag computed in the examples below
excludes the base drag and, since it is based on ideal flow theory, does not account for the
effects of viscosity or turbulence: CD fb wave ≡ CD − ( CD base + CD fb viscous ) , (11.2.112)
where the subscript fb indicates quantities pertaining to the forebody. The base drag and the
drag associated with viscosity (including turbulence effects) must be determined
experimentally or with a much more detailed numerical formulation such as grid-based
computational fluid dynamics (CFD). Note, however, that the effects of variable sound speed
due to disperse bubbles in a liquid has rarely if ever been addressed to date in grid-based
CFD computations, and the current potential flow approach is deemed much more
computationally efficient for mapping the broad parameter space associated with the key
parameters that define the problem.
Determining the drag on a slender body using the formulation presented above
requires the evaluation of three nested integrals, as seen in Eqs. (11.2.105) & (11.2.111). The
azimuthal integral in Eq. (11.2.111) is trivial for this axisymmetric case. The inner integral
over to compute Green’s function is of a class that is fraught with well-known difficulty.
Although several approximations are available, this integral was evaluated directly using two
different algorithms independently: the simple trapezoidal rule and a method based on
Richardson extrapolation via successive division of the integration interval. The upper limit
and the discretization size (for the trapezoidal rule) or the number of divisions (for the
Richardson extrapolation) were adjusted until the numerical result for the integral varied
insignificantly with further refinement. The rapid variation with axial distance of this inner
integral requires a relatively small axial step, of order Δx ∼ 10−4 L , to resolve its contribution
to the outer integral in Eq. (11.2.111). This approach is computationally intensive (and thus
is likely subject to more innovative numerical integration methods), and typical runs
required several hours on a contemporary single-processor workstation to compute the drag
for a single combination of void fraction, Mach number, and bubble radius. For the results
presented below, the radius of S1 (and the outer radius of S1 and S2) is set to be twice the
radius of the slender body.
The radius of the body chosen for the example computations presented below was
⎧ ⎡ 2 ⎤
⎪ L ⎢ 1 − sin 2 θ ⎛1 − x ⎞ − cos θ ⎥ , 0 ≤ x ≤ x
⎪ ⎜ ⎟
defined by a( x) = ⎨ 2sin θ ⎢ ⎥ 0
⎝ x0 ⎠ , (11.2.113)
⎪ ⎣ ⎦
⎩⎪ a0 , x0 ≤ x ≤ L
1− ε 2
where θ is determined by cos θ ≡ (11.2.114)
1+ ε 2
468
and the slenderness ratio is ε ≡ 2a0 / L . The expression in (11.2.113) has the useful
⎧ a = a0
properties that x = x0 ⇒ ⎨ (11.2.115)
⎩ da dx = 0
and that θ is related to the half-angle of the profile entrance, ϑ0 , by the relationship
tan ϑ0 = ( L tan θ ) ( 2 x0 ) , such that, for small profile entrance angles, ϑ0 = ( Lθ ) ( 2 x0 ) .
The computational results presented below are for a body of length L = 1m and a
slenderness ratio of ε = 0.1 , such that a0 = 0.05m . The selected body profile was cylindrical
starting at x0 = 0.5m . The values for the density and sound speeds of the gas and liquid that
were used in the computations were nearly the same as those specified in Eq. (11.2.66) in
connection with Figs.11.2.3-5, and are again typical for the case of air bubbles in water:
ρ = 1000 kg/m3 ; ρ g = ρ / 850; c = 1500 m/s; and cg = 340 m/s. (11.2.116)
Fig.11.2.12 presents the results of numerical computations showing the variation of
forebody wave drag coefficient as a function of the key parameters mentioned above. The
computational intensity of the numerical approach used to produce these plots limited the
number of parameter combinations that could be evaluated; therefore the resulting data are
somewhat sparse. Some variations are clearly smooth, but others, especially variations with
speed or Mach number, display significant “waviness,” and therefore the sparse
computational points only serve as a rough guide to the behavior – that is, values of the drag
between the computational points is not known with reliable certainty without a denser set of
computational points. In such cases, the computational points have been indicated with
markers, and the gross trends of the curves are depicted with dashed lines.
Fig.11.2.12a is a plot of the forebody wave drag coefficient versus varying void
fraction, α , for a mixture Mach number of M m = 2 and a dimensionless undisturbed bubble
radius of R0 L = 10−4 . There is little variation in the forebody wave drag over the range
plotted, suggesting that the void fraction has little effect on the wave drag, except insofar as
its important role in setting the mixture sound speed. Thus this relative constancy of the drag
with changing void fraction is understood by the qualitative discussion above: in this
parameter regime, the governing equation essentially has the form as that for the single-
phase case. The behavior of forebody wave drag with increasing mixture Mach number,
M m , at a fixed value of the void fraction, in this case α = 10−3 , is presented in
Figure 11.2.12b. Results for two values of the dimensionless undisturbed bubble radius,
R0 L = 10−4 and 10−2 , are shown. At this void fraction, the mixture sound speed is 358m/s;
accordingly, for M m in the vicinity of unity, these values of bubble radius yield values of
the quantity β defined in Eq. (11.2.81) (and, for this slender-body problem, based on the
length scale h = L ) of approximately β ( L ) = 3 × 10−6 and 3 × 10−2 , respectively. With this
latter value, the low-frequency approximation, Eq. (11.2.84), is not applicable, and the
equation governing the two-phase fluid varies considerably from that of the single-phase
case. Interestingly, this latter case displays significantly reduced drag for M m > 1 compared
with the low-frequency case, although it is still non-negligible.
To place the behavior of the forebody wave drag in a more physical context,
consider the same behavior in Fig.11.2.12b, but plotted against body speed, U , rather than
469
0.014 0.014
R0 = 0.0001
0.012 R0 =0.01
0.012
forebody wave drage coefficient

forebody wave drag coefficient
0.010 0.010
0.008 0.008
0.006 0.006
0.004 0.004
a
0.002 0.002 b
0.000 0.000
1.E-06 1.E-05 1.E-04 1.E-03 1.E-02 1.E-01 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2
void fraction mixture Mach number
0.014 12 forebody drag coefficient 0.014
R0 = 0.0001
Mach number
R0 = 0.01 0.012
0.012 10
forebody wave drage coefficient
forebody wave drag coefficient

0.010 0.010
8
Mach number
0.008 0.008
6
0.006 0.006
4
0.004 0.004
c 2 d
0.002 0.002
0.000 0 0.000
200 300 400 500 600 700 800 1.E-06 1.E-05 1.E-04 1.E-03 1.E-02 1.E-01
speed void fraction
Fig.11.2.12. The behavior of the forebody wave drag coefficient plotted versus
several key parameters: a – versus void fraction, α , for a fixed mixture Mach number of
M m = 2 and a dimensionless undisturbed bubble radius of R0 L = 0.0001 ; b – versus
mixture Mach number, M m , for a void fraction α = 10−3 and two values of the
dimensionless undisturbed bubble radius, R0 L = 0.0001 and 0.01 ; c – versus speed, U , at
the same conditions as in b; and, d – versus void fraction, α , for a fixed speed of
U = 1000m/s and a dimensionless undisturbed bubble radius of R0 L = 0.0001 . In a, the
corresponding results for single phase flow are indicated by the arrow at the left edge of the
plot. In d, the variable mixture Mach number has also been plotted. Where datasets are
plotted with dashed lines, the sparse computational points are indicated with markers;
values between markers may differ from the indicated curves. (See text.)
against the mixture Mach number. Whereas the scientist is concerned with the physics as
defined by the dimensionless parameter, the engineer is interested in the behavior with the
dimensional quantity. The void fraction is again α = 10−3 and results have been presented
for the same two values of the dimensionless undisturbed bubble radius as in Fig.11.2.12.b.
It can be seen that the drag on the body increases dramatically at speeds above the mixture
470
sound speed, but still far below the liquid sound speed. Thus, if the goal is to project an
object through water, a very significant increase in drag is to be expected at comparatively
moderate speeds – speeds well below the liquid sound speed – if the water contains disperse
bubbles, even if those bubbles are miniscule and the total void fraction is small.
Finally, consider the behavior of the forebody wave drag as the speed is held fixed at
a speed of U = 1000m/s , but the void fraction – and therefore the mixture Mach number –
varies. These results are presented in Fig.11.2.12d for the case of a dimensionless
undisturbed bubble radius of R0 L = 10−4 . The value of the mixture Mach number is also
plotted as a function of the void fraction. It can be seen that the drag peaks when the value of
the mixture Mach number is close to unity (as expected) and then falls off rather slowly as
the void fraction and the mixture Mach number both increase.
A more complete mapping of the forebody wave drag coefficient is presented in
Fig.11.2.13, which presents that quantity as a function of the dimensionless undisturbed
bubble radius, R0 L , and the mixture Mach number, M m , for the same body as discussed in
the context of Fig.11.2.12 at a fixed void fraction of α = 10−3 . This more detailed mapping
highlights an interesting behavior indeed: whereas the drag increases dramatically (as
expected) at a mixture Mach number near unity over much of the range of the computation,
the drag is also clearly dependent on the dimensionless undisturbed bubble radius.
Specifically, the forebody wave drag coefficient displays a clear ridge of elevated values in
the neighborhood of R0 L = 10−2.5 . (As was the case in Fig.11.2.12, the computational effort
in generating these results necessitated a rather sparse parameter grid, as indicated in
Fig.11.2.13, so the detailed variation between grid points is not known with certainty.) Based
on more detailed analysis of the characteristic length and time scales inherent in these
results, [Grant & Kirschner, 2006] hypothesize
that this region of maximum drag results from
excitation of the bubbles near their resonance
points, given by Eq. (11.2.82).
Fig.11.2.13.
Forebody wave drag coefficient as a
function of the mixture Mach number and
dimensionless undisturbed bubble radius for a
fixed void fraction of α = 10−3 . The sparse
computation points were at the nodes of the
indicated grid; intermediate values may differ
from the interpolations shown.
Specifically, a full accounting of the drag due to both the modified acoustic mode
and the bubble mode requires the use of the complete expression (11.2.105) for φ . However,
the similarity of mixture Eqs. (11.2.10) & (11.2.11) to their single-phase counterparts may
be exploited to gain some insight into the qualitative effect of the presence of bubbles on the
drag on a slender body. To this end, again consider the low-frequency form, Eq. (11.2.84), of
the full governing Eq. (11.2.74). As has been noted, that is the expression obtained for
single-phase flow, except that the relevant Mach number is that of the mixture, rather than
the liquid. In view of this result and relation (11.2.110), which is, also as noted above, the
same as for the single-phase case, the entire calculation for the drag of a slender body in
471
single-phase compressible flow [e.g., Ashley & Landahl, 1965, pp113-118] carries over to
the flow in a bubbly liquid.
Comparing Eq. (11.2.84) with Eq. (11.2.74), the difference between the pressure
external and internal to the bubbles may be neglected for cases where β 1 , or from Eqs.
(11.2.81) & (11.2.82), when the characteristic frequency, U L (the time it takes the body to
pass a given bubble), is much less than the bubble resonance frequency ωres . Thus, when a
low-frequency approximation can be applied to governing Eq. (11.2.74), the flow past the
body behaves much as in the single-phase case, with the Mach number determined by the
mixture (rather than the liquid) sound speed. On the other hand, detailed analysis of the
potential and pressure fields surrounding the body over the range of parameters presented in
Fig.11.2.13 indicated that the pressure field near the apex of the forebody varied
considerably over an axial region of length equal to approximately 10% of the body length
for the body selected for computation. Further analysis indicated that the characteristic
frequency associated with the quantity U 0.1L matched the bubble resonance frequency
quite closely under the conditions represented by the high drag region apparent in
Fig.11.2.13, corresponding closely to the observed value of the dimensionless undisturbed
bubble radius in that region, R0 L = 10−2.5 .
11.2.8 Summary of the Model of High-Speed Motion in a Bubbly Liquid

This section has presented a nonlinear formulation for the potential flow fields associated
with the motion of both a point source and a slender body through a medium populated with
bubbles of identical size. The importance of the dramatic reduction in sound speed in a
bubbly liquid has been noted, and an analogy with Kelvin’s problem for the wave wake of a
surface ship has been drawn based on the fact that wave speeds are frequency dependent in
both cases. A model has been developed, the experimental verification of which remains a
topic of future research.
Specifically, the governing equations have been re-derived in the context of motion
through a bubbly liquid, and Green’s functions of lowest order for the linearized equations
governing both the case of a point source and of a slender body have been developed. Two
modes of wave propagation are predicted: a bubble mode complements the usual acoustic
mode associated with single-phase flow. The associated dispersion relation has been
presented, the form of time-independent admissible waves has been determined, and the
method of stationary phase has been applied to predict the wave pattern due to the point
source for various combinations of the key parameters of the flow. The linear wave patterns
predicted by the method of stationary phase were verified via direct numerical integration,
and extended to identify the phase shift that is anticipated from the mathematical theory in
certain flow regimes. The character of the wave pattern is predicted to change depending on
the regime of the mixture Mach number, with both transverse and diverging waves
appearing in the highest and lowest speed regime, but only diverging waves appearing in the
middle regime. According to the model, the transverse waves appear downstream of the
diverging waves in the low-speed regime, and upstream in the high-speed regime.
Because of the bubble mode, wave propagation is present over all speeds, and this
theory predicts that some wave drag contribution will be present at all Mach numbers for
non-zero void fractions and bubble radii, in contrast with single-phase flows. Due to the
more complex spectral content of the resulting wave field, the pressure disturbance is present
472
somewhat upstream of a point source even for Mach numbers somewhat in excess of unity,
in contrast with the case of single-phase flow.
The mathematical and numerical results support the proposed concept of injecting a
bubble field into water as a means of investigating supersonic motion at speeds much lower
than is required for pure water. That is, these results confirm that the effective Mach number
is that based on the mixture sound speed. These results also provide guidance as to the type
of additional effects associated with the bubbles, namely, bubble oscillations, that are
anticipated, although the results also delineate regions in the relevant parameter space where
these effects should be small.
Specifically, a rule of thumb has been given delineating the value of the mixture
Mach number above which the bubble propagation mode can be ignored in predicting flow
in a bubbly liquid, thus allowing existing results for compressible single-phase flow to be
applied, provided that the Mach number with respect to the liquid is replaced with the
mixture Mach number in the governing equation.
Green’s function for the case of the linearized theory for a slender body moving
through a bubbly liquid has been integrated numerically to estimate the forebody wave drag
in an extension of the standard approach to bodies operating in compressible fluids. In
conclusion, it can be stated that wave drag due to mixture compressibility is important to the
high-speed hydrodynamics of bubbly flows, and must be accounted for in the design of
experiments. Due to the dramatic decrease in sound speed as the void fraction increases in a
bubbly liquid, engineering systems operating in such an environment will experience this
wave drag at a much lower speed than they would in an essentially pure liquid.
Future development is required to more fully understand the physics of high-speed
motion in a bubbly flow, starting with the obvious requirement for experimental verification
of the basic model. Certain physical effects that have so far been neglected – notably slip
between the liquid and gas phases, as well as thermal effects – may prove to be non-
negligible once they have been incorporated in the model and explored parametrically.
Certainly, not all of the interesting currently formulated parameter space has been explored,
nor have computational results been generated with enough density in that space to capture
the detailed behavior between the current sparse points representing the parameter
combinations tested to date. Other remaining work includes implementation of the slender-
body theory for motion through a bubbly liquid in a computationally efficient form suitable
for rapid numerical predictions of the forebody wave drag of real bodies of interest operating
in bubbly flows.
To the reader of this book, perhaps the most important unfinished business is the
formulation and implementation of a dynamic boundary condition suitable for predicting
behavior when bodies moving at high speed through a bubbly liquid are cavitating, which is
likely to be the predominant hydrodynamical condition considering the myriad nucleation
sites represented by the bubbles populating the ambient mixture.
Sec. 11.3. Micro-Scale Effects

As a final introduction to the effects of real fluids on cavitation, the following interesting
sequence of photographs is presented (Fig.11.3.1), showing high-speed entry of a tapered
projectile through the free surface of a water tank [Truscott et al., 2009]. The macro-physics
of water entry has been studied for many decades [e.g., May, 1975], thus, the features of
greatest interest in Fig.11.3.1 are those at small scale in both space and time.
473
A very clean water entry cavity is formed around the projectile and extends well
downstream. Cavity closure or “pinch-off” is followed by a bubble trail, the evolution of
which is itself interesting. Of particular interest is the behavior of the flow above the
nominally flat free surface. In addition to the macroscopic jet that rises from the water entry
point, one can discern a wave that propagates in the direction of projectile travel along the
nominally still water level of the tank. In the later sequences, small vertical jets appear along
and about the ridge of this wave. Another among the many interesting features in this
sequence is the appearance of a strong, albeit hair-like vortical structure that appears to
oscillate in strength, located below
the “knee” of the macroscopic jet
profile, extending from the
nominally flat free surface to the
point where pinch-off first occurs.
This feature – while apparently
oscillatory in strength – persists in
the same location until the last
frame in the sequence, which
represents a time at which the
projectile and its cavity have long
past the water entry point, and the
bubble trail has begun to dissipate.
Fig.11.3.1.
Micro-scale effects in high-speed
water entry. The figure shows a
time sequence of images of a
tapered projectile, fired from a 22-
caliber gun launcher, entering the
water at a relatively shallow angle
of 11◦. Images are spaced in time
by 0.0018s. Several of the many
interesting features of this flow are
discussed in the text. (All images
courtesy of T. Truscott.)
474
Ch. 12. Cavitating Flow of Marine Propellers
Ch. 12. Cavitating Flow of Marine Propellers*

Sec.12.1. Historical Notes and Testing Facilities
Cavitation of a screw propeller model two inches in diameter was first observed by Sir
Charles Parsons in 1894 during testing a propeller model of the 45-ton Turbinia powered by
a steam turbine. The model testing was carried out in a small water tunnel invented by him
for this particular purpose [Burill, 1951]. In 1910 Sir Charles Parsons built the first
cavitation tunnel of almost today’s dimensions, where propeller models up to 12” (304.8
mm) in diameter could be tested. The first cavitation tunnel in Russia was built in 1933. The
first systematic tests of a series of cavitating propellers with segment-type sections were
conducted by H.W. Lerbs in 1936 in a cavitation tunnel built specially for this purpose.
The operational principle of cavitation tunnel is well known. This experimental
installation enables one to conduct force measurements and visual observations (using the
stroboscope method) of a model of a cavitating or supercavitating screw propeller with its
advance coefficient and axial cavitation number equal to those of the full-size propeller.
Therewith, it is assumed, somewhat approximately, that such modeling ensures a similarity
of partial cavitation and supercavitation. The cavitation development level, or the local
cavitation number at blade’s cylindrical sections, depends on the hydrostatic pressure
vertical distribution, viscosity, turbulence, air content and other factors. It should be noted
that some cavitation forms can hardly be modeled in a cavitation tunnel, or require special
procedures for scaling the model results to the full-scale. Examples include cavitation of
blade tips and axial vortices as well as cavitation erosion, noise and vibration.
The most important factors (criteria) during testing the supercavitating propellers in
a cavitation tunnel include advance coefficient
J = V / nD (12.1.1)
and axial cavitation number χ = ( p0 − pv ) /(0.5ρV 2 ) , (12.1.2)
where p 0 is the hydrostatic pressure at the propeller axis level, p v is the saturated vapor
pressure at the testing temperature. Sometimes this value is taken equal to the pressure of
vapor-fluid mixture in the cavity, or the atmospheric pressure for the case of super-
ventilation of the cavity, or the gas pressure in the cavity for the case of forced ventilation of
the cavity. Here V is the flow velocity in the cavitation tunnel working section; n is the
number of revolutions of propeller per second, D is the propeller model diameter.
The local cavitation number σ for a blade’s cylindrical section, characterized by
radius r, differs form axial cavitation number (12.1.2) in terms of velocity. Instead of axial
velocity V, the local cavitation number uses relative velocity VE. Also pressure at the
propeller axis is replaced with pressure at the point of consideration porθ are taken at the
point’s σ = ( p0rθ − pv ) /(0.5ρVE2 ) , (12.1.3)
where V = V + (2πnr ) .
E
2 2 2
When a screw propeller is rotating around a horizontal axis in the gravity field, the
local cavitation number for a given point of a section varies during one revolution from a
minimum value at the upper position of the point at the given radius to the maximum value
*
Sections 12.1 – 12.3 were written by Prof. Achkinadze, St. Petersburg Marine Technical University.
475
corresponding to the point’s lowest position. For adequate modeling of the vertical
distribution of hydrostatic pressure, or more accurately, to ensure equality of the local
cavitation numbers of the propeller and its model in all blade sections in the gravity field, the
Froude numbers for the propeller and its model ( Fr = π n D / gD ≈1.0061 n D ) are to be
equal. There is an almost obvious formula for an absolute error due to ignoring the equality
of Froude numbers when determining the local cavitation number calculated at a relative
radius r = r / R of a blade in its upper position. The formula is:
Δσ = {1 /(n 2 D) − 1 /[(n* ) 2 D* ]} r g / [ J 2 + (πr ) 2 ] , (12.1.4)
where n or n∗ and D or D* are the RPM (or RPS) and diameter of the model or full-size
propeller, respectively.
For a complete modeling of Froude number and advance coefficient, the flow
velocity in a cavitation tunnel should be M times less than that of the full-size propeller,
whereas the model RPM should be M times greater than that of the full-size propeller,
where M is the model scale, i.e. the full-scale/model propeller diameter ratio. For example,
let a 1.054 m diameter propeller at 60 knots (30.87 m/s) rotates at 1465 rpm (24.42 rps). To
provide the same Froude number and advance coefficient for a 0.2 m diameter
supercavitating propeller, the model should move at 13.45 m/s in the working section and
rotate at 3363 rpm (56.05 rps). During model testing the flow velocity and model’s RPM are
often lesser that those required for the model. This can be acceptable only when the resulting
error, according to (12.1.4), does not exceed their allowable bounds.
Cavitation tunnel walls distort proper modeling because there are no such walls in
reality. The walls effect can be unacceptably high, especially in a supercavitating regime
when the flow blockage in the tunnel working section due to development of large cavities
can noticeably influence the experimental results. In addition, the free surface effect (i.e.
ventilation and wave making phenomena) is, in principle, impossible to study in
conventional cavitation tunnels. When studying supercavitating propulsors with small
advance coefficients and small cavitation numbers, the use of a cavitation tunnel becomes
impossible due to a "blockage" of the working cross section by the cavities.
Cavitation basins have been built to conduct experimental studies in such cases
[Prischemikhin, 1969; Van Manen, 1971]. The basins have sufficiently wide cross section
and a free surface, providing almost complete disappearance of the walls effect and
facilitating studying the free surface effect. There are four cavitation towing tanks (a.k.a.
Variable Pressure Model Basin, or Vacuum Towing Tank, or Depressurized Towing Tank)
in the world (USA, 1962, 55x4.6x4.6 m; Russia, 1965, 60x6x3.5 m, [Prischemikhin, 1969];
Netherlands, 1971, 240x18x8 m; China,1985, 150x7x4.5 m). The Netherland’s facility has
the biggest depressurized towing tank – more than 8 hours are needed to generate vacuum up
to 4% of the atmospheric pressure [Van Manen, 1971]. Deaeration of water is a common
problem associated with experimental studies in the cavitation towing tanks. Deaeration
leads to a decrease in the number of cavitation nuclei and a delay in early cavitation stages.
Model speed in cavitation basins is restricted by approximately 4 m/s, allowing to
test screw propeller models up to 0.2 m in diameter in the operational regimes corresponding
to full-size propeller speeds (based on equal Froude numbers) not exceeding approximately
20 knots (assuming that the full-size propeller does not exceed 1.2 m in diameter). This
diameter is far from being sufficient for experimental studies of supercavitating propellers at
476
the maximum speed regime. Therefore testing such propellers in a cavitation basin can only
be conducted disregarding Froude numbers. Unlike cavitation tunnels, tests could be
conducted at any values of advance coefficient including the blockage regime. Currently,
cavitation basins are often used to study cavitation on conventional screw propeller models
operating in a non-uniform wake behind a ship hull model. Obviously, model wake does not
fully correspond to that of a full-size ship.
To study the effects of free surface and ship hull, cavitation tunnels with free water
surface had been built in the 1960’s as an alternative to cavitation basins. The tunnels are
convenient for testing surface piercing propellers, e.g. the Free-Surface Cavitation Tunnel
K27, Berlin Technical University [Rose & Kruppa, 1991]. But it is difficult to maintain an
undisturbed free surface in the absence of a model. For this reason a new generation of
cavitation tunnels with very big working sections has been built. The world-largest and most
capable cavitation tunnel is in Memphis, TN, USA, allowing reaching 18 m/s, thus covering
the speed range of full-size propellers up to speeds of 70 kn. The cross section of its working
part is 3x3 m, allowing testing propeller models together with ship models scaled from 1/10
to 1/20, and decreasing the wall effect to an acceptable level [Etter, 2000]. A limited number
of full-scale experiments and almost en-mass propeller model testing in the numerous
conventional cavitation tunnels (there are approximately 123 tunnels in the world today
[Etter, 2000; Gorshkov et al., 2007]) constitute the bulk of the contemporary experimental
database on the supercavitating and highly cavitating screw propellers.
Sec. 12.2. Supercavitating and Highly Cavitating Screw Propellers

Discussed in this section are experimental studies, associated with the development of a
small number of hydrofoil craft fitted with supercavitating propellers. Cavitation of
propulsors, leading to erosion, vibration, hydro-acoustic noise and undesired changes of
hydrodynamic characteristics, represent a physical phenomenon hindering the efforts of
naval architects to increase speed of both conventional displacement ships and high-speed
craft with dynamic support. Sometimes, the whole set of technical problems accompanying
the occurrence and development of cavitation is called cavitation barrier.
In some cases it is possible to avoid adverse consequences of cavitation occurrence
by designing a completely non-cavitating propulsor. The proper choice of the expanded area
ratio of a screw propeller makes it possible to avoid (completely or partially) for a period
time the development of cavitation forms resulting in erosion of the propeller blades.
However, as noted in the documents of the German Navy Command dated 1932, they did
not manage to avoid strong erosive damage to the propellers of destroyers and torpedo boats
by that method [Springorum, 1932]. After 200 hours of full-speed cruising those propellers
did acquire such damages (erosion blisters size of a fist), so that their replacement became
inevitable. According to the present views [Jessup & Wang, 1997], one can ensure an
acceptable service life of screw propellers at cruising speeds up to 36 knots or more by
means of an optimal selection of expanded area ratio, RPM, and diameter, in combination
with the use of blade skew, improved profiling, as well as by smoothing the non-uniformity
of the hull wake. These measures can prevent cavitation with virtually no erosion damage.
For example, a prototype passenger hydrofoil Taifun (65t, 2x1750 hp, 100 passengers,
[Ikonnikov & Maskalik, 1987]) with automatically controlled deeply submerged foils, built
in Russia in 1971, reached a speed up to 44 knots using two non-cavitating screw propellers
mounted on struts (pushing type arrangement). However, it is often impossible to design a
477
non-cavitating propulsor. This is for example the case for a vessel designed with either or
several of the following features: high speed, screw propeller mounted on an inclined shaft,
blade tips are too close to the free water surface, too large expanded area ratio (> 1.2).
An alternative to designing a non-cavitating propulsor is naturally the idea of
designing a cavitating propulsor which is exposed to erosion and other negative
consequences of cavitation to a small extent only. This idea was first set forth in 1941 for
designing screw propellers of destroyers and similar ships by Pozdunine [1941, 1944, 1944a,
1947]. He proposed
to select the Motion of propeller
expanded area ratio, Rotation of propeller
distribution of
pitch, section
curvature and cross
sections
(Fig.12.2.1) in such
a way, that the
propeller blades of
a high-speed ship
should operate in
the regime of Leading edges
supercavitation.
More specifically,
in such a regime Fig.12.2.1. Pozdunine's
when the cavities supercavitating propeller with
formed on the D=0.2 m, P=0.26 m, AE/A0=0.4.
blades have a length
exceeding the local
chord, and therefore close at a distance behind the blade (Fig.12.2.2). Therewith, the erosion,
which usually occurs when the cavities close on the blade surface, does not occur in the
supercavitating regime. If additionally, the transient ship regime time when cavities do not
exceed the local chords is minimized, the adverse cavitation effects would not exceed
reasonable limits.
Fig.12.2.2. Propeller
operating on
supercavitating regime
[Davis & English, 1968].
Based on intensive
experimental studies in a
cavitation tunnel,
Pozdunine revealed the main difficulties hindering the realization of his idea. First, a special
wedge-type profile (Fig.12.2.3) is most suitable for the supercavitating propellers. Later on,
in Russia the wedge-type profile propellers have been called supercavitating. To distinguish
the supercavitating propellers (SCP) from a segment-type profiling a term of highly
cavitating propellers (HCP) was introduced.
478
Fig.12.2.3.
Blade section types: (a) conventional;
(b) segment; (c) wedge-shaped
(supercavitating);(d) wedge-shaped with a
spoiler (supercavitating with a spoiler).
The wedge-type profiling and the flow pattern past the SCP immediately created
several problems. First, providing the local strength of thin wedge-type leading edges of the
full-size propeller blades. Second, maintaining the necessary length and thickness in all
sections of the cavity at the full cruising regime with a certain reserve for cavitation
occurrence on the pressure side of the blade. Third, generating the sufficient thrust of the
propulsor at intermediate regimes, e.g. at the drag hump regime of a hydrofoil. The latter
problem is rather important for designing a hydrofoil ship; it will be discussed below in more
detail, as two different ways of its solution have emerged.
Experimental studies revealed that the SCP concept is feasible only for cruising
speeds higher than 50 knots, or at local cavitation numbers less than 0.045 at 0.7R. Obtaining
sufficiently long cavities, if possible at all, would require to abruptly increase angles of
attack of the blade. This would result in an inacceptable increase of cavitation drag and a
sharp drop of efficiency. Therefore, the Pozdunine’s idea was not realized for destroyers and
torpedo boats. The problem of erosion damage of torpedo boat propellers due to cavitation at
the blade root sections was successfully resolved by drilling anti-erosions holes
[Georgievskaya et al., 1984].
As recollected in [Tulin 2000, 2001; Tachmindji & Morgan, 1958; Caster, 1963],
studies of the SCP started in the United States in 1957 with a design of a wedge-type
profiled SCP model using an approximate lifting line theory [Tulin & Burkart, 1955]. For
example, thrust coefficient of 0.140 and efficiency of 0.685 were obtained for a model of a
supercavitating 3-blade propeller with expanded area ratio of 0.5 and pitch ratio of 1.57 at
advance coefficient of 1.125 and axial cavitation number of 0.3. In 1962 a team of the
Hydronautics led by Tulin successfully developed a 2-blade SCP for a 80-ton gas-turbine
powered hydrofoil Denison (60 knots, 10000 hp) [Tulin 2000, 2001]. Significant difficulties
arouse in generating a sufficient thrust at the drag hump regime. This was associated with
the flow blockage by cavities that is typical for the SCP at advance coefficients significantly
less than those corresponding to the maximum speed for the vessel in question – see below
for more detail. In the mid-1960’s there was built the 235-ton Canadian oceanic HMCS
hydrofoil Bras d'Or (60 knots, 22000 hp) with two SCP [Eames & Jones, 1971]. When
developing these propellers special attention was paid to their leading edge local strength.
In 1958, Yu.M.Sadovnikov, KSRI, tested the K series of the SCP with wedge-type
profiling and symmetric contour [Mavlyudov et al., 1982; Voitkunsky, 1985; Ikonnikov &
Maskalik, 1987], comprising 15 two-blade models and 9 three-blade models of 0.2 m
diameter. The propeller models had a relative radius of the hub 0.165, expanded area ratio
ranging within 0.34-1.11, pitch ratio varying within 1.0-1.8, and the minimal cavitation
number reached 0.3. However, such a hydrofoil ship has never been built, because the
problem of providing sufficient thrust at the drag hump regime was not solved with this
propeller series. An effective solution of this problem in the Russian version (when advance
coefficient at the drag hump regime differs only slightly from that in full cruising speed due
479
to drop in rpm) was solved by replacing the wedge-type profiling with the segment-type one
(see Fig.12.2.3). Further on, as noted, the term highly cavitating propellers (HCP) was
adopted in the KSRI (and in Russia) for supercavitating propellers having the segment-type
profiling. Thus, the distinction between the highly cavitating propellers and the
supercavitating propellers consists in the profiling, Fig.12.2.3. In particular, the wedge-type
thickness distribution along chord was replaced with a segment distribution with shifting the
max thickness by 15% from the mid-chord towards the trailing edge (CK2 series). The CK
series had thickness distribution without shifting.
Fig.12.2.4. Performance curves for СК series with segment profiling (z=3; AE/A0=0.80;
P/D=1.4; cavitation numbers from 0.2 to 1.2, after [Mavlyudov et al., 1982].
The CK series contains 28 models of 3-blade highly cavitating screw propellers with
expanded area ratio ranging within 0.65-1.10 and pitch ratio ranging within 1.0-2.2, which
had been developed and tested by VD Tsapin. Fig.12.2.4 gives the performance curves for
one of this series model. Hydrodynamic characteristics of this model are compared in Table
12.2.1 with those of the K series supercavitating propellers.
It can be seen from the table that in the supercavitating regime, when the hydrofoil
ship has a maximal speed and the cavitation number is equal to 0.3, the efficiency of the K
series propeller with wedge-type profiling is 12 - 16% higher than that of the CK series
propellers with segment-type profiling. This can be explained by a smaller thickness of the
leading edges for the wedge-type series, resulting in a decrease of cavity thickness and
cavitation drag of this series propellers. The general strength is approximately the same due
to a larger thickness of the trailing edges for the wedge-type series, although the local
strength near the leading edges is noticeably lower.
On the other hand the CK series with segment profiling has 18% higher efficiency as
compared to the "К" series at the drag hump regime, when the hydrofoil speed is about 50-
480
65% of the maximum speed, which corresponds to axial cavitation number equal to 1.0 or
slightly higher. Therewith, the advance coefficient only slightly differs from its value 1.15
corresponding to the maximum efficiency of both screw propellers at the maximum speed
regime. This case corresponds to a Russian variant of hydrofoil design, when the power
plant (with a low reserve of power) allows a reduction of propeller RPM at the drag hump
(almost proportional to the drop of translational speed of ship) and, thus, the advance
coefficient at the drag hump regime is only slightly less than that at the maximum speed.
Table 12.2.1. Comparison of characteristics of two screw propellers with an expanded area
ratio 0.81/0.8 and pitch ratio 1.4, belonging to the series K of SCP with wedge-type profiling
and a series CK of highly cavitating propellers (HCP) with segment profiling
(J - advance coefficient, KT - thrust coefficient, η0 - open water propeller efficiency).
J Axial cavitation # KT, "К" KT, "CК" η0, "К" η0, "CК" Δη0/η0, %
0.8 0.3 - - - - -
0.9 0.3 - - - - -
1.0 0.3 0.100 0.095 0.600 0.533 12.6
1.1 0.3 0.115 0.103 0.655 0.565 15.9
1.2 0.3 0.098 0.099 0.630 0.560 12.5
0.8 1.0 0.225 0.215 0.562 0.515 9.1
0.9 1.0 0.230 0.248 0.620 0.588 5.4
1.0 1.0 0.200 0.250 0.640 0.660 -3.1
1.1 1.0 0.150 0.208 0.590 0.700 -18.6
1.2 1.0 0.100 0.161 0.520 0.709 -36.3
0.8 Atmosphere 0.315 0.348 0.560 0.562 -0.4
0.9 Atmosphere 0.250 0.300 0.618 0.619 -0.2
1.0 Atmosphere 0.200 0.253 0.625 0.667 -6.7
1.1 Atmosphere 0.150 0.208 0.620 0.700 -12.9
1.2 Atmosphere 0.100 0.161 0.575 0.709 -23.3
In this case the increased efficiency of the propeller at the drag hump regime
becomes a decisive factor for choosing the CК series since the К series with its reduced
efficiency simply does not provide a sufficient thrust at the drag hump regime. (At the drag
hump regime, a hydrofoil, unlike a displacement ship, needs the thrust of propulsors to be
close or even exceeding that at the maximum speed regime).
The U.S. approach to the problem of designing a hydrofoil ship differs radically
from the Russian one. The power plant, e.g. gas turbine, is chosen in such a way as to have a
sufficient reserve of power to ensure a constant RPM of the propeller in all regimes
including the maximum speed and drag hump regimes. Then according to formula (12.1.1),
advance coefficient at the drag hump regime would decrease, as compared to that at the max
speed, by~ 50-35%, i.e. proportional to the ship speed. Then, as follows from the Table
12.2.1, (for cavitation number of 1.0 or higher) efficiency of the series K propellers with
wedge-type profiling is more than 9% higher than that of the series CK propellers with
segment-type profiling. Consequently, when using the U.S. approach, preference should be
given to the К series with wedge-type profiling. This had been realized by the American and
Canadian designers when developing the SCP for the hydrofoil Denison and oceanic
hydrofoil HMCS Bras d'Or.
481
The above comparison was made for two particular propellers of the given series,
having identical number of blades, blade contours, expanded area ratios, pitch ratios, forms
of pressure distributions, and advance coefficients, but different distribution of the section
thickness chord-wise. A more comprehensive comparison would require an analysis of both
series with regard to choosing optimal values (which are identical) of advance coefficients,
pitches, and expanded area ratios for each of the compared propellers to provide maximum
efficiency for a given cavitation number. The data required for such a comprehensive
analysis are given in Table 12.2.2. The last two columns in the table contain the coefficients,
which should be multiplied by factors identical for both compared variants (provided that
fluid density, thrust and speed are the same) to determine the optimal diameter and RPM of
the corresponding series propellers.
A straightforward analysis of the data Table 12.2.2 shows that, unlike the
comparison for the same advance coefficient and other parameters, no advantage is found for
the case considered here (max speed regime and cavitation number of 0.3) in the maximum
efficiency of either series under comparison. This is because the efficiencies of the optimal
wedge-type K series propeller and the optimal segment-type CK series propeller are
identical and equal to 0.66. It should be recognized, however, that this value is reached at
different values of advance coefficient, expanded area ratio and pitch ratio.
A further analysis shows that for the given fluid density, thrust and speed, the
optimal K series propeller would have 10% smaller diameter and 48% higher RPM as
compared with the optimal CK series propeller. The higher RPM and reduced diameter (with
other conditions being the same) constitute a substantial advantage of the K series propellers
with wedge-type profiling at the maximum speed regime.
The final choice of the series is again defined by the accepted design of the hydrofoil
ship as a whole. If a Russian approach is used, when the advance coefficient at the drag
hump regime is only slightly different from that at the maximum speed regime, the
preference should be given to the CK series with segment profile, as it follows from Table
12.2.2. The efficiency of the this series at the drag hump regime (cavitation number = 1.0,
J=1.5) is 15.7% higher than for the K series propellers.
Table 12.2.2. Comparison of characteristics of a 3-blade optimal K series supercavitating

propeller with wedge-type profiling and a 3-blade CK series highly cavitating propellers
with segment-type profiling, both at axial cavitation number of 0.3 (J - advance coefficient,
KT - thrust coefficient, η0 - open water propeller efficiency, AF /A0 - expanded area ratio,
P/D-pitch/diameter ratio).
axial K DT = 2
1 / K NT =
aim
cavitatin series
function
AF /A0 P/D J KT η0
#
J / KT KT / J 2
0.3 К maxη0 0.81 1.4 1.12 0.120 0.660 3.233 0.276
1.0 К not opt 0.81 1.4 1.12 0.140 0.610
1.0 К not opt 0.81 1.4 0.70 0.195 0.490
0.3 CК maxη0 0.95 1.8 1.50 0.177 0.660 3.565 0.187
1.0 CК not opt 0.95 1.8 1.50 0.209 0.706
1.0 CК not opt 0.95 1.8 0.90 0.290 0.457
482
Notwithstanding the indicated advantages of the K series at the maximum speed
regime, both in the case considered here and in the analysis in Table 12.21, the increased
efficiency of the propellers at the drag hump regime becomes a crucial factor for selecting
the CK series. On the other hand, if the U.S. approach is employed for designing a hydrofoil
ship, with an almost constant RPM and advance coefficient at the drag hump regime lower
than that at the maximal speed regime (by 40%, see Table 12.2.2), the preference should be
given to the K series with wedge-type profiling because its efficiency at the drag hump
(cavitation number 1.0, J=0.7) is 7.2% higher than for the СК propellers.
Thus, the comparative analysis of the propeller series differing only in section shape,
both with the same advance coefficient (as in Table 12.2.1), and for the propeller at the
optimal advance coefficient (as in Table 12.2.2), leads to the following conclusion. For
almost constant advance coefficient (Russian variant) segment profiling is preferable,
whereas for almost constant rpm (U.S. variant) the wedge-type profiling is better.
Another important conclusion can be made on the basis of the foregoing material.
When designing a supercavitating or a highly cavitating propeller for a hydrofoil ship the
project optimization at the maximum speed regime should be focused on generating
sufficient thrust at the drag hump regime. In other words, a high efficiency at the maximum
speed regime is not the only requirement when designing a supercavitating or highly
cavitating propeller for a hydrofoil ship.
Further progress in the development of supercavitating and highly cavitating
propellers in Russia was connected with a more complete utilization of computational
methods of the vortex theory. In 1970 A.A.Russetskiy and E.A.Fisher in the KSRI employed
the lifting line theory with a number of corrections, including those to approximate account
of cavities in the inter-blade space [Mavlyudov, 1982]. It is interesting to note that this
approach was semi-empirical, based on experimental data on cavity thicknesses at the blade
trailing edge [Fisher, 1966]. Obviously, the mentioned experimental materials were valid
only for the design of propellers with geometry similar to that of the tested ones. Based on
the described approach there was designed a small series of 3-blade highly cavitating
propellers СК2, comprising 4 models with expanded area ratio 0.9 and a radius-wise variable
pitch ratio, varying in the range 0.9-1.6 at 0.6R. The propeller models of this series were
specifically profiled shifting the maximum thickness and curvature of the pressure surface
15% from the mid-chord towards the trailing edge of cylindrical sections, while retaining a
segment form for the distributions near the leading and trailing edges.
The results testing the models of this series revealed no gain in the maximum
efficiency as compared with the СК series. However, it was possible to reduce relative
advance coefficient corresponding to the maximal efficiency (1.08 instead of 1.50 for the СК
series), which made it possible (in spite of a decrease in thrust coefficient) to reduce the
optimal diameter of this series propeller by 12% as compared to that (optimal and identical
in thrust) of the СК (therewith RPM increased by 56%).
Comparing with the data in 2. 12.2 for a conditional drag hump regime, it can be
seen that for a constant advance coefficient (Russian variant) the СК2 series propellers have
2.3% less efficiency than the СК series propellers. Shifting the efficiency maximum toward
smaller advance coefficients allowed one to noticeably increase efficiency of the "СК2"
series propellers at the maximum speed regime at axial cavitation number of 0.3, as
compared to the “СК" propellers, provided the advance coefficient was less than 1.2 (see
Table 12.2.3). For example, at J=1.0 the augmentation of efficiency was 18.2%.
483
In 1980 a small series of highly cavitating CК3 propellers was developed at KSRI
based on the lifting surface theory for non-cavitating propellers with cavity thickness taken
into account by a prescribed distribution of the simple layer intensity of which was
approximately determined using the previously mentioned experimental measurements of
the cavity thickness at the trailing edge [Fisher, 1966; Mavlyudov, 1982]. To provide
sufficient strength, the leading part of this series profiles was thickened in comparison with
the CК2 series. The profile was shaped segment-like, symmetric relative to the mid-chord, as
in the CК series, i.e. the idea of shifting the maximum thickness and curvature was abolished
in this series. Some data on this series are presented in Table 12.2.3. It is seen that the last of
the CK3 series is the best from all viewpoints. For example, this series propellers, which are
optimal in efficiency at the maximal speed regime, have a record-breaking efficiency of
0.68, 3% higher than that for similar К and CК series.
Table 12.2.3. Comparison of characteristics of three highly cavitating СК, СК2, СК3 series
propellers with expanded area ratio 0.95, and pitch ratios 1.4, 1.36, 1.33, respectively, (see
Tables 12.2.1 and 12.2.2 for notation).
J axial cavit. # KT , СК KT, СК2 KT, CК3 η0, СК η0, CК2 η0, CК3
0.8 0.3 - 0.115 - - 0.535
0.9 0.3 - 0.1 0.110 - 0.580 0.590
1.0 0.3 0.111 0.115 0.110 0.545 0.630 0.635
1.1 0.3 0.125 0.110 0.110 0.591 0.645 0.650
1.2 0.3 0.124 0.080 0.085 0.616 0.590 0.600
0.8 1.0 0.248 0.200 0.210 0.528 0.545 0.560
0.9 1.0 0.273 0.230 0.235 0.601 0.610 0.620
1.0 1.0 0.257 0.230 0.220 0.657 0.665 0.670
1.1 1.0 0.208 0.193 0.180 0.688 0.695 0.695
1.2 1.0 0.160 0.150 0.135 0.692 0.700 0.690
0.8 Atmosphere 0.356 0.330 0.340 0.555 0.770 0.585
0.9 Atmosphere 0.304 0.285 0.285 0.609 0.620 0.630
1.0 Atmosphere 0.257 0.238 0.230 0.657 0.665 0.670
1.1 Atmosphere 0.208 0.193 0.180 0.688 0.695 0.695
1.2 Atmosphere 0.160 0.150 0.135 0.692 0.700 0.690
The highly cavitating СК3 propellers with segment profiling were used in the early
1980’s at the world-largest oceanic hydrofoil ship Sokol (displacement ~ 465 t., 3x18000 hp
gas turbines plus 2x1100 hp diesels, 63 knot cruising speed [Rusetsky, 2000]). Six highly
cavitating propellers were mounted in counter-rotating pairs on three angular struts. At the
full speed the propeller blades operated in the supercavitation regime, though the profiling
was of the segment, not wedge, type (without shifting the maximum thickness and curvature.
The comparative tests of two sister oceanic hydrofoils, the Sokol and the Uragan,
which unlike the Sokol had only four (not six) highly cavitating CK propellers mounted on
two (not three) struts [Rusetsky, 2000]. The sea trials conducted in Russia revealed that the
Sokol had obviously better performance as compared to the Uragan. This was mainly due to
a better arrangement of wing system and a reduction of power (2x20000 hp plus 8000 hp for
the Uragan). By 1990 three Sokol-series boats had been built [Rusetsky, 2000].
484
It was also found that at a low cavitation number and a constant value of advance
coefficient, an increase of pitch for supercavitating or highly cavitating propellers, unlike
non-cavitating propellers, does not result is a significant increase of thrust coefficient. This
shows that using a greater pitch does not result in the same effect as usually expected for
non-cavitating propellers. Also, the thrust generated by a limited diameter SC propeller at
the design regime can be insufficient for propelling the ship at a given speed for any choice
of pitch. The loss of thrust can be recovered using a supercavitating propeller with a
rectilinear spoiler attached to the trailing edge in radial direction. For the experimental
verification a model of 3-blade supercavitating propeller was developed and tested in the
KSRI. It had expanded area ratio of 0.8, pitch ratio of 1.4, wedge-type profiling, and an
asymmetric contour with the trailing edge straightened for accommodating a radial spoiler,
protruding out from the blade by -3% to +5% of its chord. Tests of this mini series of
supercavitating propellers with a 2% spoiler (cavitation number 0.4, advance coefficient 1.2)
the thrust coefficient doubled as compared to that of the propeller without spoiler and
reached 0.15. Therewith, the efficiency at this advance coefficient even increased by several
percentage points. For smaller advance coefficients a drop of efficiency by 2% was
accompanied by a growth of thrust by 40-50%. The spoiler effect in the latter experiment
complies, at least qualitatively, with the results of theoretical research on a spoiler effect in
ideal-fluid mathematical models [Achkinadze & Fridman, 1995, 2000, 2000a;
Rozhdestvensky & Fridman, 1991; Baeva et al., 1981]. In particular, it was found that at the
regime of its highest effect the spoiler provides a significant decrease of cavity thickness in
the inter-blade space, as compared to a similar profile with no spoiler. The effect is due to
shifting the center of loading distribution towards the trailing edge. Positive effect of such a
shift on hydrodynamic lift-to-drag ratio of supercavitating foils at zero cavitation number
was found in [Tulin & Burkart, 1955]. In the early 1980’s the 200-ton displacement
hydrofoil ship Antaris was fitted with supercavitating propellers with spoilers as in the
above-mentioned mini-series. It reached 60 knots. In full-scale trails at an intermediate speed
of 50 knots the atmospheric air was penetrating along the inclined shafts to the propeller
blades. Therefore, supercavitating propellers with spoilers were used to generate the needed
thrust at that speed.
The story of using the supercavitating or highly cavitating propellers started
successfully in 1962 with the hydrofoil Denison and finished in the late 1980’s with the sea
trials and operation of the hydrofoil Sokol. The supercavitating and highly cavitating
propellers installed on these boats performed unreliably and required frequent repairs. Due to
this and other problems the Russian Sokol, like the U.S. hydrofoils Tucumcari and Pegas
have been re-fitted with water jets. Later on the interest to hydrofoils faded in connection
with appearance of more competitive concepts of high-speed craft, such as multi-hulls, SES
and SWATH for which the use of supercavitating propellers is hardly advisable.
Though the above-described applications of supercavitating propellers was
unsuccessful, the very idea of designing a cavitating (or ventilated) propulsor, less subject to
erosion and other negative consequences of cavitation or atmospheric ventilation, found its
further evolution in the development of several new types of propulsors. They include the
highly-skewed cavitating propellers [Georgievskaya et al., 1984, 1989] (see Fig.12.2.5),
trans-cavitating propellers [Toyoda et al., 1999; Ukon et al., 1999, 2004] (who proposed to
design a screw propeller having conventional sections (NACA or MAU) near the blade root
and gradually much more pronounced wedge-type supercavitating sections when
485
approaching the blade tip), surface piercing propellers (or semi-submerged propellers)
[Zaitsev & Maskalik, 1967; Basin & Goshev, 1963; Rose, 1993; Rose & Kruppa, 1991] (see
Fig.12.2.6 and Fig.12.2.7) and ventilated water jet [Mavlyudov et al., 1982; Ibragimova et
al., 1995] (see Fig.12.2.8). These propulsors are already in use and have good perspectives
for application on high-speed craft of different types.
Fig.12.2.5. Residual cavitation development at

various blade phase angles in a 16° oblique Fig.12.2.6. Surface piecing propellers in
flow, χ=0.65, J=1.0 [Georgievskaya et al., desighn regime: J=1.1 & 0.9; immersion
1989]. 0.7D [Kruppa, 1972].
Fig.12.2.7. Efficiency versus ship speed: Fig.12.2.8. Blade geometry of a

comparative data of surface piercing propeller ventilated water jet impeller blade
and traditional water jet [Rose, 1993]. [Mavlyudov et al., 1982].
486
Sec. 12.3. Mathematical Modeling

12.3.1. Preliminary Comments
A necessity for theoretical studies of the hydrodynamics of supercavitating propellers was
understood soon after the Pozdunine’s idea formulation, since there was no experience in
developing such propulsors. Early expectations were that supercavitating propellers would
be more efficient than the non-cavitating due to a reduction of friction on the part of blade
surface covered by a cavity. However, more detailed theoretical studies showed that for
supercavitating profiles having a wedge-type leading edge with a cavity starting at the
leading edge, no suction force occurs, and therefore, together with the elimination of friction
on the blade side under the cavity there appears a specific pressure drag called the cavity
drag. It reality, this drag component is comparable in value with the lost friction drag. As a
result a supercavitating propeller has an efficiency which is lower than that of the same
propeller at a higher advance coefficient and the same cavitation number, when partial or no
cavities rather than super-cavities there occur on the blades depending on the value of
cavitation number.
The principal theoretical problem of the new eliminated component is about the
optimal distribution of loading over the blade (both along radius and chord) with due
account of the required sizes of cavities forming on the blades. This design problem has been
solved for non-cavitating propellers. The main requirement of the supercavitating flow
regime is that the cavity length should be greater than the maximum width of the blade. As
for the cavity thickness, based on the linear theory, it should exceed the blade thickness
which is specified to comply with strength requirements.
Of a similar importance is also the problem of obtaining necessary characteristics of
a propeller analytically for any operational regime. This is a direct problem of hydrodynamic
analysis of supercavitating propellers (SCP) of a given configuration, advance coefficient
and cavitation number when Froude number is ignored and fluid is assumed weightless.
Accounting for gravity leads to an unsteady flow with a local cavitation number at a point of
cylindrical blade section is a function of time-dependent angular position of blade relative to
the propeller axis.
Both of the above problems can be solved using mathematical models of different
complexity. A number of such mathematical models have been developed, in particular [Liu
& Sulmont, 1969; Ivchenko, 1976; Sadovnikov, 1982; Yuasa, 1984; Yim et al., 1986;
Achkinadze, 1983, 1993, 2001; Achkinadze & Narvsky, 1985, 1987; Kudo & Ukon, 1994;
Kinnas, 1996, 2001; Kinnas et al., 1999; Young, 2002; Young & Kinnas, 2000, 2003] to
name a few. Moreover, a number of mathematical models were focused on surface piercing
propellers [Furuya, 1984, 1985; Barr, 1970; Vorus, 1991; Olofsson, 1996]. [Zaitsev &
Maskalik, 1967; Kruppa, 1991; Ferrando et al., 2001].
12.3.2. Regression Models for Supercavitating Propellers
The simplest albeit very actively employed mathematical models include regression analysis,
processing the test results of systematic series of the SCP and HCP models. Coefficients of
the regression polynomials presented in [Ferrando et al., 2001; Kozhukharov, 1983] were
obtained by analyzing test data for the following series of highly cavitating propellers:
• 3-blade СК highly cavitating propellers recommended by Tsapin & Sadovnikov in 1955;
• 5-blade series of cavitating propellers adapted for oblique flow operation without
487
erosion of the pressure side considered by Georgievskaya & Rusetskiy in 1965;
• 3-blade series of screw propellers [Newton & Rader, 1961];
• 3-blade series of screw propellers [Gawn & Burrill, 1957].
The regression polynomials, obtained separately for the thrust coefficient and the
moment coefficient, express these quantities as functions of advance coefficient, cavitation
number, expanded area ratio and pitch ratio. Having such a mathematical model for a given
series one can easily develop the corresponding algorithms for computing calculations
design and analytical problems. The main deficiency of such an approach consists in the
impossibility of designing a propeller with any parameters beyond the range of the series
tested in a cavitation tunnel. E.g. it is impossible to account for a non-uniformity of a flow
coming upon the propeller if the tests were conducted in the uniform flow. In addition,
designing the series and potentials for their improvements should also be discussed. These
problems cannot be solved without regression mathematical models. To solve them, different
mathematical models of vortex theory of propellers should be used. The following section
presents some models of supercavitating propellers, developed in 1985 in Leningrad
Shipbuilding Institute (now Saint-Petersburg State Marine Technical University) in
[Achkinadze, 1983; Achkinadze & Narvsky, 1985, 1987, 1989].
12.3.3. Lifting Surface Theory as a Model for Supercavitating Propellers
Let’s assume that there are Z (the number of blades) symmetrically arranged helical
surfaces (called nominal) with a common x-axis, oriented in the propeller motion direction.
Designate one of these surfaces S and consider further only this nominal surface, accounting
for the necessity of summing up all induced velocities from all remaining propeller blades
The nominal surface geometry for the x-axis is uniquely determined by pitch PS,
which is determined by a special algorithm accounting for prescribed thrust coefficient KT,
advance coefficient and parameters of the given non-uniform inciming velocity field. The
algorithm is called a Generalized Linear Model (GLM) [Fabula, 1962] and has been used
successfully for designing not only supercavitating but also non-cavitating propellers
[Achkinadze et al., 2000; Achkinadze & Krasilnikov, 1997].
In the linear theory of lifting surface, the hydrodynamic singularities are
continuously distributed over the nominal surface of the blade and the wake behind it.
According to the potential flow hypothesis, it is assumed that the considered flow has
potential in the whole flow domain except the surfaces occupied by hydrodynamic
singularities (for all blades). The flow is assumed incompressible, inviscid, gravity-free,
unbounded and steady (in a moving coordinate system attached to a uniformly rotating and
uniformly moving along the propeller’s x-axis). Then, based on the theory of potential,
knowing the potential induced by distributions of a source layer with scalar intensity q, and a
potential of surface layer of vorticity vector γ , one can integrate over all corresponding
domains and find the direct value of the perturbation velocity vector due to a simple layer:
⎛ 1 ⎛ 1 ⎞ ⎞⎟ 1 ⎛1⎞ 1 R
Wq ( X 1 , Y1 , Z1 ) = grad ⎜ − ∫ q ⎜
⎜ 4π S ⎝ R ⎠ ⎟⎟ dS = − ∫
4π Sq
q ⋅ grad ⎜ ⎟ dS =
⎝R⎠
∫
4π Sq R
q 3 dS ; (12.3.1)
⎝ q ⎠
1 γ ×R
4π S∫ R 3
and due to a vortex layer (both bound and free) Wγ ( X1, Y1, Z1 ) = dS , (12.3.2)
G
where R - radius vector directed from the integration element toward a point where the
488
velocity is calculated; Sq is a part of the nominal surface upon which the source layer is
located; SG is a part of nominal surface on which the vortex layer is located.
As known from the theory of potential, there is a jump of the perturbation velocity
normal component at the points of the nominal surface where intensity of the source layer is
not equal to zero, and there is a jump of the perturbation velocity tangential component at the
points of the nominal surface where intensity of the vortex layer is not equal to zero. With
the latter taken into consideration, we have for the perturbation velocity normal component
on the pressure side of the nominal surface (normal vector is positive when directed from the
pressure side to the suction side of the nominal surface)
Wn− = Wqn + Wγn + q / 2 ; (12.3.3)
and for the tangential component directed along the proper helical lines forming the nominal
surface, designated by index τ, the perturbation velocity component on the suction side of
the nominal surface (positive when directed toward the leading edge)
Wγτ+ = Wγτ − γ r / 2 . (12.3.4)
To preserve potential character of flow beyond the system of singularities, there
should be satisfied a known relationship between the radial and tangential components of
vector γ , which has zero normal component to the nominal surface
γ τ = ∂Γ / ∂r , (12.3.5)
where the double layer intensity, equivalent to the system of distributed continuous vortices
ξ
is Γ(r , ξ ) = − ∫ γ r (r , ξ )dξ . (12.3.6)
ξ LE
Eq. (12.3.35) is equivalent to the vortices preservation condition (zero surface divergence of
vector γ ). Both (12.3.5) and (12.3.6) are valid only for the case of absence of discrete
vortex lines and zero value of Γ at the blade profile and vortex wake. Vorticity vector is:
γ = n × grad Γ . (12.3.7)
Besides the perturbation velocities there must be taken into account a field of
transport velocities together with the moving coordinate system fixed to the rotating and
transitionally moving propeller VE and incoming flow field Vψ , so that the field of relative
velocity VR in the coordinate system fixed at the propeller can be found using the formula
VR = Wq + Wγ + Vψ − VE . (12.3.8)
Using the corresponding Euler equation integral, pressure in the fluid in the moving
propeller-fixed coordinate system can be determined, if the field of perturbation velocity is
found W = Wq + Wγ , (12.3.9)
(W + VΨ ) 2
by formula p − p0 = − ρ + ρ ⋅ VE (W + VΨ ) . (12.3.10)
2
According to the linear approach, the induced and absolute velocities of the given
incoming flow are assumed of being small values (with only the first order terms
considered), then one can derive
p − p0 = ρ ⋅ VE (W + VΨ ) = ρVEτ Wτ + ρVEτ Vψτ = ρVEWτ + ρVEVψτ , (12.3.11)
489
where index τ denotes projections of the vectors on tangents to the proper helical lines
forming the nominal surface. In the linear theory the directions of vector VE and induced
tangential velocity are coincidental within accuracy of the higher order term.
Using formula (12.1.3) for local cavitation number σ and previous equation, one
can obtain a linearized dynamic condition at the cavity boundary, which physically signifies
that the pressure at the cavity boundary equals to the pressure of saturated fluid vapor
( p = pV ), Wτ+ / VE = −σ / 2 − Vψτ / VE in SCAV , (12.3.12)
where SCAV is a part of the suction side of the
nominal surface covered by the cavity in design
regime. Note that the domain occupied by cavity
SCAV does not coincide with the blade contour
projection on nominal surface SB (see Fig.12.3.1).
When deriving (12.3.12) attention was paid
to the fluid being gravity-free, and therefore, the
pressure at infinity, which is a quantity in the local
cavitation number, coincides with the Euler integral
constant ( p 0 rθ = p0 ). Pressure surface Sp of the
supercavitating propeller is assumed free of
cavitation at the design regime. Therefore, a
kinematic condition of zero normal velocity should Fig.12.3.1. Domain nomenclature.
be satisfied on this surface (its form is to be
determined in the solution process as that of the cavity). The kinematic condition can be
written without simplification in a general way as V R ⋅ n = 0 on Sp. (12.3.13)
Now all is ready for formulating the design problem in its least general (basic) form
(Problem Statement А), when there is no freedom of choice and, consequently, no possibility
for design optimization. Problem А is aimed at finding geometry of blades and cavities for a
given loading distribution over the supercavitating propeller blades. Thus, the following are
given: pitch of the nominal surface, blade contour projection on this surface, hub diameter
(the hub is assumed to be infinite and cylindrical), advance coefficient and axial cavitation
number χ which, according to (12.1.1) and (12.1.2) and ignoring gravity, is related to σ by
obvious formula σ = χ sin 2 β , where tan β = J /(π r ) .
If the loading is prescribed then the intensity of the attached vortex layer is also
prescribed (i.e. the radial component of vector γ is given) on the nominal surface over the
blade or, in other words, circulation distributions along the radius and the blade chord are
prescribed. Unknown in this case is the intensity of source layer q modeling the cavity.
Blade thickness in this case is ignored, as the cavity is assumed covering completely the
suction side starting from the leading edge. However, since loading distribution over the
blade cannot be chosen arbitrarily, the obtained solution of problem А can give a thickness
which is unacceptable for the blade strength. This is because no additional conditions are
envisaged within the basic formulation.
After finding the cavity platform (i.e. after finding region SCAV) and determining the
intensity q of the souce layer distributed over this surface, one can easily find the form of
the blade pressure side (i.e. find pitch distribution and shape of the pressure side) as well as
all necessary hydrodynamic characteristics of the propeller under design.
490
Thus, determining the form of cavity represents a central point of problem А. This
problem can be studied using a linear theory of cavitating flows and choosing a closure
model of the open type proposed in the linear 2-D case by Fabula [1962] (a few months later
a nonlinear formulation of a similar model for the supercavitation regime was published by
Wu, and later by Bassanini for the partial cavitation regime). The choice of this model is not
incidental.
As a matter of fact, the closed cavity model in a 2D case predicts an unrealistically
large lift coefficient of a supercavitating foil when the cavity closure region is near the foil’s
trailing edge. This phenomenon is sometimes called a “Geurst paradox” [Geurst, 1956;
Achkinadze et al., 2007]. The advantage of the open model for design regimes when the
cavity length equals to or approaches unity ( σ / α = 8-10 for a flat plate) is obvious. It
should be noted that the indicated regime is most characteristic for the supercavitating
propeller blades, because the maximum efficiency of these propulsors is reached exactly at
the minimum cavity length at every section, i.e. the minimum cavity length causes no
erosion. This length, as indicated, would only slightly exceed that of the corresponding
chord. Applying the closed model in this case is either incorrect or requires special effort to
resolve the Geurst paradox. Equation (12.3.12) is the 2D singular integral equation (SIE)
with respect to intensity of source layer q with an unknown domain of integration Sq
(domain Sq for the sought solution should coincide with the domain covered by cavity SCAV).
This is the equation to be solved in Problem А.
For uniqueness of the SIE solution it is necessary to specify the class of functions for
prescribing the unknown function q and loading, i.e. given is the radial component of the
vorticity vector (i.e. intensity of the bound vortex layer prescribed within the blade). In
accordance with the open cavitation model, function q should vanish at the rear boundary of
the cavity and beyond domain SCAV. At the front boundary coinciding with the blade leading
edge, the sought function q should have a singularity expressed as a power function with
exponent of −1/4 like in a thin supercavitating 2D plate (rounded leading edge is ignored
here) Therewith, in all other points the unknown function q is assumed to be sufficiently
smooth to satisfy as an existence condition for 2D singular integrals.
A traditional approach [Kinnas et al., 1999] to treatment of the formulated problem
consists of organizing an iterative process to determine the approximate cavity shape. Due
attention has not always been paid to convergence, uniqueness (understood as a continuous
dependence of the result on the initial approximation) and efficiency of the iterative process.
In addition, the traditional approach doesn’t make it possible to conduct (by setting free
several parameters) a numerical optimization of the blade geometry with an account of
specified requirements. The latte should provide sufficient thickness and length of the cavity.
An non-traditional method of artificial variational problem using effective
procedures of linear programming was developed and applied [Achkinadze, 1974, 1983;
Achkinadze & Narvsky, 1985] for solving Problem A and other problems of supercavitating
propellers. Let’s consider this formulation of problem А for a supercavitating regime
whereby the cavities generated at the sharp leading edge have a length exceeding that of the
corresponding chord of all the blade sections (no excepts), namely:
LCAV /c > 1 for r ∈ [ rH ; 1]. (12.3.14)
It is difficult to say whether this assumption is satisfied, and so one should be able to
filter out the calculation results which do not comply with condition (12.3.14). Variational
491
(nontraditional) formulation of problem А differs from the traditional one, formulated above,
as follows:
1. integral equation (12.3.12) would now be satisfied only at the points of a given domain
SB, i.e. for the points corresponding to the blade’s suction side, namely,
− Wτ+ / VE = σ / 2 + Vψτ / VE in SB; (12.3.15)
2. for a domain beyond the blade, the integral equation is replaced by a corresponding
integral inequality − Wτ+ / VE ≤ σ / 2 + Vψτ / VE on SDET, (12.3.16)
which should be satisfied not on an unknown part of domain SCAV (beyond the blade),
but on a rather simplified stretched downstream (to embody the calculated cavity) and
located beyond the blade domain SDET, Fig.12.3.1;
3. beyond the blades for the open cavity closure model, it follows rigorously from the
convexity of stationary cavity [Gurevich, 1979] in weightless fluid
q ≥ 0 in SDET; (12.3.17)
4. behavior of both given and sought functions in the vicinity of the sharp leading edge is
as follows lim q (ξ ; r ) ⋅ (ξ − ξ 0 )1 / 4 = lim γ (ξ ; r ) ⋅ (ξ − ξ 0 )1 / 4 ≠ ∞ . (12.3.18)
ξ →ξ 0 ξ →ξ 0
The latter requirement coincides with the traditional problem formulation and has been used
when linearizing the functional of the artificial variational problem.
Note that condition (12.3.16) is universal – it is valid for the points both at the cavity
boundary and beyond the cavity, because the pressure beyond the cavity boundaries for
natural cavitation in weightless fluid should be greater than in the cavity and at its
boundaries. It should also be noted that when replacing condition (12.3.12) with a set of
requirements (12.3.15) and (12.3.16), a multitude of admissible solutions of the problem
widens considerably and becomes indefinite.
A comparison of the traditional and nontraditional formulations of problem A
reveals that when replacing equality with inequality in the dynamic boundary condition, as
indicated, the set of admissible solutions widens and becomes infinite. To obtain a unique
solution in a non-traditional formulation, one should minimize the artificial functional F over
the multitude of solutions, satisfying conditions (12.3.15)-(12.3.18). In this case the
functional is written as follows F = ∫ (Wτ+ / VE + σ / 2 + Vψτ / VE ) q dS . (12.3.19)
S B ∪ S DET
It is easy to check that according to conditions (12.3.15), (12.3.16) and (12.3.17), the
integrand equals to zero in domain SB and both multipliers in the integral are non-negative in
domain SDET, thus minF=0. Simple considerations show that the indicated minimum (zero)
value of functional F is reached exactly at the very solution of the problem A corresponding
to the traditional formulation. The functional differs from zero when within SDET there is an
area element for which both multipliers in the integrand are positive simultaneously. This
can take place if the source distribution intensity and simultaneously the pressure on this
area element are higher than that in the cavity. The latter does not correspond to the
traditional solution, for which no sinks and sources exist beyond the cavity (on its surface
the pressure is minimal and constant). But if F equals to zero, there are no such area
elements in the area in question and, consequently, the obtained solution corresponds to the
traditional approach. Thus, the described artificial variational formulation of the problem А
is equivalent to the traditional one presented earlier. It should be noted however that the
492
artificial functional is not linear. Let’s transform it to obtain its linear form considering
⎛ Wqτ Wγτ γ σ Vψτ ⎞
previous relationships: F = ∫ ⎜⎜⎝ VE + VE − 2VrE + 2 + VE ⎟⎟⎠ q dS . (12.3.20)
S B ∪ S DET
The first term can be eliminated due to the equality
∫ (W τ / V
S B ∪ S DET
q E ) q dS = 0 , (12.3.21)
which can be proved by interchanging the order of integration in the integral. One should
keep in mind that the leading edge is assumed to be sharp (according to condition (12.3.18))
and the suction force is not realized on it. Eventually one can derive a complete linear form
⎛ Wqτ Wγτ γ σ Vψτ ⎞
of the functional F= ∫ ⎜⎜⎝ VE + VE − 2VrE + 2 + VE ⎟⎟⎠ q dS . (12.3.22)
S ∪S
B DET
When proving the functional’s linearity, it is taken into account that the intensity of
the bound vortices γ r is assumed to be prescribed within the blade, i.e. within SB, and
therefore all terms in the brackets are given functions or can be calculated. The functional’s
linear form allows (after discretization) to reduce the numerical solution of the problem А in
its artificial variational formulation to the problem of linear programming. A numerical
solution of the problem А was obtained using the simplex method of linear programming for
a model of supercavitating screw propeller [Achkinadze & Narvsky, 1985, 1987]. The
propeller model, Fig.12.3.2, was made of the silumin alloy, was 0.2 m in diameter and was
tested first in the KSRI cavitation tunnel and
then in the Gdansk Institute cavitation tunnel.
Some results of calculations and experiments
presented in [Achkinadze & Narvsky, 1987] are
reproduced in Table 12.3.1 and Figs.12.3.3,
12.3.4.
Fig.12.3.2.
Supercavitating propeller model N8401
designed in LSI with cavity thickness measured:
Z=3; AE/A0=0.92; P/D(0.6)=1.28; χ=0.25;
J=1.0; KT=0.084.
At the design point the propeller thrust

coefficient was measured 0.0844 while its
design value was 0.0840; the measured efficiency was 0.543, i.e. 4% more than that obtained
by calculations. A good correlation in the forces confirmed validity of the approximate
correction, accounting for a partial loss of the integration contour when determining
circulation around a given cylindrical section of blade, related to the presence of cavities in
the inter-blade space. It should be noted that in the cavity closure model, the super-cavity
terminates, for the given cylindrical section, in its widest cross section and then forms a
helical-type stripe of a semi-infinite wake of constant width. As the cavity thickness is
unknown in advance, the interaction process is used accounting the indicated approximate
correction (two steps appear to be sufficient). In the presented calculations, the final
correction resulted in 18.6% reduction of thrust as compared to a purely linear approach.
493
Table 12.3.1. Input data (first 5 lines) and the results of verification calculations (problem
А) of the supercavitating propeller N8401(Z=3; AE/A0=0.92; J=1.0; χ =0.25; KT =0.084).
r 0.265 0.404 0.595 0.786 0.925
c/R (preset) 1.045 1.245 1.416 1.355 0.939
δ (preset) 0.087 0.060 0.037 0.027 0.030
Γ (preset) 0.0145 0.0187 0.0213 0.0195 0.0129
С L (preset) 0.1340 0.1170 0.0892 0.0680 0.0562
σ (preset) 0.1479 0.0957 0.0556 0.0352 0.0265
P/D (calculated) 1.69 1.37 1.28 1.28 1.26
δp (calculated) -0.0007 0.0073 0.0124 0.0141 0.0124
LCAV/c (calculated) 1.000 1.000 1.080 1.160 1.000
eCAV (L) /c (calculated) 0.282 0.140 0.076 0.084 0.069
eCAV (l) /c (calculated) 0.282 0.140 0.074 0.075 0.058
eCAV (l) /c (experimental) - - 0.080 0.081 -
Note: here c is the chord length of blade section, R is the propeller radius, δ is the thickness-
chord ratio of the thickest blade section, Γ is dimensionless circulation around blade section, CL
is the lift coefficient, σ is the local cavitation number, P/D is the blade section pitch ratio, δp is
maximum camber of blade pressure surface, LCAV is the cavity length, eCAV (L) is the cavity
thickness at the cavity termination point for given blade section chord.
a)
b)
c)
Fig.12.3.3. Supercavitating propeller N8401 in
design regime: (a) cavity boundaries; (b) cavity
configuration on the suction surface; (c) radial
cavity thickness on the blade trailing edge.
The main difficulties when testing were connected with thin sharp leading edges that
supposed to be stiff enough to maintain its shate under load in a cavitation tunnel. Therefore,
a loading factor was applied to preserve the leading edge singularity. When manufacturing
the model, a certain material was added to fill the gap between the cavity surface and the
blade’s pressure side in the vicinity of the leading edge. The thickness distribution of the
remaining part of section had a segment profiling, Fig.12.3.3c. The circulation distribution
494
along the radius was assumed optimal, and found in the frames of the lifting line theory
using the generalized optimum condition in [Rozhdestvensky, 1977; Achkinadze, 1989;
1993] and allowing to account for the blade’s cavitation-induced and frictional drag
distribution along radius.
Pressure distribution along chord was assumed to be almost uniform with an
increase of loading near the leading edge. Sketches of cavity development and thickness
measurements at the cavity training edge were made during the model testing in the
cavitation tunnel. The correlation with the cavity shape calculations, Fig.12.3.3b, and cavity
thickness distribution along radius, Fig.12.3.3 and Table 12.3.1, was 1.0 for the design
advance coefficient and 0.25 for the design cavitation number, which seems satisfactory. At
the design regime, a cavity began forming not at the leading edge but 5-10% of chord further
downstream. As a result of this, at the design advance coefficient the entire suction side near
the leading edge was covered by a dense bubbly cavitation, whereas a film-type cavitation
was formed at advance coefficients 5% less than the design one. There are reasons to expect
that the full-size propellers would have a film-type cavitation rather than bubble cavitation.
At design advance ratio J=1.0 the pressure side was completely cavitation-free, Fig.12.3.4.
Tests of this supercavitating propeller model have been later conducted in Gdansk
by Prof. Szantyr and Dr. Dudziak in the oblique flow conditions at axial cavitation number
of 0.4, which is noticeably higher than the design value of 0.25. However, for this off-design
regime in oblique flow the pressure surface was cavitation-free, which confirms a sufficient
margin against cavitation on the pressure side.
Efficiency of the tested propeller was close to
or a little higher than that of the СК series
propeller, though its section was noticeably
thicker (specifically, the maximum blade
section thickness/chord ratio was greater than
that for the СК-propeller by 16, 33 and 50%
for the sections with relative radius 0.6, 0.8,
0.95, respectively. Besides, the thickness near
the leading edge was in fact greater for all
cross sections where the relative radius was
less than 0.8).
Further theoretical development was
focused at obtaining such a form of the
artificial functional, which coincides with the Fig.12.3.4. Blade surface cavitation
approximate (within the lifting line theory) development at various advance ratios
expression of the propeller profile losses due for supercavitating propeller N8401
to cavitation drag of blades only. Such a (pressure surface is on the right).
functional, as shown in [Achkinadze, 1993],
has the following simple form
F1 = ∫ (σ / 2) q sin β dS . (12.3.23)
S B ∪ S DET
This functional is linear and should be minimized, although it no longer has a zero value. In
obtaining such a form of the functional a number of assumptions were made. To evaluate the
error introduced by these assumptions, a numerical solution of the problem А was obtained
for the supercavitating propeller described above. Comparison with the results obtained
495
using a rigorous form of functional (12.3.22) showed an insignificant divergence in the the
cavity shape and the pressure side of the blade in the two cases compared. This makes it
possible to solve the design problem using a formulation which is more comprehensive than
that for the problem А, i.e. with accounting for the restrictions on cavity thickness and
length, margin against cavitation on the pressure side, etc. for a given circulation distribution
along the radius. The latter can be either prescribed arbitrarily or assumed optimal, as before,
using the generalized optimum condition found in the lifting line theory [Achkinadze,
1989]).
Besides the cavity geometry the latest development allowed one to obtain an optimal
distribution of the bound vorticity layer along the chord and the corresponding form of the
pressure surface using a numerical method of linear programming. With such optimization
the induced losses and friction losses remain unchanged and the profile losses, stipulated by
the cavity drag, are minimized. The latter allows one to use the functional in form (12.3.23)
as an application of a certain variational principle (the principle of minimum cavitation
drag), having a physical sense of cavitation drag minimization of the supercavitating
propeller blade sections. The latter is valid at least for each section separately, which can be
easily seen from the cavitation drag expression for a 2D foil in the open closure cavitation
model [Achkinadze, 1993], CDCAV = σ eCAV ( L) , (12.3.24)
where eCAV (L) is cavity thickness at the termination point (for open model this is the largest
value of cavity thickness), i.e. where the source layer modeling the cavity terminates. Using
this formulation, design calculations were conducted for a systematic series of the optimal 3-
blade supercavitating propellers with segment profiling. For practical use the results were
presented as corrections to curvature and angle of attack due to the lifting surface effect on
supercavitating propellers [Achkinadze & Narvsky, 1989], similar to how it was done for the
non-cavitating propellers [Mensaas & Slaattelid, 1971]. Fig.12.3.5 reproduced from
[Mensaas & Slaattelid, 1971] shows a comparison of the mean line curvature corrections for
the blade-cavity body’s cylindrical sections of supercavitating propellers with that for
cylindrical sections of non-cavitating propellers. It can be seen that the corrections for
supercavitating propellers depend on cavitation number.
For specific values, e.g. at a relative radius of 0.79, one can see that at expanded area
ratio 0.475 the curvature correction slightly exceeds that for non-cavitating screw propeller.
For the propellers of expanded area ratio 0.950, the difference is much greater with the same
trend. The value of corrections for non-cavitating propellers are marked in Fig.12.3.5 by
horizontal dashed lines for two expanded area ratios. One may expect that with cavitation
number increasing (at thrust coefficient being the same), corrections for the supercavitating
propellers should approach similar corrections for the non-cavitating propellers.
This can be explained by a greater angle of attack that is needed for obtaining a
sufficient thickness and length of cavity for an optimal supercavitating propeller. Similar
tendency takes place for the curvature corrections, see Fig.12.3.5. With an increase in
cavitation number this requirement is more difficult to satisfy and the optimal pitch and
curvature increase. For reasonable cavitation numbers (for the propeller N8401 the ratio of
axial cavitation number to thrust coefficient, not corrected for the nonlinearity, is equal to
2.5) the corrections for curvature for cavitating and non-cavitating propellers at expanded
area ratio 0.475 differ insignificantly (less than 20%), which justifies consideration of the
blade and the cavity as a single lifting body in a flow of inviscid fluid.
496
When using the optimization methods, such as the linear programming, there arises
a need in smoothing the pressure surface, because as a result of the design calculations with
optimization, it can become excessively complicated, and the influence of this complication
on propeller efficiency would be negligibly small. The principle of minimum of cavitation
drag was applied in [Achkinadze,
1974] to solve a 2D problem of an
optimal supercavitating section. It was
found that an almost optimal shape of
the pressure surface can be obtained
using rather simple and technological
forms provided that the form, angle of
attack, and thickness distribution are
selected to satisfy the conditions of
sufficient thickness and length of the
cavity, see [Achkinadze & Fridman,
1995] for some examples of such
calculations.
Special attention should be
paid to choosing the optimal lift
coefficient with taking into account
conditions for sufficient cavity
thickness and friction drag Cf on the Fig.12.3.5. Correction for the camber of the
pressure side. For example, this blade cylindrical section and the cavity
problem was solved for a thin 2D arc at (comparative data of supercavitating and non-
zero cavitation number. In particular, cavitating propellers; KTI = ideal thrust coef.)
the following values were obtained for [Achkinadze & Narvsky, 1989].
the optimal lift coefficient, angle of
attack and curvature of the pressure side [Achkinadze & Fridman, 1995]
CLOPT = 2.548 C f , α OPT = 0.1838 CLOPT , δ p = 0.1294 CLOPT .
The approach described above is linear. As it becomes obvious in the course of time,
the employed linear approach does not adequately allow evaluating the effect of real (non
sharp) leading edge. Recent developments in the field of optimal supercavitating profiles are
associated with using a nonlinear approach and studying the influence of spoiler, wedge-type
leading edge and other nonlinearities [Achkinadze & Fridman, 2000a]. It should be noted
that improving the hydrodynamic fineness of sections, measured by a relative decrease of
inverse fineness Δε / ε leads to higher efficiency of supercavitating propellers, which can
be approximately estimated using a simple formula [Achkinadze, 1993]
Δη0 / η0 = −2ε (Δε / ε ) (12.3.25)
if the inverse fineness is taken for the section at relative radius 0.7.
It follows directly from (12.3.25) that relative improvement of the profile’s
hydrodynamic fineness gives a relative improvement of a propeller which is K/2 less
efficient (where K=1/ε is the hydrodynamic fineness of the supercavitating profile). For a
hydrodynamic fineness of about 20 there takes place a 10 times reduction of the relative gain
when recalculating from a profile to the propeller. Although an increase of the propeller
efficiency is expensive, the increase of the supercavitating propeller’s efficiency by
497
optimization of its profiles requires an improvement of their hydrodynamic fineness by at
least 10% to obtain only 1% growth in efficiency.
Sec. 12.4. Simple Mathematical Statements and Numerical Algorithms

The three preceding sections present the main problems of cavitating propellers and ways of
solving them numerically. Considered in this section are numerical solutions for a cascade of
arbitrary blades in non-cavitating as well as in partial and supercavitating flows. The
method and algorithm were proposed in [Terentiev, 2009].
12.4.1. Problem Statement for a Propeller with a Blade of High Aspect Ratio
Let the propeller rotate with angular velocity ω and move along the x-axis at velocity U.
Consider cylindrical coordinates x, r , θ . The Laplace equation in these coordinates is
∂ 2ϕ ∂ 2ϕ ∂ϕ ∂ 2ϕ
+ + + =0. (12.4.1)
∂x 2 ∂r 2 r ∂r r 2 ∂θ 2
If the aspect ratio of the blade is large then the

radial velocity components should be much smaller than the
others, and so the second and third terms in equation
(12.4.1) can be neglected. Denoting rθ = y , equation
(12.4.1) is reduced to an ordinary 2D Laplace equation
∂ 2ϕ ∂ 2ϕ
+ =0 (12.4.2)
∂x 2 ∂y 2
for a flow around a cylindrical surface of constant radius r,
that is similar to a planar flow at inlet velocity v1 = (U , ω r )
through cascade of blade sections of that radius, Fig.12.4.1.
The distance between two blades is equal to T = 2π r / n ,
Fig.12.4.1. Schematics of
where n is the number of propeller blades. As usually,
flow through cascade.
kinematical conditions ψ = const or ∂ϕ / ∂n = 0 should be
satisfied at the blade boundaries, while for cavitating flow dynamic condition
p = p0 = const at the cavity boundary must be satisfied as well. Besides, the outlet velocity
the right in infinity, v2 e −iα 2 , differs from inflow speed v1e − iα1 , which should be found by the
given forward and angular speeds and geometric parameters of the cascade. Conservation of
fluid amount and circulation connect both speeds at the left and right sides:
U = v1 cos α1 = v2 cos α 2 , (12.4.3)
Γ = Tv2 sin α 2 − Tv1 sin α1 . (12.4.4)
Γ
Whence, tan α 2 = tan α1 + , (12.4.5)
Tv1 cos α1
v2 = v12 cos 2 α1 + (v1 sin α1 + Γ / T ) 2 . (12.4.6)
If pressure at infinity at the left is denoted p1 , then the cavitation number,
σ r = 2( p1 − p0 ) / ρ v12 , or σ r = v02 / v12 − 1 could be given. Visually, cavitation number σ r
498
depends on radius r, i.e. the flow at each blade section realizes by its own cavitation number.
This problem statement of non-cavitating flow has been considered by many authors (some
references and analytical studies can be found in [Sedov, 1950; Stepanov, 1962]). Exact
analytical solutions have been obtained only for few configuration of cascade blades, but
there are many numerical methods for calculating a wide range of problems.
12.4.2. Centrifugal Pumps and Propellers with Blades of Low Aspect Ratio
There is no axial speed in centrifugal pumps, so
equation (12.4.1) is reduced to
∂ 2ϕ ∂ϕ ∂ 2ϕ
+ + = 0, (12.4.7)
∂r 2 r∂r r 2∂θ 2
i.e. the flow in this case is also identical to a 2D one.
Since the blades are arranged periodically over a
circle and the fluid comes from the axis of revolution,
the only possible flow is a point vortex-source of
strength m = 2πω R 2 + iq , where R is a determinative
radius, ω is the angular speed, q is the amount of
fluid flowing through the blades, Fig.12.4.2.
The radius can be determined from the
Fig.12.4.2. The sketch of flow following consideration given by L.I. Sedov: the
through blades on a circle. circulation along any contour at infinity out of the
cascade should be zero then the circulation of the
vortex can be found by satisfying the Kutta – Joukowski condition at the trailing edge.
Hence, the circulation of vortex is equal to the total circulation around the blades,
Γ = 2πω R = nΓ c , (12.4.8)
where Γc is the circulation around the blade.
Using the transformation z =eζ (12.4.9)
one can transform a ring grid to an ordinary linear grid with inlet velocity at its left-hand side
being v1=(q – iΓ)/2π .
12.4.3. Green Function for Periodic Flow
Consider a cascade with period T0=Teiβ
(Fig.12.4.3 shows the partial cavitating
cascade).
Fig.12.4.3.
A foil-cascade with partial cavities:
straight line b1b2 is neutral; line a1a2 is directed
towards the average speed.
The numerical methods presented in Sec. 8.1

are valid for finite numbers of bodies only, but
a cascade has infinite numbers of blades. More over, the only inlet velocity, v∞ e −α i , is in
case of finite numbers of foils, while the velocity on the right in infinity for cascade,
499
v2e −α 2i , differs from the inflow speed on the left, v1e −α1i . Thus the above-mentioned
numerical method has not been applied directly to cascades
The integral equation for a cascade can be found from equation (8.1.15) used for a
finite number of foils by increasing that number to infinity. Since the blades of cascade are
arranged as periodically located foils, the integration over all foil boundaries can be reduced
to the integration over only one foil, C. The kernel of integral can be found as a limit of the
⎛ ⎞
⎟ = − 1 Re ln ⎛ (τ − z ) ⎛1 − (τ − z ) ⎞ ⎞ .
n ∞ 2
1 1 Tk
sum Gc = lim ⎜ ln + ∑ ln ⎜⎜ ∏ ⎜ ⎟⎟
2π n →∞ ⎜⎜ τ − z k =− n z − τ − T k ⎟⎟ 2π ⎝ k =1 ⎝ k 2T 2 ⎠ ⎟⎠
⎝ k ≠0 ⎠
Considering the sine as infinite product [Gradstein & Ryzhik, 1962, 1.431,1], the product
1 ⎡ Teiβ ⎛ π ⎞⎤
can be written as Gc = − Re ln ⎢ sin ⎜ iβ (τ − z ) ⎟ ⎥ . (12.4.10)
2π ⎣ π ⎝ Te ⎠⎦
However, the inlet velocity for a single foil should be the same as the outlet one, i.e.
v1e − iα1 = v2 e − iα 2 = v∞∗ e − iγ . Hence the circulation around a foil should be zero and both inlet and
outlet speeds are directed along neutral axis b1b2 , Fig.12.4.3. Equation (8.1.15) can be
∫ v (sτ )G ( z, sτ )dsτ + c = Im(v
∗ ∗ − iγ
rewritten as c ∞e z ), z ∈C . (12.4.11)
C
Two methods of numerical solutions of cavitating flow through a cascade of any foils are
considered below: one is based on conformal mapping another on modified integral equation
like (12.4.11).
12.4.4. Conformal Mapping
As was pointed above, integral equation (12.4.11) is valid for non-circulation flow only, i.e.
speed v∞∗ e − iγ is unknown for the given cascade. Due to this conclusion, the cascade problem
can be solved using two venues: one is numerical conformal mapping and finding the neutral
direction b1b2 and speed v∞∗ e − iγ , another is an analytical solution for a flow through a plate
cascade with cavities.
Fig.12.4.4.
Flow through a plate cascade:
(a) flow on the ζ -plane; (b) unit
circle on the parametrical τ -plane.
The first is identical to a conformal

map of a foil-cascade onto a plate-cascade, Fig.12.4.4. Since the circulation around a foil is
zero, the speeds at infinity at both sides of cascade are equal, and equation (12.4.11) is valid.
The inlet function should be directed along neutral line b1b2, but it can take any value. Let
the x-component of inflow speed be unity, vx∗ = 1 , then the y-component, v∗y , is unknown and
∫ v (sτ )G ( z,τ )dsτ + c + xv

∗ ∗
should be calculated from (12.4.11) c y = y, x, y ∈ C . (12.4.12)
C
Adding to the latter equation the Kutta – Joukowski condition and the condition of non-
circulation flow, one obtains as in Sec. 8,1 the closed loop system for the speed distribution
and parameters c and Vy , which can be written in matrix form for using the BEM
500
B ⋅ V* = Y . (12.4.13)
Whence matrix-vector V = B ⋅ Y . The domain of complex potential ζ , Fig.12.4.4a,
∗ −1
represents the ζ -plane with periodic horizontal slits. The slit length is l = 0.5∑ | Vk* | lk . A
k
conformity of slit abscissa ξ and curvilinear blade abscissa S can be found from equation
d ξ = V * dS , or in discrete form ξ k +1 − ξ k = Vk*lk . Inflow speed for non-circulation flow and
its argument are equal to v∞∗ = 1 + (VN∗+ 2 ) 2 and γ = arctan(1/ VN∗ + 2 ) , respectively. The period
of cascade is T ′eiβ ′ = Tv0 ei ( β −γ ) .
12.4.5. Flow through Cascade of Plates

The second part of the method is a flow through cascade of plates, Fig.12.4.4a, with inflow
speed
⎛ dw ⎞ ⎛ dw ⎞ ⎛ dz ⎞ v1 − i (α1 −γ )
v1e − iα1 = ⎜ ⎟ =⎜ ⎟ ⎜ ⎟ = ∗e . (12.4.14)
⎝ d ζ ⎠1 ⎝ dz ⎠1 ⎝ d ζ ⎠1 v∞
The analytical solution of the flow problem is well-known and can be found in many books
[Kochin et al., 1955; Sedov 1950] and others. Further applications are presented briefly here.
The analytical solution can be found by conformal mapping the flow domain on the ζ -plain
to an inside unit circle of the parametric τ-plane, Fig.12.4.4b. The mapping function is
T ′ ⎛ iβ ′ τ − ε ετ − 1 ⎞
ζ =− ⎜ e ln − e − iβ ′ ln , (12.4.15)
2π i ⎝ τ +ε ετ + 1 ⎠⎟
where T’=Tv0, β’=β−γ; unknown parameter ε is found from equation for the split length
T′⎛ 1 − 2ε 2 cos 2 β ′ + ε 4 + 2ε sin β ′ 2ε cos β ′ ⎞
l = ⎜ sin β ′ ln + 2cos β ′ arg tan ⎟
π ⎝⎜ 1 − 2ε 2 cos 2β ′ + ε 4 − 2ε sin β ′ 1 − 2ε 2 cos 2β ′ + ε 4 ⎠⎟
(12.4.16)
( k 2 − k0 )
dw −i τ −e ik0
Speed on the ζ -plane is determined by = iAe 2 , (12.4.17)
dζ τ + eik0
where constants A and k0 are obtained satisfying the condition for inflow speed,
⎛ dw ⎞
⎜ ⎟ = v1e −iα1 . (12.4.18)
⎝ d ζ ⎠ζ =−ε
The circulation around the plate is expressed as
ε T ′(1 − 2ε cos k0 + ε 2 ) 1 − 2ε 2 cos 2β ′ + ε 4
Γ = −4 v1 sin α1 . (12.4.19)
(1 − ε 2 )(1 − ε 4 )
Speed distribution on the plate is found in parametric form
dw
v (t ) = ; ξ (t ) = Re ζ (eit ) . (12.4.20)
d ζ ζ = eit
This is totally enough for determining flow characteristics of foil cascade on the z-plane. The
circulation around the given foil is the same as (12.4.19). The outlet speed is determined by
equations (12.4.5) and (12.4.6). Speed distribution on foil can be calculated by
501
⎡ dw ⎤ dζ
v( x) = ⎢ ⎥ = v[t ( x)]v∗ ( x) , (12.4.21)
⎣ d ζ ⎦ζ =eit dz z∈C
where v is an analytical speed on the flat plate, Eq. (12.4.17); v∗ ( x) is the velocity
distribution on the foil for non-circulation flow and found numerically using BEM-method.
The resultant hydrodynamic force on a blade is
X − iY = i ρΓva e −α a , (12.4.22)
where va e −α a = (v1e −α1 + v2 e −α 2 ) / 2 is the average speed, Fig.12.4.3. It suffices to calculate
numerically a non-cavitating cascade of arbitrary blades.
12.4.6. Cavitating Flow
For cavitating flow, velocity at the cavity boundary should be constant, vv∗ = v0 = const .
The ordinate of cavity is included in velocity V * only, so the following condition at the
boundary corresponding to cavity should be satisfied: v∗ = v0 / v . (12.4.23)
Now, a numerical algorithm of iteration can be used for calculating cavity ordinates and
cavitating number by given cavity length Lc and fixed abscissa of nodal points X k . An
initial cavity boundary for partial cavitation can be used for the foil boundary or any other
curves. Iteration should be is as follows:
Y ( n −1) ⎯⎯⎯⎯(14.4.13)
→ V *( n −1) ⎯⎯⎯⎯
(14.4.21)
→ Vint( n ) ⎯⎯⎯⎯
correcting
→ V ( n) →
(12.4.24)
⎯⎯⎯⎯(14.4.23)
→ V *( n ) ⎯⎯⎯⎯
(14.4.13)
→ Y*( n ) ⎯⎯⎯⎯
correcting
→ Y(n) .
It should to noted that the analytical solution should be used on each step of iteration for
correcting speed at the cavity boundary. Besides, information about the velocity and cavity
boundary is calculated at each step, so an alteration between two steps can be graphically
seen. Coincidence of these curves as well as achieving the desired accuracy of cavitation
number, | σ ( n ) − σ ( n −1) |< ε , can be used to stop iterations.
12.4.7. Modified Integral Equation and Direct Iteration Method

The above-mentioned method of conformal map is quite complicated for numerical study
although the well known analytical solution is used. Simpler is the method based on
modified integral equation. As mentioned above (Sec. 3) integral equation (12.4.11) with
kern function (12.4.10) can be used by another inflow speed. Inasmuch as the inlet and outlet
speeds are peer entities then two integral equations could be considered
∫ v(sτ )Gc ( z, sτ )dsτ + c = Im(v1e 1 z ); z ∈ C ; ∫ v(sτ )Gc ( z, sτ )dsτ + c = Im(v2e 2 z ), z ∈ C .
− iα − iα
C C
Summing the integral equations and dividing by 2 and taking into account the average speed
in equation (12.4.22), one can obtain the required equation as
⎛ e− β i z ⎞
∫C τ
− iα1
v ( s ) G ( z , τ ) dsτ + c + Γ Im ⎜ ⎟ = Im(V1e z ) . (12.4.25)
⎝ 2 T ⎠
The latter equation differs from (12.4.11) by the last term in the left part only, and so both
integral equations can be solved numerically by the same algorithm. The second method
allows calculating speeds at nodal points and the circulation simultaneously without
analytical solution directly on the z-plane.
502

For example, the Joukowski foil a)
cascade (h = 0.1, d = 0.05, c =
0) is considered; angle of attack
α = 10 . The Riabouchinski
model was used for
calculations; the trailing plate Fig.12.4.5.
was supposed to be an element. Cascade of
Some numerical results for Joukowski foils
cascade are presented in Table (h=0.1, d=0.05,
12.4.1 and Figs.12.4.5 - 12.4.7. c=0):
Angle of attack (the angle (a) cavities for b)
between inflow speed and different spacings;
neutral axis of single foil) (b) velocity
α = 10 ; the detachment point as distribution on foil
well as the cavity length are xa at α=10°, β=90°.
=−0.993 and Lc=−0.465. All
computing was done for the
number of elements, N=200. The grid spacing effect on main parameters is shown in Table
12.4.1. The last array corresponds to single foil.
The cavities for three greed spacing, T = 1, 5 and 500, as well as speed distributions
on foil for T = 1 are shown in Fig.12.4.5. The dotted line in Fig.12.4.5b is the velocity
distribution over the foil, and the horizontal dashed line corresponds to the cavity boundary.
The solid line is the velocity distribution for zero flow around the same foil with cavity. One
can see that the dotted line in Fig.12.4.5b has a small peak close to the leading edge though
the Villat condition is fulfilled. If the detachment point moves to the leading edge then the
cavity boundary can intersect the foil. The effect angle β is
seen in Fig.12.4.6.
Fig.12.4.6.
Dependence of cavitation number σ , circulation Γ , and
pressure center xa on angle β .
It is interesting to find that two modes of computing

are possible: one is for interval 0 ≤ β < π + γ , another for
π + γ < β ≤ π . The neutral axis angle inclination γ γ is
calculated by numerical algorithm and in that case of γ ≈
−0.166. Both methods give the same results, but the second
method is easier is commonly used.
12.4.9. Supercavitating Flow
A cavity by full cavitating flow has two separate boundaries on which it is necessary to
satisfy the same value of velocity. For this purpose it is necessary to enter in addition certain
hypothesis at the end of a cavity, for example, to displace of stagnation point on a trailing
plate. The cavities past a flat plates for three values of T = 200, 10 and 7, and past the
Joukowski foil are represented on Fig.12.4.7.
503
a) a)
Fig.12.4.7.
Supercavitating cascade:
(a) cavities past a plate;
(b) cavity past Joukowski foil.
b)
504
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
Ch. 13. Applications of Cavitation Theory to Continuum

Mechanics
In this chapter the methods discussed in the previous chapters are applied to some problems
of continuum mechanics which can be used in various areas of human activity. The
discussed analytical and numerical methods can also be used in mathematics, as
demonstrated below in considering the polyharmonic functions. The problems discussed
below are different in nature but similar in mathematical and numerical methods. Mostly
new problems are considered below.
Sec. 13.1. Impulsive Impact on the Ground

13.1.1. A Short Introduction
Based on practical observations, a solid subjected to high-pressure impact action may
display the properties of a liquid within a restricted area of the impact action. This for
example is the basic assumption of the theory of high-speed penetration of metal. Action of
an explosion or high speed hammering or peenning takes very short time so that only
momentum and impulse are the essential items in the constitutive equations, i.e.
z ∇p ∇p
v = − lim ∫ dt = − . (13.1.1)
x →∞ 0 p p
Since the effect of impact envelops a relatively small region of the solid, its density could be
assumed constant, and then the velocity has potential ϕ = − P / ρ . Besides, velocity should
satisfy the incompressibility condition, so that the potential is a harmonic function. Hence,
the movement of the continuum can be considered as a steady flow of an ideal liquid.
Hydrodynamic approaches in the theory of short-time overload in a continuum have
been developed in many works. An elaborated review of these problems can be found in
[Il’insky & Potashev, 1986] which contains about two hundred references. In this section we
dwell on problems with a similarity to the flow with free boundary. The hydrodynamic
approach for studying cumulative breakdown of solid continuum was developed in
[Lavrent’ev, 1957]. It has been proposed to consider blasting impacts of charges on the
ground to determine the volume thrown out in explosion. Also, the impact-hydrodynamic
theory of explosion momentum has was considered in [Vlasov, 1957], where the whole
continuum was assumed to be a liquid and some approximated rules were proposed to
determine shell-holes; this was called fluidic model (FM). Another approach of impact
hydrodynamics, the so-called solid-liquid model (SLM), was developed in [Kuznetsov,
1977]: the continuum was assumed to be liquid only where the impact velocity exceeds
some value v0 , which marks the shell-holes boundary while the rest of continuum remains
solid. The values of v0 characterize the strength of continuums and can be determined
experimentally only. Some considerations on obtaining this value and comparing theoretical
results with experimental data are given in [Il’insky & Potashev, 1986]. It worth noting that
N.B. Il’insky and his colleagues considered a numbers of problems just in the solid-liquid
state.
Below a simple problem statement is given and a heterogeneous continuum is
considered. The solid-liquid model of the impact hydrodynamics has a similarity with
505
cavitating flow: in both models the speed over a part of the flow domain boundary of a
homogeneous continuum should be constant. There is also a difference: in the solid-liquid
model the speed in the flow domain should be greater than v0 while in cavitating flow the
speed is less than v0 . This similarity of both models makes it possible to consider pulse
impact problems by the same methods as the cavitating flow problems.
13.1.2 Homogeneous Continuum
Let an explosive be distributed uniformly over a half plane of homogeneous ground. This is
identical to a 2D problem on the z-plane with semi-infinite explosive, Fig.13.1.1. Explosive
AB hits the ground’s horizontal surface CAB with pulse P.
In the FM problem statement the whole ground under the x-
axis is assumed to be liquid and potential ϕ vanishes on
CEA ( x < 0 ) and equal to ϕ0 = − P / ρ on ADB ( x ≥ 0 ). This
problem is identical to the vortex flow problem with
P
Fig.13.1.1. Cavity due to complex potential w=− ln z . (13.1.2)
explosion of semi- πρ i
infinite charge. The streamline looks like a circle with constant speed
v = P / πρ r , where r is radius of the circle. Now, if v0
characterizes extreme rigidity of the ground then the shell-hole radius is determined as
r0 = P / π v0 ρ . (13.1.3)
In the SLM statement, the same conditions on the ground surface as well as the
condition | w( z ) |= v0 on the boundary of shell-hole (cavity) should be satisfied, but the
cavity boundary should also be streamline. All these conditions are satisfied in the above-
obtained solution. This results in important conclusions: the shell-hole appears at the end of
distributed explosive only; the characteristic dimensionless parameter is λ = P / ρ v0 r0 .
As for finite length explosives, two versions of SLM statement can be expected: one
if the length is less than 2r0 and the other otherwise. In the latter two shall-holes appear at
each end of explosive, which is determined by the same functions (13.1.2) and (13.1.3).
13.1.3. Distributed Explosive of Finite Length
Consider now a finite length distributed explosive bordering a vertical hard wall,
Fig.13.1.2a. The hard wall cannot be destroyed and the velocity is directed along the wall.
The flow can be extended through the wall (OC) so that the problem is identical to that of
explosion of an isolated band-charge of double length. As before the problem can be solved
analytically by conformal mapping. Let the flow domain on the z-plane map onto first
quadrant of unit circle, Fig.13.1.2b. The same letters denotes corresponding points. Since
potential ϕ(x) is a discontinuity of the first kind at point A then it should have logarithmic
singularity at the corresponding point A(ζ=a), so that its derivative has simple pole. Besides,
the derivative wζ (ζ ) should vanish at point C (ζ = i ) . Continuing analytically over the
whole ζ -plane, we can find that the derivative has simple poles at the symmetric points
( ζ = − a, ± 1/ a ), and simple null at point ( ζ = −i ). Hence,
506
ζ 2 +1
wζ = λ . (13.1.4)
(ζ 2 − a 2 )(ζ 2 − a −2 )
Fig.13.1.2.
Sketch of cavity on the z-plane (a), the parametric ζ -
plane (b).
Factor N is determined from condition ∫

ζ =a
wζ d ζ = −2λ . Now, function (13.1.4) is written as
2λ (1 − a 2 )a
ζ 2 +1
wζ = −i . (13.1.5)
π (ζ − a 2 )(a 2ζ 2 − 1)
2
In logarithmic function on complex velocity ω0 = ln wz , its imaginary part is

piecewise constant on real axis ( Im ω0 = −π i / 2 on AB and Im ω0 = π i / 2 on AO), and the
real part is constant on the free boundary Re ω0 = 0 and in the homogeneous continuum. The
increment of imaginary part is equal to −π i by circumvention around point A, so that it has
logarithmic singularity with a factor of –1. Continuing analytically over the axis and the unit
circle, one can find that function ω0 should have logarithmic singularities at symmetric
points with factor (-1) at point (-a) and factor (1) at points ( ±1/ a ). Beside, function ω0 (ζ )
⎛ a 2ζ 2 − 1 ⎞
at point B (ζ = 1) is equal to −π i / 2 . Hence, ω0 = ln ⎜ i 2 2 ⎟
. (13.1.6)
⎝ ζ −a ⎠
The derivative of transformation from the ζ -plane to the z-plane is expressed as
2λ (1 − a 2 ) a ζ 2 +1
zζ (ζ ) = wζ e −ω0 = − . (13.1.7)
π (a 2ζ 2 − 1) 2
The last equation can be integrated analytically
λ ⎡ ⎛ 1 − aζ ⎞ ζ ⎤
2 ⎢(
1 − a 2 ) ln ⎜ ⎟ + 2a (1 − a )
2
z (ζ ) = − 4
2 2 ⎥
. (13.1.8)
2π a ⎣ ⎝ 1 + aζ ⎠ 1− a ζ ⎦
The last function has only one unknown parameter a. It can be calculated from the condition
for unit length AO: z (a ) = −1 . (13.1.9)
The latter equation has a solution for λ > π only. For λ = π one can consider the limit
approach a → 1 . Substituting a = 1 − ε and ζ = eiβ , and then approaching ε → 0 , one
2ε eiβ
obtains from equation (13.1.8) the asymptotic expression as z (e i β ) ≈ − . Its
1 − e 2iβ + 2ε e 2iβ
right part can have a non-zero value only if variable β is infinitesimal as ε . Then we obtain
from the last equation the free boundary coordinates in parametric form as
2 2β / ε
x( β ) = − , y(β ) = − , ε,β → 0 .
1+ β 2 /ε 2 1+ β 2 /ε 2
(−1 + β 2 / ε 2 ) 2 4β 2 / ε 2
Whence ( x + 1) 2 + y 2 = + =1. Hence, the cavity for
(1 + β 2 / ε 2 ) 2 (1 + β 2 / ε 2 ) 2
λ = π is a half circle of unit radius. The same result follows from equation (13.1.3).
507
Consider another limiting case as factor λ approaches infinity. In this special case
parameter a should approach zero. Denoting m = 2λ a / π one obtains from functions
(13.1.5) – (13.1.7) the following asymptotic expressions
ζ 2 +1 1
wζ = im , wz = −i 2 , zζ = − m(ζ 2 + 1) .
ζ2
ζ
The first function means that the explosive is modeled by a dipole with momentum im. The
⎛ζ 3 ⎞
transformation looks like z = −m ⎜ +ζ ⎟. (13.1.10)
⎝ 3 ⎠
Thus the free boundary of a cavity is determined by
⎛1 ⎞ ⎛1 ⎞
x( β ) = − m ⎜ cos3β + cos β ⎟ , y ( β ) = − m ⎜ sin 3β + sin β ⎟ , β ∈ [0, π / 2]. (13.1.11)
⎝ 3 ⎠ ⎝ 3 ⎠
It is easy to find that the extreme of cavity boundary ordinate is at point ζ = ei π / 4 . If
the value of extreme ordinate is denoted as cavity depth H = − ymin , then in this asymptotic
case it is equal to H = m 2 / 3 . The cavity length is determined as the value of abscises of
point B: L = m 4 / 3 . Ratio H / L = 2 / 2 is independent of momentum m that is the
asymptote of curve H / L = f (λ ), λ → ∞ . Unfortunately, there is no means to find a relation
between factors λ and m . Each of them can be determine experimentally only. Some
considerations on obtaining these parameters are given in [Il’insky & Potashev, 1986]. If
momentum m is found assuming that formulas (13.1.8) and (13.1.10) give the same results
for point B at λ ≥ 10 . Cavity shapes vs. parameter λ are plotted in Fig.13.1.3. It shows that
the free boundary rises up near the hard wall. The same rising ground is observed under
explosion of a charge distributed over a finite length. Fig.13.1.4 shows cavity dimensions as
a function of parameter λ .
Fig.13.1.4.
Dependence of
cavity dimensions
on parameter λ .
Fig.13.1.3. Cavity shapes under

explosive distributed on unit length.
13.1.4. Concentrated Explosion

Consider explosion of a concentrated charge placed at the vertex formed by two free
boundaries of ground, Fig.13.1.5a. The value of angle α ∈ (0, π ) corresponds to uphill,
while α ∈ (π , 2π ) corresponds to a valley. In this case, a concentrated explosive is modeled
as a point dipole, but unlike a hydrodynamic dipole, it is a point of singularity as z −π / α .
Conformal mapping the shell-hole onto quarter of unit disc, Fig.13.1.5b, yields the complex
speed as well as derivative of complex potential in the form
508
1+α / π
dw ⎛ζ +i ⎞ dw Nζ
= −iv0 e −iα ⎜ ⎟ , = 2 . (13.1.12)
dz ⎝ζ −i ⎠ d ζ (ζ + 1) 2
Fig.13.1.5.
Concentrated explosive: (a) shell-hole,
(b) parametric domain on the ζ -plane.
Factor N and singularity intensity

depend on intensity of explosive and its size as
well as on the ground boundary configuration
and other physical parameters
which can be determined
Fig.13.1.6. experimentally only. E.g. for a
Shell-holes from ground with flat free boundary, the
concentrated factor can be expressed by dipole
explosive; curves 1-4
extensity m: N = 8 mv0 . Other
correspond to
α = 2π , 3π / 2, π , π / 2 dependences have been suggested
in [Il’insky & Potashev, 1986].
curves 5 and 6
The shell-hole may be
correspond to
calculated using differential
curvilinear (circular
arc) free boundaries equation zζ = wζ / wz . In
of ground. Fig.13.1.6, curves 1, 2, 4 and 6
stand for angle α = 2π , 3π /2 , π
and π / 2 , respectively; dimensionless coordinates are X = xv0 / N , Y = yv0 / N [Rodionov
and Terentiev, 1985].
13.1.5. Curvilinear Free Boundary of Ground
As in Sec. 4.1, the logarithmic function of complex speed is expanded in a series
⎡ ⎛ ζ + i ⎞2 ⎤ ∞
ω = ln ⎢iv0 ⎜ ⎟ ⎥ + ∑ A2 n −1ζ
2 n −1
. (13.1.13)
⎣⎢ ⎝ ζ − i ⎠ ⎦⎥ n =1
Coefficients A2 n −1 can be found from boundary conditions where the speed is normal to it.
dθ ds dψ π
Thus, =κ , 0 ≤σ ≤ . (13.1.14)
dσ dψ dσ 2
∞
∞
κN ∑ A2 n−1 sin[(2 n −1)σ ]
Whence ∑ (2n − 1) A
n =1
2 n −1 sin[(2n − 1)σ ] = −
4v0 (1 − sin σ ) 2
e n=1 . (13.1.15)
The coefficients can be calculated by collocation and iteration – see dashed lines 5 and 6 in
Fig.13.1.6 showing two dimensionless parameters μ = κ N / v0 = ±2.5 ..
13.1.5. Inhomogeneous Continuum

The problem can also be solved by power series expansion (see Ch. 4). Let characteristic
value ν 0 be a function of depth, i.e. function v0 = c( y ) with c(0) = 1 being given. Consider
509
function Ω(ζ ) = ω (ζ ) − ω0 (ζ ) , where ω = ln wz , ω0 is determined by (13.1.6). As an

analytic function of the whole circle, it can be expanded into Taylor power series
∞
Ω = ∑ Anζ n with real coefficients An ; all coefficients of odd indexes vanish. Besides, the
n=0
sum of coefficient should be zero. Thus the logarithmic function of complex potential is
∞
presented as ω = ∑ A2 n (ζ 2 n − 1) + ω0 . (13.1.16)
n =1
Equation (13.1.5) still remains the derivative of complex potential. If coefficients
A2n are known, the transformation can be found from differential equation dz = wζ e−ω (ζ ) d ζ ,
while unknown parameter a is calculated by the given length of distributed explosive from
equation (13.1.9). The transformation derivative at the cavity boundary with respect to
circular arc ζ = eiβ looks like z ′( β ) = ieiβ zζ (eiβ )e − R ( β ) [cos J ( β ) − i sin J ( β )] , (13.1.17)
∞ ∞
where R ( β ) = ∑ A2 n ( cos 2nβ − 1) , J ( β ) = ∑ A2 n sin 2nβ . Integrating equation (13.1.3)
n =1 n =1
can yield the free boundary.
Let the structural behavior of medium under impact is given analytically as
v0 = c( y ) , then condition V = c( y ) or inverse condition y = g (V ) should be satisfied. It is
preferable to write the latter equation in differential form y ′ = gV (V ) V ′ , where subscript V
means that the derivative is with respect to variable V while the apostrophe denotes
derivative with respect to angle β . Since the real part of ω0 on the free boundary is equal to
unity, hence V ( β ) = exp R( β ) , V ′( β ) = V ( β ) R′( β ) . Now, the coefficients can be calculated
π /2
2 y ′( β )
iteratively: A2 n = − ∫
π n 0 gV (V )V ( β )
sin 2nβ d β (13.1.18)
For example, the negative sign of parameter α in function

v0 = e −α y with inverse function g (V ) = − ln V / α denotes that
the strength of ground weakens and vice versa for the
positive sign. The cavity shapes due to an impact with pulse
λ = 5 per unit length of the ground surface are plotted in
Fig.13.1.7. It is seen that the negative sign of parameter α
affects the shape considerably.
Fig.13.1.7.
Cavity shapes due to impact over a unit length.
510
Sec. 13.2. Mathematical Modeling of Underground Sprinkling System*

13.2.1. Brief Notes on Fluid Flow in Porous Media
Underground sprinkling, unlike surface watering, has certain advantages: it delivers easily
controlled quantity water and nutrients directly to the roots of plants, is less vulnerable to
surface actions, and saves water. Such a system can also serve as drainage for removing
excessive moisture from the ground.
The real process of moisture supply in soil has a non-stationary character, depends
on many physical parameters, and is described by nonlinear differential equations
[Polubarinova-Kochina, 1977], which can in many cases be solved only numerically. The
simplest model of fluid flow in a porous medium is a stationary potential flow for which
p
Darcy’s Law is valid: v = − k ∇ h, h = + y, (13.2.1)
ρg
where v is the velocity vector; p is the pressure; k is the soil permeability; ρ is the fluid
density; g is the gravity acceleration; y is the vertical coordinate (positive upward); and h
is the piezometric head. For homogeneous isotropic soil, the permeability is constant
( k = const ). Then satisfying the incompressibility condition, one can make sure that the
piezometric head is a harmonic function. In a two-dimensional case, the flow is described by
complex potential w = ϕ + iψ with ϕ = − k h .
Risenkampf [1940] was the first who formulated and solved the underground
sprinkling problem for a single source. This and others problems, as well as respective
bibliography are presented in [Polubarinova-Kochina, 1977]. To extend the mathematical
modeling of the hydrodynamics to a range of applied problems, we consider in this section
one and two rows of sources in the ground. The soil is assumed to be homogeneous and
isotropic. The seepage and evaporation are ignored.
13.2.2. One-Row System of Perforated Pipes
Let pipes of radius R are arranged in one horizontal
row of spacing L between them at a depth of H . The
water comes in the pipe and is filtered through the
perforated surface to the ground. The water pressure at
the ground level is p0 and outside watered domain it is
equal to the atmospheric pressure pa=const.
Fig.13.2.1.
Underground flow: (a) general schematics;
(b) the w1- plane, (c) the w2- plane, (d) the ζ-plane.
Consider a steady flow in a gravity field. Generally, a part of liquid leaks out on the ground
surface and the rest flows, due to gravity, down to infinity. The total flow is periodic of
period L. A half of the periodic flow is shown in Fig.13.2.1a. The underground pipe can be
modeled first as a source at point C . The water leaks through section OA1 of length l/2 and
*
Although the effect discussed in this section does not involve cavitation, the reader will find that a free
boundary is involved in the problem, and the mathematics underlying the solution method are somewhat similar
to those presented elsewhere in this book. This section is therefore included to provide a bridge between the
hydrodynamics of cavitating flows and related topics of continuum mechanics. The same applies to Sec.13.3.
511
vaporizes into air. Curve A1A2 is the free boundary, at which the pressure is constant (pa).
Straight semi-infinite line A2 B is the contact line with the flow from the next pipe. Also, the
value problem statement for harmonic function h(x,y) is:
h = 0 on OA1 and A1 A2 ; ∂h / ∂n = 0 on BO and A1 A2 B . (13.2.2)
Besides, it should have a simple pole at source point C(−Hi).
The water-saturated domain can be easy obtained using Joukowski’s method of
considering two analytic functions. Consider complex function w1 = ϕ1 + iψ 1 with ϕ1 = − kh
and another function w2 = w1 − ikz . Then the complex variable
z = i ( w2 − w1 ) / k (13.2.3)
allows to calculate the water-saturated domain. Variation ranges of functions w1 and w2 is
shown in Figs.13.2.1b & c, respectively. Mapping these domains onto first quadrant of the
dwk Q ζ 2 + ak2
auxiliary ζ -plane, one finds derivative = , k = 1, 2 , and functions
d ζ π 1 + ak2 (ζ 2 − 1)
Q ⎛ 1 ⎞
wk = ⎜ ln f k (ζ ) − ln g k (ζ ) ⎟ , k = 1, 2 , (13.2.4)
2π ⎜ 1 + ak2 ⎟
⎝ ⎠
ζ 2 + ak2 − ζ 1 + ak2 ζ 2 + ak2 − ζ
where f k (ζ ) = , g k (ζ ) = . (13.2.5)
ζ 2 + ak2 + ζ 1 + ak2 ζ 2 + ak2 + ζ
Now, all geometric parameters can be determined from equations (13.2.3) – (13.2.5).
Substituting ζ = iη (a1 ≤ η ≤ a2 ) to functions (13.2.4) and (13.2.5), one can obtain from
(13.2.3) free boundary A1A2 in parametric form as
Q ⎛ η 1 + a1 + η − a1 η + η 2 − a12 ⎞
2 2 2
1
y ( x) = ⎜ ln − ln ⎟, (13.2.6)
πk ⎜ a η 2
+ 1 1 + a 2 a ⎟
⎝ 1 1 1 ⎠
x=
Q ⎜
⎛
arctan
η 1 + a22
−
1
arctan
η
+
(
π 1 − 1 + a12 ⎞ )
⎟
πk ⎜ ⎜ ⎟ . (13.2.7)
a2 − η
2 2
1 + a22
a2 − η
2 2
2 1 + a1
2
⎟
⎝ ⎠
Spacing between two sources and the source submergence depth are equal to double abscissa
of point A2 and to ordinate of point C, respectively:
Q⎛ 1 1 ⎞
L = 2 x( a2 ) = ⎜ − ⎟, (13.2.8)
k ⎜ 1 + a12 1 + a 2 ⎟
⎝ 2 ⎠
Q ⎛ a1 1 + a2 ⎞
2 2
( −1) 2 am
H = iz (1) = ⎜ ln +∑ ln ⎟ . (13.2.9)
π k ⎜ a2 1 + a12 m =1 1 + am2 1 + 1 + am2 ⎟
⎝ ⎠
Equations (13.2.6) and (13.2.7) allow to compute unknown parameters a1 and a2 for
given dimensionless values H/L and Cq=Q/Lk (the reduced coefficient of consumption).
Instead of coefficient Cq one can use pressure drop coefficient Cp=(p0−pa)/ρgL. The relation
between these coefficients can be obtained asymptotically following Risenkampf approach.
Let the radius of a pipe (R) be much smaller than the depth H and the length L. Then the pipe
512
section corresponds to a circle of small radius on the ζ -plane with the center at point C (1) .
Due to transformation (13.2.3), the radius of that circle is expressed as
2
(1 + a12 )(1 + a22 ) Rk ⎛ Rk ⎞
| ζ − 1|= 2π + O⎜ ⎟ . Since the asymptotic expansion of equation
a2 − a1
2 2
Q ⎝ Q ⎠
Q ⎛ a12 (1 − ζ ) 2 1 + a12 + 1 ⎞
(13.2.4) nearby the point C is w1 = ⎜ ln + ln + O (| ζ − 1|) ⎟.
2π ⎜ 2(1 + a12 ) 1 + a12 a1 ⎟
⎝ ⎠
Hence, taking into account quantity ϕ1 = − kh , yields the asymptotic expression for the
H Cq ⎛ π a12 (1 + a12 ) R 2 1 + 1 + a12 ⎞
pressure drop coefficient: C p = − ⎜ ln 2 + ln ⎟ . (13.2.10)
L 2π ⎜ (a2 − a12 )Cq L 1 + a12 a1 ⎟
⎝ ⎠
13.2.3. Range of Solvability of the Problem
The set of non-linear equations has a solution for some ranges of the given variables only.
To determine the solution existence domain, consider extreme cases. Let parameter a2
approaches infinity. In that case point A2 moves away to infinity, and the problem is
reduced to the one of a single pipe. Equations (13.2.8), (13.2.9) and (13.2.10) became:
H Cq ⎛ Cq − 1 1 Cq − 1 ⎞ C L
a1 = Cq − 1; = ⎜ ln − ln ⎟ ; C p = q ln . (13.2.11)
L π ⎜ Cq Cq 1 + Cq ⎟ 2π π RCq
⎝ ⎠
H ln 2 1 L
If Cq = 1 then a1 = 0 , and = , Cp = ln . (13.2.12)
L π 2π π R
This case corresponds to the water-saturated formation without dripping on the ground
surface, which was considered in [Risenkampf, 1940].
Another extreme case is if parameter a1 → 0 but a2 < ∞ . The water does not ooze
out through ground surface but the water-saturated formations are contiguous to each other
along vertical straight line A2 B (Fig.13.2.1a). Eqs (13.2.8) – (13.2.10) are transformed to
2Cq − 1 H Cq ⎛ 2 1 + a2 1 + a22 − 1 ⎞
2
1
a2 = ; = ⎜ ln + ln ⎟;
Cq − 1 L π ⎜ a2 1 + a 2 a ⎟
⎝ 2 2 ⎠ (13.2.13)
H Cq 4π (1 + a2 ) R 2
Cp = − ln .
L 2π a22Cq L
Equations (13.2.11) and (13.2.13) determine the solution existence domains plotted in
Fig.13.2.2; the low boundaries
are calculated by equations
(13.2.11) while the upper ones
by equations (13.2.13).
Fig.13.2.2. Existence domains

of solution.
513
The free boundaries for Cq = 2 and three positions of the

source are shown in Fig.13.2.3. The dotted line
corresponds to first extreme case when each pipe of a
system functions separately. The dashed line corresponds
to the second extreme case when the free boundary touches
the ground boundary at single point O only.
Fig.13.2.3.
Free lines for Cq = 2 .
13.2.4. Fluid Filtration of Forcing and Drainage Pipes

Pumping in excessive water can have a negative effect on vitality of plants. Therefore both
pipes can be buried together with the forcing pipe beneath the drainage pipe. This
arrangement can be modeling as system of source and sink. Generally, intensities of source
and sink can be different; a part of liquid can flow past drainage pipe, Fig.13.2.4a.
The value problem of right symmetric part of the filtration flow is as follows:
h = 0 on O1 BAO2 ; ∂h / ∂n = 0 on O1C1C2O2 (13.2.14)
Besides, intensities of the source at point C1 (−iH1 ) and the sink at point C2 (−iH 2 ) are given
as Q and q respectively. The domains of complex potential w1 and subsidiary variable w2
are shown in Fig.13.2.4b,c. Mapping these domains onto first quadrant of the auxiliary ζ -
plane, Fig.13.2.4d, one can find the transformations as
Q ⎛ ζ2⎞ q Q−q ⎛ ζ2 ⎞
w1 = ln ⎜1 − 2 ⎟ − ln(ζ 2 − 1) − ln ⎜1 + 2 ⎟ ; (13.2.15)
2π ⎝ c ⎠ 2π 2π ⎝ b ⎠
Q c −ζ q 1−ζ
w2 = ln − ln . (13.2.16)
2π c + ζ 2π 1 + ζ
Fig.13.2.4. The underground source and sink (a);

domains of variables: w1 (b), w2 (c) and ζ (d).
The derivative of these functions at point A(ia ) should vanish, thus two equations appear:
Q q Q−q
− + 2 = 0; (13.2.17)
c + a 1+ a
2 2 2
b − a2
514
Qc q
− = 0. (13.2.18)
c + a 1 + a2
22
Transformation from the z-plane to the ζ -plane can be found as

Q ⎛ ζ ⎞ Q−q ⎛ ζ ⎞ q
2
z= ln 1 + − ln ⎜ 1 + ⎟− ln(1 + ζ ) . (13.2.19)
π ki ⎜⎝ c ⎟
⎠ 2π ki ⎝ b 2 ⎠ π ki
Using equation (13.2.19), one can find all geometric characteristics of the flow and
determine unknown parameters a, b and c. Namely, the distances of source and sink are
equal to H1 = iz (c) and H 2 = iz (1) respectively, the distance between source and sink is
Q (1 + c) 2 (b 2 + c 2 ) q (1 + c) 2 (b 2 + 1)
expressed as H= ln + ln . (13.2.20)
2π k 4c (1 + b )
2 2
2π k 4c 2 (c 2 + b 2 )
If intensities Q and q as well as distance H are given, then parameters a, b and c can
be calculated from equations (13.2.17), (13.2.18) and (13.2.20). Approximate relationships
between small radiuses R1 & R2 of pipes with centers at C1 & C2 and distances H1 & H 2
as well as corresponding pressures p1 and p1 can be found in the same way [Terentiev,
1989]. Asymptotic expansions over a small neighborhood at point C1 (ζ = c) looks like
z = − H1i − iA1 (ζ − c) + O(| ζ − c |2 ) ; (13.2.21)
Q
w1 = ln(c − ζ ) + B1 + O(| ζ − c |) , (13.2.22)
2π
Q q (Q − q )c Q 2 q Q − q ⎛ c2 ⎞
where A1 = − − ; B = ln − ln(1 + c 2
) − ln ⎜ 1 + 2 ⎟ .
2π kc π k (1 + c) π k (b 2 + c 2 ) 2π c 2π 2π
1
⎝ b ⎠
Similarly, at point C2 (ζ = 1) : z = − H 2i − iAr (ζ − 1) + O(| ζ − 1|2 ) ; (13.2.23)
q
w1 = − ln(1 − ζ ) + B2 + O (| ζ − 1|) , (13.2.24)
2π
Q q (Q − q ) c Q ⎛ 1⎞ q Q−q ⎛ 1 ⎞
where A2 = − − ; B2 = ln ⎜1 − 2 ⎟ − ln 2 − ln ⎜ 1 + 2 ⎟ .
2π k (1 + c) π k π k (b + 1)
2
2π ⎝ c ⎠ 2π 2π ⎝ b ⎠
The small radiuses are equal asymptotically to R1 ≈ A1 | ζ − c | and R2 ≈ A2 | ζ − 1| .
Eliminating real parts of equations (13.2.22) and (13.2.24), and taking into account equality
ϕ = − kh yields requite equations:
p1 B Q R p2 B q R
= H1 − 1 − ln 1 , = H2 − 2 + ln 2 . (13.2.25)
ρg k 2π k A1 ρg k 2π k A2
Consider the source and sink of equal intensity ( q = Q ). Approaching q → Q in
equation (13.2.18), we obtain a 2 → c . Parameter c should differ from 1, otherwise functions
w1 and w2 vanish on the whole domain. Equation (13.2.17) could be satisfied only if b = a .
This implies that the domain of water-saturated formation becomes in this case symmetrical
relative to both vertical and horizontal strait lines. Points A and B join one point on the free
boundary with ordinate c . Parameter c is determined from condition (13.2.20):
515
( 2e − 1) − 1 .
2
c = 2eπ kH / Q − 1 − π kH / Q
(13.2.26)
The free boundary is obtained parametrically by
Q ⎛ η ⎞ Q c 1+η2
x= ⎜ arctan − arctanη ⎟ , y= ln . (13.2.27)
πk⎝ c ⎠ πk c2 + η 2
It follows from equation (13.2.26) that parameter c ≈ e −π Hk / Q → 0 as the distance approaches
infinity. Equation (13.2.27) is transformed to the well known function for free boundary of
single source [Risenkampf, 1940]:
Q π kx
y= ln cos . (13.2.28)
2π k Q
Fig.13.2.5:
Watered domain boundary for the source-sink system.
Free boundaries of the domain of water-saturated

formation around a source and sink are plotted in Fig.13.2.5,
where: x = π kx / Q, y = π ky / Q, H = π kH / Q . Risenkampf’s
solution can be obtained also from (13.2.15) and (13.2.16) by
substituting ζ = cζ ′ and approaching parameter c to zero. As a
result of asymptotic action it follows:
Q Q 1−ζ ′ Q
w1 = ln(1 − ζ ′2 ), w2 = ln , z = −i ln(1 + ζ ′) . (13.2.29)
2π 2π 1 + ζ ′ πk
13.2.5. Deformation of Free Boundary in Time
For practical watering it is worth to know deformations of the watered domain free boundary
and necessary volume. To find it, an unsteady problem for a single source or perforated pipe
should be solved. However, for real filtration factor the water penetration to the ground
happens very slowly so that the time derivative components of the general equation can be
ignored. In that case one can consider, at leas approximately, a steady flow of a single source
and calculate the free boundary deformation from ordinary differential equation of moving
water particles. If the complex potential is known, w = w( z ) , then the complex velocity is
determined as derivative u − iv = wz ( z ) . The velocity of water particles differs from that
speed by porosity factor σ , thus dz / dt = σ wz . This is the ordinary differential equation for
single particle movement. Let the known parametric solution be w1 (ζ ) and z (ζ ) , then the
dζ d ζ dw1 d ζ
equality =σ (13.2.30)
dt dz d ζ dz
yields a set of ordinary differential equations for functions ξ (t ) and η (t ) ( ζ = ξ + iη ). The
desired contour is calculated by function z (ζ ) .
In the special case for functions (13.2.29) it is easy to write differential equations
directly for coordinates in the physical z-plane that in dimensionless variables x , y and
T = π k 2σ t / Q look like
516
dx e − y sin x
= −2 y ,
dT e − 4e − y sin x + 4
(13.2.31)
dy e −2 y − 3e − y sin x + 2
= − −2 y .
dT e − 4e − y sin x + 4
Fig.13.2.6: Approximated watered domain for different times.
Initial conditions can be satisfied for the pipe of radius R = π kR / Q

and center at point y0 = −π k h / Q = − ln 2 : x 2 + ( y − y0 ) 2 = R 2 .
Indeed, the pipe radius R should not be less than distance h. The
curves for R = 0.1h at different times are plotted in Fig.13.2.6.
Sec. 13.3. Series Expansion for Flows of Limited Reynolds Number*

13.3.1. Notes on a Flow at Small Reynolds Numbers
Only few problems could be solved analytically using the complete Stokes equations for a
flow of viscous fluid. Although many problems can be calculated numerically, the problems
associated with viscous fluid have been thoroughly studied. Therefore, it is worth
considering some approximations that would reduce the equations and help obtaining
analytical or numerical solutions. The simplest approximation has been offered by Stokes
[Stokes, 1851], but his approach is invalid for unbounded domain. To overcome Stokes
paradox, Oseen [1910] linearized convective items and transformed Stokes equations to
parabolic differential equations. This approach was used in [Kaplun, 1957; Proudman &
Pearson, 1957] for calculating the resistance of a sphere and a cylinder by series expansion
on small Reynolds numbers. The same results can be received by perturbation methods [Van
Dyke, 1964]. The main results have been obtained by auxiliary simplified assumptions,
which are valid for special configuration of domains [Petrov, 1883; Sommerfeld, 1904].
Such problems are of great importance in the engineering industry in connection with
lubrication of bearing, shock-absorber and oil motor means.
Below a moving of bodies in bounded domain of arbitrary configuration using
Stokes approach is considered. As mathematical tools, the conformal map and expansion in
series (see Ch. 4) are used. The main goal is to find numerical algorithm of computing
unknown coefficients [Terentiev & Terentiev, 2002]. The problem is considered as an
example for applying the above-mentioned method only.
13.3.2. The Governing Equations
If inertial forces in incompressible fluid are insignificant, the flow is determined by
simplified Stokes equations ∇p = μ∇ 2 v, ∇ ⋅ v = 0 . (13.3.1)
Whence, ∇ 2 p = 0 or ∇ 2 (∇ × v) = 0 . In the planar case, the second equation in (13.3.1)
means the existence of stream function ψ which determines speed,
u = ∂ψ / ∂y, v = −∂ψ / ∂x , and so (∇ × v ) z = ∇ ψ . Thus the stream function should satisfy
2
bi-harmonic function ∇ 4ψ = 0 . (13.3.2)
*
See footnote to the title of Sec. 13.2 on page 511.
517
The bi-harmonic function cap be expressed as a sum of two harmonic functions [Lavrent’ev
& Shabat, 1965], ψ = Re[ z ϕ ( z ) + χ ( z )], z = x + iy . (13.3.3)
The complex speed and vorticity with pressure are obtained respectively as
u + iv = −i[ϕ ( z ) + zϕ ′( z ) + χ ′( z )] (13.3.4)
and ω = −4 Re ϕ ′( z ), p = −4μ Im ϕ ′( z ) . (13.3.5)
The latter equations show that derivative ϕ’(z) should be a single-value function while
function ϕ’(z), as well as χ’(z), can be multi-valued, but its sum ϕ(z) +χ(z) should be single-
valued on a closed graph Ck , and so the single-valued condition is
∫ ϕ ′( z )dz + ∫ χ ′′( z )dz = 0 .

Ck Ck
(13.3.6)
13.3.3. Boundary Conditions

A domain should be bounded and univalent but can by multi-connected. Adhesion
conditions should be satisfied at the boundary, i.e. the speeds of fluid and boundary should
∂ψ ∂ψ
coincide = −V , = U on C, (13.3.7)
∂x ∂y
where U and V are speed components of the boundary.
13.3.4. Hydrodynamic Force and Torque
Components of a speed deformation tensor are calculated by
∂u ∂v 1 ⎛ ∂u ∂v ⎞
ε xx = = − Im F ( z ), ε yy = = Im F ( z ), ε xy = ⎜ + ⎟ = − Re F ( z ),
∂x ∂y 2 ⎝ ∂y ∂x ⎠ (13.3.8)
F ( z ) = z ϕ ′′( z ) + χ ′′′( z ).
The strain tensor for incompressible fluid is determined by well-known dependences [Sedov,
1970] σ xx = − p + 2με xx , σ yy = − p + 2με yy , σ xy = 2με xy . (13.3.9)
Let the positive direction along boundary Сk be counter-clockwise, i.e. the fluid remains on
the right by moving along the boundary. If the normal is directed towards the fluid then its
complex representation is n = −idz / ds . Therefore, an exertion on elementary length of the
contour is dσ ( n ) dz = ipdz + 2 μ[ε xx nx + ε xy n y + i (ε xy nx + ε yy ny )]ds . Taking into account eq.
(13.3.6) one can write dσ ( n ) dz = −4μ i Im ϕ ′( z )dz + 2μ[ zϕ ′′( z ) + χ ′′( z )]dz . (13.3.10)
Hence, the resultant force and toque are
⎛ ⎞
X + iY = −4μ i ∫ Im ϕ ′( z )dz + 2μ ⎜ ∫ z ϕ ′′( z )dz + ∫ χ ′′( z )dz ⎟ ; (13.3.11)
⎜C ⎟
Ck ⎝ k Ck ⎠
⎛ ⎞
M = −4μ Re ∫ Im ϕ ′( z ) zdz − 2μ Im ⎜ ∫ zz ϕ ′′( z )dz + ∫ χ ′′( z ) zdz ⎟ . (13.3.12)
⎜C ⎟
Ck ⎝ k Ck ⎠
13.3.4. A Cylinder Moving in a Cylinder
The Loran expansion could also be applied for studying by collocation the movement of a
cylinder in a closed domain of liquid with limited high viscosity. Let internal and external
cylinders have any sections, Fig.13.3.1a. The speed of points on the moving cylinder is
518
given as U c + iVc = U 0 + iV0 + iΩ z , (13.3.13)

while the speed of outer cylinder is zero.
Fig.13.3.1.
Accordance between the domain on z-plane and the
parametric ring.
If the flow domain between these cylinders is conformal mapped onto the ring with unit
inner radius and the outer radius R, Fig.13.3.1b, the problem will be reduced to determining
two analytic functions ϕ (ζ ) and χ ′(ζ ) which also can be found as Loran series, but with
∞ ∞
complex coefficients: ϕ (ζ ) = ∑
n =−∞
anζ n + a ln ζ , χ ′(ζ ) = ∑bζ
n =−∞
n
n
+ a ln ζ . (13.3.14)
Complex velocity (13.3.4) becomes:

⎛ ∞ ⎛ b ⎞ z (ζ ) ⎛ ∞ a a ⎞ ⎞
u + iv = −i ⎜ ∑ ⎜ an + n2 n ⎟ ζ n + ⎜ ∑ (1 − n) 1−2nn ζ n + ζ ⎟ + 2 Re a ln | ζ | ⎟ .(13.3.15)
⎜ n =−∞ |ζ | ⎠ z ′(ζ ) ⎝ n =−∞ |ζ | |ζ | ⎠
2 ⎟
⎝ ⎝ ⎠
Transformation z (ζ ) and its derivative as Loran series yields
∞ ∞
ζn
z (ζ ) = ∑ cnζ n ,
n =−∞
z ′(ζ ) = ∑ (1 − n)c1−n
n =−∞ | ζ |2 n
(13.3.16)
and substituting condition (13.3.13) and functions (13.3.16) in the left part of speed
(13.3.15), yields the following equality:
∞
⎛ c1− n c ⎞
∑ ⎜ (iU 0 + V0 )(1 − n)
n =−∞ ⎝ |ζ | 2n
− Ωn n2 n ⎟ ζ n =
|ζ | ⎠
∞ ∞
c1+ m − n ⎛ b ⎞ ∞ ∞
c a
= ∑ ∑ (1 − m − n) | ζ | a + m2 m ⎟ ζ n + ∑ ∑ (n − m)(1 − m) n − m 21m− m ζ n + (13.3.17)
2( n − m ) ⎜ m
|ζ | ⎠ |ζ |
n =−∞ m =−∞ ⎝ n =−∞ m =−∞
∞ ∞
a c
+ ∑ cn −1 ζ n + 2 Re a ln | ζ | ∑ (1 − n) 1− n2 n ζ n .
n =−∞ |ζ |2n
n =−∞ |ζ |
iθ inθ
Substituting ζ = e in (13.3.17) and equating factors at e , we obtain an infinite system of
linear equations relative unknown coefficients an , bn and a . This system should be added by
a system of equations describing performance of boundary conditions on the external circle.
For this purpose it is necessary to substitute ζ = R eiθ in the right part of (13.3.17) and
equate it to zero. For two circular cylinders, one can obtain from these equations [Terentiev
& Terentiev, 2002] the well known analitic expressions in [Petrov, 1883; Joukowski &
Chaplygin, 1904; Kaplun, 1957]. But for two arbitrary cylinders, the difficulties appear
connected with conformal mapping onto a ring, with expansion in the Loran series of
transformation and with solving an infinite system of linear equations.
13.3.5. Computing Coefficients Collocation
The collocation method provides the performance of boundary conditions only in its separate
points. Unfortunately, so far the method has no strict mathematical substantiation; however,
as could see in previous section, it is simple for numerical realization and can be applied
efficiently for complex boundary conditions; it gives high accuracy and is widely used in the
519
hydrodynamics. The collocation method algorithm for forward movement and separately for
rotation was given in [Terentiev & Terentiev, 2002]. Below the collocation is applied to
calculating coefficients of series (13.3.14) for an arbitrary movement of a solid cylinder in
an fixed outer cylinder.
Keeping 2N+1 items in series (13.3.14) and numbering their coefficients from 1 up
2 N +1
to 2N+1, we can write: ϕ (ζ ) = ∑ Aζ
n =1
n
n − N −1
+ AN +1 ln ζ ,
2 N +1
AN +1
ϕ ′(ζ ) = ∑ A (n − N − 1)ζ
n =1
n
n− N −2
+
ζ
, (13.3.18)
2 N +1
χ ′(ζ ) = ∑Bζ
n= N
n
n −3 N − 2
+ AN +1 ln ζ .
It should be noted that the absolute term in first series is assumed to vanish. This term can be
included to the absolute term of the series of function χ ′(ζ ) .The complex velocity (13.3.4)
⎧ ⎫
could be written as u + iv = −i ⎨∑ An f n (ζ ) + ∑ An g n (ζ ) + ∑ Bn pn (ζ ) ⎬ , (13.3.19)
⎩ n n n ⎭
⎧ z (ζ )
⎪ z ′(ζ ) ( n − N )ζ
k − N −1
⎧ζ n − N , n ≠ N, , n ≠ N,
where f n (ζ ) = ⎨ g n (ζ ) = ⎨ pn (ζ ) = ζ n − N .
⎩ 2ln | ζ |, n = N , ⎪ z (ζ ) / z ′(ζ )ζ ,
⎩ n = N,
The total number of unknown real and imaginary parts of coefficients is equal to 8 N + 4 .
Numbering now these coefficients from 1 up to 8 N + 4 as following:
⎧ Re Ak , k = 1, 2 N + 1,
⎪
⎪ Im Ak − 2 N −1 , k = 2 N + 2, 4 N + 2,
Ck = ⎨ , (13.3.20)
⎪ Re B k −4 N −2 , k = 4 N + 3, 6 N + 3,
⎪
⎩ Im Bk − 6 N −3 , k = 6 N + 4, 8 N + 4
we may present the real and imaginary parts of complex speed (13.3.19) in the sum as
8N +4 8N +4
u= ∑ Ck Fk (ζ ), v = ∑ Ck Gk (ζ ) ,
k =1 k =1
(13.3.21)
where Fk (ζ ) and Gk (ζ ) are expressed via real and imaginary parts of the known functions
⎧ Im f k (ζ ) + Im g k (ζ ), k = 1, 2 N + 1
⎪
⎪ Re f k − 2 N −1 (ζ ) − Re g k − 2 N −1 (ζ ), k = 2 N + 2, 4 N + 2
in (13.3.19): Fk (ζ ) = ⎨ , (13.3.22)
⎪ − Im pk − 4 N − 2 (ζ ), k = 4 N + 3, 6 N + 3
⎪
⎩ − Re pk − 6 N −3 (ζ ), k = 6 N + 4, 8 N + 4
⎧ − Re f k (ζ ) − Re g k (ζ ), k = 1, 2 N + 1
⎪
⎪ Im f k − 2 N −1 (ζ ) − Im g k − 2 N −1 (ζ ), k = 2 N + 2, 4 N + 2
Gk (ζ ) = ⎨ . (13.3.23)
⎪ − Re pk − 4 N − 2 (ζ ), k = 4 N + 3, 6 N + 3
⎪
⎩ Im pk − 6 N −3 (ζ ), k = 6 N + 4, 8 N + 4
520
Dividing each of inner and outer circles of the ring into 2 N + 1 parts as Tm = 2π m /(2 N + 1) ,
we have 2 N + 1 points ζ m = eiTm on the inner circle and another 2 N + 1 points ζ m′ = R eiTm
on the outer circle. If the speed components are numbered from 1 up to 8N+4 as following:
⎧u (ζ m ), m = 1, 2 N + 1
⎪
⎪v(ζ m − 2 N −1 ), m = 2 N + 2, 4 N + 2
Vm = ⎨ (13.3.24)
⎪u (ζ m′ − 4 N + 3 ), m = 4 N + 3, 6 N + 3
⎪
⎩v(ζ m′ − 4 N + 3 ) m = 6 N + 4, 8 N + 4
and if the 8 N + 4 × 8 N + 4 matrix M is represented by array cell as
⎧ Fk (ζ m ), m = 1, 2 N + 1
⎪
⎪Gk (ζ m − 2 N −1 ), m = 2 N + 2, 4 N + 2
M mk = ⎨ (13.3.25)
⎪ Fk (ζ m′ − 4 N + 3 ), m = 4 N + 3, 6 N + 3
⎪
⎩Gk (ζ m′ − 4 N + 3 ) m = 6 N + 4, 8 N + 4
then velocity (13.3.21) can be written in matrix form V = MC . (13.3.26)
Due to the adhesion condition the fluid velocity should be given by (13.3.13), so that
(13.3.26) is the equation for unknown coefficients. The matrix equation’s solution is
C = M −1V . (13.3.27)
The coefficients of series (13.3.18) are calculated from array cell of vector-matrix C:
Ak = Ck + iCk + 2 N +1 , Bk = Ck + 4 N + 2 + iCk + 6 N + 3 , k = 1, 2 N + 1 . (13.3.28)
The procedure is valid for cylinders of arbitrary sections only if transformation z (ζ ) is
given or could be calculated.
Eccentrical cylinders. Let the inner cylinder’s axis is displaced over distance d in
the direction of its movement relative to the other cylinder’s axis. Here in the cross section
the origin coincides with the small circle’s center, then the outer circle’s center will be at –d,
Fig.13.3.1a. The flow domain between two circles can be conformal mapped onto the ring
using homographic transformation. Let the inner circle’s radius be unity and the outer
circle’s R. Since the origin ζ=0 is symmetric to infinity ζ →∝ relative to the concentric
circles then they are mapped into points z=c1 and z=c2 which should be symmetric relative
to both circles C1 and C2 simultaneously. Thus the following symmetry conditions should
be satisfied: c1c2 = R12 , (c1 = d )(c2 = d ) = R22 . (13.3.29)
Whence, coordinates c1 and c2 are determined from quadratic equation t 2 − 2bt + R12 = 0 :
R22 − R12 − d 2
c1 = b − b 2 − R12 , c2 = b + b 2 − R12 , b = . (13.3.30)
2d
( R − c )( z − c1 )
Hence, the transformation may be written in the form: ζ ( z ) = 1 2
( R1 − c1 )( z − c2 )
Since the ring’s inner radius is unity ζ(1)=1 and the outer is ζ(R2−d)=R, then the latter
c ( R − c )ζ − c1 ( R1 − c2 )
function could be inversed as z (ζ ) = 2 1 1 . (13.3.31)
( R1 − c1 )ζ − ( R1 − c2 )
521
(c1 − c2 )( R1 − c1 )( R1 − c2 )
Its derivative is z ′(ζ ) = . (13.3.32)
[( R1 − c1 )ζ − ( R1 − c2 )]2
Now, all elements of matrix M and vector V can be calculated and necessary data found.
For example, let’s consider rotation of the inner cylinder with angular velocity Ω = 1 . The
exact solution was obtained by Joukowski & Chaplygin [1904]. Here are their expressions
for the vertical force and the moment:
R22 + R12 − 2 R1 R2 cosh(η1 − η0 )
Lex = 4π ,
(η1 − η0 )( R22 + R12 ) − 2 R1 R2 sinh(η1 − η0 )
(13.3.33)
(η − η ) R 2 coth(η1 − η0 ) − 2 R1 R2 cosh(η1 − η0 ) + 1
M ex = −4π 1 0 2 ,
(η1 − η0 )( R22 + R12 ) − 2 R1 R2 sinh(η1 − η0 )
R22 + d 2 − R12 R 2 − d 2 − R12
where η0 = arccos h , η1 = arccos h 2
2dR2 2dR1
Sommerfeld’s approximations [Sommerfeld, 1904] are expressed as following:
12π dR1
Lap ≈ ,
(2( R2 − R1 ) + d 2 ) ( R2 − R1 ) 2 − d 2
2
(13.3.34)
4π (( R2 − R1 ) 2 + 2d 2 )
M ap ≈ − .
(2( R2 − R1 ) 2 + d 2 ) ( R2 − R1 ) 2 − d 2
A comparison of the results calculated in collocation manner [Terentiev, 2000] for different
numbers of collocation points with exact data is given in the Table 13.3.1 where
Sommerfeld’s approximations are in the bottom rows. It shows that the collocation approach
confirms high accuracy for as little as 15 collocation points.
Table 13.3.1: Comparison of results for different collocation numbers

R = 1.1 h = 0.05 R =1.2 h = 0.1 R=2 H = 0.5
N L -D L -D L -D
7 688.92 98.9997 187.511 54.9163 11.9761 19.7986
11 998.18 106.1163 261.847 58.0715 13.7757 20.0171
15 1014.44 106.4877 265.066 58.2062 13.8324 20.0238
21 1014.79 106.4956 265.124 58.2086 13.8337 20.0239
(13.3.33) 1014.79 106.4956 265.124 58.2086 13.8337 20.0239
(13.3.34) 967.36 96.7360 241.840 48.3680 9.6736 9.6736
Streamlines for the rotation of the inner cylinder ( R1 = 1, R2 = 2, h = 0.5 ) are plotted
in Fig.13.3.2. It shows a separation and a back flow in the domain. Loyziansky [1955; 1959]
considered both cylindrical and spherical bearings and he predicted this flow. The
streamlines for the forward movement of the cylinder are shown in Fig.13.3.2.
In conclusion it should be noted that for a movement of a cylinder with arbitrary
section the problem is reduced to two numerical problems; one is connected with
transformation, another – with solving of a matrix equation (13.3.26).
522
Fig.13.3.2.
Streamlines of a cylinder
moving in a cylinder:
(a) rotation
(b) horizontal movement.
Sec. 13.4. Application of BEM to Small Reynolds Number Flows

13.4.1. Linear Algebraic Equation Set for Polyharmonic Equation
Integral equations set (A2.18) for Polyharmonic function can be reduced, as in Sec. 8.1 using
BEM-method, to a linear algebraic equations set
(ε E + A 0 )U n −1 − B 0 Vn −1 = 0 ⎫
⎪
(ε E + A 0 )U n − 2 − B 0 Vn − 2 + A1U n −1 − B1Vn −1 = 0 ⎪
⎬, (13.4.1)
⎪
(ε E + A 0 )U 0 − B 0 V0 + A1U1 − B1V1 + + A n −1U n −1 − B n −1Vn −1 = 0 ⎪⎭
where E is unit matrix; A k , B k are matrices whose elements are computed integrating along
boundary elements functions (A2.14) and its normal derivatives; U k , Vk are vector-matrices
whose elements are values of functions (A2.16) and its normal derivatives at nodal points.
If the closed boundary is discretized by N elements then the equations set (13.4.1) is
presented as a set of N n linear algebraic equations with respect to 2N n components. For
solving this equations set, it should be given N n values of any components, and so
equations set (13.4.1) and its solution may be written in matrix form, respectively,
MX = Y, X = M −1Y . (13.4.2)
So, boundary value problems of polyharmonic functions may be classed in theory of
harmonic functions as following:
1) Dirichlet's problems if values of functions uk are given;
2) Neumann’s problems if values of its normal derivatives vk are given;
3) Mixed problems if values of functions uk and its normal derivatives vk are given on a
part of boundary for the former and on another its part for the latter.
13.4.2. Linear Algebraic Equations Set for Fluid with Limited High Viscosity
Looking back to Sec. 14.3, one can see that the stream function is the polyharmonic function
of second order, and so it can be presented as a sum of a system of the Laplace equations
concerning intensity of vortices ω ( x, y ) and a Poisson equation concerning stream function
ψ ( x, y ) : Δω = 0, Δψ = −ω , or two first equations in (A2.18) written as
523
Lω = 0, Lψ = ∫ ω (τ ) H n ( z ,τ )dsτ − ∫ ωn (τ ) H ( z ,τ )dsτ , (13.4.3)

C C
where integral operator L acts on functions ω and ψ as
ω( z)
Lω = + ∫ ω (τ )Gn ( z ,τ )dsτ − ∫ ωn (τ )G ( z ,τ )dsτ ,
2 C C
1 1 r 2 (ln r − 1)
G (r ) = ln , H (r ) = , r =| z −τ | .
2π r 8π
Denoting elements of N × 1 - matrixes P, Q, W and U accordingly as
Pn = ψ ( Z n ) , Qn = ψ n ( Z n ) , Wn = ω ( Z n ) , U n = ωn ( Z n ) , (13.4.4)
one can present the set of integral equations (13.4.3) in matrix form as
AW - BU = 0, AP - BQ + CW - DU = 0 . (13.4.5)
The elements of 2Nx2N matrixes A, B, C and D are determined in integral forms as
⎧ ln ⎛ 2⎞
⎧0.5, n = k , ⎪ ⎜ 1 + ln ⎟ , n = k ,
⎪z ⎪ 2π ⎝ ln ⎠
Akn = ⎨ n Bkn = ⎨ z
⎪ ∫ Gn ( Z k , z )dτ z , n ≠ k , ⎪ G ( Z , z )dτ , n ≠ k ,
n
⎩ zn−1 ⎪∫ k z (13.4.6)
⎩ zn−1
zn zn
Ckn = ∫H
zn−1
n ( Z k , z ) dτ z , Dkn = ∫ H (Z
zn−1
k , z )dτ z .
These integrals can be calculated numerically by Gauss formula or preferably found

1 ⎛ 1 − ξ kn 1 + ξ kn ⎞
analytically: Akn = − ⎜ arctan + arctan ⎟, k ≠ n ; (13.4.7)
2π ⎝ η kn η kn ⎠
ln ⎛ 1 ⎞
Bkn = ⎜η kn Akn − ⎡⎣(1 − ξ kn ) pkn + (1 + ξ kn ) qkn − 4⎤⎦ ⎟ , k ≠ n ; (13.4.8)
2π ⎝ 4π ⎠
lnη kn ⎛ ln ⎞
Ckn = − Bkn + ; (13.4.9)
4 ⎜⎝ 4π ⎟⎠
ln3 ⎡⎛ (1 − ξ )3 ⎞ ⎛ (1 + ξ kn )3 ⎞ ⎤
Dkn = ⎢⎜ kn
+ η kn (1 − ξ kn ) pkn + ⎜
2
⎟ + η kn2 (1 + ξ kn ) ⎟ qkn ⎥ −
128π ⎢⎣⎜⎝ 3 ⎟
⎠
⎜
⎝ 3 ⎟ ⎥
⎠ ⎦
(13.4.10)
l3 ⎡ 4 (1 + 3ξ kn2 ) 5η 2 2π ⎤
− n ⎢ + kn + ηkn3 Akn ⎥ ,
32π ⎢⎣ 9 3 3 ⎥⎦
ln2 ⎡ l 2
(1 − ξ kn ) + ηkn2 ⎤⎦ , qkn = ln n ⎡(1 + ξ kn ) + η kn2 ⎤ .

2 2
where pkn = ln
4⎣ 4⎣ ⎦
13.4.3. Determining of a Stream Function and Its Normal Derivative

Normal and tangent derivatives are derived from boundary conditions (13.3.7) and (13.3.13):
ψ n ( Z k ) = Qk = ψ x ( Z k )nx ( Z k ) + ψ y ( Z k ) ny ( Z k ),
(13.4.11)
ψ s ( Z k ) = ψ y ( Z k )nx ( Z k ) − ψ x ( Z k )n y ( Z k ).
524
yk − yk −1 x − xk −1
where nx ( Z k ) = , ny (Z k ) = − k . Integrating the second equation
lk lk
of (13.4.13) along an internal contour, one may determine the stream function on this
contour: P0 = 0.5l1ψ s ( Z1 ) , Pk = Pk −1 + 0.5 [lkψ s ( Z k ) + lk −1ψ s ( Z k −1 )] . (13.4.12)
On the external fixed contour the stream function takes constant value c
( Pk = c, k = N + 1, 2 N ), which is determined during solving the problem. Thus, after
discretization of the governing equations, the problem is reduced to calculating 4N unknown
values Wk and U k and parameter c. However equations (13.4.5) make a system of 4N linear
equations. The missing equation can be found from the condition of unique pressure along
the internal contour. Since derivative ∂p / ∂s = μωn the constant pressure condition can be
N
written as ∑U l
k =1
k k =0. (13.4.13)
Adding this equation to (13.4.5) yields the total system of equations in a matrix form of
dimension 4 N + 1 FV = R (13.4.14)
with the following elements:
⎧0, k = 1, 2 N , ⎧ −U k , k = 1, 2 N ,
⎪⎪ ⎪⎪
Rk = ⎨Vk − 2 N , k = 2 N + 1, 4 N , Vk = ⎨Wk − 2 N , k = 2 N + 1, 4 N , (13.4.15)
⎪0, k = 4N + 1 , ⎪c , k = 4N + 1 ,
⎪⎩ ⎪⎩
⎛ B1,1 ... B1, N B1, N +1 ... B1,2 N A11 ... A1,2 N 1⎞
⎜ ⎟
⎜ ... ... ... ... ... ... ... ... ... ... ⎟
⎜ B2 N ,1 ... B2 N , N B2 N , N +1 ... B2 N ,2 N A2 N ,1 ... A2 N ,2 N 1 ⎟
⎜ ⎟
Fkn = ⎜ . D1,1 ... D1, N D1, N +1 ... D1,2 N C1,1 ... C1,2 N 1 ⎟ (13.4.16)
⎜ ... ... ... ... ... ... ... ... ... ... ⎟
⎜ ⎟
⎜ D2 N ,1 ... D2 N , N D2 N , N +1 ... D2 N ,2 N C2 N ,1 ... C2 N ,2 N 1 ⎟
⎜ l ... l1 0 ... 0 0 ... 0 0 ⎟⎠
⎝ 1
The first 2N elements of the solution V = F -1 R (13.4.17)
determine the normal derivative regarding vortices at nodal points U k = −Vk (k = 1, 2 N ) ;
here the first N elements correspond to the internal cylinder, while other N elements
correspond to the external one; the following 2N elements determine the values of vortices
Wk = −Vk − 2 N , and the last component gives the value of stream function on the external fixed
cylinder.
13.4.4. Resulting Force and Torque
To calculate the speed-deformation tensor, partial derivatives of stream function are to be
defined. This can be done by differentiating boundary conditions (13.3.7) and (13.3.13)
along boundary C and to taking advantage of vorticity. Hence, we have a linear system of
equations:
525
⎧ n yψ xx − nxψ xy = −Ω n y ,
⎪
⎨ n yψ xy − nxψ yy = Ω nx , (13.4.18)
⎪
⎩ψ xx + ψ yy = − ω .
Whence, the necessary partial derivatives of second order are defined
ψ xx = −ω nx2 + Ω(nx2 − ny2 ) , ψ yy = −ω n y2 − Ω(nx2 − n y2 ) , ψ xy = −ω nx n y + 2Ωnx n y .
Using these expressions one can find from (13.3.8) and (13.3.9) the speed-deformation
tensor and then the stress tensor, and finally all forces acting on the boundary; e.g. the stress
is calculated by f x = pnx + μ (ω − 2Ω)n y , f y = pn y − μ (ω − 2Ω) nx . (13.4.19)
For calculated normal derivative ωn , the tangent derivative ps = μωn and pressure at nodal
l ω l + ωk −1lk −1
points of all element are determined as p0 = μω0 0 , pk = pk −1 + μ k k . (13.4.20)
2 2
Components of resulting force and toque are calculated accordingly as:
N N
X = ∑ ( f x ) k lk , Y = ∑ ( f y ) k lk ; (13.4.21)
k =1 k =1
N
M = ∑ ⎡⎣( f y ) k X k − ( f x ) k Yk ⎤⎦ lk . (13.4.22)
k =1
It is seen that the dynamic viscosity coefficient is included into the resulting force and
moment as a factor. If dimensional sizes are related to fluid density ρ , square of
characteristic speeds V, and a linear size of the moving cylinder (for example radius а), the
resulting force and toque will be inversely proportional to Reynolds number R e = ρVa / μ .
Therefore, each characteristic quantity, ρ, V, a and μ, could be put unity. The calculated
values of forces and torque determine their dimensionless values increased on Re. By non-
stationary movement the time appears only via kinematic conditions, i.e. hydrodynamic
characteristics do not depend on the background movement and are determined by kinematic
conditions only.
Rotation of an eccentric circular cylinder in a cylinder. Let the external fixed cylinder’s
center be shifted relative to the rotating internal cylinder’s center upwards on distance d.
Their radiuses are R and unity, respectively (Fig.13.3.1). For comparison two different
arrangements have been calculated: 1) R = 2, d = 0.9 and 2) R = 1.1, d = 0.5. Due to small
distance between surfaces of the cylinders, both cases are quite inconvenient for numerical
methods. Therefore, the number of elements on each cylinder should be rather high. In the
first case, calculations were carried out for two element numbers: N = 100 with
X = 39.19, M = −40.76 and N = 150 with X = 39.25, M = −40.80 . The exact values
calculated by equations (13.3.33) and Sommerfeld’s approximation (13.3.34) are equal
accordingly to X e = 39.32, M e = −40.85 and X s = 27.70, M s = −26.88 . In the second case
calculations were carried out only for N = 150 : X = 1017, M = −106.46 . The exact and
approximate data are accordingly: X e = 1015 , M e = −106.50 and X s = 967.4, M s = −96.74 .
From here it is seen that the offered numerical method allows receiving high accuracy, but
the Sommerfeld approximation diverges noticeably from the exact one even for small gaps.
526
Plotted in Fig.13.4.1 are the results of numerical calculations for rotating cylinders
of an elliptic section with their centers shifted by d = 0.9 relative to the center of the
external circle of radius R = 2 . The speed of rotation is assumed unity. The plot shows the
dependence of resulting force and torque on turning angle of the elliptical section with
aspect ratio 0.5 (radius of external cylinder R=2, distance between centers d=0.9). It is seen
that components X and Y depend strongly on turning angle, however the resulting moment M
varies little (dashed line). It should be taken into account
at designing of any mechanisms with a pin and bearing
parts. At a small deviation of the pin shape from a
circular cylinder, the loadings varied sharply causing fast
deterioration of the bearing and creating noise and
fluctuations.
Fig.13.4.1.
Forces and torque vs. turning angle of internal cylinder
of elliptic section with aspect ratio 0.5.
Also calculated was the rotation of a square-

section cylinder with unit sides in a square-section box with sides = 2, as shown in the insert
in Fig.13.4.2. Torque in this case is approximately expressed as
M = −0.4cos 2α − 4.91 . (13.4.23)
Fig.13.4.2.
Resulting force and torque for an asymmetric piston
moving in a box.
Fig.13.4.2 shows a plot of drag and torque for an

asymmetric piston moving horizontally in a rectangular
box. Each side of the rectangular boundaries was divided
into 25 elements so that the domain boundary was
approximated by 200 elements. One can see that the
maximal drag occurs when the piston moves directly in the middle of the box. If the piston
deviates aside, the torque is generated to turn it clockwise for the case as shown the plot.
Numerous calculations verify a high efficiency of the boundary element method. The
numerical algorithm allows to calculate a very close sizes of internal and external bodies, at
that the curve of drag remains smooth almost to the touch of bodies, while the torque
deviates from smooth curve even for d > 0.08 . When approaching d→0, the piston touches
the cylinder’s wall and the problem is reduced to that of piston moving in the middle of the
cylinder with its vertical sides doubled. The point X(0.1) = – 734 was calculated as a half of
the drag of such piston.
In a conclusion the considered methods and algorithms are quite effective for
numerical calculations and could be applied to movement problems of more than one body.
These methods may be used also in the two-dimensional theory of elasticity.
527
Appendices
Appendices
App. 1. Boundary-Value Problems for Singular Functions
A.1.1. The Keldysh – Sedov Solution
In many applied problems, one needs to develop functions having isolated singularities at
prescribed points of a region or its boundary; for example, in cavitating flow problem, in
flow problems around profiles with prescribed singularities, etc.
Boundary value problem solutions for analitic functions have been obtained by many
authors. A mixed problem was solved in [Signorini, 1916] for analytic function in a class of
bounded functions. The solution was extended in [Keldysh & Sedov, 1937)] for integrable
singularity at a boundary using different domains: half plane, strip, and ring. Another way of
the Keldysh-Sedov solution was obtained by Muskhelishivili [1953] and for solving many
other value problems. A modification of the Keldysh-Sedov formula is presented below for
solving mixed boundary problems as well as Schwartz problems in the class of functions
with prescribed singularities [Terentiev, 1971b].
Let’s consider a value problem for function ω ( z ) = u − iv satisfying periodic mixed
conditions with period π at sides y = 0 and y = πτ i / 2 of a strip πτ / 2 wide. Let M be a set
of intervals ak , bk ( k = 1,..., p ) and cs + πτ i / 2, d s = πτ i / 2( s = 1,..., q) on a horizontal sides of
a rectangle of length π , where real part u of function ω ( z ) is given, and let N be a set of
complementary intervals where imaginary part v is prescribed.
The solution is sought in a class of functions having poles of given orders at points
of the rectangle, not including the endpoints of intervals in M. At the endpoints of these
− m+ 1
intervals, the function may have a singularity of the form ( z − zn ) 2
. In particular, if
desired function ω ( z ) is to have integrable singularities at the endpoints of intervals, we
must take m = 1 . Applying the functions
p
ϑ1 ( z − bk ) q
ϑ4 ( z − d s )
g ( z) = ∏
k =1 ϑ1 ( z − ak )
∏ ϑ (z − c ) ,
s =1
(A1.1)
4 s
ϑ1 (0)ϑ1 ( z + c) 1⎡ p q
⎤
A( z ) = , c = ⎢ ∑ (bk − ak ) + ∑ (d s − cs ) ⎥ , (A1.2)
ϑ1 (c)ϑ1 ( z ) 2 ⎣ k =1 s =1 ⎦
the Keldysh – Sedov solution
1 ⎡ ⎤
F ( z) = ⎢ ∫ u (t ) g (t ) A(t − z )dt − i ∫ v(t ) g (t ) A(t − z )dt ⎥ , (A1.3)
π ig ( z ) ⎣ M N ⎦
has periods π & πτi/2, satisfies the given boundary conditions at the rectangle sides, is
bounded at points ak & cs +πτi/2, and has an integrable singularity at points bk & ds +πτi/2.
A.1.2. Singular Solution on Rectangle Domain
Consider auxiliary function χ ( z ) = [ω ( z ) − F ( z ) ] g ( z ) . It is easy to see that Im { χ ( z )} = 0 on
straight line y = 0 , Im { χ ( z )e − ci } = 0 on line y = πτ / 2 , and that the latter function has only
poles in the z -plane, excluding the endpoints of the intervals in M . Besides,
528
Appendices
χ ( z + π ) = χ ( z) ,
χ ( z + π ) = e2 ci χ ( z ) , i.e., function χ ( z ) is a periodic function of second
kind with multipliers μ1 = 1 and μ 2 = e 2 ci . Consequently [Tannery & Molk, 1898], it can
be presented as a linear combination of A( z ) -function and their n-th derivatives, A( n ) ( z ) , as
n
⎡ mr ⎤ λ
⎡ vr
πτ ⎤
χ ( z ) = ∑ ⎢ ∑ Brm A( m −1) (α r − z ) ⎥ + ∑ ⎢ ∑ Crv A( v −1) ( β r − z + ) ⎥ eci +
r =1 ⎣ m =1 ⎦ r =1 ⎣ v =1 2 ⎦
. (A1.4)
εr
⎡ γ
⎤
+ ∑ ⎢ ∑ Drε A(ε −1) ( zr − z ) + Drε A(ε −1) ( zr − z ) ⎥
r =1 ⎣ ε =1 ⎦
Here coefficients Brm and Crv are real; Drε and Drε are complex conjugates; z = α r ,
πτ i
z = βr + , z = zr are the given points, at which function ω ( z ) has poles of orders mr ,
2
πτ i
vr and ε r . If z = α r and z = β r + coincides with endpoints of intervals of M , then
2
function ω ( z ) will have singularities at these points of the form
1 1
− vr + − vr −
⎛ πτ i ⎞ 2
− mr +
1
− mr − ⎛ πτ i ⎞ 1
2
( z + ar ) , ⎜ z − cr − ⎟
2
, ( z + br ) 2
, ⎜ z − dr − ⎟ .
⎝ 2 ⎠ ⎝ 2 ⎠
The desired function is now presenting as: ω ( z ) = F ( z ) + χ ( z ) / g ( z ) . (A1.5)
n λ γ
Thus the problem solution contains exactly χ = ∑ mr + ∑ vr + 2∑ ε r arbitrary constants.
r =1 r =1 r =1
A.1.3. Periodic Solution in the Half-Plane

From formulas (A1.1) – (A1.4) one can easily obtain, by approaching the limit ( τ → ∞ ), a
mixed boundary problem solution of periodic functions on the half-plane as a special case.
Constants c , d s , cs are then arbitrary. Setting c = π / 2 in expressions (A1.1) – (A1.4), we
p
sin( z − bk )
get g ( z) = ∏ sin( z − a )
k =1
, A( z ) = cot z , (A1.6)
k
n
⎡ mr ⎤ γ
⎡ εr ⎤
χ1 ( z ) = B + ∑ ⎢ ∑ Brm cot ( m −1) (α r − z ) ⎥ + ∑ ⎢ ∑ ( Drε cot (ε −1) ( zr − z ) + Drε cot (ε −1) ( zr − z ) ) ⎥ , (A1.7)
r =1 ⎣ m =1 ⎦ r =1 ⎣ ε =1 ⎦
λ
where B = ∑ Cr1 is an arbitrary constant. The solution is given in the form of (A1.5).
r =1
A.1.4. A Solution in the Half-Plane

Changing the arguments of the sine and tangent in (A1.6) and (A1.7), one may obtain the
mixed boundary problem solution for a half-plane
⎛ z − ak ⎞ ⎧ 1 p ⎡ k p ⎛ z − bk ⎞ udt ⎛ t − bk ⎞ udt ⎤
p b ak +1 p
ω ( z) = ∏ ⎜ ⎟ ⎨ ∑ ∫ ∏⎜ ⎢ ⎟ − i ∫ ∏ ⎜ ⎟ ⎥+B+
k =1 ⎝ z − bk ⎠ ⎩ π i k =1 ⎢ ak k =1 ⎝ z − ak ⎠ t − z ⎝ t − a ⎠ t − z ⎥
⎣ bk
s =1 k
⎦ (A1.8)
n m r
Brm γ ε r
⎛ Drε Drε ⎞ ⎪ ⎫
+ ∑∑ + ∑∑ ⎜ + ⎟ ⎬ , a p +1 = a1
r =1 m =1 ( z − α r )
ε
r =1 ε =1 ⎝ ( z − zr ) ( z − zr )ε ⎠ ⎪⎭
529
Appendices
From equations (A1.1) – (A1.5) one may also derive a mixed boundary problem solution for
a strip with boundary conditions on lines y = 0 and y = π / 2 . Using the imaginary Jacobi
transformation τ = 1/ τ ′ , and approaching τ ′ → 0 , we get:
p
sinh( z − bk ) q
cosh( z − d s ) sinh( z + c)
g ( z) = ∏
k =1 sinh( z − ak )
∏ cosh( z − c ) ,
s =1
A( z ) =
sinh(c )sinh( z )
, (A1.9)
s
n mr λ νr
χ 2 ( z ) = B + ∑∑ Brm coth ( m −1) (α r − z ) + ∑∑ Crν coth (ν −1) ( β r − z ) +
r =1 m =1 r =1 ν =1
εr
(A1.10)
γ
⎧ ⎫
+ ∑ ⎨∑ ⎡⎣ Drε coth ( ε −1)
( zr − z ) + D rε coth ( ε −1)
( zr − z ) ⎤⎦ ⎬ ,
r =1 ⎩ ε =1 ⎭
λ ν
⎡ ⎤
( )
n
B = ⎢ ∑ Br1 + ∑ Cr1 + ∑ Dr1 + D r1 ⎥ coth c .
⎣ r =1 r =1 r =1 ⎦
The solution, as before, is determined by equation (A1.5). It should be noted that the solution
for a strip by second equality in (A1.2) satisfies condition u (+∞) = u (−∞ ) . If c is arbitrary,
then u (+∞) ≠ u (−∞) , and the solution contains at least one arbitrary constant B.
A.1.5. A Singular Solution of Schwartz’s Problem

Using previous formulas one can obtain solutions of Schwartz’s boundary problem in the
class of singular functions. Find double periodic functions ω ( z ) = u − iv with periods π and
πτ i due to the given imaginary part on horizontal sides y = 0 and y = πτ / 2 of the
fundamental rectangle. Let’s consider first the mixed boundary problem for auxiliary
function ω ( z ) = u − iv , which satisfies the boundary conditions as follows:
• within interval ( −ε , ε ) : u = u1 ( x)
• within intervals ( −δ , − ε ) and (δ , ε ) : v = 0
• within interval (δ , π − δ ) : v = v1 ( x)
• within interval (πτ i / 2, π + πτ i / 2 ) : v = v2 ( x)
where u1 ( x) is an arbitrary bounded function; ε and δ are infinitesimal with ε / δ → 0 at
δ → 0 . Outside a circle of radius δ with center at point z = 0 , the following expansions are
ϑ ′( z ) 1⎡ ϑ ′( z ) ⎤
valid: g ( z) = 1 − ε 1 + O (ε 2 ) , A( z ) = ⎢1 + ε 1 + O (ε 2 ) ⎥ ; (A1.11)
ϑ2 ( z ) ε⎣ ϑ2 ( z ) ⎦
on the interval ( −ε , ε ) g (t ) = i (ε − t )(ε + t ) + O(ε 2 ) . (A1.12)
Expanding mixed value problem solution (A1.1) – (A1.5) in power series on ε , taking into
account (A1.11) and (A1.12), and approaching δ→0 yields the Schwartz’s problem solution
π π
1 ⎡ ϑ ′ (t − z ) ϑ4′ (t ) ⎤ 1 ⎡ ϑ ′(t − z ) ϑ1′(t ) ⎤
ω ( z ) = ∫ v2 (t ) ⎢ 4 − ⎥dt − ∫ v1 (t ) ⎢ 1 − ⎥dt + χ ( z ), (A1.13)
π0 ⎣ ϑ4 (t − z ) ϑ4 (t ) ⎦ π0 ⎣ ϑ1 (t − z ) ϑ1 (t ) ⎦
where function χ(z), as before, is given by (A1.4). Besides, functions v1(x) and v2(x) should
satisfy the equation obtained by asymptotic approach as double-periodicity condition:
π π n λ γ
1 1
∫2v
π
dt −
0
∫ 1 ∑ r1 ∑ r1 ∑ ( Dr1 + Dr1 ) = 0 .
v dt −
π
B −
0
C −
r =1 r =1
(A1.14)
r =1
530
Appendices
Solution (A1.13) together with (A1.4) and (A1.14) is a generalization of the well known
Villat’s solution fofr singular functions.
A.1.6. A Singular Solution of Schwartz’s Problem on other Domains
The Schwartz’s problem solution for other domains can determine similarly:
• for a half-plane
∞
1 v1 (t ) n mr
Brm γ εr
⎛ Drε Drε ⎞
ω ( z) = − ∫ dt + B + ∑∑ + ∑∑ ⎜ ε
+ ⎟; (A1.15)
π −∞ t − z r =1 m =1 ( z − α r )
m
r =1 ε =1 ⎝ ( z − zr ) ( z − z r )ε ⎠
• for a half-plane with periodic boundary conditions
π
1 dt
ω ( z ) = − ∫ v1 (t ) + χ1 ( z ) ; (A1.16)
π0 tan(t − z )
• for a strip of width π / 2
∞ ∞
1 1
ω ( z ) = ∫ v2 (t ) th(t − z ) dt − ∫ v1 (t ) cth(t − z )dt + χ 2 ( z ) , (A1.17)
π −∞
π −∞
where functions χ1 ( z ) and χ 2 ( z ) are determined by expressions (A1.7) and (A1.9).

Equation (A1.14) corresponds to equality u (−∞ ) = u (∞ ) .
App. 2. Integral Relationships

Most of numerical methods have been based on integral equations. Therefore, it is worth to
present briefly some auxiliary features of integral equations.
A.2.1. Green’s Identities
Let D be a domain bounded by closed surface C, and D be a domain including C. Let U and
G are unique and double differentiable functions in D . With the assumptions the second
Green’s identity is valid [Smirnov, 1957]
∫ (U ∇ G − G∇ U )dv − ∫ (UGn − GU n )ds = 0 ,
2 2
(A2.1)
D C
where Gn and U n denote the derivatives with respect to exterior normal to the boundary at
integrated points. Let z and τ are points in D , and function G ( z ,τ ) depends on distance
r =| z − τ | in the following way:
1 ⎛1⎞
• for a two-dimensional plain domain G ( z ,τ ) =
ln ⎜ ⎟ ; (A2.2)
2π ⎝ r ⎠
1
• for three-dimensional domain G ( P, Q ) = . (A2.3)
4π r
The third Green’s identity is expressed from (A2.1) in the form
LU = − ∫ ∇ 2U (τ )G ( z ,τ ) dvτ , (A2.4)
D
where L is function operator LU = c ( z )U ( z ) + ∫ U (τ )Gn ( z ,τ )dsτ − ∫ U n (τ )G ( z ,τ )dsτ . (A2.5)

C C
Factor c(z) takes a value of 1, if point z is an interior point z ∈ D ; or 1/2 if z is located on a
531
Appendices
smooth boundary z ∈ C , or β / 2π if z is vertex of angle with opening angle β (for a plane
domain), or is a half spherical angle of a cone (for three-dimensional domain).
Let points z0 ( x0 , y0 ) ∈ C , τ (ξ ,η ) ∈ C for a two-dimensional plane be determined by
curvilinear abscissas s0 , s , and boundary C be determined by functions x(s) and y(s), then
x0 = x( s0 ), y0 = y ( s0 ) and ξ = x( s ), η = y ( s ) . Green’s function (A2.4) and its normal
derivative could be expressed as follows:
1
G=− ln ⎡⎣ ( x − ξ ) 2 + ( y − η ) 2 ⎤⎦ ,
4π
(A2.6)
1 ( x − ξ ) y ′( s ) − ( y − η ) x′( s )
Gn = ∇G ⋅ nτ = − .
2π ( x − ξ )2 + ( y − η )2
Consider now extreme case: s → s0 . Substituting s = s0 + ε and then approaching ε → 0 ,
one obtains the extreme values as
1 y ′′( s0 ) x′( s0 ) − x′′( s0 ) y ′( s0 )
G ( s0 , s0 + ε ) ≈ − ln | ε |, Gn ( s0 , s0 ) = . (A2.7)
2π 4π
It follows that function G ( s0 , s) has a logarithmic singularity while its normal derivative is
continuous at point s = s0 . Since both numerator and denominator of second function (A2.6)
are infinitesimals of second order, computer can indicate it as a singularity. Therefore,
function Gn ( s0 , s) should be calculated by expression (A2.6) on boundary C except small
neighborhood of s0 , on which the function is equal to (A2.7).
A.2.2. Harmonic Functions

Greens’s integral identity (A2.2) does not carry any information on function f ( z ) . For its
determination, one should satisfy some equation in the domain and certain conditions on the
boundary. Namely, a harmonic function ϕ ( z ) satisfies Laplace’s equation ∇ 2ϕ = 0 , so that
the Green’s identity (A2.2) becomes Lϕ = 0 . (A2.8)
Satisfying boundary conditions, one obtains integral equation (A2.8) for this function or for
its derivative on the boundary. Substituting G = 1 into identity (A2.1), one can obtain
equality as ∫ ϕn ds = 0 . C
(A2.9)
Hence, the function given for Neumann’s condition cannot be arbitrary; it should satisfy
condition (A2.9). Besides, since normal derivative ϕn = (ϕ + const.) n then function could be
determined with the accurate within a constant. It should be noted that equation (A2.8) is
valid for infinite three-dimensional domain as well if expressions | z | ϕ ( z ) and | z |2 | ∇ϕ ( z ) |
are bounded at infinity (| z |→ ∞ ) [Smirnov 1957].
The harmonic function can have a logarithmic singularity at infinity on two-dimensional
plane [Wrobel, 1987], i.e., | ϕ ( z ) |≈ ln | z |, | ∇ϕ ( z ) |≈| z |−1 , | z |→ ∞ . (A2.10)
Indeed, equation (A2.8) is valid for bounded domain with boundary C and sphere CR of a
radius R. Since the first terms of the expansions of ϕ ( z )Gn ( z0 , z ) and ϕn ( z )G ( z0 , z ) are
equal, then their difference is small as | z |−2 ln | z |, (| z |→ ∞) . Hence,
532
Appendices
lim
R →∞ ∫ [ϕ (τ )G ( z,τ ) − ϕ
CR
n n (τ )G ( z ,τ )] ds = 0 .
A.2.3. Poly-Harmonic Functions

A poly-harmonic function can be represent as a set of integral equations [Terentiev, 2007].
Consider the poly-harmonic equation for both two- and three-dimensional domains:
∇ 2 nU = 0 . (A2.11)
To solve problems involving the latter equation, it is desirable to develop a contour integral
similar to that replacing the right-hand integral of Green’s identity (A2.4). For that, let us
consider poly-harmonic equations of an arbitrary order with polar and spherical symmetries,
1 d dGk 1 d dGk 1 1
respectively: r r = ln (for two-dimensional domain), (A2.12)
r dr dr r dr dr 2π r
1 d 2 dGk 1 d 2 dGk 1
r r = (for three-dimensional domain). (A2.13)
2
r dr dr 2
r dr dr 4π r
Integrating each ordinary differential equation one can obtain a particular solution for the
1 r 2k ⎛ 1 k 1 ⎞
planar problem (equation A2.15) as Gk = ln + ∑ , (A2.14)
2π 4k (k !) 2 ⎜⎝ r m =1 m ⎟⎠
and for the three dimensional problem (equation A2.16) as
1 r 2 k −1
Gk = , (A2.15)
4π (2k )!
where subscript k means the harmonic function of kth order, ∇ 2 k Gk = G ( r ) . It is obvious
that ∇ 2 Gk = Gk −1 , ∇ 2G1 = G ( r ) . Denoting
∇ 2u = u1 , ∇ 4u = u2 , … , ∇ 2( n −1) u = un −1 ,
(A2.16)
G = G0 , u = u0 , H k = ∂Gk / ∂n, vk = ∂uk / ∂n .
the right-hand integral of equation (A2.4) can be presented as
n −1
∫ ∇ u G (r )dVτ = ∑ ∫ (uk ∇ Gk − ∇ uk Gk ) dVτ .

2 2 2
(A2.17)
D k =1 D
Using now the Green’s identity (A2.1) we obtain instead of the Green’s equation (A2.4), a
n − k −1
set of n integral equations: ε uk = ∑ ∫ (v
m=0 C
k +m
)
Gm − uk + m H m ds, (k = 0, n − 1) . (A2.18)
It makes sense to classify the boundary-value problems of poly-harmonic functions similar

to harmonic functions in compliance with the boundary conditions as Dirichlet's problem (all
functions uk are given), or as Neuman’s problem (all normal derivatives vk are given), or as
a mixed problems (functions uk and vk are partly given).
A.2.4. Basic Problem of the Theory of Poly-Harmonic Functions

The basic value problem is: to find poly-harmonicс function u ( x, y ) of the n-th order n ≥ 1
satisfied in domain D to equation ∇ 2 n u = 0 continuously together with derivatives up to the
(n − 1) th order inclusive in closed domain D ⊂ D ∪ C and satisfied on boundary C to
conditions: u C = f 0 ( s ), ∂ k u / ∂ n k C = f k ( s ), s ∈ C , (k = 1, n − 1) . (A2.19)
533
Appendices
Let boundary C is determined by functions x = x( s ), y = y ( s ) which together with
its derivatives of the (n − 1) th order are continuous.
The tangent and the normal to boundary C have the following components:
τ = ( x′, y ′), n = ( y ′, − x′) (A2.20)
N = (n ⋅ ∇), S = ( τ ⋅ ∇),
S1 = S,
Let's enter differential operators S 2 = S1S + ∂S1 / ∂s, (A2.21)
S k = S k −1S + ∂S k −1 / ∂s.
∂u ∂u
The normal and tangent derivatives are = Nu , = Su . (A2.22)
∂n ∂s
When finding derivatives, it is necessary to mean that vectors (15.1.15) are functions of
variable s only while u is a function of coordinates x and y , and derivatives on boundary
are defined as limiting values. Therefore the mixed derivatives on a normal and curvilinear
coordinate are defined by differential operators (A2.21) and (A2.22), for example,
∂k u
= NiS j u . (A2.23)
∂n ∂s
i j
Let's consider poly-harmonic equations consistently from the first up to the third order.
Harmonic Function
A harmonic function satisfies the Laplace’s equation, ∇ 2 u = 0 . The first boundary condition
in (A2.19), u = f 0 ( s ) , is the Dirichlet’s one.
Bi-Harmonic Function.
In that case the function should satisfy bi-harmonic equation ∇ 4 u = 0 and also the first two
boundary conditions in (A2.19), u C = f ( s), v C = f1 ( s ), s ∈ C .
Values of functions u1 and v1 at nodal points are calculated by two first equations in
(A2.19). The partial derivatives are unknown; they can be found from linear equation set
x′u x + y′u y = u y ′u x − x′u y = v , or in matrix
M 0 U 0 = V0 , (A2.24)
⎛ x′ y ′ ⎞ ⎛ ux ⎞ ⎛u ⎞
where M0 = ⎜ ⎟, U0 = ⎜ ⎟ , V0 = ⎜ ⎟
⎝ y ′ − x′ ⎠ ⎝ uy ⎠ ⎝v⎠
Poly-Harmonic Function of Third Order
Differential equation Δ 3u = 0 should be solved. Using operators (A2.21), one can find the
unn = y ′2 u xx − 2 x′y ′u xy + x′2 u yy
following equations: uss = x′2u xx + 2 x′y ′u xy + y ′2u yy + x′′u x + y ′′u y
uns = x′y ′u xx + ( y ′2 − x′2 )u xy − x′y′u yy + y ′′u x − x′′u y
The left parts of the equations can be calculated from the three first conditions (A2.19), and
so adding equation, u xx + u yy − u1 = 0 , one obtains a set of four linear equations that can be
534
Appendices
written in matrix form M1U1 = V1 , (A2.24)

⎛ 1 −1 0 −1 ⎞ ⎛ 0 ⎞ ⎛ u1 ⎞
⎜ 2 ⎟ ⎜ ⎟ ⎜ ⎟
0 y ′2 −2 x′y ′ x′ ⎟ ⎜ unn ⎟ , U = ⎜ u xx ⎟ .
where M1 = ⎜ , V =
⎜ 0 x ′2 2 x′y ′ y ′2 ⎟
1
⎜ uss − x′′u x − y ′′u y ⎟ 1
⎜ u xy ⎟
⎜ ⎟ ⎜⎜ ⎟⎟ ⎜⎜ ⎟⎟
⎝ 0 x′y ′ y ′2 − x ′ 2 − x′y′ ⎠ ⎝ uns − y ′′u x + x′′u y ⎠ ⎝ u yy ⎠
Whence function u1 is calculated, and so other unknown functions v1 , u2 , v2 are calculated
by three first equations in (A2.19).
In the same way, the next boundary values could be found and desirable functions as
well as function u ( s ) can be calculated for any polyharmonic equation. For instance, for
differential equation ∇8u = 0 from four first conditions in (A2.19), functions u0 , v0 , u1 , v1
could be found, and then the values at nodal points of desirable functions u2 , v2 , u3 , v3 are
calculated by four first equations (A2.18). Thus the basic problem of the theory of
polyharmonic functions are reduced to a mixed value problem for a system of poly-harmonic
functions uk , vk . Using the BEM, the latter equations and integral equations system (A.2.18)
can be reduced to a single global system of linear algebraic equations.
A.2.5. Green’s Equation for Axisymmetric Functions
Let variables x, y , θ be cylindrical coordinates with radius y. Since axisymmetric function
U (τ ) depends on x and y only, then the double integrals in Green’s operator (A2.5) with
function (A2.3) can be reduced to single integrals:
2π
1 dθ kK (k )
F ( z ,τ ) = ∫ =
4π 0 ( x − ξ ) + ( y − η cos θ ) + η 2 sin 2 θ 2π yη
2 2
, (A2.25)
yη
where k =2 .
(x −ξ ) + ( y +η )
2 2
The derivative with respect to normal n(nξ , nη ) could be calculated as

⎡ E ( k ) ⎛ ∂k
1 ∂k ⎞ kK (k ) ⎤
Fn (τ , z ) = ( nτ ⋅ ∇τ F ) = ⎢ 2 ⎜
nξ + nη ⎟ − nη ⎥ , (A2.26)
2π yη ⎣1 − k ⎝ ∂ξ ∂η ⎠ 2η ⎦
where E (k ) and K (k ) are the full elliptic integrals of first and second orders, respectively.
∂k k 3 ( x − ξ ) ∂k k k3 ( y +η)
The partial derivatives are = , = − . (A2.27)
∂ξ 4 yη ∂η 2η 4 yη
Thus Green’s operator (A2.5) for axisymmetric function is expressed as
LU = c( z )U ( z ) + ∫ U (τ ) Fn ( z ,τ ) y (τ )dsτ − ∫ U n (τ ) F ( z ,τ ) y (τ )dsτ . (A2.28)
C1 C1
Here C1 is non-closed generatrix of axisymmetric domain on axial section (y≥0, η≥0).

Function (A2.6) and its partial derivatives have uncertainty 0/0 at the intersection of the x-
axis and a generatrix. Approaching η→0 while y≠0, one obtains asymptotic functions:
1 x −ξ
F ( x, y; ξ ,0) = , Fξ ( x, y; ξ ,0) = 3
, Fη ( x, y; ξ ,0) = 0 . (A2.29)
2 (x − ξ ) + y
2 2
2 (x − ξ ) + y
2 2
535
Appendices
The latter could be useful by integrating along the generatrix including the
intersection point. If integrals calculated using Gauss’ approach, which includes no end
points, then the computing can proceed with no problems. Another singularity has function
(A2.6) by coinciding coordinates z and τ on boundary C1 . Using the above-mentioned
asymptotic approach, one may obtain the following asymptotic expressions:
ε2 ε 4
k2 ≈1− 2 , k′ = 1− k 2 ≈ , K (k ) ≈ ln ,
4 y ( s0 ) 2 y ( s0 ) k′
∂k ε2 ⎛ x′( s0 ) ⎞
≈ 2 ⎜ y ′( s0 ) x′′( s0 ) − x′( s0 ) y ′′( s0 ) − ⎟
∂n 8 y ( s0 ) ⎝ y ( s0 ) ⎠
Thus functions (A2.25) and (A2.26) are presented asymptotically as
1 8 y ( s0 )
F ( s0 , s0 + ε ) ≈ ln , (A2.30)
2π |ε |
1 ⎛ x′( s0 ) 8 y ( s0 ) ⎞
Fn ( s0 , s0 + ε ) ≈ ⎜ y ′( s0 ) x′′( s0 ) − x′( s0 ) y ′′( s0 ) + ln ⎟. (A2.31)
4π y ( s0 ) ⎝ y ( s0 ) eε ⎠
Expressions (A2.30) show that function F ( s0 , s ) as well as G ( s0 , s) has a
logarithmic singularity by approaching s → s0 . Such singularity has also its normal
derivative Fn ( s0 , s ) while normal derivative Gn ( s0 , s ) is bounded.
It should be noted that parameter k approaches 1 as s approaches s0 so that elliptic
integral K (k ) has a logarithmic singularity. Besides, function (A2.26) has uncertainty as
ε 2 / ε 2 . All these facts can decrease the calculation accuracy. Therefore, asymptotic
expressions (A2.30) and (A2.31) should be applied by development of numerical algorithms.
App. 3. Singular Functions

Integral relationships in two previous sections are valid in an interior of the domain for
continuous and bounded functions only. In the exterior the functions may have logarithmic
singularity at infinity. Problems considered in Chapter 1 show that functions in the flow
could have singularity as multiple poles both at inner points and at infinity. Some of integral
identities for singular functions are considered below.
A.3.1. Generalization of Cauchy’s Identity
Consider function w( z ) which has a multiple pole of the nth order at point za and a single
pole at boundary point zc . Besides, at infinity it becomes infinity of the m-th order. Let a
n
Ak
leading parts be as follows: w( z ) ~ ∑ , z → z0 ;
k =1 ( z − z0 )
k
n
w( z ) ~ ∑ Bk z k , z→∞; (A3.1)
k =1
D
z → zc , zc ∈ C .
w( z ) ~
,
z − zc
The value of complex function at an arbitrary point is determined by Cauchy’s identity
536
Appendices
1 w(τ )
c( z ) w( z ) = ∫
2π i δ z τ − z
dτ , (A3.2)
where δ ε is the boundary of an infinitesimal neighborhood, which includes point z and

belongs under an analytical domain; c ( z ) is the same as in equation (A2.5). Deforming curve
δ ε one can reduce to integrating along boundary C and to residues of the function at poles
1 ⎛ w(τ ) w(τ ) w(τ ) 1 w(τ ) ⎞
c( z ) w( z ) = ⎜∫ dτ − ∫ dτ + ∫ dτ − ∫ dτ ⎟ . (A3.3)
2π i ⎜⎝ C τ − z z0
τ −z ∞
τ −z 2 δ c τ − z ⎟⎠
The last integral in (A3.3) is obtained as
w(τ ) D
∫z τ − z dτ = −2π i z − zc . (A3.4)
c
In a small neighborhood of point za , function ( τ − z ) −1 can be expanded in the power series

1 ⎛ τ − za ⎛ τ − za ⎞ ⎞
2
1
= ⎜1 + +⎜ ⎟ + …⎟ .
τ − z za − z ⎜ z − za ⎝ z − za ⎠ ⎟
⎝ ⎠
Considering the series and the leading part of the function at the pole, the second integral in
w(τ ) n
Ak
(A3.3) is expressed as ∫z τ − z dτ = −2 π i ∑
k =1 ( z − z0 )
k
. (A3.5)
0
1⎛ ⎞
2
1 z ⎛z⎞
Near the infinite points, τ → 0 , expansion = ⎜1 + + ⎜ ⎟ + ⎟ is valid.
τ − z τ ⎜⎝ τ ⎝ τ ⎠ ⎟
⎠
w(τ ) m
Hence, ∫r =∞ τ − z dτ = 2π i ∑ k =1
Bk z k . (A3.6)
Substituting obtained integrals (A3.4) – (A3.6) into (A3.3) yields the generalized Cauchy’s
1 w(τ )
identity for meromorphic function: c( z ) w( z ) = ∫
2π i C τ − z
dτ + F ( z ) (A3.7)
n
Ak m
D
with F ( z) = ∑ + ∑ Bk z k + . (A3.8)
k =1 ( z − z0 )
k
k =0 2 ( z − zc )
Variable z in (A3.7) and (A3.8) belongs to inner points. If z belongs to boundary C then the
left part of identity (A3.7) should be multiplied by ε , which determines the value between
the tangents, β = 2πε .
A.3.2. Generalization of Green’s Identity

Consider a meromorphic function w(z) with the above- mentioned singularities.
Meromorphic function w( z ) = ϕ ( x, y ) + iψ ( x, y ) is determined by two conjugate harmonic
functions with singularities corresponding to (A3.1). These are the functions for which Just
Green’s integral identities are generated. Using polar coordinates r, θ and expressions
dτ dr
τ − z = r eiθ , = + idθ , one obtains
τ −z r
537
Appendices
1 w(τ ) 1 ⎛ dr ⎞ i ⎛ dr ⎞
∫
2π i C τ − z
dτ =
2π ∫ ⎜⎝ ϕ dθ + ψ
C
⎟+
r ⎠ 2π ∫ ⎜⎝ −ϕ
C
r
+ ψ dθ ⎟ .
⎠
(A3.9)
dr ∂ ∂ ⎛ 1⎞
The following relationships are fulfilled along boundary C: = ( ln r ) ds = − ⎜ ln ⎟ ds
r ∂s ∂s ⎝ r ⎠
and, due to analytic property of the function, ln(τ − z ) = ln r + iθ ,
∂θ ∂ ∂ ⎛ 1⎞
dθ = ds = ( ln r ) ds = − ⎜ ln ⎟ ds .
∂s ∂n ∂n ⎝ r ⎠
Substituting the latter relationships into (A3.9) and then into (A3.7), one obtains two integral
identities as follows: c ( z )ϕ + ∫ ϕ Gn ( z ,τ )ds + ∫ ψ Gs ( z ,τ )ds = Re F ( z ) , (A3.10)
C C
c( z )ψ + ∫ ψ Gn ( z ,τ )ds − ∫ ϕ Gs ( z ,τ )ds = Im F ( z ) . (A3.11)

C C
The second integrals are improper and should be calculated as principal values. They
could be integrated by parts if ϕ and ψ are single-value functions. Equations (A3.10) and
(A3.11) in that case yield the generated Green’s identities
Lϕ = Re F ( z ) , (A3.12)
Lψ = Im F ( z ) , (A3.13)
where L is Green’s operator (A2.5).
A.3.3. Properties of Integrals
Consider interior of a domain with closed boundary C. If function ϕ = const then conjugate
function ψ is also constant, then function F(z) and the second integral in (A3.10) or (A3.11)
vanish. Hence, ε + ∫ Gn ( z ,τ )ds = 0 . (A3.14)
C
If the domain is exterior then for constants ϕ and ψ (e.g. w = a + ib ) function F ( z ) = a + ib
and the integral equations yield ε + ∫ Gn (τ , z )ds = 1 . (A3.15)
C
It should be noted that identities (A3.14) and (A3.15) are well known and could be
found integrating along a closed curve. They are correct for arbitrary closed curve C. If the
domain is multiply-connected and bounded by a finite number of closed curves Ck , then
equalities (A3.14) and (A3.15) are valid for each curve. Namely, if the domain is exterior for
all curves, then the equalities for each curve should be fulfilled as
⎧1, z ∈ Ck , τ ∈ Ck
ε + ∫ Gn ( z ,τ )ds = ⎨ . (A3.16)
Ck ⎩0, z ∉ Ck , τ ∈ Ck
Properties of the integral and of integral identities are very useful for a construction of
numerical approaches and algorithms.
A.3.4. Integral Equations for the Symmetric Case
Consider symmetric boundary C relative to the x-axis and the horizontal flow with unit
speed at infinity. In that case the complex potential at infinity w( z ) ≈ z , ( z → ∞) . Here the
arbitrary constant for the potential is assumed to be zero.
538
Appendices
Satisfying kinematical condition ψ = 0 at boundary C yields from (A3.10) and
(A3.11) two integral equations εϕ + ∫ ϕ Gn ( z ,τ )ds = x , (A3.7)
C
− ∫ ϕ Gs ( z ,τ )ds = y . (A3.8)
The second integral equation after integration by part is reduced to

∫ ϕs (s)G( z,τ )ds = y .
C
(A3.9)
Equations (A3.7) – (A3.9) can be applied for studying the flow without circulation past an
arbitrary plane body. The last equation (A3.9) allows obtaining the velocity at the boundary
directly; therefore it is more preferable than others.
App. 4. Generalized Analytic Functions

A.4.1. Determination of Generalized Analytic Functions
As was pointed above the methods of the theory of analitic functions of complex variable are
the mostly powerful tools in the hydrodynamics. Unfortunately, complex functions could be
applied for two-dimensional planar flow if it is determined by two conjugate functions,
which satisfy analytic conditions ∂ϕ / ∂x = ∂ψ / ∂y and ∂ϕ / ∂y = −∂ψ / ∂x . The potential and
stream function of axisymmetric flow satisfy similar conditions as follows:
∂ϕ 1 ∂ψ ∂ϕ 1 ∂ψ
= , =− , (A4.1)
which differ from the analytic conditions by factor 1/y only. Pologiy [1965] has considered a
p-analitic function, F ( z ) = U ( x, y ) + iV ( x, y ) , which satisfies two equations as
∂U 1 ∂V ∂U 1 ∂V
= , =− . (A4.2)
∂x p ( x, y ) ∂y ∂y p ( x, y ) ∂x
Such functions are denoted as generalized p-analitic functions with characteristic function
p(x, y). The p-analitic functions were applied to studying three-dimensional flows with free
water surface in [Goman, 1972; Popov, 1979; Goman & Popov, 1980; Goman, 1984]. A
systematic use of generalized analytic functions for various axisymmetric flows has been
done in [Terentiev, 1993; Zhitnikov, 1993; Dimitrieva & Terentiev, 2002; Terentiev &
Dimitrieva, 1998; Terentiev et al., 1997].
A.4.2. The Pologiy Transformation
For characteristic function p = x k Pologiy gave an integral transformation that defines the
dependence of p-analitic function F(z) on ordinary analytic function f(z) = u(x, y) + v(x, y).
For characteristic function p = y k , Pologiy’s transformation can be presented as
1
F ( z ) = ∫ f ( z ′)[( −1) k y1− k + i ( z ′ − x)]g k − 2 ( z ′)dz ′,
2Γ (A4.3)
g ( z ′) = ( z ′ − z )( z ′ − z ) ,
where line Γ joints the self-conjugate variables z = x − iy and z = x + iy . Functions v( x, y )
and V ( x, y ) on the real x-axis are assumed to be equal to zero, i.e.
539
Appendices
v( x,0) = 0, V ( x,0) = 0 . (A4.4)
− (1+ ε )
If the analytic function for z → ∞ satisfies condition f ( z ) ≈ z where ε is an
arbitrary positive value, then line Γ can be arbitrary deformed and complex integral
transformation (A4.3) can be written as two real transformations
x + i∞
U ( x, y ) = (− y )1− k Im ∫z
f ( z ′) g k − 2 ( z ′)dz ′,
x + i∞
(A4.5)
∫ f ( z ′)( z ′ − x) g ( z ′)dz ′ .
k −2
V ( x, y ) = − Im
z
Function g ( z ′) is not uniquely defined; therefore a single-value branch cut must be chosen
for it, for instance, Im g ( z ′) > 0 within interval ( z , x + i∞ ). One can determine after
substituting new variable z ′ = z + iη into integrals (A4.5) and then differentiating, that
functions (A4.5) satisfy equations (A4.2). The partial derivatives are written form as
x + i∞
∂U
= (− y )1− k Im ∫ f ′( z ′) g k − 2 ( z ′)dz ′,
∂x z
x + i∞
(A4.6)
∂V
= − Im ∫ f ′( z ′)( z ′ − x) g ( z ′)dz ′ .
k −2
∂x z
Since function g ( z ′) approaches difference ( z ′ − x )

as ordinate y approaches zero ( y → 0) , then the integrand
Fig.A4.1. Axisymmetric for function U ( x, y ) in (A4.7) have a single pole for k = 1
domain with leading part u ( x,0) /( z ′ − x) . Consider two closed
curves at both sides of straight line x = const. Integral along
each curves vanish because the integrand is an analitic function. Deforming these curves into
the two other curves, as shown by points in Fig.A4.1, one can obtain two formals for the
same value of function U ( x, y ) on the real axis:
π dz ′
U ( x,0) = u ( x,0) − Im ∫ f ( z ′) , (A4.7)
2 C1
z′ − x
π dz ′
or U ( x,0) = − u ( x,0) + Im ∫ f ( z ′) . (A4.8)
2 C2
z′ − x
Here curves C1 and C2 are the boundaries of the definitional domain, for instance, the
generatrix for axisymmetric bodies. Functions U ( x, y ) and V ( x, y ) for k = 1 could be
considered as a potential and a stream function of some axisymmetric flows. Therefore,
transformations (A4.5) convert an axisymmetric problem to determine ordinary analitic
functions. Further, we shall deal with the p-analitic functions of the characteristic
y 2 m +1 , m = 0,1, 2, . If point z belongs to boundary C, then integrals (A4.6) – (A4.8)
could be reduce to integrating along the boundary
z z
1
U = − 2 m Im ∫ f ( z ′) g 2 m −1 ( z ′)dz ′, V = Im ∫ f ( z ′)( z ′ − x) g 2 m −1 ( z ′)dz ′ . (A4.9)
y x0 x0
Point x0 could be the extreme left (or right) intersection of the boundary with the x-axis. For
540
Appendices
the partial derivative one can obtain similar expressions as (A4.7) – (A4.9), in which
function f ( z ′) should be replaced with its derivative f ′( z ′) = df ( z ′) / dz ′ .
A.4.3. Behavior of the p- Analytic Functions

Consider now right extreme intersection point x0 so that interval ( x0, ∞ ) belongs to the
defined domain. Approach point z to x0 along the boundary. Let be the angle between the
tangent of the boundary at point x0 and the x-axis equal to α . Then variable z could be
presented as z = x0 + ε exp(iα ) for ε → 0 . The first integral in (A4.9) is transformed to
α
2m −1/ 2
U ( x0 ,0) = −u ( x0 ,0)
sin α 0
2m ∫ (cosτ − cos α ) m −1/ 2 cos(m + 12 )τ dτ ,
which are expressed analytically [Gradstain & Ryzhik, 1962] as

2m −1/ 2 π
U ( x0 ,0) = −u ( x0 ,0) 2 m sin m αΓ(m + 12 ) Pm− m (cos α )
sin α 2
π sin m α
Using the formulas Γ(m + 12 ) = m (2m − 1)!! , Pm− m (cos α ) =
2 (2m)!!
one obtains a final asymptotic expression as
π (2m − 1)!!
U ( x0 ,0) = − u ( x0 ,0) . (A4.10)
2 (2m)!!
It should be noted that the asymptotic behavior of function (A4.10) does not depend on a
way of approaching point x0 and, therefore, valid on any points x of interval ( x0 , ∞ ).
A.4.4. Three-Dimensional Domain

Laplace’s three-dimensional equation is expressed in cylindrical coordinates on the form
∂ 2ϕ ∂ 2ϕ 1 ∂ϕ 1 ∂ 2ϕ
+ 2 + + = 0. (A4.11)
∂x 2
∂y y ∂y y 2 ∂θ 2
If function ϕ is presented in trigonometric sequence as
∞ ∞
ϕ = ∑ ϕm ( x, y )cos mθ + ∑ g m ( x, y )sin mθ , (A4.12)
m=0 m =1
then functions ϕm and g m satisfy the same differential equation

∂ 2ϕ m ∂ 2ϕ m 1 ∂ϕ m m 2ϕ m
+ + − 2 =0. (A4.13)
∂x 2 ∂y 2 y ∂y y
Introduce auxiliary function U m ( x, y ) as ϕ m = y mU m , so that it satisfies the differential
∂ 2U m ∂ 2U m 1 + 2m
equation of the form + + Um = 0 . (A4.14)
∂x 2 ∂y 2 y
Introduce another function denoted as conjugate function, which satisfies the equation
∂ 2V m ∂ 2Vm 1 + 2m
+ − Vm = 0 . (A4.15)
∂x 2 ∂y 2 y
It is easy to see that functions U m ( x, y ) and Vm ( x, y ) satisfy the p-analytical conditions as
541
Appendices
∂U m 1 ∂V ∂U m 1 ∂V
follows: = 1+ 2 m m , = − 1+ 2 m m . (A4.16)
Complex function Fm ( z ) = U m + iVm is a p-analitic function with the characteristic function
p = y1+ 2 m and can be expressed using integral transformation (A4.3) through ordinary
analytic function f m ( z ) . Therefore, the three-dimensional problem of the potential flow of
an inviscid and incompressible fluid can be converted to find ordinary analytic functions of
complex variable.
A.4.5. Flow around Axisymmetric Bodies
For example, consider an arbitrary flow around an axisymmetric body. In this case, potential
(A4.12) is presented as ϕ = ϕ0 ( x, y ) + ϕ1 ( x, y ) cosθ . (A4.17)
The first item determines the longitudinal flow of axisymmetric body and the second one –
the transverse flow. Corresponding functions U 0 and U1 are determined by (A4.5),
x + i∞
f 0 ( z ′)
respectively, U 0 ( x, y ) = Im ∫ z
g ( z ′)
dz ′ , (A4.18)
x + i∞
1
U1 ( x, y ) = − Im ∫ f1 ( z ′) g ( z ′)dz ′ , (A4.19)
y2 z
The total potential of longitudinal and transversal flows are obtained as
Φ 0 = x + U 0 ( x, y ) and Φ1 = [ y + yU1 ( x, y )]cos θ . (A4.20)
3 3
R 2⎛R⎞
It is easy to verify that functions f0 ( z) = −and f1 ( z ) = − ⎜ ⎟ (A4.21)
πz 2
π⎝z⎠
determine exact potentials of a flow around a sphere with radius R. The longitudinal and
transverse velocities of a flow are assumed to be unity. For arbitrary axisymmetric body,
unknown functions f 0 ( z ) and f1 ( z ) should be found from integral equations, which appear
satisfying kinematic conditions on the generatrix. These equations can be solved numerically
by conformal mapping of the flow domain onto exterior of unit circle of the ζ-plane. The
transformation z(ζ) and unknown functions f0(ζ) and f1(ζ) can be found by expansion into
series with real coefficients as follows:
∞ ∞ ∞
a b c
z (ζ ) = ∑ nn , f 0 (ζ ) = ∑ nn+1 , f1 (ζ ) = ∑ nn+ 2 . (A4.22)
n =−1 ζ n =1 ζ n =1 ζ
Numerical results for the ellipsoid were obtained in a collocation manner keeping finite
numbers of terms, 5 and 10, in the last two series of (A4.22). The numerical data [Terentiev,
1993] have been compared with exact solutions given in [Loyziansky, 1959]; the tolerance
for velocity and speed potential was about 0.001 for both longitudinal and transverse flows.
App. 5. Calculation of Integrals and Special Functions

One of basic phases in numerical modeling is numerical calculation of integrals, especially,
singular integrals and special functions. Currently all kinds of numerical methods are
elaborated, some of them could be computed rather easily. Some of the methods are included
in computer system and could be run by all users. Ordinary non-singular integrals can be
numerically calculated, using Gauss quadrature [Krylov, 1967]. The quadrature do not
542
Appendices
include limits of integration, hence it can be used for numerical calculations of functions
with integrable singularities such as logarithmic, or fractional exponent, or even Couchi’s
singularity. However, accuracy and computer time could not always satisfy the user. It is
therefore important to find an analitic expression of the singular integral, or first to transform
the integrand and so forth. Some of such techniques are considered below.
A.5.1. Matrix Analytic Expressions in BEM for Constant Elements
Let ln , α n and ( X n , Yn ) be a length, inclination angle, and the middle point of the nth
element, respectively. Let ξ and η be the local coordinates with origin ( X n , Yn ) and ξ is
directed toward the element. If ( X k , Yk ) is any point in the fluid domain then the Green’s
function in local coordinates ξ , η is
⎛ ⎞
ln ⎜ 2 ⎟,
G (ξ ,η , ζ kn ) =
ln (A5.1)
4π ⎜ l (ξ − ξ )2 + (η − η ) 2 ⎟
⎝ n kn kn ⎠
where ζ kn = ξ kn + iη kn is coordinate of point ( X k , Yk ) in the local coordinates. Its real and
2
ξ kn = ⎡⎣( X k − X n ) cos α n + (Yk − Yn ) sin α n ⎤⎦ ,
ln
imaginary parts are as follows: (A5.2)
2
ηkn = ⎣⎡(Yk − Yn ) cos α n − ( X k − X n ) sin α n ⎤⎦ ,
ln
x +x y + yn x − xn −1 y − yn −1
X n = n −1 n , Yn = n −1 , cos α n = n , sin α n = n ,
where 2 2 ln ln (A5.3)
ln = ( xn − xn −1 ) 2 + ( yn − yn −1 ) 2 .
As far as segment [ zn −1 , zn ] is mapped into section [– 1, 1] on the real ξ -axis then the
normal to the section is opposite to the imaginary η -axis, thus normal derivative
Gn ( z , z ′) = −2∂G (ξ ,η ) / ln ∂η . On the ξ -axis, η = 0 , the Green’s function and its normal
derivative are as follows:
⎛ ⎞
l 2 ηkn
G (ξ , ζ kn ) = n ⎜ ln ⎟ , G (ξ , ζ ) = − (A5.4)
4π ⎜ l (ξ − ξ ) + η 2 ⎟ ( )
n kn
( )
2
2
2π ξ − ξ + η 2
⎝ n kn kn ⎠ kn kn
Integrating functions (A5.4) along the ξ -axis in interval [– 1, 1], one obtains the terms of
matrix (15.5) as follows:
⎧ 0, ηkn = 0, k ≠ n;
⎪
⎪ 1 ⎡ ⎛ 1 − ξ kn ⎞ ⎛ 1 + ξ kn ⎞ ⎤
Akn = ⎨− ⎢arctan ⎜ ⎟ + arctan ⎜ ⎟ ⎥ , ηkn ≠ 0, k ≠ n; (A5.5)
⎪ 2π ⎣⎢ ⎝ ηkn ⎠ ⎝ ηkn ⎠ ⎦⎥
⎪1/ 2, ηkn = 0, k = n;
⎩
ln ⎡ ⎛ ln ⎞ (1 − ξ kn ) ln pkn + (1 + ξ kn ) ln qkn ⎤
Bkn = ⎢1 − ln ⎜ ⎟ − + πηkn akn ⎥ , (A5.6)
2π ⎣ ⎝2⎠ 4 ⎦
543
Appendices
pkn = (1 − ξ kn ) + ηkn2 , qkn = (1 + ξ kn ) + η kn2 .

2 2
where (A5.7)
The terms of matrix B coincide with expression (A5.6), if ( X k , Yk ) are coordinates of kth
nodal point. The terms of matrix A coincide with akn but the diagonal terms are equal to
Akk = akk + 1/ 2 .
A.5.2. Linear Elements

Analytical expressions can be obtained by considering two adjoining elements, n, and, n + 1,
and two local coordinates should be considered. Then the terms of matrices a and B
calculated as sum of integrals are
1 1
akn = akn ′ = ∫ N1 (ξ )Gn (ξ , ζ kn )d ξ , ak′′,n +1 = ∫ N 2 (ξ )Gn (ξ , ζ k , n +1 )d ξ ;
′ + ak′′,n +1 , akn (A5.8)
−1 −1
1 1
Bkn = bkn′ + bk′′, n +1 , bkn′ = ∫ N1 (ξ )G (ξ , ζ kn )d ξ , bk′′,n +1 = ∫ N 2 (ξ )G (ξ , ζ k ,n +1 )d ξ . (A5.9)
−1 −1
Analytical expressions of the integrals are as follows:
1 + ξ kn η
′ =
akn akn − kn ( ln pkn + ln qkn ) , ak′′,n +1 = ak ,n +1 − ak′ ,n +1 ; (A5.10)
2 8π
1 + ξ kn lξ ⎛ l ⎞ l
bkn′ = bkn − n kn ⎜ 2ln n − 1⎟ + n ( qkn ln qkn − pkn ln pkn ) , bk′′, n +1 = bk , n +1 − bk′′, n +1 . (A5.11)
2 8π ⎝ 2 ⎠ 32π
The diagonal terms of matrix A can be calculated using off-diagonal elements [Wroubel,
1987]. Identity (13.14) is present in a matrix form as
AI = 0, (A5.12)
where I is a unit vector. As above, the equation indicated the sum of all the elements of A in
a row should be zero, so that the diagonal elements can be calculated using all the remaining
N
elements of the row Akk = −∑ Akn , k = (1, N ) (A5.13)
n =1
n≠ k
The last identity is very useful for numerical calculations of the matrices.
A.5.3. Theta Functions
There are many designations and presentations of theta-functions. Jacob’s symbols
[Whittaker & Watson, 1927] are most convenient for applications. Theta functions are
determined on the rectangle with sides π and πτ i , and therefore, they depend on parameter
τ also. Unlike [Whittaker & Watson, 1927] we use real parameter τ . There are four kind of
theta-functions: ϑ1 ( z | τ ), ϑ2 ( z | τ ), ϑ3 ( z | τ ), ϑ4 ( z | τ ) , which are connected each other by a
number of relationships. For numerical calculations, it is preferable to present theta-
functions in a form of infinite products, because the factors contain variable z only as a
trigonometric function cos 2z . Functions ϑ1 ( z | τ ) and ϑ2 ( z | τ ) contain also factors sin z
and cos z , respectively. The infinite product of ϑ3 ( z | τ ) is
∞
ϑ3 ( z,τ ) = G∏ (1 + 2q 2 n −1 cos 2 z + q 4 n − 2 ) , (A5.14)
n =1
544
Appendices
∞
where q = e −πτ , G = ∏ (1 − q 2 n ) . (A5.15)
n =1
Peak value of cos2z in the rectangle is about e2πτ , hence the factors of product (A5.14) are
about q 2 n −3 . Parameter q < 0.05 for τ ≥ 1 , so that already third factor in (A5.14) differs from
unity by a small value about 10−5 and the fourth factor by 10−7 . Therefore, it is enough to
keep in the product the first three or four factors. Other three theta-functions could be
presented also as infinite products, but they can be calculated using expression (A5.14) with
⎛ π πτ i ⎞
ϑ1 ( z | τ ) = −iq1/ 4 eizϑ3 ⎜ z + + |τ ⎟;
⎝ 2 2 ⎠
relationships , (A5.16)
⎛ πτ i ⎞ ⎛ π ⎞
ϑ2 ( z | τ ) = q e ϑ3 ⎜ z +
1/ 4 iz
| τ ⎟ ; ϑ4 ( z | τ ) = ϑ3 ⎜ z + | τ ⎟
⎝ 2 ⎠ ⎝ 2 ⎠
For τ < 1 the imaginary Jacob’s transformations should be used. Thus the theta-functions
could be calculated together with (A5.16)by formula
⎧ϑ3 ( z ,τ ) , τ ≥1
⎪ − z 2 / πτ
ϑ3 ( z,τ ) = ⎨ e ⎛ zi 1 ⎞ (A5.17)
⎪ ϑ3 ⎜ , ⎟ , τ <1
⎩ τ ⎝τ τ ⎠
A.5.4. Elliptic Integrals
Many problems can be solved using full elliptic integrals of the first and second orders
π /2 π /2
dα
K (k ) = ∫ and E ( k ) = ∫ 1 − k 2 sin 2 α dα . (A5.18)
0 1 − k 2 sin 2 α 0
The second integral is bounded for parameter k ∈ [0,1] while the first integral is unlimited as
k → 1 . Elliptic integrals can be expanded into various asymptotic series [Gradstain &
π ⎛⎜ ⎛1− m ⎞ ⎞
2n
M
Ryzhik, 1962], namely: K1 ( m, M ) = 1 + ∑ an ⎜ ⎟ ⎟, (A5.19a)
1 + m ⎜ n =1 ⎜⎝ 1 + m ⎟⎠ ⎟
⎝ ⎠
1 ⎛ 16 ⎞ ⎛ M
⎞ M n
mn
K 2 (m, M ) = ln ⎜ ⎟⎜1 + ∑ an m n ⎟ − 2∑ an ∑ , (A5.19b)
2 ⎝ m ⎠ ⎝ n =1 ⎠ n =1 k =1 2k (2k − 1)
π (1 + m ) ⎛⎜ ⎛1− m ⎞ ⎞
2n
M
an
E1 (m, M ) = 1− ∑ 2 ⎜
⎜ ⎟ ⎟, (A5.20a)
4 ⎜ n =1 (2n − 1) ⎝ 1 + m ⎟⎠ ⎟
⎝ ⎠
M
2n ⎛ 1 ⎛ 16 ⎞ 1 n
2 ⎞
E2 ( m, M ) = 1 + ∑ an m n ⎜ ln ⎜ ⎟ + −∑ ⎟ , (A5.20b)
n =1 2n − 1 ⎝ 2 ⎝ m ⎠ 2n(2n − 1) k =1 2k (2k − 1) ⎠
2
⎡ (2n)! ⎤
where m = 1 − k , an = ⎢ 2 n ⎥ .
2
(A5.21)
⎣ 2 n !⎦
The expression of an leads to overfilling for a large number of n due to the factorials. It can
2
⎛ 2n − 1 ⎞
be calculated by a recurrent formula a0 = 1, an = an −1 ⎜ ⎟ (A5.21a)
⎝ 2n ⎠
545
Appendices
Elliptic integrals (A5,18) are obtained from equations (A5.19) and (A5.20) by limit M → ∞ .
Series (A5.19a) and (A5.20a) converge at m ∈ (0,1] but number M should be increased for
sufficient accuracy if parameter m approaches zero. In that case, series (A5.19b) and
(A5.20b) should be used. Computing shows that the elliptic integrals can be calculated as
⎧ K (m, M ) if m ≥ 0.2
follows: K ( m, M ) = ⎨ 1 ; (A5.22)
⎩ K 2 (m, M ) if m ∈ (0, 0.2)
⎧ E ( m, M ) if m ≥ 0.1
E ( m, M ) = ⎨ 1 . (A5.23)
⎩ E2 (m, M ) if m ∈ (0, 0.1)
These formulas give ratio error mostly as 10−5 even for M = 3. This is usually enough. If a
problem needs higher accuracy, number M should be increased. Elliptic integrals can be
calculated by approximated polynomials [Banerjee & Butterfield, 1981] as well:
M M
K (m) ≈ ∑ [an m n + bn m n ln(1/ m), E (m) ≈ 1 + ∑ [cn m n + d n m n ln(1/ m)] . (A5.24)
n =0 n =1
The coefficients for M = 4 are given in Table 16.1 after [Banerjee & Butterfield , 1981].
Table A5.1. Coefficients of the appreciated polynomials.
n an bn cn dn
0 1.38629436112 0.5 - -
1 0.09666344259 0.12498593597 0.44325141463 0.24998368310
2 0.03590092383 0.06880248576 0.06260601220 0.09200180037
3 0.03742563713 0.03328355346 0.04757383546 0.04069697526
4 0.01451196212 0.00441787012 0.01736506451 0.00526449639
Dependencies of ratio error ( ε ) on parameter m corresponding to equations (A5.24)

as well as to equations (A5.22) and (A5.23) are given in Fig.A5.1. It is shown that the
approximated polynomials (A5.24) ensure the ratio error at the at least 10−8 . Equations
(A5.22) and (A5.23) by M = 8 provide also the same
ratio error but for smaller interval around the joint points
(m = 0.2) and (m = 0.1); outside intervals the error is
much less. Since the error is small and equations (A5.24)
need relatively little computer time, thus the
approximated polynomials are convenient for using in
the numerical hydrodynamics. Equations (A5.22) and
(A5.23) should be applied to theoretical analysis and
also to computing by high accuracy.
Fig.A5.1.
Dependence of the ratio on parameter m
A.5.5. Singular Integrals
The Cauchy’s singularities as well as others appear in the theory of boundary value
problems. There are many approaches for computing integrals of such singularities. Here
another approach is proposed based on the variance property that the value problem solution
for a slightly changing boundary differs little from the exact one [Lavrent’ev & Shabat,
1965]. Namely, if δ is an area between the given and changed boundaries then the variance
546
Appendices
solution will differ from the exact one also as δ only. Therefore, changing the boundary
curve by a small value one can cancel the singularity and apply standard integral sums. For
b
f (s)
example, consider an integral J ( x) = ∫ dx (A5.25)
a
s−x
where f(s) is a function with any integrated singularities. Integral J c ( x) = Re ( J ( x + δ i ) )
should differ from J(x) by a small value of order O(δ ) outside infinitesimal vicinities of
singular points of function f(s). But then integral Jc(x) has no singularities and can be
computed by any standard quadrature sum. Some singular integrals from [Gachov, 1963]
and their exact expressions Ja(x) and absolute error ε = max | J a ( x) − J c ( x) | , computed by
δ=10−7 are presented in Table A5.2
Table A5.2. Integrals, their exact expressions, and approaching errors on shown intervals.
J ( x) J a ( x) ε x ∈ ( a, b)
1
1 1− t t + 3
2
7
π∫
dt − x2 + x − 10−4 (−0.97,1)
−1
1+ t t − x 2
1
1 t4 x
π∫
dt x3 +
−1 1 − t (t − x)
2
2 (−0.99, 0.99)
2π
1 ⎛ t⎞ t−x π −x
2π ∫ ln ⎜⎝ 2sin 2 ⎟⎠ ctg
0
2
dt
2 10−5 (0.01π ,1.99π )
2(arccos x − π ) + 3arccos x
1
1 ln[(1 − t ) 2 (1 + t )3 ]
π∫
dt (−0.97, 0.97)
1 − t (t − x)
2
1 − x2 10−4
−1
1 ⎛ 1 ⎞
1 1
1 ⎛ 2 1− t ⎞
∫−1 t − x ⎝ −∫1 s − t ds ⎟⎠dt
⎜ ⎜ ln
2 ⎝ 1+ t
−π 2 ⎟
⎠ 10−4 (−0.99, 0.99)
A.5.6. Approximation of Boundary-Value Problems

The approximation can be applied to computing boundary
value problems. General analytical solutions of value
problems could be easily written but obtaining particular
solutions is connected with solving a set of nonlinear
equations. Applying the approximation allows formalizing
computations and obtaining numerical results with high
accuracy. For example, consider a conformal mapping of
upper half-plane Im z ≥ 0 onto inner rectangle
− K , K , K + iK ′, − K + iK ′ , (Fig.A5.2). The symbols are as Fig.A5.2. Conformal map
in [Lavrent’ev & Shabat, 1965]. The boundary values of of rectangle (a) onto upper
transformation w = u + iv , due to Fig.A5.2 are piecewise half-plane (b).
constant, thus the transformation can be found from mixed solution (A1.8)
w( z , a ) : g ( z , a) F ( z , a) + K ′i , (A5.26)
547
Appendices
where g ( z , a) = ( z + 1)( z − a ) /( z − 1)( z + a) ;

1 ⎡ ⎛ ⎞ ⎤
a −1 1
dt dt dt
F ( z, a) = ⎢ ⎜∫
K − ∫ ⎟ − iK ′ ∫ ⎥.
π i ⎢⎣ ⎝ 1 g (t , a)(t − z ) − a g (t , a)(t − z ) ⎠ −1
g (t , a )(t − z ) ⎥⎦
The transformation has unknown parameter a, which can be found from the condition of
non-singularity at point z=1, i.e. F(1,a)=0. Calculating the integral using a quadrature sum
can yield the singularity. The approximation is wc ( z ) = w( z + iδ , a) , and parameter a is
found from equation F (1 + iδ , a ) = 0 ; δ = 10−5 . The exact analytical transformation is
dτ
z
K
wa ( z ) = ∫
K ( k ) 0 (1 − τ )(1 − τ 2 / a 2 )
2
. (A5.27)
Fig.A5.3.
Comparison of computing by approximated
function and by exact formulas (points).
Fig.A5.3 shows dependence of function

u(x)=Rewc(x) for low side on interval −1<x<1
computed by the approximation; the points are
computed by analytical formulas (A5.27).
Calculations show that error ε =| wa ( x) − Re wc ( x) |
within interval (−1, 0.05) doesn’t exceed 10−3.
A.5.7. Cubature Formula for Triangular Domain
If function f (ξ ,η ) is continuous and differentiable of the given order in triangular domain,
the double integral of the function with respect to this domain can be calculated by the
1 1−η n
cubature formula J =∫ ∫ f (ξ ,η ) dξdη = ∑ Ai f (ξi ,ηi ) . The coordinates and weighting
0 0 i =1
coefficients obtained in [Hammer et al., 1956] are given in Table A.5.3.
Table A.5.3. Parameters to Hammer’s cubature formula.

n error ξi ηi Ai
2
1 O(h ) 1/3 1/3 ½
½ ½ 1/6
3 O(h3) 0 ½ 1/6
½ 0 1/6
0.333,333,33 0.333,333,33 0.112,500,00
0.797,426,99 0.101,286,51 0.062,969,59
0.101,286,51 0.797,426,99 0.062,969,59
6
7 O(h ) 0.101,285,51 0.101,285,51 0.062,969,59
0.059,715,87 0.470,142,06 0.066,197,07
0.470,142,06 0.059,715,87 0.066,197,07
0.470,142,06 0.470,142,06 0.066,197,07
548
Appendices
App. 6. Velocity Field as a Distribution of Vortices Over the Boundary

Consider the steady flow of an unbounded ideal fluid past a closed two-dimensional body.
Such a flow may be described by a velocity potential Φ . One may then introduce a
disturbance velocity potential, φ , such that Φ = U ⋅r +φ (A6.1)
where U represents the freestream velocity and r = xi + yj is the position vector. This
disturbance potential satisfies the condition that φ → 0 as r → ∞ (A6.2)
Hence, from Green’s second identity it can be shown that φ may be represented by a
distribution over the fluid boundary, ∂D , of sources and normal dipoles of the form
1 ⎪⎧ ∂φ ∂ ⎡⎣ln ( R ) ⎤⎦ ⎪⎫
φ= ∫ ⎨ ln ( R ) − φ ⎬ ds (A6.3)
2π ∂D ⎩⎪ ∂n ∂n ⎭⎪
If the circulation about the body is zero, then φ may be expressed as a unique distribution of
solely sources or a unique distribution of solely dipoles (see [Lamb, 1932, p59-61).
If, however, the circulation about the body is nonzero, then the velocity potential
becomes multi-valued and a barrier or auxiliary flow boundary must be introduced into the
fluid to render it single-valued once again (see Fig.A6.1). In fact, the circulation about the
body is exactly equal to the jump in potential across this barrier. Thus, if φ is to be
expressed in terms of boundary distributions of
sources and normal dipoles, then there must
exist a constant distribution of normal dipoles on
the barrier with the density of the dipole moment
equal to the circulation.
Fig.A6.1.
Auxiliary flow boundary, after [Uhlman, 1983].
It is no longer possible to express φ solely in terms of sources. It is, however,

possible to use Lamb’s interior flow arguments to express φ in terms of normal dipoles
1 ∂ ⎡ ln ( R ) ⎤⎦ Δφ ∂ ⎡⎣ln ( R ) ⎤⎦
alone. Thus φ =− ∫
2π ∂D
( )
φ −φ ⎣
∂n
ds +
2π ∫B ∂n
ds (A6.4)
∂φ ∂φ
where φ is the potential in the domain D c such that = on ∂D (A6.5)
∂n ∂n
and Δφ is the jump in potential across the barrier. One may now introduce the multi-valued
function Θ = arc tan ( y / x ) (A6.6)
This is the conjugate harmonic function of ln ( R ) and hence the Cauchy-Riemann equations
∂ ⎡⎣ln ( R ) ⎤⎦ ∂Θ
state that = (A6.7)
∂n ∂s
1 ∂Θ Δφ ∂Θ
Equation (A6.4) may therefore be written as φ = −
2π ∫ (φ − φ ) ∂s ds + 2π ∫ ∂s ds
∂D B
(A6.4)
1 ∂
One may then perform an integration by parts to yield φ=
2π ∫ ∂s (φ − φ ) Θds ,
∂D
(A6.9)
549
Appendices
which expresses the velocity potential in terms of a distribution of vortices over the body
boundary ∂D . The barrier now serves the purpose of defining which branch of the multi-
valued function Θ is under consideration.
The simplest manner of demonstrating that the flow about a body with nonzero
circulation can be represented in terms of a boundary distribution of vortices alone is to
consider the stream function Ψ . If U and V denote the x and y components of the
freestream, respectively, then one may define a disturbance stream function, ψ , by
Ψ = (Uy − Vx ) + ψ (A6.10)
This disturbance stream function satisfies the conditions necessary to apply Green’s second
1 ⎧⎪ ∂ψ ∂ ⎡ln ( R ) ⎤⎦ ⎫⎪
identity and hence may be written as ψ = ∫ ⎨ ln ( R ) − ψ ⎣ ⎬ ds (A6.11)
2π ∂D ⎩⎪ dn ∂n ⎪⎭
The above representation expresses Ψ in terms of boundary distributions of vortices and
tangential dipoles. If there is no flux from the body boundary, then the body boundary is a
streamline and the stream function is single valued. The arguments of Lamb concerning
interior flows may then be applied. In particular, consider an interior stream function ψ . In
1 ⎧⎪ ∂ψ ∂ ⎡⎣ ln ( R ) ⎤⎦ ⎫⎪
the region D , ψ satisfies 0=
2π∂D ⎩
∫ ⎨⎪ dn ln ( R ) −ψ∂n
⎬ ds
⎪⎭
(A6.12)
where ∂ ∂n = ∂ ∂n . If one now requires that ψ = ψ on ∂D (A6.13)

then equations (A6.11) and (A6.12) may be combined to yield
1 ⎧ ∂ψ ∂ψ ⎫
⎬ ln ( R ) ds in D
2π ∂∫D ⎩ ∂n ∂n ⎭
ψ= ⎨ − (A6.14)
which was to be shown.

It should be noted that whether or not the velocity potential (or stream function) is
multi-valued, the associated velocity field is single-valued. Hence, since the present work
deals only with the velocities, the question of multi-valuedness does not arise.
App. 7. 2D Mass and Momentum Control Volume Analyses

The control volume for this analysis with the associated boundary designations is depicted in
Fig.A7.1. Employing this control volume plus the surface at infinity, C∞ (where
C∞ = lim CR and CR is the contour at radius R ), we may perform a control volume analysis
R →∞
of the conservation of mass using ∫ ρ ( n j u j ) dS = 0 (A7.1)

Cb + Cc
+ C j + C∞
Assuming, for large r , that the total potential may be approximated as

Q Γ ⎛1⎞
Φ ≈ U∞ x + ln ( r ) + θ + O⎜ ⎟ (A7.2)
2π 2π ⎝r⎠
where Q is the source strength seen in the far field and Γ is the circulation, we find that
Q x Γ y Q y Γ x
u = U∞ + − ; v= + (A7.3)
2π r 2
2π r 2
2π r 2
2π r 2
550
Appendices
Q Γ
and ur = U ∞ cos (θ ) + ; uθ = −U ∞ sin (θ ) + (A7.4)
2π r 2π r
we find that the contributions to equation (A7.1) then become
∫ ρ ( n j u j ) ds = 0; ∫ ρ ( n j u j ) ds = 0; ∫ ρ ( n j u j ) ds = ρU jet h jet ∫ ρ ( n j u j ) ds = ρ Q
Cb Cc Cj C∞
(A7.5)
where U jet = U ∞ 1 + σ . Hence we find that Q = −h jetU jet (A7.6)
n
C∞
Fig.A7.1.
Axisymmetric
control
volume with
boundary
Cc
designations.
n Cj
Cb
inflow
Cc
Performing
the momentum analysis using ∫ ρ ui ( n j u j ) ds + ∫ ni ( p − pc ) ds = 0 (A7.7)
Cb + Cc Cb + Cc
+ C j + C∞ + C j + C∞
then yields the integrals

∫ ρ u ( n j u j ) ds = 0; ∫ ρ u ( n j u j ) ds = 0; ∫ ρ u ( n u ) ds = − ρU
2
j j jet h jet ;
Cb Cc Cj
(A7.8)
3
∫ ρ u ( n j u j ) ds = ρU ∞ Q
C∞ 2
and ∫ ρ v ( n u ) ds = 0; ∫ ρ v ( n u ) ds = 0; ∫ ρ v ( n u ) ds = 0; ∫ ρ v ( n u ) ds =
Cb
j j
Cc
j j
Cj
j j
C∞
j j
1
2 ρU ∞ Γ ; (A7.9)
p − pc = ( p∞ − pc ) + 12 ρ ⎡⎣U ∞2 − ( ur2 + uθ2 ) ⎤⎦

and since ⎡ UQ (A7.10)
⎛ Q ⎞ ⎛ Γ ⎞ ⎤
2 2
UΓ
= ( p∞ − pc ) − 12 ρ ⎢ cos (θ ) − sin (θ ) + ⎜ ⎟ ⎜+ ⎟ ⎥
⎣⎢ π r πr ⎝ 2π r ⎠ ⎝ 2π r ⎠ ⎦⎥
551
Appendices
∫ n ( p − p ) ds = D; ∫ n ( p − p ) ds = 0; ∫ n ( p − p ) ds = 0;
Cb
x c
Cc
x c
Cj
x c
we find that (A7.11)

∫ nx ( p − pc ) ds = − 12 ρU ∞ Q
C∞
∫ n ( p − p ) ds = L; ∫ n ( p − p ) ds = 0; ∫ n ( p − p ) ds = 0;
Cb
y c
Cc
y c
Cj
y c
and (A7.12)
∫ n ( p − p ) ds =
C∞
y c
1
2 ρU ∞ Γ
so that L = − ρU ∞ Γ; D = ρ h jetU 2jet − ρU ∞ Q = ρ h jetU ∞2 ⎡⎣(1 + σ ) + 1 + σ ⎤⎦ (A7.13)
Γ ⎛ h jet ⎞ ⎡ 1 ⎤
or, in dimensionless form Cl = −2 ; Cd = 2 ⎜ ⎟ (1 + σ ) ⎢1 + ⎥, (A7.14)
U∞c ⎝ c ⎠ ⎣ 1+σ ⎦
where we note that h jet = h jet (σ ,α ; f ) where f represents the shape of the hydrofoil.
App. 8. Axisymmetric Mass and Momentum Control Volume Analyses

The control volume for this analysis n
with the associated boundary
designations is depicted in Fig.A8.1. SR
R
Fig.A8.1. Sc
Axisymmetric control volume with
boundary designations, after
[Uhlman, 2006]. Sb n
d dj
Inflow Sj
Employing this control volume plus

the surface at infinity, S∞ (where
Control Volume
S∞ = lim S R and S R is the surface at
R →∞
radius R ), we may perform a control

volume analysis of the conservation
of mass using ∫∫ ρ ( n j u j ) dS = 0
Sb + Sc
(A8.1)
+ S j + S∞
Assuming, for large r , that the total potential may be approximated as

Q ⎛1⎞
Φ ≈ U∞ x − + O⎜ 2 ⎟ (A8.2)
4π r ⎝r ⎠
where Q is the source strength seen in the far field, then the velocity components are
Q x Q y Q z
u = U∞ + ; v= ; w= (A8.3)
4π r 3 4π r 3 4π r 3
or, employing the spherical coordinates x = r cos (φ ) , y = r sin (φ ) cos (θ ) , z = r sin (φ ) sin (θ )
552
Appendices
ur = U ∞ cos (φ ) + Q /(4π r 2 ) : uφ = −U ∞ sin (φ ) ; uθ = 0 . (A8.4)

We then find that the contributions to equation (A8.4) then become
π
∫∫ ρ ( n u ) dS = 0; ∫∫ ρ ( n u ) dS = 0; ∫∫ ρ ( n u ) dS = ρU
Sb
j j
Sc
j j
Sj
j j jet
4
d 2jet ; ∫∫ ρ ( n j u j ) dS = ρ Q (A8.5)
S∞
π
where U jet = U ∞ 1 + σ . Hence we find that Q = −U jet d 2jet (A8.6)
4
Performing the momentum analysis using
∫∫ ρ ui ( n j u j ) dS +
Sb + Sc
∫∫
Sb + Sc
ni ( p − pc ) dS = 0 (A8.7)
+ S j + S∞ + S j + S∞
p − pc = ( p∞ − pc ) + 12 ρ ⎡⎣U ∞2 − ( ur2 + uθ2 + uφ2 ) ⎤⎦

where ⎡ U ∞Q (A8.8)
⎛ Q ⎞ ⎤
2
= ( p∞ − pc ) − ρ ⎢ cos (φ ) + ⎜
1
2 ⎟ ⎥
⎢⎣ 2π r ⎝ 4π r ⎠ ⎥⎦
2 2
then yields the integrals

π
∫∫ ρ u ( n u ) dS = 0; ∫∫ ρ u ( n u ) dS = 0; ∫∫ ρ u ( n u ) dS = − ρU
2
j j j j j j jet d 2jet ;
Sb Sc Sj
4
(A8.9)
4
∫∫S ρ u ( n j u j ) dS = 3 ρU ∞Q;
∞
∫∫ n ( p − p ) dS = D; ∫∫ n ( p − p ) dS = 0; ∫∫ n ( p − p ) dS = 0;
Sb
x c
Sc
x c
Sj
x c
and (A8.10)
∫∫ nx ( p − pc ) dS = − 13 ρU ∞Q;
S∞
π π
so that D=ρ d 2jetU 2jet − ρU ∞ Q = ρU ∞2 d 2jet ⎡⎣(1 + σ ) + 1 + σ ⎤⎦ (A8.11)
4 4
2
D ⎛ d jet ⎞ ⎡ 1 ⎤
or, in dimensionless form Cd = 1 = 2⎜ ⎟ (1 + σ ) ⎢1 + ⎥, (A8.12)
2 ρU ∞ 4 d 1+σ ⎦
2 π 2
⎝ d ⎠ ⎣
where we note that djet=djet(σ;f) where f represents the shape of the cavitator. Note that for a
disk cavitator with long cavities djet/d≈0.454. For small σ Eq. (A8.12), in agreement with
well-known empirical results, becomes Cd≈0.82(1+σ) (A8.13)
App. 9. Axisymmetric Green Functions

Axial Inflow. Following [Uhlman, 2006], the source and dipole potentials may be defined in
terms of complete elliptic integrals of the first and second kinds. A cylindrical coordinate
system is used as shown in Fig.A9.1 with x=(x,r,α) and ξ=(ξ,ρ,θ). The axisymmetric
Green’s function can then be written
1
G = (A9.1)
(x − ξ ) + r + ρ 2 − 2r ρ cos(α − θ )
2 2
553
Appendices
Fig.A9.1. z y
( x, r , α )
Geometry definition.
α (ξ , ρ ,θ )
For purely axial inflow, α may be assumed θ
to be zero without loss of generality (the

case of a non-zero crossflow component is
addressed below). The potential induced at x
x due to a source ring is then
π π π
ρ dθ ρ dθ
Ψ= ∫ G ( x , y ) ρ dθ =
−π
∫π
− ( x − ξ ) + r + ρ − 2r ρ cosθ
2 2 2
= ∫π
− A − B cos θ
(A9.2)
and the dipole strength is

π π
∂G ( x , y ) ∂ ⎛ 1 ⎞
Ψn = ∫ ρ dθ = ∫ ⎜ ⎟ ρ dθ =
−π ∂ n −π ∂n ⎝ A − B cos θ ⎠
π π
(A9.3)
∂ ⎛ 1 ⎞ ∂ ⎛ 1 ⎞
= nξ ∫ ⎜ ⎟ ρ dθ + nρ ∫ ⎜ ⎟ ρ dθ
−π ∂ξ ⎝ A − B cosθ ⎠ −π ∂ρ ⎝ A − B cos θ ⎠
A = (x − ξ )2 + r 2 + ρ 2
where (A9.4)
B = 2r ρ
After [Uhlman, 2006], we now define generalized integrals as
π
cos m θ
J nm ( A, B ) = ∫ n dθ (A9.5)
−π ( A − B cos θ )
2
1
which satisfy the recursion formula J nm = ⎡⎣ AJ nm −1 − J nm−−21 ⎤⎦ (A9.6)
B
4 K (κ ) 4 E (κ )
and where J10 = ; J −01 = 4 A + BE (κ ); J 30 = . (A9.7)
A+ B ( A − B) A + B
In Eq. (A9.7), K (κ ) and E (κ ) are first and second kind complete elliptic integrals,
2B
respectively [Abramowitz & Stegun, 1970] and κ is defined as κ = (A9.8)
A+B
Using these expressions, we find that the ring source induced potential may be expressed as
π
ρ dθ 4ρ
Ψ= ∫ = ρ J10 = K (κ ) (A9.9)
−π A − B cos θ A+ B
and the ring normal dipole induced potential becomes
π ⎛
π
ρ ( x − ξ ) dθ ρ ( ρ − r cosθ ) ⎞
Ψ n = − nξ ∫ − n ρ ∫ ⎜ ⎟dθ
− π ( A − B cos θ )
3
2 ⎜
−π ⎝ ( A − B cos θ ) ⎠
3
2 ⎟
4ρ ( x − ξ )
= − nξ ρ ( x − ξ ) J 30 − nρ ρ ( ρ J 30 − rJ 31 ) = − E (κ )nξ (A9.10)
( A − B) A + B
−
2
( A − B) A + B
{ }
⎡( x − ξ ) 2 + (r − ρ ) 2 ⎤⎦ K (κ ) − ⎡⎣( x − ξ ) 2 + r 2 − ρ 2 ⎤⎦ E (κ ) nρ
⎣
554
Appendices
In evaluating the source potential when the field point and integration point coincide
(or nearly coincide), it’s necessary to consider the logarithmic singularity in the first kind
elliptic integral, K (κ ) . The logarithmic character of the source potential is
⎛ 8r ⎞⎡ t 1 ⎛ s2 ⎞ ⎤
Ψ = ln ⎜ ⎟ ⎢2 − − 2 ⎜⎜ + t 2 ⎟⎟ + +O(s 3 , t 3 )⎥ , (A9.11)
⎜ 2 ⎟ r 2r ⎝ 2
⎝ s + t ⎠ ⎢⎣ ⎦⎥
2
⎠
where s = x − ξ and t = r − ρ . In addition, when the field point lies within the domain of
integration, we determine the dipole self-influence using the relation
∂Ψ ∂Ψ
∫∫S ∂n dS = 4π − ∫∫ ∂n dS . (A9.12)
i
∑ Sj
i≠j
Inflow with a Crossflow Component. When there is a component of the inflow velocity in
the crossflow direction, as is the case for heave or pitch motions, the axisymmetric solution
may still be applied. For that case, it can be shown that the source and dipole strengths on
the boundary are proportional to cosθ [Hess & Smith, 1964]. We may then include the θ
dependence explicitly by defining the potential as follows ϕ j = ϕ j cos θ . ( )
(A9.13)
If we then absorb the cos (θ ) in the definition of the source and dipole potentials, we can
show that the original boundary integral equations apply, but with ϕ j in place of ϕ j and with
new definitions for the source and dipole potentials. The induced potential due to the ring
π
ρ cosθ dθ
source then becomes Ψ= ∫ (A9.14)
−π A − B cos θ
and that due to the ring normal dipole is
π
∂ ⎛ 1 ⎞
Ψn = ∫ ⎜ ⎟ ρ cosθ dθ
−π ∂n ⎝ A − B cos θ ⎠
π π
(A9.15)
∂ ⎛ 1 ⎞ ∂ ⎛ 1 ⎞
= nξ ∫ ⎜ ⎟ ρ cos θ dθ + nρ ∫ ⎜ ⎟ ρ cosθ dθ
−π ∂ξ ⎝ A − B cos θ ⎠ −π ∂ρ ⎝ A − B cosθ ⎠
Using equations (A.6) and (A.7), it can be shown that
ρ 4ρ ⎡ A ⎤
Ψ = ρ J11 ( A, B ) = ⎡⎣ AJ10 − J −01 ⎤⎦ = K (κ ) − A + B E (κ ) ⎥ (A9.16)
B B ⎢⎣ A + B ⎦
Similarly, the two components of the ring normal dipole induced potential may be evaluated
π π
∂ ⎛ 1 ⎞ cos θ dθ
∫−π ∂ξ ⎜⎝ A − B cos θ ⎟⎠ ρ cos θdθ = ρ(x − ξ )−∫π (A − B cos θ )32 = ρ(x − ξ )J 3 =
1
as (A9.17)
ρ(x − ξ ) 4 ρ (x − ξ ) ⎡ A ⎤
= ⎡⎣AJ 3 − J 1 ⎤⎦ =
0 0
E (κ ) − K (κ )⎥
B B A + B ⎢⎣ A − B ⎦
π π
∂ ⎛ 1 ⎞ r ρ cos θ − ρ cos θ
2 2
∫−π ∂ρ ⎜⎝ A − B cos θ ⎟⎠ ρ cos θdθ = −∫π (A − B ) 2

3 dθ =
and (A9.18)
B 2 2 ⎡ A (A − ρ 2
) − 2 r ρ
2 2
⎤
= J 3 − ρ 2J 31 = ⎢ E (κ ) − (A − ρ 2 )K (κ )⎥ .
2 rρ A − B ⎣ A−B ⎦
555
Appendices
As before, care must be taken in evaluating the source potential when the field point
and integration point coincide (or nearly coincide). Again extracting the logarithmic
behavior of the source potential and expanding the multiplier in a Taylor series about
⎛ 8r ⎞⎡ t 1 ⎛ 3s 2 ⎞ ⎤
( s, t ) = ( 0,0 ) yields Ψ = ln ⎜⎜ 2 2 ⎟⎟ ⎢2 − + 2 ⎜⎜ + t 2 ⎟⎟ + O(s 3 , t 3 )⎥ . (A9.19)
⎝ s + t ⎠ ⎣⎢ r r ⎝ 2 ⎠ ⎦⎥
∂Gcf ∂Gaxi ∂H
To obtain the dipole self-influence terms, we note that = + , (A9.20)
∂n ∂n ∂n
∂H ∂Gcf ∂Gaxi
where = − (A9.21)
∂n ∂n ∂n
Gcf is the cross-flow Green’s function, and Gaxi is the Greens function for axial inflow. We
∂Gcf ∂G ∂H
can then write ∫∫S ∂n dS = ∫∫S ∂naxi dS + ∫∫S ∂n dS , (A9.22)
where S is the surface of the complete closed body. The integral on the left hand side can be
broken into two components: the integral over the self-influence panel and the integral over
∂Gcf ∂Gaxi ∂H ∂Gcf
the rest of the surface, thus ∫∫ dS = ∫∫ dS + ∫∫ dS − ∑ ∫∫ dS . (A9.23)
S ∂n S ∂n S ∂n j ≠i S ∂n
i j
∂Gaxi
Furthermore, we can make use of the fact that ∫∫
S
∂n
dS = 4π (A9.24)
∂Gcf ∂H ∂Gcf
so that ∫∫ dS = 4π + ∫∫ dS − ∑ ∫∫ dS (A9.25)
Si
∂n S
∂n j ≠i S ∂n
j
The second term on the right-hand-side of (A9.25) can be written

∂H ∂H ∂H
∫∫S ∂n dS = nξ ∫∫S ∂ξ dS + n ρ ∫∫S ∂ρ dS (A9.26)
To evaluate this expression, we note that

∂ ⎛ cos θ − 1 ⎞
π
2(x − ξ )
∫−π ∂ξ ⎜⎜ A − B cos θ ⎟⎟ ρ cos θdθ = ρ(x − ξ ) ⎡⎣J 3 − J 3 ⎤⎦ = r A + B ⎡⎣E (κ ) − K (κ )⎤⎦ (A9.27)
1 0
⎝ ⎠
∂ ⎛ ⎞
π
cos θ − 1
∫π ∂ρ ⎜⎜ ⎟⎟ ρ cos θ dθ = ρr ⎡⎣J 3 − J 3 ⎤⎦ − ρ ⎡⎣J 3 − J 3 ⎤⎦
2 1 2 1 0
and − ⎝ A − B cos θ ⎠ (A9.27)

=
1
ρr A + B ⎣
⎡ ( ) (
B + 2(A − ρ ) E (κ ) − B − 2(A − ρ ) K (κ )
2 2
)
⎤
⎦
Special treatment is also required to evaluate these expressions when the field point and
integration point are close to one another to ensure that the logarithmic behavior of the
second kind elliptic integral is accurately captured.
556
References
References
Abelson, HI 1970 Pressure measurements in the water-entry cavity. JFM 44, #1, p129-144.
Abramowitz, M., Stegun, IA, (Eds.) 1972 Handbook of Mathematical Functions. Applied
Mathematics Series, 55, National Bureau of Standards, Washington, 364p.
Achkinadze, AS 1974 Linear problem of supercavitating wing and problem of optimal wing's
shape (variation approach). Trans. Scientific Soc. v.217, p139-164 (in Russian).
Achkinadze, AS 1983 Application of mathematical programming methods in the linear theory of
cavity flows. Proc. 12-th SMSSH, Varna, v.2, p32/1-19 (in Russian).
Achkinadze, AS 1989 Design of optimal screw propellers, turbines and freely rotating
turbopropellers adapted for radially nonuniform swirled flow. Proc. PRADS'89, v3, p124/1-1.
Achkinadze, AS 1993 Design calculation of noncavitating and highly cavitating propellers using
vortex theory with load and radially nonuniform inflow taken into account. Doctoral
dissertation, St-Petersburg Marine Technical Univ., (in Russian)
Achkinadze, AS 2001 Supercavitating propellers. Preprint: Von Karman Inst. for Fluid
Dynamics, RTO AVT/VKI Special Course: "Supercavitating Flows", Belgium.
Achkinadze, AS et al. 2007 Hydromechanics. (with Besydovsky, AR, Vasileva, VV, Kornev, NV,
Faddeev, YI) “Morkoi Vestnk”, SPb, 552 p. (in Russian).
Achkinadze, AS, Fridman, GM 1995 On some aspects of design of supercavitating foils and
propellers. Variation and asymptotic approach. Proc. PROPCAV'95, p163-174.
Achkinadze, AS, Fridman, GM 2000 Optimal profiles for supercavitating propeller screws with
spoilers and fixed angle of frontal edge taper. Applied hydromechanics, Inst. of
Hydromechanics, NAS of Ukraine, vol 2 (74), number 3, pp.5-16 (in Russian).
Achkinadze, AS, Fridman, GM 2000a Optimal Sections for Supercavitating Propellers with
Spoiler and Preset Leading Edge Angle. Proc. NCT’50 Int’l Conf. on Propeller Cavitation,
Univ. of Newcastle, Newcastle upon Tyne, p263-274.
Achkinadze, AS, Krasilnikov, VI 1997 A generalized optimum condition for wake adapted screw
propellers. Proc. SNAME "Propellers/Shafting'97, Virginia Beach, pp.22/1-22/28
Achkinadze, AS, Krasilnikov, VI, Stepanov, IE 2000 Hydrodynamic Design Procedure for
Multi-Stage Blade-Row Propulsors Using Generalized Linear Model of the Vortex Wake.
Trans SNAME "Propellers/Shafting'97", Virginia Beach, p20/1-20/21.
Achkinadze, AS, Narvsky, AS 1985 Supercavitating propeller design equation in lifting surface
theory and method of its solution. Proc. 14th SMSSH, Varna, vol. 1, p19/1-21.
Achkinadze, AS, Narvsky, AS 1987 Calculation of cavitating propellers based on the lifting-
surface theory. Proc. 4th Int’l. Conf. IMAEM, Varna, vol. 5, p. 158/1-15.
Achkinadze, AS, Narvsky, AS 1989 Supercavitating propeller lifting-surface corrections. Proc.
18th SMSSH, Varna, vol. 1, p. 19/1-4.
Acosta, A 1961 The effects of a longitudinal gravity field on the super-cavitating flow over
wedge. J. Appl. Mech., 1.
Adams, M, Sears, W 1953 Slender Body Theory – Review and Extension. J. Aeronautical
Sciences, p85-98.
Afanasiev, KE et al. 1986 Finite and boundary element investigation of the evolution of free
surfaces in connection with the unsteady motion of bodies in an ideal incompressible fluid.
(with MM Afanasieva, AG Terentiev), MZG, # 5, p8-13; English transl. in Fluid Dynamics,
21
557
References
Afanasiev, KE et al. 1989 Deformation of gas bubbles in a liquid. (with MM Afanasieva, AG
Terentiev)Proc. Topical Problems in Hydrodynamics. ChSU, p4-10.
Afanasiev, KE et al 2006 Numerical simulation of free surface flows by the SPH and MPS. (with
AE Iliasov, RS Makarchuk, AY Popov) Comput. Technol. v.11, 9, p26-44 (in Russian)
Afanasiev, KE, Goudov, AM 2001 Information technologies in numerical calculations:
Educational manual. Kemerovo: KemSU. 203 p
Afanasiev, KE, Grigorieva, IV 2002 The investigation of buoyant gas bubble dynamics near an
inclined wall. Proc. Int. Sum. School “High Speed Hydrodynamics”. Ch., p111-118.
Afanasiev, KE, Samoilova, TI 1995 Procedure of using the boundary element method in
problems with free boundaries. Calculating Technol., 7, № 11, P. 19-37.
Afanasiev, KE, Stukolov, SV 1998 Moving of a solitary wave onto an inclined shore. Bul. Omsk
Univ., Omsk, № 3, P. 9-12.
Afanasiev, KE, Stukolov, SV 1999 On the existence of three solutions for a supercritical steady
flow of a heavy liquid over obstructions. PMTF, 40, # 1, p27-35. Engl. transl. in Fluid Dyn.
Afanasiev, KE, Stukolov, SV 2002 Solution of non-linear problems in the hydrodynamics of an
ideal fluid with free boundaries by the method of boundary elements. Proc. Int. Summer Sci.
School “High Speed Hydrodynamics”, Comp. Publication, Cheboksary-Washington, p41–53.
Afanasiev, KE, Terentiev, AG 1984 Application of the finite element method in problems with
free boundaries. In «Dynamics of a Continuous Medium with Time-Dependent Boundaries»
(ed. Terentiev). ChSU, p8-17 (in Russian).
Ahlberg, JH et al. 1967 The theory of splines and their applications. (with EN Nilson, JL Walsh)
Academic Press.
Ahuja, V et al. 2001. Simulation of Cavitating Flows Using Hybrid Unstructured Meshes. (with
A Hosangadi, S Arunajatesan) J. Fluids Eng., 123, pp331-338
Amromin, EL 1985 On cavitation flow calculation for viscous capillary fluid. Fluid Dyn., 20,
p891-897
Amromin, EL 1986 Cavitation Flow Calculations for a Viscous and Capillary Fluid. Plenum
Publishing, New York, NY; translated from MZG, 6, p45-51, 1985.
Amromin, EL 2002 Scale Effect of Cavitation Inception on a 2D Eppler Hydrofoil. J. Fluids
Eng., 124, p186-193.
Amromin, EL 2007 Determination of Cavity Detachment for Sheet Cavitation. J. Fluids Eng.,
129, p1105-1111.
Amromin, EL, Ivanov, AN 1981 Ideal cavitation and cavitating flow by real conditions. In «High
Speed Hydrodynamics» ChSU, p3-13 (in Russian).
Amromin, EL, Ivanov, AN 1982 Determination of Cavity Separation Points on Body in Viscous
Capillary Fluid. Sov. Physics-Doklady, 262, p823-826.
Amromin, EL et al. 2010 Viscous Drag Reduction by Air Cavity as an Inverse Ideal Fluid
Problem: Achievements and Issues. (with G Karafiath, B Metcalf) Proc. 28th Symp. on Naval
Hydrodyn., Pasadena, CA.
Arakeri, VH 1975 Viscous Effects on the Position of Cavity Separation from Smooth Bodies.
JFM, Vol. 68, Part 4, p779-799.
Arakeri, VH, Acosta, AJ 1976 Cavitation inception observation on axissymmetric bodies at
supercritical Reynolds numbers. J. Ship Res., 20, p40-50.
Arndt, RAE 2002 Cavitation in Vortical Flow. Ann. Rev. Fluid Mech. 34, p143-176.
558
References
Arndt, RAE 2010 Sheet/Cloud Cavitation Revisited. Proc. 13th Int’l Symp. Transport Phenomena
and Dynamics of Rotating Machinery, Honolulu, HI.
Ashley, H, Landahl, M 1965 Aerodynamics of Wings and Bodies. Dover Publications, NY.
Auslaender, J 1962 The linearized theory for supercavitating hydrofoils operating at high speeds
near a free surface. J. Ship. Res., v. 6, No 2.
Baeva, MA et al. 1981 Hydrodynamic characters of a cavitation profile with spoiler. (with MY
Mizina, YM Sadovnikov) Trans. Krylov Scientific Society, vol.358, p.44-60 (in Russian)
Banerjee, PK, Butterfield, R 1981 The boundary element method for engineers. McCraw-Hill.
Barr, R 1970 Supercavitating and superventilated propellers. Tran. SNAME, Vol.78.
Basin, AM, Goshev, GA 1963 Experimental investigation of semi-submerged propeller
characters. Trans. LIWT, Vol. X.
Batchelor, GK 1970 An introduction to fluid dynamics. Cambridge Univ. Press.
Bazhenov, VG et al. 1984 High speed impact of visco-elastic thin shells construction on the
surface of a compressible fluid. (with AV Kochetkov, SV Krylov, AG Ugodchikov) Izv. AN
USSR, MTT, No. 5, p161-169 (in Russian).
Belotserkovskii, OM, Davydov, YM 1971 Non-stationary method of finite particles for gas-
dynamic calculations. J. Vych. Math. & Math. Physic, II, #1, pp.182-207 (in Russian)
Belotserkovskii, OM, Davydov, YM 1973 Numerical approach for investigating some transsonic
flows. Lecture Notes in Physics, Springer-Verlag, v. 19, pp.25-32 (in Russian).
Belotserkovskii, OM, Nisht, MI 1978 Separation and continuous flow of ideal fluid around thin
wings. M. Nauka, 351p (in Russian).
Bervy, NO 1894 On flow of a liquid with free boundary in the gravity field. Proc. Society of
Natural Science, v. VI & VII, M. (in Russian).
Betz, A, Petersohn, E 1931 Anwendung der Theory der freien Strahlen. Ing. Archiv, Bd. 2.
Billet, ML, Holl, JW 1981 Scale effects on various types of limited cavitation. J. Fluids Eng.,
103, p405-414.
Birkhoff, G 1956 Stability of spherical bubbles. Applied Math., v.13, No.4.
Birkhoff, G 1960 Hydrodynamics. Princeton Univ. Press, Princeton
Birkhoff, G, Goldstine, H, Zarantonello, E 1953 Calculation of plane cavity flows past curved
obstacle. Rend. Seminar. Mat. Univ.Politec. Torino. V.13. p205-224.
Birkhoff, G, Zarantonello, EH 1957 Jets, Wakes, and Cavities. Academic Press, NY.
Blake, JR, Gibson, DC 1987 Cavitation bubbles near boundaries. Ann. Rev. Fluid Mech., 19,
99—124.
Bloch, E 1969 A finite difference method for the solution of free boundary problems. Reports
NYU-NYO-1480-116.
Bloch, EA 1969a Numerical solution of free boundary problems by the method of steepest
dencent. J. Phys. Fluids, v.12, No.12, p129-132.
Bonder, J 1936 Sur un cas de mouvements plans a deux sillages. Ann. Acad. des Sciences Techn.
a Varsovie, Vol. III, Polish
Bourgoyne, DA et al. Time-averaged flow over a hydrofoil at high Reynolds number. JFM, 496,
p365-404.
Brebbia, CA 1978 The boundary element method for engeneers. Pentech Press, UK.
Brebbia, CA Telles, JCF, Wrobel LC 1984 Boundary element techniques. Springer-Verlag
Brennen, CE 1969 A numerical solution of axisymmetric cavity flows. JFM., 37, #4, p671-688.
559
References
Brennen, CE 1969a Some viscous and other real fluid effects in fully developed cavity flows.
chapter in “Cavitation State of Knowledge”. ASME, New York, NY, p141-147.
Brennen, CE 1969b The Dynamic Balances of Dissolved Air and Heat in Natural Cavity Flows.
JFM, 37, 1, p115-127.
Brennen, CE 1970 Cavity surface wave patterns and general appearance. JFM, 44, 1, p33-49.
Brennen, CE 1995 Cavitation and Bubble Dynamics. Oxford Univ. Press, NY.
Brewer, WH, Kinnas, SI 1997 Experiment and Viscous Flow Analysis on a Partially Cavitating
Hydrofoil. J. Ship Res., 41, p161-171.
Brillouin M 1911 Les surfaces de glissement de Helmholtz et la resistance des fluides. Ann.
Chemie et phys. T. 23.
Brodetsky, S 1923 Discontinuous fluid motion past circular and elliptic cylinders. Proc. Roy.
Soc. London, S.A. V. CII, No A718.
Buford, R Koehler, Jr, Kettleborough, CF 1977 Hydrodynamic Impact of Falling Body upon a
Viscous Incompressible Fluid. J. of Ship Research, Vol. 21, No. 3, p165-181.
Buivol, VN 1980 Slender Cavities in Flows with Perturbations. Nauk. Dumka, Kiev, (In Russian.)
Buivol, VN Kapankin, EN, Zhuravlev, EN 1973 Distributed motion of thin three-dimensional
cavities. Proc. IUTAM Simp. «Non-steady Flow of Water at High Speeds». L. (in Russian).
Burill, WC 1951 Sir Charles Parsons and Cavitation. Trans. Inst. of Marine Eng., Vol.LXIII, #8.
Campbell, IJ, Hilborne, DV 1958 Air Entrainment behind Artificially Inflated Cavities. Proc.
Second Symp. on Naval Hydrodynamics, Washington, DC.
Capodano, P 1969 Sur le mouvement d’un profil sons une surface libre. C.r. Acad. Sci., v. 286, 3.
Caster, EB 1963 TMB 2, 3 and 4-bladed supercavitating propeller series. DTMB Report 1637.
Castillo, L Wang, X, George, WK 2004 Separation Criterion for Turbulent Boundary Layers via
Similarity Analysis. J. Fluids Eng., 126, p297-304.
Ceccio, SL, Brennen, CE 1992 Dynamics of sheet cavitieson bodies of revolution. J. Fluids Eng.,
114, p93-99.
Chahine, G, Duraiswami, R, 1998 Interaction between cavitating bubbles and vertical structures.
Proc. #rd Symp on Cacitation, v.1, p93-98, Grenoble.
Chahine, GL, Hsiao, CT 2000 Modeling 3D unsteady sheet cavities using a coupled UnRANS-
BEM Code. Proc. 23rd Symp. on Naval Hydrodynamics
Chaplygin, SA 1899 On problem of jets in incompressible liquid. Proc. Society of Natural
Science. V. X, 1, M. (in Russian).
Chaplygin, SA 1902 On gas jets. MGU Print, 121p (Transl. Engl. Tech. Memo. Washington, DC,
No 1063, NACA, 1944, 112p.).
Chaplygin, SA 1914 The theory of latticed wing. M., Math. articles, XXIX (in Russian).
Chaplygin, SA, Golubev, VV 1935 To the theory of slat and flap. Ann. TsAGI, v171. (in Russian)
Chaplygin, SA, Minakov, AP 1930 Theoretical calculation of an operation of a turbine.
(Generalization of Joucowski solution). M., Trudy TsAGI, 41 (in Russian).
Chaplygin, YS 1941 Gliding on fluid of finite pepth. PMM, Vol. V, No 2 (in Russian).
Charny, IA 1963 Underground hydro-gasdynamics. Gostehizdat, (in Russian).
Cherepanov, GP 1963 Flow of ideal fluid with free surfaces in two and three connected domains.
PMM, v. XXVII, No 4.
Charny, GG 1988 Gas dynamics. M. Nauka, 424 p.
560
References
Chou, YS 1974 Axisymmetric Cavity Flows Past Slender Bodies of Revolution. J. Hydronautics,
8, 1, p13-18.
Cisotti, U 1921 Idromeccanica piana. Milano, v. 1, 2.
Cohen, H, Sutherland, CD, Tu, YO 1958 Wall effects in cavitationg hydrofoil flow. JSR, v.1, #3,
p31.
Cole, JD 1968 Perturbation Methods in Applied Mathematics. Blaisdell Publ. Co, Waltham, MA.
Cole, RH 1950 Underwater explosions. Princeton Univ. Press
Cox, PN, Clayden, WA 1955 Air Entrainment at the Rear of a Steady Cavity. Proc. Sympos at
Nat. Phys. Lab., London.
Cueto-Felgueroso, L et al. 2004 On the Galerkin formulation of the Smoothed Particle
Hydrodynamics method. (with I Colominas, G Mosqueira, F Navarrina, M Casteleiro) Int’l J.
Numerical Methods in Engineering, 60, 1475 - 1512.
Cuthbert, J, Street, R 1964 An Approximate Theory for Supercavitating Flow About Slender
Bodies of Revolution. Lockheed Missiles & Space Co Report, TM81-73/39,., Sunnyvale, CA.
Dang, J, Kuiper, G 1998 Re-entrant jet modeling of partial cavity flow on two-dimensional
hydrofoils, J. Fluids Engineering, Vol. 121, p773-780.
Davis, BV, English, JW 1968 The evolution of a fully cavitating propeller for a high-speed
hydrofoil ship. Proc. 7-th ONR Symp. on naval hydrodynamics, Rome, August.
Dimitrieva, NA 1996 Numerical investigation of cavitating flow about cylindrical and spherical
surface. Proc. Dynamics of Continuous Medium with Free Boundary. ChSU, p78-91.
Dimitrieva, NA, Terentiev, AG 1998 Theoretical investigation of cavitating flows. Proc. Third
Int’l Sym. on Cavtation. Grenoble, France, p275-280.
Dimitrieva, NA, Terentiev, AG 2002 An application of the p-analytical functions to investigate
cavitating flows. Proc. Int. Summer Scientific School “High Speed Hydrodynamics”.
Cheboksary, p119-126.
Dobrovolskaya, ZN 1969 On some problems of similarity flow of fluid with a free surface. JFM,
v. 36, part 4.
Dommermuth, DC, Yue, DKP 1987 Numerical simulation of nonlinear axisymmetric flows with
a freesurface. Fluid Mech., Vol. 178, p195-219.
Drew, DA, Passman, SL 1999 Theory of Multicomponent Fluids, App. Math. Sci., 135, Springer,
New York.
Eames, MC, Jones, EA 1971 HMCS Bras d’Or. The Naval Architect, RINA, April.
Efremov, II & Roman, VM 1973 Influence of a free surface and walls on characteristics of a
cavity foil. Proc. IUTAM Symp. Non-steady flow of water at high speeds, L.-M., 165-172 (in
Russian).
Efremov, II 1969 Calculation of supercavitating hydrofoils near a free boundary. Applied Mech.
and technical physics. Kiev, 3.
Efremov, II 1974 Linear theory of cavitating flow. Kiev, Naukova Dumka, 156p (in Ukrainien).
Efros DA 1946 Hydrodynamic theory of plane cavitating flow. DAN USSR, v.51, #4, p267-270.
(in Russian).
Egorov, IT Sadovnikov, YM, Isaev, II, Basin, MA 1971 Artificial cavitation. Sudostroenie Pubs,
L., 280p.(in Russian).
Elizarov, AM et al. 1994 Inversed boundary problems in air hydrodynamics. M., FML.
Eppler, R 1954 Beitrage zur Theorie und Anwendung der unstetigen Zstromungen. J. Rat. Mech.
Anal. V. 3.
561
References
Epshteyn, LA 1961 Determination of the Amount of Gas Needed to Maintain a Cavity behind a
Body Moving at Low Fr Number. Trudy TsAGI, No. 824, Moscow, Russia.
Epshteyn, LA 1963 Motion of inclined plate under the free surface. PMM, v. XXVII, No. 4, p4,
p735-738.
Epshteyn, LA 1970 Methods of Dimensional Analysis and Similarity Theory in Problems of Ship
Hydrodynamics. L., Sudostroenie, 207p. Transl. by Naval Intel. Support Center.
Eroshin, VA 1980 Hydrodynamic forces by impact of blunt bodies onto surface of compressible
liquid. (with NI Romanenkov, VV Serebryakov, YL Yakimov) MZG, No 6, p44-51. English
transl. in Fluid Dynamics
Etter, RJ 2000 State of the art - cavitation test facilities and experimental methods. Proc. NCT’50
Int’l Conf. on Propeller Cavitation, Univ. of Newcastle, UK.
Fabula, AG 1962 Thin-airfoil theory applied to hydrofoils with a single finite cavity and arbitrary
free-streamline detachment. J. Fluid Mech, Vol.12, Part 2.
Fedotkin, IM, Machinsky, AC 1983 Supercavitation method of water desalination. Proc. Conf.
on Problems of global water budget. Vladivostok, p.124-126.
Ferrando, M, Scamardella, A, Bose, N, Liu, P, Veitch, B 2001 Performance of a family of
surface piercing propellers. RINA, W283, p1.
Filippov, VI 1987 The Influence of Compressibility on the Geometrical Sizes of Cavities. In:
Interactions of Bodies in Fluid with Free Boundaries, ChSU, p115-122, (in Russian).
Fine, NE 1992 Nonlinear Analysis of Cavitating Propellers in Nonuniform Flow. PhD Thesis,
Dept. of Ocean Engineering, MIT, Cambridge, MA.
Fine, NE, Kinnas, SA 1993 A Boundary Element Method for the Analysis of the Flow Around 3-
D Cavitating Hydrofoils. J. Ship Res., Vol.37, No.1, p213-224.
Fisher, EA 1966 Experimental determination of the screw propeller cavitation size. Trans. KSRI,
Vol.231.(in Russian)
Forbes, LK, Schwartz, LW 1982 Free surface flow over a semicircular obstruction. JFM V.114.
P. 229-314.
Frankel, FI, Karpovich, EA 1948 Slender Body Gasdynamics. Gostechizdat. M., 175p.
Fridman, GM 2002 Matched asymptotics for nonlinear problems of the theory of jets in an ideal
fluid. Proc. Int. Sum. Scien School HSH-2002, Ch, p139-146.
Fridman, GM, Uryadov, AK 2004 Cavitating flat plate with stagnation zone in the spoiler
vicinity. Proc. Int. Sum. Scien School HSH-2004, Ch, p83-90.
Furuya, O 1984 A performance prediction theory for partially ventilated propellers. Proc. 15th
Symp. on Naval Hydrodynamics.
Furuya, O, 1985 A performance prediction theory for partially ventilated propellers. J. Fluid
Mech, 151, p. 311-335.
Galanin, AV 1974 On influence of singularities on the lift of airfoil in bounded stream of fluid.
Problems of applied Math. and Mech. ChSU, v3, p25-39 (in Russian).
Galanin, AV et al. 1979 Initial penetration of deformed shell into compressible fluid. (with VA
Gusev, SS Saikin) In «Non-station moving of bodies in a fluid» ChSU, p25-36 (in Russian).
Galanin, AV et al. 1982 Penetration of bodies into liquid of finite depth. (with VM Mikhailov, NP
Porfiriev, AG Terentiev) Proc. Dynamics of Continuums with free boundaries. CSU, 45-61
(In Russian).
Galanin, AV, Gusev, VA 1978 Cavitating flow of over thin wedge of a weighty fluid over thin
wedge in a longitudinal gravity field. Proc. Jet and cavitations stream. Ch., p19-27.
562
References
Galanin, AV, Terentiev, AG 1984 Boundary value problems of linear hydrodynamics. CSU.
Galin, LA 1947 Impact on solid body on a compressible fluid surface. PMM, 11, No. 5, p547-550
(in Russian). English transl. in Appl. Math & Mech
Garabedian, PR 1956 Calculation of axially symmetric cavities and jets. Pac J.Math, 6, p611-689
Garabedian, PR 1971 Calculation of Riabouchinsky cavity flows. Proc. IUTAM Symp. Non-
steady flow at high speeds. L., Nauka, p197-199.
Gawn, R, Burrill, LC 1957 Effect of cavitation on the performance of a series of 16” model
propellers. Trans. INA, Vol.99, p690-728.
Georgievskaya, EP 1970 Cavitation damage of screw propellers and methods of removal. L.
Sudostroenie.
Georgievskaya, EP 1984 Design principles of hydrofoil screw propellers with improved erosion
characteristics. (with MA Mavludov, IV Salazkin) Shipbuilding Problems, 41, p86-89 L. (in
Russian)
Georgievskaya, EP 1989 A new generation of cavitating propellers and their practical application
to "Kometa" hydrofoil. (with MA Mavludov, V Hadjimikhalev, P Kozhukharov) Proc. 4-th
PRADS, Varna, Bulgaria, Vol.3, p123/ 1-11.
Geurst, JA 1956 Linearized theory for partially cavitated hydrofoils. ISP, v.6, #60, p369-384.
Geurst, JA 1959 Der Zusammenhang zwishen den linearen und nicht-linearen Theorien der
Stroemungen mit Kavitation. ZAMM,Bd. 39, No. 9/11, S. 430-432.
Geurst, JA 1959a Linearized theory for partially cavitated hydrofoils. Int. Shipb. Prog., v.6, # 60.
Geurst, JA 1960 Linearized Theory for Fully Cavitating Hydrofoils, Int’l Shipbuilding Progress,
Vol. 7, No. 65
Geurst, JA 1960a A linearized theory for the unsteady motion of a supercavitating hydrofoil. Inst.
Appl. Math., Technol. Univ. Rep. No. 22, Delft.
Geurst, JA, Verbrugh, PJ 1959 A Note on Camber Effects of a Partially Cavitated Hydrofoil.
Int’l Shipbuilding Progress, Vol. 6, No. 61
Gilbarg, D, Rock, HH 1946 NOL Memo 8718.
Gilbarg, D 1960 Jets and cavities. In: Handbuch der Physik, Springer-Verlag, 9, p311-445.
Gingold, RA, Monaghan, JJ 1977 Smoothed particle hydrodynamics: theory and application to
non-spherical stars. Monthly Notices of the Royal Astronomical Society 181, 375-389.
Gogish, LV, Stepanov, GY 1979 Turbulent Separated Flows. Nauka, M. (in Russian).
Golubev, VV 1935 On action of a wing with a sucking of boundary layer. Technical articles of
TsAGI. M., No. 45 (in Russian).
Goman, OG, Popov, VV 1980 To solving of axisymmetric potential problems. In “Dynamics of
continuum with free boundaries” ChSU, p55-61.
Goman, OG 1972 Remark on the problem solving for axisymmetric flows from plane-parallel
one. MZG, No 2, p157-158.
Goman, OG 1984 On the problem of connection plane-parallel and three-dimensional flows.
Proc. Conf. Dynamics of continuum with moving boundaries, CSU, p43-53.
Gopalan, S, Katz, J 2000 Flow Structure and Modeling Issues in the Closure Region of Attached
Cavity. Phys. Fluids, 12, p895-911.
Gorelov, DN 1975 Theory of a wing in non-stationary flow. Novosibirsk, Press of Univ.
Gorshkov, AS, Rusetskiy, AA, Borusevich, VO 2007 Cavitation tunnels. KRSI, 52p. (in Russian)
Gradstain, IS, Ryzhik, IM 1962 Tables of integrals, sums, series and products. M. GIFML.
563
References
Grant, JR, Kirschner, IN 2003 High-Speed Motion in Bubbly Flows. Proc. CAV2003, Osaka
Grant, JR, Kirschner, IN 2005 Supercavitation in a Bubbly Mixture: Toward a Slender Body
Theory and Implications for Drag. Proc. 2nd Int’l Symp. on Seawater Drag Reduction
(ISSDR2005), Busan, Korea.
Grant, JR, Kirschner, IN 2006 Steady High Speed Motion of an Axisymmetric Slender Body in a
Bubbly Liquid: Behavior of the Drag Coefficient. Proc. CAV2006, Wageningen.
Grant, JR et al. 2004 High-Speed Motion in Bubbly Flow: Comments on Drag. (with IN
Kirschner, JS Uhlman) Proc. 2004 (HSH2004, NANI, Cheboksary
Gridneva, VA, Kuleshov, VP, Shachtmeyster, LI 1979 Solving of impact problem with free and
contact boundaries by applying of method of finite particles. Tomsk Univ. Dep. in VINITI,
p722-79, (in Russian).
Grigoryan, SS 1959 An approximated method of separation flow past axisymmetric body. PMM,
Vol. 23, No 5, M., p951-953.
Grilli, ST, Skourup, J, Svendsen 1989 An efficient boundary element method for nonlinear water
waves. J. Eng'g Analysis with Boundary Elements. CMP, UK, Southampton. v.6, #2, p97-107.
Gubriy, VI 1969 Direct problem of cascade of supercavitating foils. Hydrodynamics, v.15, p16-
25, Kiev, Naukova Dumka (in Russian)
Gurevich, MI 1947 On the jet model of flow around flat plate. Rep. of TsAGI, # 612. (in Russian)
Gurevich, MI 1947a Semi-body of finite resistance in subsonic flow. J. App. Math.& Mech.,
#653, p1-12.
Gurevich, MI 1953 The drag of cylinder and wedge by low cavitation numbers. Proc. Moscow
Inst. of fish industry, V. (in Russian)
Gurevich, MI 1979 The theory of jets in ideal fluids. 2nd Ed. M., Nauka. 536 p. (in Russian),
English transl of the 1st Ed. NY – London, Academic Press, 1965.
Gurevich MI, Haskind, MD 1953 Jet flow of contour with small oscillatory movement. PMM,
Vol. XVII, No 5.
Gusev, VA 1974 Flow of a plate cascade by using Efros (re-entrant jet) model. Coll. Problems of
Math. and Mech. ChSU., 3 (in Russian).
Gusev, VA, Terentiev, AG 1970 On flow of plate with full cavitation. Proc. of konf. Problems of
hydrodynamics and low-temperature plasma, Ch. CSU, p3-10 (In Russian)
Gusev, VA, Terentiev, AG 1981 Depth and surface closure of ventilated cavity by entry of a
body into gravity liquid. In «High Speed Hydrodynamics», ChSU, p48-58 (in Russian).
Guzevsky, LG 1975 The calculations of the axisymmetric flows with free boundaries. Doklady
AN USSR, Vol. 225, N 2, p269-271 (in Russian).
Guzevsky, LG 1979 Numerical Analysis of Cavity Flows. Preprint 40-79, Inst. of
Thermophysics, Siberian Branch, RAN, Novosibirsk
Guzevsky, LG 1982 The flow of gravity fluid of finite depth around obstacles. Proc. Conf. on
Dynamics of Continuum with Free Bounaries, ChSU, p61-69 (in Russian)
Guzevsky, LG 2002 Analytical Approximations of Riabouchinsky Cavity Flows past a Disk.
Proc. Int’l Summer Scientific School on High-Speed Hydrodynamics, NANI, Cheboksary.
Haack, C, Gravert, P, Schlegel, V 1981 The modeling of extreme gravity waves: an approach
towards a numerical wave channel. Proc. «Computational Modeling of Free and Moving
Boundary Problems». Southampton, UK, p91-104.
Hammack, JL 1973 A note on tsunamis: their generation and propagation in on ocean of uniform
depth. Fluid Mech, 60, p769-799. Not referred to in the text
564
References
Hammer, PC 1959 Numerical evalution of multiple integrals. In: Numerical Approximation (ed.
RL Langer), Univ. of Wisconsin pres, p99-15.
Harlow, FH, Welch, JE 1966 Numerical study of large-amplitude free surface motions. Phys.
Fluids, v. 9, No. 5, p842-851.
Hart, DP, Brennen, CE, Acosta, AJ 1990. Observations of cavitation on a three dimensional
oscillating hydrofoil. Proc. ASME Cavitation and Multiphase Flow Forum, FED-98, NY,
p49-52.
Helmholtz, H 1868 Ueber discontinuirliche Fluessigkeitsbewegungen. Monatsber. Koenig. Akad.
Wissenschaften, Berlin.
Hess, JL, Smith, AMO 1964 Computation of Non-Lifting Potential Flow about Arbitrary Three-
Dimensional Bodies. J. Ship Res., 8(2), p22-44.
Hoperskov, AE 1963 On flows with formation of closed cavitaties. PMTF, # 5. (in Russian)
Hromadka, TV, Lai Ch 1987 The complex variable Boundary Element Method in Engineering
Analysis. Springer-Verlag.
Hrubes, JD 2001 High-Speed Imaging of Supercavitating Underwater Projectiles. Experiments in
Fluids, 30, 1, p57-64, Springer-Verlag, Berlin.
Ibragimova, TB, Mavlydov,MA, Rousetsky, AA 1995 Basic principles of propulsor efficiency
comparisons. Proc. 3rd Int’l Conf. on Fast Sea Transportation.
Iga, Y, Nonmi, N, Goto, A, Shin, BR, Ikohagi, T 2003 Numerical Study of Sheet Cavitation
Breakoff Phenomenon on a Cascade Hydrofoil. J. Fluids Eng., 125, p643-650.
Ikonnikov, VV, Maskalik, AI 1987 Unique feature of hydrofoil ships design and formation.
Sudostroenie Pubs, L. 318 p. (in Russian)
Il’insky, NB, Potashev, AV 1986 Boundary-value problems of explosion theory. Kazan SU, 179p
(in Russian)
Ivanov, AN 1960 Cavitating flow of hydrofoils or wings. Izv. AN USSR, OTN, 6.
Ivanov, AN 1961 On influence of gravity on hydrodynamic characteristics of hydrofoils.
Sudostroenie, L., No 2, p10-12.
Ivanov, AN 1980 Hydrodynamics of Supercavitating Flows. Sudostroyenie, L., 237p. (in
Russian).
Ivchenko, VM 1966 Non-stationary problems of super-cavitated foil. Proc. Hydrodynamics of
wing surfaces. Kiev, Nauk. Dumka.
Ivchenko, VM 1976 Hydrodynamics theory of supercavitating pumps and turbines. MZG №1,
p153-158 (in Russian).
Jaswon, M 1963 Integral equation methods in potential theory. Proc. Royal Soc, Ser.A. 275
Jessup, SD, Wang, HC 1997 Propeller design and evaluation for a high speed patrol boat
incorporating iterative analysis with panel method. Trans. Prop/Shafting'97 Symp., Virginia
Beach, p11/1-20.
Johnson, VE 1958 The influence of depth of submersion. Aspect ratio and thickness of
supercavitating hydrofoils operating at zero cavitation number. Proc. 2nd Symp. Naval
Hydrodynamics. Washington, DC
Johnson, 1961 Theoretical and experimental investigation of supercavitating hydrofoils operating
near the free surface. NASA, TPR-93.
Joukowski, NE 1890 Modification of Kirchhoff method for determination of a two-dimensional
flow for given constant speed on unknown streamline. Math. Sbornik, v.XV, SPb, p121-278
(in Russian).
565
References
Joukowski, NE 1891 Determination of a flow by given some conditions on a streamline. J.
Russian Physical and Chemical Society, M. V. XXII (in Russian).
Joukowski, NE, Chaplygin, SA 1904 On friction of a lubricating layer between pin and bearing.
Col. Phys. Sciences. M. 34p. (also Chaplygin, SA Selected works. M. “Nauka”, pp289-304).
Kaplun, S 1957 Low Reynolds number flow past a circular cylinder. J. Math. Mech. 6. p595-603.
Karlikov, VP, Sholomovich, GI 1966 Method of numerical calculation of influence of tannel
wall on cavitating flow of foils in the tunnel. MZG, No 4, p89-93.
Karlikov, VP, Tolokonnikov, SL 2004. On possible models of cavity closing. MZG, 2, p133-139.
Karman, T 1929 The impact of sea plane floats during landing. NACA T N 321.
Karman, T, Rubach, H 1912 Ueber den Mechanismus des Fuessigkeits- and Luftwiderstand.
Physikalische Zeitschrift, V. 13, 48.
Kawaguti, M 1953 Discontinuous flow past a circular cylinder. J. Phys. Soc. Jap, 8(3), p403-406.
Kawakami, E, Williams, M, Arndt, REA 2009 Investigation of the Behavior of Ventilated
Supercavities. Proc. CAV2009, Ann Arbor, MI.
Kedrinskij, VK 2000 Hydrodynamics of explosion. Experiments and models. Novosibirsk:
Siberian Branch of RAS, 434 p. (in Russian)
Kedrinskij, VK, Lavrent’ev, MA 1983 Models in problems of the unsteady currents with free
boundaries. Problems of Math. and Mech. Novosibirsk: Science, p97-116. (in Russian)
Keldysh, MV, Lavrent'ev, MA 1935 To the theory of oscillating wing. Tech. Notes of TsAGI.
Keldysh, MV, Sedov, LI 1937 Effective solving some boundary problems for harmonic function.
DAN USSR, 16: 1, p7-10.
Keller, AP 1998 Cavitation Scale Effects: A Representation of Its Visual Appearance and
Empirically Found Relations. Proc. NATO/AGARD Fluid Dynamics Panel Workshop on High
Speed Body Motion in Water, AGARD Rpt 827, p30—1-10.
Keller, AP 2001 Cavitation Scale Effects Empirically Found Relations and the Correlation of
Cavitation Number and Hydrodynamic Coefficients. Proc. Cav-2001 Symp. Pasadena
Kermeen, RW 1961 Experimental investigation of three -dimensional effects on Cavitating
hydrofoils. J. Ship Res., 5, #2, 22.
Kevorkian, J, Cole, JD 1996 Multiple Scale and Singular Perturbation Problems. Springer-
Verlag, New York, NY.
Kiceniuk, T 1954. An experimental study of the hydrodynamics forces acting on a family of
cavity-producing conical bodies of revolution inclined to the flow. CIT Hydrodynamics
Report E-12.17, Pasadena, CA
Kim, JW, Bai, KJ, Ertekin, RC, Webster, WC 2002 A strongly-nonlinear model for water waves
in water of variable depth - the irrotational Green-Naghdi model. Proc. OMAE 02.
Kim, S 2009 A numerical study of unsteady cavitation on a hydrofoil. Proc. 7th Symp on
Cavitation, paper #56, Ann Arbour, MI.
King, AC, Bloor, MIG 1990 Free streamline flow over curved topography. Quart. Appl. Math. V.
48. No 2, p281-293.
Kinnas, SA 1991 Leading-Edge Corrections to the Linear Theory of Partially Cavitating
Hydrofoils. J. of Ship Research, Vol. 35, No. 1, p15-27.
Kinnas, SA 1996 An Int’l consortium on high-speed propulsion. MT, Vol.33, #.3, p203-210.
Kinnas, SA 1998 The prediction of unsteady sheet cavitation. Proc. 3th Simp. “Cavitation”,
Grenoble, v.1, p19-36.
566
References
Kinnas, SA 2001 Super-cavitating hydrofoils and propellers: prediction of performance and
design. Preprint: Von Karman Inst. for Fluid Dynamics, RTO AVT/VKI Special Course:
"Supercavitating Flows", Belgium.
Kinnas, SA, Fine, NE 1991 Analysis of the Flow Around Supercavitating Hydrofoils with
Midchord and Face Cavity Detachment. J. Ship Res., Vol.35, No.3, p198-209.
Kinnas, SA, Fine, NE 1993 A Numerical Nonlinear Analysis of the Flow around Two- and
Three-Dimensional Partially Cavitating Hydrofoils. JFM, Vol.254, p151-181.
Kinnas, SA, Kosal, EM, Young, J 1999 Computational techniques for the design and analysis of
super-cavitating propellers. Proc. FAST'99,
Kinzel, MP et al. 2007 Computational Investigation of Air Entrainment, Hysteresis and Loading
for Large-Scale, Buoyant Cavities. (with JW Lindau, J Peltier, F Zajaczkowski, T Mallison,
RF Kunz, R Arndt, M Wosnik) Proc. 9th Numer. Ship Hydrodynamics. Ann Arbor, MI.
Kirschner, IN 1997 Supercavitating Projectile Experiments at Supersonic Speeds. Abstract in
Proc. NATO/AGARD Fluid Dynamics Panel Workshop on High Speed Body Motion in Water,
AGARD Report 827, Kiev, Ukraine.
Kirschner, IN 1998 Supercavitating Projectile Experiments at Supersonic Speeds. Proc.
NATO/AGARD Fluid Dynamics Panel Workshop on High Speed Body Motion in Water,
AGARD Report 827, p35—1-3.
Kirschner, IN 2001 Results of Selected Experiments Involving Supercavitating Flows. Proc.
Supercavitating Flows, VKI/RTO, Brussels. (also Lecture Notes for the RTO AVT/VKI Special
Course on Supercavitating Flows, von Karman Inst. for Fluid Dynamics, Rhode Saint
Genèse, Belgium.
Kirschner, IN 2006 A Robust Solver for Delay Differential Equations. Proc. 2006 Conf. High-
Speed Hydrodynamics & Numerical Simulation, Kemerovo State Univ., Kemerovo, Russia.
Kirschner, IN, Arzoumanian, SH 2008 Implementation and Extension of Paryshev’s Model of
Cavity Dynamics. Proc. Int’l Conf. on Innovative Approaches to Further Increase Speed of
Fast Marine Vehicles, Moving Above, Under and in Water Surface, SPb.
Kirschner, IN et al. 1995 Supercavitating Projectiles in Axisymmetric Subsonic Liquid Flows.
(with JS Uhlman, AN Varghese, IM Kuria) Proc. ASME & JSME Conf, Cavitation
&Multiphase Flow Forum, FED210, Hilton Head, SC
Kirschner, IN et al. 2001 Numerical Modeling of Supercavitating Flows. (with NE Fine, DC
Kring & JS Uhlman) Lecture Notes for the RTO AVT/VKI Special Course on Supercavitating
Flows, von Karman Institute for Fluid Dynamics, Rhode Saint Genèse
Kirschner, IN et al. 2002 Control Strategies for Supercavitating Vehicles. (with DC Kring, AW
Stokes, NE Fine, JS Uhlman). J. Vibration and Control, 8, 2, Sage Publications, London
(Also in 8th Int’l Symp. on Nonlinear Dynamical Systems, Blacksburg, VA.)
Kirschner, IN et al. 2006 High-Speed Projectile Behavior in the Bubbly Wake. (with JR Grant,
JS Uhlman) Torpedo Defense issue of J. Underwater Acoustics.
Kirschner, IN et al. 2009 A Simple Approach to Estimating Three-Dimensional Supercavitating
Flow Fields. (with RE Chamberlin, SH Arzoumanian) Proc. CAV09, Pap. 145, Ann Arbor, MI
Kirschner, IN, Rosenthal, BJ 2006 The Effects of Added Mass on the Initiation of a Supercavity
using a Pressurized Canister. Proc. CAV2006, Wageningen.
Kiselev, OM 1969 On the problem of a gas bubble in a plane ideal fluid flow. AN USSR, MZG,
#4, 13.
Knapp, RT, Daily, J, Hammit, FC 1970 Cavitation. New York, McGraw-Hill.
567
References
Kochin, NE 1949 Hydrodynamic theory of cascades. Gostehizdat, M. (in Russian).
Kochin, NE, Kibel, IA, Roze, NV 1955 Teorethical hydrodynamics. M., 560p.
Kojouro, LA 1980 Calculation of axisymmetric stream flows by Riabouchinsky scheme. #5:
p109-115. (in Russian).
Kolscher, M 1940 Unstetige Stromungen mit endlichem Totwasser. Luftfartforschung, Bd XVII,
5.
Konhauser, P 1984 Berechnung zweidimensionaler Totwasserstromungen um vorgegeben
Konturen. Dissertation. Institut A fur Mechanik der Univrsitat Stuttgart.
Korobkin, AA, Pukhnachov, VV 1988 The initial stage of water impact. Ann. Rev. Fluid. Mech.
20, p159-195.
Kotlyar, LM, Terentiev, AG 1970 A partial cavitating flow around a plate in a gravity field.
Proc. Seminar of value problems, Kasan State Univ., pp178-182 (in Russian).
Kotlyar, LM, Troepolskaya, OB 1975 Non-symmetric cavitating flow of gravity fluid. J.
Prikladnaya Mechanika, Kiev, Vol. XI, 9, pp.73-81 (in Russian).
Kozhukharov, P 1983 Investigation and design on cavitating screw propellers operating in
oblique flow. Ph. D. Thesis, Leningrad, LKI (now SMTU).
Krasnov, VK 2002 The Movement of an Axisymmetrical Solid with Formation of a Cavity.
Proc. Int’l Summer Scientific School on High-Speed Hydrodynamics, NANI Cheboksary.
Krasnov, VK, Kuznetsov, YV 1989 The application of the boundary integral equations to the
calculation of the axisymmetric and plane flows in a channel. Proc. Actual problems of
hydrodynamics ChSU., p71-75 (in Russian).
Kreisel, G 1946 Cavitation with finite cavitation numbers. Admiralty Res. Lab. Rep. R 1/H/36.
Krikunov, YM 1966 Differentiation of integral with logarithmic singularity and Couchi’s
integrand. Proc. Boundary Value Problems. Kazan State Univ., 3, p68-74 (in Russian).
Kruppa, C 1972 Testing of partially submerged propellers. 13th ITTC Report of cavitation
committee, Apendix V.
Krylov, VI 1967 Numerical calculation of integrals. M. “Nauka” (in Russian).
Krylov, VV 1963 On a method of solution of axisymmetric jet problems. Proc. TsAGI. Issue
897, 40 p. (in Russian).
Kudo, T, Ukon,Y 1994 Calculation of supercavitating propeller performance using a vortex-
lattice method. Proc. 2nd Int’l Symp. on Cavitation, p403-408.
Kuria, IM, Daily, J, Hammit, FC 1997 Compressible Cavity Flows Past Slender Non-Lifting
Bodies of Revolution. Proc. ASME & JSME Fluids Engineering Annual Conf., Cavitation
and Multiphase Flow Forum, FEDSM97-3262, Vancouver, BC.
Kutta, WM 1911 Ueber ebene Zirkulationstroemungen nebst flugtechnischen Anwendungen.
Sitzungsberichten der Bayer. Akad. Muenchen. S. 108.
Kuznetsov, AV 1962 Cavitating flow of the plate in a jet of finite depth. Izv. AN USSR, Mech.
and Mash., No 1.
Kuznetsov, AV 1964 On the model of cavitating flow. Proc. «Inverse value problems» Kazan
Univ., v. I.
Kuznetsov, AV 1969 Flow of imponderable fluid with free boundary around a plate. PMTF, # 6.
Kuznetsov, AV 1975 Unsteady perturbation flow of a liquid with free boundaries. Kazan Univ.
144 p.
568
References
Kuznetsov, AV 1977 Entry of thin foil into water. Proc. Seminar “Boundary Value Problems”.
Kazan Univ., 14 (in Russian).
Kuznetsov, AV, Manevich, AS 1979 Gravity effect in water entry of thin foil. Proc. NKI,
Nikolaev, NKI, 152 (In Russian).
Kuznetsov, AV, Manevich, AS 1979a Vertical Entry of an Inclined Slender Symmetric Profile
into a Compressible Fluid. Trans. AN USSR, ser. MZG, 5 (in Russian), see also Fluid Dyn..
Kuznetsov, AV, Terentiev, AG 1967 On some model of partial cavitating flow. Izv. Vuzov,
Math., 11, p43-46 (in Russian).
Kuznetsov, BG et al. 1967 Calculation of the cavity shape in the gravity field with taking into
account of surface tension.(with VN Shepelenko, NN Yanenko) Izv. Sib. Branch of AN USSR,
ser. Tech., 3, #13 (in Russian).
Kuznetsov, YV 1977 Partial cavitating flow of a plate cascade. Coll. Problems of Math. and
Mech. ChSU., 5, p51-60 (in Russian).
Kuznetsov, YV, Terentiev, AG 1980 Flow of weightless fluid around a plate under free surface.
Izv. AN USSR, MZG, No. 1, p158-162 (in Russian).
Lamb, H 1932 Hydrodynamics, 6th ed. Dover Publications, New York.
Larock, BE, Street, RL 1965 Rieman–Hilbert problem for fully cavitating flow. JSR. v.9, #. 3.
Larock, BE, Street, RL 1967 A nonlinear solution for a fully cavitating hydrofoil beneath a free
surface. JSR., v. 11, No. 2.
Larock, BE, Street, RL 1967a A non-linear theory for a full cavitating hydrofoil in a transverse
gravity fieald. JFM, v. 29, No 2, p317-336.
Lavrent’ev, MA 1957 Cumulative charge and a principe its action. UMN, Vol. XII, No 4(76),
p41-56 (in Russian)
Lavrent’ev, MA, Shabat, BV 1965 Methods of theory of functions on complex variable. M.:
“Nauka”, FML, (in Russian).
Lavrent’ev, MA, Shabat, BV 1977 Problems of hydrodynamics and their mathematical models.
M. Science, 407p
Leehey, P 1973 Influence of a free surface and walls on characteristics of a cavity foil. Proc.
IUTAM Symp. Also in: Non-steady flow of water at high speeds, M., 277-298.
Lenau, CW 1963 A nonlinear theory for supercavitating flow past a wedge in a longitudinal
gravity field. Dept.of C. E. Tech. Rep. 28, Stanford, CA.
Lenau, CW, Street, RL 1965 A nonlinear theory for symmetric supercavitating flow in a gravity
field. J. of Fluid Mech., v. 21, part. 2.
Levi-Civita, T 1907 Scie e leggi di resistenza. Rend. Circolo Math. Palermo. -Vol. XXIII.
Levinson, N 1946 On the asymptotic shape of the cavity behind an axiallysymmetric nose
moving through an ideal fluid. Annals of Math., vol.47. No.4, p704-731.
Levkovski, YL, 1973 Structure of cavitation streams. L.: Sudostroenie (in Russian)
Lighthill, MJ 1951 A new approach to thin aerofoil theory. Aero. Quart., 3, p193-210.
Lindau, JW, Kunz, RF, Boger, DA, Stinebring, DR, Gibeling. HJ 2002 High Reynolds Number,
Unsteady, Multiphase CFD Modeling of Cavitating Flows. J. Fluids Eng., 124, 3, p607-617.
Lindau, JW, Kinzel, MP, Stinebring, DR, Kunz, RF, Boger, DA, Noack RW 2006 The Effects of
a Free-Surface on Computed Turbulent, Super and Partially Cavitating Flows. Proc. 6th Int’l
Symp. on Cavitation, MARIN, Wageningen.
Liu, GR, Liu, MB 2003 Smoothed particle hydrodynamics: a meshfree particle method. World
Scientific.
569
References
Liu, PLF, Liggett, JA 1986 Boundary element formulations and solution for some non-linear
water wave problems. Development in Boundary Element Methods, p171-190.
Liu, WK, Jun, S, Sihling, DT, Chen, Y, Hao, W 1997 Multiresolution Reproducing Kernel
Particle Method for Computational Fluid Dynamics. Int’l J. Num. Method in Fluids, 24, 1–25.
Logvinovich, GV 1959 Initial moving of a body in a liquid by cavitation. Proc. Hydrodynamics.
M. TsAGI, p3-40 (in Russian).
Logvinovich, GV 1959a Body entry into fluid and moving with developed cavitation. Proc.
Hydrodynamics. M., TsAGI, p119-139. (in Russian)
Logvinovich, GV 1969 Hydrodynamics of Free-Boundary Flows. Nauk. Dumka, Kiev, Transl.by
the Israel Program for Scientific Translations, Jerusalem 1972.
Logvinovich, GV Yakimov, YL 1973 High velocity entry of bodies into a liquid. Proc. IUTAM
Sym. Non-steady Flow of Water at High Speeds, L., p85-92 (in Russian).
Logvinovich, GV et al. 1985 Free-Surface Flows. (with VI Buivol, AS Dudko, SI Putilin, YR
Shevchuk) Nauk. Dumka, Kiev. (in Russian).
Logvinovich, GV, Serebryakov, VV 1975 On Methods of Calculation of Slender Axisymmetric
Cavity Shapes. J. Hydromechanics, 32, p47-54, Kiev (in Russian).
Longuet-Higgins, MS, Cokelet, ED 1976 The deformation of steep surface waves on water. I. A
numerical method of computation. Proc. Roy. Soc. Lond. A 350, p1-26.
Longuet-Higgins, MS 1980 On the forming of sharp corners at a free surface. Proc. Royal Soc.,
London, A371, p453-478.
Loyziansky, LG 1955 Hydrodynamic theory of spherical bearing. PMM, M., p531-540.
Loyziansky, LG 1959 Mechanics of liquid and gas. M. GIFML.
Lucy LB 1977. A numerical approach to the testing of fusion process. Astron. J. 82/12, 1013-
1024.
Luu, T, Sulmont, P 1969, Design of supercavitating propellers on the basis of lifting surface
theory. J. Ship Res., 1969, v.16, №176, p116-124.
Mackerle, J, Brebbia, CA 1988 The boundary element reference book. WIT Press, 392p.
Mackie, AG 1962 A linearized theory of the water entry problem. Qurt. J. Mech. & Appl. Math.
15, 137-151.
Maklalov, DV 1997 Nonlinear problems of potential flow with unknown boundaries. M., Yanus-
K, 280 p. (in Russian).
Marchuk. AG, Chubarov, LB, Shokin, YI 1983 A numerical modulation of tsunami waves,
Nauka, Novosibirsk, (in Russian).
Matveev, KI, Burnett, T, Ockfen, A 2009 Study of Air-Ventilated Cavity Under Model Hull on
Water Surface. Ocean Eng., 36, 12-13, p930-940.
Mavlydov, MA, Rousetsky, AA, Sadovnikov,YM, Fisher, EA 1982 Propulsors for high speed
ships. 2nd Ed., Sudostroenie Publ. House, L., 280p.(in Russian)
May, A 1975 Water Entry and the Cavity-Running Behavior of Missiles. SEAHAC TR75-2,
Naval Sea Systems Command, Arlington,VA (also in Naval Surface Weapons Center, White
Oak Laboratory, Silver Spring, MD).
Mazur, VY 1966 Motion of a circular cylinder near a vertical wall. Izv. AN USSR, MZG, No. 3,
p75-79; English transl. In Fluid Dynamics I 1966).
McKenney, EA, Brennen, CE 1994 On the dynamics and acoustics of cloud cavitation on an
oscillating hydrofoil. Proc. ASME Symp. Cavitation and Gas-Liquid Flows in Fluid
Machinery and Devices, FED-190, NY, NY, p195-202.
570
References
McLeod, EB 1955 The explicit solution of a free boundary problem involving surface tension. J.
Rat. Mech., 4, 557. 3.
Meijer, MC 1959 Some Experiments on Partly Cavitating Hydrofoils. Int’l Shipbuilding
Progress, Vol. 6, No. 60.
Mensaas, K, Slaattelid, OH 1971 Lifting-surfaces corrections for 3 bladed optimum propellers.
ISP, vol.18, №208, pp.437-452.
Michel, R 1951 Etude de la Transition sur les Profiles d’Aile: Etablissement d’un Critére de
Determination de Point de Transition et Calcul d la Trainee de Profile Incompressible. Tech.
Report 1/1578A, ONERA.
Michel, JM 1973 Ventilated cavities. Proc. IUTAM Symp “Non-steady flow of water at high
speeds”, M., Nauka, p343-360.
Mikhailov, VM 1986 A moving of a plane with variable speed in perfect incompressible liquid.
Proc. Topical problems of cont. mech. (Ed. .Terentiev). ChSU, 70-79 (in Russian).
Mikhailov, VM 1987 Inclined entry of a thin plane in perfect incompressible liquid. Proc.
Interaction of foils in the fluid with free surfaces. (Ed. Terentiev). ChSU, 77-86 (in Russian) .
Mikhailov, VM 1988. Inclined entry of a thin plane of large span in perfect incompressible
liquid. Proc. Hydrodynamics of bounded stream. ChSU, 93-101 (in Russian).
Mikhailov, VM, Nikitina, GV 1980 On inclined penetration of thin foil with ventilated cavity
into ideal weightiness fluid. In “Dynamics of cont. medium with free surfaces”. CSU, p96-
112 (in Russian).
Mokry, M 1990 Complex variable boundary element method for external potential flows. AIAAJ.
V. 127. pp.1-11.
Monaghan, JJ 1992 Smoothed particle hydrodynamics. Ann. Rev. Astron. & Astrophysics 30,
543-574.
Monaghan, JJ, Thompson, MC, Hourigan, K 1994 Simulation of free surface flows with SPH. J.
of Comp. Physics, 110, 399-406.
Moran, JP 1961 The vertical water-exit-and-entry of slender symmetric bodies. J. Aerospace Sci.,
28, 803-812.
Morris, JP, Fox, PJ, Zhu, Y 1997. Modeling low Reynolds number incompressible flows using
SPH. J. Comp. Physics, 136, 214-226.
Münzer, H, Reichardt, H 1950 Rotational Symmetrical Source-Sink Bodies with Predominantly
Constant Pressure Distributions. A.R.E. Translation 1/50.
Murty, TC 1981 Seismic Waves of Tsunami. L., Hydrometeoizdat, 488p (in Russian)
Muskhelishivili, NI 1953 Singular integral equations. P. Noordhoff N.V. Groningen-Holland.
Nakos, DE, Sclavounos, PD 1990 Ship Motions by a Three Dimensional Rankine Panel Method.
Pros. 18th Naval Hydrodynamics Conf., Ann Arbor, MI.
Nekrasov, AI 1947 Theory of a wing in non-stationary flow. AN USSR. Press
Nekrasov, AI 1959 Flow around the Zhuckovski’s airfoil with a source and a sink on the
boundary. Proc. of Steklov Mathematical Inst.. No 53, pp 236-265 (in Russian).
Newman, JN 1977 Marine Hydrodynamics. MIT Press, Cambridge, MA.
Newton, RN, Rader, HP 1961 Performance data of propellers for high-speed craft. Trans. RINA,
Vol.103.
Nikitin, VV, Porfiriev, NP 1985 Crossing of free boundary of two gravity liquids by a thin body.
Proc. High Speed Hydrodynamics. CSU 101-10. (in Russian)8
571
References
Nikitin, VV, Terentiev, AG 1980 Non-stationary moving ofthin foil in fluid with inner free
boundaries. Proc. Dynamics of Continuum wit Free Boundaries, CSU, p113-123 (in Russian)
Nikitin, VV, Terentiev, AG 1982 Horizontal motion of a thin foil in stratified fluid. Proc.
Dynamics of continuum with free boundaries. ChSU, p112-118
Nikitin, VV, Terentiev, AG, 1985 Intersection of free boundary of two liquids by a wedge with
Kirchhoff cavity. Proc. Interaction of bodies with free fluid boundaries, Ch. p93-100 (in
Russian)
Nishiyama, T 1970 Lifting-line theory of supercavitating hydrofoil of finite span. ZAMM, 50,
645-653.
Nishiyama, T, Khan, O 1981 Compressibility Effects on Cavitation in High-Speed Liquid Flow
(Transonic And Supersonic Liquid Flows). Bull. Jap. Soc. of Mech. Eng., 24, 190.
Nishiyama, T, Kobayshi, H 1969 Finite cavity flow of axial symmetry. Tech. Reports, Tohohu
Univ. 34, No.1, p173-185.
Numachi, FI, Chida, J 1958 Cavitation tests on hydrofoils profiles of simple form. Rep. Inst.
High Speed Mech. Japan, v. 9.
Oba, R 1964 Theory of supercavitating hydrofoils at arbibtrary cavitation coefficient. Science
Rpt. Of Res. Inst., Tohoku Univ.,Ser B, High Speed Mech. 20p.
Ogilvie, TF, Tuck, EO 1969 A Rational Strip Theory for Ship Motions. Pt I. Report # 013, Dept.
of Naval Architecture and Marine Engineering, Univ. of Michigan.
Olofsson, N 1996 Force and flow characteristics of a partially submerged propeller. Doct.
dissertation, Chalmers Univ. of Technology, Goteborg.
Oseen, CW 1910 Über die Stokes'sche Formel und über eine verwandte Aufgabe in der
Hydrodynamik. Arch. Math. Astronom. Phys. V. 6, No. 29.
Otta, SP 2003 Blow-down Theory: Blow-down of a Pressurized Tank. Course notes for
Aerospace Engineering, AerE 311L, Gas Dynamics Laboratory, Iowa State University, Ames,
IA, USA, http://www.public.iastate.edu/~aero311l/blow-down.pdf
Otta, SP 2003 Blow-down Theory: Blow-down of a Pressurized Tank. Course notes for
Aerospace Engineering, AerE 311L, Gas Dynamics Laboratory, Iowa State University, Ames,
IA, USA, http://www.public.iastate.edu/~aero311l/blow-down.pdf
Ovsyannikov, LV 1967General equasions and examples. The problem of non-steady movement of
liquid with free boundaries. Novosibirsk, Nauka (in Russian)
Panchenkov, AN 1975 Theory of acceleration potential. Novosibirsk, Nauka, 244p. (in Russian)
Parkin, BR 1957 Fully cavitating hydrofoils in nonsteady motion. Eng. Div. No. 85-2, CalTech.
Paryshev, EV 1978 A System of Nonlinear Differential Equations with a Time Delay, Describing
the Dynamics of Non-Stationary, Axially Symmetric Cavities. Trudy TsAGI, #1907, p3-16
Paryshev, EV 1985 The principle of cavity expansion independence as a method of investigating
the non-starionary cavitating flows. Proc. TsAGI, 2256p43-50 (in Russian)
Paryshev, EV 2002 The Plane Problem of Immersion of an Expanding Cylinder through a
Cylindrical Free Surface of Variable Radius. Proc. the Int’l Summer Scientific School on
High-Speed Hydrodynamics, NANI, Cheboksary
Paryshev, EV 2006 Approximate Mathematical Models in High-Speed Hydrodynamics. J.
Engineering Mathematics, 55, p41-64.
Pavlova, NA, Terentiev, AG 2010 Computer simulation of flow past arbitrary hydrofoils with
partial cavitation. MZG, No 3, p33-42. (Transl. Fluid Dyn. 2010, v45, #3, p369-377)
572
References
Peregrine, DH, Cocelet, ED, McIver, ED 1980 The fluid mechanics of water approaching
breaking. Proc. 17th Conf on Coastal Engineering, ASCI, 1, p512-528.
Pernik, AD 1966 Problems of cavitation. “Sudostroenie”, L., 439p.
Petrov, AG 1981 Direct variational method of analyzing the flat and axisymmetric cavities. DAN
USSR, v.257, #6, p1323-1327. (in Russian)
Petrov, AG 1986 Asymptotic expansions for slender axisymmetric cavities. PMTF, #5, p45-49.
Petrov, AG 2007 Asymptotic expansions for axially symmetric cavities. Eur. J. Appl. Math. 16,
No. 3, 319-334
Petrov, AG 2008 Quadrature formulas for periodic functions and their application to the
boundary element method. Comput. Math. & Math. Phys. 48, No. 8, 1266-1283.
Petrov, AG, Smolyanin, VG 1993 Calculation of unsteady waves on the surface of a heavy fluid
of finite depth. Applied Math. & Mech., 57, № 4, p137-143.
Petrov, NP 1883 Friction in machines and influence on it of greasing oils. J. of Eng., M.
Plesset MS, Chapman, RB 1971 Collapse of an initially spherical vapor cavity in the
neighborhood of a solid boundary. JFM, 47, 283--290.
Plotkin, A 1978 Leading-Edge Correction for Supercavitating Flat-Plate Hydrofoil. J. Fluids
Engineering, Vol. 100, p276-280
Pologiy, GN 1965 An extension of the theory of analytic functions of complex variables. Kiev
State Univ., 424 p (in Russian).
Polubarinova-Kochina, P 1977 Theory of moving of underground water. M. Nauka, 664p
Popov, VV 1979 The problem of impact of a floating cone. Hydromechanics, v.40, p28-35.
Porfiriev, NP 1978 Vertical moving of a thin body to free surface of a gravity liquid. Proc. Jet
and cavitating flow, and problems of the theory of controls. CSU 86-99 (In Russian).
Porfiriev, NP 1981 Vertical moving of a thin axisymmetric body to free surface of a gravity
liquid. Proc. High Speed Hydrodynamics. CSU 100-109 (In Russian).
Porfiriev, NP, Troeshestova, DA 1996 Generation of gravity waves by sudden bottom removal at
finite interim. Proc. 4th HSH-96, Ch. NANI, p.149-158 (in Russian)
Poruchikov, VB 1964 Impact of the disk on a ideal compressible fluid surface. PMM, M., 28, No.
4, p 797-800 (in Russian), English transl. in Fluid Dynamics
Poruchikov, VB 1973 Penetration of a Cone into a Compressible Fluid. PMM, 37, 1, p84-93, (in
Russian), English transl. in Fluid Dynamics
Pozdunine, VL 1941 Problems of ship propellers. Soviet Science, No. 2 (in Russian).
Pozdunine, VL 1944 On the working of supercavitating screw propellers. DAN USSR, Vol.
XXXIX, № 8, p334-339. (in Russian), English transl. in Fluid Dynamics
Pozdunine, VL 1947 The effectiveness of supercavitating propellers. Izv. OTN AN USSR, No. 10.
(in Russian), English transl. in Fluid Dynamics
Prishchemikhin, YN 1969 Cavitation tank with remote testing operation. Trans. 12th ITTC
Prosnak, W, Kucharzyk, P 1959 The influence of the ground on the aerodynamic properties of an
airofoil with jet flap. Nadpitka zarchiwin mechaniki stoswanej. Warsaw, v. 11.
Prosperetti, A 1987 The equation of bubble dynamics in a compressible liquid. Phys. Fluids 20,
p3626-3628
Proudman, I, Pearson, JRA 1957 Expansions at small Reynolds numbers for the flow past a
sphere and a circular cylinder. JFM 2. p237-262.
573
References
Pyhteev, GN 1956 Solving of inverse problem of cavitating flow around curvilinear arc. PMM, v.
XX, No 3. (in Russian) English transl. in Fluid Dynamics
Qin, Q, Song, CCS, Arndt, REA 2003 A Virtual Single-Phase Natural Cavitation Model and its
Application to Cav2003 Hydrofoil. Proc. 5th Symp. on Cavitation, Paper 1-004.
Rayleigh, L. 1917 On the pressure developed in a liquid during the collapse of a spherical
cavity. Phil. Mag., 34, 94--98.
Reichardt, H 1946 The Laws of Cavitation Bubbles at Axisymmetric Bodies in a Flow. Ministry
of Aircraft Production Volkenrode, MAP-VG, Reports and Translations 766, ONR, Arlington.
Reisman, GE, McKenney, EA, Brennen, CE 1994 Cloud cavitation on an oscillating hydrofoil.
Proc. 20th ONR Symposium on Naval Hydrodynamics, p328-340.
Reisman, GE, Wang, YC, Brennen, CE 1998 Observations of shock waves in cloud cavitation.
JFM, 355, p255-283.
Riabouchinsky, D 1920 On steady fluid motion with free surfaces. Proc. London Math Soc. 19,
ser. 2.
Riegels, F 1948 Das Umstromungsproblem bei inkompressiblen Potentialstromungen. Ing. Arch.,
Bd. 16, p373-376; 1949, Bd. 17, p94-106.
Risenkampf, BK 1940 hydraulics of underground water, pt.1. Saratov Univ. v.15, No 5, p3-93.
Roberts, WB 1980 Calculation of Laminar Separation Bubbles and their Effect on Airfoil
Performances. AIAA J., 18, p25-31.
Roman, VM, Makaseev, MV 1984 Calculation of cavity shape past supercavitating wing of finite
span. Proc. Dynamics of continuoues fluid with non-stationary boundaries. CSU, 103-109 (in
Russian).
Rose, JC 1993 Combination surface propeller and water jet systems. Proc. 3rd Int’l Design Symp.
on Yacht and Small Craft Design, Chianciano Terme, Italy.
Rose, JC, Kruppa, CFL 1991 Surface piercing propellers - methodical series model test results.
Proc. FAST'91, Trondheim.
Roshko, A 1954 New hodograph for free-streamline theory. NACA Tech.Note 3168.
Roshko, A 1955 On the wake and drage of bluff bodies. J. Aeronaut. Sci., Vol. 22, No 2.
Rowe, A, Blottiaux, O 1993 Aspects of Modeling Partially Cavitating Flows. JSR, 37, p39-50.
Rozhdestvensky, KV 1977 Cavitation. Textbook, Sudostroenie Pubs. L., 247p (in Russian)
Rozhdestvensky, KV 1979 Method of intertwisted asymptotic development in hydrodynamics of
a wing. Sudostroenie, L.. 208p. (in Russian)
Rozhdestvensky, KV, Fridman, GM 1991 Matched asymptotics for free surface lifting flows with
spoilers. In: Matched Asymptotics in Hydrodynamics, SIAM. Philadelphia, p499-517.
Rudzki, MP 1893 Ueber eine Klasse hydrodynamischer Probleme mit besonderen
Grenzbedingungen. Math. Ann., No 50.
Rusetsky, AA, Zhuchenko, MM, Dubrovin, OV 1971 Marine propellers. Sudostroenie, L. 287p.
Rusetsky, AA 2000 Alar ships. Maritime J., Sudostroenie Pubs., #2, p46-49.
Sadovnikov, YM 1982 Supercavitating propellers-design calculation using lifting-surface theory.
In book: Fast Ferry Propulsors, Sudostroenie Pubs., 280 p(in Russian).
Sagomonyan, AY 1959 Falling flat plate on a compressible fluid. Vestnik MGU, Math. & Mech.,
No. 2, p49-53 (in Russian).
Sagomonyan, AY 1974 Penetrating. M. Univ. Press. 299p.
574
References
Salimov, RB 1970 Some basic problems of boundary deformation of the theory of analytic
functions and applications to hydrodynamics. Kazan, Milit. Eng. School Press, 210p.
Sato, K, Tomita, Y 1998 Non-spherical motion of two cavitation bubbles produced with/without
time delay near a boundary. Proc. 3rd Int. Symp. on Cavitation. v.1, p63-68.
Saurel, R, Cocchi, JP, Butler, PB 1999 Numerical Study of Cavitation in the Wake of a
Hypervelocity Underwater Projectile. J. Propulsion and Power, 15, 4, p513-523.
Sautreaux, C 1893 Sur une question d’hydrodynamique. Ann. Sci. Ecole Norm. super. v.10
Savchenko, YN 2001 Experimental investigation of supercavitating motion of bodies. Proc.
«Supercavitating Flows» VKI/RTO, Brussels.
Savchenko, YN et al. 1985 Phenomenon of gas capture by cavitating bubbly flow around
obstacls. (with VN Semenenko & SB Osipenko) Proc. HSH’85 Conf., CSU Press, p108-116
(in Russian)
Savchenko, YN, Semenenko VN, Serebryakov, VV 1993 Experimental Research in Subsonic
Cavitating Flows. DAN of Ukraine, N.2, p64-68, (in Russian).
Savchenko, YN et al. 2000 Theory of Stable Model Motion with Ventilated and Unventilated
Supercavities. (with VN Semenenko, SI Putilin, YD Vlasenko, VT Savchenko, YI Naumova)
IHM-NASU Review for ONR - Europe, Inst. Hydromechanics, Ukr. NAS, Kiev
Schiebe, ER, Wetzel, JM 1961 Ventilated cavities on submerged three-dimensional hydrofoil.
Uni. of Minnesota, AF Hydaulic Lab. Paper No 36, B.
Schlichting, H 1979 Boundary-Layer Theory. transl. by J. Kestin, McGraw-Hill, NY.
Schmidt, SJ, Thalhamer, M, Schnerr, GH 2009 Inertia Controlled Instability and Small Scale
Structures of Sheet and Cloud Cavitation. Proc. CAV2009, Ann Arbor, MI
Schmieden C 1929 Die unstetige Stroemung um einen Kreiszylinder. Ing. Arch. Bd.12, h.5
Schnerr, GH, Sezal, IH, Schmidt, SJ 2008 Numerical Investigation of Three-Dimensional Cloud
Cavitation with Special Emphasis on Collapse Induced Shock Dynamics. Phys. Fluids, 20.
Schubauer, GB, Skramstad, HK 1948 Laminar boundary-layer oscillations and transition on a flat
plate. NACA report 909
Sedov, LI 1934 On an impact of solid body floated on a surface of a liquid. Proc. TsAGI, M., v.
187 (in Russian).
Sedov, LI 1935 Falling of a wedge on the water surface. Proc. TsAGI, v.52 (in Russian).
Sedov, LI 1936 Theory of non-stationary planning and movement of a wing with trailing
vortexes. Proc. TsAGI, 252 (in Russian).
Sedov, LI 1950 Two-dimensional problems in hydrodynamics and aerodynamics. M.-L. 444 p.
(English Translated, Wiley, XV 427 p.). 1, 1
Sedov, LI 1970 Mechanics of continuum. M. Nauka, 492p.
Seidel, W 1952 Bibliography of numerical methods in conformal mapping. Nat. Bureau Stand.
Appl. Math. Ser. Vol. 18, p269-280.
Semenenko, VN 2001 Artificial Supercavitation: Physics and Calculation. Lecture Notes for the
RTO AVT/VKI Special Course on Supercavitating Flows, von Karman Institute for Fluid
Dynamics, Rhode Saint Genèse, Belgium.
Serebryakov, VV 1973 Asymptotic solution for slender axisymmetric cavity. J. DAN, Ukr. SSR,
No.12, p1119-1122. (in Russian)
Serebryakov, VV 1974 Ring model for calculation of axisymmetric flows with developed
cavitation. J. Hydromechanics, Kiev. No.27, p25-29. (in Russian)
575
References
Serebryakov, VV 1976 About one variant of the equations of the principle of independent cavity
expansion. J. Hydromechanics, Kiev., No.34, p45-48. (in Russian)
Serebryakov, VV 1977 On statement of linearized problems of axisymmetrical supercavitation in
unsteady flows. In: “Math. methods of hydrodynamics flow study. Kiev, p58-62. (in Russian)
Serebryakov, VV 1986 Dependencies to limit an unsteady flow over slender axisymmetric
bodies. J. Bionics, Kiev. No.20, p21-32. (in Russian)
Serebryakov, VV 1990 Asymptotic solutions of axisymmetric problems of the cavtational flow
under slender body approximation. In: Hydrodynamics of high speeds, ChSU, p99-111. (in
Russian)
Serebryakov, VV 1992 Asymptotic Solutions of Axisymmetric Problems for Sub- and
Supersonic Separated Flows at Zero Cavitation Numbers. DAN of Ukraine, 9, p66-71, (in
Russian).
Serebryakov, VV 1998 Asymptotic approach for problems of axisymmetric supercavitation
based on the slender body approximation. Proc. 3rd Int. Symp. on Cavitation, Grenoble, V.2,
p61-70.
Serebryakov, VV 2002 The models of the supercavitation prediction for high speed motion in
water. Proc. Sci. School: HSH-2002, Cheboksary, p71-92. (in Russian)
Serebryakov, VV 2007 Compressibility Effects for Very High Sub-, Trans- and Supersonic
Speeds of Motion in Water. Proc. APM2007, SPb, p378-390, (in Russian).
Serebryakov, VV 2009 Physical-mathematical bases of the principle of independence of cavity
expansion. Proc. CAV2009, Ann Arbor, Michigan, p14.
Serebryakov, VV et al. 2008 Subsonic, Transonic, and Supersonic Motion in Water with
Supercavitation. (with IN Kirschner & GH Schnerr) Proc. SuperFAST 2008, MTU, SPb.
Serebryakov, VV, Schnerr, GH 2003 Some Problems of Hydrodynamics for Sub- and Supersonic
Motion in Water with Supercavitation. Proc. 5th Int’l Symp. CAV2003, Osaka.
Serebryakov, VV, Schnerr, GH 2004 Sub-, Trans-, and Supersonic Flows in Water with
Supercavitation. Proc. 2004 High-Speed Hydrodynamics Int’l Summer Scientific School
(HSH2004), NANI, Cheboksary.
Shahverdy, GG 1984 Investigation of penetrating of solid body into compressible fluid by using
the conservative method of finite methods. In «Dynamics of Elastic and solid bodies by
interaction with liquid». Tomsk, p170-179 (in Russian).
Shepelenko, VN 1968 To the calculation of cavitation flows. PMTF, #5, p100-105 (in Russian)
English transl. in Fluid Dynamics
Sherykhalina, NM, Zhitnikov, VP 2002 Application of extrapolation methods of numerical
results for improvement of hydrodynamics problem solution. CFD J. v 11, #2, p155-160
Sheshukov, EG 1968 To problem of a movement of two foils close to free boundary of two
fluids. Proc. Seminar on Value Problems. Kazan State Univ., 5, p262-271 (in Russian).
Sidorov, OP 1958 The solution of a problem on a flow of a body of rotation. Trans. Kazan
Aviation Institute, XXXVIII, Kazan, p23-42 (in Russian)
Signorini, A 1916 Sorga un problema al contorno nella teoria delle funzioni di variable
complessa. Annali di Mathematica. 3 serie, t. 25.
Silberman, E, Song, CS 1961 Instability of Ventilated Cavities. JSR, 5, p13-33
Singhal, AK, Athavale, MM, Li, H, Jiang, Y 2002 Mathematical Basis and Validation of the Full
Cavitation Model. J. Fluids Eng., 124, p617-624.
576
References
Skalak, R, Feit, D 1966 Impact on the surface of a compressible fluid. Trans ASME, J. Eng. Ind.,
v.88, No. 3, p325-331.
Slezkin, NA 1951 Plane flow of an ideal fluid past a gas-filled shell. Uch. Zap. M. State Univ.
Mekh. 3, 61.
Smirnov, VI 1957 Сourse of high mathematik. M. v. II, IV.
Snider, AD 1971 Numerical solution of nonlinear boundary value problems using reflection.
Reports NYU-NYO-1480-167.
Sommerfeld, A 1904 Zur hydrodynamischen Theorie der Schmiermittelreibung. Zeitschrift fur
Mathematik und Physik, No. 50.
Soyama, H et al. 1998 Useful correlations for cavitating jet. Proc. Third Int. Sym. on Cavitation,
Grenoble. V. 2, pp.147-156.
Springorum, K, 1932 Hydromechanische Probleme des Schiffsantriebs. Hamburg, 331 p.
Stepanov, GY 1962 Hydrodynamics of cascades of turbo-machines. FM, M. 512p (in Russian)
Stepanov, GY 1998 Design of flow-separation free bodies in combination with a propulsor. Proc.
Problems of Modern Mech. M. State Univ. p109-117.
Stoker, JJ 1957 Water Waves. NY, Interscience Publishers.
Stokes, GG 1851 On the effect of the internal friction of fluids on the motion of pendulums.
Trans. Camb. Phil. Soc. V. 9. Part. II, p8-106.
Street, RL 1963 Supercavitating flow about a slender wedge in a transverse gravity field. JSR 7, 1
Sturova, IV 1990 Numerical calculations in the problems of generation of plane surface waves.
Preprint 5. Computer Center, Siberian Branch of AS USSR, Krasnoyarsk. 48p.
Subkhanculov, GI, Khomyakov, AN 1990 The application of the boundary elements method to
the calculations of the axisymmetric cavities. High Speed Hydrodynamics (ed. Terentiev),
ChSU, p124-132 (in Russian).
Susumi, T Nagamiga, T, Takemouti, J 1933 The lift on a plate placed near a plane wall with
special reference to the effect of the ground upon the lift of a monoplane aerofoil. Aeronaut.
Res. Inst. Tokyo Imp. Univ., v. 8, No. 97, p1-60.
Symm, GT 1963 Integral equation methods in potential theory II. Proc. Royal Soc., Ser. A, 275.
Tachmindji, AJ, Morgan, WB 1958 The design and estimated performance of a series of
supercavitating propellers. Proc. 2nd ONR Symp. on Naval Hydrodynamics.
Taic, OG 1985 Axisymmetric cavities. J. Bionics, Kiev, #19, p73-80. (in Russian)
Takahira H, Yasuda A 1998 Effects of bubble-bubble/bubble-wall interaction on the nonlinear
oscillations of bubbles. Proc. 3rd Int. Symp. on Cavitation, Grenoble,
Tannery, J, Molk, J 1898 Elements de la theorie de functions elliptiques. Paris, Vol. 3.
Telste, JG 1986 Inviscid flow about a cylinder rising to a free surface. JFM #182, p149-168.
Terentiev, AG 1962 To the theory of a ring wing. Ann. post-grad. papers, Kasan State Univ.,
p203-215.
Terentiev, AG 1964 Cavitating flow around flat plate. Izv. VUZ, Math. #6, p159-167 (in Russian)
Terentiev, AG 1967 The flow of a plate cascade with full cavitation. MZG, No 2, p52-58.
Terentiev, AG 1970 Partial cavitating flow of a plate. Izv. Vuzov, Math., 11, p112-118
Terentiev, AG 1971a To linear theory of cavitating flow of obstacles. Proc. Problems of appl.
math. & mech. ChSU, p3-35 (in Russian).
Terentiev, AG 1971b On the solution of a mixed boundary value problems. DAN USSR. v196,
57-60; English transl. in Soviet Math. Dokl. 12 1971).
577
References
Terentiev, AG 1972a To solving of linear problem of cavitating flow around curvilinear arc. Izv.
AN USSR, MZG, M.,#1, p34-38. English transl. in Fluid Dynamics
Terentiev, AG 1972b Jet flow around thin hydrofoil in a bounded stream. Izv. AN USSR, MZG,
M., # 2, p137-139. English transl. in Fluid Dynamics
Terentiev, AG 1974 Boundary problems of linear theory of jets. Proc. Symp. “Mech. of
continuum fluids and related problems of analysis”, Tbilisi, Metsnierba, II, p255-268. (in
Russian)
Terentiev, AG 19776 On non-linear theory of cavitating flows. Izv. AN USSR, MZG, # 1, p158-
161 (in Russian). English transl. in Fluid Dynamics
Terentiev, AG 1977a Non-linear theory of cavitating flows. In «Problems of Appl. Math. &
Mech.» ChSU, 5, p138-185 (in Russian).
Terentiev, AG 1977b Inclined entry of a thin body into incompressible liquid. Izv. AN USSR,
MZG, No. 5, p16-24 (in Russian). English transl. in Fluid Dynamics
Terentiev, AG 1979 Inclined entry in ideal non-gravity liquid of thin body with ventilated cavity.
Izv. AN USSR, MZG, # 3, p66-76 (in Russian), see also Fluid Dynamics.
Terentiev, AG 1981a Mathematical problems on cavitation. ChSU, 131 p. (in Russian).
Terentiev, AG 1981b To the theory on unsteady moving of bodies in a liquid. In «High Speed
Hydrodynamics» ChSU, p126-138 (in Russian).
Terentiev, AG 1985 Evaporation effect on cavitating flow around a body. In «Interaction of
bodies with separation boundaries of continuous medium» ChSU, p113-122 (in Russian).
Terentiev, AG 1987 On momentum and kinetic energy of a fluid by impact and entry of solid
body. In «Interaction bodies in fluid with free boundaries» ChSU, p106-114 (in Russian).
Terentiev, AG 1989 Nonstationary motion of bodies in a fluid. Proc. Steclov Inst. of Math., 186.
(Translated 1991, Issue 1 p211-221.)
Terentiev, AG 1993 Applying of generalized analytic functions in hydrodynamics. In “Problems
of high speed hydrodynamicsChSU p10-25
Terentiev, AG 1994 Numerical investigations in hydrodynamics. Izv. Academy of Science,
Chuvash Republic, N 2, p61-84 (in Russian).
Terentiev, AG 1994a On the infinite region in the boundary element method. Proc. Boundary
Element Method in Fluid Dynamics. Southampton, Comp. Mech. p103-109.
Terentiev, AG 1996 Impulse inducing of tsunami waves. Proc. 6th Scientific School «High Speed
Hydrodynamics» Cheboksary, p178-184 (in Russian).
Terentiev, AG 1997 Wave generation by impulse moving of the bottom. Proc. 4th Int. Conf.
«Moving Boundaries» Southampton, pp307-314.
Terentiev, AG 1998 Equilibrium hole- and point vorteces in a liquid flow. Izv. NANI,
Cheboksary. No. 5, p66-84 (in Russian).
Terentiev, AG 2000 Generalization of Green equation and application in numerical
hydrodynamics. Proc. Int. Sc. Conf. Modeling, computing and design under indeterminacy
conditions, Ufa St. Avia Tech. University, p82-89. (in Russian).
Terentiev, AG 2002 Problems in the theory of high-speed hydrodynamics. Proc. Int. Summer
Scient. School “High Speed Hydrodynamics”, Cheboksary, p11-27.
Terentiev, AG 2003 Riabouchinsky model in the cavitating theory. Izv. NANI, Cheboksary, 3,
p47-56.
Terentiev, AG 2004 Gravitational effects in cavitation. Proc. 2nd Int. Sum. Sc. School, Ch, 73-80.
Terentiev, AG 2005 Numerical modeling of cavitating flows. Proc. FAST’2005.
578
References
Terentiev, AG 2006 Numerical analysis of cavitating flows direct iterative manner. Proc.
CAV2006,Vageningen, 15p.
Terentiev, AG 2009 Numerical investigation of cavitating flow through the cascade of arbitrary
foil. Proc. CAV2009, Ann Arbor, Michigan, Paper No 151, 5p
Terentiev, AG, Afanasiev, KE 1987 Numerical methods in hydrodynamics. ChSU
Terentiev, AG, Chechnev, AV 1985 Numerical investigation of penetrating of the plate and disk
into compressible liquid. Trans. AN USSR, MZG, No. 2, p104-107 (in Russian). English
transl. in Fluid Dynamics
Terentiev, AG, Chechnev, AV 1989 Penetrating of the plate and disk of finite mass into
compressible liquid. Trans. AN USSR, MZG, # 2, p177-179 (in Russian). English transl. in
Fluid Dynamics
Terentiev, AG, Dimitrieva, NA 1998 Theoretical investigation of cavitating flows. Proc. Third
Int. Symp. on Cavitation, Grenoble. Vol. 1, p275-280.
Terentiev, AG et al. 1987 Simulation of unsteady free surface flow problems by the direct
boundary element method. (with KE Afanasiev, MM Afanasieva) Proc. IUTAM Sym.
Advanced Boundary Element Methods. San Antonio, TX, Springer-Verlag, p427-434.
Terentiev, AG et al., 1997 An application of analytic functions to axisymmetric flow problems.
(with VP Zhitnikov, NA Dimitrieva) Appl. Math. Modelling. Vol. 21, No. 2, p91-96.
Terentiev, AG et al. 2005 Riabouchinsky cavity modele in flow dynamics. (with NA Dimitrieva,
NA Pavlova) Proc. FAST’05.
Terentiev, AG, Kartuzova, TV 1995 Numerical investigation of a system wing foils by boundary
element methods. Proc. Actual Problems of Math. and Mech. ChSU, p108-116 (in Russian).
Terentiev, AG, Kartuzova, TV 1996 Numerical investigation of a flow over aerofoil near the
bottom. Izv. NANI, Cheboksary, v. 6, p94-104 (in Russian).
Terentiev, AG, Lasarev, VA 1969 Cavitating flow of bounded fluid around the plate. In
«Physical and technical problems» ChSU, p89-101.
Terentiev, AG, Makarov, VV 1999 Whole vorteces in a liquid closed to solid and free
boundaries. Izv. NANI, Cheboksary. No. 5, p66-84 (in Russian).
Terentiev, AG, Mikhailov, VM 1979 Non-stationary moving of the slender bodies in an ideal
fluid. Proc. Non-stationary moving of bodies in a fluid. Cheboksary, p111-148.
Terentiev, AG, Nikitina, GV 1983 Deformation of free boundary of the water by moving nearer a
rectangular body. In «Dynamics of continual fluid of free boundaries» ChSU, p117-126.
Terentiev, AG, Pavlova, NA 2006 Numerical analysis of cavitating flows by direct iterative
manner. Proc. CAV2006, Vageningen, 12p.
Terentiev, AG, Smirnova, TN 1998 Application of boundary element to calculate of a
panitratical airfoil. Izv. NANI ChR, 5, p85-95.
Terentiev, AG, Smirnova, TN 2000 Flow around penetrable foil near a boat. Izv. NANI,
Cheboksary. No 4, p61-69 (in Russian).
Terentiev, AG, Terentiev, AA 2002 Moving of a cylinder in a viscous fluid by small Reynolds
number. NANI, Ch., No 2, p44-62.
Terentiev, AG, Vishnevsky, VA 1970 Flow of a plate cascade with partial cavitating. Proc.
Problems of hydrodynamics and low-temperature plasma. CSU, p11-20. (in Russian)
Thill, C et al. 2005 Project energy-saving air-lubricated ships. Proc. 2nd Int’l Symp. on Seawater
Drag Reduction, Busan, Korea
579
References
Timman, R, Newman, JN 1962 The Coupled Damping Coefficients of Symmetric Ships. JSR,
5(4), p34-55.
Timman, RA 1958 A general linearized theory for cavitating hydrofoils in nonsteady flows.
Proc. Symp. Naval Hydrodynamics. Washington, DC, p559-582.
Todorashko, GT 1977 Solution of the nonlinear problem of the flow with the developed
cavitation behind the solid of revolution. Probl. of Applied Math. & Mech. CSU # 5, p195-
205 (in Russian).
Tomita, Y, Blake, JR, Robison, PB 998 Interaction of cavitation bubble with a curved rigid
boundary. Proc. Third Int. Symp. on Cavitation. V. 1, pp 51-56.
Tomita, Y, Sato K, Shima A 1994 Interaction of two laser-produced cavitation bubbles near
boundaries. Bubble Dyn. Inter. Phen. p 33-45.
Toyoda, M et al. 1999 Study on theoretical design method of trans-cavitating propeller. J. SNA
of Japan, Vol. 186, pp.41-49. (in Japanese)
Truscott, TT, Beal, DN, Techet, AH 2009 Shallow Angle Water Entry of Ballistic Projectiles.
Proc. 7th Int’l Symp. on Cavitation, Ann Arbor, MI.
Truscott, TT, Techet, AH 2009 A spin on cavity formation during water entry of hydrophobic
and hydrophilic spheres. Physics of Fluids, 21, 121703 2009.
Tuck, EO 2000 Numerical solution for unsteady two-dimensional free-surface flows. Proc. 11th
Biennial Comp. Tech. & Appl. Conf., World Scientific, p43-46.
Tulin, MP 1953 Steady two-dimensional cavity flows about slender bodies. DTMB, Rep.834.
Tulin, MP 1961 Supercavitating Flows. In Handbook of fluid dynamics. p12-24 thru 12-46
Tulin, MP 1964 The shape of cavities in supercavitating flows. Proc. Eleventh Int. Congress of
Applied Mech. Munich, Germany, pp.1146-1155.
Tulin, MP 1964a Supercavitating flows - small perturbation theory. JSR, Vol.7, #.3, p16 – 37.
Tulin, MP 1965 Supercavitating flows – small perturbation theory. Proc. IUTAM Sym. p403-439.
Also J. Ship. Res., v. 7, No. 3, 1964.
Tulin, MP 1998 Cavity Shapes in Three Dimensions. Proc. Fluid Dynamics Panel Workshop on
High Speed Body Motion in Water, Kiev, published as AGARD-R-827, Advisory Group for
Aerospace R&D, NATO, Neuilly-sur-Seine, France.
Tulin, MP 1999 On the Shape and Dimensions of Three-Dimensional Cavities in Supercavitating
Flows. Fluid Mechanics and its Applications, in: Fascination of Fluid Dynamics, Kluwer
Academic Publishers, Dordrecht, The Netherlands.
Tulin, MP 2000 Fifty years of supercavitating flow research in the U.S.: personal recollection.
Applied Hydromechanics, Kiev, Vol.2, 3, p100-107 (in Russian).
Tulin, MP 2000a On the theory and modeling of real cavity flows. Proc. Int. Symp. on
Cavitation. Osaka.
Tulin, MP 2001 The history and principles of operation of supercavitating propellers. Preprint:
Von Karman Inst. for Fluid Dynamics, RTO AVT/VKI Special Course: "Supercavitating
Flows", Belgium.
Tulin, MP 2003 On the theory and modeling of real cavity flows. Proc. CAV2003, Osaka, 14p.
Tulin, MP, Burkart, MP 1955 Linearized theory for flows about lifting foils at zero cavitation
number. DTMB Report C-638, Feb.
Tulin, MP, Hsu CC 1977 The theory of leading edge cavitation on lifting surface with thikness.
Tech. Paper, Hydronautics, MD, USA, 28p.
580
References
Tumashev, GG, Nuzhin, MT 1965 Inverse value problems and its applications. Kasan Univ.
333p. (in Russian), see also Aksentiev et al 1982 The Theory of Inverse Boundary Value
Problems for Analytical Functions and its Applications. J. Math. Sciences,v18, p479-515
Uhlman, JS 1978 A Partially Cavitated Hydrofoil of Finite Span. J. Fluids Engineering, Trans
ASME, Vol. 100, No. 3, p353-354
Uhlman, JS 1983 The Surface Singularity Method Applied to Partially Cavitating Hydrofoils.
PhD Thesis, MIT, Dept of Ocean Eng., Cambridge, MA
Uhlman, JS 1987 The Surface Singularity Method Applied to Partially Cavitating Hydrofoils.
JSR, Vol. 31, No. 2, p107-124
Uhlman, JS 1989 The Surface Singularity or Boundary Element Method Applied to
Supercavitating Hydrofoils. JSR, Vol. 33, No. 1, p16-20
Uhlman, JS 2006 A Note on the Development of a Nonlinear Axisymmetric Reentrant Jet
Cavitation Model. JSR,. Vol. 50, No. 3, p259-267.
Uhlman, JS, Fine, NE, Kring, DC 2001 Calculation of the Added Mass and Damping Forces on
Supercavitating Bodies. Proc. 4th Int’l Symp.CAV2001, Pasadena, CA.
Uhlman, JS, Varghese, AN, Kirschner, IN 1998 Boundary Element Modeling of Axisymmetric
Supercavitating Bodies. Proc. 1st Symp. on Marine Appl. of CFD, Hydroacoustic Technology
Center, McLean, VA.
Ukon, Y, Kudo, T, Fujisawa, J, Matsuda, N 1999 Experimental evaluation of trans-cavitating
propellers. J. SNA of Japan, Vol. 186, pp.51-58. (in Japan)
Ukon, Y, Kudo, T, Fujisawa, J, Sasaki, N 2004 Computational design of trans-cavitating
propellers and experimental evaluation of their performance, Proc. 25-th Symp. on Naval
Hydrodyn., St.John’s, p34-48.
Vaidyanathan, S, Senocak, I, Wu, J, Shyy, W 2003 Sensitivity Evaluation of a Transport-Based
Turbulent Cavitation Model. J. Fluids Eng., 125, p447-455.
Van Dyke. M 1964 Perturbation Methods in Fluid Mech. Academic Press, NY – London, 310p.
Van Dyke, M 1975 Perturbation Methods in Fluid Mechanics. The Parabolic Press, Stanford.
Van Manen, JD 1971 The effect of cavitation on the interaction between propeller and ship's hull.
Proc. IUTAM Symp. L., p305-328.
Van Wylen, GJ, Sonntag, RE 1973 Fundamentals of Classical Thermodynamics. 2nd Ed., John
Wiley & Sons, Inc., New York, NY
Vanden-Broeck, JM 1989 Bow flows in water of finite depth. Phys. Fluids A, 1(8):1328-1330
Varghese, AN, Uhlman, JS, Kirschner, IN 1997 Axisymmetric Slender-Body Analysis of
Supercavitating High-Speed Bodies in Subsonic Flow. Proc. 3rd Int’l Symp. on Performance
Enhancement for Marine Applications, Newport, RI.
Varghese, AN, Uhlman, JS, Kirschner, IN 2005 Numerical Analysis of High-Speed Bodies in
Partially Cavitating Axisymmetric Flow. J. Fluids Eng, 127, p41-54.
Vasiliev, VN 1977 Full cavitating flow of a plate cascade. Coll. Problems of Math. and Mech.
ChSU, 5, p3-14 (in Russian).
Vasin, A 1998 Supercavities in compressible fluid. Proc. 3rd Int. Sym. on Cavitation,.2, p3-8.
Vasin, AD 1996 Calculation of Axisymmetric Cavities Behind a Disk in Subsonic Flow of a
Compressible Fluid. Trans. RAN, ser. MZG, 2, p94-103, (in Russian).
Vasin, AD, Paryshev, EV 2001 Immersion of a Cylinder in a Fluid through a Cylindrical Free
Surface. Fluid Dynamics, 36, p169-177.
581
References
Vaz, 2005 Modelling of Sheet Cavitation on Hydrofoils and Marine Propellers using Boundary
Element Methods. PhD Thesis, Univ. Technica de Lisboa, Instituto Superior Technico.
Verhagen, JHG 1967 The impact of a flat plate on a water surface. JSR, v11, p211-223.
Villat, H 1911 Sur la resistance des fluids. Ann.Sci. Ecole Norm. super. (3 t.28.
Villat, H 1915 Sur la ecoulement des fluids pesants. Ann. Sci. Ecole Norm. super. V. 32.
Vinje, T, Brevig, P 1981 Numerical simulation of breaking waves. Adv. Water Res. v.4. p77-82.
Vishnevsky, VA, Kotlyar, LM, Terentiev, AG 1974 Influence of the gravity on cavitating flow
over obstacles. Problems of App. Math. and Mech”, CSU, p9-24.
Vishnevsky, VA, Galanin, AV 1975 On influence boundaries of a fluid on cavitating flow of a
hydrofoil. Problems of App. Math. and Mech. CSU. V.4, p14-30.
Vlasenko, YD 2002 Experimental Investigations of Supercavitating Flows at Subsonic and
Transonic Velocities. Proc. 2002 HSH2002, NANI, Russia.
Vlasov, OE 1957 Fundamental theory of explosion action. M. Military Eng. Academy, 408p
Voinov & Petrov 1976 Moving of bubbles in fluid. Summary of Sci. Tech. Mech, of Liquid and
Gas, v.10, p86-147.
Voinov, OV, Voinov, VV 1975 Numerical method of calculating non-stationary motions of an
ideal incompressible fluid with free surfaces. Sov. Phys. Dokl., 20, 179--180.
Voit, SS 1967 Review of works on the theory of tsunami waves, fulfilled in USSR. Physics of
atmosphere and ocean, v.3, No 11, p1158-1165
Voitkunski, YI (Ed) 1985 Ship Theory Handbook. Vol. 1-3, Sudostroenie Pubs, L., 540 p
Voronin, VV 1983 Investigation of 2D gas cavity near free boundary of gravity fluid. Sci. Notes
of CAHI, v.14, #2, p37-42. (in Russian)
Vorus, W 1986 Ambient supercavitaties of slender bodies of revolution. JSR. 3, #3, p215-219.
Vorus, WS 1991 A Theoretical Study of the Use of Supercavitation/Ventilation for Underwater
Body Drag Reduction. VAI Tech. Rrt, Vorus & Assoc., Inc., Gregory, MI.
Wade, RB 1967 Linearized theory of partially cavitating cascade of flat plate hydrofoils. Appl.
Sci. Res., v. 17, No 3.
Wade, RB, Acosta, A 1966 Experimental Observations on the Flow Past a Plano-Convex
Hydrofoil. J. of Basic Engineering.
Wagner, H 1932 Ueber Stoss und Gleitforgange an der Oberflache von Flussigkeiten. ZAMM, v.
17, No. 3. p194-215.
Wang, DP 1977 Water entry and exit of fully ventilated foil. J. Ship Res., 21, 1.
Wang, DP 1979 Oblique water entry and exit of a fully ventilated foil. J. Ship Res., 23, 1.
Wang, QX 1998 The Evolution of a Gas Bubble Near an Inclined Wall. Theor. Comput. Fluid
Dynamics. 12, p29-51.
Weinig, F 1932 Die Ausdehnung des Kavitationsgebietes. In «Hydromechanische Probleme des
Schiffsantriebs». Hamburg.
Weinig, F 1935 Die Stromung um die Schaufeln von Turbomaschinen. J. Springer.
Weissinger, J 1956 Zur Aerodynamik des Ringfkuegels in inkompressibler Stroemung. Z. f.
Flugwissenschaft, 4, H. ¾, pp.141-150.
Whittaker, ET, Watson, GN 1927 A course of modern analysis. Cambridge, Univ. Press. 3
Widnal, SE 1966 Limiting gravity waves in water of finite depth. Phil. Trans. Roy. Soc. London,
A302, N1466, P. 139-188.
Wirz, HJ (Ed.) 1979 Numerical methods in fluid dynamics. Hemispher,Pub, Washington-London
582
References
Woods, LC 1955 Usteady plane flow past curved obstacles with infinite wakes. Proc. Roy. Soc.,
S.A. v.229, No 1177.
Woods. LC 1961 The theory of subsonic plane flow. Cambridge Univ. Press.
Wosnik, M, Arndt, REA 2009 Control Experiments with a Semi-Axisymmetric Supercavity and
a Supercavity-Piercing Fin. Proc. CAV2009, Ann Arbor, MI
Wosnik, M, Qin, Q, Arndt, REA 2006 Identification of Large Scale Structures in the Wake of
Cavitating Hydrofoils Using LES and Time-Resolved PIV. Proc. 26th Symp. on Naval
Hydrodyn., Rome
Wosnik, M, Schauer T, Arndt, REA 2003 Experimental Study of a Ventilated Vehicle. Proc.
CAV2003, Osaka.
Wrobel, LC 1987 Potential problems. In Top Boundary Elements Res. Com. Mech. Pub. UK, v. 3,
p235-264.
Wrobel, LC 1991 Boundary element simulation of axisymmetric cavity flow problems. Proc.
Conf. “Computational Modelling of Free and Moving Boundary Problems”. Computational
Mech. Publications, Southampton, Boston, p161-171.
Wu, TY 1956 A note on the linear and nonlinear theories for fully cavitated hydrofoils. Bull.
Amer. Phys. soc., 1, No. 7.
Wu, TY 1957 A linearized theory for nonsteady cavity flows. Eng. Div. Rep. #85-6, CalTecn.
Wu, TY 1958 Unsteady supercavitating flows. Proc. Symp. Naval Hydrodynamics. Washington,
DC, p293-315.
Wu, TY 1962 A wake model for free-streemline flow theory. JFM, v. 13, No. 2.
Yahnke, B, Emde, F 1952 Tafeln Hohere Funktionen. Springer Verlag.
Yakimov, YL 1968 On axisymmetric flow around body of revolution by small caitation number.
PMM, M., Vol. XXXII, No 3
Yakimov, YL 1973 Influence of the atmosphere by falling of bodies on the water. Izv. AN USSR,
MZG. No. 5 (in Russian) English transl. in Fluid Dynamics
Yakimov, YL 1983 Thin cavity in a compressible fluid. In «Problems of contemporary Mech.».
Moscow Univ., M. Pt. 1, p63-73 (in Russian).
Yakimov, YL 1983a On Energy integral for motion with small cavitation numbers and form
cavities to the limit. MZG, #3, p67-70. (in Russian) English transl. in Fluid Dynamics
Yamaguchi, H, Kato, H 1983 Non-linear theory for partially cavitating hydrofoils. J. Soc. Nav.
Arch. Jap., 152, p117-124.
Yanke, E, Emde, E, Losch, E 1968 Special Functions. Nauka. M.
Yas’ko, NN 1993 Calculation of the plane and axisymmetric Riabouchinsky cavitational flows
by the direct boundary element method. Comp. Modelling of Free and Moving Boundary
Problems II. Comp. Mech. Publications, Southampton, Boston, p145-151.
Yim, B 1964 On a fully cavitating two-dimensional flat plat hydrofoil with non-zero cavitation
number near a free surface. Hydronautics Techn. Rept., 463-4.
Yim, B 1971 Investigation of gravity and ventilation effects during the water entry of thin foils.
Proc. IUTAM Sym. “Unsteady Flow of Water at High Speed,” L..
Yim, B 1973 Investigation of gravity and ventilation effects during the water entry of thin foils.
Proc. IUTAM Sym. “Unsteady Flow of Water at High Speed,” L.., p471-489.
Yim, B, Dobay, G, Larimer, G, Peck, J 1986 Supercavitating propellers-design theory and
experimental evaluation. Trans. Int’l Symp. on Cavitation, Sendai, Japan, p239-245.
583
References
Young, YL 2002 Numerical modeling of supercavitating and surface-piercing propellers. Doct.
dissertation, Univ. of Texas Austin, TX.
Young, Y, Kinnas, S 2000 Prediction of Unsteady Performance of Surface-Piercing Propellers.
Proc. SNAME P/S 2003 Symp., Virginia Beach, VA, p7/1-9.
Young, Y, Kinnas, S 2003 Numerical Analysis of Surface Piercing-Propellers. Proc. SNAME P/S
2003 Symp., Virginia Beach, VA, p4/1-20.
Yuasa, H 1984 Application of the vortex lattice method to the threedimentional theory of a
cavitating propeller. J SNA of Japan, Vol.156, p69-81.
Zaitsev, NA, Maskalik, AI 1967 Fatherland hydrofoils. 2nd Ed, L. Sudostroenie, 364p.
Zhitnikov, VP 1993 Solution of axisymmetric problem of the flow over the body in tube with the
help of the methods of the theory of functions of complex variables. Problems of High Speed
Hydrodynamics. ChSU, p107-117 (in Russian).
Zhitnikov, VP, Sherykhalina, NM, Sherykhalin, OI 2000 Postcritical regimes in the nonlinear
problem of vortex motion under the free surface of a weightable fluid. J. Appl. Mech. & Tech.
Physics, v41, #1, 0021-8944/00/4101-0062. Kluwer Academic/ Plenum Publishers, p62-68.
Zhitnikov, VP, Terent'ev, AG 1982 Jet flow of an ideal fluid past a flexible shell. Izv. AN USSR.
MZG, No. 6. 43. English transl. in Fluid Dynamics
Zhitnikov, VP, Terent'ev, AG 1984 Continuous flow past a flexible shell. Izv. AN USSR. MZG, #
5, p15-20. English transl. in Fluid Dynamics
Zhitnikov, VP, Terent'ev, AG 1988 Non-linear problem of cavitating flow around a flat plate
with capillary tension on the free boundary. In «Hydrodynamics of bounded streams» ChSU,
p51-58 (in Russian).
Zhitnikov, VP, Terentiev, AG 1993 Axisymmetric flow of ideal fluid past a gas bubble. Izv. AN
USSR. MZG, No. 5, p98-103 (in Russian) English transl. in Fluid Dynamics
Zhuravlev, YF 1973 Perturbations Methods in 3D Stream Flows. Proc. CAHI, #1532, 23 p.
Zienkiewicz, OC & Cheung, YK 1967 The finite element method in structural and continuum
mech. McGraw-Hill, London.
Zigangareieva, LM, Kiselev, OM 1998 Subsonic Flow of a Plate in a Compressible Fluid at
Small Cavitation Numbers. Trans. AN USSR, MZG, 4, p94-104, English transl. in Fluid Dyn.
Zuykov, YP, Shepelenko, VN 1967 On one problem of cavitation. Izv. Sib. Branch AN, ser.
Tech. Nauk, 3, No 13 (in Russian).
584
Subject Index
Subject Index
NOTE: The subject index refers to page numbers. Dash line stands for all page numbers in-
between; in this case the first (and second) digits of the last page number can be omitted if it is
the same as that of first number. Tilda sign ~ in front of or after a sub-entry stands for the
wording of the preceding entry, one ~ per entry.
acoustic propagation mode, ~ mixture, 446

added mass 207, 280, 289, 303, 304, 312, ~ mode, 445, 450, 457-9, 468, 472
319, 419, 423 ~ moving in the channel, 145-6
advection 14, 17, 280 ~ number density, 445-7
adverse pressure gradient, 422 ~ pressure, 308-10, 373
air ~ propagation, 453, 455, 473
~ cavity craft, 426 ~ radius, 307-15, 373,376, 450, 458-72
~ content, 428, 442, 475 ~ resonance frequency, 463, 472
~ effects 259, 262 ~ shape 143, 148
approximation spherical ~, 304, 312-5, 374
far-field ~ 454 ~ trail, 474
low-frequency ~ 463; 469, 472 ~ wave, 457-8, 460
Sommerfeld’s ~s 522, 526 canister
asymptotic matching, 197, 434 ~ -nozzle-bubble system, 313
axisymmetric ~ volume, 306
~ bodies, 11, 145, 198, 247, 256, 279, isentropic ~ blow-down, 306
330-1, 342, 412-3, 419, 422, 439-41, 540- pressurized gas ~, 280
2 cavitating
~ cavities, 14, 16, 202, 203, 298, 337, 412 ~ plate beneath free surface and in
expansion of an ~ cavity, 304 unbounded domain, 101
~ flexible bubble, 154 ~ ~ between the free surface and the
body bottom, 99
~ of displacement, 435 ~ ~ in a flow out of a channel, 95
ITTC ~, 432, 441 ~ propellers, 475, 498
slender axisymmetric ~, 429, 439 ~ sphere in an axisymmetric water tunnel,
slender ~, 203, 206-9, 282, 287-9 337
boundary layer ~ finite-span wings, 254
laminar ~, 128, 178, 431, 437, 442 fully ~, 37, 45, 56, 95, 113, 284, 347
turbulent ~, 435, 442-3 ~ ~ flat plate, 37
wall ~, 434, 437 ~ ~ hydrofoil beneath a free surface, 95
boundary problem highly ~ propellers, 480, 485
mixed ~, 231, 528 partially ~ cascade of plates, 110
~ ~ solution for a half-plane, 529 cavitating flow (around or in)
bow shock, 23, 459 ~ circular cylinder, 127
bubble ~ disk, 337
~ boundary, 146, 372-3, 375 ~ flat plate, 30, 37
~ ~ self-intersect, 155 ~ ~ in a channel, 98
cavitation ~, 11-2, 429, 495 ~ ~ near free surface, 91
cloud of ~s, 13 ~ foil with an interceptor, 348
~ collapces, 28 ~ slender body, 180
~ dynamics, 280, 303, 307-9, 376, 447 ~ system of arbitrary bodies, 346
585
Subject Index
~ wedge ~ ~ of revolution, 286, 330-2, 338, 353,
~ ~ in a longitudinal gravity field, 132 439-40, 542
~ ~ inside another wedge, 276 ~ collapse 372, 387
~ in curvilinear arc, 180 ~ closure, 13-6, 25-6, 41-3, 265, 402-3,
~ in a channel with 416, 433-6, 474, 491-3
~ ~ flat sides, 103 ~ ~ at a depth, 265
~ ~ variable cross section, 275 ~ cloud, 118, 269, 374
~ cone, 337 ~ detachment point, 26, 286, 394, 429,
~ ring wing, 347 443
~ spherical segment, 142 ~ flapping on a free surface, 387
~ of compressible fluid, 277-8 ~ flow past a circular arc, 127
~ through a flat plate cascade, 107 ~ in a bounded flow domain, 173
~ sphere, 142 ~ in a channel, 76, 174
~ through a cascade of foils, 500 ~ in a gravity field, 267, 290
~ using small perturbation approaches, ~ in a jet, 173
185 ~ in longitudinal gravity field, 175
~ with capillary tension on cavity shapes, ~ transverse gravity field, 176
146 ~ inception, 26, 294, 426, 430
partially ~, 45, 110, 113, 139, 414, 438 ~ initiation by a pressurized gas canister,
cavitating model with 280
~ artificial solid boundary at the cavity ~ on a straight wall in rotational fluid
end, 151 motion, 178
~ single spirals, 26, 30, 91, 108, 132, 136, partial ~ on Joukowski hydrofoil, 184
337, 347 ~ platform, 490
~ rotation fluid motion 56 single ~ in a rotational fluid flow, 179
cavitation vaporous ~ 24, 257, 280, 427
back-bubble ~,429 ventilated ~,
~ barrier, 477 ~ vortices 114, 118
~ basin 476 ~ with loss of gas, 299
bubble ~, 11-2, 495 ~ with negative cavitation number, 79
~ collapse, centrifugal pumps, 499
~ tunnel 106, 475, 593 characteristic length, 16-20, 24, 431, 438-
~ inception, 41, 463-4, 471
~ index 13 circulation 29, 33
cloud ~,12 ~ around a circular contour 364
full ~ 12, 49, 183, 347 ~ ~ Joukowski airfoil 364
ideal ~, 430 Clauser’s form, 438
natural ~ 11, 24, 492 cloud of bubbles, 13
partial ~ , 50, 61, 140, 181-4, 321, 342, coalescence, 447
389, 412-4, 475, 491 coating
sheet ~, 12, 432, 442 hydrophilic/hydrophobic ~, 426-7
tip ~, 12 coefficient
unsteady ~, 12 advance ~, 475-9, 481-5, 488-90, 495
vaporous ~ 11, 19, 428, 441 buoyancy ~, 373, 377-8
ventilated ~ 11, 24, 66, 428 convective heat transfer ~, 18
cavitator, 15-20, 203-11, 280-9, 293-5, 312- drag ~, 36-7, 43-4, 76-8, 105-7, 127-8,
21, 409-12, 427-30, 449, 553 144, 171-5, 199, 270-2, 381-2, 410-1,
cavity (ies) 469-71
~ as an ellipsoid, 198-202, 205-7, 281
586
Subject Index
lift ~, 43-5, 48, 86-8, 95, 181-4, 217, 256, derivative of complex potential, 54, 89, 91,
268-9, 344-5, 348-52, 400, 497 99, 119, 193, 215, 276, 510
~ of normal force 103-5, 109, 140, 144, design
~ of resultant force, 51, 59, 94, 100, 185 ~ of a hydrofoil with given pressure
pressure ~, 197-202, 250-1, 392-3, 407, distribution on the boundary, 166
434, 461, 466-7 ~ of supercavitating hydrofoil, 189
surface tension ~, 21 dessinence, 426
thrust ~, 488-9, 496 detachment point, 26, 70, 126, 144, 179,
torque ~, 43-5, 192, 229, 268-9 201, 229, 394, 406
ventilation ~ 13-7, 321, 426 ~ on foil, 31
virtual mass ~, 244 dimensionless parameters, 14,16, 115, 135,
compressibility of fluid, 28, 66, 167 146-8, 154, 295, 371, 464, 470
concept of cavitation within viscous layers dispersion, 449-50, 465, 472
= CCVL, 10, 430-5, 437-42 displacement
condition ~ amplitude, 420
Brillouin ~, 30, 153, 156, 343-4, 354-7 ~ discontinuity, 11
Brillouin-Villat ~, 398, 428-30, 433, 440 ~ rate, 369
closed-cavity ~, 39 ~ thickness, 58, 435
Courant’s ~, 380 element(s)
Courant-Friedrichs-Lewy ~ 385 arc ~, 31
dynamic ~, 29, 103, 144, 213, 240, 498 constant ~, 322-3, 543
dynamic boundary ~, 389-90, 394-6 high-order ~ 322
existence ~, 437, 491 linear ~, 322, 544
Joukowski – Kutta ~, 11, 324-5, 334, 342, quadrilateral ~, 376
347, 350, 353, 363-4 triangular ~, 375-6
kinematic (al) ~ 29, 83, 84,114, 146, 161, entry
240, 360, 490 ~ into gravity fluid, 248
Kutta- Joukowski ~, 107, 402, 499, 500 ~ of a thin foil, 245
linearized dynamic ~, 490 ~ ~ a wedge with straight sides, 233
Tulin’s ~, 176 ~ ~ thin wedge into water, 265
conservation of mass, 445-6, 550, 552 inclined ~ of a plate with ventilated
constant cavity, 235, 268
Euler ~ 69, 229 non-symmetric ~ into water, 267
polytropic ~, 305, 463 partial ~ ofa plate with ventilated cavity,
consumption, 63, 347, 512 268
converging nozzle, 304 perpendicular ~ of a wedge, 234-5
cubic splines, 338, 356 perpendicular ~ of inclined plate, 268
cylindrical coordinates, 162, 204, 330, 498, perpendicular ~ of thin foils, 230
535, 541 vertical ~ of a wedge, 237
dam breaking, 386-7 vertical ~ of a rectangular plate, 255
Darcy’s law, 511 water ~ of a blunt body, 246
deformation equation
3D ~ of bubble surface, 376 bi-harmonic ~, 534
~ of a single bubble near free boundary, blow-down ~s, 280
373 continuity ~, 159, 382
~ of gas bubbles, 372 Euler ~s, 161, 489
density, 13, 20, 135, 159, 198, 262, 283, Green’s ~, 323, 339, 349, 533
304, 360, 430, 445, 505, 549 ~ ~ for axisymmetric functions, 535
587
Subject Index
Laplace ~, 28, 159, 162, 285-6, 323, 498, ~ ~ around the wedge in a jet, 76
523 ~ ~ of two single spirals, 76
momentum ~, 159, 438, 446 laminar ~, 18, 26
Navier-Stokes ~s, 382 near-critical ~, 368
Poisson ~, 523 non-choked ~, 304, 310
Rayleigh ~, 44, 51 non-steady cavitating ~, 221
Stokes ~s, 382, 517 outer ~,434-7, 441-2, 446
Tulin’s asymptotic ~, 44 ~ past flexible tubes, 146
equilibrium cavity vortices ~ through a cascade of cavities, 171-3
~ in a stream, 114 ~ through plate-cascade, 107-14, 499-507
~ in a jet stream against a wall, 117 perturbation ~ between two bodies, 350
equipotential lines, 41 perturbed ~ around a foil, 273, 349
erosion, 11, 118, 372, 475-9, 485, 488 potential ~, 28, 158-9, 293, 323, 433-4,
evanescent, 451-2 460, 472, 488, 511, 542
evaporation single-phase ~, 20, 446, 450, 461, 471-3
~ from cavity boundary, 259, 269-72 subcritical ~ 368
~ parameter, 272 supercritical ~ 368
excavation, 158 turbulent ~ 18, 20, 431, 442
expansion of a spherical bubble, 307 two-phase ~, 450, 467
explosion in soil, 158 unsteady ~ with free boundaries, 235, 365
flow vortex-type ~ of incompressible fluids,
~ between two concentric circles, 351 160
bubbly ~,22, 444 fluid
circular ~ from a point vortex 94 compressible ~, 158, 277, 278, 378, 451,
compressible ~, 19, 447-9, 462-3, 467-8, 473
472 ~ density, 28, 38, 133, 482, 526
continuous ~ around a given profile, 166 ~ filtration, 514
flat ~ of heterogeneous fluid, 161 ~ viscousity, 28
~ along a rectangular step-bottom, 33 heterogeneous ~, 161-2
~ around homogeneous ~, 162, 164
~ ~ a thin wing foil, 167 ideal ~, 25, 190, 430, 432, 549
~ ~ ellipsoid of revolution in a pipe, incompressible ~ 158, 160, 243, 252, 278,
353 354, 367, 385, 517, 542
~ ~ foil between two large bodies, 351 Newtonian viscous ~, 382
~ ~ penetrable circular cylinder, 351 real ~, 24-6, 118, 155, 426-8, 438, 473
~ ~ plate with a spoiler, 195 weightless ~, 136, 161, 265, 492
~ ~ slender body of revolution, 201 foil
~ ~ small curved arc, 187 ~ between a free surface and a wall, 352
ideal cavitating ~, 26 finite-span ~, 252
~ in a channel with converging straight rectangular ~, 254
boundaries, 350 force
~ in a gravity field, 131-2, 161-2, 290, Archimedes ~, 134, 268
511 capillary ~, 145
incompressible ~, 279, 282 dynamic pressure ~, 13-4, 16, 19
inviscid ~, 28, 419, 432, 437, 488 gravitational ~, 14, 16, 19-20
irrotational ~, 419, 432, 445-7 inertial ~, 14, 16, 19-21, 156, 263, 517
Kirchhoff ~ 43, 71, 72, 105, 109, 124, resultant ~, 33, 42, 48, 87, 105, 261, 332,
152, 191-4, 271 518
~ ~ along a flexible shell, 152
588
Subject Index
static ~, 134 Gauss (Gaussian) quadrature, 438, 542
formula Geurst paradox, 491
Hammer’s cubature ~, 376, 548 Geurst’s symbols, 181
Keldysh-Sedov ~, 528 gravity (gravitational)
Monaghan ~, 384 ~ acceleration, 17, 24, 131, 138-9, 158,
Reichardt ~ 341 176, 248
Sedov’s ~, 215, 246 ~ effects, 14, 132, 141, 156, 247, 267,
Fourier series, 122, 242, 289, 354 284, 293
free boundary (-ies), ~ field, 53, 131-2, 136, 175-6, 248
~ deformations in a gravity field, 358 longitudinal ~ ~, 79, 131-2, 135-6, 175-
~ surface, 11, 21, 56, 83, 91, 99, 101, 6, 248, 277
145, 158, 239, 352, 366 normal ~ ~, 162
~ effect, 276 transverse ~ ~, 53, 132, 136, 176, 321,
~ vortices, 212, 215 360, 411
frequency, vertical, 299
bubble resonance ~, 463, 472 ~ waves, 455
cavity oscillation ~, 297 Green’s identities, 322, 332, 337, 365, 531,
peak bubble response ~, 463 538
function ground explosions, 158
Bessel ~s, 453, 464 Harlow’s approach, 379
beta- ~, 233 high-speed prosseses, 28
bi-harmonic ~, 517, 534 hydro-acoustic noise, 477
elliptic ~, 85, 119, 261, 355 hydrofoil in finite depth stream, 363
gamma- ~, 223 hydroplanes, 158, 164
generalized analytical ~ of complex impact
variables, 142 ~ hydrodynamics, 505
generalized p-analytic ~, 539-42 ~ of floating thin and blunt bodies, 244
Green’s ~, 247, 252, 323, 375, 415, 459, inertial effects,
464, 499 influence of free surface 56, 91
Hankel ~, 221, 228, 452-4 inner solution,
Heaviside ~, 248, 329 intake diffusers 331
hyper geometric ~, 234 integral
integral exponential ~, 130 Bernoulli ~ 28, 42, 105, 361
kernel ~, 384 Blasius-Chaplygin ~ 86
logarithmic ~ of complex velocity, 29, 39, Cauchy ~ 63, 177, 358, 359
66, 139, 195, 274 Cauchy type singular ~ 81, 177, 546
meromorphic ~, 66, 180, 323, 537 full elliptic ~ 163, 217, 535
polyharmonic ~, 505, 523, 535 Schwarz ~ 151, 261
pressure ~, 212 Volterra’s type ~ 238
stream ~, 32, 116, 142, 160, 198, 213, integral equation
235, 259, 324, 363, 517, 550 Abel’s ~ 238
Theodorsen ~, 221, 227 boundary ~ 337, 555
theta ~s, 85, 119, 544 ~ as Volterra form of first kind 214
time-depended Green ~, 247 modified ~ 500, 502
Wagner ~, 235 momentum ~
gas singular ~
~ content, 22 integral identities for singular functions 536
~ turbine, 481, 484 interaction
weightless ~, 159
589
Subject Index
fluid-fluid ~ 386 reproducing kernel particle ~, 384
~ of two bubbles 372, 374 Runge-Kutta ~, 386
high-speed impact ~s of solid bodies ~ smoothed particle hydrodynamics
158 (SPH), 382-7
interfacial force density, spline ~, 95
isothermal, stroboscope ~, 475
jet-induced ejection of gas from cavity 300 virtual particle ~, 385
Joukowski hydrofoil near a circular cylinder Michel criterion, 440
349 model
Kármán-Poulhausen differential form, asymptotic ~s of axisymmetric
Kirchhoff’s cavity supercavitating flows, 203
~ in a channel, 107 CCVL ~ - see concept of cavitation within
~ in boundless domain, 238, 241 viscous layers
inclined plane with ~ near a wall, 107 Chaplygin – Kolscher ~, 129
Kronecker symbols, 356, 383 closed ~, 54, 491
Lagrangian double-spiral ~, 56, 61
~ particles, 383 double rectilinear ~, 132
~ variables, 379 Efros – Gilbarg ~, 56
Laplace’s law, 146 fluidic ~ (FM), 505
Laurent series, 143 fully-coupled ~, 316
Legendre polynomial, 202 generalized linear ~, 488
lift 11, 86, 90, 140, 218, 326, 335, 430 Helmholtz ~, 357
line of speed discontinuities, 212 Joukowski ~, 58, 103, 107
liquid with limited high viscosity, 518 Kirchhoff ~, 125, 129, 221, 271, 381
Logvinovich’ principle , 281 ~ of radial flow expansion, 206
mass matrices, 11, 56, 376, 441, 475 ~ with single spirals, 26, 91, 136
material derivative, 446, 448, 452 open cavity ~, 30
meniscus, 437 partial cavitating ~ with close cavity, 54
method(s) re-entrant ~, 52, 55
analytical extension ~ 239 re-entrant straight-lines ~, 53
approximate (approximation) ~, 259, 280 Riabouchinsky ~, 51, 132, 199, 202, 337,
asymptotic ~ in cavitating flow 158 347, 412
boundary elements ~ (BEM), 322, 437, second Tulin's ~, 56
527 solid-liquid ~ (SLM), 505
complex boundary elements ~, Tulin’s double-spiral ~, 61
collocation ~, 125, 155, 519 turbulence ~,
direct iterative ~, 127, 131, 137 Wu – Fabula ~, 59, 60
~expansion in series, 122, 141, 517 κ-ε turbulence ~, 431
finite particles ~, 378 motion
Levi-Civita ~, 354 longitudinal ~ of a symmetrical
matched asymptotic expansions ~ deformable body, 215
(MAEM), 190, 195, 197, 205, 282-287 ~ of curvilinear deformable arc, 215
numerical-analytical ~ for arbitrary- ~of fully immersed wedge, 238
shaped foils 354 ~ of partially submerged slender body,
~ of conformal mapping, 62 230
~ stationary phase, 454, 459-60, 472 ~ of oscillating hydrofoil, 219
~ of weighted residual functions, 322 ~ of plate with arbitrary speed, 218
~ perturbation velocity, 217, 462, 468-9 vertical ~ of a circular cylinder in a fluid
with free surface, 366
590
Subject Index
~ of plate near a wall, 260 Silberman-Song-Logvinovich ~, 19-20,
~ of sphere in a fluid with free surface, 426
366 Strouhal ~, 20, 220
movement turbulent Prandtl ~, 18
horizontal ~ of half cylinder, 367 vaporous cavitation ~, 19, 441
~ of body perpendicular to free boundary, Weber ~, 21-2, 147-8, 152-5, 426, 430-2,
240 436-7, 440-2
~ of thin body under a free surface above numerical conformal mapping 332, 500
a solid bottom, 212 outer solution, 205, 285, 461-2
~ of foil with Kirchhoff cavity, 241 paradoxes of unsteady cavity, 221
~ of plate with acceleration, 224 penetrable hydrofoils 332
non-stationary ~ of thin bodies, 212 ~ near a bottom, 334
pulse ~ of plate with Kirchhoff cavity, 221 penetration of a body into a fluid, 230
unsteady ~ of plate with Kirchhoff cavity, permeability, 511
221 Petrov’s presentation, 202
unsteady ~ of a wing of finite span, 253 plate
multi-sheet Riemann plane, 338 artificial ~ 26
Münzer-Reichardt profile, 304, 313 dropping a ~ on a flat bottom, 259
non-cavitating flow systems, 433 inclined ~ with Kirchhof’s cavity in
number compressible fluid, 278
Atwood ~, 20 partially-cavitated, ~ 160
axial cavitation ~, 475, 495 ~ in a free stream, 100
Biot ~, 18 ~ near a bottom in a stream flowing out of
Bond or Eötvös ~, 21 a channel, 98
Buyvol ~, 20, 426 ~ perpendicular to the free stream 40
Campbell-Hilbert ~, 19, 426 rectangular ~, 255-6
capillary ~, 21 supercavitating flat ~ in a transverse
Cauchy ~, 19 gravity field, 136
cavitation ~, 28 point disturbance, 444-5, 459
cavitation inception ~, 441-2 point of zero friction, 438
cavities Froude ~, 14, 290, 321, 411 Pologiy integral relationship, 143
cavities Richardson ~, 17 polynomial splines, 383
Courant ~, 381 porous medium, 511
Eckert ~, 17 potential
Euler ~, 16, 19 complex ~, 29, 62, 129, 177, 195, 332,
Froude ~, 14, 20, 132, 138, 176, 247, 360, 511, 538
266, 290, 321, 371, 411, 476, 487 Lennard-Jones ~, 385
Galileo ~, 21 ~ of acceleration, 212
incipient cavitation ~, 430, 441-2 perturbation ~, 419-22, 436, 448, 462
Laplace (Suratman) ~, 22 velocity ~, 131, 142, 158, 244, 288-9,
local cavitation ~, 475, 487, 494 293, 333, 419, 433, 446-7, 461-2, 466-7,
Mach ~, 19, 22-3, 159-60, 277—9, 305, 549-50
380, 426, 443-5, 450, 456-60, 466-9, 481 pressure
mixture ~ ~, 466-72 critical bubble ~, 308
Prandtl ~, 18 hydrostatic ~, 268, 274, 373, 475
Reynolds ~, 14, 18, 20, 24, 430-43, 517, vapor ~, 13, 16, 24, 373, 441, 475
526 principle
Richardson ~, 17 ~ of cavity expansion independence, 209
Roshko ~, 20 ~ of cavitation drag minimum, 497
591
Subject Index
problem flow ~ 26, 118
droplet ~, 385 laminar ~ 26, 431
Helmholtz ~, 29 ~ point, 26, 48, , 223, 344, 434-5
impact ~s, 158, 244, 506 turbulent ~ 128
inner ~, 286, 435, 461 viscous ~, 433
inverse value ~s, 338 simple layer, 484, 488
Kelvin’s ~, 444, 472 single spirals, 26, 30, 35, 42, 53-4, 70, 76,
Kirchhoff ~ 43, 187 90, 108, 132, 142, 337, 347
mixed value ~, 34, 83, 133, 179, 183, 230, singularity
235, 272, 535 ~ at the trailing point, 30
Neumann ~, 259 Cauchy ~, 75, 334, 359
nonlinear ~, 56, 95, 185, 191, 235, 343, integrable ~, 30, 168, 528
362, 403, 437 slenderness ratio, 339, 469
plane ~ of cavitating flow, 29 small disturbances of a vortex flow, 160
plate to water entry ~, 380 small perturbed potential flows, 158
Schwarz ~, 231, 259, 370, 528, 530 smooth attachment of cavities, 344
transient ~s of heat transfer, 18 soft cylindrical shell 148
underground sprinkling ~, 511 source
variational ~, 491 point ~, 390, 400-1, 429, 453-5, 459-60,
Wagner’s ~, 215 464-5, 472-3
propeller, ~ distribution, 198, 200, 285, 374, 389,
~ with a narrow blade 498 422, 461, 492
screw ~ 331, 475, 485, 496 ~ strength, 389, 393-4, 400-1, 418, 436,
surface-piercing ~ 11 550
Rayleigh-Taylor instabilities, 20 speed
re-circulating water tunnel, 440 inlet ~, 28, 82, 109, 200, 324, 360, 501,
Rectangular Step-Bottom, 33 503
re-entrant jet, 14, 43, 52, 108, 157, 258, mixture sound ~, 448-50, 455-7, 467-9,
321, 411, 427, 436 473
regime sonic ~, 28, 277-8, 444
drag hump ~, 479 sound ~, 19, 22, 159, 380, 384-6, 428-9,
intermediate toroidal shedding ~, 16 444-50, 455-7, 466-9, 471-3
toroidal shedding ~, 14, 16, 321 supersonic ~, 22, 459
twin vortex ~, 14-6, 19-21, 321 wave ~, 444, 455, 460, 472
Riabouchinsky wall, 26, 43, 397-8, 409-15 spoiler effect, 485
Riegels factor, 170 stagnation point, 37, 51, 85, 116, 148, 185,
ring stabilizers, 331 334, 348, 421, 503
rotation of an eccentric circular cylinder in a streamlines, 30, 41, 56, 116, 121, 257, 364,
cylinder, 526 431, 522
scheme ~ near cavity end, 30, 41
cavity closure ~, 430 submerged gun, 22-3
leap-frog ~, 385 supercavitating
Schwartz reflection, 192 ~ foils of high performance, 189
Schwarz principle of analytical extension, ~ high-speed undersea vehicles, 11, 426
359 ~ projectile, 13, 19, 22, 444, 455-8, 463
separation ~ propellers, 475, 485, 496
~-reattachment zone, 433-7 ~ vehicles, 11, 319
boundary layer ~,26, 128, 433-7 supercavitation
592
Subject Index
axisymmetric ~, 279, 407 two cavities on a foil, 346
~ experiments, 459 uniform turbulent cavitation concept
vaporous ~, 280 (UTCC), 431
surface vaporization, 280
~ irregularity, 437 variation principles of conformal mapping
~ tension, 12, 21, 145, 156, 257, 303, 373, 322
426, 448, 463 variable,
system asymptotic ~, 36
coupled tank-nozzle-bubble ~, 306 complex ~, 29, 62, 123, 185, 358, 512
one-row ~ of perforated pipes, 511 self-conjugate ~, 143, 539
~ of hydrofoils, 328 velocity,
~ of the ellipsoid of revolution and a ring edge ~, 397, 435, 437-8
wing, 331 complex ~, 43, 62, 89, 117, 132, 147, 166,
Taylor series, 39, 67, 71, 80, 122-3, 196, 195, 230, 275, 519
232, 274, 322, 361, 556 conjugate complex ~, 62, 355
theorem slip ~, 446
Cauchy ~, 64, 122 ventilation, 14-8, 238, 303-4, 319-21, 475-6
Dirichlet’s ~, 122 Villat’s solution for singular functions, 531
Schwarz’s ~, 38 viscose stress tensor, 383
Weierstrass’ ~, 122 viscosity, 18, 20-1, 24, 28, 178, 265, 383-7,
theory 429-30, 523
boundary layer ~, 434 void fraction, 22, 431, 443-5, 449, 455-60,
сhoked-flow ~, 304 465-73
lifting line ~, 479, 483, 495 vortex layer, 217, 488
lifting surface ~, 484, 488 vortex near a flat bottom, 115
linear ~, 45, 164, 295, 491 wake,
linear ~ of cavity stability and Kelvin ~, 429, 445, 449, 456-7
oscillations, 295 turbulent ~, 56, 259
nonlinear ~, 151, 164, 372 walls effect, 476
~ of analytical functions, 142, 247, 539 wave(s)
slender body ~, 203, 283, 293, 453, 460 breaking ~ 367
wave ~, 444 diverging ~,429, 445-8, 472
third Green’s identity, 531 ~ drag, 14, 429, 444, 467-73
torque, 33, 42, 86, 140, 166, 219, 237, 518, ~ field, 444, 458, 460, 472
525 generation (propagation) of ~, 369
transform ~ number, 449-51, 455, 458
Charny ~, 162 ~ pattern, 429, 444-5, 454-60, 472
Fourier-Bessel ~, 453, 464 shock ~, 23, 382, 444, 448, 457, 459
Fourier-Hankel ~, 452-3, 464 solitary ~, 361-2, 368
partial ~, 465 ~ system, 429, 444, 456-8
transformation transverse ~, 455-8, 472
Fourier ~, 250, 449 wettability, 437, 439
Galilean ~, 284, 286 wing
imaginary Jacobi ~, 530 rectangular ~, 254-6
Pologiy’s ~, 539 ring ~, 163, 330, 347
trapezoidal rule, 468 ~ of finite span, 256
tsunami, 369-72
turbulence stimulator, 442
593
Name Index
Name Index
NOTE: The subject index refers to page numbers. Dash line stands for all page numbers in-between;
in this case the first (and second) digits of the last page number can be omitted if it is the same as
that of first number.
Abelson, HI, 209 Bloor, MIG, 354 Clauser, FH, 438

Abramowitz, Milton, 454 Blottiaux, O, 430 Clayden, WA, 114, 295
Achkinadze, AS, 59, 60, Bobylev, DK, 72, 100 Cokelet, ED,365
475, 485-8, 491, 493, Bolotin, VV, 274, 276 Cole, JD, 200, 283
495-7 Bonder, J, 106 Cole RH, 278
Acosta, AJ, 134, 164,176, Bourgoyne, DA, 431, 440 Cox, PN, 114, 295
275, 400, 431, 442 Brebbia, CA, 322, 384 Courant, 385
Adams, MC, 203, 467 Brennen, CE, 12, 14, 27, Cueto-Felgueroso, L, 382,
Afanasiev, KE, 322, 358, 121, 142, 144-5, 203, 388
361-3, 367-8, 372-4, 376, 211, 257, 283, 304, 337, Cuthbert, J, 283,
382, 386-7 342, 372, 411-2, 427-8, d’Alembert, JR, 43, 449
Afanasieva, MM, 366 430, 441, 443, 449, 463, Dang, J, 343
Ahlberg, JH, 356 Brevig, P, 368 Davis, BV, 478
Ahuja, V, 431, Brewer, WH, 432 Davydov, YM, 378
Amromin, EL, 24, 26, 33, Brillouin, LM, 26, 29-30, Dimitrieva, NA, 128, 131,
145, 151, 426, 428-9, 153, 156, 343-5, 356-7, 142, 144-5, 203, 274-6,
432, 436, 438-42 398, 428-30, 433, 440 357, 539
Arakeri, VH, 398, 430-1, Brodetsky, S, 125, 127 Dobrovolskaya, ZN, 235,
433, 436, 439, 442 Buford, R, 265 265
Arndt, REA, 15, 156, 256- Buivol, VN, 19, 265, 284, Dommermuth, DC, 372
8, 321, 414, 441, 287, 426 Drew, DA, 443, 445-6, 449,
Arzoumanian, SH, 284, 290 Burill, WC, 475 463
Ashley, Holt, 283, 447, Burkart, MP, 479, 485 Duraiswami, R, 372, 374
449, 46-1, 466-8, 472 Butterfield, R, 322, 376, Efremov, II, 91, 164, 254;
Auslaender, J, 91, 164, 546 389-90
Baeva, MA, 485, Campbell, IJ, 19, 426 Efros, DA, 26,, 52, 56, 422
Banerjee, PK, 322, 376, Capodano, P, 83, Egorov, IT, 173, 199, 339
546 Caster, EB, 479 Elizarov, AM, 338-9
Barr, R, 487 Castillo, L, 435 English, JW, 478
Basin, AM, 486 Ceccio, SL, 441 Eppler, R, 58, 127
Batchelor, GK, 24-5, 278, Chahine, GL, 372, 374, 432 Epshteyn, LA, 25, 83, 114,
384 Chaplygin, SA, 79, 86, 88, 295, 298, 300
Bazhenov, VG, 378 100, 107, 129, 277-9, Etter, RJ, 477
Belotserkovskii, OM, 378-9 519, 522 Fabula, AG,
Bernoulli, Daniel, 28, 332 Chaplygin, YS, 100 Feit, D, 378
Bessel, F, 453, 464 Chapman, RB, 372 Falkner, VM,
Bervy, NO, 132 Charny, IA, 162 Fedotkin, IM, 269, 271
Betz, A, 108-9 Chechnev, AV, 23, 378-9, Filippov, VI, 23
Billet, ML, 442 381 Fine, NE, 415, 417
Birkhoff, G, 31, 56, 66, Cherepanov, GP, 83 Fisher, EA, 483-4
106, 198, 200, 207, 354, Cheung, YK, 322 Forbes, LK, 368
429-30 Chida, J, 60 Fourier, JBJ, 122-3, 188,
Blake, JR, 372 Chou, YS, 283 242, 250, 284, 287-9,
Bloch, EA, 203, 337, Cisotti, U, 26, 51, 114 354, 449, 452-3, 464
594
Name Index
Frankel, FL, 203 Grigoryan, SS, 201 Karpovich, EA, 203
Fridman, GM, 191, 197, Grilli, ST, 368 Kartuzova, TV, 88, 330
337, 348, 485, 497 Gubriy, VI, 108 Kato, H, 430, 439
Friedrichs, 385 Gurevich, MI, 29, 31, 52, Katz, L, 431
Froude, William, 14-7, 20, 65-6, 72, 76, 79, 82-3, Kawaguti, M, 127
23-4, 132-5, 138-41, 176, 87, 100, 127, 136, 185, Kawakami, E, 320
247, 250-2, 266-8, 283, 198, 205, 246, 492 Kedrinskij, VK, 365, 387
285, 290, 292, 321, 358, Gusev, VA, 52, 108, 176, Keldysh, MV, 192, 195,
361-2, 364, 366, 368, 265 219, 528
371-2, 411, 426, 476-7, Guzevsky, LG, 142, 199, Keller, AP, 426, 430, 442
487 203, 206, 284, 293, 337- Kelvin, WT, 429, 444-5,
Furuya, O, 487 9, 342, 361-2 449, 455-7, 472
Galanin, AV, 88, 91, 158, Haack, C, 368 Kermeen, RW, 254
176, 247, 250, 378 Hammer, PC, 548 Kevorkian, J, 283
Galin, LA, 378 Hankel, H, 221, 228, 452-3, Khakimov, A, 260
Garabedian, PR, 203, 283, 464 Khan, O, 23
285-6, 304, 312, 316, Harlow, FH, 379 Khomyakov, AN, 142
337, 342 Hart, DP, 257 Kiceiuk, T, 411
Gauss, CF, 367, 384, 438, Haskind, MD, 185 Kim, JW, 365
524, 536, 542 Helmholtz, H, 29, 177, 357 Kim, S, 431
Gawn, R, 488 Hess, JL, 425, 555 King, AC, 354
Georgievskaya, EP, Hilbert, David, 19, 426 Kinnas, SA, 337, 394, 415,
Gerlach, 100 Holl, JW. 442 417-8, 432, 487, 491
Geurst, JA, 43, 48, 164, Hoperskov, AE, 114 Kinzel, MP, 431
181-2, 221, 399-400, 491 Hromadka, TV, 358 Kirchhoff, 43, 51, 56-7, 69,
Gibson, DC, 372 Hsiao, CT, 432 71-2, 76-7, 83, 99, 105-9,
Gilbarg, D, 26, 52, 56 Hsu, CC, 185 124-9, 152-3, 187, 191-5,
Gingold, RA, 382 Ibragimova, TB, 486 221-4, 238-41, 271-2,
Gogish, LV, 438 Iga, Y, 431 381
Golubev, VV, 88, 332 Il’inskiy, NB, 158, 505, Kirschner, IN, 22-3, 279,
Goman, OG, 142, 539 508-9 284, 287, 290, 303, 337,
Gopalan, S, 431 Ikonnikov, VV, 477, 479 412-4, 429, 444, 458-9,
Gorelov, DN, 219 Isaacs, R, 207 462, 471
Gorshkov, Alexey, 440, Ivanov, AN, 29, 136, 142, Kiselev, OM, 23, 146
477 144-6, 151, 164, 185, Knapp, RT, 29, 341-2, 372
Goshev, GA, 486 337, 342, 439-40 Kochin, NE, 107, 120, 501
Gradstein, IS, 69, 177, 234, Ivchenko, VM, 212, 487 Kojouro, LA, 142, 144-5,
236-7, 255, 500, 541, 545 Jaswon, M, 322 337
Grant, John R, 429, 443-4, Jessup, SD, 477 Kolscher, M, 79, 129
462, 471 Johnson, VE, 91, 190 Konhauser, P, 354
Green, George, 63, 100, Joukowski, NE, 58-9, 103, Korobkin, AA, 230
247-8, 252-5, 322-5, 332, 107, 131-2, 145-7, 150, Kotlyar, LM, 53, 132, 141
337-9, 344, 349-50, 358, 158, 163-7, 171, 183-4, Kozhuhariv, P, 487
365, 370-1, 374-6, 387, 275, 324-36, 342, 345, Krasnov, VK, 142, 284,
397, 402, 407-8, 415, 347, 349-51, 353, 357-8, 337
422, 424, 444, 450, 452- 363-4, 402, 499-504, Kreisel, G, 26, 52, 422
3, 459-66, 468, 472-3, 512, 519, 522 Krikunov, YM, 264
499, 531-8, 543, 549-50, Kaplun, S, 517, 519 Kronecker, 356, 383
453, 456 Karlikov, VP, 56, 276 Kruppa, C, 477, 486-7
Gridneva, VA, 378 Karman, TV, 118, 120-1, Krylov, VI, 542
Grigorieva, IV, 374, 376 158, 438 Krylov, VV, 142, 337
595
Name Index
Kucharzyk, P, 88 Loyziansky, LG, 198, 287, Newton, RN, 488
Kudo, T, 487 522, 542 Nikitin, VV, 164, 239, 241-
Kuiper, G, 343 Lucy, LB, 382 2, 248
Kuria, IM, 283 Mach, E, 19, 22-3, 159, Nikitina, GV, 263-4, 268
Kutta, WM, 107, 167, 324- 277-9, 305, 380, 381-2, Nishiyama, T, 23, 203, 254
5, 334, 342, 347, 350, 426, 428, 443-5, 450, Nuzhin, MT, 166, 185, 338
353, 363-4, 386, 390, 456-60, 463-4, 466-73, Numachi, FI, 60
396-8, 402, 404, 415, Machinsky, AC, 269, 271 Oba, R, 164
499-500 Mackerle, 322 Ogilvie, TF, 420
Kuznetsov, AV, 23, 52-3, Mackie, AG, 247, Olofsson, N, 487
70, 83, 87, 108, 141-2, Makarov, VV, 114 Oseen, SW, 517
185, 229, 251, 268, 505 Makaseev, MV, 254 Otta, SP, 304
Kuznetsov, BG, 337 Maklakov, DV, 91, 95, Ovsyannikov, LV, 385
Kuznetsov, YV, 83, 87, 91, 354-5, 357, 362 Panchenkov, AN, 212, 219,
94, 337 Manevich, AS, 23, 251 253-4
Lai, C, 358 Marchuk, AG, 372 Parkin, BR, 136, 221
Lamb, H, 549-50 Maskalik, AI, 477, 479, Paryshev, EV, 203, 211,
Landahl, M, 283, 447, 449, 486-7 284, 290, 293, 295-6, 298
460-1, 466-8, 472 Matveev, KI, 426 Passman, SL, 443, 445-6,
Laplace, PS, 21-2, 28, 146, Mavlyudov, MA, 479-80, 449, 463
148, 158-9, 162, 248, 483-4, 486 Pavlova, NA, 274, 349, 352
250, 282, 285-8, 323-4, May, A, 211, 304, 410-1, Pearson, JRA,
419, 452, 498, 523, 532, 473 Peregrin, DH, 368
534, 541 McLeod, 146, 149-50, 155 Pernick, AD, 372
Larock, BE, 30, 56, 91, 95, McKenney, EA, 257 Petersohn, E, 108-9
132 Maijer, MC, 399 Petrov, AG, 198, 202-3,
Lasarev, VA, 57, 91, 95 Mensaas, ,K, 496 274, 365, 368. 372
Lavrent'ev, MA, 62, 67, Meshersky, 100 Petrov, NP, 517, 519
122, 185, 219, 322, 365, Michel, JM, 295, 298, 302, Plesset, MS, 372
387, 505, 518, 546-7 440 Plotkin, A, 193
Leehey, P, 254 Mikhailov, VM, 216, 220- Pologiy, 142-3, 154, 539
Lenau, CW, 132 1, 253, 268 Polubarinova-Kochina, P,
Levi-Chivita, T, 125, 127, Minakov, 107 511
354 Mokry, M, 358 Popov, VV, 142, 539
Levinson, N, 198, 205 Molk, J, 529 Porfiriev, NP, 247-8, 372
Levkovsky, YL, 376 Monaghan, JJ, 382-7 Poruchikov, VB, 23, 378
Lewy, H, 337, 385 Moran, JP, 247 Pozdunine, VL, 190, 478-9,
Ligget, JA, 365 Morgan, WB, 479 487
Lighthill, MJ, 170, 394 Morris, JP, 382, 385 Potashev, AV, 158, 505,
Lindau, Jules W, 431-2 Münzer, H, 283, 286, 304, 508-9
Liu, GR, 382, 385 312-3, 316 Poulhausen, K, 438
Liu, MB, 382, 385 Murty, 372 Prandtl, L, 18, 158, 164, 90,
Liu, PLF, 365 Muskhelishivili,, NI, 528 277, 279, 435
Liu, T, 487 Nakos, DE, 423 Prishchemikhin, YN, 476
Liu, WK, 382, 384 Narvsky, AS, 60, 487-8, Prosnak, W, 88
Logvinovich, GV, 19-20, 491, 493, 496-7 Prosperetti, A, 447
31, 198, 203, 208-9, 245, Nekrasov, AI, 125, 219, Proudman, I, 517
247, 265, 280-7, 290-1, 332 Pukhnachov, VV, 230
293, 295, 321, 381, 426 Newman, JN, 283, 285, Pyhteev, GN, 52
Longuet-Higgins, MS, 362, 288, 417, 420, 449, 454 Qin, Q, 431
365 Newton, I, 357, 382
596
Name Index
Rayleigh, L, 20, 43, 51, Schwartz, LW, 38-9, 65, Tachmindji, AJ, 479
280, 303-4, 307-8, 376, 67, 151, 192, 231, 240, Taic, OG, 209
447 259, 261, 359, 368, 370, Takahira, H, 372
Reichardt, H, 283, 286, 528, 530-1 Tannery, J, 529
304, 312-3, 316, 341-2 Sears, WR, 203, 467 Techet, AH, 426
Reisman, GE, 257 Sedov, LI, 26, 88, 100, 107, Telste, JC, 367
Rethy, M, 72, 100 111, 158, 164, 167-8, Terentiev, AA, 517, 519-20
Reynolds, Osborne, 14, 18, 192, 195, 212, 215, 217, Terentiev, AG, 23, 26, 29-
20, 22, 24-5, 426, 430- 243, 246, 499, 501, 518, 30, 35, 47, 51-4, 57-9,
43, 517, 523, 526 528 73, 78, 83, 87-8, 91, 96,
Riabouchinsky, D, 26, 43, Semenenko, VN, 284 105, 108, 113-4, 128,
516, 132, 141, 144, 199, Serebryakov, VV, 23, 198, 132, 141-2, 144, 146,
202, 337-40, 343, 347, 203-7, 209, 279, 283-4, 148-9, 151, 155, 158,
397-8, 409-10, 412-3, 342 163-4, 173, 180, 184,
415, 433, 435, 503 Shabat, BV, 62, 122, 322, 203, 212, 214, 216, 220-
Richardson, LF, 17, 468 518, 546-7 1, 230, 239, 241-2, 245,
Riegels, F, 170 Shahverdy, GG, 378 259-60, 263-5, 270, 274-
Risenkampf, BK, 511-3, Shepelenko, VN, 142, 337 5, 279, 284, 322-3, 330,
516 Sheshukov, EG, 239 332-3, 337-8, 349, 357,
Roberts, WB, 435 Sholomovich, GI, 276 361, 365-8, 370, 372-3,
Rock, HH, 26, 52 Sidorov, OP, 330 376, 378-9, 381, 387,
Roman, VM, 254 Signorini, A, 528 498, 509, 515, 517, 519-
Rose, JC, 477, 486 Silberman, E, 19, 20, 426 20, 522, 528, 533, 539,
Rosenthal, BJ, 303 Singhal, AK, 431 542
Roshko, A, 20, 58, 127 Skalak, R, 378 Thill, C, 24
Rowe, A, 430 Skan, SW, 437 Timman, R, 221, 420
Rozhdestvensky, KV, 190- Skramstad, HK,431 Todorashko, GT, 142, 337
1, 197, 348, 485, 495 Slaattelid, OH, 496 Tolokonnikov, SL, 56
Rubach, H, 118 Slezkin, NA, 146 Tomita, Y, 372, 374
Rudzki, MP, 132 Smirnov, VI, 122, 531-2 Toyoda, M, 485
Rusetsky, AA, 331, 483-4, Smirnova, TN, 332-3 Troepolskaya, OB, 53, 132
488 Smolyanin, VG, 365, 368 Troeshestova, DA, 372
Ryzhik, IM, 69, 177, 234, Snider, AD, 337 Truscott, TT, 426-7, 473-4
236-7, 255, 500, 541, 545 Sommerfeld, A, 517, 522, Tsapin, VD, 480, 487
Sadovnikov, YM, 479, 487 526 Tuck, EC, 365, 387, 420
Sagomonyan, AY, 230, Song, CCS, 19-20, 426 Tulin, M, 26, 30, 36-7, 44,
378, 381 Sonntag, RE, 304 51, 56-8, 60-1, 66, 71,
Salimov, RB, 185 Springorum, K, 477 91, 95, 118, 132, 134,
Samoilova, TI, 358, 361, Stegun, Irene A, 454, 554 136, 138, 141, 158, 164,
367 Stepanov, GY, 58,60, 107, 170-1, 176-9, 185, 190,
Sato, K, 372, 374 111, 128, 499 192, 203, 265, 283, 479,
Sautreaux, C, 132 Stoker, JJ, 248 485
Savchenko, YN, 23, 279, Stokes, GG, 382, 517 Tumashev, GG, 166, 185,
284 Street, RL, 30, 56, 91, 95, 338
Schiebe, ER, 254 118, 132, 136, 141, 283 Uhlman, JS, 203, 284, 395-
Schlichting, H, 437-8 Stukalov, SV, 358, 362-3, 401, 409-14, 422-3, 425,
Schmidt, Steffen J, 258 368, 372, 382 549, 552
Schmieden, C, 125, 127 Sturova, IV, 368 Ukon, Y, 485, 487
Schnerr, GH, 23, 258 Subkhanculov, GI, 142 Uryadov, AK, 348
Schubauer, GB, 431 Sulmont, P, 487 Vaidyanathan, S, 431
Symm, GT, 322
597
Name Index
Van Dyke, M, 170, 191, Voronin, VV, 203 Wu, TY, 30, 58-60, 108,
199, 203, 283, 434, 517 Vorus, WS, 203, 283, 487 221, 354, 491
Van Manen, JD, 476 Wade, RB, 113, 164, 400 Yakimov, YL, 198, 201,
Van Wylen, GJ, 304 Wagner, H, 158, 164, 215- 203, 265, 274, 279
Vanden-Broeck, JM, 361-2 7, 235, 246-7 Yamaguchi, H, 430, 438
Varghese, AN, 23, 203, Wang, HC, 477 Yanke, E, 228
283-4, 412-4 Wang, DP, 268 Yas’ko, NN, 142, 337
Vasiliev, VN, 108 Wang, OX, 377 Yasuda, A, 372
Vasin, A, 23, 279 Watson, GN, 52, 85, 96, Yim, B, 91, 230, 247-9,
Vaz, GNVB, 337, 376 261, 544 251, 268, 487
Verbruch, PJ, 182 Weber, M, 21-2, 147-8, Young, Y, 487
Verhagen, JHG, 265 152, 154-5, 426, 430-2, Yuasa, H, 487
Villat, H, 32, 125, 132, 436-7, 440-2 Yue, DKR, 372
354-5, 398, 428-30, 433, Weierstrass, 122 Zaitsev, NA, 486-7
440, 503, 531 Weinig, F, 26, 51, 107 Zarantonello, EH, 31, 56,
Vinje, T, 368 Weissinger, J, 163 198, 200, 207, 429-30
Vishnevsky, VA, 91, 113, Welch, JE, 67 Zhitnikov, VP, 142, 146, 149,
132 Wetzel, JM, 254 151, 155, 363, 539
Vlasov, OE, 505 Whittaker, ET, 52, 85, 96, Zhuravlev,YF, 203, 209,
Vlasenko, YD, 23 261, 544 Zienkiewicz, OC, 322
Voinov, OV, 273, 372 Widnall, SME, 254 Zigangarieva, LM, 23
Voinov, VV, 372 Williams, M, 362 Zuykov, YP, 337
Voit, SS, 372 Woods, LC, 185, 332
Voitkunsky, YI, 479 Wosnik, M, 256, 258, 320
Volterra, V, 214, 238, 246, Wrobel, LC, 342, 532
268
598
Books by Backbone Publishing Co.
P.O. Box 562, Fair Lawn, NJ 07410, USA
Tel: 201-447-1834; FAX: 201-670-7892; www.backbonepublishing.com
alfred.tunik@gmail.com
Engineering Books
The HYDRODYNAMICS of CAVITATING FLOWS by A.G. Terentiev, I.N. Kirschner,
J.S. Uhlman
ISBN 978-09742019-5-5, Trade Cloth, 598 pages, 7.5x10”
CYLINDRICAL SHELL FOUNDATIONS by Y.A. Pronozin, A.D. Gerber
RELIABILITY of STRUCTURES. ANALYSIS and APPLICATIONS by V. Raizer
SHALLOW WATER and SUPERCRITICAL SHIPS by A. Lyakhovitsky
SMALL WATERPLANNE AREA SHIPS by V. Dubrovsky, K. Matveev, S. Sutulo
HANDBOOK ON PLASTIC ANALYSIS IN ENGINEERING by L. Belenkiy
FATIGUE ANALYSIS OF SHIP STRUCTURES by S. Petinov
ISBN 0-9644311-8-1, Trade Cloth, 267 pages, 7.5x10"
SHIPS WITH OUTRIGGERS by V. Dubrovsky
ISBN 0-9742019, soft cover, 88 pages, 7x10"
MULTI-HULL SHIPS by V. Dubrovsky & A. Lyakhovitsky
ARC WELDING OF ALUMINUM & MAGNESIUM ALLOYS by Ryabov et al.
ISBN 0-9644311-7-3, Cloth, Hardbound, 160 pages, 7x9"
WELDING STRESS RELIEF BY EXPLOSION TREATMENT by Petushkov et al.
ISBN 0-9644311-1-4, Cloth, Hardbound, 179 pages, 7x9",
Arctic and Antarctic Sea Ice Navigation Series
ATLAS OF ICE AND SNOW OF THE ARCTIC BASIN AND SIBERIAN SHELF SEAS
by I.P. Romanov
ISBN 0-9644311-3-0, hardbound, 275 pages, 11x17"
ATLAS OF ARCTIC ICEBERGS. The Greenland, Barents, Kara, Laptev, East-Siberian,
and Chukchi Seas and the Arctic Basin by V. Abramov
ISBN 0-9644311-4-9, soft cover, 70 pages, 11x8.5"
ATLAS OF ANTARCTIC SEA ICE AND ICEBERGS by A.A. Romanov
ISBN 0-9644311-6-5, soft cover, 176 pages, 9x11"
1994-1995 Northern Sea Route Directory of Icebreaking Ships & 1996 Update
ISBN 0-9644311-0-6, soft cover, 217 pages, 8.5x11”

Hydrodynamics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Hydrodynamics

Uploaded by

Copyright:

Available Formats

The HYDRODYNAMICS

Edited by Alfred Tunik

Cover design by Jarek Kupsk

Printed in the United States of America

© 2011 Backbone Publishing Company

Library of Congress Control Number: 2011924385

Backbone Publishing Company

Alfred Tunik, PhD

Ch. 1. A Physical Description of Cavitation

The physics of cavitating flows emphasizes macro-scale effects. The micro-physics

Sec. 1.2. Dimensionless Quantities Describing Cavitating Flow

therefore its effectiveness in supporting the cavity. If such cooling is

cavity that is typically characterized by relatively low thermal conductivity to a

momentum transport is that due to dissipation. For most cavitating flows of

Fig.1.2.2 provides examples of cavitating flows involving extreme values of two of

Tail-slap scar Tip

Sec. 1.3. Main Assumptions in the Theory of Cavitating Flow

1.3.2. Flow at Cavity Closure

Sec. 1.4. Summary

Ch. 2. Main Boundary-Value Problems of Cavitating Flow

domain D, kinematical condition ( vn = ∂ϕ / ∂n = 0 ) at the total domain boundary and

Since differential dϕ is negative in this case, differential dx should be negative as

From this it follows that the derivative vanishes at

Sec. 2.2. Simple Examples of Cavitating Flows

v∞ = 1 / 1 + σ where σ is a given cavitation number.

2.2.2. Flow around a Cavity

2.2.4. Numerical Results

Fig.2.2.3. Comparison nonlinear Fig.2.2.4. Cavity shapes for nonlinear

Sec. 2.3. Fully Cavitating Flat Plate

2.3.1. Complex Potential

2.3.2. Derivative of the Potential

2.3.3. Logarithmic Function of Complex Velocity

• and the closed-cavity condition ∫ zζ (ζ )dζ = 0 .

Integral (2.3.11) is equal to 2πiB1, so the closed-cavity condition is

The cavity, the

2.3.6. Hydrodynamic Characteristics

2.3.8. Numerical Calculations

Fig.2.3.4. Coefficient of resultant Fig.2.3.5. Center of pressure for

Sec. 2.4. Partially Cavitating Flow

2.4.2. Complex Function Derivative

Transformation from the z –plane to the ζ- plane is written in a differential form as

Fig.2.4.6a shows that

The results presented here

Sec. 2.5. Review of Cavitating Flow Models

• the inlet velocity at infinity wz (c + id ) = −1/ 1 + σ ; (2.5.6)

Fig.2.5.4. Cavity shapes of

2.5.6. Cavitating Models with Rotational Fluid Motion

Sec. 2.6. Open Models of Cavitating Flow

attack angle α=30° and cavitation number σ = 0.5 is shown in Fig.2.6.4a.

Hence, Re[2π iA( a, c, d ) B (a, c, d )] = 0 , (2.6.16)

Fig.2.6.8. Two possible

Ch. 3. Method of Conformal Mapping

Proof. Choose the real and imaginary parts of integral as follows

2. An analytic function at inner points is expressed as contour integral (Cauchy integral)

Some theorems are given below without proof.

3.1.4. Analytic Extension through Boundary

as ( z − za∗ ) − λ . The negative values of λ are considered similarly.

Sec. 3.2. Cavitating Flow around a Flat Sided Wedge

Fig.3.2.1. a) Cavitating flow past a wedge; b) complex w-plane; c) parametric ζ-plane.

Fig.3.2.4. Velocity distribution over Fig.3.2.5. Dependence of drag

Sec. 3.3. A Wedge in a Channel and in a Jet

Besides, function ω ( w) has the only singularity as ( w − b) −1/ 2 at point B(b).

3.3.2. Analytical Solution of the Problems

Integral (3.3.8) can be calculated analytically