Professional Documents
Culture Documents
Hydrodynamics
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
The image of cloud cavitation in the flow past a fixed hydrofoil in the front cover
background is courtesy of Roger Arndt
ISBN-13 978-09742019-5-5
ISBN-10 0-9742019-5-2
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.
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
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]).
14
Ch. 1. A Physical Description of Cavitation
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
Ch. 1. A Physical Description of Cavitation
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
Ch. 1. A Physical Description of Cavitation
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
Ch. 1. A Physical Description of Cavitation
18
Ch. 1. A Physical Description of Cavitation
20
Ch. 1. A Physical Description of Cavitation
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
Ch. 1. A Physical Description of Cavitation
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.
Fram e 1
Shock
P rojectile Fram e 2
base +100 µs
Shock
Glassy
Fram e 3 cavity Base
+200 µs
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
Ch. 1. A Physical Description of Cavitation
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.
24
Ch. 1. A Physical Description of Cavitation
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].
27
Ch. 2. Main Boundary Problems of Cavitating Flow
29
Ch. 2. Main Boundary Problems of Cavitating Flow
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
Ch. 2. Main Boundary Problems of Cavitating Flow
32
Ch. 2. Main Boundary Problems of Cavitating Flow
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.
Fig.2.2.2.
Cavitating flow
along a
rectangular
step bottom:
a) streamlines;
b) spiral at the
cavity end.
35
Ch. 2. Main Boundary Problems of Cavitating Flow
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
Ch. 2. Main Boundary Problems of Cavitating Flow
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
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)
38
Ch. 2. Main Boundary Problems of Cavitating Flow
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)
39
Ch. 2. Main Boundary Problems of Cavitating Flow
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
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
Ch. 2. Main Boundary Problems of Cavitating Flow
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.
41
Ch. 2. Main Boundary Problems of Cavitating Flow
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.
42
Ch. 2. Main Boundary Problems of Cavitating Flow
⎛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
Ch. 2. Main Boundary Problems of Cavitating Flow
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 ) ⎟⎠
2.3.9. 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
Ch. 2. Main Boundary Problems of Cavitating Flow
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
Fig.2.4.1. (a) Partially cavitating flow; (b) the w-plane; (c) the parametric plane.
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
Ch. 2. Main Boundary Problems of Cavitating Flow
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
Ch. 2. Main Boundary Problems of Cavitating Flow
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].
2.4.5. Numerical Results
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
Ch. 2. Main Boundary Problems of Cavitating Flow
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
Ch. 2. Main Boundary Problems of Cavitating Flow
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.2.4.7.
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 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
Ch. 2. Main Boundary Problems of Cavitating Flow
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 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
Ch. 2. Main Boundary Problems of Cavitating Flow
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
Ch. 2. Main Boundary Problems of Cavitating Flow
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.
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
Ch. 2. Main Boundary Problems of Cavitating Flow
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
Ch. 2. Main Boundary Problems of Cavitating Flow
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.
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
Ch. 2. Main Boundary Problems of Cavitating Flow
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
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∞
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.
61
Ch. 3. Method of Conformal Mapping
1
Conditions (3.1.1) were derived first by D'Alambert in 1752 and then by Euler in 1755 [Lavrent’ev
& Shabat, 1965].
62
Ch. 3. Method of Conformal Mapping
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)
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)
64
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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)
⎣ ⎦
3.2.7. Numerical Results
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
Ch. 3. Method of Conformal Mapping
Fig.3.2.6. Dependence of cavity length and width on cavitation number and wedge angle.
73
Ch. 3. Method of Conformal Mapping
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)
74
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
( |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
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
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
Fig.3.3.6. Linearized problem of cavitating flow of a thin wedge in a channel (solid lines)
and a jet (dashed lines).
Fig.3.4.1.
Flow in the z-plane (a) and in the w-plane (b).
79
Ch. 3. Method of Conformal Mapping
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.
80
Ch. 3. Method of Conformal Mapping
−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 .
3.4.4. Parametric Solution
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
81
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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.
3.4.5. Numerical Results
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.
83
Ch. 3. Method of Conformal Mapping
⎛ 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
Ch. 3. Method of Conformal Mapping
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.
3.5.5. Numerical Results
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
87
Ch. 3. Method of Conformal Mapping
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.
88
Ch. 3. Method of Conformal Mapping
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
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°.
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
Ch. 3. Method of Conformal Mapping
∫ 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
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
98
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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.3.8.8.
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
Ch. 3. Method of Conformal Mapping
ln( v0 / v∞ )
102
Ch. 3. Method of Conformal Mapping
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.
Fig.3.9.1.
Cavitation flow in a channel (a) - the z-plane view;
(b) - parametric rectangle on the ζ -plane
105
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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.
(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.3.10.4 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
Ch. 3. Method of Conformal Mapping
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.
110
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
⎛ 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
Ch. 3. Method of Conformal Mapping
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 .
113
Ch. 3. Method of Conformal Mapping
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.
114
Ch. 3. Method of Conformal Mapping
115
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
∫ 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].
118
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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
Ch. 3. Method of Conformal Mapping
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.
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
1
Series (4.1.1) was derived for real functions first by B. Taylor and systematically used by Maclaurin.
122
Ch. 4. Method of Expansion in Series
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.
123
Ch. 4. Method of Expansion in Series
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
Ch. 4. Method of Expansion in Series
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
Ch. 4. Method of Expansion in Series
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
Ch. 4. Method of Expansion in Series
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.
128
Ch. 4. Method of Expansion in Series
129
Ch. 4. Method of Expansion in Series
yC + = 2acπ e + C 2 ,
−2 c
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
Ch. 4. Method of Expansion in Series
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.
131
Ch. 4. Method of Expansion in Series
132
Ch. 4. Method of Expansion in Series
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
Ch. 4. Method of Expansion in Series
∞
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
Ch. 4. Method of Expansion in Series
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
Ch. 4. Method of Expansion in Series
136
Ch. 4. Method of Expansion in Series
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
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
Ch. 4. Method of Expansion in Series
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
Ch. 4. Method of Expansion in Series
Fig.4.4.3.
Relationship between boundary
length and Froude number.
Fig.4.4.4.
P partially cavitating flow:
(a) flow on the z-plane;
(b) parametric circle on the ζ-plane.
139
Ch. 4. Method of Expansion in Series
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 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
Ch. 4. Method of Expansion in Series
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.
141
Ch. 4. Method of Expansion in Series
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.
4.5.1. Problem Statement
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.
142
Ch. 4. Method of Expansion in Series
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
Ch. 4. Method of Expansion in Series
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.
145
Ch. 4. Method of Expansion in Series
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.
4.6.1. Problem Statement
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
Ch. 4. Method of Expansion in Series
(
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
Ch. 4. Method of Expansion in Series
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.4.6.2.
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
Ch. 4. Method of Expansion in Series
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
Ch. 4. Method of Expansion in Series
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
Ch. 4. Method of Expansion in Series
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
Ch. 4. Method of Expansion in Series
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
153
Ch. 4. Method of Expansion in Series
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 ).
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
Ch. 4. Method of Expansion in Series
∞ ∞
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
Ch. 4. Method of Expansion in Series
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.
157
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
160
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
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).
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
Ch. 5. Asymptotic Method in Cavitating Flow
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 )
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
Ch. 5. Asymptotic Method in Cavitating Flow
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)
167
Ch. 5. Asymptotic Method in Cavitating Flow
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
168
Ch. 5. Asymptotic Method in Cavitating Flow
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 ⎠
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
Ch. 5. Asymptotic Method in Cavitating Flow
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.
170
Ch. 5. Asymptotic Method in Cavitating Flow
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
171
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
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.
173
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
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.
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.
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
Ch. 5. Asymptotic Method in Cavitating Flow
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;
Fig.5.3.8.
Cavity shape in a transverse
gravity field.
178
Ch. 5. Asymptotic Method in Cavitating Flow
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)
180
Ch. 5. Asymptotic Method in Cavitating Flow
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
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
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
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
187
Ch. 5. Asymptotic Method in Cavitating Flow
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
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
Ch. 5. Asymptotic Method in Cavitating Flow
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 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 )
Fig.5.5.3.
High
performance
foils.
*
This section was prepared jointly with Кirill V. Rozhdestvensky.
190
Ch. 5. Asymptotic Method in Cavitating Flow
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
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
Ch. 5. Asymptotic Method in Cavitating Flow
(
cavity, ω1+ = i 1 − ( x + 1) / x , ) x ∈ (0, ∞) , and the cavity boundary is found integrating
192
Ch. 5. Asymptotic Method in Cavitating Flow
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.
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
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
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
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).
200
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
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
(s ′)
2
s ′′ ln s +
2s
+
d
dx
( (
I (s ′) − s ′ ln 1 − x 2 )) = σ2 . (5.7.23)
202
Ch. 5. Asymptotic Method in Cavitating Flow
†
This section was prepared by Prof. Vladimir V. Serebryakov.
203
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
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
206
Ch. 5. Asymptotic Method in Cavitating Flow
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.
207
Ch. 5. Asymptotic Method in Cavitating Flow
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
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
Ch. 5. Asymptotic Method in Cavitating Flow
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
Ch. 5. Asymptotic Method in Cavitating Flow
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 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
Ch. 5. Asymptotic Method in Cavitating Flow
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
212
Ch. 6. Unsteady Cavitating Flows
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)
213
Ch. 6. Unsteady Cavitating Flows
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
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 .
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
Ch. 6. Unsteady Cavitating Flows
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
Ch. 6. Unsteady Cavitating Flows
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
Ch. 6. Unsteady Cavitating Flows
∫0 S − s′ γ (s′)ds′ = π U α , (6.1.26)
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)
218
Ch. 6. Unsteady Cavitating Flows
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 .
219
Ch. 6. Unsteady Cavitating Flows
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
Ch. 6. Unsteady Cavitating Flows
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)
{
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)
( )
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
Ch. 6. Unsteady Cavitating Flows
( )
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
0
∫ s−x 2
β1 ( x) dx = s1/ 4
s + s +1 (6.2.17)
223
Ch. 6. Unsteady Cavitating Flows
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
⎛
⎜
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
Ch. 6. Unsteady Cavitating Flows
⎛ 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)
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
Ch. 6. Unsteady Cavitating Flows
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.
226
Ch. 6. Unsteady Cavitating Flows
⎛ (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)
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
⎛ 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
Ch. 6. Unsteady Cavitating Flows
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
(
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
Ch. 6. Unsteady Cavitating Flows
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.
229
Ch. 6. Unsteady Cavitating Flows
230
Ch. 6. Unsteady Cavitating Flows
( 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
Ch. 6. Unsteady Cavitating Flows
∂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)
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)
232
Ch. 6. Unsteady Cavitating Flows
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
Ch. 6. Unsteady Cavitating Flows
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)
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
Ch. 6. Unsteady Cavitating Flows
⎧ ⎛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
Ch. 6. Unsteady Cavitating Flows
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 .
237
Ch. 6. Unsteady Cavitating Flows
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
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
238
Ch. 6. Unsteady Cavitating Flows
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
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
Ch. 6. Unsteady Cavitating Flows
240
Ch. 6. Unsteady Cavitating Flows
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 ) ⎠
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
Ch. 6. Unsteady Cavitating Flows
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.
242
Ch. 6. Unsteady Cavitating Flows
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.
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
Ch. 6. Unsteady Cavitating Flows
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
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
Ch. 6. Unsteady Cavitating Flows
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
Ch. 6. Unsteady Cavitating Flows
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.
248
Ch. 6. Unsteady Cavitating Flows
∞ ∞
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
∞
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
Ch. 6. Unsteady Cavitating Flows
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
250
Ch. 6. Unsteady Cavitating Flows
∞
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)
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
Ch. 6. Unsteady Cavitating Flows
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.
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 ⎠
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
Ch. 6. Unsteady Cavitating Flows
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.
∂
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
Ch. 6. Unsteady Cavitating Flows
⎡ ( 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.
256
Ch. 6. Unsteady Cavitating Flows
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.)
257
Ch. 6. Unsteady Cavitating Flows
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
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
Ch. 7. Approximation Method
⎧ 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
Ch. 7. Approximation Method
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)
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.
Thus, partial derivative of stream function at the free boundary is expressed as Cauchy’s
263
Ch. 7. Approximation Method
1
∞
ϕ t ( x′, t )
integral ψ t ( x, t ) = −
π ∫
x′ − x
−∞
dx′ . Since the kinematical condition at the free boundary
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
Ch. 7. Approximation Method
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.
265
Ch. 7. Approximation Method
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)
s2 − z 2
s
ϕ3t ( x′) x′dx′
w3t ( z , t ) =
πi ∫ ( x′
0
2
− z 2 ) s 2 − x ′2
. (7.2.20)
267
Ch. 7. Approximation Method
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
Ch. 7. Approximation Method
Fig.7.2.6.
Vertical entry of
fully ventilated
plate:
(a) – lift
coefficients;
(b) - torque
coefficients.
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
Ch. 7. Approximation Method
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
Ch. 7. Approximation Method
(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
Ch. 7. Approximation Method
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.
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.
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
Ch. 7. Approximation Method
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.
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
Ch. 7. Approximation Method
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
Ch. 7. Approximation Method
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.
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
∇ 2ϕ = 0
u(x) ⋅ n(x) = 0 p ( x ) = pc
0.5
W
y0 g
z
x
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
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
287
Ch. 8. Supercavity as a Dynamical System
∂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
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
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 ⎠
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
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
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.
*
This section was prepared by Emil Paryshev
293
Ch. 8. Supercavity as a Dynamical System
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 )
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
Ch. 8. Supercavity as a Dynamical System
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.
295
Ch. 8. Supercavity as a Dynamical System
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 ⎠
296
Ch. 8. Supercavity as a Dynamical System
297
Ch. 8. Supercavity as a Dynamical System
298
Ch. 8. Supercavity as a Dynamical System
Fig.8.5.1.
Stability boundaries at dimensionless time:
solid lines stand for critical value αcr;
dashed lines stand for stable value αj.
299
Ch. 8. Supercavity as a Dynamical System
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).
300
Ch. 8. Supercavity as a Dynamical System
301
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
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 .
303
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
(γ +1)
Δ⎛ 2 ⎞ 2(γ −1)
and Γ =⎜ ⎟ = Γ + Γ− 2Γ+
. (8.6.13)
⎝ γ +1⎠
306
Ch. 8. Supercavity as a Dynamical System
(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− ⎦
308
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
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 ⎥⎦
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
Ch. 8. Supercavity as a Dynamical System
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
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
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
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.
314
Ch. 8. Supercavity as a Dynamical System
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
Ch. 8. Supercavity as a Dynamical System
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.
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
Ch. 8. Supercavity as a Dynamical System
no vehicle envelopment:
blow-down equation
coupled system
critical time
Fig.8.6.6. Simulation results for the notional supercavitating vehicle and operating
conditions in Table 8.6.3 for a canister volume of 300 cm3 and an orifice diameter of 2.4mm.
Fig.8.6.7. Simulation results for the notional supercavitating vehicle and operating
conditions in Table 8.6.3 for a canister volume of 500cm3 and an orifice diameter of 4.0mm.
317
Ch. 8. Supercavity as a Dynamical System
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.
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
Ch. 8. Supercavity as a Dynamical System
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.
Notional system
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
Ch. 8. Supercavity as a Dynamical System
equation (i.e. ignoring the effects of cavity expansion dynamics on the gas discharge
process).
Notional system
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
Ch. 8. Supercavity as a Dynamical System
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.)
321
Ch. 9. Numerical Methods
constant or linear, is O (ln2 ) or O ( N −2 ) . This justifies using constant elements in the BEM.
∂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
Ch. 9. Numerical Methods
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
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
Ch. 9. Numerical Methods
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)
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
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
Ch. 9. Numerical Methods
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
Ch. 9. Numerical Methods
Fig.9.1.4.
Speed distribution
on Joukowski foil:
solid lines
correspond to
numerical data,
points – to
analytical results.
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
Ch. 9. Numerical Methods
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
Ch. 9. Numerical Methods
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
Ch. 9. Numerical Methods
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.
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
Ch. 9. Numerical Methods
Thus the speed potential and velocity can be found. Namely, the speed on the circle ζ = eiσ
333
Ch. 9. Numerical Methods
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
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)
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.
9.2.5. Numerical Results
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
Ch. 9. Numerical Methods
(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
Ch. 9. Numerical Methods
339
Ch. 9. Numerical Methods
(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)
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 )
interpolating functions;
3) correct the wetted boundary by given C and the cavity boundary by moving
b
th
point, the n iteration is determined by the ordinates of the cavity nocks, y = y . (n) ∗( n )
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
Ch. 9. Numerical Methods
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.
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.
342
Ch. 9. Numerical Methods
∗( 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
Ch. 9. Numerical Methods
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.
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
Ch. 9. Numerical Methods
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
Ch. 9. Numerical Methods
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
Ch. 9. Numerical Methods
Fig.9.4.10.
Fully cavitating
curvilinear foil.
Fig.9.4.11.
Ring wing with partial cavities
347
Ch. 9. Numerical Methods
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.
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.
348
Ch. 9. Numerical Methods
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
Ch. 9. Numerical Methods
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
Ch. 9. Numerical Methods
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).
351
Ch. 9. Numerical Methods
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
Fig.9.5.7.
Joukowski foil near
free boundary, with
and without a
cavity.
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
Ch. 9. Numerical Methods
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.
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)
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
Ch. 9. Numerical Methods
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
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
Ch. 9. Numerical Methods
⎛ 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
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.
9.6.6. Numerical Results
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
Ch. 9. Numerical Methods
a) b)
c)
358
Ch. 9. Numerical Methods
determined by Cauchy integral via its boundary value
1 w( z ′)
w( z ) = ∫
2π i C z ′ − z
dz ′ . (9.7.1)
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
Ch. 9. Numerical Methods
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
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
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
Ch. 9. Numerical Methods
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
363
Ch. 9. Numerical Methods
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.9.7.5a&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
Ch. 9. Numerical Methods
365
Ch. 9. Numerical Methods
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
366
Ch. 9. Numerical Methods
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
Ch. 9. Numerical Methods
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).
∫
−∞
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)
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
Ch. 9. Numerical Methods
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].
372
Ch. 9. Numerical Methods
373
Ch. 9. Numerical Methods
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].
374
Ch. 9. Numerical Methods
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
Ch. 9. Numerical Methods
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
Ch. 9. Numerical Methods
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
0.6
0 2 4 6 8 10 t
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
Ch. 9. Numerical Methods
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.
378
Ch. 9. Numerical Methods
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
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
Ch. 9. Numerical Methods
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 − .
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
Ch. 9. Numerical Methods
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.
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
Ch. 9. Numerical Methods
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
Ch. 9. Numerical Methods
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
384
Ch. 9. Numerical Methods
⎧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
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 .
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.
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
Ch. 9. Numerical Methods
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
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=0.1017s
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
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
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
Ch. 10. Method of Singularity in Cavitating Flow
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 .
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
10. Method of Singularity in Cavitating Flow
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
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
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
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
10. Method of Singularity in Cavitating Flow
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
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.
395
10. Method of Singularity in Cavitating Flow
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].
396
Ch. 10. Method of Singularity in Cavitating Flow
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].
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
10. Method of Singularity in Cavitating Flow
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
10.2.3. Hydrodynamic and Geometric Characteristics
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
Ch. 10. Method of Singularity in Cavitating Flow
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
10. Method of Singularity in Cavitating Flow
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
10.3.1. Integral Equations
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
Ch. 10. Method of Singularity in Cavitating Flow
( 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
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
10. Method of Singularity in Cavitating Flow
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
⎧2π , ( x, y ) ∈ S
⎪
Green’s function and β = ⎨π , ( x, y ) ∈ ∂S (10.4.2)
⎪ 0, ( x, y ) ∈ S c
⎩
402
Ch. 10. Method of Singularity in Cavitating Flow
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
10. Method of Singularity in Cavitating Flow
( 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
∂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
Ch. 10. Method of Singularity in Cavitating Flow
∂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.
10.4.2. Discretization and Solution of the Integral Equations
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
−
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
j ∋ C j ∈C foil Cj ∂n j ∋ C j ∈Ccav ∂n j Cj +
j ∋ C j ∈Ccav 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
405
10. Method of Singularity in Cavitating Flow
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
‐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
Ch. 10. Method of Singularity in Cavitating Flow
Fig.10.4.5. 0.20
y
⎧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
⎡ ∂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
Ch. 10. Method of Singularity in Cavitating Flow
∂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 ∈∑
j ∋ C j ∈Ccav ∂n j C j
∫ nx rds
C foil + C jet C j
(10.5.8)
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
10. Method of Singularity in Cavitating Flow
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
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
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)
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
411
10. Method of Singularity in Cavitating Flow
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.
412
Ch. 10. Method of Singularity in Cavitating Flow
∂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 − ∑ ∫
j ∋ C j ∈Ccav ∂n j C j
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)
j ∋ C j ∈Ccav C j
∂G
+U ∑ ∫ ( x − x ) ∂n ds 0
j ∋ C j ∈Ccav C j
⎡ ∂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 ∈C∑foil +Cwall C∫ nx rds
j ∋ C j ∈Ccav ∂n j C j
(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
10. Method of Singularity in Cavitating Flow
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].
Fig10.6.3.
Illustration of non-
realizable cavity cases.
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.)
414
Ch. 10. Method of Singularity in Cavitating Flow
F= ∫∫ npdS
S
(10.6.7)
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.
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
10. Method of Singularity in Cavitating Flow
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.
10.7.2. Discretization and Solution of the Integral Equations
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
Ch. 10. Method of Singularity in Cavitating Flow
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 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
10. Method of Singularity in Cavitating Flow
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.
10.7.3. Hydrodynamic and Geometric Characteristics
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
Ch. 10. Method of Singularity in Cavitating Flow
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.
†
This section has been adapted from material presented, in part, in [Uhlman et al, 2001].
419
10. Method of Singularity in Cavitating Flow
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
Ch. 10. Method of Singularity in Cavitating Flow
∂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. Method of Singularity in Cavitating Flow
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
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
Ch. 10. Method of Singularity in Cavitating Flow
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
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
Ch. 10. Method of Singularity in Cavitating Flow
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
426
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
428
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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.
x
This section was jointly developed with Dr. Edward Amromin based on [Amromin, 2007].
429
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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].
432
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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.
433
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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 .
438
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
439
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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.
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
440
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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.
xi
Dr. John R. Grant made significant contributions to the mathematical formulation of the model
discussed in this section.
443
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
448
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
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
M = 9 );
2
m
0.3
0.0
0.0 0.5 1.0 1.5 2.0
k squared
452
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
455
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
continues to increase,
radial distance
radial distance
0.1 0.1
-0.3 -0.3
radial distance
-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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
0.2 0.2
radial distance
radial distance
cm=997m/s,
radial distance
radial distance
0.1 0.1
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
axial distance axial distance
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
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.)
459
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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)
464
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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)
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
(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).
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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.
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
0.014 0.014
R0 = 0.0001
0.012 R0 =0.01
0.012
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
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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
Ch. 11. Selected Effects in the Modeling of Cavitating Flows in Real Fluids
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 .
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
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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.
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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.
482
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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.
486
Ch. 12. Cavitating Flow of Marine Propellers
487
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
(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
Ch. 12. Cavitating Flow of Marine Propellers
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.
493
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
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
Ch. 12. Cavitating Flow of Marine Propellers
optimization of its profiles requires an improvement of their hydrodynamic fineness by at
least 10% to obtain only 1% growth in efficiency.
498
Ch. 12. Cavitating Flow of Marine Propellers
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.
499
Ch. 12. Cavitating Flow of Marine Propellers
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.
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 ) .
501
Ch. 12. Cavitating Flow of Marine Propellers
⎡ 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.
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
Ch. 12. Cavitating Flow of Marine Propellers
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
506
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
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).
2λ (1 − a 2 )a
ζ 2 +1
wζ = −i . (13.1.5)
π (ζ − a 2 )(a 2ζ 2 − 1)
2
Fig.13.1.4.
Dependence of
cavity dimensions
on parameter λ .
508
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
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.
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 ..
Fig.13.1.7.
Cavity shapes due to impact over a unit length.
510
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
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
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
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
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
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).
513
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
Fig.13.2.3.
Free lines for Cq = 2 .
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
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
Qc q
− = 0. (13.2.18)
c + a 1 + a2
22
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
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
( 2e − 1) − 1 .
2
c = 2eπ kH / Q − 1 − π kH / Q
(13.2.26)
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.
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.
*
See footnote to the title of Sec. 13.2 on page 511.
517
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
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
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)
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
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
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
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
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
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
(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.
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
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
Fig.13.3.2.
Streamlines of a cylinder
moving in a cylinder:
(a) rotation
(b) horizontal movement.
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 .
524
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
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
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
⎧ 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.
13.3.5. Numerical Calculations
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
Ch. 13. Applications of Cavitation Theory to Continuum Mechanics
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.
Fig.13.4.2.
Resulting force and torque for an asymmetric piston
moving in a box.
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
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.
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 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
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).
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 .
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)
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
yη
where k =2 .
(x −ξ ) + ( y +η )
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.
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)
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 .
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
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)
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.
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
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 ′ .
541
Appendices
∂U m 1 ∂V ∂U m 1 ∂V
follows: = 1+ 2 m m , = − 1+ 2 m m . (A4.16)
∂x y ∂y ∂y y ∂x
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.
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
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 .
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
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)
547
Appendices
Fig.A5.3.
Comparison of computing by approximated
function and by exact formulas (points).
548
Appendices
∂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)
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)
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∞
∫ n ( p − p ) ds = D; ∫ n ( p − p ) ds = 0; ∫ n ( p − p ) ds = 0;
Cb
x c
Cc
x c
Cj
x 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
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.
Fig.A8.1. Sc
Axisymmetric control volume with
boundary designations, after
[Uhlman, 2006]. Sb n
d dj
Inflow Sj
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 ) − ρ ⎢ cos (φ ) + ⎜
1
2 ⎟ ⎥
⎢⎣ 2π r ⎝ 4π r ⎠ ⎥⎦
2 2
∫∫ 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)
553
Appendices
Fig.A9.1. z y
( x, r , α )
Geometry definition.
α (ξ , ρ ,θ )
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
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
⎝ ⎠
∂ ⎛ ⎞
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
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.
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.
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
ISBN 978-09742019-8-6, Trade Cloth, 121 pages, 6.5x9”
RELIABILITY of STRUCTURES. ANALYSIS and APPLICATIONS by V. Raizer
ISBN 978-09742019-7-9, Trade Cloth, 146 pages, 7.5x10”
SHALLOW WATER and SUPERCRITICAL SHIPS by A. Lyakhovitsky
ISBN 978-09742019-5-5, Trade Cloth, 272 pages, 7.5x10”
SMALL WATERPLANNE AREA SHIPS by V. Dubrovsky, K. Matveev, S. Sutulo
ISBN 978-09742019-3-1, Trade Cloth, 255 pages, 7.5x10”
HANDBOOK ON PLASTIC ANALYSIS IN ENGINEERING by L. Belenkiy
ISBN 978-09742019-2-8, Trade Cloth, 1055 pages, 7.5x10”
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
ISBN 0-9644311-2-2, Trade Cloth, 496 pages, 7.5x10”
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”