ne-dimensional

wo-phase Flow
Graham B. Wallis
Associate Professor of Engineering
Thayer School of Engineering
Dartmouth College
One-dlmensional Two~phase F'low

Copyright© 1969 by McGraw-Hill, lnc. All rights reserved.

Printed in the United States of America. No part of this
publication may be reproduce-el, stored in a retrieval system,
or transmitted, in any form or by any mea ns, electronic,
mechanical, photocopying, recording, or otherwise, without
the prior written permission of the pub!isher.

Library of Congress Catalog Card Number 75-75170

07-067942-8

34567890 KPKP 79876

Preface

The purpose o! this book is to make a thorough presentation o! the basic

techniques far analyzing one-dimensional two-phase flmvs and to show
how thcy can be applied to a wide variety o! practica! problems. The
subject has immense importance in a large variety of traditional cngi-
neering disciplines. I t is ripe far development to the point where it
can rank on a par \\ith compressible fluid fiow and boundary layer theory
as a mature branch of fluid mcchanics. Eventually sorne of the simpler
material should diffuse down into early undergraduate courses.
Since it would take many volumes to do justicc to all aspects of
two-phase flows, the scope o! this book has been !imited to flows which
are essentially one Wmensional 1 occurring in ducts and channels or on
continuous surfaces. The use of velocity and concentration profiles per-
mits variations across the direction of flow. l\!Iultidimensional phe-
nomena are introduced only ,vhen they are needed to put the one-dimen-
sional assumptions in perspectivc or to provide results which are needed
befare the problem can be adequately specified. For example, the two-
dimensional flo\v field around and within a bubble rising in a liquid is
V
 PREFACE

not derived or discussed, but the resulting depcndence of velocity on

diameter is quoted. Of course1 many of the one-dimcnsional equations
and techniques are readily extended to the general case by substituting
the appropriate vectors for scalars in the mathematics.
In the interest of brevity, attention has been confined almost
entirely to the fluid mechanics of hvo-phase flo,v. Detailed cliscussions
of heat and mass transfer phenomena and of the mechanics of nucleation
during phase change have been omitted. The reader ,vho is interested
in certain problems of boüing 1 condcnsation, flashing, frcezing, combus-
tion1 and flows ,vith chemical reaction will therefore find that he may
need to gather sorne supplementary information befare a complete
analysis can be performed.
Most of the analytical developments have been seleeted for their
generality and uscfulness for predicting paramcters which are of engi-
neering interest. Typical variables of concern are pressure and density
variations along a duct, rates of discharge from containers1 liquid film
thickness 1 spatial variations in the concentrations of the phases 1 flmy
patterns, and the factors ,vhich limit the performance of equipment.
The book is designed for use both as a university text and as a
reference for engineers and researchers. It is divided into hvo parts.
Part lJ consisting of Chaps. 1 to 7, is concerned with analytical tech-
niques which have a broad and quite general validity. These techniques
are illustrated by examples from various fields of engineering. The
emphasis is on generality, the mastering of key concepts, and the develop-
ment of analytical skill, familiarity, and facility.
Part 2 is organized around particular phase combinations, flmv
regimes, and practica! applications. For exampleJ the spccific topics of
fluidization and sedimentation are considered as subsections under the
heading of "Fluid-partiele Systems" which is the title of Chap. 8. An
entire chapter is devoted to each of the major gas-liquid flow patterns.
Ali of the techniques which are developed in Part 1 are used in Part 2
when they are appropriate for thc particular situation. Specific calcula-
tion proceduresJ correlations, flm:v regime boundaries 1 and the individual
idiosyncrasies of certain systcms are discussed. The emphasis is on the
prediction of useful engineering parametcrs.
Single-phase flows are diseussed at any length only in Chap. 6 in
arder to clarify the concepts of wave motion and interaction. lVIany
textbooks in fluid mechanics are weak in this area. 1-i'ew1 if any, make
any reference to continuity ,vaves. lVIost of them fail to point out that
compressibility and gravity waves are examples of a general class of
dynamic waves ,vhich are produced by forces resulting from any form of
concentration gradient.
In order to preserve a reasonable size, many theoretical develop-
PREFACE vil

ments are rclegated to the status of problems in which the route is indi-
cated or the answer is provided but the dctails of the derivation are
omitted. Thc many fascinating aspects of multidimcnsional, multiphase
flo1n are also cxcluded except 1Yhen they have special relevance for
completing the description of particular phenomena.
The reader is assumed to possess a basic knmvledge of fluid
mechanics and thermodynamics and to be familiar ,vith the methods for
deriving conscrvation equations from suitable control volumes. No
attempt is made to reitera te the derivation of fundamental theorems and
in most cases merely thc results are written down. I t may be useful
for students who are unfamiliar ·with the particular form of sorne of these
relationships to derive them from first principles and gain confidence in
their validity.
From the academic point of vimv the book is probably too long
far a single course. Parts 1 and 2 can be used for a two-term sequence
for graduate studcnts in mechanical, civil 1 or chcmical engineering or
with a particular interest in fluid mechanics in a flexible curriculum. A
more clementary course 1 perhaps suitablc for a senior elective follmving
a general fluid mcchanics course, can be constructed by combining
Chaps. 1 to 4 with relatcd material from Chaps. 8 to 12. The more
advanced tapies of ·wave motion, unsteady flm,v, velocity profiles, and
interfacial phenomena can then be reserved for a later coursc. Honors
students who are more interested in the overáll thcorctical picture might
study Part 1 in detail, leaving Part 2 for independent readíng.
For the professional engineer and researcher, Part 1 provides a
thcoretical perspective ,vithin which he can vimv the entire scope of the
subject and perhaps broaden his interests. Part 2 provides the answers
to more specific practical problems.
I have not tried to ,vrite a scientific or technological history. A
detailed account of who did ,vhat, when 1 and hmv has been avoided.
The book is in tended to be read on its own and to be intelligible without
frequent referencc to the pub!ished !iterature. Bibliographical material
iR cited only in arder to give credit where credit is due or to indicate
,vhere the reader may find more specific and detailed information.
The text has developed from parts of the leeture notes for snmmer
courses which I oflered jointly with .John G. Collier at Dartmouth College
(1965, 1966, 1968), The University of Glasgow (1967), and Stanford Uni-
vcrsity (1967). This material has been considerably reorganized to be
more suitable for use by students and by those ·who are unfamiliar 1vith
the field. Numerous worked examples and problems have been added to
give physical meaning to the theoretical concepts and to promote an
awarencss of practical applications.
I am very much indebted to my students, .J. Michael Turner,
viii PREFACE

Andre\V Portcous, Philip E. i\Ieyer1 Dale E. RungeJ Stephen S. l\Iac Vean,

Stanley F. Birch, Gary Grulich, Thomas E. Brady, and Donald A. Steen
,vho worked problems, searchcd the literature, wrote computer programs,
and performed experiments ,vhich contributed immensely to the text.
Thc \York would also have becn impossiblc without the administrative
and clerical support of the Thayer School of Engineering at Dartmouth
College, in particular thc thoroughly reliable and painstaking secretarial
work of Edith G. Henson.
The part of Chap. 3 describing entropy generation in separated
flov,c is mostly taken from my articlc in the J nternational J ournal of H eat
and Mass Transfer (vol. 11, pp. 459-472, 1968) and is rcprinted by per-
m1ss10n. I am also indebted to the Institution of Chcmical Engineers
(London 1 England) for allmving me to quote the material on sedimenta-
tion in Chap. 8 from my paper at thc Symposium on the Interaction
between Fluids and Particles, London, 19G2.
The choice of nomenclature ,vas mostly dccided during conversa-
tions ,vith Dr. Novak Zuber in 1964, during which ,:re tried to develop a
consistent set from thc prolific symbols used by numerous authors in
many fields.

GRAHAM B. w ALLIS
Contents

Preface v
list of Symbols xvii

par! one ANALYTICAL TECHNIQUES

1 lntroduction 3

1.1 What Is Two-phase Flow? 3

1.2 Methods of Analysis 5
Correlations 5
Simple Analytical Models 5
Integral Analysis 6
Differential Analysis 6
Universal Phenomena 6
1.3 Flow Regimes 6
1.4 N otation 9
Simple Definitions 9
Properties 14
X CONTENTS

Pressrne Drop 14
Coordinates 14
Units 15
Problems 15
References 16

2 Homogeneous Flow 17

2.1 Introduction 17
2.2 One-dimensional Steady Homogeneous Equilibrium Flow 18
Further Development of the Momentum Equation 23
2.3 The Homogeneous Friction Factor 26
Laminar Flow 26
Turbulent Flow 28
2.4 Pressure Drop in Bends, Tees, Orifices, Valvesi Etc. 35
2.5 Unsteady Flow 35
Problems 37
References 41

3 Separated Flow 43

3.1 Introduction 43
3.2 Steady Flow in which the Phases Are Considered Together but Their
Velocities Are Allowed to Differ 43
Continuity 44
Momentum 44
Energy 45
Evaluation of Wall Shear Stress and Void Fraction 49
Flow of Boiling Water in Straight Pipes 55
3.3 One-dimensíonal Separated Flow in which the Phases Are Considered
Separately 61-
Continuity Equations 61
Momentum Equations 61
3.4 Flow with Phase Change 64
3.5 Flow in which Inertia Effects Dominate 68
3.6 Use of the Concept of Entropy Generation to Evaluate the Coefficient
" 73
3. 7 Energy Equations 80
Problems 82
References 88

4 The Driftaflux Modei 89

4.1 Introduction 89
4.2 General Theory 90
4.3 Gravity-dominated Flow Regimes with No Wall Shear 90
4.4 Corrections to thc Simple Theory 97
4.5 Sign Conventions and Identification of Components 1 and 2 101
4.6 Unsteady Flow 103
CONTENTS

Problems 103
References 105

5 Velocity and Concentration Profiles 106

5.1 Introduction 106
5.2 Qualitative Aspects 107
5.3 Differential Analysis 108
Velocity Profiles in Single-phase Flow 108
Velocity Profiles in Two-phase Flow 109
5.4 Integral Analysis 115
5 ..5 More Complex Methods of Analysis 117
Problems 118
References 121

6 One~Dimensional Waves 122

6.1 Introduction 122
6.2 Continuity Waves in Single-phase Flow 123
The Formation and Stability of Continuity Shocks 127
Stability of Continuity Waves 130
The Effect of a Source of Matter 130
6.3 Continuity Wavcs in Incompressible Two-component Flow 133
6.4 Dynamic Waves 135
Dyna.mic Waves in Single-component Flow 135
Examples of Dynamic Waves in Single-component Flow 136
Long Wavcs in a Canal of Constant Width 136
Waves in a Homogeneous Compressible Fluid 187
Dynamic Waves in Incompressible Two-component Flow in a
Constant Area Duct 137
An Example of Dynamic Waves in Incompressible Two-component
Flow; Waves in a Rectangular Horizontal Duct 139
"'T
The Effect of Compressibility on Dynamic aves in Two-componcnt
Flow 141
The Effect of Phase Change 145
6.5 The Interaction between Dynamic and Continuity Waves 146
Single-phase Flow 146
Incompressible Two-component Flow 149
6.6 Dynamic Shock \Vaves 152
Normal Compressibility Shocks 152
Oblique Shock Waves 154
Relaxation Phenomena 155
Problems 156
Reterences 160

7 !nterfacial Phenomena 161

7.1 Introduction 161
7.2 Velocity Boundary Conditíons 161
7.3 Stress Bounda.ry Conditions 162
Surface-tension Effects 162
xii CONTENTS

7.4 The Effect of Phase Change on Interfacial Stresses 165

7.5 Further Effects 169
Problems 169
References 172

par! lwo PRACTICAl APl'LICATIONS

8 Suspensions of Particles in Fluids 175

8.1 Introduction 175
8.2 One-dirnensional Vertical Flow of a Uniform Incompressible
Dispersion with No Wall Friction 176
General Theory of Uniform Steady Flow 176
Terminal Velocity of a Single Particle 176
Evaluation of the Index n 178
Forces on the Pa.rticles and the Fluid 179
8.3 Particulatc Fluidization 183
The Minimum Fluidization Velocity 183
Pressure Drop through a Fluidized Bed 183
Fixed Bed 183
Incipient Fluidization 185
The Fluidized State 186
Summary of Caleulation Procedures far Particulate Fluidized Beds 187
Stationary Bed 187
l\ifoving Bed 187
8.4 Unstea.dy Flow in Particle Dispersions 189
Propaga.tion of Continuity Waves 189
8.5 Batch Sedimentation 190
A "Generalized" Representation of Batch Sedimentation 194
8.6 Particle-particle Forces 201
8.7 Unsteady Flow in the Presence of Particle-pilrticlc Forces 204
8.8 Stability of Fluidized Systems 205
8. 9 Compressible Flow of Particle Suspensions 207
One-dimensional Stcady Flow 207
Homogeneous Equilibrium Flow 209
Limiting Gases of Nonequilibrium Flow 210
Velocity Equilibrium, Thermal Insulation 210
Thermal Equilibriumi Velocity Insulation 210
Thermal and Velocity Insulation 210
Similar Solutions for Constant Fractional Lag 211
Perturbation Techniques 213
Shock Waves 213
The Reia.xation Zone 215
Oblique Shocks 218
Two- and Three-dimensional Effects 218
Other Effects 219
8.10 Additional Force Components in Rapidly Accelerating Flows 219
CONTENTS

Apparent Mass 219

The Basset Force 221
8.11 Friction Characteristics of Particle Suspensions 222
Laminar Flow 224
Turbulent Flow 228
Pneumatic Transport 228
8.12 Nonuniform Particle Distribution 228
Stratification 229
Gravitational Effects 229
Symmetrieal Radial Concentration Variations 230
Channeling or Spouting in Fluid.ized Beds 230
Periodic Flows 230
Slugging 230
Wave Formation in Stratified Flow 231
Aggregative Flows 231
Flocculation 231
Bubbling 232
8.13 Percolation Theory 234
Problems 235
References 239

9 Bubbiy Flow 243

9.1 Introduction 243

9.2 Bubble Formation 244
Bubble Formation at an Orifice 244
Formation of Bubbles by Taylor Instability 246
Formation of Bubbles by Evaporation or l\!Iass Transfer 247
The Influence of Shear Stresses on Bubble Size 247
9.3 One-dimensional Vertical Flow of a Bubbly Mixture without Wall
Shear 247
The Rise Velocity of Single Bubbles 24.8
The lnfluence of Containing Walls 251
Infiuence of Vibrations 252
The Influence of Void Fraction 252
Modifications to .the Simple Theory to Take Account of Variations in
Concentration and Velocity 255
9.4 Unsteady Flow 256
9.5 Special Problems Associated with the Bubbly Flow Regime 260
Bubble Size 260
Agglomeration and Fracture of Bubbles 260
Bubble Growth and Collapse 261
9.6 Friction and Momentum Flux in Bubbly Flow 262
9.7 The Velocity of Sound in Bubbly Mixtures 264
9.8 The Limits of the Bubbly Flow Regime 265
9.9 Isotherma.l Homogeneous Flow of Gas-liquid Mixtures in Straight
Pipes 269
 CONTENTS
xiv

9.10 Isothermal Homogeneous Flow with Area Change Only 271

Use of the Equations of Motion for Both Components 274
9.11 Shock Waves 274
Problems 274
References 279

10 Slug flow 282

10.1 Introduction 282

10.2 General Theory 282
Bubble Dynamics 282
Bubble Velocity 283
Void Fraction 284
Pressurc Dro p 284
10.3 Vertical Slug Fiow 285
Rise Velocity of Single Bubb1es in Stagnant Liquid 285
Inertia Dominant 285
Viscosity Dominant 287
Surface Tension Dominant 287
The General Case 288
Use of the Bubble Velocity in the Drift Flux l\lodel 291
Improvements to the Simplified Theory 292
Correction for Long Buhbles 294
Viscosity Effects 299
10.4 Horizontal Slug Flow 299
Bubble Velocity 299
Void Fraction 302
Pressure Drop 302
10.5 Slug Flow in Inclined Pipes 304
10.6 The Limits of the Slug-flow Regime 307
10.7 P.ressure Oscillations in Slug Flow 312
Problems 313
References 314

11 Annular IFlow 315

11.1 Introduction 315
11.2 Horizontal Flow 316
The Boundaries of the Annular and Stratified Regimcs in Horizontal
Flow 316'
Correlations 317
Separated Flow, Annular Geometry Model 317
The Interfacial Shear Stress 318
The Wall Shear Stress 323
Evaluation of Pressure Drop and Void Fraction 324
Extension to the General Case 326
Improvements to the Theory 326
Integral Analysis 326
The Liquid Film 326
CONTENTS XV

The Gas Core 329

Differential Analysis 330
11.3 Countercurrent Vertical Annular Flow 380
FallÍng Film Flow 331
Stability of Falling Films 335
11.4 Flooding 336
Empirical Flooding Corrclations 336
TUTbulent Flow in Both Components 336
Viscous Flow in the Liquid 339
Prediction of Flooding from the Separate Cylinders Model 343
Turbulent Flow 343
Viscous Flow in the Liquid 344
11..'.J Vertical Upwards Cocurrent Annular Flow 345
The Boundaries of the Vertical Annular Flow Regime 345
The Slug-annular Transition 345
Criteria for Upwards or Downwards Flow in a Liquid Film 346
"Bridging" of the Gas Core 347
"Entrainment" Measurements Using a Sampling Probe 347
Comparison of Void Fraction Data with Theory 348
Prcssure-drop Measurernents 360
Discussion 351
The Annular-mist Transition 351
Correlations for Predicting Void Fraction and Prcssure Drop 351
The Dartmouth Corrclation for Void Fraction 351
The Modified 1\lartinelli Correla.tion 353
Simple Flow Models 354
Separate-cylinders Model 354
Homogeneous Model 355
Separated-flow Model 355
lmprovcments to the Separated-flow J\fodel 360
The Liquid Film 360
The Gas Core 363
Thc Gas-liquid Interface 366
Additional Effects 367
Problems 368
References 372

12 Drop Fiow 375

12.1 Introduction 375
12.2 Single-drop Formation 376
12.3 Atomization 376
12.4 Drop Size Spectra 378
12.5 The Terminal Velocity of Single Drops in a Gravitational Field 381
12.6 One-dimensional Vertical Flow without Wall Friction 382
12.7 Flooding in Drop Flow 382
12.8 Drop Fluidization 384
12.9 Pressure Drop in Forced Convection 385
xvi CONTENTS

12.10 Entrainment 386

Qualitative Observations 386
The Effect of Inlet Conditions and Tube· Length 388
Definition of a Critical Gas Velocity 388
Prediction of the Critica! Gas Velocity 890
Droplet Concentration and Velocity Distributions 391
Problems 393
References 396

Appendix 398
lndex 403
 st of Symbols

A arca; a constant or coefücient

a amplification factor in Eq. (6.124); a constant or coefficient; acceleration ·
E body force field; damping factor; a constant or coefficient; correction factor
b body force per unit volumc; breadth; index; constant
e a constant; cocfücient of apparcnt mass
CD drag coefficient
e, friction factor
e, defined in Eq-~ (4.21)
e,, e, correction factors
e velocity of dynamic wave relative to average or ·weighted average velocity;
velocity of sound; specific heat; coefficient of consolidation
molal concentration of non-condensable gas
'"e, vclocity of compressibility wave
C 0 ¡,., Ce, wave velocities in homogcneous and stratifred flow
e, specific heat at constant pressure
specific heat at constant volume
'"D diameter of pipe
D; bubble diameter
D, drop diarneter
Do orifice diameter
xvii
xvili LIST OF SYMBOLS

energy dissipation per unit volume

"'
d particle diameter
E defined by Eq. (8.80); fraction or percent entrainment
e base of natural logarithms; thermodynamic energy per unit mass
F force per unit volume of entirc flow field
F force; function defined in Eq. (11.25)
f force per unit volume of component
G mass flux
g acceleration due to gravity
H height; sign of shear stress
h enthalpy; heat transfer coefficient; height
J volumctric flux
]12 or in drift flux
K a constant; Bankoff parameter [Eq. (5.49)]; velocity ratio [Eq. (8.125)]
k a constant; thermal conductivity
k, roughness
L length; fractional thermal lag [Eq. (8.126)]
l mixing length
ln natural logarithm
M Mach number; mass exchange per unit volume per unit time
m mass flux in boiling or condensation; index; mass flow rate ratio; fraction
of cylindrical bubble cross section occupicd by liquid; a cocfficient
N dimensionless in verse viscosity; index
n index in correlations; velocity ratio
p pressure
!1p pressure drop
p perimeter
Q volumetric flow rate
q volumetric flow rate per unit width
q, heat transfer rate
R radius, radius of curvature; gas constant
radius
vector coordinatc of a point in space
pipe or duct radius
parameter defined in Eq. (3.37)
s source of mass
s entropy
T tem perature
11T temperature difference
t time
u wave velocity
u wave velocity relative to V 0 ; local velocity
u' friction velocity
characteristic transverse velocity in the Reynolds flux model
component velocities
unperturbed component velocities
average velocity; single-phase velocity
relative velocity
drift velocity of component 1 relative to the volumetric average
average single-phase velocity
weighted average velocity [Eq. (6.75)]
UST OF SYMBOLS xix

Vw continuity wave velocity

>w continuity wave velocity relative to average or ,veighted average vclocity
V, shock-wave velocity
V specific volume
'l) volume
w mass flow rate
w work
w, shaft work
w, shear work
X generalized sedimentation coordinate [Eq. (8.68)]; J\,fartinelli parameter
[Eq. (3.31)]
X quality [Eq. (1.7)]
y generalizcd sedimentation coordinate
y spatial coordinate; distan ce from wall
z coordinate in direction of motion
void fraction; volumetric fraction of component 2
"~ contact angle (degrees); shock angle; source of matter per unit length;
dimensionless pressure drop
~ isentropic exponent
film thickncss; volumetric flow rate ratio
' volumetric fraction of component 1; Reynolds flux
r fraction of dissipated energy transferred to component 2
limiting viscosity at high rates of shear; fraction of force due to phase
" change acting on component 2
o angle to vertical; wedge angle
),. ,vavelcngth; parameter defined by Eq. (10.57)
µ viscosity
kinematic viscosity, µ/p
'
~ natural constant
p density
surface tension
"
T shear stress
f heat flux; function; velocity potential
f¡, f, Martinelli parameters [Eqs. (3.24) and (3.25)]
<> shape factor
w frequency
Q correction factor (Fig. 3.13); reaction frequency [Eq. (2.92)]
f a function defined by Eq. (8.18)
r mass flow per unit width

Subscripts

1 component 1; state 1; location 1

2 component 2; state 2; location 2
a air
A acceleration
b bubble
e continuous phase; curvature; core in annular :fl.ow
d drop; discontinuous phase
e exit; effective value of
f liquid or fluid
XX UST OF SYMBOLS

F friction
g gas
G gravitational
i component i; interface; inlet; inside; inflexion; quantum state
j relative to or pertaining to average volumetric flux j
m mixcd mean in homogeneous fiow; val u e at tube e en ter in single-phase flow
n,N normal
o initial or boundary value; orífice; outside; zero quality; zero flow rate; stagna-
tion; relative to stationary fluid or particles
p pipe; particle
s salid; slug; shock; entropy
tangential; turbulent
w wave; wall
y yield
00 single particle; drop or bubble in an infinite medium; limiting value
TP two phase
when the concentration of the continuous phase is E
0 at anglc 0 to vertical

Note: Subscripts in Chap. 6 denote partial clifferentiation while subscripts outside

brackets enclosing partial deriva ti ves denote variables which ar-e kept constant

accel acceleration
crit critica!
mf minimum fluidization
frict friction
max maximum
min minimum

Dimensioniess Groups

Fr Fraude number
M Mach number
Nu Nusselt number
Pr Prandtl number
Re Reynolds number
We Weber number
Dp)i[Dg(p1 - p2)]½
N;

NA,

y gµ'
Property group: -
•'P
UST OF SYMBOLS xxi

[gD(p¡ - p,)]'"
32j¡µ¡
D2g(p¡ - p,;)
-(dp/dz) - p 0 g cos 0
g(p¡ - pg)
j,µ, (Po_)¼
u PI

(;)¼
~refixes

d differential
a partial differential
small change
Ll negative increment
V gradient operator
V divergen ce
dr element of volume at r
dA element of area

Superscripts
, dimensionless form of; at point where the 1V1ach numher is unity
modified; perturbed; in new frame of reference
+ dimensionless form of velocity profiles

Abbreviations

Btu British thermal unit

cfm cubic feet per minute
cm centimeter
fps feet per second
ft feet
hr hour
lb pound
mm millimeter
psia pounds per squarn inch absolute
ºF degrees Fahrenheit

Some Uncommon Abbreviations Used in IReferences

AEC U .S. Atomic Energy Commission
AEEW Atomic Energy Establishment, \Vinfrith, UKAEA
AERE Atomic Energy Research Establishment, UKAEA
ANL Argonne N ational Labora.tory
CISE Centro Informazioni Studi Esperienze, 1Y1ilan, Italy
EURAEC Euratom-Atomic Energy Commission Joint Program
NYO New York Operations Officc of U.S. Atomic Enorgy Commission
UKAEA United Kingdom Atomic Energy Authority
One-dimensional Two-phase Flow
one

Analytical Techniques
 1
lntroduction

U WHAT IS TWO-l'HASE Fl!JW?

A phase is simply one of the states of matter and can be either a gas, a
liquid, or a salid. M ultiphase fiow is the simultaneous flow of severa!
phases. Two-phase fiow is the simplest case of multiphase flow.
The term two-component is sometimes used to describe flows in which
the phases do not consist of the same chemical substance. For example,
steam-water flows are two-phase, while air-water flows are two-component.
Sorne two-component flows (mostly liquid-liquid) consist of a single-phase
but are often called two-phase flows in which the phases are identified
as the continuous or discontinuous components.
Since the mathematics which describe two-phase or two-component
flows are identical, it does not really matter which definitions are chosen.
The two expressions will therefore be treated as synonyms in most
developments in this book.
There are many common examples of two-phase flows. Sorne, such
as fog, smog, smoke, rain, clouds, snow, icebergs, quicksands, dust storms,

'
• ONE:.DIMENSIONAL TWO-PHASE FLOW

and mud, occur in nature. Others 1 such as boi1ing 1vater1 tea making,
egg scrambling, salad tossing, j am spreading, cream whipping1 sugar
stirring, and sphaghetti twirling, are frequent occurrences in kitchens
and dining rooms.
Several everyday processes involve a sequence of different two-phase
flow configurations or flow patterns. In a coffee percolator, for example,
the water is first boiled to form steam bubbles, alternate slugs of liquid
and vapor then rise through the central tube, and the hot water percolates
through the coffee grounds and eventually drips down into the pat.
When beer is poured from a bottle, the rate of discharge is limited by the
rise velocity of slug-fiow bubbles in the neck; subsequently bubbles
nucleating from defects in the walls of the glass rise to form a pleasing
foam at the surface. Bread and cakes begin with a multiphase mixing
process, are cooked with the release of bubbles, except when the appro-
priate ingredient is forgotten, and are eventually consumed orally in one
of the most common multiphase phenomena of ali.
The subtle blend of flavor, texture, and temperature that is achieved
in a perfect IVIartini is the result of ski!lful control of a two-phase chemical-
eng1neenng process.
Biological systems contain very few pure liquids. Body fluids, such
as blood, semen, and milk, are all multiphaseJ containing a variety of cells,
particles, or droplets in suspension. Their behavior can be described
by much the same equations as are used for analyzing paints, inks, pastes,
and nuclear fuel slurries.
A more technical example can be taken from the familiar area of
fire prevention and control. Almost without exception the various
methods of fire extinguishing are all multiphase processes 1 involving
sprays, jets, foams, or powders. Even the extinguishers which use pure
gas cannot be analyzed without considering the flash evaporation which
occurs as the material is expelled from the high-pressure storage cylinder.
l\.1oreover, the fires themselves usually result from a reaction between
solid or Iiquid fuels and oxygen in the air, produce smoke and steam,
which are invisible unless they are two-phase, and cause death by irritat-
ing the nose and throat until the victim drowns in his own multiphase
secretions. Deliberate fires in boilers, automobile engines, and rockets
are designed to burn two-phase dispersions.
Examples are equally profuse in the industrial field. Over hall of
all chemical engineering is concerned with multiphase flows. l\.1any
industrial processes such as power generation, refrigerationJ and distilla-
tion depend on evaporation and condensation cycles. The performance
of desalination plants is limited by the "state of the art" in two-phase
technology. Steelmaking, paper manufacturing, and food processing ali
conta.in critica! steps which depend on the proper functioning of multi-
 JNTRODUCTION 5

phase devices. lHany problems of air and water pollution are due to
unwanted two-phase flows.

L2 METHODS OF ANALYSIS

Two-phase flows obey al! of the basic laws of fluid mechanics. The
equations are merely more complicated or more numerous than those of
single-phase flows.
The techniques for analyzing one-dimensional flows fall into several
classes which can conveniently be arranged in ascending arder of sophisti-
cationi depending on the amount of information which is needed to
describe the flow, as shown in the following paragraphs.

COITT:RELATIONS

Gorrelation of experimental data in terms Of chosen variables is a con-

venient way of obtaining design equations with a minimum of analytical
work. The crudest correlations are mere mathematical exercises, readily
performed with modern computers, while more advanced techniques use
dimensional analysis or a grouping of several variables together on a
logical basis.
A virtue of correlations is that they are easy to use. As long as
they are applied to situations similar to those that were used to obtain
the original data, they can be quite satisfactory, within statistical limits
-vvhich are usually known. However, they can be quite misleading if used
indiscriminately in a variety of applications. Furthermore 1 since little
insight into the basic phenomena is achieved by data correlation, no
indication is given of ways in which performance can be improved or
accuracy of prediction increased.
In general, correlations will be avoided in this book unless they
possess a viable claim to generality. Those which are quoted will be
dimensionless and have sorne theoretical basis or have been tested against
a variety of data.

5EMPLE ANALYTJCAl MODELS

Very simple analytical models which take no account of the details of the
flow can be quite successful, both for organizing experimental results and
for predicting design parameters. For example, in the homogeneous model
the components are treated as a pseudofluid with average properties,
without bothering with a detailed description of the flow pattern. A
suspension of droplets in a gas, a foam, or the stratified flow of a gas
over a liquid are al! treated exactly alike. In the separated-fiow model
the phases are assumed to flow side by side. Separate equations are
written for each phase and the interaction between the phases is also
6 ONE-DIMENSJONAL TWO-PHASE FLOW

considered. In the drift-j/.ux model attention is focused on the relative

motion.
Each of these simple models is accorded an entire chapter in Part 1
and is used extensively throughout Part 2 of this book.
INTEGRAL ANALYS!S

A one-dimensional integral analysis starts from the assumption of the

form of certain functions which deE¡cribe, for example, the velocity or
concentration distributions in. a duct. These functions are then made to
satisfy appropriate boundary conditions and the basic fluid-mechanics
equations in integral form. Similar techniques are quite commonly used
for analyzing single-phase boundary layers.
OiFFERENTIAL ANALYSIS

In a differential analysis the velocity and concentration fields are deduced

from suitable differential equations. Usually, following the one-dimen-
sional flow idealization, the equations are written for time-averaged
quantities, as in single-phase theories of turbulence. fdore sophisticated
versions of the theory may even consider temporal variations.
Efforts will be made throughout the book to show how the various
levels of analysis are related. U sually the more complex theories lead
to the inclusion of additional effects and the pre::liction of numerical values
of correction Jactors, which can be applled to the simpler theories in arder
to increase their accuracy. The complex theory may also lead to an
analytical rather than empírica! relationship between the important
variables. Thus these sequential levels of analysis resemble a pyramid
in which the broader and more general theories serve to -support the more
approximate and simpler techniques.
UNIVERSAL PHIENOMENA

In addition to this hierarchy of analytical methods, there is a class of

very powerful techniques based on universal phenomena that are inde-
pendent of the flow regime, the analytical model, or the particular system.
Typical of these methods are the various theories of wave motion and
extremum techniques for obtaining the locus of the limiting behavior of a
system. These ideas thread their way through the various chapters and
help to bind them together into a uniform conceptual framework.
Understanding of these concepts should develop as each new applica-
tion is described.

1.3 FLOW REGIMES

The price that is paid for a greater accuracy in prediction of results is an

increase in complexity. In two-phase flow the amount of knowledge
!NTRODUCT/ON
'
which is needed in arder to perform a detailed analysis is often surpris-
ingly great. For example, in studying the motion of a single gas bubble
rising in a stagnant liquid, one is concerned with all of the following·effects:

Inertia of the gas and the liquid

Viscosity of the gas and the liquid
Density difference and buoyancy
Surface tension and surface contamination

The last item above is itself extremely complicated since 11 contamination"

can take the form of dirt, dissolved matter, or surface-active agents.
Heat transfer and mass transfer to the bubble also alter its motion.
Perhaps the first step in rendering this hydralike problem tractable
is to break it up into various regimes which are each governed by certain
dominant geometrical or dynamic parameters.
Part of the definition of the flow regime is a description of the
morphological arrangement of the components, or flow pattern. The fiow
pattern is often obvious from visual or photographic observations but is
not adequate to define the regime completely because of additional dis-
tinguishing criteria, such as the difference between laminar and turbulent
flow or the relative importance of various forces. In arder to keep the
terminology manageable, the numerous imaginative expressions which
have been used throughout the literature to describe flow patterns will
not be quoted. It is far simpler to restrict classifications to the mor-
phological flow patterns (far example, bubbly, slug, annular, and drop
flow in gas-liquid systems) and create further subdivisions into distinct
regimes ,vithin each of these classifications. Hybrid flow patterns,-
usually representing a region of transition from one pattern to another,
are denoted by hyphenated expressions (thus, slug-annular and annular-
drop flows). Sorne synonyms (e.g., 11 fog" or 11 mist" instead of "drop")
may be used when perfunctory repetition of a single word becomes
monotonous.
Asan example of the complexity of two-phase flows, Fig. 1.1 shows
a sequence of flow patterns occurring in an evaporator as more and more
liquid is converted to vapor. Obviously different parts of the evaporator
require different methods of analysis, and the problem of how one regime
develops from another has to be considered also.
Numerous authors have presented flow-pattern and flow-regime
maps in which various areas are indicated on a graph for which there are
two independent coordinates. For a given apparatus and specified com-
ponents this is readily done in terms of the flow rates, as shown in Figs.
1.2 and 1.3. However, since the flow reg1me is governed by about a
 ONE-DIMENSIONAL TWO-PHASE FLOW

Vapor

.4
...

Fig. 1.1 Approximate sequence of flow patterns ín a

vertical tube evaporator.

dozen variables, a two-dimensional plot is quite inadequate for general

representation. In this text the criteria for regime boundaries, when
they are known, will be discussed in detail as each regime is analyzed.
When these criteria are represented on a two-dimensional plot for par-
ticular applications, it will be found that the a.reas covered by a particular
regime change drastically both in size and shape, as variables such as
pressure and diameter are changed. For certain combinations of param-
eters, whole regimes disappear from the map altogether.
1NTRODUCT!ON
'
U NOTATION

Befare proceeding \Yith thc analysis of two-phase flow it will he necessary

to define sorne o! the relevant terminology. Although a detailed list o!
nomenclature js given at the front of this volume, it may require sorne
explanation. In addition, a certain familiarity with the simple relation-
ships among sorne of the parameters will enable the analysis to be under-
stood more rapidly.

SlMIPLE DEFJNJTIONS

The two components are distinguished by subscripts 1 and 2 in general,

or by subscripts J and g for a liquid-gas systcm or J and s for a fluid-solid
system. Component 2 is usually chosen to be the dispersed phase or the
lighter phase in a stratified flow.
The total mass rate al flow (in pounds per sccond) is represented by
the symbol W. The total flow is thc sum of the component flows.

Drop-annular

100 f - - - - - - - - - - - - + - - + - - - - - - + - - - - - - - - - - - - ,
t
Homogeneous

S!ug

Fig. 1.2 Flow-pattcrn boundaries for vertical upflow of air an<l water at
15 psia in a 1-in.-diam tube deduced from equations in the text.
10 ONE-DIMENSIONAL TWO-PHASE FLOW

□ = Plug and churn flow

• = Intermediote region
Annulor flow
o = Nonwetting
., = No dísturbonce waves
v = Pulses
400
.a. = Disturban ce waves with
small ripple waves
• t.. = Disturbance waves with
large ripple waves

V
V

o V
•
•
o
V
V V
•
V

V
• • •
V
• •
V
•• • Disturbonce waves
and ripple woves
•
V
• • •
• • • • •
•
t,,. a t:.. t.
•
• "
• ---,,_-<,_~~
100 ¡...:•:...,.~-=•:...,.~•-·....:•;___
Intermediote region

Plug and churn flow

º~----~------'----------'
O 100 200 300
Water flow rote, lb/hr

IFig. 1.3 Various 11regimes" or subdivisions of the annular

flow pattern for cocurrent upward flow of air and water in a
l}i-in.-diam pipe at 15 psia. (Hall-Taylor and Hewitt. 1)

Therefore
W - W1 + W, (1.1)
The volumetric rate of flow (in cubic feet per second) is represented
by the symbol Q. The following relationships are obvious:
Q - Q¡ + Q, (1.2)

Q¡
W1 (1.3)
p¡

Q, w, (1.4)
p,
fNTRODUCTION 11

Every part of the flow field is occupied by one or other component.

If a represents the fraction of an element of volume which is occupied
at a.ny instant by component 2, then 1 evidently, if the element is chosen
small enough, a can only be O or l. However, for most purposes a volume
much larger than the discrete particles (drops or bubbles) is cbosen, and"
then represents an average volumetric concentration. Usually a is
measured as an average over the whole flow cross section and a sufficient
length o! duct to eliminate local fluctuations. Thus, if a pipe al length
L and cross-sectional area A is suddenly isolated by closing valves at
both ends, the contents can be analyzed and the total volume 'ü2 of com-
ponent 2 in the pipe can be determined. The average value o! " is then

(1.5)

Often it is not possible to measdre (a) over a long length of pipe

beca.use the flow is not uniform. In this ca.se a large number of instan-
taneous readings over a length óL give the time average of a at a given
location. The average value of a both in space and time is then
( ) ~ JJ a(r,t) dr dt
(1.6)
ª Jdrfdt
U sually the symbol a is used loosely to represent an average volumetric
concentration without bothering to define exactly how the average is to
be taken. Thus, extra care is needed when periodic phenomena and
nonuniform concentrations are importan t. In gas-liquid flows, a usually
represents the void Jraction, or volumetric concentration of the gas.
It is often convenient, particularly in boiling or condensing applica-
tions, to have a measure of the fraction of the total mass flow across a·
given area which is composed of each component. The quality is there-
fore defined as
W,
X = w (1. 7)

Evidently x is subject to averaging laws when flow is unsteady or non-

uniform. The average is to be taken over a specified surface and for a
period of time. Therefore

() = JG,dAdt (1.8)
x JG dA dt

The symbolj is used to represent volumetricflux (in feet per second)

or volumetric flow rate per unit area. The flux is really a vector quantity
but at the leve! o! sophistication which is appropriate at present, j will
be used exclusively to represent the scalar component in the direction
of motion along a pipe or duct. The flux is related to the local compo-
12 ONE-OIMENSIONAL TWO-PHASE FLOW

nent concentration and velocity as follows:

j, = (1 - c,)v, (1.9)
J2 = Cl'V2 (1.10)

The total local flux is

j = j, + j, (1.11)
The following results are self-evident
Q, = Jj, dA (1.12)
Q,=Jj,dA (1.13)
The average volumetric flux across an area A is then

(J2.) -- A
Q, (1.14)

Unless variations across the flow are being considered, the brackets are
usually omitted from Eq. (1.14).
The mass flux is represented by the symbol G (in pounds per hour
per square foot).
Clearly, for a small element in which the density of each component
can be regarded as uniform:
G1 = Pd1 ( 1.15)
G2 = PÓ2 (1.16)
G = G, + G, (1.17)
The average mass flux of component 2 across an area A is

(G,) = Á 2
(1.18)

Although the most general description of two-componcnt flow

involves a consideration of the three-dimensional and temporal variation
of al! the above quantities, we shall usually be content with the one-
dimensional flow assumptions and work entirely in terms of averages
across the duct. Under certain circumstances, however, when large vari-
ations across the duct occur, the theory will prove inadequate and more
detailed analysis will be necessary.
Sorne useful relationships for one-dimensional flow are summarized
below.
j, = ~ (1.19)
. Q, (1.20)
J2 = A
. Q, + Q, (1.21)
J = A
iNTRODUCT!ON
"
J1
V¡ = 1- a (1.22)

V2 = J2
-a ( 1.23)

G - w,
1---::r (1.24)
W,
G, =--::r (1.25)
w, = Q,p, (1.26)
w, = Q,p, (1.27)
J1 Q1 V¡ 1- a
)2 = Q2 = v; -Ci-
(1.28)
G, w, 1 - x
(1.29)
G, = W, = -x-

From Eqs. (1.26) through (1.29),

1- X V¡ PI 1- a
(1.30)

The relative velocity is defined as

V21 = (v2 - V¡) = -V12 (1.31)
Drift velocities are defined as the difference between the component
velocities and the average as follows:
Vlj = V¡ - J (1.32)
V2j = V2 - j (1.33)
The drift flux represents the volumetric flux of a component relative.
to a surface moving at the average velocity, i.e.,
j 21 = a(v, - j) (1.34)
j,, = (1 - a) (v 1 - j) (1.35)
Substituting Eq. (1.11) into Eq. (1.34) and using Eq. (1.10), we
obtain
j 21 = j 2 - a(j 1 + j 2) = j,(1 - a) - aj, (1.36)
Similarly,
j 12 = j 1 a - (1 - a)j, (1.37)

Therefore
(1.38)

This symmetry is an important and useful property of the drift flux.

14 ONE-DIMENSIONAL TWO-PHASE FLOW

Substituting for j, and j, in Eq. (1.37) by using Eqs. (1.22) and

(1.23), we get
(1.39)
Therefore the drift flux is proportional to the relative velocity.
Any system of units may be used in the above equations as long
as the requirements of consistency and compatibility are satisfied.
Velocities which are characteristic of the overall flow, such as wave
velocities, are represented by capital letters with appropriate subscripts.
For example, the continuity wave velocity is given the symbol Vw and
the shock-wave velocity the symbol V,.
PROPIERT!ES

To distinguish between the symbol for volume and the symbol for
velocity 1 the former is written in script form. Thus, the volume of a
bubble is V, and the specific volume of gas is v,. Care should be taken
in distinguishing these symbols; for example VJo is the liquid velocity
relative to the gas, whereas v10 represents the change in speci:fic volume
on vaporization. The subscript convention is reversed in the case of
thermodynamic properties; thus v 10 = v 0 - v 1 and h!G = ha - h¡. t
There should be no difficulty in identifying the common symbols
for properties such as p for density and µ, for viscosity. Surface tension
is not so widely used and does not have a standard symbol; it will be
represented by cr with the units of force per unit length. Both enthalpy
and heat-transfer coefficient have the symbol h but the context should
make the distinction clear.
PRESSURE DROP

The symbol for pressure drop in a pipe is t:.p. On the other hand dp/dz
represents the rate at which the pressure increases with distance in the z
direction. Therefore if z denotes the coordinate clown the axis of a pipe
measured in the direction of flow, the pressure drop over a length L
will be
t:,p = - [L dp dz (1.40)
)o dz
COORDINA.TES.

Since the symbol x has already been chosen to represent the quality, it
will be avoided for use in deseribing the coordinate system. Usually z
will be the coordinate measured in the direction of flow and y the coordi-
nate measured from a boundary 1 such as a wa.11. The radial distance
from a pipe axis will be denoted by r.
t In hand work it is often convenient to use the density rather than specific volume
as the basic variable in arder to avoid possible confusion between the v's and the v's.
lNTRODUCTION 15

ur-nns
All equations are written in a consistent dimensional form and are suitable
for use with any convenient system of units. Superfluous factors which
represent the ratio between various conventional units have been omitted
entirely. Any student who has learned to copo with the plethora of units
(which h,ave evolved by historical accident) in elementary fluid mechan.ics
and thermodynamics courses should have developed sufficient maturity
to have no difficulty using the equations in this book.

PROBLEMS
:LL A bubbly mixture fl.ows in a 1-in.-diam pipe. The gas flow rate is 30 cfrn and the
bubble velocity is dctermined photographically to be 100 fps. What is the void
fraction? \Vhat is the liquid velocity if the liquid fl.ow rate is 5 cfm?
1.2. 300 lb/hr of air at 70ºF and 20 psia fl.ow together with 300 lb/hr of water in a
1.25-in.-diam pipe. What is the overall volumctric flux j? If the drift flux j 0 ¡ is
10 fps, what are the average velocitics of the phases?
1.3. A steam-water mixture with 1 % quality flovvs at atmospheric pressure in the
riser of a coffec pcrcolator. The void fraction is measured to be 80%. What is the
ratio of the average steam velocity to the average water velocity?
1.4. In a certain fluid-solid systern (a quicksand) the drift flux is related to the vol-
umetric concentration 1c of the fluid by the equation
Ús = fü 3 (1 - t) fps
If a flux j¡ = 1 fps of fluid flows upward through a stationary bcd of particles,
what is the value of €? If the particles are spheres, which havc a random packing
value of €o = 0.4, is the system "fluidized" or do the particlcs rcst on one another?
1.5. Express Gin terms of the quality, the individual phase vclocitics, and thc densities.
1.5. Express j in terms of the individual rnass flow rates, the pipe diameter, and thé
phase den.sities.
:u. Show that the drift flux is independent of the motion of an observer.
1.8. On a graph of j 2 versusj 1, for phases with given properties, show lines of constant j,
constant G, and constant x. Can lines of constant a be drawn? Why not?
:U.t On thc graph of )2 v:.ersus ji, show lines of constant a, if
(a) ::'.! = const
v,
(b) v12 = const
(e) ]12 = const
(d) fo - ka(I - a)•
1J.O. Express the momentum flux in one-dimensional flow in terms of G, x, a, and the
densities of the phases. For what value of a will the momcntum flux be a minimum
if G and x are constant? For what values of a will G be a maximum if x and the
momentum flux are fixed?
1.11. Solve Prob. 1.10 using the kinetic-energy flux instead of the momentum flux.
1.12. Show that for incompressible flow in a constant-area duct, j is constant with
position. 1 although the individual fluxes may vary.
 ONE-DIMENS!ONAL TWO-PHASE FLOW
"
1.13. Show that the drift flux is zero when a = O or a = l.
1.14. Prove that j2v12 = v2V1J-
1.15. A glass is filled to the brim with draft beer. After it has stood for a while until
bubble activity ccases, the glass is observcd to be 70% full. What was the original
void fraction? Estimate the quality of the mixture issuing from the tap.
1.16. Estímate the maximum value of a in a stable foam which was made from bubbles
1 mm in diameter and has drained for a very long time until thc liquid filaments are
a few moleculcs thick.
1.17. What value of a corresponds to a close-packed array of spheres?
1.18. Water at 100 psia enters a straight evaporator tube. If the velocity ratio
Vu/Vf is constant at the valuc 2.5, and the mass flux is 2 X 10 5 lb/(hr)(ft2), what are
the values of void fraction and momentum flux when x = O, 0.1, and 0.5?
1.19. In u particular vertical flow regime the relative velocity is constant and equal to
Vo. Draw the lines of constant a on a graph of j 1 versus j2. Show that the lines
envelop a curve in the quadrant whichrepresents countercurrent fiow. This envelope
is the "fl.ooding linc" outside which operation is impossible. Show that the equation
of the fl.ooding line is

REFERENCE

l. Hall-Taylor, N., and G. F. Hewitt: AERE-R3952, UKAEA, 1962.

 2
Homogeneous Flow

2,1 INTIIODUCTION

Homogeneous flow theory provides the simplest technique for analyzing

two-phase (or multiphase) flows, Suitable average properties are deter-
mined and the mixture is treated as a pseudofluid that obeys the usual
equations of single-component flow, Ali of the standard methods of fluid
mechanics can then be applied,
The average J)roperties which are required are velocity, thermody-
namic properties (e.g., temperature and density) 1 and transport properties
(e,g,, viscosity), These pseudo properties are weighted averages and are
not necessarily the same as the properties of either phase, The method
of determining suitable properties is often to start with more complex
equations and to rearrange them until they resemble equivalent equations
of single-phase flow. For example, the virtual viscosity of an emulsion
can be obtained by analyzing the three-dimensional flow field in the two
components. In another case, the apparent properties for a gas-particle
mixture are found from the separa.te equations for each component by
assuming a class of similar solutions (Chap, 8, p, 212),
17
" ONE-DJMENS!ONAL TWO-PHASE FLOW

Differences in velocity, temperature, and chemical potential between

the phases will promote mutual momentum and heat and mass transfer.
Often these processes proceed very rapidly, particularly when one phase
is finely dispersed in the other, and it can be assumed that equilibrium is
reached. In this case the average values of velocity, temperature, and
chemical potential are the same as the values for each component and we
have h01nogeneous equilibiiitm flow. The resulting equations are simple
and easy to use, but it is often advisable to check the validity of the
equilibrium assumptions by using the more accurate theories which will
be presented in later chapters or by performing a detailed analysis of other
transport processes which are beyond the scope of this book. For exam-
ple, rapid acceleration and pressure changes make the equilibrium theory
inaccurate for describing the discharge of flashing steam-water mixtures
through short nozzles or ori:fices. I t is necessary to consider the rates of
bubble nucleation and growth in the superheated liquid. N onequilibrium
effects also occur when supercooled vapor condenses in high-velocity steam
or when particles of salid propellant are bumed in the nozzle of a rocket
eng1ne.
In sorne cases the use of homogeneous theory is obviously inappro-
priate. For example, countercurrent vertical flows, which are driven by
gravity acting on the different densities of the phases, cannot be described
by a suitable "average" velocity.
This chapter will develop homogeneous flow theory using a sufficient
number of examples to illustrate the techniques. fviore specific and
detailed applications will be found in later chapters.

2.2 ONE-lllllllENSIONAl STEADY HOMOGENEOUS EQl!lllBRIIJM FLOW

The basic equations for steady one-dimensional homogeneous equilibrium

flowin a duct are:
Continuity w= PmVA = const (2.1)
dv dp
Niomentum w dz ~ -A dz - PTw - Apmg cose (2.2)

Energy dq, - dw
dz dz
~ W .'!:._·
dz
(h + 11_2
2
+ gz' ) (2.3)

In the above equations A and P represent the duct area and perim-
eter, Tw is the average wall shear stress, dqe/dz is the heat transfer per unit
length of duct, z0 is the vertical coordinate, and 0 is the inclination of the
duct to the vertical. W ork terms are assumed to be zero in the energy
equation in most cases. It is often possible to use the momentum and
energy equations in integral form when one is only interested in changes
between particular points in the duct.
HOMOGENEOUS FLOW 19

Equation (2.2) is often rewritten as an explicit equation for the pres-

sure gradient. Thus
dp P W dv (2.4)
dz -- - A Tw - A dz - Pmg cos 0

The three terms on the right side can then be regarded as frictional,
accelerational, and gravitational components of the pressure gradient.
Since engineers (being pessimists) are mostly interested in prcssure dropsi
the following definitions are usually adopted.

- (ªp)dz ,
= F'_ Tw
A
(2.5)

_ (dp)
dz A
= W dv
A dz
(2.6)

- (~nG = Pmg COS 0 (2.7)

The total pressure gradient is then the sum of the comp.onents, as

follows:

(2.8)

In addition to the above equations we usually have sorne knowl-

edge about the equation of state for tbe components. For a steam-water
mixture, for example, the steam tables or lVIollier chart can be employed.
Far a mixture of a gas and a salid, an equ.ivalent equation of statc can be
derived by assuming equilibrium at all times between the components or
by making other appropriate assumptions.
The mean density can be expressed in various ways. In terms ·of
the volume fraction a, it is

Pm = ap, + (1 - a)p1 (2.9)

whcreas in terms of. the quality or mass fraction specific volumes are addi-
tive. Therefore

~=-='+1-x (2.10)
Pm P2 Pt

The mass of each component per unit volume can be expressed in

terms of a or x to give the following equations:

XPm = ap2 (2.11)

(1 - x)pm = (1 - a)p1 (2.12)

For homogeneous steady f:low with velocity equilibrium, the void

 ONE-D!MENSIONAL TWO-PHASE FLOW
'º
fraction and quality are

(2.13)

(2.14)

An example of simple homogeneous flow can be taken from the case

of isentropic expansion of a steam-water mixture through a nozzle with
no frictiÓn or heat transfer at the walls. For expansion between two pres-
sures at constant entropy the final state can be determined on a Mollier
chart. Therefore the enthalpy change and final density can be deter-
mined. The exit velocity then follows from the integral of Eq. (2.3) and
the area from Eq. (2.1). The technique is exactly the same as the equiv-
alent technique that is used for predicting the expansion of dry steam.
Errors, however, are likely to be introduced because of nonequilibrium
effects. The water and steam are not in thermal equilibrium nor do they
have equal velocities in a rapid expansion. Friction and heat transfer
between the phases also introduce irreversibility which violates the con-
stant-entropy assumption.

Example 2.1 Dry saturated steam from a large container at 100 psia is expanded
through a frictionless adiabatic nozzle to a final pressure of 15 psia. What
are the exit velocity and flow per unit area?
Solution From steam tables, for isentropic expansion as above, we find that the exit
wetness is 10. 7 percent. Therefore the exit spccific volume and enthalpy are:
v, - (0.893)(26.32) + (0.107)(0.0167) - 23.5 ftS/lb
h, - 1150.7 - (0.107) (969.6) - 1047 Btu/lb

The inlet enthalpy is hi = 1187.3 Btu/lb.

Gravitational terms are negligible in Eq. (2.3), which reduces for adiabatic
flow to the equation

t - h, -
V '
h, - 140.3 Btu/lb

Therefore
v, - (2 X 140.3 X 32.2 X 778)'" - 2660 fps
The mass :fl.ow per unit area is derived from Eq. (2.1).
61
_: - ::-;~ ~ - n3lb/(ft')(sec)

Example 2.2 Formulate equations for the one-dimensional frictionless expansion of a

dilute gas containing a dispersion of small solid particles. Assume homogeneous
flow with uniform velocity, and consider the two limiting conditions which
bound the possible thermal behavior, namely,

l. No heat transfer between the gas and the particles

2. Thermal equilibrium at all times between the components
HOMOGENEOUS FLOW 21

Using these results, calculate the exit velocity, upstrearn stagnation

pressure, and exit gas temperature for expansion of a mixture of 2 lb of sand
[specifi.c heat 0.21 Btu/(lb)(ºF)] pcr pound of air to a lVIach numbcr of 2 at
15 psia. The upstream stagnation temperature is 1500ºF.
Solution The simplest method of solution is to derive the equation of state and
adiabatic expansion law for a pseudo gas consisting of thc sand and the air
combined.
Let 1' be thc air temperature and let all the air properties be denoted by
subscript a.
For 1 pound of the air alone the UBual equation of state is
(2.15)
If the particle density is very large compared with the air density and if the
two-phaSe mixture contains m pounds of particles per pound of air, the apparent
density (or specific volume) will be related to the air density (or specific volume)
by the approximate equations (see Prob. 2.1)

(2.16)

Using Eq. (2.16), Eq. (2.15) can be rewritten to give an equation of state for
the pseudo gas consisting of both air and particles, as follows:

R" 1' (2.17)

pv=l+m
Therefore the effect of adding the particles is simply to modify the appropriate
value of R in the perfect-gas equation.
The expansion law will depend u pon the mutual heat transfer. If there
is no heat transfer and no friction bebveen thc air and the particles (since they
are assumed to have the same velocity), thc air will expand isentropically
according to the law
P?-Jo.'Yª = const (2.18)
Since 1 +mis a constant in Eq. (2.16), Eq. (2.18) is equivalent to the relation
p?-J'Y" = const (2.19)
Therefore for the zero heat-transfer case the mixture behaves exactly as a
pseudo gas with the same value of the isentropic exponent )'a as for air alone,
but with a modificd value of R = Ra/(I + m).
On the other hand, if the gas and particles are always in thermal equilib-
rium, the total eritropy of them both taken together stays constant. If the
particles have a specific heat of e Btu/ (lb) (ªF), their entropy in crease accompany-
ing a transfer of heat dQ to the air is
dQ dT
dsp=-T=emT per lb of air (2.20)

The entropy in.crease for the air is

dQ dT dp
dsa = T = epa T - Ra p (2.21)

Combining Eqs. (2.20) and (2.21) we get

dT dp
dsa + dsp = T (epa + me) - Ra p = O (2.22)

"1
22 ONE-DIMENSIONAL TWO-PHASE FLDW

Therefore the pseudo gas obeys the expansion law

r-1pc"t- 1i1-r = const (2o23)
where
7 - 1
(2024)
~ Cpa + me
Solving Eq. (2.24) for 1' and using thc relation Ra = Cpa - Cva, we have
Cpa + me (2025)
7 -
Cva + me
This result could also have been deduced by realizing that the effective speci:fic
heats at constant pressure and constant volume are
Cpa + me and
e,..,.+ me
l+m l+m
The effect of mutual heat transfer is therefore to modify the isentropic
exponent to the value given by Eq. (2.25). The pseudo gas therefore behaves
as if it possessed a value of R = Ra/(l +
m) anda value of 1' given by Eq.
(2.25), but in other ways obeys all of the well-known one-dimensional flow
relationships of gas dynamics that can be taken from a standard text such as
Ref. l. Alternatively these results can be deduced from Eqs. (2.1), (2.2), and
(2.3), neglecting gravity and friction and assuming no externa! heat transfer.
For the numerical example we have to deal with a pseudo gas with a
value of
5 3
R - ~ - 17º8 (ft)(lb,)/(lbm)(°F) anda value of 7 - L4

for assumption 1 and

0241 + (2)(021)
for assumption 2
7 - 0º173 + (2) (0º21) - Ul
The stagnation pressure follows from the equation
7 - 1 )1'1(1'-l)
Po = P• ( l +- -
2
M.' (2026)

and the exit air temperature from the equation

T, - To (1 + 7; 1M;r' (2o27)

The exit velocity is

v.= M. ·,/rRT. (2028)
Putting in the appropriate numbers we get:

1. For no mutual heat transfer

Po = 117 psia T, - l090ºR - 630ºF
V, - 1870 fps
2. For equilibrium between the components
Po = 109.5 psia T, - 1605ºR - 1145ºF
V, - 2020 fps
HO!VlOGENEOUS FLOW
"
f!Jfl:THER D!EVELOPM!ENT OF THE MOMENTUM EQUATION

The momentum equation can be developed further by expressing the wall

shear forces in terms of a friction factor and a hydraulic mean diameter
D. The average wall shear stress is
(2.29)
The frictional pressure gradient is then

- (dp)
dz F
v'
= 2 CtPm D (2.30)

A convenient modification to Eq. (2.30) can be made by substitution in

terms of the volumetric and mass flow rates. Thus
. Q, + Q,
(2.31)
V= J = A

PmV = G = W, ÁW, (2.32)

Therefore
_ (dp) = 2C1 Gj (2.33)
dz F D
Alternatively we may cboose to work in terms of the specific vol-
umes of the components and the quality. From Eqs. (2.10) and (2.32)
we have
G
v = -
Pm
= G[xv 2 + (1 - x)v,] = G(v 1 + xv 12 ) (2.34)

Using Eqs. (2.31) and (2.34) in Eq. (2.30) then gives

- (dzdp) , 2C1G2
= ---y¡- (v, + xvn) (2.35)

Since the mass fl.ow rate is constant and each phase has the same velocity,
the accelerational pressure gradient in Eq. (2.6) becomes

(2.36)

Substituting for v from Eq. (2.1) in (2.36) gives

(2.37)

Expanding the differential we find

_(dp)
dzA-
_G' !!_dzpm
(_-1:_) _ G dA
PmAdz
2
_.!:_ (2.38)
24 ONE-DIMENSIONAL TWO-PHASE FLOW

Further, from differentiating Eq. (2.10),

-dzd(l)
-Pm -- -dx(l
- - -p11) + Xdzd(l)
dz P2
- -
P2
+ (J - X) -
dz
d(l)
-
Pi
(2.39)

or, in terms of the specific volumes of the phases,

-d ( -1 ) =
dz Pm
't/12-
dx
dz
+ X -dv2
dz
+ (1 - x) -dv1
dz
(2.40)

In the two-phase region (e.g., vapor-liquid) far apure substance, V¡ and

Va are only functions of pressure. For a two-component mixture, like-
wise, V1 and V2 can be expressed in terms of thc pressure as long as the
thermodynamic path can be determined. Equation (2.40) can then be
rewritten as

!}__
dz
(Je) = v
Pm
12
dx
dz
+ dp [x dv, + (l
dz dp
_ x) dv 1 ]
dp
(2.41)

The acceleration pressure drop, in terms of quality, flow rate 1 aúd prop-
erty variations, is now faund from Eq. (2.38), with the use of Eq. (2.41),
to be

_ (dp)
dz A
= G' ¡vi, dx
dz
+ dp
dz
[x dv, + (l
dp
_ x) dv 1 ]
dp
- (V1 + X't/12) -A1 -dA¡
dz
(2.42)

The gravitational pressure drop is faund in terms of quality by sub-

stituting far Pm from Eq. (2.10) in (2.7). The result is

- (dp) dz a
= g cos e - -1- -
V1 + xv12 (2.43)

Combining Eqs. (2.35), (2.42), and (2.43) in the farm of Eq. (2.8) and
rearranging, we eventually obtain an expression from which the pressure
gradient can be calculated as follows:
2C¡ dx 1 dA g cos e
dp -D G2 (vi + XV12) + G Vi, d-2
- G (Vi + XV12) -A -d + +
Z
2
Z V¡ XV12

- dz = 1 + G' [x dv, + (l _ x) dv,]

dp dp
(2.44)

This equation can be expressed in many other ways by making substitu-

tions in terms of other chosen variables. However, the form of the equa-
tion will be retained and the physical significance of each term will be
unchanged. In factJ any one-dimensional steady-flow analysis will even-
HOMOGENEOUS FLOW
"
tually be found to lead to an equation of the form
dx 1 dA
dp
e,+ c,dz + cAATz + c,g cose
- dz (2.45)
1 - M'

In this equation CF¡ Cx, CA, and C(J are infiuence coeffiáents, which express
the effect of friction, phase changeJ area change, and gravity on the pres-
sure gradient. The term M 2 in the denominator has the same signifi.cance
as the square of the lV[ach number in single component flow. Comparing
Eq. (2.44) with (2.45) and using Eq. (2.32) we can therefore deduce that
the velocity of a compressibility wave in a homogeneous two-phase mix-
ture is
e = (- Pm [X!; + (1 -
2
x) : ] ¡-¡; (2.46)

This result will be deduced more directly in Chap. 6. Using Eqs. (2.9),
(2.11), and (2.12) an alternative expression in terms o! a is found to be

e = \ [ap, + (1 - a)p1] [ ap 2 ( - !;)+ (1 - a)p1 ( - :,1)] ¡-¡;

(2.47)
The pseudo velocities al sound in the components taken alone and
following the thermodynamic path consistent with the two-phase flow may
be defined as
1
C12 = dp¡ =
dp Pl-2
( - · dv
dpl )- (2.48)

1_dp,
e, - -dp -_ P2
_, (-dv,)-1
--
dp
(2.4Q)

Using these definitions, Eq. (2.47) can be rearranged to give

(2.fiü)

In cases where p1c1 2 >> P2C2 2 and P1 >> p2 (e.g., an air-water mixture at
atmospheric pressure) this equation reduces to the approximate expression
c2 = !!__'!_ C22
(2.51)
p 1 a(l - a)
Evidently the velocity of sound in a homogeneous mixture can be
far less than the velocity of sound in the gas alone. A minimum occurs
at a = 7'2 and for air and water at atmospheric pressure, this gives a sonic
velocity of about 70 fps.
Usually the rate of change of quality is calculated from the energy
equation by equating heat transfer to latent heat changes. However 1 if
 ONE-DIMENSIONAL TWO-PHASE FLOW
"
significant "flashing"t occurs because of pressure changes, the quality is
not only a function of enthalpy, and a more correct way to proceed is as
follows. Let the quality be a function of both enthalpy and pressure.
Then for a given thermodynamic system we have
x ~ x(h,p) (2.52)
WhenceJ by differentiation,

dx _ (ªx) dh
dz - ah p dz
+ (ªx)
ap h
dp
dz
(2.53)

Moreover
1
(2.54)
h12

Equation (2.44) now becomes

-2C1 G'(V1 + ) + G' dh

V12 - - G'( V1 + )l dA + g cos 0
D
XV12 -
h 12dz
XV12 -~
A dz v1 + xv 12

1 + G [x av,
2
ap
+ (l - x) av,
ap
+ v (ªx) ]
1,
ap,
(2.55)

Many alternative forms of Eq. (2.55) could have been obtained by

using different thermodynamic properties in Eq. (2.52).
In practice the condition far a :VIach number of unity is more likely
to be given by Eq. (2.46) than by the equivalent result which would be
obtained by using the whole denominator of Eq. (2.55). This is because
of nonequilibrium effects which preclude rapid phase change caused by
sudden pressure changes.
If kinetic-energy effects are appreciable, the enthalpy gradient can-
not be calculated directly from the energy equation, and a more complex
form of Eq. (2.55) results (see Prob. 2.25).
N onequilibrium phase-change effects can be analyzed by assuming
an effective thermodynamic path to replace Eq. (2.52). This might con-
sist of only allowing a given fraction of the equilibrium quantity of vapor
to be formed, or of a more thorough analysis in which the actual quality
is related to nonequilibrium heat- and mass-transfer processes. 14

2.3 THE HOMOGENEOUS FRICTION FACTOR

LAMINAR FLOW

Many methods have been proposed for evaluating the two-phase homo-
geneous friction factor, C1, which is the only empirical parameter in Eqs.
t The term jlashing is usually used to describe vapor formation caused by pressure
changesi whereas boiling refers to vapor formation as a result of heat addition.
HOMOGENEOUS FLOW
"
(2A4) and (2.55). In laminar flow the simplest technique is to find a
suitable "virtual viscosity" for the mixture. For example, a theoretical
solution for a suspension of fluid spheres at low concentrations is

(2.56)

where the subscript 1 refers to the continuous phase. If the suspension

consists of salid particles, µ, is very large and Eq. (2.56) becomes Ein-
stein1s equation 3
µ = µ1(l + Z.5a) (2.57)
If, on the other hand, the emulsion consists of bubbles containing gas of
a low viscosity, the result is
(2.58)
Unfortunately, Eqs. (2.56), (2.57), and (2.58) are only valid at concen-
trations below about 5 percent for which the change in viscosity is small.
Numerous rheological models for taking account of larger values of a and
particles of various shapes and sizes have been proposed and will be dis-
cussed in later chapters. :i.víany two-phase mixtures are nonnewtonian.
Often the details of the two-phase flow pattern are not known and
an idealized rheological model cannot be defined. Faced with the neces-
sity of choosing sorne expression for the visoosity, many workers have
chosen averages which fit the limiting cases in ,vhich either phase is not
present. Sorne common expressions for gas-liquid flow are

!µ = -"'-
µ(}
+ 1 µ¡- x McAdams' (2.59)
µ = xµ, + (1 - x)µ¡ Cicchitti 5 (2.60)
Dukler' (2.61)

It is often convenient to relate the viscosity, friction factor, and

friction pressure. drqp in two-phase flow to the equivalent values for
single-phase flow of one of the phases alone. Far example, from Eq.
(2.59) in laminar flow

-µ =
µ¡
[ µ¡
x-
µ(}
+ (1 - a) ]-1 (2.62)

Far flows with change of phase, if the subscript fo is used to denote the
case in which liquid flows in the same pipe with the same mass velocity
as the combined flows, we have, using Eq. (2.62) for laminar flow
28 ONE-D!MENSIONAL TWO-PHASE FLOW

The ratios between the frictional pressure gradient for the two-phase flow
and the frictional pressnre gradients for related single-phase flows are
usually known as two-phase multipliers and are denoted hy the symbol
ef., 2 with appropriate subscripts, for example,

-(dp/dz)r
(2.64)
</,¡o' = -(dp/dz)F10
If q,1 ,' can be determined, then the first term in the numerator of Eq.
(2.44) or (2.55) can be replaced by

(2.65)

TURBULENT FLOW

Single-phase friction factors in tnrbulent flow are usually correlated in

terms of the Reynolds number and the pipe roughness. Except in
extreme cases the actua] value <loes not differ by more than a factor of
2 from the rule-of-thumb estímate of C1 = 0.005. In commercial situa-
tions where pipes are subject to corrosion, distortion, and scaling, the
accuracy with which pressure drops can be computed in single-phase flow
is often no better than 25 percent and it would be presumptuous to expect
a better correlation in the case of two-phase flow.
The three common alternatives for estimating two-phase turbulent
friction factors are

l. Use a constant value for all conditions. A good choice is

C1 = 0.005 (2.66)
Figure 2.1 shows a comparison bet,veen this prediction and sorne
data taken in the high-velocity annular-mist flow regime. Under
these conditions a considerable fraction of the liquid flow is entrained
in the form of droplets, and homogeneous theory provides a reason-
able approximation.
2. Use a friction factor ca.lculated from sorne equivalcnt single-phase
flow. For example, for low-quality vapor-liquid mixtures it may
be assumed that the friction factor is the same as it would be if the
total mass flow (liquid plus vapor) flowed entirely as liquid. The
appropriate Reynolds number is

Re1 =
GD
~ (2.67)
µ¡
and it is readily shown that

</,¡o 2 = 1+X(;: - 1) (2.68)

HOMOGENEOUS FLOW
"
1 1 1 1 1

•
V
1
º'eov io •
A
A
A
• 1 •
D ◊
0
Dffl ~
◊
ª"1 oa
◊
"' A
•
A
A
0
A o
0.0050- 00
El O
• • □a
G i V ♦

.8V
V V o
•o •
-E 1 V
e Meyer and Turner
o
·~ 11
0.548 I.D. copper tube Gas mass flow rotes, slug / sec
-~
~ p=l otmos Meyer
0.0025 -
Cocurrent □ ir /water upflow il 0.266 X 10- 3 O 0.95] X 10- 3 -
A 0.384 X " a 1.007x
Gas mass flow rotes, slug/sec
h. 0.50] X " 111 1.324 X
Turner
□ 0.607 X V 1.640 X
1 0.130x10- 3
◊ 0.200 X
• 0.726 X
V 0.286 X

º~---~'---~'---~'----'~---~'---~
'--~
O 1 2 3
Wf a: liquid mass flow rote, 10- 3 slug/sec

rig. 2.1 Homogeneous friction factor for annular-mist flow. (Meyer and Wallis. 7 )

This method was used by McAdams et al. 4 and Owens. 8 Typical

results are shown in Fig. 2.2 which also represents the kind of
accuracy which can be achieved using this method. At high quali-
ties approaching the pure vapor condition, it is more accurate to
replace Eq. (2.68) by an analogous expression based on the ali-vapor
flow condition. ·
3. Substitute one of the expressions for equivalent viscosity in the
Reynolds number and use the single-phase friction-factor charts.
For example, Blasius' equation for smooth pipe flow is

C¡ = 0.079 Re-0• 25 (2.69)

Using Eq. (2.59) and the definition of cf,¡ 0 2 in Eq. (2.64), it is readily
found that

(2.70)

The predictions of Eq. (2. 70) for water are tabulated in Table 2. la.
Similar tables can be drawn up for cryogenic or refrigerant fluids to
enable rapid design predictions to be made.
An alternative procedure for calculating frictional pressure drop is
 '" ONE-D!MENSIONAL TWD-PHASE FLOW

200,--------,---,--,-,--,----rT77---,--,,
Ou □ lity, x
o O -0.1 +30%
/4 1/
a - 0.12-0.2 / 1/
□ -0.3-0.5_+-~io~ff'----
-10%

1/
/
10"'--''------'--L----'----'---'---'--'-L..L---~
10 20 40 60 80 100 200
Measured two-phase pressure gradient, psi/ft

Fig. 2.2 P.redicted versus measured two-phase pressure

gradient for 0.118 in.-diam 1 30-in.-long tube. Water at
pressures from 71 to 374 psia, and qualities from 4 to
50 percent. (Owens. 8 )

Table 2.1a Vaiues of fjJ¡ 0 2 for steam-water mixtures

predicted by the homogeneous model [IEq. (2.70)]

Steam Pressure, psia

qUality,
wt% 14.7 100 500 1000 1500 2000 2500 3000 3206
--- --- ---
1 16.21 3.40 1.44 1.19 1.10 1.05 1 .04 1.01 1 .o
5 67.6 12.18 3 .12 1.89 1.49 1 .28 1 .16 1 .06 1 .o
10 121.2 21.8 5.06 2.73 l. 95 1.56 1.30 l. 13 1 .o
20 212.2 38.7 7.8 4.27 2.81 2.08 1.60 1.25 1 .o
30 292.8. 53.5 11. 74 5.71 3.60 2.57 1.87 1 .36 1 .o
40 366 67.3 14.7 7.03 4.36 3.04 2.14 1 .48 1 .O
50 435 80.2 17.45 8.30 5.08 3.48 2.41 1.60 1.0
60 500 92.4 20.14 9.50 5.76 3.91 2.67 l. 71 1.0
70 563 104.2 22.7 10.70 6.44 4.33 2.89 1.82 1.0
80 623 115. 7 25.1 11.81 7.08 4.74 3.14 l. 93 1 .O
90 682 127 27.5 12.9 7.75 5.21 3.37 2.04 1 .o
100 738 137.4 29.8 13.98 8.32 5.52 3.60 2.14 1 .o
HOMOGENEOUS FLOW 31

Table 2.1b Values of ,:P¡o 2 for steam•water mixtures

based on Martinelli's emplrical correiation

Sleam Pressure, psia

quaiity,
wl % 14.7 100 500 1000 1500 2000 2500 3000 3206
---- - - - - - - ----

1 5.6 3.5 1.8 1.6 1.35 1.2 l. 1 1 05 1 .00

5 30 15 5.3 3.6 2.4 1 . 75 1.43 1 . 17 1 .00
10 69 28 8.9 5.4 3.4 2.45 l. 75 1 30 1 .00
20 150 56 16.2 8.6 5.1 3.25 2. 19 l. 51 1 .00
30 245 83 23.0 11.6 6.8 4.04 2.62 1.68 1.00
40 350 115 29.2 14.4 8.4 4.82 3.02 1.83 1.00
50 450 145 34.9 17.0 9.9 5.59 3.38 1 .97 1 00
60 545 174 40.0 19 .4 11. 1 6.34 3.70 2 10 1 00
70 625 199 44.6 21.4 12.1 7.05 3.96 2.23 1 00
80 685 216 48.6 22.9 12.8 7 70 4.15 2.35 1 00
\ÍO 720 210 48.0 22.3 13.0 7.95 4.20 2.38 1 00
100 525 130 30.0 15.0 8.6 5.90 3.70 2.15 1 .00

to avoid use of the friction factor and to use correlations for the two-phase
multipliers instead. The details of this technique will be left to the next
chapter. Far the particular case of boiling water the rcsults of such a
correlation scheme, evolved by lVIartinelli and N elson, 9 are shown in
Table 2.lb for comparison with the predictions of Eq. (2.70). Quite large

100-----------------------
◊ G = 3.22 x 10 6 lb/hr-ft 2
6.G = 2.47 x
oG=l.8lx
□ G = 1.29 x
v G=0.84 x Martinelli
and Nelson

---
- - _,,,,, ~Homogeneous
1 L - - - - - - - = - - _ : º : __ _ _ _ __!__ _ _ _ _ _ __ j
0.001 0.01 0.1
model

Vapor qu □ lity, x

Fig. 2.3 Dependence of ,:P10 on mass flux for water at 1000 psia_ (i1fus-
cettola.10)
32 ONE-DlMENS!ONAl TWO-PHASE FLOW

differences between the tables are evident. Which of them is the more
correct depends on the flow regime. The Martinelli-N elson predictions
tend to be better for separated flows, whereas homogeneous theory is
better far dispersed flows. Far example, Fig. 2.3 shows a drift from the
JVIartinelli-N elson predictions toward homogeneous theory as mass
velocity, and hence entrainment, is increased in high-speed steam-water
flows.

Example 2.3 Formulate equations for predicting the pressure drop during the
homogeneous fl.ow of a boiling liquid in a round tube of constant arca with a
uniform heat flux. N eglect kinetic- and potential-energy terms and the effects
of flashing and compressibility (these assumptions are only valid at high pres-
sures and low velocities when the overall pressure drop is small compared with
the absolute pressure). Assume a constant friction factor.
Solution First we relate the rate of heat addition per unit length to the heat flux,
as follows
dq, D (2.71)
dz=1rct,

where <pis the heat flux (in British thermal units per hour per square foot).
Using the definition of G, and remembering that W is constant, Eq. (2.3)
becomes, with the given assumptions,
dh 4$
(2.72)
d, - GD
Substituting Eq. (2.72) into Eq. (2.55) we get, neglecting flashing and
compressibility,

_ dp = 2C¡ QZ(v¡ + XV¡u) + 02 v 1u it_ + g cos 8 (2.73)

dz D htu GD 'ti¡ + X'tl¡u
The enthalpy of the mixture, if pressure changes are small, is given by the
equation

h = h1 + xh 1u (2. 7 4)

Equation (2.72) is now used in conjunction with Eq. (2.74) to give

dx 4$
(2.75)
dz = GDh,u
Since <pis constant, Eq. (2.75) may be integrated from the inlet value of x, to
the point z = z to give
~ 4z
X= Xi+ Gh,(J D (2.76)

Substituting for Eq. (2.76) in Eq. (2.73), denoting the density at inlet by Pi,
and integrating over the tube length L, it is eventually found that

.6. _ 2L C1G
p - J5 ----¡;;-
2
+ (2L)
D
2
C G v,u
f rp htu
+G
<J>
V1u
h¡g D
4L
Gh¡, ¡
+ g cos 0 4q,v,(J n
(1 + ~V¡,p;
Gh1u
4L)
D (2.77)
HOMOGENEOUS FLOW 33

In the special case where Xi = O an alternative form of Eq. (2.77) in terms

of the quality at z = L is

t:.p
= 2C1LG2v1
D
(l + v 1ªx)
'V¡2
+ 02'VJuX + LgX'Vjg
cos 8 ln (l + 'V¡g x)
'ti¡
(2.78)

This problem is discussed in greater detail by Owens. 8 In applying Eq. (2.77)

it is most advisable to check the assumptions sincc compressibility and kinetic-
energy effects rapidly become important as mass vclocities are increased and
flashing must also be considered whenever the relative change in pressure is
appreciable.

Exampie 2.4 Saturated water atara.te of G1 = 2 X 10 5 lb/(hr)(ft 2 ) enters the bottom

of a vertical evaporator tube >] in. in diamcter and 5 ft long. The tube receives
a heat flux of 2 X 10 5 Btu/(hr)(ft2), and thcre are no heat losses. Calculate
the pressure drop through the evaporator for inlet pressures of (a) 350 psia and
(b) 1000 psia. Assume a constant friction factor of 0.005.
Solution Assuming homogeneous flow theory this is simply an application of Eq.
(2.77),
(a) From steam tables, at 350 psia, v 10 = 1.3064 ft3/lb, v 1 = 0.01912
11'/lb, and h1 , - 794.7 Btu/lb.
From Eq. (2.77)

(2)(120)(0.005)(2 X 10')'(0.01912) + (240)'(0.005)(2 X 10')'(1.3064)

ilp (32.2)(3600)'(144) (3600) '(794. 7) (32.2) (144)
+ (2 X 10')'(1.3064)(4)(120)
794. 7 (32.2) (93600)'(144)
(32.2)(2 X 10')(794.7)(,-54) [ (2 X 10')(1.3064)(480) ]
+ (4)(2 X 10')(1.3064)(32.2)(144) In 1 + (2 X 10')(794.7)(0.01912)
- 0.015 + 0.315 + 0.505 + 0.165 psi
= 1.02 psi
Since this value is small compared with 350 psia, the assumption of constant
fluid properties is justified.
We should ELlso check for compressibility effects and confirm that the
exit steam is not superheated.
The exit quality, from Eq. (2.76), is

f 4L (2 X 10')(480)
x, - Gh1 , D - (2 X 10')(794.7) - O.&o 5

The compressibility and flashing effects are checked by evaluating the

terms in the denominator of Eq. (2.55). From steam tables we find

a:; - -3.8 X 10-'ft"/(lb)(psi)

dv 1 ~ 0
dp
OX)
( ap h
~ 0 at X = 1
-3.9 X 10- 4 ps1- 1 at X= Ü
-2 X 10- 4 psi- 1 at X = 0.5
34 ONE-DIMENSIONAL TWO-PHASE FLOW

The largest numerical value of the bracket in the denominator therefore

occurs at x = 0.6. The value of the whole denominator is
_ (2 X 10")'(0.6) (3.8 X 10-,) ~ O •
1 998 0
(144)(32.2) (3600)' .
Thc assumptions are thereforc justified. However, if both G and ,:P ,vere
increased by a factor of 10, the denominator would become 0.85 and could not
be put equal to unity without introducing significant error.
(b) At 1000 psia the relevant property va.lucs are V¡g = 0.424 ft 3 /lb,
v, = 0.0216ft 3/lb,h1 g = 649.5Btu/lb. Puttinginthcnumbersas aboveweget
0.424 794.7 .
i1pA ~ 0.525 1. ¡j";¡-g_S ~ 0.208 ps,
3064
O 015 0.0216 0.424 794.7 O 43 .
f1p, ~ . 0.01912 + o. 3 15 1.3064 649.5 ~ .l ps,
t:.pG = 0.111 ln 17.3 = 0.317 psi
The total pressure drop is thcrcfore 0.208 + 0.143 + 0.317 = 0.658 psi.

!Example 2,5 Air and water at 70ºF flow in a vertical pipe of diameter 0.98 in. at
ratcs W 1 = 31.3 X 10- 3 slug/scc and TVa = 0.583 X 10- 3 slug/sec. The exit
pressure is 14.7 psia and the pressure 18 in. upstream is found to be 15.0 psia.
How does this compare with thc predictions of homogencous flow theory with
C1 = 0.005?
Solution Since the pressure drop is low, it is a reasonahle approximation to assume
constant property valucs in calculating the gravitationa.l and frictional pressure
drops. The acceleration component is due to the expansion of thc gas. Since
W¡ » Wa, the thermodynamic path is probably isothcrmal (see Examplc 2.2).
Equation (2.44) is first put into a convcnient forro by utilizíng Eqs. (2.31)
and (2.32) to eliminate x, using Eqs. (2.11) and (2.12). Since the area change,
quality change, and compressibility of the liquidare essentially zero, we obtain
2C¡jG/D + g cos 0[apa + (1 - a)p¡]
(2 79)
1+ G)aP/l(dv /dp)
11

Since the gas expansion is isothermal, we have

dv, 1
(2.80)
dp PP,
The variables of intercst are cvaluated as follows:

j - Q¡ i Q, - 50.3 ft/sec
W1 +W
G ~ A ' - 6.08 slugs/(sec)(ft')

a - Q¡ ~ Q, ~ 0.939
Making the requisitc substitutions in Eq. (2.79) we find
_ dp _ (2.17 X 10-')(2.22 X 10-s) _ _, ·¡·
dz - 1 - 0.136 - 5 .08 X 10 psi m.

Since the prcssure gradient is so Iow, the velocities hardly change down
thc duct and the overa.U prcssure drop is found by multiplying the gradient by
HOMOGENEOUS FLOW 35

the lcngth, as follows:

1'p = (5.08 X 10-,)18 = 0.092 psi

The error in pressure-drop prediction is mostly due to the poor cstimate

of liquid fraction which is given by homogeneous theory at thcsc values of the
:flow rates. The flow pattern is, in fact, slug or annular and the gas flows much
faster than the liquid. The experimental liquid fraction was actua.lly 0.23,
,vhich is much greater than the value of 0.061 which homogeneous theory pre-
dicts. Errors of this order of magnitude are tobo expected when using a theory
for a flow regime to which it <loes not apply. Thc scparatcd-fl.ow model is far
more appropriate in the present case. Slug-flow theory or a thcory which is
valid in the transition rcgion from slug fl.o,v to annular flow is even more accu-
rate, as will be shown in Chaps. 10 and 1 L

In many cases the solution to a problem cannot be found in closed

form and it is necessary to put equations such as (2.55) or (2. 79) in terms
of finite differences and to use numerical methods of integration.
Note that the denominator of Eq. (2.79) together with Eq. (2.80)
can be combined to show that the square of the l\Iach number in an
isothermal homogeneous gas-liquid system is
M' = Gj, (2.81)
p

U PRESSUl!E llROP IN BENDS, TEES, ORIFICES, \IALVES, ETC.

The usual way of calculating pressure drop through pipe fittings in single-
phase flow is to replace the fitting by an equivalent length of pipe. The
same procedure can be applied to two-phase flows. The equivalent pipe
lengths tend to be somewhat longer in the two-phase flow case.
Several problems relating to pressure-drop prediction in nozzles
and orífices will be found at the end of this chapter.

2.5 UNSTEAl!V FLOW

Homogeneous theory can be extended to unsteady flows by including

time-dependent terms in the equations of continuity, momentum, and
energy. Far one-dimensional flow these equations are, in differential
form
Continuity (2.82)

lVIomentum (2.83)

Energy

(2.84)
36 ÜNE-DIMENSIONAL TWO-PHASE FLOW

These equations can be combined in numerous ways, depending

upon the conditions of interest. A particularly useful development is
to substitute the identity
(2.85)
into Eq. (2.84) and expand the differentials to get

(
h + 1!_'
2 at + a(pmv)]
) [ªPm oz + Pm (ªh
at + Vah)
az
+ PmV ( av
at + av
az + g cos e)
V
ap + A1 (ªª'
= at az - aw)
az (2.86) ·

Making use of Eqs. (2.82) and (2.83) then gives the result

ah+ V ah= _1_ (ªP + V ap) + VI'_Tw + _l_

at uz Pm at az A Pm Apm az
(ªª" - aw)
az (2.87)

The use of this equation may be illustrated by considering the case

of flow in a high-pressure straight-tube evaporator in which pressure
changes and viscous dissipation are small compared with the other energy
terms during a transient. Let the tube diameter be D, the wall heat
flux </,, and !et there be no shaft work. Equation (2.87) then becomes

(2.88)

If pressure changes and their effect on properties are negligible, the

enthalpy and density may be expressed as
h = h1 +
xh1, (2.89)
1
- ='ti¡+ X'tl¡g (2.90)
Pm
Substituting Eqs. (2.89) and (2.90) in Eq. (2.88) eventually yields

ax+ V ax =
.at az
('/J¡
'tifo
+ x) \l (2.91)

where
íl = 4'11¡,q, (2.92)
Dh1,
Equation (2.91) expresses the propagation equation for quality
changes and can be integrated to obtain the dynamic response of a boiler
channel. The left-hand side of the equation is the substantial time
derivative of the quality and represents the time rate of change of quality
for a given fluid particle (lagrangian viewpoint). The quantity íl has the
dimensions of inverse time and can be called a reaction frequency. If a
given particle is identified by the time to at which it starts to evaporate,
HOMOGENEOUS FLOW 31

Eq, (2,91) can be integrated to give

x(t,t 0) - V¡ (eº«-<,l - 1) (2,93)

V¡,

Therefore the quality of a given fluid particle increases exponentially

with time, Furtber developments along these lines lead to prediction
of the transient response of evaporators and condensers. 11

PROBLEMS
2.1. Deduce Eq. (2.16) from Eq. (2.10) and state what conditions are necessary for
an error of less than 5% in Eq. (2.16).
2.2. Salve Example 2.2 if the required exit velocity is (a) 1000 fps; (b) 1500 fps; and
(e) 2500 fps,
2.3. Suppose that a homogeneous two-phase flow between parallel plates is assumed
to consist of a large number of parallcl shects of the two phases oriented in the direction
of the plates. Show that the equivalent viscosity is given by

.!_=~+1-a
µ µ2 µ1

2.4. Salve Prob. 2.3 if the sheets of fluid are oriented perpendicular to the platea and
show that

µ = aµ2 + (I - a)µ1

2.5. Compare the equiva.lent viscosities predicted by Eqs. (2.59), (2.60), and (2.61) and
the results of Probs. 2.3 and 2.4 for steam-watcr mixtures of 0.1, 1, and 10% quality
at 10, 100, and 1000 psia.
2.6. For air-water flow in u 2~in.-diam pipe at 70ºF and 30 psia, calculate the friction
factor by using the various techniques in the text for the following conditions:

_I
1_'1_,1_p_s 1 ~-I 10 l~~-1-º-~

j~, fps ~ 10 n 10 1 100 1 200

2.7. Still dusty air is sampled by a stagnation pitot tube mounted on a jeep traveling
at 50 mph over the desert. The air contains its own weight of sand per unit wcight.
What is the measured stagnation pressure? State clearly all the assumptions which
are made.
2.8. A vertical tubular test section is installed in an experimental high-pressure water
loop. Thc tube is 0.4 in. ID and 7 ft long and is uniformly heated with 100 kw of
power. Saturated water enters at the base at 1000 psia and with a flow rate of 1000
lb/hr. Calculate the frictional, gravitational, accelerational, and total pressure drops
using the various equations in the te:xt. Compare the prediction for total pressure
drop with the measured valuc of 8 psi.
2.9. Estimate the critica! flow rate per unit arca for a stcam-water mixture, 26.9%
quality, ata pressure of 125 psia. The mcasured value is 1265 lbm/(ft2)(sec).
2.10. Air and water flow from a large tank through a converging nozzle with exit area
of 1 in. 2 • The air flow rate is 10 lb/hr, the water flow rate is lOlb/sec, and thefluids
are intimatcly mixed. The temperature is 70ºF and the externa! pressure is 14.7 psia.
 ONE-DIMENSIONAL TWO-PHASE FLOW

What is the pressurc in the tank? (Use Bernoulli's equation for the homogeneous
mixture and ncglect wall shear and gravity.)
2.11. Consider the isentropic, adiabatic homogeneous equilibrium flashing discharge
of carbon dioxide from a large storage cylinder at 1000 psia and 80ºF. What are the
quality, velocity, density, and flow per unit arca as a function of pressure? What
is the critical pressure at which choking occurs? Assume that choking corrcsponds
to the maximum flow per unit area.
2.12. Consider low-velocity horizontal laminar flow of a gas-liquid mixture in a pipe.
Let the gas volumetric fl.ow rate be Qu and the liquid volumetric flow rate Q¡. The
pressure drop for the liquid alone in the pipe is D.p 1. If the two-phase pressure drop
is D.pTP, use equation (2.58) to show that, if p¡ » Pu
t,,pTP _ l = 2Qu
D.p¡ Q¡

2.13. If the liquid is dirty, small gas bubbles tend to behave like solid spheres. In
this case, solve Prob. 2.12 using Eq. (2.57) to show that

.Ó.PTP _ l = 3.5 Qu
D.p¡ Q¡

2.14. Solve Prob. 2.12 for turbulent :flow assuming that the friction factor is the same
as for the liquid alone. Do the solutions to Probs. 2.12, 2.13, and 2.14 suggest a
method of plotting two-phase flow data? (See Reí. 12.)
2.15. A coal-water slurry is pumped in a horizontal 2-in.-diam pipe overa distance of
100 ft at an average velocity of 10 fps. The slurry is nonnewtonian with a limiting
viscosity at high shear rates of 0.01 lb/(ft)(sec). It consists of 45% by volume of
coal with a density of 85 lb/ft3. Single-phase flow tests in the same pipe gave the
following expression for the turbulent flow friction factor
C¡ = 0.026Re- 0 · 12
What is the pressure drop? What difference would it makc if the viscosity of water
at 7üºF were used in the Reynolds number instead of the limiting viscosity of the
slurry?
2.16. It has been suggestecl that the intensity of ",vater hammer" in hydraulic lines
could be significantly decreascd by suspendíng small air bubbles in the fluid. Discuss
the possibilities of thís idea.
2.17. Develop equations for describing the adiabatic compressible flow of a dusty gas
in a long horizontal pipeline, assuming a constant value of the friction factor (Fanno
line).
2.18. Develop equations for describing the :flow of a dusty gas in a long pipeline if
there is no friction but a constant wa.11 heat flux (Rayleigh line).
2.19. Combine the solutions to Probs. 2.17 and 2.18 to generate the normal shock
relationships for a homogeneous dusty gas.
2.20, Consider two-phase flow through the nozzle shown in Fig. 2.4. Assuming no
phase change and incompressible homogeneous fl.ow show that

and
Pi - p3
G,'
2p¡
HOIVIOGENEOUS FLOW
"
Pa
1

Fig. 2.4 Flow of a gas-liquid mix-

ture through a nozzle. (Prob.
2.20.)
u
How are thesc results alt.cred if the nozzle is sharp-edged and produces a vena contracta?
2.21. Figure 2.5 shows a proposed method for controlling the flow of oil through a
nozzle valve by varying the air flow supplied to the set of hypodermic needles. VVhat
is the relationship bctwcen the air and oil flow ratcs Wa and W¡?
222. Diehl has measured the pressure drop f:::.p for two-phase flow over tube banks.
I1.is results were correlated by plotting f:::.p/t::.p; versus Q¡p¡/(Q¡ + Qg)Pu, where f:::.p;
was the pressure drop for the same mass flow (W¡ + Wu), of gas a.lone. Show that
homogeneous flow theory, assuming a friction factor equal to the ali gas friction factor.,
is quite closc to the da.ta presented in Fig. 2.6.
2.2~. Show that the void fraction propagation equation analogous to Eq. (2.91) is

Deduce that, if to is the time at which an element of fluid starts to evaporate and n
is constant, the density of the same fluid element aftcr time t will be

2.24. Derive the normal shock-wave relationships for an isothermal, homogeneous

bubbly mixture. (See Chap. 9.)
2.2!t Equation (2.55) is not adequate for predicting the pressure gradient in evaporators
if the kinetic energy of the fluid is apprecia.ble, since in this case thc cnthalpy cannot'
be determined dircctly from a hcat balance. Use the energy cquation to show that,

---+
Oil
-->-
Wr

t
Air Wg

fig. 2.5 A two-phase flow control valve. The air and

oil stagnation temperatures and stagnation pressures
are IO0ºF and 400 psia. Assume choking at the throat
where A* = 0.1 in. 2•
40 ONE-DIMENSIONAL TWO-PHASE FLOW

o o
o
10-' 1-------+----t---=.,,j------t-------j
6p
o
o
D.p;

10-2 e - - - - - - - ¡ - - - - - - t - - - - - - - - - j - - - - - º - - , ~ c i ° - - - - - - - - j
o Air-woter l otm pg/pr- 0.0012
□ Pentane-pentane 44 otm pg/p¡- 0.0212
10-3~---~~---~----~----~---~
10-2 10° 10 1
_{)_[_ Pf
Or+Og pg

Fig, 2.6 Diehl's correlation 13 of two-phase pressure drop for turbulent

horizontal crossflow through tube bank with 45º layout. (Prob. 2.22.)

in the absencc of potcntial energy and shaft ,vork effects, Eq. (2.55) becomes

dp
- d,
1 + G2 [X
av,
-
ap
+ (1 - x) -av,
ap
+ V12 (ªx) ]
--
ap 1,.
'Í'

where

\11lhat is thc choking condit.ion in this case?

2.26. Compare the order of magnitude of the various terms in Eqs. (2.44) and (2.55)
for vertical upward steam-water fl.ow at 10% quality with G = l0 6 lb/(hr)(ft2) and
a wall heat flux of 10 6 Btu/(hr)(ft 2) in a 1-in.-diam pipe with a taper of Hoo in./in.
at pressures of I, 10, 100, and 1000 psia. Is the correction introduced in Prob. 2.25
important under any of these conditions?
2.27. Investigate the effect on the thrust of a jet engine oí injecting small, solid
particles into the hot gases before expanding them through a nozzle.
2.28. What is the discharge rate from a round-edged 1-in.-diam hole in the pressure
vessel of a nuclear reactor containing saturated water at 2000 psia? Assume choked
isentropic, adiabatic homogeneous flow and the following various circumstances: (a,)
Equilibrium flow; (b) H of the amount oí vapor predicted by the equilibrium assump-
tion everywhere; (e) :0,{o of the amount of vapor prcdicted by the equilibrium assump-
tion¡ (d) no vapor formation.
2.29. Air and water flow upward in a 10-ft-long, 2-in.-diam vertical pipe and discharge
into the atmosphere at 14.7 psi. Assuming homogcneous isothermal flow, ata tem-
perature of 7üºF calculate the inlet pressure for the following volumetric fluxes meas-
HOMOGENEOUS FLOW 41

Falling
film

Water
at
70ºF

Fig. 2.7 Air-lift pump demonstration.

(Prob. 2.30.)

ured at atmospheric pressure:

j¡, fps 0.5 1 10 ~5 30 35 40 40

--- ---- ------- - - - - - - - - - ----
j., fps 1 2 2 10 30 35 40 60

Under what conditions is the exit flow choked? What is the cxit pressure under
choking conditions? Use numerical techniqucs.
2.30. Figure 2. 7 shows an air-lift pmnp demonstration for use in a laboratory. Predict
pressure drop in thc riser as a function of the air and water flow rates Wg and W¡. If
h = 1 ft and H = 2 ft, what is the relationship between W g and W ¡? Under what
conditions is the power expended to run the air compressor per pound of water
pumped a minimum?
2.31. Solve Prob. 2.30 for the case where h = 50 ft and H = 100 ft and the pipe
diameter is increased to 6 in. Use numerical techniques.
2.32. Under conditions of rapid phase change, thermodynamic equilibrium as repre--
sented by Eq. (2.52) is not obtained. Assume that enthalpy is constant in a rapid
expansion but that the quality is given as a function of time, thus

Explore the one-dimensiol;}.al fl.ow relationships, assuming sorne simple expressions for
this function such as power laws, exponentials, series expansions, etc.

REFERENCES

l. Shapiro, A. H.: "The Dynamics and Thermodynamics of Compressible Fluid

Flow," The Ronald Press Company, New York, 1953.
2. Taylor, G. L: Proc. Roy. Soc. (London), vol. A148, p. 141, 1932.
3. Einstein, A.: Ann. Phys., vol. 4, p. 289, 1906.
4. McAdams, W. H., et al.: Vaporisation lnside Horizontal Tubes. II. Benzene-Oil
Mixtures, Trans. ASME, ·vol. 64, p. 193, 1942.
5. Cicchitti, A., et al.: Two-phase Cooling Experirnents-Pressure D.rop, Heat Trans-
fer and Burnout Measurements, Energi Nucl., vol. 7, no. 6, pp. 407-425, 1960.
., ONE-DIMENSIONAL TWO-PHASE FLOW

6. Dukler, A. E., et al.: Pressure Drop and Hold-up in Two-phase Flow, A. I. Ch. E. J.,
vol. 10, no. 1, pp. 38~51, 1964.
7. Meyer, P. E., and G. B. Wallis: Co-current Upwards Annular-mist Flow, AEC
Rept. NYO-3114-10, Sept., 1965.
8. Owens, W. L., Jr.: Two-phase Pressure Gradient, International Developments in
Heat Transfer, Am. Soc. Mech. Engrs., papcr 41, vol. 2, pp. 363~368, 1961.
9. Martinelli, R. C., and D. B. Nelson: Prediction of Pressurc Drop during Forced
Circulation Boiling of Water, Trans. ASME, vol. 70, p. 695, 1948.
10. Muscettola, M.: Two-phase Pressure Drop, UKAEA Rept. AEEW-R284, 1963.
11. Wallis, G. B., and J. H. Heasley: Oscillations in Two-phase Systems, Trans.
ASME, ser. C, vol. 83, p. 363, 1961.
12. Wallis, G. B.: Sorne Hydrodynamic Aspects of Two-phase Flo,v and Boiling.
Part III. International Developments in Heat Transfer, Am. Soc. Mech. Engrs.,
paper 38, vol. 2, pp. 330-340, 1961.
13. Diehl, J. E., and C. H. Unruh: Two-phase Pressure Drop for Horizontal Cross-
flow through Tube Banks, Am. Soc. Mech. Engrs., paper 58-HT-20, 1958.
14. Simpson, H. C., and R. S. Silver: Theory of One-dimensional Two-phase, Homo-
geneous, Non-equilibrium Flow, Symp. Two-phase Fluid Flow, London, Inst.
Mech. Engrs., pp. 45-56, 1962.
 3
Separated Flow

l.l INTRODUCTION
The separated flow model takes account o! the fact that the two phases
can have differing properties and different velocities. It may be devel-
oped with various degrees o! complexity. In the most sophisticated
version, separate equations of continuity, momentum, and energy are
written far each phase and these six equations are sol ved simultaneously,
together with rate eqµations which describe how the phases interact with
each other and with the walls o! the duct. In the simplest version, only
one para.meter, such as velocity, is allowed to differ for the two phases
while conservation equations are only written for the combined flow-.
When the number o! variables to be determined exceeds the available num-
ber of equations, correlations or simplifying assumptions are introduced.

3.2 STEAOY FLOW IN WHICH TI-IE PHASES ARE CONSIDERED

TOGETHER BUT THEIII VELOCITIES ARE ALLOWED TO lllFFEI!
Suppose that one of the assumptions of homogeneous equilibrium flow
is relaxed in order to allow different velocities far the two phases. The
43
44 ONE-DIMENS!ONAL TWO-PHASE FLOW

conservation laws for mass, momentum, and energy in steady one-dimen-

sional flow can then be derived in terms of the two velocities V1 and V2-
Alternatively1 the extra degree of freedom of the system can be accounted
for by introducing both the void fraction and the quality into the equa-
tions. Sorne of the most useful forros of these equations will Ilow be
developed.

CONT!NUITY

N ormally, no mass is added to the flow from outside the duct and the
overall mass flow rate is constant. Therefore

W ~ W1 + W, ~ const (3.1)

In the absence of phase change both W 1 and W, are constant individually.

The mass flow rates can be expressed in terms of other variables in
many ways. Far example, if the areas of cross section of the two streams
are A1 and A2, we have

W1 = p¡V1A1 (3.2)
YV2 = P2V2A2 (3.3)

The mass flux of each stream is then, fron1 the relationships derived in
Chap. 1,

G1 ~ p1V1(l - a) (3.4)
G2 = P2V2a (3.5)

By using the definition of x, the aOove two equations can be used to give
two alternative expressions for the overall mass flux as follows:

1 - °' (3.6)
G = P1V1 -~
1- X
G = pzV2 ~ (3.7)
X

MOMENTUM

:V[any alternative forms of the momentum equation can be derived by

manipulating relationships among a, x, G1 v 11 v2 i and other variables.
For steady flow in a round pipe, for example, one version is

dp 47-w
- dz - D + G dzd [xv, + (1 - x)v1] + [ap 2 + (1 - a)p 1] cos 0

(3.8)
SEPARATED FLOW

ENERGY
The energy equation is conveniently written 1n terms of the quality.
Thus
_1_
Wdz
(dq, - dw)
dz
~ 11,__ [xh, + (1 - x)h1]
dz
12
+ 11,__
dz
[x v,' + (1 -
2
x) "
2
] + g CDS 0 (3.9)

In order to solve the above set of equations, two further relationships

are required besides knowledge about the relationships between thermo-
dynamic properties. While it is more satisfactory to derive these rela-
tionships by analyzing each component separately, a common technique
is to use empirical correlations for T w and a in terms of the flow rates,
fluid properties, and geometry. The method of solution is then mostly
determined by the farm of these correlations.
When using correlations to fill the gaps in the theory, it should be
remembered that they are usually established far adiabatic flow with
low-pressure gradients. Under conditions of rapid phase change and
acceleration they may be in serious error.
A common way of correlating the wall shear stress in gas-liquid
flow was faund by Martinelli and coworkers. 1.2 The actual two-phase
shear stress is expressed as a factor <p 2 times the wall shear stress which
would occur in a related single-phase flow. :For example, one method
that was introduced in the previous chapter relates the wall shear stress
to the stress which would act if ali the mass flow were composed of liquid.
This is evidently motivated by problems of boiling and condensation.
Defining (C1)10 as the appropriate friction factor far the all-liquid flow,
\Ve have

_(dp)dz ,
~ 2(C1)10G'v1</>10 2
D
(3.10)

The void fraction correlation is often expressed 1 for a vapor-liquid

mixture of a given substance, in the form
a ~ a(p,x) (3.11)
In this case it is an exercise to show that the equivalent of Eq. (2.44) is
_ dp ~ (2(C1 ) 10 G v1 0
2 2
,j,¡ + G' dx 1(2xv, _ 2 1 - x 'ti¡)
dz D dz " 1 - "

+ (!:), [i~ =:;:'ti¡ - :: 'ti,]) - ~2 ~~ [ f V,+ (~ = ~2 'ti¡]

+ g cos 0 ( ~
V¡
+ 'llo-"'))
(
l + 02 ¡~av, + (1
adp 1-aap
2
- x) av¡ + (ª") [(l - x)' 'tlr _ x
ap,(l-a) 2 a'
2
v,J¡)- 1

(3.12)
 ONE-DIMENSIONAL TWO-PHASE FLOW

For the solution of equations such as this, one is virtually forced

to use numerical methods.
The denominator of Eq. (3.12) may be equated to zero to obtain
the condition for "choking" in the absence of flashing, using Eqs. (3.6)
and (3.7) as follows:

+ (ª°'
-a )
2
a--¾
V
Cg
+ (1 v¡'
- a) --¡
C¡ p X
(p¡v¡' - p,v,') = l (3.13)

However, this equation is misleading since the term (ao:/ap), is

usually derived from a correlation which was obtained at moderate values
of the pressure gradient when frictional forces dominated the inertia
terms. A more rigorous treatment that considers the separate equations
of motion of the two streams will be considered later (p. 68). If inertia
effects domínate, the choking condition is found to be

;, (e~' - v~') + 1
: o: =O (e~, - v~') (3.14)
Equations (3.13) and (3.14) are compatible only for the particular
case in which
(3.15)

Example 3.1 Consider the discharge of fl.ashing liquid through a nozzle. Let the
initial state be saturated liquid at a high pressure and possessing negligible
kinetic energy. At a much lower pressure downstream let tbe quality be x.
Assume adiabatic equilibrium flow and neglect gravitational effects. What is
the maximum possible flow rate per unit area at the downstream point?
Solution Let the enthalpy change from the initial to the final condition be ti.h. From
Eq. (3.9), in view of the assumptions, we have
(3. 16)

The mass flow rate per unit area can be expressed in terms of the quality by
eliminating a between Eqs. (3.6) and (3.7) to obtain

(3.17)

Denoting vu/v 1 by the symbol n and eliminating v 1 2 from Eqs. (3.16) and (3.17),
we get

2 ti.hG- 2 = (
-X
npu
+-
j --
Pf
")' (xn 2 + I - x) (3.18)

If the thermodynamic path is known, ti.h and x are fixed and Eq. (3.18) can be
used to predict the influence of n, the velocity ratio, on the mass flux G. In
particular, G will be a maximum when

=2nx ( -X
npu
+J-x)'
-Pf-
- 2(xn' + 1 - x) - 2X
n pu
(X- + J-x)
npu
- -
P!
(3.19)
SEPARATED FLOW
"
that is, when

(3.20)

The first par.entheses can never be zero, and the only nontrivial solution of Eq.
(3.20) is

or !'R =
v¡
(P_j_)"
P~
(3.21)

and the maximum value of Gis then, from Eq. (3.18),

G _ (2 t,h)H
me, - [(! - x)/p¡% + x/p,%p, (3.22)

Variations in n close to the value given in Eq. (3.21) have small effects on the
value of G.
If choking is regarded as a condition of maximum possible discharge
through a given nozzle exit area, then it can be argued that this condition corre-
sponds to the value of G given by Eq. (3.22). l\foody 3 has had sorne success in
plotting Eq. (3.22) versus available data for steam-water flows as is shown in
Fig. 3.1. Since x does not vary much with the thermodynamic path and t:.h
is a maximum for an isentropic process, the maximum value of G should occur
if the flashing is reversible and isentropic.
In the graphs shown in Fig. 3.1 the abscissa is the measured pressure Pm
at the exit of the nozzle. Since the flow was choked, this pressure was not the
same as the receiver pressure. Figure 3.2 gives value of Pm as a function of the
upstream stagnation condition, p 0 and ho, based on the additional criterion
that Gin Eq. (3.22) is to be maximized as a function of the pressure p2. Thus
p.,. corresponds to the further condition

when p2 = Pm (3.23)

-,
u
-1:
x~o.2 •'
~ u
'-
_2103 lc------~i.:IJ!'$<''.____----d ---:::: 10 3 ec------+..-~,,_____---1

x- 0.4
•
..o

10'~--L---'----'--'--'J..L~--L---'----'--W...U~ 10'~--L.-L-L..U..LLl.L.__ _L_._L_L.LJ..Ll.lJ

IO 100 l000 10 100 1000

P,w, psia
'
Fig. 3.1 Comparison between measured choked flow rates of steam-water mixtures
and Eq. (3.22) for two ranges of steam quality, x. (Afoody. 3 )
48 ONE-DIMENSIONAL TWO-PHASE FLOW

T T T T
1 '
1
1 1 1 1 1

P0 =3206.2 psio Critica! pressure]

-
2000 L
~ -,;" - 2800' ~
/ p,.. 2400
' -
/ Po- 2000 \
./ \
1000 Po 1600
-
I -
800 - Po 1200 '\ -
/ ~ -
Po 1000 o
600 -
-
/ Po 800
-oe
o
-

-
o el o
_Q
'[400 - {!f-- Po' 600 -
it--- Po 500 18-
""
~
o
-
.~
o
--0 1

Po 400
1t
,.,
¡-o
-

:...._!? r-._ 1
~ 200 - -ó / Po 300 1~ -

a. .!'!/ o
-~
g
-
Jt- ¡.__ ) o 200
~ -
J? /
8 100 I
s L -
1
/- -
-
80 L
L -
Po IDO 1
60 L -

L
1 1 -
1 1
40 ' -
/l.....__ 1
Po 50 1 -
'
1 1
20 L
L
l__ Po~ 25 Psia Stagnation pressure
1
1
-

10 ' ' ' ' ' ' ' ' ' '
o 200 400 600
800 1000 1200 1400
Stognation enthalpy, h0 , Btu /lbm

Fig. 3.2 Local static pressure and stagnation properties at maximum steam-water
fl.ow rate in terrns of upstrcam stagnation conditions. (1,11oody. 3 )

If the receiver pressure is less than Pm; the flow is assumed to be choked and G
is calculated from Eq. (3.22) using thermodynamic properties appropriate to Pm•
In general, the theoretical valucs given in Fig. 3.2 tend to be above the values
observed by Fauske 4 who found that, to a good approximation, PmlPo ,,;::, 0.55.
In spitc of its impressive correlation of data, this theory of maximum
flow ata given kinetic energy (or the equivalcnt mínimum kinetic energy ata
given fl.ow rate) is by no means correct in detail. Much lcss vapor is produced
than is predicted from the equilibrium assurnption. Thc velocity ratio is not
given accurately by Eq. (3.21). Part of the success of Eqs. (3.22) and (3.23)
is caused by the fact that they prcdict an extremum of extremes which is not
too sensitive to the initial assumptions.
SEPARATED FLOW .,
i!VALUATlON Of WALl SHEAIR STRESS AND \IOID IFRACTiON

Equations (3.1), (3.8), and (3.9) cannot be solved, in general, without

additional expressions for the wall shear stress (oi_Jrictional pressure drop)
and void fraction. One possibility is to use ""'the homogeneous flow
assumptions, iri which case the equations reduce to those which were
given in the previous chapter. An alternative usually associated with
the separated flow model is the correlation scheme which was developed
by :VIartinelli and others 1 •2 in the period of 1944-1949.
Perhaps the simplest way to approach the correlation is via the
frictional pressure drop. For given flow rates of the gas and the liquid
it is assumed._that we know how to calculate the pressure gradient which
would occur if either fluid were flowing alone in the pipe. Denote these
pressure gradients by (dp/dz), and (dp/dz) 1 . Now define new variables
in terms of the ratio between the observed pressure gradient dp/dz with-
out phase change, acceleration, or body force effects, and the values for
each component alone. Thus
dp/dz
(3.24)
(dp/dz),
dp/dz
(3.25)
(dp/dz) 1
When there is no gas flow, the following results apply.
1
-=O (3.26)
</>.'
Whereas when there is no liquid flow we have, similarly,
1
-=O (3.27)
<l>t'
Furthermore, at the critical point where the phases are indistinguishable
the relationships between q, 1 and </>, are
For laminar flow,
1 1
-+-=l (3.28)
<p¡2 cp/
For turbulent flo,v obeying Blasius' smooth-pipe law,

(:!)" + (:,)" = 1 (3.29)

A very simple model of separated flow can also be developed by

assuming that the two phases flow, without interacÜon, in two horizontal
separa te cylinders and that the areas of the cross sections of these cylinders
add up to the cross-sectional area of the actual pipe. The pressure drop
50 ONE-DIMENSIONAL TWO-PHASE FLOW

1.0 "'--
',--AII gas, a=l

0.8 X 2 =0.1, a;;,0.9

0,6
(/:1, a~0.8
/
(3.24}
Eq. [3.28}

0.4 n=l

0.4 0.6
1/,¡,/
Fig. 3.3 l\!Iartinelli's correlation scheme represented on
the basis of Eq. (3.30).

in each of the imagined cylinders is the same as in the actual flow, is due
to frictional effects only, and is calculated from single-phase flow theory .
.This separate-cylinders model5 resembles l\tlartinelli's original formulation
but has the virtue that it can be pursued to an analytical conclusion,
whereas lVIartinelli was content with a correlation. The results of this
analysis are (see Probs. 3.3, 3.4, and 5.12 and Example 3.2):

(2-)1/n
<l>t' +
(2-)1/n ==
q,,2 1
(3.30)

where n == 2 for laminar flow, 2.375 to 2.5 for turbulent flows analyzed
on a basis of friction factor, and 2.5 to 3.5 for turbulent :f:lows calculated
on a mixing-length basis. Equation (3.30) is compatible with Eqs. (3.26)
to (.3.29), which are shown in Fig. 3.3.
Equation (3.30) gives a one-parameter family of curves which can
be used to fit experimental data.

Example 3.2 Develop the separate-cylinders model far turbulent fl.ow, assuming a
constant friction factor for both phascs.
Solution Let the gas fl.ow in a cylinder of radius rQ and the liquid in one with -.:::,dius r¡.
The gas and liquid volumetric fractions will then be

a=-
r,'
ro'
where r 0 is the radius of the actual pipe.
SEPARAT!:::D FLOW
"
For the gas the pressure grndicnt in the imagined cylinder is

- dp = e, Pu (b)2 = C¡p/]ji __!_

dz ru a ro a%
and is the same as the actual pressure gradient.
The :first factor on the right-hand side is identical with ( -dp/dz)u;
thereíore, from Eq. (3.24)
1
'Pu2 = a%

Similarly for the liquid

' - 1
fÍJJ - (1 - a)~i

Eliminating a between these expressions, we get

which is Eq. (3.30) with n = 2.5.

In order to avoid having the unknown two-phase pressure drop in

both the correlating parameters it is convenient to divide them to yield
a new variable. Thm,

X' = ,f,,' = (dp/dz)¡ (3.31)

q,¡2 (dp/dz),
X 2 gives a measure of the.degree to which the two-phase mixture behaves
as the liquid rather than as the gas. Thus, one might expect sorne
relationship between X 2 and the void fraction as indicated in Fig. 3.3.
Because of the much greater specific volume of the gas, the void-fraction
variation is not symmetrical about X = l and tends to be lopsided at
low pressures, as shown in the figure.
Since the pressure drop can be calculated from laminar- or turbulent-
flow theory for both the liquid and the gas, four different combinations
are possible. In annular flow the gas is seldom viscous if the liquid is
turbulent, and therefore this combination has not been studied exten-
sively. J\1artinelli's data for Iow pressures for the other three permuta-
tions are shown in Figs. 3.4, 3.5, and 3.6 together with the correlating
line which was drawn by the original authors.
A reasonable correlation for ali flow regimes is given by Eq. (3.30)
with n = 3.5, as shown in Fig. 3. 7; n = 4 gives a better correlation in
the turbulent-turbulent regime.
Martinelli's empirical correlation for void fraction at low pressures
is shown in Fig. 3.8. The line can be represented very well by the equation
a = (l + Xº·s)-o.m (3.32)
 ONE-DIMENSIONAL TWO-PHASE FLOW
"
100
..;J:
- - - Mortinelli correlation (Ref. l)
o. - - - Eq.(3.30), n~4
e
7' • Dota points
~
~
~
.•
o.
10
e
,Q
~
•E .:···~--
o :_:'f;,,..... =:

1 L___L_.L..I....LLLlil---'-...L-'--Ll_lll_ _j__j_c._L_LI_Ll_u__ _L__LL.L-'.LLLl

0.01 0.1 10 100
Mortinelli parameter, X

Fig. 3.4 Comparison between pressure drop predicted by Eq. (3.30) with n = 4 and
empirical results of Martinelli for turbulent-turbulent flow. 6

1ooc--------l----------+------+------A
..;J: Martinelli correlation (Ref. 1)
o. Eq. (3.30), n ~ 3.5
7'
e Data points
t
~
~
~IOe------------1-------+---~~"'-+-----------=
_.,
e
·2e
•E
o
---
1 =_j__j__l_LJ..l-1.l.L-"---'-_LL.L.LLl.lL_L_.L..J__j_J_-1.J..11c-_j__j__LLJ..l.l,!-l
0.01 0.1 10 100
Mortinellí parometer, X

Fig. 3.5 Comparison between pressure drop predicted by Eq. (3.30) 1vith n = 3.5
and empirical results of Martinelli for viscous-liquid turbulent gas fiow. 6
SEPARATED FLOW

IOOf-------+-------+------+------rl
Martinelli correlation (Ref.1)
~ Eq. (3.30), n = 2.75
a. Eq. (3.30), n = 3.5
e Data points
"~
~
e
~ 101----------l---------+-----=f"---+---------I
...
e
-~e
•E
o
1 L==mr-::-~~uul__J____LLJ..ij~____L_j_j_~-
º·º1 0.1 10 l00
Mortinelli porameter, .X

Fig. 3.6 Comparison betwecn pressure drop predicted by Eq. (3.30) and empirical
results of Martinelli for viscous-viscous flow. !i

100
~ Eq. (3.30), n =3.5
-e. Dota points
a.
e
"E
~
E
.
a. 10
~
e
o
·a
e
•E
o
1
O.DI 0.1 1 IOO
Martinelli porameter, X

Fig. 3.7 Comparison betwcen pressure drop predicted by Eq. (3.30) with n = :3.5
and empirical results of Martinelli for all flow regimes. 5
54 ONE-DIMENSIONAL TWO-PHASE FLOW

o
cr
:o

0.01 L__L_.L__LL.J...LilL-----'-------'----'-_u_-'--'--u_-.L___L._.LJ...LC.lli-------'~----'-...LC~
0.01 0.1 1 10 100
Martinelfi porameter, X

Fig. 3.8 Martinelli's correlation for void fraction. (Ref. l.)

Although Martinelli's correlation was specifically derived for hori-

zontal flow without phase change ar significant acceleration, it is often
used to calculate both the void fraction and frictional pressure drop for
insertion into Eq. (3.12), even when these other effects are not negligible.
This procedure leads to progressively increasing errors as the frictional
component of pressure drop decreases in proportion to the other terms.
The lVIartinelli correlation in essence balances frictional shear stresses
versus pressure drop. lHore elaborate methods are required when body
forces and inertia forces are significant.

IExample 3.3 Solve Example 2.5 using the Martinelli-Lockhart correlations.

Solution (working in cgs units) The Rcynolds number for the liquid flowing alone
in the pipe is found to be 23,000; therefore the flow is turbulent and the friction
factor is 0.005. The pressure drop is calculated to be
dp) _ (2)(0.005)(94.5) '(1.0) _ .d / ,
(a)
( - dz f - 2.48 - 36 ynes cm

Similarly for the gas alone the Reynolds number is found to be 24,000,
giving a turbulent friction factor of 0.005 anda pressure drop

(
- dp) ~ (2)(0.005)(1.4 X 10')'(1.25 X 10-•J ~ 9 _4 dynes/cm' (b)
dz u 2.48
The value of the Martinelli parameter Xii is

Xu ~ [(dp/dz)i]" ~ (~)" ~ 1.96 (e)

(dp/dz), 9.4
From Figs. 3.4 and 3.8,
(1 - a) ~ 0.31 (d)
 SEPARATED FLOW

The friction pressure drop is therefore

!1p, - 1; ( - ~;), L - (40) (9.4) (45.6) (1.4 X 10-•) - 0.24 psi (e)

The gravitational pressure drop is

f::.pG -
_ [(0.31)(62.5) + (0.67)(0.076)] 15
. -_ 020
. .
pS! (f)
144
There is a problem in- calculating the acceleration pressure drop because it is
very sensitive to the accuracy with which small changes in a can be calculatcd.
However, it is unlikely to be greater than the homogeneous flow acceleration
pressure drop. Therefore we use Eq. (2.81) of Example 2.5 to get an expected
overestimate of the pressure drop. Thus

!1p 0.24 + 0.20 _ O 51 .

0.866 - · psi

These predictions for liquid fraction and pressure drop both exceed the
measured values (0.23 and 0.3 psi). This is because the actual flow regime is
sli.lg-annular, liquid bridges move with the gas velocity, and the liquid fraction
is therefore decreased. Example 10.5 shows how an accurate solution can
be obtained.

FlOW OF BOILING WATER IN STRAJGHT PIPES

Because of its practical importance, the flow of boiling water has perhaps
been studied more than any other two-phase system. Correlations based
on the Martinelli method are highly developed and tables and curves
have been generated for convenient use. The values fot <p10 2 which result
from the Martinelli-Nelson 2 correlation were already presented in Table
2.lb. A graph which allows a rapid prediction of void fraction as a
function of quality and pressure is shown in Fig. 3.9. These empírica!
relationships, together with Eq. (3.12) and the assumption of thermal
equilibrium, enable pressure-drop predictions to be made by numerical
integration.
In many cases considerable simplification is possible. For example,
in straight pipes wi_th no area change (constant G), Eq. (3.8) can be
integrated to give
4 (L
t,p - D Jo Tw dz + G[xv, + (1 - x)v¡]t

+ g cos e Jt [ap, + (1 - a)p¡] dz (3.33)

For constant heat flux and negligible work, kinetic- or potential-

energy changes, or property variations, Eq. (3.9) gives in the two-phase
reg10n

(3.34)
56 ONE-DIMENSIONAL TWO-PHASE FLOW

o"-------'-------'-------"----'---
o 20 W 60 80 100
Quality, x, per cent

Fig. 3.9 Void fraction versus quality for steam-water

flow at various pressures. (Martinelli and Nelson. 2)

0% exit quality
=;:,..,.¡,-90
100 ªº'~---+---------0
70
60
50
40

l. 0 L__[__l__l_l_lll_lL____l___i_Ll_l_Lw___J""°LLL.llil
10 100 1000 10,000
Pressure, psia

Fig. 3.10 q;,10 2, from Eq. (3.36), as a function of pressure

and quality for water. (J.11 artinelli and N elson. 2)
SEPARATED FLOW 57

100 c-----,--,-rrrm-,---,-----,TTTTTrr---,-,,rTT1,1

100% exit quolity

90 1

80
lübc'>'2's~~770 - - - - - ¡ - - - - - - j
60
50
40

0.01 L _ - C _ L _ L _ ~ = - ~ ~ ~ = ~ ~ ~ ~ ~ ~
10 100 1000 10,000
Pressure, psio

F'ig. 3.11 r 2 , from Eq. (3.37), as a function of pressure and

exit quality for water. (Martinelli and N elson. 2)

showing that the quality increases linearly with distance. Letting x = O

at z = O, assuming that C1o is constant, using Eqs. (2.5) and (3.10), and
assuming constant properties, Eq. (3.33) becomes

Dp¡
2
D.p = 2C10 G L
x )o
(.!: ["
</>io' dx) G'
ap(}
+ p¡ 1 - a
1~ +
_!_ [ (1 - x)' 1 ] ¡
+ g cos 0 Jt [(l - a)p1 + ap,J dz (3.35)
The various terms in parentheses can be evaluated from the lVIartinelli-
N elson correlation. This has been done for water for the first two terms
and the results are shown in Figs. 3.10 and 3.11 in terms of the parameters

1'10' = _!:X )o
(" </>10' dx (3.36)

r, = _!_ [x' P_J + (1 - x)~ - 1] (3.37)

Pt a Po 1- a
The final term in Eq. (3.35) has to be calculated by integrating the results
58 ONE-D!MENSIONAL TWO-PHASE FLOW

Table 3.1 Revised values of Martinelli factors as given by Thom 6

Pressure, psia
Steam
quality, 250 600 1250 2100 3000
wt%
efi¡o 2 ;¡,,02 f 102 ;¡,,02 efito 2 ;¡,,02 <'P,02 ;¡,,02 <'P,02 ;¡,,02
--- --- --- ---
1 2.12 1.49 1.46 1.11 1.10 1 03 .....
5 6.29 3.71 2.86 2.09 1.62 1.31 1.21 1 .10 1 .02
10 11. 1 6.30 4.78 3.11 2.39 l. 71 1.48 1 21 1.08
1.06
20 20.6 11.4 8.42 5.08 3.77 2.47 2 02 1.46 1.24
1.12
30 30.2 16.2 12.1 7.00 5.17 3.20 2 57 l. 72 1.40
1.18
40 39.8 21.0 15.8 8.80 6.59 3.89 3 .12 2.01 1.57
1.26
50 49.4 25.9 19.5 10.6 8.03 4.55 3.69 2.32 l. 73 1.33
60 59.1 30.5 23.2 12.4 9.49 5.25 4.27 2.62 1.88 1.41
70 68.8 35 2 26.9 14.2 10.19 6.00 4.86 2.93 2.03 1.50
80 78.7 40.1 30.7 16.0 12.4 6.75 5 45 3.23 2.18 1.58
90 88.6 45.0 34 5 17.8 13.8 7.50 6 .05 3.53 2 33 1.66
100 98.86 49.93 38.30 19.65 15.33 8.165 6.664 3.832 2.480 1.740

from Fig. 3.9. These graphs provide a very rapid method for estimating
pressure drops for boiling water.
Altemative (generally better) values for </>¡ 0 2, é/J¡ 0 2, a, and p¡r 2 for
high-pressure water were deduced by Thom, 6 using much more data than
the original _i\'.Iartinelli-Nelson correlation, and are shown in Tables 3.1
and 3.2.

Table 3.2 Revised values of Martinem factors as given by Thom 6

Pressure, psia

Steam
quality, 250 600 1250 2100 3000
wi %

T2pJ T¡p¡
" " " np¡
" T2pJ
" Tzpf

- - - - - - - - - ---· - - - - - - - - - - - - - - -
1 0.288 O .4125 0.168 0.2007 0.090 0.0955 0.0476 0.0431 O. 0213 0.0132
5 0.678 2 .169 O. 512 1.040 0.340 0.4892 0.207 0.2182 0.102 0.0657
10 0.816 4.620 0.690 2. 165 0.521 l. 001 0.355 0.4431 0.193 O.1319
20 0.910 10.39 0.833 4.678 0.710 2.100 0.553 0.9139 0.350 O.2676
30 0.945 17.30 0.895 7.539 0.808 3.292 0.679 1.412 0.480 0.4067
40 0.964 25.37 0.930 10. 75 0.866 4.584 0.767 1.937 0.589 0.5495
50 0.975 34.58 0.952 14.30 O. 908 5.958 O. 832 2.490 0.682 0.6957
60 0.984 44.93 0.967 18. 21 O. 936 7 .448 0.881 3.070 0.763 O. 8455
70 0.990 56.44 0.979 22.46 0.959 9.030 0.920 3.678 0.834 O. 9988
80 0.994 69.09 º·988 27.06 0.976 10.79 0.952 4.512 0.895 1.156
90 0.997 82. 90 0.995 32.01 0.989 12.48 0.978 5.067 0.951 l. 316
100 1 98.10 1 37.30 1 14.34 1 5.664 1 1.480

1 ..• . . .•..··
 SEPARATED FLOW
"

G = 10 6 lb/hr-ft 2

3.5
2

0.1

o60Lo~o~1l_LL1.Uo11.01
0~0~1l_LL.Wo.JJ_too"'1_1__L.1...LWJ0~.0~1--:r:~;=;:i:;¡¡;o~.1~¡¡¡¡

Sodium,ºF .!cc--cec-~~Sc~=-,,:
1200
Potassium, ºF
1000
1400
Property index, (µ,/µ 0 )º·1/(p,/p9 )

1600

1200
1800 2000

1400
Freon-22,ºF 40

1600 1800 2000

~--~~~~~
80100 140 180 205

Rubidium, ºF
1000 1200 1400 1600 1800
Mercury, ºF
600 800 1000 1200 1400
Water, ºF ',---~c'-,---~'c---c'cc-----,,-,----~
212 328 467 545 636 705

Fig. 3.12 Baroczy's correlation 7 of c/J¡o 2 for G = 10 6 lb/(hr)(ft2).

A further extension of the method by Baroczy 7 enables it to be used

for fluids other than water over a wide range of mass velocities. The
parameter <f>10 2 is expressed as a function of a property index (µ¡/ µ,)º· 2 (p,/ p1 )
for a given value of G ~ 10 6 lb/(hr)(ft') (Fig. 3.12). For other mass
velocities, this value of rf>io 2 is multiplied by a correction factor íl shown
in Fig. 3.13 in arder to get the appropriate value of 1> 10 2 . The wbimsical
fluctuations in the curves in Fig. 3.13 provide an cxcellent stimulus for
mirth among those skeptics who are unimpressed by grand correlation
schemes. However, they do enable a design engineer to get on with the
job of making performance predictions in the absence of any better
information.
60 ONE-DIMENSIONAL TWO-PHASE FLOW

1.6,--,-------,-------,-----~--------,

G= 0.5 x 10 6 ond G= 3 x 10 6

1.4
Quality, %
10_./

1.2

100
1.0

0.8

Cl 0.6
-ie
4
·.e
0.4 ~ - ~ - - - - - - - ~ - - - - - - - ' - - - - - - - - - - - ' - - - - - - - - ~
o
El8,--,-------,-------,-----~-----~
•o
~
,,. G=0.25 x 10 6 and G = 2 x 10
6
o
~ 1.6

1.4

1.2

1.0

0.8

0.6
40 to 60

0.0001 0.001 0.01 0.1

Property index, {µ 1 /µg)º· 2/(p¡/pg)

Fig. 3.13 Mass flux correction factor versus property index for Baroczy correlation. 1
SEPARATED FLOW
"
3.3 ONE-OIMENSIONAL SEPARATE!l FLOW IN WHICH THE PHASES ARE
CONSlllEl!ED SEPARATELY

In order to avoid the extensive use of correlations to make predictions

for separated flmvs, it is necessary to use more equations to relate the
variables. This can be done by writing separate conservation laws for
the componcnts, rather than by using only the equations for the entire
mixture. Thus increased accuracy is bought at the price of increased
complexity.
At this stage it is convenient to keep these equations in a very
general form which will provide the basis for a unified treatment in later
chapters. This will involve the introduction of symbols and concepts
which w:ill probably appear rather abstract, with a meaning which is
difficult to grasp until more specific cases are treated. Examples will
be included in order to help to make this meaning clear.
CONTJNUITY EQUATiONS

The equations of conservation of mass, or continuity equations, can be

written quite generally in differential form:

ata [p1(l - a)] + V · (p1(l - a)v1] = S12 + S1 (3.38a)

a
at (p,a) + V • (p,av,) = -S12 + S2 (3.38b)

Sú is a source term which represents the mass rate of phase change per
unit volume. S1 and S2 are external sources of matter and are almost
invariably zero.
In one-dimensional form, Eqs. (3.38) become, after integration
across the duct,

(3.39a)

(3.39b)

Corresponding integral versions of these equations can be derived in

exactly the same way as they are in the usual single-phase flow cases.
MOMENTUM EQUATIONS

The momentum equations, equations of motion, or N ewton's law for

the two phases can be written quite generally 1n three-dimensional
vector form:
0 1
P1 ( ~
0
+ v1 • Vv1) = b1 + f1 - Vp (3.40a)

p, (°; 2
+ vv,) = b, + f, -
V2. Vp (3.40b)
52 ONE-DIMENSIONAL TWO-PHASE FLOW

b1 and b2 are the body forces, per unit volume of that component 1 which
act on each component. Vp is sorne average gradient of pressure or bulk
stress which is suitably defined and is usually the thermodynamic pressure
of one or both of the fluid phases. f 1 and f 2 are simply "what is left
over." Since Eqs. (3.38) and (3.39) are merely force (or momentum)
balances, the f's are introduced to keep the accounts straight and the
way in which they are evaluated will depend on the particular flow regime
and conditions of the problem.
For example, if Eqs. (3.38) and (3.39) refer to elements of incom-
pressible newtonian fluid containing only one component which is not
undergoing phase change, the usual results of viscous fl.ow are obtained,
thus
f1 = µ.1V 2v1 (3.41a)
f2 = µ,2V 2V2 (3.41b)
To obtain a quasicontinuum model of two-phase flow 1 one usually
calculates the f's using an element o! the flow which is larger than the
particles, drops, or bubbles that occupy the flow field. In this case the
fs represent the average total surface forcei per unit volume, that is not
contained in the pressure gradient. The f's contain components due to
hydrodynamic drag 1 apparent mass effects during relative acceleration 1
particle-particle forces, forces dueto momentum changes during evapora-
tion or condensation 1 and so on. The evaluation of the f's in specific
cases often requires considerable care. For example, most "drag" forces
on particles suspended in fluids are deterrnined experimentally and con-
tain both the effects of Vp and f; in addition, phase change modifies
hydrodynamic drag, while bubble or droplet shapes, and hence drag
forces, are altered by the surrounding force field.
F1 and F, are defined as the equivalent of the f's per unit volume of
the whole flow field. Thus
F1 = Ji (1 - a) (3.42)
F, = J,a (3.43)
If these forces are entirely due to mutual hydrodynamic drag, action and
reaction are equal and we have
F1 = -F 2 = F 12 (3.44)
In the case of one-dimensional flow, Eqs. (3.38) and (3.39) are
resolved in the direction of motion to give

Pl (ªVi
at + V1 ~"_i
az ) = +
b1 Íi - ap
az (3.45)

P2 (ªv,at + v, ª"az2) = b, + !2 - ap
az (3.46)
SEPARATED FLOW
"
Example 3.4 In annular flow, liquid flows as a film on the wall of a pipe w-hile gas
flows down a central cylindrical core. For a pipe of diameter D let thc ínter-
facial and wall shcar strcsses be Ti andrw. Assuming symmetrical vertical flow
with the positive dircction mcasured upward, derive the values of Ff, F 0 , ji,
f¡¡, bJ, and b0 and hcnce the equations of motion.
Solution The diameter of the gas corc is D ~ ' where a is the void fraction.
Therefore
4Ti,v;;
--D-

and

Thc corresponding valucs of ff and fe are

A1so

bJ = -Pfg

and

The equations of motion are then, from Eqs. (3.45) and (3.46),

- -p¡g + D(! 4__ a) (r, y';; - Tw) - :

ap
D v: - az
4Ti
-p¡¡g -

IExample 3.5 The Carman-Kozeny equation for the frictional"prcssure drop during
viscous flow through a packed bed of spheres with diamctcr d and void fraction
" is

_ (ªP)
az F
= lSO µJi Jo
d2
(1 - e)
eª
2

where }¡o is the fluid flux relative to the par ti eles. E has been chosen to rcprcsent
the liquid fraction bccause a is more commonly used to denote the volumetric
fraction of the discontinuous phasc. Deduce the values of J1 and f.. The
subscripts f and s refer to the fluid and thc salid, respectively. What are the
componcnts of fs due to the fluid and the particles?
Solution Sincc wc are only concerned with frictional pressurc drop, the inertia and
gravitational terms are dropped from Eqs. (3.45) and (3.46) to give

1 -f, - (ªP)
az F
 ONE-DIMENS!ONAL TWO-PHASE FLOW
"
The force f., on the particles is rnade up of two parts, one duc to the fluid, (f.) 1 ,
and the other dueto thc particles, (f.).. Thus

The mutual force between the fluid and the particles obeys Eq. (3.44).
Therefore

Combining these relationships we find

180µ¡j¡o (1 -
Ít ~ - --d,- - - , , -
~v
(!) = 180 µ¡j¡o I - "'
." 1 d2 "'2

(f,,), = -180 µ1{º I ~ E

Once the values of the f s have been determined, thcy can be used to salve more
complex problems (such as those encountered in :fluidization and soil compac-
tion) in ,vhich other terrns in Eqs. (3.45) and (3.4G) are not negligible.

Example 3.6 In a fiuidized bed particles are supported by an upward flow of fluid
around them and interparticle forces are negligible. Use the results of the
previous example to deduce thc fluid flux necessary to cause fluidization in a
bed with void fraction E. What is the pressure gradient through the bed in
this case?
Solution The condition for fluidization is

Use Eqs. (3.45) and (3.46) without the inertia terms to obtain

- dp - p¡g - ISOµ¡j¡o (I - e)2 = O

dz d2 1c 3

- dp - Ps(] + ISO µ¡}¡o I - e = O

dz d2 e2

Solving these equations simultaneously we get

. d2g(p. - p1) eª
J /O = 180µ¡ 1 - E

dp
- dz ~ g[,pt + (1 - ,)p,]

The pressurc gradient is simply equal to the weight of thc bed, solids plus fluid,
per unit depth of the bed.

3,4 FLOW WITH PHASE CHANGE

The general momentum cquations for one-dimensional separated flow in

a duct in the presence of phase change will now be devcloped. To be
more specific about the f's, !et the drag forces from the duct walls on
components 1 aud 2 be Fw1 and F wz, per unit volume of the flow, rcspec-
SEPARATED FlOW
"
tively. Let the drag force betwcen the components be F12 acting on
component 1 in the direction of motion and in the opposite direction on
component 2. Since the two components have differing velocities, any
phase change will result in a change of momentum. The mass rate of
phase change per unit Iength is W dx/ dz and the velocity change is v2 - v1 •
Therefore the force that is necessary to account for momentum increase
due to phase change per unit volume of the duct is
Wdx dx
(v 2 - v1) A dz = (v 2 - v1)G dz (3.47)

It is not clear at this juncture how much of this force is to be charged to

stream 1 and how much to stream 2. In fact, this assignment will depend
on the process (e.g., boiling or condensation) and on the interaction
between the pbase change and the hydrodynamic mechanism which gives
rise to F12- For example, the drag force on an evaporating droplet will
depend on its rate of vaporization.
To be quite general, therefore, we ascribe a fraction r¡, of the force
represented by Eq. (3.47), to stream 2 and a fraction 1 - ry to strcam l.
The choice of ~ may be different for different systems.
The two momentum equations of steady flmv are now, if gravity is
the only body force,
dv1 dp 1 - ~ dx
p1V1 dz - dz - p1g cos 0 + F121 -_
F.1
a -
1 _ a (v2 - v1)G dz
(3.48)
dv 2 dp F 12 + F w2 ~ ) dx
p,v, dz = - - - p,g cos 0 - - - - - - - (v, - v1 G - (3.49)
dz a a dz
Equations (3.48) and (3.49) can be combined in severa! ways. The
following two results are particularly convenient.
Multiply (3.48) by 1 - a and (3.49) by a and add. The result is
the equation of rnotion for the mixture and could have been obtaincd by
considering the momentum balance for both components taken together,
thus
dv1 dv, dp ( ) ]
(1 - a)p1V1 dz + ap2V2dz = - dz - g cos 0 [p1 1 - a + p 2a
dx
- (Fwl + F w2) - (v2 - V1)G dz (3.50)

Furthermore, the last term on the right-hand side can be combined

with the left-hand side, using Eqs. (3.2) through (3.5) and (3.47) to give
1 d dp
A dz (W1v1 + W,v,) - dz - g cos 0 [p1(l - a) + p,a]
- (Fwl + Fw2) (3.51)
which is plainly the momentum balance for the mixture.
 ONE-DIMENSIONAL TWO-PHASE FLOW
"
If, on the other hand, Eq. (3.49) is subtraeted from Eq. (3.48), the
result is
F wl
g cos 0 (P2 - Pl) - ---
1 - a
+ -Fa
wZ

+ a(lF12- a)
-Gdx(l-~_i)(v,-v1)
dz 1 - a a
(3.52)

This equation does not inelude the pressure gradient and could be thought
of as the relative motion equation since it describes the difference between
the rates at which the two phases are gaining kinetic energy.
Usually only one component is in contact with the wall and the
appropriate wall shear stress can be obtained from a correlation. The
drag force between the components is a function of the relative velocity
and can also be estimated. However, the detailed solution of the result-
ing equations is often quite a formidable problem.

Examp!e 3,7 Use the Martinelli correlation scheme for gas-1iquid flow to deduce the
values of F Ju and F wf, if F wo is assumcd to be zero duc to the liquid ,vetting thc
cntire wall of the duct (annular flow).
Solution 1\fartinelli's correlation represcnts a balance between the F's and the pres-
sure gradient, since it <loes not include the effects of inertia, gravity, or phase
change. From Eqs. (3.48) and (3.49), therefore,
F1u dp
Fwl =~ = - dz

It appears most reasonable to express the force on the liquid in terms of liquid
properties and the force on the gas in terms of gas properties. Using Eqs.
(3.24) and (3.25) we therefore have

Fw1 = 11(!:)
2

F1g = a1(! (!~) 2

g

If 1 1 2 and 1i are given as functions of a by Martinclli's correlation, F wf and

F¡q can be expressed in terms of known parameters, leaving a to be determincd
by solution of Eqs. (3.48) and (3.49), even when gravity and inertia are not
negligible.

The momentum equations can be combined with the continuity

equations to give results which parallel corresponding developments
which are common 1n gas dynamics. Consider first component 2.
Write the identity
dv2 d(p2v2) dp,
P2V2 dz = V2 --------¡¡;¡- - V2
2
dz (3.53)

K ow, the change in the density of phase 2 will be related to the change
of pressure and the particular thermodynamic path which the components
SEPARATED FLOW 67

are following. For example, in the two-phase region, for apure substance
in equilibrium, the densities are only a function of pressure. For p,article-
gas systems a ''virtual equation of state" may be appropriate, as discussed
in Example 2.2.
In any case we may write

(3.54)

where it is understood that ap/iJp, is evaluated for the prevailing

conditions.
Define a pseudo-sonic velocity Cz for component 2 by the equation

Cz
2
= -iJp (3.55)
iJp,

Combining Eqs. (3.53), (3.54), and (3.55) and using the result in
Eq. (3.49), we get, alter rearrangement,

+ p,g cos e + +a Fw2

2
dp ( v2 ) d F12
- dz 1 - e,' ~ v, dz (p,v,)

+ -a~ (v v,)G ~
dx
2 -
dz
(3.56)
Equation (3.3) is now rewritten as
Wx = Aap2V2 (3.57)
and is differentiated to yield
1 dx _ 1 dA + 1 da + 1 d(p 2v,)
(3.58)
x dz A dz a dz p2V2 dz
Eliminating d(p,v,) /dz between Eqs. (3.56) and (3.58) then gives the result
dp
dz
1
pzvz 2
(i v,') _c2 2 -
1 dx
x dz
1 da _ 1 dA.
a dz A dz
1 [ p g cos 0 + F + F + -n (v,
+ --, 2
12 w2
- v, ) G dx]
~d (3.59a)
pzVz a a z

The similar equation for the other phase is

dp 1
- dz p 1v1 2
(i -
v1 ') _
c1 2 - -
1 dx
1 - x dz
+ 1 da 1 dA
1 - a dz - A dz
1 [ p g cos 0 -
+ -- P1V1 2
1
F12 -
1 -
Fw1
a
+ -11 --- -~a (v, - dx]
v,)G~d
z
(3.59b)

The above equations resemble the general one-dimensional steady flow

equations of single-component flow, except for the effects of phase change
and the additional degree of freedom which is introduced by the variable
a. "Choking" occurs in the individual components when V1 = c1 or
 ONE-D\MENSIONAL TWO-PHASE FLOW

V2= c2, but this <loes not nccessarily correspond to the "compound
choking" of the combined flmvs since a is free to adjust to local conditions.
In order to investigate compound choking ,ve elimina te da/ dz
between Eqs. (3 ..57) and (3.58). The result is

- :r: [p,:,, (1- ~::) + \~/ (1 - ~::)] ~ - ~ :1

+ g cose(-"'-+
2
V2
1- ") F (-1- __1_) Fw
V1
2
+ 12
p2V2
2
p1V1
2
+ 2
P2V2
2
+ Fw 12
p1V1

+ dx [" - 1 - a + G(v, - v1) (1 - 2~ + ____'I__,)] (3.60)

dz x l - x p1V1 P2V2

As long as F12 and dx/dz are independent of the pressure gradient,

the choking condition is

(3.61)

Since all the factors except those in parentheses are positive 1 it is evident
tha.t one of the ratios 1VJ 2 2 = v22/c 2 2 and llI 1 2 = v1 2/c1 2 must be less than
unity ·while the other is greater than unity. Thus one stream is super-
sanie while the other is subsonic.
The choking condition is considerably modified if flashing (phase
change as a result of pressure change) is significant. This effcct intro-
duces a dependence on pressure gradient in the terms on the right-hand
side of Eq. (3.60) which invol ve dx/ dz.
In general, if the thermodynamic path is known (ar assumed), then
ax/ ap can be evaluated and we may write
dx ax dp (3.62)
dz ~ ap dz
Combining Eq. (3.62) with Eq. (3.59a) the choking condition in the
presence of flashing is found to be

----"-2 (1 - M,')
P2V2
+ 1P1V1
- "(1 - M12
2)

_ ax [" _ 1- a + G(v, _ vi) (1 - n + ____'I__,)] (3.63)

ap X 1 - X p1V1 P2V2 2

This result depends on the value of n except in the particular case where
(3.64)

3.5 FlOW IN WHICH INERTIA EFFECTS OOMINATE

In certain cases in which a two-component separated flow is accelerated
rapidly through a nozzle, the inertia and pressure-drop terms dominate
SE?AílATED FLOW
••
Eqs. (3.48) and (3.49) entirely. In this case thc velocity changes of the
components can be related. In steady one-dimensional flow without
phase change, for example, if the f's are zero, we have
dp dv, dv,
- dz = p¡V¡ dz = v dz
p2 2 (3.65)

In particular, if both components start with a low velocity and density

changes are small, their final velocities after expansion are related by
the simple expression

~= ✓~ (3.66)

This expression has no universal validity but is approximately true under

conditions of rapid expansion at low Mach numbers.

Example 3.8 Air and water flow from a large tank through a converging nozzle
having an exit area of 1 in. 2 • The flow rates measured at atmospheric pressure
are 2000 in. 3 /sec of air and 27 in. 3 /sec of water. What are the pressurc in the
tank and the exit velocities of the cornponents if the externa! pressure is atmo-
spheric (14.7 psia) and the temperature is 70ºF? Neglect wall shear stresses.
Solution The problem as stated above is indeterminate becáuse of the lack of detail
about thc method of mixing the components and the resulting flow pattern.
The best that can be done is to perform a limiting analysis considering the two
extreme cases: (1) no forces acting between thc components (i.c., Ji = h = O);
(2) la.rge forces suppressing relative motion (i.e., homogeneous flow).
Making assumption 1, wc use Eq. (:3.65) and first assume negligible
density change for the air (this can be taken care of later if necessary). Inte-
grating Eq. (3.65), we get
2 2
!ip = P1V1 = p2V2
(3.67)
2 2
Let component 2 be thc air and let its volumetric conccntration in the exit
of the nozzle be a. In the usual wa.y we have
j,
V1 = 1 - a (3.68a)
j,
V2 =- (3.68b)
a

Substituting Eqs. (3.68a) and (3.68b) into Eq. (3.67) gives

2
Pd1 Pd2 2
(1 - o:)2 a2
(3.69)

whence

1 - a =
a
J;!.
J2
('-')½
p2
(3.70)

Therefore

(3.71)
 ONE-DIMENSIONAl TWO-PHASE FLOW

For this particular problem

j1 = 27 in./sec j2 = 2000 in. /sec
p, -62.5lb/fl' e, - 0.076 lb/ft'

Therefore
1 1
ª - 1 + 2 ½000(62.5/0.076)), - 1.388 - 0.7Z
and 1 - a = 0.28.
Substituting these values in Eqs. (3.68a) and (3.68b) we get

Vi = O~:S = 96.5 in./sec V2

2000 .
= 0 _72 = 2780 m./sec

The pressure drop, from Eq. (3.67), is then

62.5 (96.5)' 1 O 43 .
"-P - (32.2)(2) 12 144 - · 5 psi,
or, alternatively (as a check),

0.076 (2780)' .
(2) (32.2) 144 - 0.4 39 psi

Using assumption 2 we havc v 2 = v 1 = j = 2027 in./sec. Since there

exist forces bctween the components in this case, we must use the momentum
equation for both components taken together, i.e., Eq. (3.51), which, for the
homogeneous flow case with no area change, reduces to

dv-_dp (3. 72)

0
dz dz
or, using Eq. (2.32),
dv dp
p.,.vdz = - dz (3.73)

Integrating, and assuming incompressible flow for the moment, we get

1
2 p.,.v2 = !ip (3.74)

which is simply Bernoulli's equation applied to the homogeneous mixture.

Expressing Eq. (3.74) in terms of j and G we find
jG (2027)1(2000)(0.076) + (27) (62.5)]
D.p - 2 - (32.2)(144)'(2)
= 2.8 psi
Note the lar ge difference between the answers which are predicted by
the two methods. Since it is possible to ach.ieve almost any value of pressure
drop in this range by suitable design of thc way in which the fluids mix, it is
evident that there is often great latitude for design of two-phase fl.ow apparatus.
For instance, if one· were interested in achieving minimum pressure drop, for
economic reasons, it would be advantageous to allow the air and water to
separate in the tank and ensure that the interface was on the levcl of thc nozzle
so that a stratified flow could take place with little friction betwccn the com-
ponents. On the other hand, if a large pressure drop is desired (for instan.ce,
to restrict leakage from the container) it is bctter to mix the components
SEPARATED FLOW 71

thoroughly and form a mixture which would be essentially homogeneous. It

is unlikely that bubbly flow could be stable at such high void fractions. A
droplet dispersion would be needed. The air vclocities derive4 in this example
are actually barely suffi.cient to produce e:ffective atomization.

Example 3.9 Show how to analyze the characteristics of an isentropic, adiabatic

cpnvergent-divergent nozzle which is to carry a stratified flow of air and water.
Neglect friction, phase change, and droplet entrainment. In particular, show
how the maximum possible air flow rate is related to the rate of water discharge
for given upstream conditions and a fixed throat geometry.
Solution The maximum rate of discharge will be governed by compound choking
at the throat.
Let the upstream stagnation conditions be po¡ for the liquid and p 0 g 1
T og for the gas.
Since the water is incompressible, it will obey Bernoulli's equation and
its velocity at the place where the pressure is p will be

(3.75)

The corresponding flow rate per unit area will be

(3 76)

Differentiating Eq. (3.76) we obtain, since W 1 is constant,

(3.77)

The gas will expand isentropically in a nozzle which has an area variation
determined by the total area minus the liquid area. In terms of the local
pressure this area is given by the well-known equation 8

(3.78)

where -y is the isentropic exponent for the gas.

In principle, Eqs. (3.76) and (3.78) contain all that is necessary to predict
the relationships between flow rates and overall area for the passage. At the
throat, under choked conditions, the gas is supersonic and the liquid subsonic.
The overall qualitative appearance of the arca variation with length is shown
in Fig. 3.14. Note that the stagnation pressurcs of the two streams are not
necessarily identical, although the local pressures are assumed to be the same.
The equation corresponding to Eq. (3.77) for thc gas is, from Eq. (3.59a),

(3.79)

Assuming air to be a perfect gas we can replace pgvg2 by -ypM 0 2 in the

usual way to get

_I__ dA, _
Ag dz -
(-¡-
JJ;Ji
_i) 'YP
_l__ dp
dz (3.80)
72 ONE-DIMENSIONAL TWO-PHASE FLOW

Fig. 3.14 Separated gas-liquid flow through a converging-diverging

nozzle under chokcd conditions.

Now, at the throat dA/dz is zero. Therefore

dA¡ + dA, ~ O (3.81)

dz dz

From Eqs. (3.77) and (3.80), therefore

A ( I ) sAr
u Mi - I + 2(po¡/p - 1) = O (3.82)

which is another form of Eq. (3.61).

J\foreover, if A* is the throat area we also have

A,+ A 0 =A* (3.83)

Eliminating A 0 from Eqs. (3.82) and (3.83) gives

(3.84)

In addition we have, from Eq. (3.76),

(3.85)

Equations (3.84) and (3.85) can now be solved simultaneously to obtain

the pressure at the throat and the area of thc liquid stream. The othcr vari-
ables can then be deduccd by substitution in the appropriate equations.
A method for visualizing the result graphically is shown in Fig. 3.15.
Equations (3.84) and (3.85) are plotted to give A 1 as a function of p. Equa-
tion (3.84) is independent of the liquid flow rate. M" is only a function of the
pressure ratio for the gas stream, as follows:

(3.86)
 SEPARATED FLOW

p
Pg'

fig. 3.15 Simultaneous solution of Eqs. (3.84) and (3.85)

of Example 3.9.

A 1 is equal to zero in Eq. (3.84) where kl/ = 1, and this corresponds

to choking of a fl.ow of gas alone. As Mu is increased abovc unity, p decreases
and A¡ increases as shown in the figure.
Equation (3.85), on the other hand, shows A, increasing with p. The
intersection then moves to the left as W 1 is increased. The limit occurs when
liquid fills the duct entirely, the throat pressure is zero, and there is no gas flow.
Knowing p and Au from Fig. 3.15, we can calculate the gas flow rate
from Eq. (3.78).
The simplest procédure for deriving the locus of maximum flow rates is
to use the throat pressure as a parameter. First calculate A 1 from Eq. (3.84).
Then use Eq. (3.85) to get W 1 . Use Eq. (3.83) to find A, and Eq. (3.78) to
compute wg.
The allowable flow rates will then lie to the left of the lines shown in Fig.
3.16, where the boundary is calculated as indicated above.
The method of solution can be compared with similar methods of treat-
ing separated compound gas flows. 9

3.6 USE Of THE COIIICEPT OF ENTROPY GENERATIOIII

TO EVALUATE THE COEFFICIEIIIT n

Up to now we have deliberately left the quantity n, which apportions the

effects of phase change between the components, as a variable which can
be selected according to the conditions of interest. A particularly in-
triguing development is based on the requirement that the flow should be
isentropic. 10 We shall see that this enables us to assign a definite value
to r¡. For convenience in description we shall refer to the phases as
1
' liquid" and "vapor" when it is necessary to give physical significance

to the results.
" ONE-DIMENSIONAL TWO-PHASE fLOW

40

'-"• 20 Hi-H----f---\-----/'+---\--------+c

±'l':z

~
101-----+'f------,,!'--\---+-_,.----'\~-~
~o~. 1,
o
-o
~o
ÜL---~--~-"----~--'~~-~~--~
O LO 2-0 3.0
~ ' lbm/(sec)(in.2 )

Fig. 3.16 Relationship between air and water flow rates for strati-
fied critical flow with equal stagnation pressures (p 0¡ = p 00 ).
(D. A. Sullivan, Dartmouth College.) k = 1.4, R 0 = 53.3 (ft)(lb)¡/
(lbm(°R), T 0 =·530ºR, PJ = 62.4 lb,,./ft3. (lVIª)i, a. 1, and A* are
the values of the gas-phase Mach number, void fraction, and
total area at thc throat undcr conditions of compound choking.

First, write the energy equation for the two phases taken together,
neglecting any shaft or shear work, as follows:

_-1:_ dq,
W dz
= _rJ,__
dz
[xh, + (1 - x)h, + ),~xv,' + ),f(l - x)v,'] + g cos 0
(3.87)
SEPARATED FLOW 75

If Ss denotes a sourcc of entropy per unit length caused by irreversi-

ble processes and heat transfer, an entropy balance for steady one-
dimensional flow gives

d
S, - W dz [xs 2 + (1 - x)s 1 ] (3.88)

Equation (3.51) is rewritten by using the definition of x to yield

d dp
G dz [xv, + (1 - x)v 1 ] - dz - g cos 0 [ap 2 + (1 - a)pi]
- (Fwl + F w2) (3.89)
Now, from thermodynamic theory it is known that, for a homo-
geneous pure substance in equilibrium,
1
T ds - dh - -dp (3.90)
p

For a two-phase mixture, moreover,

+ (1 + (1 X 1 - X)
T d[xs, - x)s1] - d[xh 2 - x)hi] - (-P2 + -P1- dp
(3.91)
Although Eq. (3.91) is usually deduced far thc particular case of a
homogeneous mixture, it is an algebraic relationship between thermo-
dynamic properties which is valid far any change of state far which the
phases remain in equilibrium. Therefare, we may divide Eq. (3.91) by
dz, substitute from Eqs. (3.87), (3.88), and (3.89), and make use of the
relationships between x and a to eventually obtain

TS, - ªJi' + a(l - a) (v 2 - ¡

v1)A g cos 0(p1 - p2)

+ ½[ G(v2 - v1) (i ~ a - ~) !: - P2 ª;;' + PIª;;']}

+ GA(Fw1 + Fw2) (-"'-P2 + 1 Pl- X) (3.92)

Furthermore, the final term in Eq. (3.92) can be rearranged as fallows:

GA(Fw1 + Fw,) (-"'-P2 + l P1- X)- A[av2 + (1 - a)v1](Fw1 + Fw,)

- [v1 + a(v, - v1)]AFw1 + [v, - (1 - a)(v, - V1)]AFw2

v1AFw1 + v,AFw, + Aa (1 - a)(v v (i F:._

1 2
- F;
2 - 1) " - )

(3.93)
 ONE-DlMENSIONAL TWO-PHASE FLOW
"
Combining Eqs. (3.92) and (3.93) yields

TS, = ~~• + v,AFw, + v,AFw,

+ a(l - a)(v, - v,)A (g cos 0(p 1 - p 2) + 1Fwi
- a
- Fwz
a

+ !2 [G(v, _ v,) (-1- _!)adzd_x_ _ p, dv,dz + p, dv,']} 2

(3.94)
1-a dz

Equation (3.94) expresses four mechanisrns of entropy generation:

l. Heat transfer
2. W all shear on the Iiquid
3. Wall shear on the gas
4. Relative motion

The entropy generation dueto relative motion is a new phenomenon

to those who are used to working with single-phase flows and is often
overlooked. It is represented by the final term in Eq. (3.94) and is zero
only if one of the factors in that term is zero. Taking the factors in
arder, we have the following conditions for zero entropy generation due
to relative motion:

" = o i.e., single-phase Iiquid

(1 - a) = O i.e. single-phase gas
1

i.e., homogeneous flow

The only nontrivial result is found if the square bracket is equated to

zero and rearranged to give
dv, + + 1 G(v, - v,) dx + F w1
P1V1 dz QP1 cos 0 2 1 - a dz 1 - a

= P2V2
dv, +
~ QP2 COS 0
+ -1 G(v, - v,) -dx + F-w,- (3.95)
dz 2 a dza

To discover the physical significance of Eq. (3.95), we let both sides equal
A, take (1 - a) times the left-hand side and add it to a times the right-
hand side. The result is
dv, dv 2 ) dx
A = (1 - a ) p 1v1 dz + ap,v 2 dz + (v, - v, G dz
+ g COS 0 [(1 - a)p1 + ap,] + Fw1 + Fw, (3.96)
Comparison with Eq. (3.50) reveals the simple result that

(3.97)
SEP/i,RATED FLOW 77

Equation (3.95) is therefore equivalent to the two equations, which are

equations of motion for each phase, in the absence of interfacial shear,
as follows:
dv, _ dp _ pig cos " _ F wl _ l G(v 2 - v 1) dx (3.98)
piv, dz - dz v 1 - a 2 1 - a dz

dv, - dp - p,g cose - Fw, _ l G(v, - v,) dx

p,2.v 2 dz = dz a 2 a dz (3.99)

Thus it is seen that, in arder to conserve entropy, the force represented

by Eq. (3.47) has to be shared equally by the components, each of which
experienccs a reaction -7~G(v, - v,)(dx/dz) per unit volume of the duct.
Since the terms in the equations of motion represent forces per unit
volume of the specific component, this force must be divided by the
volumetric concentration of each, as appropriate.
For isentropic flow, therefore, the value of ~ should be chosen to
be 7f The above equations are equally valid for evaporation or con-
densation and are perfectly symmetrical. At first sight it may appear
very strange that the liquid (or solid) and the vapor should share the
reaction which is dueto vaporization. ::.\1ost authors are wont to attrib-
ute all of this reaction to the vapor stream during evaporation and to
the liquid stream during condensation. The explanation 1 however, should
be sought in the two-dimensional effects which have been obscured by
thc one-dimensional idealizations. In fact, the local velocities of the
liquid and vapor streams tangential to the interface will be equal and
there will be velocity variations over the flow field. If solutions could
be obtained to the detailed three-dimensional flow problem, this would
give a more complete representation of reversible evaporation and
condensation.
The conclusion that the force represented by Eq. (3.47) should be
shared equally between the components could also have been derived
from consideration of reversibility. Unless the force is shared equally,
the equations differ in the cases of evaporation and condensation and the
fluid cannot be caused to return to its initial state by reversing either
process in a symmetrical way.
Colloquially this result could be expressed as follows: "If you are
going to use a one-dimensional flow model, and you are going to require
that there be no entropy generation, then you must include the final
terms in Eqs. (3.98) and (3.99) in order to keep al! the relevant equations
consistent.''
N ow, suppose that there is irreversible momentum transfer taking
place between the vapor and liquid streams. This could be imagined
to resemble the process proposed by Osborne Reynolds 17 to describe
turbulent transfer phenomena involving M units of mass per unit volume
 ONE-DIMENSIONAL TWO-PHASE FLOW

per unit timeJ which alternately contact the vapor and liquid and share
momentum with each. The resulting force on the liquid will be

F12 - (v, - v,)M (3.100)

in the direction of motion. The corresponding rcaction on the gas is

-F12 - -(v, - v,)iVI (3.101)

The total force on the liquid per unit volume of the flow field, resulting
from the relative motionJ assuming that the two processes of reversible
and irreversible momentum transfer can be superposed linearly, is

(v, - v,) (M - 72G~:)

and on the gas, similarly,

(v, - v,) ( -M - .l-2G ~:)

An alternative description can be developed in terms of mass and
momentum fluxes at the two-phase interface. If the interfacial perimeter
is Pi in a pipe of arca A, then the surface area per unit volume is P/ A.
If , 0 is defined to be the irreversible Reynolds fiuxt per unit area of the
interface, then we have
MA (3.102)
Eo = pi
The rate of vaporization per unit area of the interface is
GA dx (3.103)
m - - - = m12 = -m21
P, dz
The net forces on the liquid and vapor per unit volume can then be
rewritten as

F, (v, - v,) ~(,o-%;) (3.104)

F, - -(v, - v,) ~(,o+;) (3.105)

Dividing Eqs. (3.104) and (3.105) by 1 " and "respectively, to

obtain the force per unit volume of each component, and incorporating
t The concept of Reynolds flux was first proposed by Osborne Reynolds 17 in 1874.
It has been used by Nusselt, 18 Silver, 19 - 21 and extcnsively by Spalding 11 to analyze
problems of combustion and simultaneous heat and mass tra.nsfer. Applications to
two-phase fl.ows are comparatively recent. 10 , 22 - 24 The subjcct is sadly neglectcd
in many textbooks.
SEPARATED FLOW
"
the result in the equations of motion (3º98) and (3º99), wc obtain
dp Fw1
- - - p¡g cos 0 - - -
dz 1 - a

- A(/~ a) (v, - v2) (,o - ;) (3º106)

- -dp - pzg CDS 0 - -Fw2 - -P, (V2 - V1
) ( + -m)
Eo
dz a Aa 2
(30107)
These equatiüns are quite symmetrical and are valid for phase change in
either direction, dcpending on the sign of ni. The last term in each
could be regarded as representing a modified interfacial shear stress (and
momentum transfer) as a result of phase changes. An alternative deriva-
tion of this same result will be given in Chapo 7º
The interpretation of these results is interesting from the point
of view of irreversible thermodynamics. The term involving m involves
orderly mass transfer and does not contribute to the entropy production.
The Reynolds flux EoJ however, is a measure of inherent disorderly mass
exchange, irreversibility, and entropy productionº Use of Eqsº (3º106)
and (3º107) in Eqº (3º94), far example, yields the result

TS, = 1; + v1Fw1 + V2Fw2 + (v, - v,) 2eoP, (30108)

Note that the terms involving wall shear stress in Eqº (3º108) are
not necessarily positive. In vertical slug flow, for cxample, the second
term can be negative while the third term is zeroº The second law of
thermodynamics, hmvever, requires that the final term should be Iarge
enough to ensure that the net dissipation is positive.
It is likely that a similar analysis could he performed to account
for entropy production due to energy and mass transfer between the
phases across finite differences of temperature and Planck potentiaL
Such a treatment would extend the methods o! Spalding 11 to two-phase
flow and could provide a very powerful technique far dealing with multi-
component chemical reaction, combustion, and other practically impor-
tant prohlemsº The Reynolds flux concept would provide the key to a
unified trea tment.
The last term in Eqº (3º108) could be regarded as a force (per unit
volume) 1 (v2 - v1)E0Pi, times a finite velocity difference, V2 - v1, which
agrees with the usual ideas of entropy production in linear systems.
Since cntropy production due to interphase friction (in a flow with
initially equal velocities) is zero in the extremes of infinito Eo (homoge-
neous flow) and zero Eo 1 it will evidently be a maximum somewhere in
between. For ex&.{nple, in the case of the expansion from stagnation
 ONE-DIMENS!ONAL TWO-PHASE FLOW

conditions of tvvo incompressible fluids with no vvall. shear stress and

negligible phase change or gravitational effects, entropy production is
zero when either v,/v 2 ~ 1 ar v,/v, ~ (p 2/p 1) 1', from Eq. (3.66).

3.7 ENERGY EQUATIONS

General ene-dimensional energy conservation equations can be written
for the two phases by considering a control volume of length dz in the
duct. Far component 2, for example, we have

:t [Aap, (e + "f)] + :z [Aap v (h + v 2 2

2
2 )]

~ (q, + qm - w, - w,)A - Ap,aV2(/ cos e (3.109)

The symbols qe, qm, -ws, and -w.,. represent average rates of energy
addition per unit volume of the duct due to heat transfer, mass transfer,
shaft work) and shear work. Since these processes may all occur both
between the components and between each component and the duct wall,
the general formulation becomes extremely cumbersome. Since this
text is concerned primarily with fluid mechanics1 rather than with heat
and mass transfer, we shall avoid a general discussion and development
of tbe energy equations. Simplified versions will be introduced wben
they are necessary and appropriate.

IExample- 3.10 What is the one-dimensional equation governing interna! energy

changes for steady flow of particles in a gas stream? Neglect mass transfer,
heat transfcr, and friction with the walls and shaft work.
Solution Let thc particles be component 2. Since the flow is steady and there is no
mass transfer
(3.110)
Equation (3.109) then reduces to

p20:v2 [ d~ ( h2 + v~2) + g cos O] = q, - 'Wr (3.111)

Since the particles are incompressible and gravity is the only body force,
Eq. (3.46) can be put into the form
d e,'
p2 [ dz 2 + g cos 0 + dzd (pvz) ] = fz (3.112)

Combining Eqs. (3.111) and (3.112) and using the thermodynamic relation
+ pvz, we get
h2 = e2

p2cxV2
de,
( dz
h)
+ P2 = q. - Wr (3.113)
or
de2 _ q. - Wr - F 2V2
(3.114)
dz - p2aV2

in view of Eq. (3.43).

SEPARATED FLOW
"
If the mechanical work terms, w,.. and F 2v 2, are small compared with the
heat-transfer term q., there is no difliculty in evaluating dez/dz as long as q. can
be determined from heat-transfer theory. In particular, if the heat-transfcr
coeflicient between thc particles and thc fluid is h, the temperatures are T 1
and T2, and the particles are spherical with diameter d and a constant specific
heat c2, we have

(3.115)

and, therefore
dT, 6h(T, - T,)
(3.116)
dz Gzp2V2d

In general, w,.. and Fzv2 are not necessarily small, nor are they equal and
opposite, as the following example shows, and thc proper evaluation of the
energy relationships can be quite tricky.

Example 3.11 Consider one-dimensional steady flow with no mass transfer. Assume
no shaft work or heat transfer, but let there be a force F 12 acting between the
components per unit volume of the flow .field (conventionally acting on compo-
nent 1 in the direction of its motion). Discuss the energy equations for the
components, neglecting gravity.
Solution The only term surviving on the right-hand side of Eq. (3.109) and its
equivalent for the other phase is the shear-work term and wc obtain, under the
speci.fied conditions, apparently

(3.117)

(3.118)

Now, the rate at which component 2 is doing work is F 12v2. Howcver, compo-
nent 1 only receives shear work amounting to F 12V1. Therefore, due to the
relative motion there is a "dissipation" of work, per unit volume, equal to

(3.119)

This dissipation is equal to a source of entropy times the temperature and

appears as "heat," but there is no way of telling, at this level of sophistication,
how to apportion this energy between Eqs. (3.117) and (3.118) in order to
balance accounts.
One way around this difficulty is to increase the levcl of sophistication
and to considcr a diffcrential analysis which <loes not allow discontinuities in
velocity. This leads to a study of three-dimensional temperature and velocity
variations, "recovery factors," and other considerations which are beyond the
scope of a one-dimensional analysis.
The simpler solution is to apportion the dissipation between the compo-
nents as if it were a heat source. Thus we write

(3.120)

(3.121)
 ONE-DIMENSIONAL TWO-PHASE FLOW
"
r is merely a convenient parametcr which fulfills much the same general func-
tion as nin Eqs. (3.48) and (3.49).
Note that the right-hand side of both Eqs. (3.120) and (3.121) could be
written as a force times a weighted mean velocity, .tv2 +
(1 - !;)v 1 , which
could be regarded as an effective interface velocity.
In cases where hcat, mass, and momcntum transfer occur simultaneously,
a rigorous treatment of the energy exchange processes becomes very difficult.
A convenient concept for simpli:fication purposes is the Reynolds flux which
will be discussed in Cbap. 7.

l'ROBlEMS
3.1. Use a one-dimensional flow control volume approach to derive thc momentum
equations for component 2 in the forro

where F~ is the total resultant surface force per unit volume of the duct acting on
component 2.
Combine this equation with the continuity equation in order to get Eq. (3.40b).
How do you account for the emergence of an apparent component of force due to
phase change?
3.2. Prove that, for the separate cylinders model in laminar flow, n = 2 in Eq. (3.30).
3.3. Prove that, for the separatc cylinders model in turbulent flow, n = 2.5 if the
friction factor is assumed to be constant, whereas n = 2.375 if the Blasius equation
is valid.
3.4. Solve Prob. 2.10 using separated-flow theory.
3.5. Consider homogeneous flow thcory with a constant friction factor. Rearrange
the frictional pressure-drop prediction in terms of the Martinelli parameters. For
various valucs of the density ratio, plot ,:/:,¡¡ versus X and compare with Fig. 3.4.
3.6. Solve Prob. 3.5 using other assumptions about the friction factor in both laminar
and turbulent flow.
3.7. Using the results of Examplc 3.6 and Eqs. (3.48) and (3.49), show how to extend
Martinelli's correlation to vertical flow when incrtia and phase-change effects are
negligible. What will the curves of constant a or constant dp/dz look-like on a graph
of (dp/dz), versus (dp/dz) 1?
3.8. In a straight pipe, instead of using the energy equation in Example 3.1, it is
possible to start from the momentum equation. If the total wa.11 friction force divided
by the pipe cross-sectional area is F, show that

f.p - F = G[xv,, - (1 - x)v¡]

Using the continuity equation, deduce that G is a maximum at a givcn value of F if
v.,/v¡ = (p¡/p¡¡)¼. Furthermore, show that the greatest maximum value of G occurs
when F = O and the flow is isentropic. What is this absolute maximum value of G
in terms of b.p and x?
Compare the predicted values for this maximum with thc predictions of Exam-
ple 3.1 for specific cases of the discharge of initially saturated water. The results
should not be very different.
This model of minimum momentum flux was initially proposed by Fauske. 4
SEPARATED FLOW
"
3.9. Rather than maximizing the discharge in terms of the slip ratio for given momen-
tum or energy fluxes, it is perhaps more reasonable to solve the energy and momentum
equations simultaneously. Do this by plotting the mass flux G versus v 0 /v¡ for both
the energy and momentum methods (Example 3.1 and Prob. 3.8). For a typical
case (e.g., saturated water at 2000 psia expanded to 1200 psia) <loes it make much
difference whether one uses the intersection of the curves or their maxima to makc
predi~tions?
3.10 VVhat is the "best" value of nas a function of pressure in Eq. (3.30) in order to
fit the empirical curves in Fig. 3.17. Express the resultas a correlation of n versus
the reduced pressure (pressure divided by critica.! pressure), a representation which
might justify cxtrapolation to fluids other than water.
3.11. A h0rizontal pipeline, 7.75 in. ID, 11,317 ft long, with an inlet pressure of 1022
psia, was passing 5484 barrels per day of oil and 50,000 lb /hr of gas. The line tempera-
ture vrns S0ºF. The density and viscosity of the oil were 6.499 lb/gal and 0.574 cp,
respcctively. Thc density and viscosity of the gas were 3.48 lb/ft3 and 0.014 cp,
respectively. Determine the two-phase pressure drop in this line. The measured
value was 32 psi.
3.12. (a) Estima.te the critical flow rate for a steam-water mixture, quality 26:9%
by weight, issuing from a horizontal pipe with 0.269 in. ID at a pressure of 125
psia. Compare the value calculated with that determined by Fauske 12 [1265.4
lb/ (ft') (scc)].
(b) Calcula.te the pressure distribution over the last 4 ft of the tube for a flow
of 1265.4 lb/(fV)(sec) and compare this with that measured by Fauske 12 and shown
in Fig. 3.18. The inlet pressure is 350 psi and the inlet quality 19.04% by weight.
Assume a fluid enthalpy decrease dueto heat losses of 2.5 Btu/(lb)(ft) run of tube.
Use thc Baroczy correlation for the frictional pressure-drop component.
3.13. Solve Prob. 2.29 using the Martinelli correlations.
3.14. Solvc Prob. 2.20 using the separated-flow model and suitable assumptions.
3,15. Solve Prob. 2.30 using the Baroczy correlations.

100~-----~----~-----~------

14.7 psia
500
1000
'Pf 101------+-- 1500'-----+------
2000
2500
3000
3206

1.0 !O 100
[x,,]°'sss
Fig. 3.17 Martinelli and Nelson's empirical dependence of cp 1 on pressure and
X for water. 2
 ONE-DIMENS!ONAL TWO-PHASE FLOW

Flow 1265.4 lb/ft 2 -sec

0
300
·a;
o.
io
~ 200f----------1---------+----"--C
~
o.
-~
u
Critico! pressure -
E
V"! 100
EJ of tubel

º~--~---'---~--~--~
O IO 20 30 40 50
Test section length, in.
Fig. 3.18 Pressure variation during flashing in a tube.
(Fauske. 12 )

3.16. Consider condensation in a round duct of diarneter D. Let thcre be no direct

contact betwcen the vapor and the wall and ]et the condensatc film be so thin that
v 1 « Vu anda = l. From Eq. (3.59) show that the pressure gradient is

dp
p¡¡g COS 8 + p va2 [ - A1 dz
9
dA
+ (1 + 11);1 Ji
dx
+ 2C,,]
D
- dz 1 - ¡)1¡¡2

where C1 ; is a suitably defined interfacial friction factor.

Evaluate the pressure gradient for steam at 90ºF traveling upward at 500 fps
in a 1-in.-diam straight vertical pipe if C ¡; = 0.005 and thc condensing rate is 10 lb/
(hr)(ft2). Consider the two limiting cases 7J = O and 7J = H-
Suppose that the pipe is tapered so that dDjdz = 0.05; what is the pressure
gradient?
Solve the same problem for steam velocities of 1000 and 1500 fps.
3.17. Cons-ider uniform condensation in a strnight horizontal tube at a low val u e of
Mach number. Assume constant properties, an inlet vapor velocity of V 1 and an
outlet velocity of V 2 in a pipe of length L. Using the results oí Prob. 3.16 show that
the pressure recovery is 23

If the pipe is short and condensation is complete, show that the predicted pres-
sure recovery is

By applying Bernoulli's equation to the central streamline for the vapor, show
that this result is only valid if the centerline velocity is (1 + 17 )Hi times the average.
Is this compatible with the usual turbulent flow velocity profiles if 7J }~? =
SEPARATED FLOW

3,18. The final term in Eq. (3.106) could be taken to represent the effect of modified
interfacial shear stress as a result of phase change. Show that this shcar stress is
given by

Ti = (V2 - V1) ( Eo - ~)

If evaporation is rapid enough, m/2 can be greater than 1c 0 • One may argue that
negatiVe shear stress would be unreasonable so that for high rates of vaporization
we must have (Ea - m/2) =
O. At these rapid rates of vaporization, inertia and
phase-change terms are usually much larger than thc friction and body-force terms.
Show that in this case, for gas-liquid flow, Eqs. (3.106) and (3.107) become
dv 1 dp
p¡V¡ dz - d,

P v dvg = _ dp _ Vg - v¡ G dx
ogdz dz a dz
For evaporation in a straight duct (G = const) deduce that, if p¡ is constant

,J_ [p x 1
2
_ 1 (1 - x) 2 + (1 - x)2] ~
0 (3.122)
dz Paª 2 (1 - a) 2 1 - a
This is the result obtained by Levy 13 in bis momentum exchange model which is
reasonably valid for rapid evaporation in pipes. (It is equivalent to assuming n = l.)
For small values of x during boiling or fl.ashing from initially pure liquid, deduce
that the slip ratio is approximately

~=
V¡
e~ e.t_)"
2 Po
Integrate Eq. (3.122) from x = O and deduce a relationship between x and a
at any point in the duct.
3.19. The choking condition in the presence of flashing is given by Eq. (3.62). Assume
that Eq. (3.63) is approximately valid, so that the actual value of '11 is irrelevant.
Furthermore, assume p¡ » pg and c¡ 2 »e/in gas-liquid fl:ow. Show that the mass
flux is then given by 1 º

ª' -- e~ - ax~)-,
p¡¡2ac¡¡2
2
ap ªPo
whilc a is givcn by Eq. (3.71).
3.20. From the results of Prob. 3.19 deduce the maximum possible values for the mass
flux of water as a function of pressure for x = 0.03, 0.1, 0.5, and 0.8. The results
should be virtually indistinguishable from those presented by Levy 14 and shown in
Fig. 3.19.
3.21. For expansion from stagnation conditions ,vith constant propcrties and no grav-
itational effects, entropy production is zero if vi/v2 = 1 (homogeneous flow) or
vi/v2 = (p2/p1)~'1. (frictionless separatcd flow) throughout. Zivi 15 claims that the
maximum (actually he called it a minimum) rate of entropy production occurs if
vi/v 2 = (p2/p 1)Vi everywhere. Do you agree?
3.22. Sho,v that in separated flow:
(a) The entropy gcneration per unit volume due to interfaeial effects alone is

S. = v, - v, { F12
--T- + G dx
dz [ (1 - 11)v1 + 11v2 - v,
-- +-
2
v,]}
. ONE-D!MENSIONAL TWO-PHASE FLOW

1000

600
400

-~200
n x=0.8
/
¡ 100
e
~
60
40

20
/ '
10
1000

600
400
j
.g 200 '/
n
x=0.7~ . /..
.; .·.·
~ 100
e
~ 60
,/··
✓., ..
.....
40
/'
/'
20 , /
/
10'='"~~--c"-~~~J_Ll"c--c-~~~ 4000
70 100 200 400 1000 4000 70 100 200 400 l000
2
Critica! flow rote, lb/sec-ft

Fig. 3.19 Critica! flow rates of steam-water mixtures. Dashes are Levy's predictions. 14
Dots are data points. These results should agree with the solutions to Probs. 3.19 and
3.20.

Deduce that the entropy increase due to phase change depends on the difference
between an "interface velocity/' (1 - r¡)v 1 +
11v2, and the arithmetic mean velocity
(v, +
v,)/2.
(b) the above equation reduces to

Deduce that the entropy generation due to phase change is zero if r¡ = },i and that
for evaporation we must have r¡ 2:: H and for condensation, r¡ ~ H-
 SEPARATED FLOW
"
3.23. Show that the interfacial forces per unit duct volume in Eqs. (3.48) and (3.49)
can be written as

Deduce that entropy generation is dependent on the "odd" forces, which act in oppo-
site directions on componcnts 1 and 2 and not on the "even" forces. Show that Eo
is proportional to the odd-force component, whercas m is proportional to the even
component. Relate to general ideas of reversibility in thermodynamics.
3.24. How is r¡ related to the quantity t in Example 3.10? Are the two interface
velocities, !;vz + (1 - t)v1 and r¡V2 + (1 - r¡)v1, the same? How are thcse related to
boundary layer or mixing phenomena on each side of the interface?
3.25, A vertical tubular test section is to be installed in an experimental high-pressure
water loop. The tube is 0.400-in. ID and 12 ft long heated uniformly over its length.
An estimate of the pressure drop across the test section is required as a function of
the flow rate of water cntering the test section at 400ºF and 1000 psia.
_(a) Calculate the pressure drop over the test section for a water flow of 2 gpm
with a power of 100 kw applied to the tube using (i) the homogeneous model; (ii)
the Martinelli-Nelson model; (iii) the Thom correlation; and (iv) the Baroczy
correlation.
(b) Estimate the pressure drop versus flow-rate relationship over the range of
2 to 15 gpm for a power of 100 and 200 kw applied to the tube using (i) the Martinelli-
N elson correlation; (ii) the Baroczy correlation.
3.25. Water and air flow together through a 350-ft section of horizontal 1-in.-ID
smooth pipe. The flow rateS are 1000 and 15 lb/hr of water and air, respectively.
The discharge end of the pipe is at a pressure of 1 atm and the system is isothermal
ata temperature of 68ºF. Calculate the pressure drop over the pipe work using the
Lockhart-Martinel1i correlation.
3.27. For separated incompressible flow of two fluids through a nozzle, show that

where !::.p1 and !::.p2 are the pressure drops for each component·flowing alone through
the nozzle. Compare with Eq. (3.30) and the Martinelli correlation.
(Murdoch 16 found that t::.p¾ = l.26t::.p 1 ½ + t::.p//1. for air-water and steam-
water mixtures.)
3.28. The one-dimensional flow model ignores variations across the channel. As a
result, errors are introduced due to the various different "averages" which can be
derived by integration over the flow cross section. Show how correction factors can
be applied to the various equations in the text to allow for these effects. Explore
the possibility for improving the theories in this way.
3.29. Solve Prob. 2.9 using (a) the minimum kinetic-energy flux theory, from Example
3.1, (b) the mínimum momentum flux theory from Prob. 3.8, (e) the momentum
exchange model from Prob. 3.18, and (d) the results of Prob. 3.19.
3.30. Rework Exarnple 3.9 for the case of (a) two incompressible liquids, (b) two com-
pressible gases, and (e) multiphase flow of n compressible fluids. 0
 ONE-DIMENS10NAL TWO-PHASE FLOW
"
3.31, Solve Example 3.5 for the case of vertical flow. Show that when gravitational
effects are considered, the expressions for f¡ and (f,) 1 are unchanged but

Show that this result is compatible with the solution to Example 3.6.

REFERENCES
l. Lockhart, R. W., and R. C. Martinelli: Chem. Eng. Progr., vol. 45, p. 39, 1949.
2. Martinelli, R. C., and D. B. Nelson: Trans. AS111E, vol. 70, p. 695, 1948.
3. Moody, F. J.: Trans. AS1"l1E J. Heat Transfer, vol. 87, No. 1, pp. 134-142, 1965.
4. Fauske, H. K.: ANL Rept. 6633, 1962.
5. Turner, J. M.: Ph.D. thesis, Dartmouth College, Hanover, N.H., 1966, also
AEC Repts., J. M. Turner and G. B. Wallis, Rept. NY0-3114-6, 1965, and
G. B. Wallis, Rept. NYO-3114-14, 1966.
6. Thom, J. R. S.: Intern. J. Heat 111ass Transfer, vol. 7, pp. 709-724, 1964.
7. Baroczy, C. J.: A. I. Ch. E. J. preprint no. 37, Eighth National Heat Transfer
Conference, Los Angeles, Calif., 1965.
8. Shapiro, A. H.: "The Dynamics and Thermodynamics of Comprcssible Fluid
Flow," The Ronald Press Company, New York, 1953.
9. Bernstein, A., W. H. Heiser, and C. Revenar: Trans. ASME J. Appl. Mech.,
vol. 34, pp. 548-554, 1967.
10. Wallis, G. B.: Intern. J. Heat Mass Transfer, vol. 11, pp. 445-472, 1968.
11. Spalding, D. B.: "Convectivc Mass Transfer," McGraw-Hill Book Company,
New York, 1963.
12. Fauske, H. K.: Inst. 111ech. Eng. Symp. Two-phase Flow, paper 10, Feb., 1962.
13. Levy, S.: Trans. ASME J. Heat Transfer, vol. 82, pp. 113-124, 1960.
14. Levy, S.: Trans. ASME J. Heat Transfer, vol. 87, pp. 53-58, 1965.
15. Zivi, S.M.: Trans. AS111E J. Heat Transfer, vol. 86, pp. 247-252, 1964.
16. Murdoch, J. W.: Trans. ASME J. Basic Eng., vol. 84, pp. 419-433, 1962.
17. Reynolds, O.: Proc. Lit. Phil. Soc., Manchester, England, vol. 14, pp. 7-12, 1874.
18. Nusselt, W.: Z. Ver. Deut. Ing., vol. 60, pJ). 102-107, 1916.
19. Silver, R. S.: Nature, vol. 165, p. 725, 1950.
20. Silver, R. S.: Fuel, vol. 32, pp. 121-150, 1953.
21. Silver, R. S.: Proc. 3d Intern. Heat Trans. Conf., A. l. Ch. E., New York, 1966.
22. Silver, R. S.: Proc. Inst. Mech. Engrs., vol. 178, pp. 339-376, 1963.
23. Wallis, G. B.: Proc. Inst. 111ech. Engrs., vol. 180, pp. 27-35, 1965-1966.
24. Silver, R. S., and G. B. Wallis: Proc. Inst. Mech. Engrs., vol. 180, pp. 36-40,
1965-1966.
 4
The Drift-flux Model

U INTRODUCTION

The drift-flux model is essentially a separated-flow model in which atten-

tion is focused on the relative motion rather than on the motion of the
individual phases. Although the theory can be developed in a way which
is quite general, it is particularly useful if the relative motion is deter-
mined by a few keyparameters and is independent of the flow rate of
each phase. For example, in bubbly flow at low velocities in large
vertical pipes, the relative motion between the bubbles and the liquid is
governed by a balance between buoyancy and drag forces; it is a function
of the void fraction but not of the flow rate. Drift-flux theory has wide-
spread application to the bubbly, slug, and drop regimes of gas-liquid
flow as well as to fluid-particle systems such as fluidized beds. It pro-
vides a starting point for extension of the theory to flows in which two-
and three-dimensional effects such as density and velocity variations
1

across a channel are significant. It is also the key to a rapid solution of

1

unsteady-flow problems of sedimentation and foam drainage.

••
 ONE-DIMENSIONAL TWO-PHASE FLOW
'"
4.2 GENERAL TIIEORY

The drift flux j 21 was introduced in Chap. 1 where it was shown to repre-
sent the volumetric flux of either component relative to a surface moving
at the volumetric average velocity j. It can be expressed in terms of
the relative velocity by using Eq. (1.39),

(4.1)

or in terms of the component fluxcs by using Eq. (1.36),

(4.2)

Since j = j 1 + j 2, Eq. (4.2) can be expresscd in the alternative forms

j, = (1 - a)j - J21 (4.3)
j, = aj+ in (4.4)

Equation (4.3) shows that the volumetric flux of component 1 is the sum
of the volumetric concentration times the average volumetric flux and a
flux - j 21 = j¡ 2 due to the relative motion. Equation (4.4) is a similar
statement far component 2. The drift flux is therefore analogous to the
diffusion flux in the molecular diffusion of gases and provides a convenient
way of modifying homogeneous theory to account for the relative motion.
Indeed, all of the properties of the flow, such as void fraction, mean
density, and momentum flux can be expressed as the homogeneous flow
value together with a correction factor or an additional term which is a
function of the ratios of }21 to the component fluxes. For example, the
void fraction is, from Eq. (4.2),

(4.5)

The mean density is

Pm -
- j,p, +· j,p, + (Pl - P2) j,,
------:- (4.6)
J J

When }21 is zero these results reduce to the homogeneous :flow values.

4.3 GRAVIH-DOMINATED FLOW REGIMES WITH NO WAll SHEAR

Drift-flux theory is particularly convenient for analyzing fl.ow regimes in

which gravity (or sorne other body force) is balanced by the pressure
gradient and the forces between the components. For vertical flow,
THE DRIFT-FLUX MODEL
"
Eqs. (3.45) and (3.46) then reduce to

0= - d p - gPl
dz
+ 1___!__i,__
- a
(4.7)

dp F12
0= - dz - gp 2 - ~ (4.8)

Subtracting Eq. (4.8) from Eq. (4.7) we get

F12 = a(l - a)g(p1 - P2) (4.9)
In the absence of wall effects, F12, the mutual drag force per unit volume,
is a function of the properties of the components, their geometry, the
void fraction, and the relative motion. Far a given system, therefore,
F12_is a function only of a and )21. Thus, in view of Eq. (4.9), j 21 must
be only a function of a and the system properties and we can write
)21 = )21 [a, properties of system] (4.10)
Evidently 1 in the absence of infinite relative velocities, )21 must
go to zero ata = O anda= 1, in view of Eq. (4.1).

Exampie 4.1 Using the results of Example 3.4 determine the relationship betwecn
Jsf and E for a vertically moving fluid-solid system to which the Carman-Kozeny
equation applies.
Solution Let component 1 be the fluid and let E = 1 - a. From Example 3.4

F fs = _ 180;:j¡o (1 ~ E) 2 (4 _1l)
Jio denotes the fluid flux when Js = O. Therefore, from Eq. (4.2),
j,¡ = -(] - ,)j¡o (4.12)

Combining the above relationships with Eq. (4.9) we obtain

d2g(Ps - p¡)<E3
(4.13)
180µ¡
This equation is reasonably valid for packed beds with values of E below
0.6. It is not a reasonable exprcssion for JsJ at high values of E since it does
not go to zero at t' = l. Since the sign convention has been adopted that the
upward direction is positive, the negative value of j.J indicates that the par-
ticles "drift" downward relative to the average motion if p. > Pi- Other sign
conventions are possible as long as consistency is maintained.

For a given system )21 can be plotted as a functión of a, using Eq.

(4.10), on a graph such as Fig. 4.1. Then, if the overall flow rates Q1
and Q2 are specified in a given situation1 )1 and )2 can be calculated from
Eqs. (1.19) and (1.20), and Eq. (4.2) predicts a linear relationship between
}21 and a. In fact, Eq. (4.2) represents a line joining the points a = O,
i21 = )2 and a = 1, j21 = -j¡ in Fig. 4.1. The intersections between
 ONE-D!MENSlONAL TWO-PHASE FLOW

- IJ,1 3

0
0

,..-Eq. (4. 10)

CD

-IJ,1,
CD Cocuccent flow
G) Countercurrent flow
Eq. 14.2 1
G) Flooding point
{
@ No solution

Fig.4.1 GraphicalsolutionofEqs. (4.2) and (4.10).

this line and the curve determine the values of a which are obtained in
practíce. 1- 3
This graphical method of solving Eqs. (4.2) and (4.10) is particularly
convenient as a means of visualizing the effects of changing the flow rates
Q 1 and Q2 since behavior in cocurrent and countercurrent flow in either
directÍon can be predicted simply by moving a straight cdge.
Figure 4.1 has been drawn for the particular case of small bubbles
suspended in a liquid flowing in a vertical pipe. It shows that for cocur-
rent upward or downward flow there is always a possible solution; for
countercurrent flow with the gas flowing downward there are no solu-
tions; while for countercurrent flow with the gas flowing upward there
are either two solutions or none depending on the magnitude of the flow
rates. ln the case of a dispersion of solid particles in a fluid thej21 versus
a curve passes through a sudden discontinuity at the point where the
particles pack together randomly and form a compact bed. If the
THE DRlFT-FLUX MODEL

particle concentration is denoted by a, then there is a maximum allowable

value of " as shown ín Fíg. 4.2. In thís figure the drift flux has been
rendered dimensionless by dividing by the terminal velocíty v. of a single
particle in an infinite fluid. The curve shown represents a typical equa-
tion whích has been found to correlate a wide variety of data.

(4.14)

n is a function of a suitably defined Reynolds number, and n - 3 is an

intermediate value for fluid-particle systems. a is the volumetric con-
centration of the dispersed component in general.
The actual value of "mox at which the particles pack together
depends on the particle shape and the nature of the interparticle forces.
In the extreme case of flocculated suspensions ama:x: can be as low as 0.1.
For hard spheres the range is usually 0.58 < "m•x < 0.62, depending on
the way in which the packing is achíeved. Tapping or shaking the bed

0.125
"o" ,<.
Li◊ º...Rt~ o ~l>.
"9, +,'o
0.100 t> Od -a
t>v t> ~oo: J;
0
% v t> + <1° ~h.. ] Particles
/xv v vt> ◊ l>. cfbx u
supported
0.075 >!r ◊ <1 o□ JI from above
?,t> vt> l>. o@ /
1" 9o ,o.'\<,
'•
0.050
ca
r ••~"" <1 'cio;!

r7 1> +~ ó 6 P-articles just

J t>"v o1i fluidized
t t>v◊ /
t>17 % Porticles
t>~ o supported
~,&; O ,..,... from below

O L_:______Jc__ _J__ _...L_ __L__ _L__ _l_J_C:":.cmc,oe,,

o 0.1 0.2 0.3 0.4 0.5 0.6

Material Diameter Material Dio meter

Socony beads 0.129" v Seasand 0.0147"
□ Socony beads 0.174" t> Sea sond 0.0219"
o Gloss beads 5mm <1 Sea- sand 0.0393"
o Glass beads 0.0201" + Crushed rock 0.0557"
6
Lead shot 0.0505" - Eq. (4.14) {n =3)

Fíg. 4.2 Dimensionless drift flux plotted versus particle volumetric

concentration for fluidization of various materials with \Va ter. (Data
of W ilhelm and K wauk. 4)
 DNE-DIMENSIONAL TWO-PHASE FLOW

1;
l
Water,
1 /p,
......, 0.05 0.05

\
\
\ - /1 Naphtha,
\ lps
' ' ',
o~------~-----'~
o 0.5
Concentrotion, a:: _..,.._
Fig. 4.3 Derivation of the drift-flux-concentration. curve
from Blanding and Elgin's flooding data. 6

will promote a higher value of a:max· The curve of J21/v,,,, versus a shown
in Fig. 4.2 represents a balance between fluid dynamic drag and buoyancy.
amax is the point at which particle-particle forces become significant.
If a exceeds °'m.x then it is necessary to return to Eqs. (4.7) and (4.8)
and include terms to describe the particle-particle interaction. If the
particles are completely inflexible and incompressible, then the packed
bed with a particle concentration of amax must be supported from above
or below 1 as indicated in the figure, depending on the value of J21/vei';l. For
very flexible particles 1 such as bubbles, the particle-particle forces can
be so small that Eq. (4.14) gives a good representation up to values of a
very el ose to unity.
The limit of operation in countercurrent flow is known as "flooding"
and occurs when the line representing Eq. (4.2) is a tangent to the curve
of j21 versus a. If the flow rate of either phase is increased beyond this
point, no steady flow solution is possible and a change in behavior must
occur. There can either be a change in flow regime or a rejection of
excess material at the ends of the flow passage.
The J21-versus-a curve can be assembled from its tangents if the
corresponding flow rates at the flooding points are known. Figure 4.3
shows this procedure applied to the flooding data of Blanding and Elgin 5
for a liquid-liquid (water-naphtha) system in a vertical pipe.
Another way of representing the different modes of operation is
to rewrite Eq._ (4.2) in the form
a . 1 .
J2 - 1 - a Jt +1 - a Jn (4.15)
THE DRIH-FLU)( MODEL

For a given system the last term in this equation is a function of a in

view of Eq. (4.10). Therefore, the curves of constant a on a plot of j,
versus J1 will be straight lines o! slope a/ (1 - a). The intercepts on the
axes will be J21/ (1 - a) and -j,¡/ a. The various regions of operation
are shown on a graph of this kind in Fig. 4.4, which corresponds to the
j 21-versus-a relationship shown in Fig. 4.1. These conclusions are illus-
trated by Deruaz's data' _for the bubbly regime o! gas-liquid flow in
Fig. 4.5.
Using Eq. (4.14) and the geometrical definition of flooding shown
in Fig. 4.1, it is readily shown that the corresponding flow rates at the
fiooding point are given parametrically as functions of a by the equations

]2
.
= )21 - a
dj21 = a 2v n(l -
a:-¡; a
) n-l
00 (4.16)

J1 - . -
-J21 (1 - a) -~21 - -(1 - an)v (1 - a )n (4.17)
da •

IExample 4.2 For bubbly flow of a particular mixture it is found that n = 2 and
v"' = 1 fps. VVhat is the relationship between Qf and Qa for flooding in a ve~ti-
cal 6-in.-diam pipe?

f, t
Cocurrent
upflow

Limited region of
countercurrent
flow

Flooding
line

Operotion
Cocurrent impossible
downflow

Fig. 4.4 The various regions of operation in one-dimensional

vertical flow in terms of the component fluxes for a flow regime
in which the drift flux is a function of a but independent of
j1 and j2. The particular case Pr > p2 is shown.
 ONE-DIMENSIONAL TWO-PHASE FLOW
"
40

0 20
0

10 20 50 100
Liquid flux, j 1, cm /sec --+-

a= 0.10
X a= Ó.] 5
o a= 0.20
+ a = 0.25

Fig. 4.5 Deruaz's data 6 for flow of bubbly mixtures in vertical pipes with
8- and 10-cm 2 cross section.

Solution From Eqs. (4.16) and (4.17), using the given values of n and v 00 , and identi-
fying thc gas as the discontinuous component 2, wc have

Ju = 2a 2 (1 - a) fps
ir - - (1 - 2a) (! - a)' fps
The area of cross section of the pipe is (ir/4)()°'2) 2 = 0.196 ft 2 • Therefore
Q 1 = 11.8j¡ (in cubic feet per minute) and Qa = 11.SJv- The following table
gives the predicted values at the flooding points:

a o 0.1 0.2 0.3 0.4 0.5

Q1, cfm 11.8 7.64 4.41 2.31 0.85 o

Q 11 , cfm o 0.21 0.76 1.49 2.26 2 95

It is assumed that no agglomeration of bubbles, leading to a change in

fl.ow regime, occurs.
THE DRIFT-FLUX MODEL
"
U CORRECTIONS TO THE SIMPLE THEORY

The simple one-dimensional gravity-dominated theory does not take

account of variations in concentration and velocity across the cross
section. The quantities a, v1 and v2, and J1 and J2 are merely aver-
ages 3:cross the dfrection of flow. Continuity and dynamic equations
written in terms of these averages are not necessarily identical with the
rigorous formulation of these equations as integrals over the flow field
(e.g., the product of averages is not the same as the average of a product).
This problem is already familiar to students of single-phase fluid mechanics
,vho know, for example, that the momentum flux in a pipe with a velocity
profileJ but uniform density, is

pfv 2 dA - Ap(v') (4.18)

and that this is not the same as

where the ( )'s denote averages over the cross section defined by the
equation
(X) - JX dA (4.20)
A

However, just as in single-component flow the ratio between Eqs.

(4.18) and (4.19) is usually expressed as a correction factor to the one-
dimensional theory1 so in two-component flow the ratio of suitable inte-
grals is also used to define parameters which are equal to unity for truly
one-dimensional flow and are not far from unity in the general case.
Of particular usefulness is the distribution parameter C0, introduced
by Zuber and Findlay 7 and defined as
(aj)
e, - (a)(j) (4.21)

Co represents the ratio of the average of the product of flux and concen-
tration to the product of the averages.
A convenient definition of an average velocity of phase 2 is obtained
from Eq. (1.23)

v, - (.i,) (4.22)
(a)

Averaging Eq. (4.4) term by term across the duct, we have

(4.23)
 ONE-DIMENS!ONAL TWO-PHASE FLOW
"
Combining Eqs. (4.22), (4.23), and (4.21) leads to

v, = Go(j) + U212
(a)
(4.24)

Note that Eq. (4.23) represcnts a strict equation of averages across the
channel. The quantity ii 2 in Eq. (4.22), however, is not equal to (v,) in
general since (v,) is related to the flux and conccntration distributions by
the equation

(4.25)

The average velocity represented by ii2 is usually more convenient to use

than (v,) because it can be directly related to the overall flow rate Q,
and the volumetric mean concentration (a) by means of Eq. (4.22).
(a) is a convenient average volumetric concentration because it is
readily related to a simple experiment in which a section of the duct is
suddenly isolated and the proportion of the volume occupied by compo-
nent 2 is determined. (a) is also the quantity which is usually measured
by 1'-ray scanning techniques.
In terms of the overall volumetric flow rates, Q, and Q,, the follow-
ing equations are valid:

(j) = Q, 1Q, (4.26)

( J2.) = Q,
A (4.27)

(J1.) = Q,
A (4.28)

In view of Eqs. (4.22), (4.26), and (4.25), Eq. (4.24) can be expressed
as
~ = Co Q, + Q, + (jn) (4.29)
(a)A A (a)
Therefore the volumetric mean value of a is
(a) = Q, - A(j,,) (4.30)
C,(Q, + Q,)
If (j, 1) is small compared with Q,! A, Eq. (4.30) reduces to
1 Q, (4.31)
(a ) = Co Q, + Q,
The second factor in Eq. (4.31) is the same as we should obtain from
homogeneous theory. Thus the effect of concentration variations, but
not the effect of relative velocity, is to multiply the mean concentration
Ti-!E DR!FT-FLUX MODEL .,
calculated from homogeneous theory by a correction factor I/C 0 • This
correction factor is the same as the jloiu parameter K uscd by Bankofr8
and the Armand parameter 9 which is used in the Russian literature.
The quantity Ü21) cannot be evaluated in general ,vithout a knowl-
edge both of the dependence of jn on a and a.lso the variation of a across
the channel. Simplification is, however, possible in two cases.
Case 1 j 21 independent of a, i.e. 1
j21 = const (4.32)
The average value of j21 is then equal to this constant value.
Case 2 ]21 varies linearly with a, i.e.,
}21 - b, + b1a
where b0 and b1 are constants.
In this case
(j,1) - b, + b1(a) (4.34)
and Eq. (4.30) becomes
(a) - Q, - Abo
C,(Q1 + Q,) + b1 (4.35)

Equation (4.35) is applicable, for example, to the slug-flow and churn-

turbulent bubble-flow regimes in gas-liquid systems.
Note that the local fluxes j in the above context denote the time
average of the volumetric rate of flow across a given area. If time
variations in the fluxes are importantJ more complicated equations will
arJSe.

Example 4.3 An important use of simple one-dimensional drift-flux theory is for

the interpretation of experimental results. In this exa.mple the technique is
applied to the data of Smissaert for air-water flow in a vertical 2-in.-diameter pipe
reported in Ref. 10. The subscript 1 refers to the water and 2 to the air.
First the data are used to calculate }21 from Eq. (4.2) and the rcsults are
plotted up as a function of a for the lowest and highcst liquid rates as in Fig. 4.6.
The assumption that j21 is only a function of a: is seen to be quite good, but there
is a noticeable drift with j1. Furthermore, the value of } 21 appears to become
vcry large at about a = 0.8.
Explanation for these observations is sought in terms of the parameter
G0• Equation (4.22) is combined with Eq. (4.24) to give

(4.36)

The "averaging" signs have been dropped for convenience. N ow it is notíced

that the high values of j21 in Fig. 4.6 correspond to large values of j 2( ;:c;:3Q fps).
In order for j 21 to stay small at thesc flow rates, it is necessary for (1 - Coa:) to
 ONE-DIMENS\ONAL TWO-PHASE FLOW
'"'
4

Aír ond water upflow

J21 vs a'.
C0 =1
o

3
/ 1, fps
◊ 0.1
o 0.2
t o 0.8
• 1.0
--~o. 2

0.4 0.6 0.8

Q!~

Fig. 4.6 Data of Smissaert 1 º plotted as drift flux versus

concentration, assuming C O = 1.

approach zero. Therefore a approaches the value 1/Co. Looking again at

Fig. 4.6 we see that the limiting value for a is about 0.8, giving a value for C o
of about 1.25.
Substituting G 0 = 1.25 in Eq. (4.36) and recalculating j 21 for all of tbe
data, we obtain theresult shown in Fig. 4.7. The data scatter around a straight
line, therc is no systematic tcndency with ji, and the drift flux j21 does not take
on excessively large values.
Having brought the data together by the use of the parameter Co, we
now test for the value of nin Eq. (4..14). Plotting hi/a versus (1 - a) on log
paper we expect a line of slope n and intercept v at a = O. Figure 4.8 shows
00

that n is approximately zero and the value of v,,, is about 1.2 fps. A change in
flow regime is indicated as the value of a approaches 0.8. The scatter is of the
same order of magnitude as the experimental repeatability.
In this example it was fortunate that the method of introducing the com-
ponents led to a range of bubble sizes conducive to a constant value of v"".
Thus it was possible to bring the data together by adjusting the parameter Co
alone. In the case of the nitrogen-mercury results reported by Smissaert in
the same report, a trend of v"' as a function of j1 is evident, although an average
value of v"" = 2.5 fps is within about 10 percent of all his data.
THE DRtFT-FLUX MODEL m
Since one has three parameters to play with (C 01 v.,, and n), there are the
usual pitfalls to be expected in correlating data in this way without any refer-
ence to the flow pattern or other important parameters such as bubble size.
Of course the best technique is to make additional observations during the
experiment which motívate an independent assessment of these parameters.
This topic will be taken up again in later chapters when the individual flow
regimes are discussed in detail.
The data have been plotted in Figs. 4. 7 and 4.8 in a way which realis-
tically represents the scatter and does not flatter the correlating scheme. Man y
published comparisons betwecn data and correlations are plotted in such a way
as to give the illusion of reduced scatter.

U SIGN CONVENTIONS AND IDENTIFICATION OF COIIIPONENTS 1 ANO 2

Since velocities and fluxes may be in either direction in one-dimensional

flow, sorne sign convention is necessary. It <loes not matter which system
is used as long as it is applied consistenti y.
Usually one is interested more in describing the motion and concen-
tration of one component than the other. In fact, throughout this

1.0

i1, fps ' 1

1
□
o 0.1
◊ 0.2
0.8
" 1 "
0.3
◊ 0.4 □

◊
ol"
4 0.6 49) ◊
◊ ◊
rP 9
" 0.8 1
□ 1.0 o
1
j 0.6
◊

- o.
.--,;;
◊

0.4

1 Probable

0.2,-----;--t--------¡--------,--~I cho-nge of
I reglme
1
o
1

o~-~-~-~-~-~----'--~--'---'
O 0.2 0.4 0.6 0.8

fig. 4.7 lVIodified l'epresentation of Fig. 4.6 using C O = 1.25.

,., ON._E-D1MENSIONAL TWO-PHAfE FLOW
10~--~-~~~~~~---~~

! ¡□ n=O
,.E- ¡oocwo%
~ 1 r------6~~~ ~ -~"-º-+--------,
·~ 1
□I
Cbl
o 0.1
1
V 1 ◊ 0.2
□ 0.3
Probable 1
◊ 0.4
chonge in 1 17 0.6
regirne 1
ª 0.8
1 □ 1.0
1

Fig. 4.8 Determination of n for

(1-a) ~ data in Fig. 4. 7.

chapter we have "favoritized" component 2 by calling its volumetric

concentration a and leaving the concentration of 1 to be 1 - a. W e
have further selected j,1 rather than j,, far analytical purposes although
the defining equations are quite symmetrical. l\1oreover 1 we later stated
that component 2 was to be chosen as the discontinuous component.
The reason for this is that one is usually inclined to view the relative
velocity vn or the terminal velocity v«:, as the velocity of a particle in a
fluid medium rather than the velocity of the fluid which will bring particles
to rest. Thus a convenient sign convention is often that in which veloci-
ties measured in the direction of drift of component 2 are positive 1 in
which case}21 is always positive. Alternatively if a force balance is used
to deduce j,1, as in Example 4.1, it may be useful to define the positive
direction in terms of the direction of the gravitational field,
In later chapters it will be found convenient sometimes to use the
inverse descriptíon and to focus attention on the motion of the continuous
component 1 for example 1 when a fluid flows around approximately sta-
tionary particles in a fluidized bed. For tbis purpose we define , as the
volumetric concentration of the continuous component where

, - (1 - a) (4.37)
 THE DRIFT-FLUX MODEL 103

and work in terms of j12 rather than j21, The analytical techniques
remain unchanged.

4.6 UNSTEAD'i' FLOW

The drift flux is a very useful parameter for analyzing unsteady flows by
using the theories of wave motion which will be presented in Chap. 6.

PROBLEMS
4.1. Using the drift-flux model show that the average phase velocities are

Deduce an expression far the momentum flux in a duct in one-dimcnsiona1 flow.

4.2. Air and water flow in a vertical duct at 70ºF and 36 psia. The drift flux is given
as a function of void fraction by
iuf = 0.8 a(l - a) 2 fps
What is the void fraction far the following values of j¡ and j., in feet per second?

JI 0.5 1 0.5 2 5

0.5 0.5 1 2 5

Far these conditions, compare the momentum flux and void fraction calculated
from drift-flux theory with the values which would be predicted from homogeneous
theory and the Martinelli correlation in a 2-in.-diam pipe.
4.3. Air is bubbled uniformly through stagnant water. VVhat is the relationship
between volumetric air flux and void fraction if j~ 1 = v a(l - a)2? What is the
00

significance of the two possible values of void fraction? If the depth of the water
before the bubbling is 100 cm and v,., = 25 cm/sec, what is the height of the two-phase
mixture during bubbling as a function of the flow rate? When_ does floo4ing occur?
4.4. Sketch the curves of j 2 versus a as a function of )1 and j1 versus a as a function
of h far a drift flux given by Eq. (4.14). Identify the conditions of flooding, cocurrent
and countercurrent flow.
4.5. Show graphically how the void fraction depends on the fluid flow rate in a fluidized
bed for different particle fluxes, both upward and downward. What regions of th.e
graph correspond to cocurrent and countercurrent, upward and downward flow?
Over what regions of the graph is a = a,uax if the particles are unrestrained by the
ends or the walls of the duct?
4.6. A flooding experiment yielded the following data far an air-water system.

-j¡, cm/sec 2 4 5 9 11 15 18
-----1 --- --- ------ --- --- ---
j~, cm/sec 9 7 6 4 3 2 1
 ONE-DIMENSIONAL TWO-PHASE FLOW
'"'
Derive the j¡¡¡-vcrsus-a relationship. What "best" values of v.,, and nin Eq. (4.14)
describe this relationship?
4.7. What does the graphical technique described in Fig. 4.1 become in the limiting
case oí homogeneous flow with v1 = vz?
4.8. One way which has been suggested for determining the values of C 0 andj21 is to
plot j2/a versus j. For what value oí nin Eq. (4.14) will this relationship be linear?
What characteristics of the line will then determine C 0 and v"'? How does the drift
velocity of component 2, V2J, depend on a in this case?
4.9. For the churn-turbulent regime of bubbly flow and for the slug-flow regime,
Vaj is approximately constant. If the flow is isothermal show that the choking condi-
tion is

j_,
p
(a 11 +
1 +G¡Va;.¡Ji. ) - 1

Prove tbat, at the same mass flux of the components, the pressurc at which choking
occurs will be lower in this case than in homogeneous fl.ow. Assume Ca = l.
4.10. Solve Prob. 4.9 if Co is constant but is not equal to l.
4.11. When drift occurs the various expressions for the homogeneous flow frictíonal
pressure drop are not identical; in particular we can choose between the relationships

Compare these various predictions in terms of the relationship between j21 and
the individual fluxes. Under wha.t conditions will the predictions be most sensitive
to the assumptions?
4.12. When drift occurs, the momentum flux differs from the homogeneous theory
prediction. Express the momentum flux in terms of the ratios between )21 and the
component fluxes, the mass flow rates and the overall flux. Show that the ratio
between the momentum flux with drift and that without drift is
G,/(1 + j,,/j,) + G,/(1 - jn/j,)
G1 +G2
4.13. Starting from Eqs. (3.39a) and (3.39b) and using the definition of il in Eq. (2.92),
show that, in the presence of drift, the void propagation equation for constant fluid
properties in vapor-liquid flow is

-da + J. aa- + -J¡¡J

o . = [ -V¡ + (l - a)] n
at az az Vfa

If j 11 ¡ is only a function of a and the system properties, use the relationship

a.
az1of =
ªªaz ªªa .
Ju!

to prove that voids propagate with the continuity wave velocity

and grow exponentially with time.

4.14. Deduce the relationship between )21 and a from the data shown in Fig. 4.5.
Comment on the form of the result.
THE DR1FT-FLUX MODEL 105

4.15. Solve Prob. 2.30 if the drift flux is given by the expression Jut = l.0a fps.
4.16. Solve Prob. 2.31 if Jut = 1.0a fps.
4.17. How do the lines shown in Fig. 4.4 compare with the prcdictions of (a) homo-
geneous flo,v, (b) the Martinelli correlation for viscous-viscous flow.
4.18. Under subcooled boiling conditions in vertical upflow, C 0 can be less than unity
(Why?) while J11 , is positive. Investigate undcr what circumstances the void fraetion
in the ~uct can exceed the predictions of homogeneous thcory.
4.19. Calculate the value of C 0 if O'. and J. va.ry as power laws across a circular pipe.
Discuss what ranges of values are reasonable under various conditions.-
4.20. Suppose that bubbles supplied to a vertical pipe do not have a uniform size.
If a probability distribution for bubble size can be estimated and J11¡ can be expresscd
in terms of bubble size and the local valuc of a, how should one modify drift-fiux
theory in terrns of appropriate averages?

REfERENCES
l. Wallis, G. B.: Paper No. 38, Proc. Inlern. IIeat Transfer Conf., ASME, Boulder,
Colo., vol. 2, pp. 319-340, 1961.
2. Wallis, G. B.: Symp. Interaction Fluids Particles, Inst. Chem. Engrs., London,
pp. 9-16, 1962.
3. Wallis, G. B.: Symp. Two-phase Flow, Inst. lvlech. Engrs., paper no. 3, pp. 11-20,
1962.
4. Wilhelm, R. H., and ]\-1. Kwauk: Chem. Eng. Progr., vol. 44, pp. 201-217, 1948.
5. Blanding, F. H., and J. C. Elgin: Trans. A. J. Ch. E., vol. 38, pp. 305-335, 1942.
6. Deruaz, R.: ANL transl. 61, 1964 (from the French, Centre D'Etudes Nuclcaires
de Grenoble, note TT#165, April, 1964).
7. Zuber, N., and J. Findlay: Trans. ASivlE J. Heat Transfer, ser. C, vol. 87, p. 453,
1965.
8. Bankoff, S. G.: Trans. AS.lv[E J. Heat Transjer, ser. C, vol. 82, p. 265, 1960.
9. Armand, A.: UKAEA, AERE transl. 828, 1959 (lzvest. Vsesoyuz. Teplotekh. Inst.,
no. 1, pp. 16-23, 1946).
10. Smissaert, G. E.: ANL Rept. 6755, 1963.
 5
Velocity and Concentration
Profiles

5.1 INTROllUCTION
The next step in sophistication, beyond the simple nlumped" models
of homogeneous and separated flow, is the consideration of velocity and
concentration profiles across the duct. This is still a quasi-one-dimen-
sional description of the flow because local velocities are allowed only in
the principal direction of the motion. Any motion across the duct
is either neglected or absorbed into parameters 1 such as "eddy
diffusivity," which account for turbulent mixing. In turbulent flow
the velocity and concentration profiles are averages over a long period
of time.
The maj or use of velocity and concentration profiles in this book
will be for motivating correction factors which can be applied to the
simpler homogeneous or separated-flow models in arder to increase their
accuracy. In sorne cases it will be possible to derive analytical expres-
sions for correlating parameters, such as friction factor and two-phase
multipliers 1 rather than relying on the empiricism of previous chapters.
,.,
VELOCITY AND CONCENTRATION PROFILES 107

A further use of the theory will be the derivation of uselul dimensionless

groups.

5.2 QUALITATIVE ASPECTS

The concept of velocjty profiles in single-phase flow is a familiar one.

Certain simple cases, such as flows in circular pipes, ha.ve been investi-
gated in great detail.
In two-phase flow the situation is far more complicated for the
following reasons:

l. The concentration of a particular phase may be neither uniform nor

symmetrical. Suspended particles tend to settle out in horizontal
pipes. Liquid films stick to walls and are thicker at the bottom than
at the top of the pipe. Bnbbles rise preferential!y near the middle
of a vertical pipe. A series of large bubbles 1 or slugs, gives rise to
large periodic variations in concentration.
2. The uconcentration" is not usually an adequate means of describing
the local two-phase properties. Large particles may sink to the
bottom of a pipe whereas small ones remain suspended. Very small
particles (below about 1 ¡,. in diameter) follow the fluid streamlines,
whereas large ones may not be significantly diverted within the
length of the apparatus or may undergo many collisions with con-
taining walls. Several flow configurations, such as drops 1 bubbles,
and liquid films, may coexist at the same time.
3. Although there is no relative motion at the interface between the
phases, there may be average relative motion on a scale which is
larger than the distance at which a continuum description is mean-
ingful. For example, the average droplet velocity in a suspension
is not necessarily the same as the veloeity of the suspending fluid.
Thus there exist two velocity profiles, one for each phase, as shown
in Fig. 5.1.

The general solution of these problems appears quite hopeless. A

diseussion of ali the work which has been done in this field would itself
require a complete book. Al! that will be done in this chapter is to
derive a few simple results that are extensions of the homogeneous and
separated-flow models and which provide material that will be useful in
later ehapters.
V elocity and concentration profiles can be analyzed using either a
clifferential or an integral technique. In the former, the velocities and
concentrations are derived as solutions to differential equations which are
deduced from a study of small elements of the fluid field. In the latter,
108 ONE-D\MENSIONAL TWO-PHASE FLOW

Gos-phase velocities:

-
-----o-- Air alone

--D-

~
With 10
With 20
With 30
lb glass
lb glass
lb gloss

Solid-phase velocities:
-----fr- With 1O lb gloss
---o- With 20 lb glass
With 0.75 lb MgO
0.2 With 1.7516 MgO
With 2.7516 Mgü

o c__ _¡__ __L__ _J __ ___l_ __J

o 0.5 1.0 1.5 2.0 2.5

Radius, in.

fig. 5.1 Velocity distributions of gas and particles in a duct. (Results of

Soo, 1'rezek, Dimick, and Hohn.streiter. 1 )

the form of the profiles is assumed and the dynamic and geometrical
conditions are satisfied in integral form.

5.3 DIFFERENTIAL ANALYSIS

VELOCITY F'ROFJLES IN SINGLE-PHASE flOW

lh single-phase flows the "concentration" is usually uniform over the

cross section. Fluid properties may va:ry normal to the flow due to
temperature and composition changes, but it is usually possible to assign
suitable local values of viscosity and density for use in the equations.
The well-known method for deriving a "universal)) velocity profile
for round pipes is first to calculate the shear stress distribution from a
force balance, then relate the shear stress to the velocity gradient and
finally integrate the velocity gradient to get the profile.' Sorne useful
parameters are the friction velocity u* ,and dimensionless forms of the
velocity and distance from the wall, as lollows:

u*= (~y· (5.1)

V
u+= (5.2)
u*
pu*y
y+ - (5.3)
/L
VELOCITY ANO CONCENTRATION PROFILES 109

Close to the wall the flow is assumed to be laminar and the shear stress
is related to the velocity profile by the equation
dv
7
= µ dr (5.4)

For flows with uniform density in round pipesi an approximate solution

for small values of y is
u+= y+ (5.5)
and empirically this is found to be valid up to y+~ 5.
Numerous alternative expressions can be chosen to relate the shear
stress to the velocity gradient in the regían of turbulent flow. Perhaps
the simplest is in terms of a mixing length l introduced by Prandtl, 3
as follows:
r = ¡2 dv I dv 1
(5.6)
P dy dy

If it is further assumed that Z varies linearly with y, we have

l = ky (5.7)
In round pipe.s these equations Iead to a Iogarithmic dimensionless
velocity profile
1
u+=7clny++C (5.8)

Empirically it is found that k ~ 0.4, C "-' 5.5, and that Eq. (5.8) is valid
for y+> 30.
In between y+ = 5 and y+ = 30 there is a buffer ]ayer in which the
shear stress can be represented by a hybrid equation
dv
T = (µ. + ,p) dy (5.9)

ande is given by an .equation such as Deissler's 4

(5.10)
nis an empirical dimensionless constant with a value of about 0.1.
If the flow is nonnewtonian, solutions can be derived in the laminar
regime by replacing Eq. (5.4) by the appropriate relationship between
shear stress and rate of strain.

VIELOClTY PROFILES IN TWO-PHASE FLOW

The techniques which were described above can be applied to two-phase

(or multiphase) flows as long as the flow is locally homogeneous. One
velocity profile is then determined for both components together. In
110 ONE-DlMENSIONAL TWO-PHASE FLOW

order to solve the equations the values of p µ., and l are needed as a func-
1

tion of position. Usually the homogeneous density and equivalent vis-

cosity are given as functions of void fraction or concentration a by the
methods of Chap. 2. Thus, the definition o! a flow regime in terms of a
concentration profi1e and the properties of the components enables p
and µ. to be calculated everywhere. The mixing length presents more
difliculty because comprehensive empirical expressions for it in two-phasc
flow have not yet been established. Two-phase mixing lengths are
sometimes greater and sometimcs smaller than in single-phase flows.
For example, a study of the gas core in annular flow by Gill, Hewitt,
and Lacey 5 revealed that Eq. (5.8) was valid but with values of k which
depended on the flow rates 1 as shown in Fig. 5.2. The corresponding
gas velocity profiles differed noticeably from the single-phase flow value
(Fig. 5.3) in contrast to the results for a particle-fluid system shown in
Fig. 5.1. General conclusions have yet to be drawn from studies of this
type.
In any case a computation scheme can be set up which ·will readily
accept information about property and mixing-length variations across
the channel as they become available. For axially symmetric steady

280
vii,, vlj, Left-hand, Right-hand,
lb/h, lb/h, side side

240 300 500 • V

500
300
50
1250
•• o
□

.
-l'
200

>-
·Go 160 ◊-

o> /
e
□
1120
□
u
.3
80
•
□

40 /.

o
0.1 0.2 0.5 1.0
y/ro
Fig. 5.2 Logarithmic plots of gas velocity profiles in
annular-mist flow. (Results of Gill, Hewitt, and Lacey. 5)
VELOCITY AND CONCENTRATION PROFILES m
2so~--,----.---,-------,---,----,---,------,----,--,

240e---------_,,-c-v,rs,.,,-------

200 J /-x~X t:,,--'

-~~ª~
- ---~~
✓-&:::==

--ff--- Ainate, 500 16/h,---"-"-',cx:-

f
O

Watenate, p9 , \~
lb/he lb/ft 3 1
80f---- O 0.090
30 0.09040 X
200 0.09695 o
40 500 0.1063 8 o
IOOO O. 1 1 8 1 8

o~~--c"-:-~--c'-:--~--:c'--:-~--:c'-;c-~--,,
O 0.2 0.4 0.6 0.8 1.0
Ratio, y/D

Fig. 5.3 Effect of injected water ratc upon thc air velocity
profile in annular-mist flow. (Gill, Hewitt, and Lacey. 5 )

flows without stratification, the shear stress distribution in a round pipe

at an angle e to the vertical is found, by balancing the forces on the
cylinder of radius r shown in Fig. 5.4, to be

T = -2r ( - -dp - p,g g(p1 - p,) cos e

cos 0) - --~,::---- Jc' (1 - a)2r dr
dz 2r o
(5.11)
W e now introduce the dimensionless parameters

* 4T
7
= Dg(p1 - p,) (5.12)
ó.p* = -dp/dz - p2 g cose
g(p1 - P2) (5.13)
2r
r* = D (5.14)

and express Eq. (5.11) in the dimensionless form

T* = cos-0
r* t,,p* - -
r* o
fc''
(1 - a)2r* dr* (5.15)
112 ONE-Dl!VIENSIONAL TWO-PHASE FLOW

p+'op

Verticol'"2----
direction

Fig. 5.4 :Force balance on an elementary

cylinder used to derive Eq. (5.11).

This shear stress distribution differs markedly from the familiar linear
variation of single-phase flow. Depending upon the flow pattern,
maxima and minima, changes of sign, and changes of curvature are all
possible as the two terms in Eq. (5.15) interact.
Following Turner 6 we can also express Eqs. (5.4) and (5.6) in
dimensionless form by defining the parameters

(5.16)

(5.17)

(5.18)

(5.19)

The subscript i refers to either phase and the phase subscripts are chosen
so that p1 > p 2. The properties or parameters with no subscripts apply
to the local homogeneous average. For laminar flow, Eq. (5.4) becomes

(5.20)

whereas for turbulent flow, from Eq. (5.6)

dvf = _ HR, l7 *l" (5.21)

dr* l*
H denotes the sign of the shear stress.
VELOCITY AND CONCENTRATION PROFILES m

N in Eq. (5.20) is a function of the local homogeneous viscosity

and density and can be written as

N = N, - (p)½ --µ1
Pl µ
= N, -(p)½ -,,,
P2 µ
(5.22)

From Eq. (2.9) R, can be expressed as a function of the local value of a

and the density ratio p1/ p,.
From any of Eqs. (2.56) to (2.61) the ratio µ/µi can be expressed
in terms of a and the viscosity ratio µ¡/ µ2. l\.Ioreover

N 2(P')"
µ1 = N1 (5.23)
µ2 P2

Combining these results we eventually find that N is a function of N 1,

N 2, p1/ P2 and a.
The velocity profile is determined by integrating Eqs. (,5.20) and
(5.21), using Eq. (5.15) and information about the variation of a and l*
as functions of the radius r*. The volumetric flow rate of each phase is
obtained in dimensionless form by integrating the velocity profile. Thus

(5.24)

where ai is the volumetric concentration of phase i and

j; = }iPiH[gD(p1 - p 2)J-H (5.25)
In general, if the flow regime is specified so that the local value of a can
be related to the average value (a), we have

J.¡·• -- Ji·• ( !:J..p * (a )

I
Pt
1 -,
N l¡ N 2,0) (5.26)
p, .

Example 5.1 Determine expressions for j¡ and j; in laminar vertical annular gas-
liquid flow.
Solution For vertical flow cos 0 = l. Sin ce the flow is annular, the local val u e
of o: is unity in the gas core and zero in the liquid film. In terms of the overall
void fraction therefore 1 Eq. (5.15) becomes
In the core r* = r* !lp* (5.27)

In the film r* = r* t..p* - -

r*
¡ f,v-;;:,, Z1rr* dr* (5.28)

Evaluating the integral in Eq. (5.28) we find

In the film r* = r*(ó.p* - 1) +~

r'
(5.29)

Substituting Eq. (5.29) into Eq. (5.20) with i = f we obtain

dr*
-1
-dvf = - N
8
[ r*(.ó.p* -1) +-
r*
ª] (5.30)
n• ONE-D\MENSIONAL TWO-PHASE FLOW

Tntegration from r* to the wall, r* = 1, v; = O, yields

V¡
* = gN 1 [ (t,p' - 1)(1 - ,")
- a ln r*
]
(5.31)
2

Substituting this value into Eq. (5.24) and performing the integration we have,
since ª i = 1 in the film and zero in the core,

(5.32)

The interface velocity is derived from Eq. (5.31) by putting r* 2 = a and can be
put in terms oí the dimensionless gas velocity
root oí the density ratio, thus
v:
by multiplying by the square

'
(v,), - Ni
~
16
(P');'
~
p¡
[(t,p' - 1)(1 - a) - a ln a] (5.33)

Now Eq. (5.27) is substituted into Eq. (5.20) to show that in the gas core

dv: N,
- - r* iip* (5.34)
dr* 8

Integrating from r* to Va and using Eq. (5.33), we obtain the gas velocity
profile

v' - N, flp* (r" - a)

g 16
+ Ni
16
(P_,)¼
PI
[(flp' - 1)(1 - a) - a ln a] (5.35)

Using this in Eq. (5.24) with Cl.i = 1 in the core and zero in the film, we find
eventually that

., N,32 {a
Ju = - 2 t:..p* + 2a -N1
Nu
(P')¼ [(6.p* - )
-
Pt
1 (1 - a) - a ln al ) (5.36)

Both Eqs. (5.32) and (5.36) are special cases of Eq. (5.26).

Far a particular combination of components and flow regime in a

given pipe, Eq. (5.26) expresses the two dimensionless flow ratesj¡ andj;
as functions of t;p* and (a). These two equations must be solved simul-
taneously if t;p* and (a) are to be deduced from known values of the
flow rates.
Even when a theoretical solution to these equations cannot be
obtained, the dimensionless parameters which have been derived can
form the basis far a rational correlation scheme. The dimensionless
parameters can also be combined in numerous ways to form new group-
ings when convenient.

IExample 5.2 In the separated-flow model of annular vertical flow it has been sug-
gested that the wall friction factor should be correlated versus the liquid film
Reynolds number, Re 1 = j¡p¡D/µ¡. Show how this description is related
to Eq. (5.26).
VELOCITY ANO CONCENTRATlON PROFILES m

Solution The Reynolds numbcr is readily cxpressed in terms of the previous dimen-
sionless groups as
Re¡ = jJ1'l¡ (5.37)
The dimensionless wall shear stress, from Eq. (5.15), is
r! = t:.p* - (1 - a) (5.38)
The wall friction factor_ can be expressed as follows:
Tw(l - a) 2 r!(l - a) 2
C¡w = }ip¡j¡i
2·>P
J¡
(5.39)

Therefore
[l>p* - (1 - a)](! - a)'
C1w = .*
2]¡
2 (5.40)

A correlation expressing C fw as a function of Re 1 thus implies thc existencc

of a relationship between jf, N1, t:.p*, anda and is a special case of Eq. (5.26).

5.4 INTEGRAL ANALYSIS

To perform an integral analysis the form of the velocity and concentration
profiles is assumed rather than calculated. U nknown parameters are
either determined empirically ar by making the profiles satisfy certain
integral equations and boundary conditions. A familiar example from
single-phase flow is the von Kármán integral technique far analyzing
boundary layers. The method has not been widely used far two-phase
flows, and only a few useful results are available.
The Bankoff variable density model far round tubes is based on the
assumption of local homogeneity and no relative velocity. The velocity
and concentration are assumed to vary according to power laws. 'I'hus
_v_ = (}!_)1/m (5.41)
Vm ro

..':_ =
am
(]!_)'In
ro
(5.42)

Vm and am are the values at the center of the tube, y is the distance from
the wall, and ro is the pipe radius. From Eq. (1.6) the average value
of a is
(a) = ~ ('"
1rr 0 Jo
C<m (]!_)'
r 0
in 21r(ro - y) dy
(1 + n)(l + 2n) (5.43)

The mass flux of phase 2 is

G, = -1-2
1rr 0
f'"o avp,21r(ro - y) dy

= e<mVmPZ (m + n + mn)(n + m + 2nm) (5.44)

116 ONE-DIMENSIONAL TWO-PHASE FLOW

and the mass flux of phase 1 is

G1 l - X
G, = - x - (5.46)

Equations (5.44) and (5.45) can be combined to yield

l + P2 1 - x = (rn + n + nrn) (n + rn + 2nm) (5.47)
P1 X n'(l + rn)(l + 2rn)c,m
Finally, elimination of "m between Eqs. (5.43) and (5.47) shows that a
is related to x by the equation

(5.48)

where
2(rn n + +
rnn) (rn + n + 2rnn) (5.49)
K = (n +
1)(2n +
l)(m + 1)(2rn + 1)
Overa wide range of values of n between 0.1 and 5 and m between 2 and 7,
K only varíes between 0.5 and l.
One interpretation of Eq. (5.48) is obtained by realizing that

P2 1 - X=:/): (5.50)
Pl X )2

Therefore, Eq. (5.48) is equivalent to

1
a=K.
J1
+'·J, (5.51)

and K is seen to be identical with the inverse of the parameter C 0 which

was introduced in Chap. 4. The result of the analysis is a means for
correcting homogeneous theory to allow for two-dimensional effects.
An advantage of having a model of this type is that it can be used
to derive expressions for numerous quantities, such as momentum or
kinetic-energy flux in terms of the parameters m and n. Furthermore,
1

since a is always zero at y = O (that is phase 1 must "wet" the wall) the
wall shear stress can be related to the wall velocity gradient. The form
of Eqs. (5.41) and (5.42), however, as well as the "locally homogeneous"
VELOClTY ANO CONCENTRAT\ON PROFILES m

assumption limits thc number of flow regimes to which this technique

can be applied.
The parameter K has been the basis far a variety of correlations
relating void fraction to quality in gas-liquid flow.

Example 5.3 Show that, if K is constant, a tends to a limiting value as the quality
is 'increased in an evaporator. How can greater values of a be achievcd'!
Sol-ution As the quahty is increased thc ratio (1 - x)/x tends to zcro and Eq. (5.48)
prcdicts that a tends to a hmiting value cqua.l to K. If this value is to be
exceeded, the flow regime must change to onc which has a highcr valuc of K.

Equation (5.41) can be useful even when Eq. (5.42) is quite inap-
propriate, for cxample in annular gas-liquid flow. Assume that the
liquid film can be regarded as behaving as part of an equivalent single-
phase flow which fills the pipe. Far turbulent flow m ~ 7 at low Reyn-
olds numbers and the average velocity of the equivalcnt single-phase
flow is found by integration to be
(5.52)
If the void fraction is °' the average velocity in the liquid film is

V¡ ~
7í1o
.
1
'(l - a ) !c"(l-yol
o
Vm
(Y)'" 21r(ro -
-;--
1o
y) dy (5.53)

whence
(1 - V,:,:)"(1 + ~fy;:)
Vf = 4 J'6 OVm ------'--é----~~~----'--
1 - (Y_
(5.54)

Therefore, in view of Eq. (5.52)

v_¡ ~ (1 - -y;:)~' (l + H-v;:) (5.55)

ii (1 - a)
The wall friction factor is known as a function of ii from single-phase
flow theory and the wall shear stress is
Tw ~ C¡w ½P¡V 2 (5.56)
By substitution in Eq. (5.56) from Eq. (5.55), the wall shear stress can
now be related to the average liquid velocity and calculated directly,
if the void fraction is known. This approach will be discussed in more
detail in Chap. ll.

5.5 MORE COMPLEX METHODS OF ANALYSIS

Velocity and concentration profiles in fully developed steady flow are

perhaps the simplest examples of two-dimensional effects. A vast num-
"' ONE-OIMENSIONAL TWO-PHASE FLOW

ber of more complex problems have practical importance, including two-

phase jets, developing profiles in inlet regions of pipe flow, boundary
layer problems of film boiling, condensation, ablation and combustion,
flows around bends and in centrifugal separators, interfacial waves,
droplet break-up phenomena, and motion of large bubbles in ducts and
nozzles.
Detailed development of equations which are suitably sophisticated
to solve these problems is beyond the scope or intention of this text.
However, the predictions of multidimensional analysis will often be
quoted when the results are needed for incorporation into the one-dimen-
sional theories. For instance, the analytical derivation of the equation
governing the rise velocity of a large bubble in a vertical pipe will not
be quoted but the result will be used extensively.

PROBlEIIIS
5,1. Derive Eq. (5.8) from Eqs. (5.6) and (5.7), using the single-pha.se shear stress
distribution in a round pipe for a fluid of constant density.
5.2. Show how the shear stress distribution in vertical annular gas-liquid flow depends
upon the values of i::,.p* anda.
5.3. Deduce Eq. (5.51) from (5.41) and (5.42) and the definition of C0 in Chap. 4
[Eq. (4.21)]. Show tbat Co - K-,_
5.4. Evaluate the Bankoff parameter K for n = 0.1, 1, and 5 and m = 2, 4, and 7.
5.5. Show that the dimensionless pressure drop
t,,p' - -(dp/d,) - p,g
g(p¡ - pg)
for a gas-liquid system in vertical flow represents the reading on a manomcter con-
nected to unit length of pipe, containing liquid of density p¡ below gas with density p0.
5.6. What dimensionless group would you expect to govern the rise velocity of a large
bubble of gas in an inviscid fluid?
5.7. What dimensíonless group would you expect to govern the rise velocity of a large
bubble oí gas in a very viscous fluid?
5.8. Combine the results oí Probs. 5.6 and 5.7 to show that the rise velocity can be
represcnted for liquids of any viscosity by plotting vt versus 1V ¡ where v¡ is a dimen-
sionless bubble velocity
v¡ = Vb(P1/<í[gDo(p¡ - Po)]~¼
What relationship between N 1 and v:
represents the bubble in a viscous fluid?
5.9. In gas-liquid vertical upflow the definition of i::,.p* is
t,,p' - (-dp/d, - p,g)
g(p¡ - pg)
Consider laminar or turbulent flow oí the gas or the liquid alone in a vertical pipe.
For the usual laminar-flow relations, and for a given turbulent friction factor, obtain
the following expressions relating i::,.p* to other dimensionless groups. The coordinate
z is measured upward.
VELOCITY AND CONCENTRATION PROFILES 119

(a) Laminar flow of liquid alonc,

32"*
- 1 + --'1L
Nr
(b) Laminar flow of gas alone,
32"*
b..p* = ~
N,
(e) Turbulent flow of liquid a1one,

t.p* = 1 + 2C¡j;2
(d) Turbulent flow of gas alone,

t::.p* = 2C1 j; 2
5.10.Armand 10 suggests that K = 0.833 in horizontal flow. Predict the dependen ce
of void fraction on quality for steam-water mixtures at 500, 1000, and 2000 psia and
compare with the Martinelli and Thom correlations.
s.n. Perform an integral analysis of the liquid film in laminar annular flow, starting
from the single-phase flow velocity distribution

Show that V¡ = ii(l - a) and hence that the Martinclli para.meter ,f>¡ has the value
1
ef>r = 1 - a

!U.2. By assuming a constant value of mixing length, integrate Eq. (5.6) in single-
phase flow to show that

t.p* = (7l*j*)2

and hence that

l* = (2C1 )Hi
7

Develop the same analysis for the separate-cylinders model, asSuming that l* is the
same in each cylinder and equals the single-phase flow value. Show that in this case
n = 3.5 in Eq. (3.30).
!l.13. Equations (5.32) and (5.36) can be used to plot j; versusJ;for constant values
of CT. Show that the lines of constant a are linear. VVhat is the locus of points with
constant !::.p*? Where is the flooding locus? Compare with Fig. 4.4.
For air and water at 70ºF and 15 psia in a 1-in. pipe, evaluate 1V1 , NY, and py/p¡.
Compare the jJ-versus-j; plot with the bubbly flow results shown in Fig. 4.5.
5.14. Determine an expression for momentum flux in a pipe from the Banko:ff model.
Compare with homogeneous flow theory.
5.15. lf the single-phase friction factor C 1 is related to the Reynolds number p¡iiD/µ¡
by the Blasius equation,
120 ONE-DIMENS\ONAL TWO-PHASE FLOW

deduce the relationship between jf, N¡, t:.p*, and a in horizontal flmv, using the
results of Examplc 5.2.
Show that the result is almost exactly thc same as assuming that thc wall
friction factor defined by Eq. (5.39) is rclatcd to the film Heynolds numbcr Re¡ by a
similar equation

Ctw = A(Pri;D)-H
5.16. Solve Example (5.1) far the case of horizontal flow. Show that

-32j:
-- = a 2 N1
+ 2--:- (P")
- l< a(l - a)
Nat..p* Ng PI
32j;
--- ~ (1 - a)'
N1 t:J.p*
The left-hand sides of these equations are the same as the ]'viartinelli parameters for
viscous :flow. For various valucs of (µ¡/µa), deduce thc rclationships between ,P¡
and rf,~, a and X, and compare with Figs. 3.6 and 3.8.
5.17. Use Eq. (5.42) and the values of local density and viscosity from Chap. 2 to
analyze laminar gas-liquid flow in horizontal, vertical, and inclined pipes, using a
differential tcchnique.
5.18. Show that the axial velocity profile
m-rrz 1ry
v~ = - 2apu cos 2a

and the radial velocity distribution

m . 1ry
Vy = -Sln-
2-a
Pu

can represent frictionless flow váth constant properties in bctween parallel plates
distance 2a apart when uniform condensation is occurring on each plate at a mass
rate m pcr unit area. 9 z is the axial coordinate and y the coordina.te from the center-
line normal to thc plates. VVhat is thc pressure as a function of z and y? Show that
the streamlines are normal to the walls and that thc ccnterline dynamic head is twice
the average. Deduce this latter conclusion by making an overall momentum balance
and comparing with Bernoulli's equation for the center strcamline.
5.19. Show that the solution to Eq. (5.18) for small va.lues of z is an approximation
to the fl.ow represented by the velocity potential

et,= _'!!!cosh~cos"!J/.
Pu 2a 2a
Show that this solution represents irrotational fl.ow whereas the solution to Prob.
(5.18) represents fl.ow with constant rotation of each fluid element. Determine the
pressure field in the duct.
5.20. Show that frictionless flow in a tube of radius r 0 ,vith uniform condensation on
the walls can be represented 9 by the velocity field
1rmz 1rr 2
Vz = Puro cos 2ro2
2
mro . 1rr
Vr = --sm--
pgr 2ro 2
VELOC!TY AND CONCENTRATION PROFILES 121

Sketch the velocity distributions and compare with the results of Olson and Eckert. 8
Show that the centerline dynamic head is twice the average. Determine the prossurc
:field as a function of r and z.
5.21. Discuss how the solutions to Probs. 5.18, 5.19, and 5.20 can be used to represent
reversible evaporation and condensation in a duct. Thesc results may have com-
mercial importance for the design of efficient two-phase separation equipment for
distillation, desalination, and water purification and for incrcasing thc efficiency of
thermodynamic engines employing evaporation and condensation (e.g., refrigerators,
power plants).
5.22. Show how it is possible, in the presence of condensation on the walls of a duct,
to have either positivo, negative, or zero vmll shear stress depending on the shapc of
the velocity profile.

REFERENCES

J. S00 1 S. L. 1 G. J. Trczek1 R. C. Dimick, and G. F. Hohnstrciter: I & EC Funda-

mentals, vol. 3, pp. 98-106, May, 1964.
2. Schlichting, H.: "Boundary Layer Theory," 4th ed., chap. 20, J\.JcGraw-IIill
Book Company, N ew York, 1960.
3. Prandtl, L.: "Essentials of Fluid Dynamics," p. 118, Blackie and Son, Ltd.,
Glasgow, 1952 (transl. of "Führer durch die Strümungslchre," 3d cd., ViC\veg
u. sohn, Braunschweig, 1949).
4. Deissler, R. E.: NACA Tech. Notes, no. 2129, 1950; no. 2138, 1952; and no. 3145,
1959.
5. Gill, L. E., G. F. Hewitt, and P. M. C. Lacey: Chem. Eng. Sci., vol. 19, p. 665,
1964.
6. Turner, J. M.: Ph.D. thesis, Dartmouth College, Hanover, N.H., 1966.
7. Bankoff, S. G.: Trans. ASM'E J. Heat Transfer, ser. C, vol. 82, .p. 265, 1960.
8. Olson, R.M., and E. R. G. Eckert: Trans. AS111E J. Appl. lvfech., ser. E, vol. 88,
pp. 7-17, 1966.
9. Wallis, G. B.: Trans. ASME J. Appl. Mech., ser. E, pp. 950-953, 1966.
10. Armand, A.: UKAEA, AERE transl. 828, 1959 (Izvest. Vsesoyuz. 'l'eplotekh. Inst.,
no. 1, pp. 16-23, 1946).
 6
One-dimensional Waves

6.1 INTl!ODUCTION
This chapter will be devoted to the development of one-dimensional wave
theory for both single-phase and two-phase flows. Single-phase flow is
included because many of the results which are needed for later develop-
ments are, strangely enough, not to be found in any well-known textbook
on the subject. As the concepts will be new to most readers, many
examples will beused to illustrate thephysical meaning of the mathematics.
Wave theory is a very powerful technique for analyzing unsteady
flows and transient response. I t also explains the processes such as
choking and flooding which limit the performance of equipment. In
sorne cases flow-regime changes can be attributed to instabilities which
result from wave amplification.
W aves can either propagate continuous changes in sorne variables
or can involve a step change or a finite discontinuity. The latter will
be called shock waves, or "shocks" for short. Both waves and shocks
can exist in many forros depending on the important physical processes
122
ONE-DIMENSIONAL WAVES
"'
which cause them. T,vo of the most important classes are continuityt
and dynamic waves. Continuity waves are a quasi-steady-statc phc-
nomenon and occur whenever there is a relationship betwcen flow rate
and concentration. One steady-state value simply propagates into
another one and there are no dynamic effects of inertia or momentum.
Dynamic waves, on the other hand, depend for their existence on forces
which will accelerate material through the wave as a result of concentra-
tion gradients. Since both the densities of the phases and their volu-
metric concentration are forms of cüncentration, many different kinds
of waves can occur within the two majar classes. In addition, the inter-
action between the various waves determines which, if any, dominates
the motion and can also govern the stability of the flow.
l\1any two-phase wave phenomena, particularly interfacial waves,
are two- and three-dimensional in character and will'not be discussed in
this chapter.

6.2 COIIITIIIIUITY WAVES 1111 SINGLE-PHASE FLOW

Continuity waves occur whenever the steady equilibrium flow rate of a

substance depends on the amount of that substance which is present.
For example, the flow rate of water in a river depends on the depth, the
flow rate of cars on a highway depends on the traffic density, and the
rate at which doctors interview patients depends on the number who are
waiting outside their door. More specifically, the flow rate per unít
width of a viscous fluid draining down a vertical wall ís given as a function
of thickness by the equation
q - g(p¡ - p,)ó' (6.1)
3µ¡

q is the flow rate and ó, the thickness, is the amount of fluid that is present.
In arder to achieve sorne generality and to lead on to a variety of
later developments, let us assume that the flow rate is expressed as a
suitable flux j. Denote the amount of substance by a general variable a
which can be called a concentration. a can be expressed in any units
that are consistent with the definition of j such as pounds per cubic foot,
cubic feet per cubic foot, molecules per cubic centimeter, cars per mile,
liquid depth, patients per room, etc. Ali that is required is that j should
t I first heard of a "continuity wave" indirectly from someone who had heard Professor
A. H. Shapiro use the term ata Massachusetts Institute of Technology seminar. This
name is more descriptive than "kinematic" waves, which is Lighthill and Whitham's
expression for the same thing, and is less easily confused with "dynamic" waves.
Only very simple ideas of continuity are needed in order to understand what is going
on and it is high time that the subject was given a place in elementary fluid mechanics
texts.
124 ONE-DlMENSlONAL TWO-PHASE FLOW

a h
1 1
1 1
1
1
1 a+8a

1~ 1 _______;,,,,-j+8j
" 1

1
1
a b

Fig. 6.1 Propagation of a continuity wave.

be a function of a, and should equal V times a, where V is the average

velocity.
The continuity wave velocity is derived by considering continuity
of matter across the control volume moving with the wave and bounded
by the lines aa and bb in Fig. 6.1 which shows a wavc propagating with
velocity Vw from a region where the concentration is a to a region where
the concentration is a + óa. Relative to the wave front, as much matter
is approaching as is departing. Therefore
j - aVw =j + ój - Vw(a + óa) (6.2)
whence

Vw = (:!\ (6.3)

If the change in concentration across the wave is finite, exactly the same
-reasoning leads to the result

V, = (!: =~,\ (6.4)

Subscripts denote different sides of the wave front. Such a wave will
be called a continuity shock wave.
The subscript f denotes that an equilibrium of forces is maintained
on both sides of the wave and that there are no inertia effects. For
example 1 continuity waves often occur in systems where gravity is bal-
anced against dra.g forces. Removing the inertia terms and the pressure
gradient from Eq. (3.45), written for a single-phase flow, we have
(6.5)
If b is a constant body force, then the drag force per unit volume will
also be a constant as long as the flow is not accelerating. The fact that j
is a function of a implies that this constant drag force is a function of
DNE-DIMENSIONAL WAVES 125

both j and a. In fact, for a specific system, the drag law may be known
in the form
f = f(j,a) (6.6)
Alternatively, since
j = aV (6.7)
Equation (6.6) is equivalent to
f = J(V,a) (6.8)
Using Eq. (6.7) the continuity wave velocity from Eq. (6.3) can be
expressed in the alternative form

(6.9)

which shows that the wave velocity exceeds the average velocity by the
amount a(aV /aa) 1 . One advantage of this formulation is that the result
is independent of the units in which a is measured.
If only the relationship (6.8) is knowu explicitly, we may write
(aJ/aa)v
(6.10)
(aJ/aV).
Using subscripts to denote partía! differentiation, Eqs. (6.9) and
(6.10) can be combined to give

Vw = V - "j: (6.11)

The following examples will illustrate the physical significance of

these results.

Examp!e 6.1 Calculate the continuity wave velocity for a viscous fluid flowing down
a vertical wall. Show that this velocity is three times the average velocity.
Deduce the surface profile of a film which flows clown the wall of a vessel which
is draincd from the bottom aftcr being initially full.
Solution ó represents the amount of liquid per unit width and can be identifÍed with
a. q is the flow per unit width and can be identified with j. The average
velocity o:f the liquid, from Eq. (6.7), is
V = !J. = g(p¡ - p~)oi (6.12)
o 3µ¡

The continuity wave velocity is derived from Eq. (6.3). Thus

V = iJ!1 = g(p¡ - p~)o2 (6.13)
w ªº µ¡

It follows from Eqs. (6.12) and (6.13) that

V. - 3V (6.14)
 ONE-DIMENSIONAL TWO-PHASE FLOW
"'
As the vessel drains, continuity waves will propagate values of film
thickness, each with the appropriate velocity given by Eq. (6.12). If the
surface is fl.at at z = O when the draining starts, waves corresponding to all
values of /5 start there at t = O. During the subsequent motion the larger
values of ó will propagate faster, in view of Eq. (6.13). After a time ta wave
has gane a distance

z = V.,,,t (6.15)

and the position of each wave can be represented in the zt plane by straight
lines (Fig. 6.2b).
Substituting from Eq. (6.13) we have

z = g(p¡ - p¡¡)t 152 (6.16)

µ¡

which is the equaticin of the surface profile at time t (as deduced by Jeffreys 2
and shown in Fig. 6.2a).

Example 6.2 A viscous fluid is being poured down a wall ata steady rate correspond-
ing to a thickness 01 • The flow rate is suddenly reduccd to a value correspond-
ing to a new thickness i'i2, in steady flow. Describe what happens to the liquid
surface profile.
Solution Waves will propagate from z = O exactly as in the previous example.
Only waves corresponding to values of ó between 01 and 02 will exist. The
initial step change will spread out as it moves down the wall (Fig. 6.3).

The two previous examples illustrate how the motion of continuity

waves enables the "end conditions" to be propagated throughout the flow.
This mechanism has a general significance. If continuity waves do not
propagate in this way1 the end conditions are unable to exert any influence.

---- t=O
z =O

VW¡t s,

z s, J'
t
o.,.,..x:c-,._'o" Vw/
i _,i (ll

s,

s, z

(al (b)
F'ig. 6.2 Draining of a liquid film down a wall. (a) Film surface
profile after time t. (b) Wave lines in the zt plane.
ONE-DIMENSIONAL WAVES
"'
z=O

s, Values of 8
between 81 and 82
8 t

z
lal [6)

Fig. 6.3 Behavior of a falling liquid film after the flow rate is suddenly
reduced. (a) Surfacc profile after time t. (b) Wave lines in thc zt plane.

The limiting condition occurs when the continuity wave1~ are brought to
rest and is analogous to choking due to a IVIach number of unity in the
presence of compressibility waves.

Exampie 6.3 The friction factor in turbulent flow of a falling liquid film is approxi-
mately constant. What is the ratio between the continuity wave velocity and
the average liquid velocity?
Solution The wall shear stress is

(6.17)
Identify li, the thickness, as the concentration; then f, the force pcr unit volume
is
f =~ = C1pV2
(6.18)
' 2,
Using Eq. (6.11) we have
-C 1 µV 2 /2li 2
v.~v-, c,pv¡,
whence

V.~ 3V (6.19)
2

Therefore the wave velocity is one and one-half times the average velocity.

THE FORMATION AND STABIUTY OF CONTINUITY SHOCKS

If a fast continuity wave overtakes a slower one, they will agglomerate

and form a shock. This process is represented on the zt plane by the
intersection of the wave propagation lines to form a new line with an
intermediate slope.
m ONE-DIMENS!ONAL TWO-PHASE FLOW

v, V¡
0 v.,
V,
v., (i)
Unstoble shock

V, v,
-vw2
(i)
v., v,
0
Stable shock
Concentration, a--,,,..
[o) [b)
Fig. 6.4 The dírection of stable shock propagation deduced from the curvature of
the flux-concentration curve.

In order for a shock to maintain a steep front and not degenerate

into a series of smaller shocks and waves, it is necessary that continuity
waves run toward the shock on both sides. These effects determine
whether a stable shock can occur and, if so, whether it propagates into
fluid at state 1 or state 2. Without tms consideration the direction of
the shock would be indeterminate from Eq. (6.4).
For example, Fig. 6.4a shows a continuity shock represented on the
curve off versus a. In view of Eq. (6.4) the shock velocity is given by
the slope of the chord joining points 1 and 2. The corresponding con-
tinuity wave velocities in the fluid adjacent to the shock are given by the
slopes of tangents to the curve at 1 and 2. Figure 6.4b shows that state
2 must occur on the positive side of the shock if it is to be stable.

Example 6.4. On long uniform highways it is found experimentally that there is a

relationship between the number of cars per mile (concentration) and the :flow
rate in cars per hour. There is no flow when there are either no cars at all
(a = O) or when the cars are bumper to bumper (a = amax)- There is a
maximum flow rate Ums,x) at sorne intermediate value of a. Suppose that a
steady flow of cars has existed for a long time on a highway with a value of
j1 < imax and that an accident occurs at z0 which temporarily reduces the
allowable flow rate in its vicinity to a value j 2 less than j 1• What happens and
how can it be predicted from the /-versus-a relationship?
Solution Figure 6.5a shows j as a function of a. Since the steady flow j1 has been
established for a long time, it has probably propagated downstream and there-
fore has the lowcr of the two possible values of a (the one corresponding to the
positive slope of the curve). When the accident limits the flow to jz, a new
value of a is propagated back up the highway in the negative direction. The
curve must therefore have a negative slope at the point which represents this
situation, and so the larger of the two values of a corresponding to j2 occurs.
Of\lE-DIMENSIONAL WAVES 129

The conditions for a stable shock between 0:1 and 0:2 are satisfied and this wave
moves back up the highway with the negative velocity V, shown in the figure.
After the obstruction is removed at time t', the initial shock wave at the
scene of the accident splits up into continuity waves corresponding to a continuous
spectrum of values of o:. Those to the left of the maximum in thc curve
propagate forward and those to the right propagato backward. The fastest
forward-moving wave corresponds to o: = o:~ since this is the traffic density
which was set up by the flow j 2 downstream of the accident. The fastest
backward-moving wave propagates 0:2 - óo: up the highway until it coalesces
with the shock at time t = t11 • As further waves join this shock it "weakens,"

h
e

t□ n- 1 V/
--; 1

! _________ _
J,

a, cors / mil e
(o)

.E t
¡.e
,,

t'

z,
Position, Z
(h)

Fig. 6.5 Traffic density variations following a temporary

restriction on a highway. In (b) shock waves are repre-
sented by heavy lines and continuity waves by thin lines.
130 ONE-DIMENSlONAL TWO-PHASE FLOW

swings around in thc tz planc and asymptotically degenerates toan infinitesimal

continuity wave corresponding to a 1. The fan of waves bctween O!: and a1 in
Fig. 6.M eventually merges with the shock from a~ to a 1 which has propagated
downstream from the accident.
The traffi.c density as a function of t and z is readily determined from
Fig. 6.5b. In the large open areas of thc gra.ph the density is constant, ,vhcreas
in the region covered by the fan of lines the density varies continuously.

STABILlTY OF CONTlNUITY WAVES

As long as the wavelength is sufficiently long, it is always possible to

draw the sections aa and bb in Fig. 6.1 far enough away from the steeper
parts of the wave front so that steady flow can be assumed. However,
a quasi-steady-flow analysis, which assumes that the net forces on the
flowing substance are everywhere in balance, is unable to account far the
acceleration which must occur somewhere in the wave. Another mecha-
nism must be postulated to account far this acceleration. In many cases
this can be shown to be due to a force that is produced by concentration
gradients (i.e., a compressibility effect). The effect of concentration
gradient is conveniently analyzed by considering dynamic waves which
will be discussed in a later section.

THIE EHECT OF A SOURCE OF MATTER

Continuity wave theory requires modi:fication if there is an external

source of matter being added to the flow (for example, tributaries dis-
charging into a river, cars entering a highway from driveways, condensa-
tion on aliquid film). In this case we modily Fig. 6.1 to include a source
of matter {3, per unit length, and, far variety, consider the stationary
control volume of length oz shown in Fig. 6.6. If the units of /l are wisely
chosen we have

/ioz + j (j+oj)+~7•z (6.20)

whence

(6.21)

__ -,
1
1
1
1 a [ --+-j+'oj

l____ l___ J Fig. 6.6 Control volume for analyzing

continuity waves when there is a source
1+---dz of matter f:l per unit length.
ONE-DIMENSIONAL WAVES 131

If j is a function of a, aj/az can be replaced by (aj/aa)(aajaz) and use of

Eq. (6.3) gives

ª"+vª"=/3
at w az
(H.n)
The left-hand side of Eq. (6.22) is the total time derivative o! a far a
coordinate system moving with velocity V w• Therefore

(da) df rooving with V w

= f3 (6.23)

If ¡3 is zero, cont]nu]ty ·waves propagate values of a unchanged, as befare.

Otherwise the waves will grow ar decay, depending on the sign al /3.

Example 6.5 Salve Example 6.1 if there is uniform condensation on the film surface
at a constant rate f3 (volume of liquid per unit length per unit ·width per unit
time).
Solution Identify ,x with ó, the film thickness. Then Eq. (6.23) bccomes far a given
wave

(6.24)
Since f3 is constant, Eq. (6.24) integrates to predict that
8 - óo = Mt - lo) (6.25)
where the subscript O represents initial conditions which serve to identify the
wave.
Since the wave velocity is Vw, we have, from Eq. (6.13), again far a
given wave
dz
(6.26)
dt
Combining Eqs. (6.24) and (6.26) gives

da da/dt
(6.27)
dz dz/dt
and, on integration,

(ó' - óo')g(p¡ - Po) _ "( )

µ -1--'Z-Zo (6.28)
3
Equation (6.28) gives the wave paths in the óz plane. Elimination of ó between
Eqs. (6.25) and (6.28) gives the wave line in the zt plane in terms of the initial
parameters. Thus
3
[80 + Mt - t,)]' = '•' + µ~(, - zo) (6.29)
g(p¡ - Pu)

On the other hand, elimination of 00 between Eqs. (6.2ñ) and (6.28) gives the
value of ó as a function of z at a given time and is the surface profile,

(6.30)
m ONE-D\MENSIONAL TWO-PHASE FLOW

z
¡ t

-8

Eq. (6.33)
z
(a) (b)
Fig. 6.7 Tra.nsient draining of a viscous-liquid film in the presence of uniform con-
densation. (a) Surface pro-file and wave paths. (b) VVavc lines.

For the present problem there are two families of waves, those which
start at t 0 = O from zo = O with all values of ó and those which start at later
times from zo = O with the value óu = O. For the first family we have, from
Eqs. (6.28) to (6.30):
Wave path in óz plane:

ó3 = óoª + 3µ(3z (6 31)

g(p¡ - Pu)

Wave Iine in zt plane:

(óo + f3t) 8 = óo 3 + g(p¡3µf3Z

- Pa)
(6.32)

Surface pro.file:

o' - (o _ ~/)' + g(p¡3µ~Z

- Pu)
(6.33)

For the second family the wave paths and the surface profile are coincident and
are represented by

óª = 3µ{3z (6.34)
g(p¡ - p~)

which is the equation of the film profile in the final steady state. In the zt
plane the wave lines are the parallel curves, deduced from Eq. (6.29),

3µz ]" (6.35)

t - to+ [ g (Pr - )"'
Pu ,_,

These results are illustrated in Fig. 6.7.

ONE-DIMENSlONAL WAVES m

6.3 CONTINUITY WAVES IN INCOMPRESSIBLE TWO-COMPONENT fll!W

Consider two incompressible components 1 1 and 2, flowing in a duct of

constant cross section. For continuity reasons thc ovcrall fiow rate and
hence the mean volumetric fluxj is constant throughout the duct (although
it may Vary with time due to changes in the end condition). Denoting
the volumetric fluxes of the componcnts by J1 and J2, we have
j = j, + j, (6.36)
Let the volumetric concentration of component 2 be a. The equilibrium
condition can then be expressed as before:
f(j¡,j,,a) = const (6.37)
The technique used to derive Eqs. (6.3) and (6.4) may then be
applied in exactly the same way to show that the continuity wave and
shock-wave velocities are

(6.38)

(6.39)

To visualizc the meaning of Eqs. (6.38) and (6.39), consider that

Eq. (6.37) is plotted to give a as a function of j2, for various values of j 1
(Fig. 6.8). In gas-liquid flow, far instance, this would be a plot o! void
fraction versus gas rate far various liquid rates. In going from point 1
to point 2 on the graph across a wave one must follow a lineJ1·+ ]2 = const

fig. 6.8 Continuity waves in

incompressible two-component } 1 "" constant
flow. The continuity \Va-ve ve-
locity is given by the slope of
the tangent to the curvej1 + j2
= const. The shock vefocity
is the slope of thc chord j.oining
points on the same ,curve. J,
 ONE-DIMENSIONAL TWO-PHASE FLOW

from Eq. (6.36). Equations (6.38) and (6.39) then give the required
wave velocities directly. Graphically, the continuity wave and shock
velocities are represented by the slopes o! chords and tangents to the
curves J1 + J2 = const.
An important modification can be made by focusing attention on
the drift flux j21 which was introduced in Chap. 4.
Substituting Eq. (4.4) into Eqs. (6.38) and (6.39), we obtain

Vw = j + (ªj")
ªª j,f
(6.40)

V, = j + [(j21)i - (j21)2] (6.41)

a1 - a2 i.!

Equations (6.40) and (6.41) are always valid. They are particularly
useful in cases where the drift flux is only a function of concentration and
does not depend on the overall flux. In this case the wave velocities
differ from the volumetric average velocity by an amount which depends
only on the value o! a and the system properties. The relationship
between j21 and a then lulfills the same lunction for two-phase flow as
the j-versus-a relationship does for single-phase flow. Wave and shock
velocities can be represented by the slopes of tangents and chords, as
before.
In later chapters continuity wave theory will be applied in detail
to explain unsteady one-dimensional behavior in bubble columns, foams,
sedimentation, and fluidization.

Example 6.6 For a bubble column the expressionfor iat is determined empirically to be

(6.42)

It is also known that foams with a > 3,'i are very unstable and lead to rapid
bubble bursting and agglomeration.
A depth H of liquid is put into a vertical vessel of constant diameter.
Gas is then bubbled through with a flux j 0 = v,,,/8. What is the new height of
the bubbly mixture? If the flow rate of gas is changed slightly, how long does
it take before the whole column has adjusted to the new conditions?
Solve the same problem for a gas flux of v /4.
00

Solution Since it is zero we have, from Eq. (4.2) with component 2 identified as the
gas,
j 0 ¡ = (1 - a)j0 (6.43)
Combining this with Eq. (6.42) gives
j 0 = V"'a(1 - a) (6.44)

With j 0 = v.,.,/8, the solutions to Eq. (6.44) are a = H ± V2/4. Normally

the lower value of a will occur at the bottom of the column and the higher value
at the top (because continuity waves propagate in from the ends). However,
ONE-DIMENS!ONAL WAVES 135

we are told that thc foam is unstablc so only a = H - y'2¡4 occurs. Since
liquid is conserved, the new height H 1 is givcn by the expression
H'-H
--------¡¡,-=a

Therefore

H' ~ ~H¡ ~ J,L +~ ~ 1.17H (6.45)

- ª 2 2/4

"\Vhen the flow rate is changed slightly, the new conditions will propagate
with the continuity wave velocity given by Eq. (6.40). Since J = Ju we have 1
using Eq. (6.42)

Vw ~ j, + ,.(1 - a)(l - 3a) (6.46)

With the given values of Ju and the calculated valucs of a, therefore

Vw = 0.605v.., (6.47)

The time taken by the wave to traverse the whole column and propagate
the new value of a is

H' H
- ~ 1,935- (6.48)
Vw V"'

If the value of iu had been v,,,/4, the solution to Eq. (6.43) would have
been a = :!,~ which is a double root corresponding to flooding. Equation (6.46)
then predicts V w = O, showing that waves cannot propagate at all, thus giving
another explanation for the mechanism of flooding.

5.4 DYNAMIC WAVES

DVNAMIC WAVES IN SINGLE-COMPONENT now
Dynamic waves occur whenever a net force on a flowing substance is
produced by a concentration gradient. Let this force per unit volume
be denoted by f. For a concentration gradient oa/ oz the net force pro-
ducedJ as long as we stay in the range over which a linear relationship
applies, is
a¡ ] º" (6.49)
f = [ a(aajaz) az
Using a subscrípt notation for derivatives and replacing aajaz by
the more convenient Va, Eq. (6.49) can be rewritten

(6.50)

Consjder the case shown in Fig. 6. 9 in which a dynamic wave is

traveling with velocity e relative to stationary fluid or particles. The
 ONE-DIMENSIONAL TWO-PHASE FLOW
"' Wove velocity e relative to the fluid

Fig. 6.9 Propagation of a dy-

namic wave.

wave may be brougbt to rest by imposing a velocity - e on the whole

system. Thcn the continuity equation through the wave is

e da+ a dV ~ O (6.51)
dz dz
and the equation of motion ls
dV da
pe dz ~ Íva dz (6.52)

Eliminating dV /dz and da/dz from Eqs. (6.51) and (6.52) we obtain

c2 = _ afv(:t. (6.53)
p

Therefore the wave velocity is given by the expression

e ~- + - -
p
- ( -afva)" (6.54)

Relative to the fluid, dynamic waves run in both directions with speed e,
whereas continuity waves run in only one direction. The condition that
waves exist at all is that fva is negative. Physically this means that the
system under consideration resists compression or expansion (otherwise
it would either explode or collapse catastrophically if left to itself in an
environment at constant pressure).

EXAMPLES OF DYNAMiC WAV'ES IN SJNGlE-COMPONENT FlOW

long waves in a canal of constant width The concentration can be

measured in terms of volume per unit width per unit length, i.e., depth.
Therelore a can be identified with the water depth y in the canal. Thus

a= y (6.55)

The force on a vertical slice of fluid, width b, caused by a small

change in depth óy is easily lound to be

F ~ -pgyb (óy) (6.56)

ONE-D!MENSIONAL WAVES m

Therefore the force per unit volume averaged over the section is
F dy
f = ybdz = -pg dz (6.57)

Hence, using Eq. (6.50),

Íva = -pg (6.58)

Substituting Eqs. (6.58) and (6.55) into Eq. (6 ..54) gives at once

e= ±(gy)l' (6.59)
which is a well-known result.

Waves in a homogeneous compresslbie fluid The concentration a 1s

conveniently measured in terms of mass per unit volume, i.e.,

a= p (6.60)
The force on a fluid element per unit volume is
f _ dp _ ap dp
(6.61)
- - dz - - ap dz

Hence, from Eq. (6.50),

ap
Íva = - ap (6.62)

Substituting Eqs. (6.62) and (6.60) into Eq. (6.54) then gives tbe
familiar expression

e= ± (::y• (6.63)

The derivative ap/ap must be evaluated along a specified path.

For gases the isentropic path is usually appropriate; however, 1n sorne
special cases, other paths (e.g., isothermal) may be applicable.

DYNAMIC WAVES IN INCOMPRESSIBLE TWO-COMPONENT

IFLOW IN A CONSTANT AREA DUCT

Consider two components flowing steadily in a straight duct of constant

arca with velocities v, and v,. Apply a velocity - U to the whole system
to bring any dynamic waves to rest. The veloclties in the new frame of
reference are
V~ = V1 - U (6.64)
u (6.65)
 ONE-D!MENS10NAL TWO-PHASE FLOW

The continuity equations are

d
dz [v;(I - a)]= O (6.66)
d (6.67)
dz (v;a) = O

and the equations of motion are, from Eqs. (3.45) and (3.46),

- -dp
dz
+Ji+ b, (6.68)

- -dp
dz
+!, + b, (6.69)

Subtracting Eq. (6.69) from Eq. (6.68) we get

P1V1
' dz
dv; - p2v , dv¡ f
2 dz = 1 -
f2 + b1 - b2 (6.70)

Dynamic waves occur when the right-hand side of Eq. (6.70)

depends linearly on the concentration gradient, thus
da
f, - !, + b, - b, = -f,ª dz (6. 71)

Substituting Eq. (6.71) into Eq. (6.70) and using Eqs. (6.66) and
(6.57) to eliminate dv;Jdz and dv;Jdz gives
12 12
~
1 - a
+ P2V2
a
+ Íva = O (6.72)

Expressing v~ and v~ in terms of the original velocities from Eqs.

(6.64) and (6.65) leads to a quadratic in U,

U' (__!!!.__
1-a
+ p,_)
a
_ 2 U (__l'
1-a
IJ!t_ + P2V2)
a
2 2
+ _l'l1!_1_
1 - a
+ p,v,
a
+fa = O (6.73)
"

whence

(6.74)
Defining a weighted mean velocity (which is not the same as the
velocity of the center of gravity) by
Vo = v,p,/(1 - a) + V2P2/a (6.75)
p 1 /(1 - a) + p /a
2
ONE-DIMENSIONAL WAVES 139

and a dynamic wave velocity by

(6.76)

Equation (6.74) becomes

U= V0 ± e (6. 77)
Dynarn:ic waves therefore move relative to the weighted mean
velocity V 0 with a velocity ± e given by Eq. (6.76).
From Eq. (6.76) it can be seen that the quantity fva must not only
be negative but must also be sufficiently large to overcome the destabiliz-
ing effect of the relative motion.

AN EXAMPLE OF DYNAMIC WAVES IN INCOMPRESS!IBLE TWO-COMPONENT

tf'LOW; WAVES IN A RECTANGULAR HORIZONTAL DUCT

Consider two fluids flowing with velocities V1 and v2 .in a horizontal duct
of depth H and uniform width. There are no body forces acting in the
direction of flow; however, under the inflnence of a gradient in concentra-
tion, the variation of hydrostatic pressure across the duct will give rise
to forces f in the direction of flow which are not contained in the mean
pressure gradient.
Let the pressure at the top of the duct be p. Then the net force
per unit width on the element of lighter fluid shown shaded in Fig. 6.10 is
F, = (1 - a)H óp (6. 78)
However, for the heavier fluid the net force is

F2 = aH[óp - g(p2 - p,)H óa] (6.79)

-----+ e

H
r

j
(D

@
(1-a)H

"H
li,,.
1

1
1 1

k---dz---.1
Fig. 6.10 Dynamic ,vave propagation in two incom-
pressible stratified components in a horizontal duct.
 ONE-D\MENSIONAL TWO-PHASE FLOW

Subtracting the total force per unit volume of fluid 2 from the force
per unit volume of fluid 1, wc obtain
óp (1 - c,)H c,H
Íi - f, - óz (1 - c,)H - c,H óz [óp - g(p, - p,)H fo] (6.80)

Therefore
f, - !, - (p, - p,)gH -aª"z (6.81)

The quantity fva is therefore, lrom Eq. (6.71),

Íva - - (p, - p,)gH (6.82)
From Eq. (6.76) the velocity of dynamic waves relative to V 0 is then

e- ±
-(v,-v,) 2
[ a/ P2 + (1 - e,)/ p1 + (p, - p,)gH
]l'( Pl
1 - a
p,)-;,
+ ;;-
(6.83)
The relative velocity is seen to have a destabilizing effect because
it decreases the dynamic wave velocity. In fact, for sufficiently high
relative velocity c2 becomes negative and the flow is unstable. This
occurs when
1
(v, - v2) 2 > (p, - p1 )gH (-" + -
P2 Pl
ª) (6.84)

The above equations become invalid far wavelengths smaller than

the duct dimensions because the one-dimensional approximation breaks
clown. However, they can be shown to be the limit for long wavelengths
of the two-dimensional flow solution given by Lamb. 3
Further analysis of the flow of stratified fluids is developed by
Long. 4- 6 Of particular interest is his discussion of solitary waves with a
hyperbolic secant profile. For waves with an amplitude ó we define
the following nomenclature.

(6.85)

(6.86)
(6.87)
An equation given by Long 6 far the condition at the wave crest is

Ji'
(1 - a) 3
(1 - ~) + jf' (1 + •*) -
1 - a a 3
a
1 (6.88)

The condition under which Eq. (6.88) is satisfied by ali values of o*, i.e.,
waves of any amplitude, is

(6.89)
ONE-DI MENS!ONAL WAVES 141

In this case Eq. (6.88) reduces to

"*2 '*2
.Ji +lL3 = 1 (6.90)
(1 - a) 3 a
Now, if we !et a vary in Eq. (6.90) the resulting relationships
betweenj¡ andj, will be ellipses with an envelope which bounds the range
of allowable flow ratcs. The envelope is determined by differentiating
Eq. (6.90) with respect to a to get
'*2 "*2
Ji -lL=o (6.91)
(1 - a)' 4 a

which is identical to Eq. (6.89) (if it were not, there would be a lack of
consistency).
Eliminating a from Eqs. (6.90) and (6.91) leads to tho result
(6.92)
This equation represents the maximum possible flow rates of the
components that are possible without the formation of stationary waves
of indeterminate amplitude. Thus Eq. (6.92) defines the locus of the
flow rates at "flooding 11 of the channel.
The above derivation is of interest because of the remarkable simi-
larity between Eq. (6.92) and empírica! flooding correlations far vertical
annular flow (Chap. 11), although the dynamics are quite different in
the latter case and gravity <loes not act directly as a restoring force on
the interface.

THE EFFIECT OF COMPRESSIBlllTY ON DYNAMIC WAVES IN TWO-COMPONENT FLOW

When compressibility effects are important 1 the propagation of dynamic

waves is governed by the gradients of three "concentrations,n the
densities of the two components as well as the volumetric concentration
of one of them.
The continuity Eqs. (6.56) and (6.67) now become

Íz [v(p 1 (1 - a)] = O (6.93)

Íz (v;p,a) = O (6.94)

Changes in density are related to changes m pressure by the

equations
clp, _ clp / dz
dz - ap/ap, (6.95)
dp 2 _ dp/dz
dz - ap/ap, (6.96)
 ONE-D1MENSIONAL TWO-PHASE FLOW
'"
From Eqs. (6.93) and (6.95), therefare,

_l_ dv; _ _ 1_ da + 1 dp = O (6.97)

v[ dz 1 - a dz p,(ap/ap 1) dz

and from Eqs. (6.94) and (6.96), similarly,

_l_ dv; + ~ da + 1 dp = O (6.98)

v; dz a dz p2(ap/ap,) dz
Making the previous assumption that the f's in Eqs. (6.68) and (6.69)
depend on concentration gradient, and ignoring body forces, we have
, dv; f da
dp
P1V1 dz - Iv,x+ dz
dz = O (6.99)

, dv; f da + dp O (6.100)
P2V2 dz - 2va dz dz =
Equations (6.97) to (6.100) are faur equations far faur unknowns
and are compatible only if
1 -1 1
o
í71 1 - " p,(ap/ap,)
1 1 1
o p,(ap/ap2)
=0 (6.101)
v'2
p1vf o "
-fiva 1
o P2V; -hva 1
Evaluating the determinant and multiplying throughout by v,'
and v~ we obtain

If the compressibilities are zero, the previous result, Eq. (6.72),

is obtained. If, on the other hand, the f's can be neglected (as, far
example, in stratified flow at low relative velocity), the result is

_':_ (1 - ___v;2__)
p2v? ap/ap2
+ 1- "(1 - -----1'
ap/ap1
L) = O p1v?
(6.103)

in agreement with Eq. (3.61).

If the relative velocity is zero and both components have the
velocity v', Eq. (6.103) reduces to
a 1- a
v'2
-+--
= --~P_2_ _~P_1_ __ (6.104)
a 1 - a
p,(ap/ap,) + p,(ap/ap,)
ONE-DIMENSIONAL WAVES

Therefore compressibility waves move with velocities

U=V±c" (6.105)

where Ces denotes the velocity of a compressibility wave in stratified flow,

with no relative velocity which is, from Eqs. (6.104) and (6.105),

(6.106)

where c1 2 and c2 2 are the velocities of compressibility waves in each com-

ponent separately. Equation (6.106) shows that the wave velocity lies
between the values of c1 and c2.
It should be stressed that the ve!ocity e" is not the velocity of a
compressibility wave in a homogeneous mixture in which all relative
velocity is suppressed by means of the forces between the components.
If the f's are zero in Eqs. (6.68) and (6.69) it is impossible for both the
conditions v; = v; and dv;/dz = dv;/dz to be satisfied. For truly homo-
geneous flow, therefore, there must be sufficient friction between the
components. If there are no externa! forces apart from this interaction,
then
fi(l - a) + f,a = O (6.107)

since the action and reaction between the components are equal and
opposite. Treating thc f!ow as homogeneous and adding 1 - " times
Eq. (6.68) to " times Eq. (6.69) and neglecting body forces, we obtain

[(1 - a)p, + ap,]v , dv'

dz
dp
= - dz (6.108)

which is simply the equation of motion of a fluid with mean density Pm

given by Eq. (2.9)
Eliminating d'é'/dz from Eqs. (6.97) and (6.98) and assuming homo-
geneous flow, we get

_I
v' dz
dv' + dp
dz
[ '
p,(ap/ap,)
+ p,(ap/ap,)
1 - ' ] = O (6 109)
·
From Eqs. (6.108) and (6.109) it follows at once, in agreement with
Eq. (2.50), that
2 2 (6.110)
V~ = Cch =

where Cch is the compressibility wave velocity in a strictly homogeneous

mixture. Unlike Cca Cch does not necessarily Iie between C1 and c2 and in
1
 ONE-DlMENSlONAL TWO PHASE FLOW

sorne circumstances may be far less than either. If pz is much less than
p1 and c,2 is less than c1 2, Eq. (6.110) becomes

The lowest value of Cch is therefore at a = ½ and is

(6.112)

For air and water at atmospheric pressure, for example c2 :::::: 1100 fps,
p,/ Pl ~ 0.0012, and (c,,)min is then equal to 75 fps.
If only the void fraction of a two-component mixture is specified,
there is obviously a problem in deciding whether to use Eq. (6.106) or
Eq. (6.110) or, indeed, sorne compromise betwecn the two. Qualita-
tively, one would expect Eq. (6.110) to be true for a fine dispersion of
bubbles in a liquid, whereas Eq. (6,106) should apply when two fluid
streams flow side by side with no drag forces between them. Transient
drag forces would probably be both amplitude and frequency dependent,
and thercfore the wave velocity would be a function of these variables
as well as the properties of the componcnts. At present, knowledge of
these phenomena is very limited.

Example 6.7 In Ref. 7 an equation is derived for the speed of sound in a bubbly
mixture as follo,vs:

(a)

where

(b)

(e)

(d)

and Ra is the gas constant for the gas. Deduce this result from Eq. (6.110).
Solulion Neglecting the liquid compressibility, so that p1c1 2 can be regarded as large
compared with p2c2 2 , and rewriting Eq. (6.110) in gas-liquid nomencla.ture we
find that

(e)

where

Pm = ªPo + (1 - a.)p¡ (f)

ONE-DJMENSIONAL WAVES

The void fraction in homogencous flow is related to the volumetric :flow-

rate ratio as follows:

(g)

The mass flow-rate ratio is given in terms of the void fraction by the
equation

Wu cxp 0
~ - W1 - (1 - a)p 1 (h)

Using Eqs. (h) and (J) we obtain

(i)

Substituting Eqs. (i) and (g) into Eq. (e) gives

e' -
-Cul+f3
, -~- (1 + ')'ó (j)

Now, for the adiabatic compressibility wave in the gas alone we have

(k)
Substituting Eq. (k) into Eq. (j), taking the squarc root and using Eq. (e)
cventually gives the desired result

(l)

It is also readily shown, as in Rcf. 7, that the specific heats for the mixture
in thermal equilibrium are
C¡ + /3Cpg
Cp= l+/3 (m)
C¡ + /3Cvg
,.- 1+~ (n)

The isentropic exponent far expansion with the two phases in continua!
thermal equilibrium is therefore

(o)

This equation is equivalent to Eq. (2.25) of Example 2.2. Equation (l) will
therefore be true only if there is no mutual heat transfer; if equilibrium is
maintained 1 'Y' from Eq. (o) should replace 'Yu·

THIE EFFECT OF PHASE CHANGE

If phase change can occur during the passage of a wave, then the wave
speed depends on the degree to which equilibrium is achieved. In a very
dispersed steam-water flow, for example, the value of ap/ap for the homo-
geneous mixture can be determined from steam tables or a property chart
and the sonic velocities plotted as shown in Fig. 6.11. Many nonequilib-
 ONE-D!MENS10NAL TWO-PHASE FLOW

p, psio T, ºF
3206 706
---3000-695___ ------ -
2000-635 1, 2000
3000
1000- 5 4 5 - - - - - - - - -
100

10
~
·e 100-- 327
o

"
>
u
·t¡
o
ou
<(
___
,_ 10-193

102

1/
0.1

Ouality, x

fig. 6.11 Theoretical values for the velocity of sound in equilibrium, homogeneous
steam-water mixtures. (Karplus. 8 )

rium effects complicate this situation and bedevil devices such as low-
pressure flash evaporators. Surfac-e-tension and nucleation effects can
also be important.

6.5 THE IJIITERACTIOJII BETWEEN D'i'JIIAIIIIC AJ\lll COIIITIJIIIJITY WAVES

SINGLE-PHASE FLOW

Dynamic and continuity waves occur simultaneously in many practical

systems. The condition that both kinds of wave shall exist is simply
that the force per unit volume on the fluid or particles shall be a function
of all three variables-the velocity, the concentration, and the concentra-
tion gradient-i.e.,

f=J(v,a,!;) (6.113)

In uniform steady flow without acceleration the equilibrium condition is

ªª=o
az (6.114)
ONE-DIMENSIONAl WAVES
"'
For unsteady flow the continuity and momentum equations are

ª"'
at + v ª"'
az + "'av
dZ
= o (6.115)

av + v av)
' p ( at az
= f +b (6,116)

Now, consider an initially steady flow satisfying Eq. (6.114) with

concentration a everywhere and with a coordinate system chosen so that
V = O. Fór a smaJI perturbation a' and v from equilibrium, Eqs. (6,115)
and (6.116) become

ª"''
at + "'av
az = o (6.117)
av aa'
at = f,v + J.a + !va a;¡
¡
P (6,118)

where the subscripts denote partial differentiation as before.

Taking the z derivative of Eq. (6.118) and using Eq. (6.117) to
eliminate v from the equations yields
a2ar <i 2a' ao/ ªª'
P at' + af,. ¡¡;;, - f,----¡¡¡ + aj. ai = O (6.119)

Using the results derived previously w, can define the following

quantities which are ide.ntical with the contir.:uity and dynamic wave
velocities relative to the average velocity:

Vw = Vw - V= (6,120)

(6,121)

Since in almost any conceivable system fv is negative (otherwise

particles would tend to accelerate without limit), we can also define

E = - fp (6.122)

where Bis a positive quantity which represents "damping."

Equation (6.119) can be rewritten with the help of Eqs. (6,120) to
(6.122). Thus

B'o/
f)t 2
_ C
2 o'a"
f)z 2
+E(º"''at + Vw
aa')
f)z
= O (6.123)

The physical significance of Eq. (6,123) 1s most easily shown by

substituting the expression
(6,124)
'" ONE-DIMENSIONAL TWO-PHASE FLOW

which represents a wave with frequency w/21r and velocity U and which
grows or decays with time) depending on the sign of a. The quantities
a and w are, by definition, real. Performing the substitution we o btain,
after separating real and imaginary parts and solving for a and w 2,

(6.125)

(6.126)

From Eq. (6.126) it can be shown that the requirement that w 2

should be positive limits the value of U 2 to values between vw 2 and c2 •
Therefore, if Vw 2 is greater than c2, Vw/U will be greater than unity for
waves moving in the direction of Vw, and these waves will grow. On
the other hand, if Vw 2 is less than c2, vw/U is always less than unity and
ali waves will attenuate. The stability of the flow is governed entirely
by the relative magnitude of the dynamic and continuity wave velocities,
e and Vw. If continuity waves overtake dynamic waves, the flow is
unstable. Since the coordinate system has been chosen with V = O,
Vw must be expressed relative to the average velocity.
A familiar illustration of the above conclusion is given by the
analogous behavior of traffic streams. Lighthill and Whitham 1 have
shown how continuity w:1ve theory may be used to describe unsteady
flow in a stable stream of traflic, and their concluding remarks suggest
that instabilities similar to "roll waves" may occur in sorne cases. We
can see qualitatively what occurs by considering the dynamic waves that
are the result of the response of a driver (and bis brakes and accelerator)
to changes in local gradients of concentration. If brakes are bad, the
road surface is slippery, or if vision is obscured by dense fog then the
dynamic wave velocity is decreased. If it is decreased below the con-
tinuity wave velocity the result is a "pile up" involving large numbers
of cars. Similar effects occur under normal conditions if the drivers are
in a hurry and drive too closely together so that the continuity wave
velocity is increased beyond the dynamic wave velocity. An experienced
driver wj}l adjust his speed and separation from other cars intuitively
so that the dynamic wave velodty is always a uf actor of safety" greater
than the local continuity wave velocity.
It can also be checked that Eqs. (6.125) and (6.126) lead to the usual
results in the case where one or other wave motion only is present. For
example, if B = O, all frequencies are possible if and only if c2 = U 2 and
in th:is case a = O, and there is no amplification or attenuation (dynamic
waves). On the other hand when w beco mes very small (long waves),
vw 2 = U 2 and waves propagate unchanged with the velocity vw since a is
then zero from Eq. (6.125) (continuity waves).
ONE-DIMENSIONAL WAVES

!Example 6.8 Examine the stability of long one-dimensional waves on a viscous film
falling clown an inclined plane.
Solution Let the plane make an angle 0 with the vertical. The components of
gravity a.long and perpendicular to the plane are g cos 0 and g sin 0. Substitut-
ing these values into Eqs. (6.12), (6.13), and (6.59) for a film of thickness ó,
we find

V= gcos 0(p1 - p 11 )ó 2
(6.127)
3µ1
2
V.., = g COS 0 (p¡ - Pu)Ó
(6.128)
µf
e= ±(ógsine)H (6.129)

Instability will occur if the continuity waves overtake the downward-moving

dynamic waves, that is, if
Vw > V+ e (6.130)
Substituting from Eqs. (6.127) to (6.129) the condition for instability is found
to be

2g cos 0 iPJ - P11)ó2 > (óg sin 0)H

(6.131)
"1

Alternativcly, substituting from Eq. (6.127) and rearranging,

4 Vó(p¡ - py) > 3 tan 0 (6.132)
µf

The left-hand side of Eq. (6.132) is the usual expression for the Rcynolds number
of a falling film if p¡ » Pu·
Brooke, 9 by a more cxact method, obtained the factor 1 % rathcr than
3 on the right-hand side'of Eq. (6.132).
The instability is manifested by the appearancc of roll waves such as those
which are observed on windows and inclined streets after rainfall and on wa.11s
which áre painted too thickly.

INCOMIPRESSIBLE TWO-COMPONENT FLOW

Consider the flow of two incompressible components without change of

phase in a constant-area duct. The continuity equations for the com-
ponents are

ª"'
7ii + ª"'
az + a 7iz
V
2 av, o (6.133)
aa - "1 aa
- at az + (l ) av1
- "' 7ii
= 0 (6.134)

The equations of motion are, as usual,

p¡ av, + V¡ Ji
( at ª"') -
az + b, + J,
ap (6.135)

p, av, + v,
( 7ii ª"')
7ii -
az + b, + J,
ap (6.136)
150 ONE-D1MENSlONAL TWO-PHASE FLOW

Subtracting Eq. (6.136) from Eq. (6.135) we get

p¡ (ª:t +V¡ª::)- P2 (ª:t + v, ~:) = b, - b, + f1 -f, (6.137)

The right-hand side of Eq. (6.137) may be replaced by a new variable f.

Thus
f = -[(b, - b,) + (f, - J,)J (6.138)
f is a function of the velocity of both components, the concentration and
the concentration gradient (and possibly other derivatives of velocity
or concentration), i.e.,

(6.139)

Considering a small perturbation vi, v2 , a', óf from a uniform steady

flow V,, V,, a and retaining only the first-order terms in Eqs. (6.133),
(6.134), and (6.137), we get
da' + V aa' + a av 2 = O (6.140)
at 2
az az
aa' aa' av1
- -
at
- V1 -
az
+ (1 - a) -
az = O (6.141)

p¡ ( av,
at +V 1 ª"')
az - .2 (ª"'
at + V ª"')
p oz = -óf 2 (6.142)

The quantities v1 and v2 can be eliminated from the above equations

by differentiating Eq. (6.142) with respect to z and substituting from Eqs.
(6.140) and (6.141). The final result is

at' 1 - a a az at 1 - a
¡,,)
a2a' (___1'1_ + + 2 a2a' ( V 1p 1 + V,p,)
a
+ a a' (
2
p1V, + V,') =
2
p, a(óf) (6 _143 )
Oz 2 1 - a a az
The quantity óf is made up of contributions dueto each perturbation
as follows:
óf = f.a' + f,,v1 + f,,v, + Íva ª:,' (6.144)

Differentiating Eq. (6.144) with respect to z and substituting for av¡/az

and av,/az from Eqs. (6.140) and (6.141) gives

a(óf) = aa' ( ~ _ f,,)

az at l - a a
+aa'az (f.+ Vd,, 1 - a
_ V,!,,)
a
ONE-mMENSIONAL WAVES m

Equation (6.145) may be reconciled with general continuity wave

theory by writing f as a function of the velocity of one component and
the total volumetric flux (or volumetrie mean velocity) j. Thus
f = f'(V,,j,a) (6.146)
where
j = V,(1 - a) + V,a (6.147)
The derivatives in Eq. (6.145) can be evaluated in terms of Eq. (6.146)
as follows:
aj'
f. = (V, - V,) aj + aj'
aa (6.148)

f,, = -a
ar a¡'
+ "-a.J (6.149)
v,
f., = (1 - a) aj
ar (6.150)

Substituting Eqs. (6.148) to (6.150) into Eq. (6.145), we have

(6.151)

(6.152)

The factor multiplying aa' / az in Eq. (6.152) is seen with the help of
Eq. (6.146) to be

V2 + a av,)
(-a-
a f',i
aa
= ~a (V 2a) 1 ,i = -a
1 (ªj')a t',i
(6.153)

and therefore represents the continuity wave velocity V w that was derived
previously, Eq. (6.38).
Equations (6.152) and (6.153) can be substituted into Eq. (6.143)
to give an equation o! the lorm

+ B (ªª'
2 2 2
aat'
0/
+ 2V, aza a'at + A aaz'a' ao/)
7it + Vw az = O (6.154)

The quantity Vo is a weighted mean velocity given [see Eq. (6.75)] by

Vo = V,p,/(1 - a) + V,p,/a (6.155)
p 1/(1 - a) + p /a 2

and the specific values o! A and B are

A _ p1V1 2/(l - a)+ p,V,'/a + Íva (6.156)
- p,/(1 - a) p 2 /a +
B = -f.Ja + f.J(l - a) (6.157)
p,/(1 - a) + p /a2
152 ONE-DIMENSIONAL TWO-PHASE FLOW

Equation (6.154) is interpreted by making the substitution

(6.158)

then defining ali wave velocities in terms of their differences from V,.
Thus
u= v, u + (6.159)
V,,,= Vo + Vw (6.160)

Then a new quantity is defined which is the same as the square of the
dynamic wave velocity given in Eq. (6.76).

c2 = Vo 2 - A (6.161)

The eventual result is

(6.162)

2 B2 u2vw2 - u2
w =--
2
(6.163)
4 u u 2 - c2
The qualitative conclusions that can be drawn from Eqs. (6.162)
and (6.163) for two-component flow are exactly the same as those which
were drawn from Eqs. (6.125) and (6.126) far single-component flow.
Dynamic waves move with velocity ± e relative to the weighted average
velocity V 0 defined by Eq. (6.155). Instability results when Vw 2 > c2 in
which ca~e waves grow in the direction of Vw at a rate governed by Eq.
(6.162). If c2 is negative, the flow is always unstable.

6.6 DYNAMIC SHOCK WAIIES

The theory of dynamic shock waves can be derived from the continuity,
momentum, and energy laws across finite discontinuities. The normal
shock o! gas dynarnics and the hydraulic jump in hydraulics are both
special cases of dynamic shock waves.

NORMAL COMPRESSIBIUTY SHOCKS

Consider a stationary normal shock in a two-phase gas-liquid :flow as

shown in Fig. 6.12. The continuity equation across the shock is simply

G1 = G, (6.164)

The momentum equation is

(6.165)
ONE-DIMENSlONAL WAVES m

The energy equation is

[ G1 ( h1 + "{) + G, ( h, + ";')] 1

= [ G1(h1 + "{) + G,(h, + "l)], (6.166)

Thermodynamic relationships will enable further manipulations to be

performed between pressure, enthalpy, and density.
The solution of these equations in detail is left far later chapters.
However, as an example, we shall develop the solution far the simple
case of the isothermal homogeneous shock wave with only one of the
components compressible and obeying the perfect gas laws. Denote the
compressible component by subscript g and the other by subscript f
(although it could be a solid).
In this case Eq. (6.165) reduces to

p, + GJ¡ = p, + Gj, (6.167)

Furthermore, since only the gas is compressible,

j, - j¡ = (j,)2 - (j,)i (6.168)

Combining Eq. (6.168) with Eq. (6.167) gives

p, - p, = G[(j,), - (j,)i] (6.169)

Moreover, far isothermal flow of the gas

p,(j,)¡ = p,(j,), (6.170)

Combining Eqs. (6.169) and (6.170) we find

p, = G(j,), (6.171)
p, = G(j,)¡ (6.172)

Gf1 ----------J>,. ------f¡,. G1 2

Gg\__.___......... _______... Gg 2
h11,h 1g,P1 P2, h21,h2g
Vfl -----:,..- _______... Vf 2

vg, ----------J>,. -----f1,. V 2 g

Fig. 6.12 A normal dynamic shock wave in two-phase

gas-liquid flow.
154 ONE-DIMENSIONAL TWO-PHASE FLOW

In view of Eq. (2.81) the isothermal homogeneous Mach numbers before

and after the shock are

M,' = G(j,), (6.173)

p,
M,2 = G(j,), (6.174)
p,

Combining Eqs. (6.171) to (6.174) leads to the pleasantly simple

result
M12 = _l_ = (j,)i = 1'_2 (6.175)
M,' (j,), p,

Eddington 10 has confirmed these relationships for gas-liquid flow

at about 50 percent void fraction.

OBUQUE SHOCK WAVES

Oblique shocks can be treated in the usual way by resolving the motion
along and normal to the wave line.
Consider, for example, the isothermal homogeneous oblique shock,
as shown in Fig. 6.13.
Along the w.ave the velocity is unchanged. Therefore

j, cos /3 = j, cos (/3 - 0) (6.176)

Normal to the wave we have, from the velocity triangles,

}1 sin /3 = }IN (6.177)

j, sin (/3 - 0) = j,N (6.178)

J,

Fig. 6.13 An oblique two-phase, dynamic shock

wave.
ONE-D!MENSIONAL WAVES 155

From Eq. (6.175) for the normal components we find

(6.179)

The elimination of j,N between Eqs. (6.178) and (6.179) yields

(j,),N . . ( O) .
. , /3 = J1N
M 1 , sm cot /3 tan /3 - - JfN (6.180)

Now, since all the normal incident fluxes are equal to sin {3 times
the fluxes in the original flow direction, rearrangement of Eq. (6.180)
leads to an implicit expression far the shock angle in terms of the known
incident parameters and the wedge angle 0:

(j,) 1 = ~ cos f3
M 1 J,
sin f3 tan (/3 - e) - sin' /3 (6.181)
Jf

The derivation of further interesting quantities, such as the pressure ratio,

follows at once by using the value of /3 to rcsolve the incident flow normal
to the wave.
These results obtain good confirmation from the work of Eddington. 10

RELAXATION PHENOMENA

Since velocitles and thermodynamic properties change very rapidly across

a shock wave, the two phases may take a significant time to come to
equilibrium again. The mechanism of return to equilibrium is known
as relaxation. If relaxation is slow, the overall shock thickness can be
considerable.
Far example, conslder a suspension of small spherical particles_in a
fluid. Let the dilution be such that the fluid properties, velocities, and
temperature remain constant while the particles come to equilibrium.
Denoting the particles by the subscript s we have, far acceleration in
laminar flow, following the path of each particle (Lagrangian viewpoint),
p;;rd 3 Dv, d( )
- - Dt = 31rµo Vg - Vs (6.182)
6
whence

d'p, Dv, ( )
l8µo Dt = Vo - Vs (6.183)

The particle time constant far velocity relaxation is therefare

(6.184)
156 ONE-D1MENS10NAL TWO-PHASE FLOW

Similarly, for thermal relaxation we have

c,p,,rdª DT, = 2 k d(T - T) (6.185)

6 Dt tr ª 0 $

and the time constant for thermal rela.JCation is

t _ d 2psCs (6.186)
T - 12k,
For example, for aluminum particles [p, = 169 lb/ft3, e, = 0.21 Btu/
(lb) (°F)] of diameter 0.001 in. in air at 68ºF, t, = 5.45 msec and tT = 4.93
msec. In this time the particles will go severa! feet in a gas stream with
velocity about 1000 fps.

PROBLEMS
6.1. On a particular highway the relationship between j (cars per hour) and a (cars
per mile) is

j ~ ~ (120 - a)

What is the speed limit? ,vhat is the maximum capacity oí the highway? If a flow
of 1000 cars per hour is stopped at a traffic light for 2 min and then relcased, what
happens?
6.2. What is the maximum capacity of the highway in Prob. 6.1 if the traffic light
operates continuously, being alternately red for 1 min and green for 1 min indefinitely?
6.3. The equation in Prob. 6.1 represents each lane of a three-lane highway. If the
flow rate is 4000 cars/hr and one lane is closed by an accident for H hr, how far back
up the highway is the influence of the accident felt?
6.4. One function of the dean of a college faculty is to circulate interesting documents.
These are sent sequentially to faculty members in alphabetical order. If the rate at
which these are read by each professor is proportional to the number in his in-tray
raised to sorne power n, what happens
(a) If the dean only issues documents on Mondays.
(b) If one faculty member is away for a month.
Is any special value of n particularly desirable?
6.5. If thc flow rate in Example 6.2 is increased suddenly back to the original value,
show that a shock wave is formed. How should the flow rate be varied in order that
ó may vary as a function of time as a symmetrical triangular wave at the top of the
wall? How will this wave change its shape as it propagates? When will the first
shock wave forro?
6.6. Solve Example 6.3, using Eq. (6.3) rather than Eq. (6.11). What is the answer
if the friction factor is proportional to the Reynolds number to the power -n?
6.7. Prove that a stable continuity shock cannot occur between two points on the
j-versus-a curve if the line joining these points cuts the curve at sorne intermediate
point.
6.8. Develop continuity wave theory for single-phase flow in a duct with variable
area. Show that increase in area causes the waves to attenuate, and vice versa.
ONE-DIMENS10NAL WAVES m
6.9. Show that, if the friction factor is constant, turbulent falling films are unstable if

cot 0 > 2C1

6.10. From the result of Prob. 6.9 find the critical Froude number for the formation of
white water in rivers.
6.11. Solve Prob. 6.9 if the friction factor varies as the Reynolds number to the power
-n.
6.12. Derive the relationship between film thickness and flow rate relative to the
ground for a laminar liquid film on a belt of unit width moving verticaJly ,vith speed
V. VVhat is the maximum possiblc flow rate? For flow rates less than the maximum
there are two possible values of thickness. 11 Which values of thickness occur before
and after an obstacle such as a sharp edge which is jnserted at right angles to the flow?
Why?
6.13. Consider a liquid film on the inner wall of a rotating drum. If the drum speed
is V and viscous and gravity forces alone are important, derive the relationship
bctween circumferential liquid flow rate per unit width and film thickness for various
positions around the drum. Assume laminar flow. Show that there are two possible
configurations of the liquid film, as shown in Fig. 6.14, depending on the amount of
liquid which is present. What is the reason for the change in configuration? At
what point in regime 2 is the continuity wave velocity zero? In terms of the drum
dimensions and fluid propcrties, what is the critical ámount of liquid which is needed
to bring about regime 2?
6.14. Solve Prob. 6.13 for turbulent flow with a constant value of C1 . Under what
conditions will centrifuga! and inertia forces be significant?
6J.5. What is the continuity wave velocity in homogeneous incompressible two-phase
flow in a constant-area duct?
6.16. Steam condenses at a coilstant rate on the inside of a rotating drum and is
removed at circumferential locations (rotating with the drum) distance L apart, wherc
L is much less than the radius of the drum. As the drum rotates the condensate film
sloshes to and fro between the removal points. If the film is laminar and inertia
and centrifuga! forces

- V

Regime 1 Regime 2

Fig. 6.14 Two regimes of liquid film behavior inside a rotating

drum. (Prob. 6.13.)
 ONE-DIMENSIONAL TWO-PHASE FLOW
'"
are small, show by idealizing the sinusoidal gravitational field (resolved along the
film) by a square wave that:
(a) There are either three or four distinct groups of waves on the film, depending
on whether T, tb.e time for one-half revolution, is greater than or less than
T _ [
º -
3Ln-µ 1
2g(32(p¡ - Po)
]¡,
(b) The dimensionless mean film thickness 8/L can be represented as a function
o! T /T, and ~T /L.
(e) No shock wave is formed.
Represent the various regimes of waves on both the z5 and zt planes and explain
the behavior in detail.
6.17. Plot the homogeneous compressibility wave velocity as a function of a for air-
water mixtures at 70ºF and 15 1 100, and 1000 psia.
6.18. Equation (6.92) predicts flooding dueto the growth of stationary waves. Moving
waves can also be unstable in view of Eq. (6.84). Show how Eq. (6.84) leads to
lower limits on the flow rates if flooding is to be avoided in a long horizontal duct.
In short ducts, only those waves which are almost stationary will grow enough
to cause flooding.
6.19. Substitute

into Eq. (6.123) and salve for a' and w·'. Interpret the result physically and compare
with Eqs. (6.125) and (6.126). Is the stability condition unchanged?
6.20. Far a liquid film on a vertical surface, e = O and the motion is always unstable.
Consider a turbulent falling film on which there are disturbances of wavelength
A = 27rn0. Show that, if n is large enough for the one-dimensional idealizations to
apply, the value of U given by Eq. (6.126) is

lf n = 10 and C1 = 0.01 1 show that waves will grow by a factor of e in a distance of

about 92 times the film thickness.
6.21. Salve Prob. 6.20 for laminar flow and show how the wave amplification depends
on the Reynolds number.
6.22. A normal shock wave is moving at 200 fps and the conditions in the stationary
air-water mixture ahead of it are p = 15 psia, T = 75ºF, and a = 0.3 (a is the gas
void fraction). What are the conditions behind the shock? If the shock "bounces"
normally off a stationary wall, what values of panda are set up near the wall?
6.23. A flow oí helium and water with a = !,~, T = lOOºF, and p = 20 psia impinges
squarely on a wedge with an included angle 2tJ of 60º. Predict the shock angle and
the pressure behind the shock. The approach velocity is 400 fps.
6.24. Using the definition of flooding from Chap. 4, show that the continuity wave
velocity is ahvays zero at the fl.ooding point.
6.25. In constant property two-phase fl.ow with phase change in a duct of constant
area show that continuity waves propagate so that

(ª")
~
dl moving with V w
~ [-v, + (1 - o) ] ¡¡
't! Jo

where f.! is defined by Eq. (2.92). Compare with Eq. (6.23).

ONE-D1 MENSIONAL WAVE
"'
6.26. The following four Mach numbers can be defined by forming the ratio of the
phase velocities and the velocities of compressibility and dynamic waves:

Show that Eq. (6.102) can be.rearranged to the form

and that if Eq. (6.107) is satisfied this reduces to

6.27. If the relative velocity is small campa.red with c1 and c2, show that the wave
velocity in stratified flow is
U = V2Cl'/p2 + V1(l - Q')/p1 ±
Cl'/p2 + (1 - Q')/p1 Cea

6.28. Derive Eq. (6.88) starting from Bernoulli's equation and the condition of pressure
equality at the interface.
6.29. Show that the two-component flow results in this chapter reduce to the single-
component flow results when only one component is present.
6.30. The system described in Example 6.6 initially contains a depth H of pu.re liquid.
The steady gas flux iu = v..,/8 is then suddenly turned on and kept constant thereafter.
If there is no bubble bursting or agglomeration, describe in detail what happens.
The system extends upward to a height severa! times greater than H, at which point
any foam is allowed to overflow. There are at least three stages in the process.
6.31. Under what conditions can continuity waves result from the flow or flux being a
function of thermodynamic density rather than concentration? Develop continuity
wave theory for the case where both densities and volumetric concentration influence
the flow rate.
6.32. In slug flow, large cylindrical gas bubbles and liquid plugs alternate in series.
What is the mean sonic velocity 12 for this flow pattern? How does it compare with
the predictions of homogeneous and separated flow [Eqs. (6.110) and (6.106)]? Is
the slug-flow acoustic velocity always greater than in the other flow patterns?
6.33. (a) Use Eqs. (6.164), (6.165), and (6.166) to develop a theory of dynamic shock
waves in equilibrium single-component vapor-liquid flow.
(b) Steam at 10 psia and 90% quality is flowing in a duct and enters a stationary
dynamic shock wave. Determine the velocities on each side of the shock as a function
of the pressure on the downstream side.
6.34. (a) Use Eqs. (6.164), (6.165), and (6.166) to show how to analyze l<condensation
shocks" which occur when droplets nucleate in supercooled vapor traveling at high
speeds.
(b) Salve Prob. 6.33b if the steam is dry and supercooled but has enthalpy
corresponding to 90% quality. How does this problem differ from Prob. 6.33b?
Assume thermodynamic equilibrium behind the shock.
160 ONE-DlMENSIONAL TWO-PHASE FLOW

REFERENCES
l. Lighthill, M. J., and G. B. VVhitham: Proc. Roy. Soc. (London), vol. 229A, p. 281,
1955.
2. Jeffreys, H.: Proc. Cambridge Phil. Soc., vol. 26, pp. 204-205, 1930.
3. Lamb, Sir Horace: uHydrodynamics," 6th ed., p. 371, Dover Publications, Inc.,
New York, 1945.
4. Long, R. R.: Tellus, vol. 5, no. 7, pp. 42-57, 1953.
5. Long, R. R.: Tellus, vol. 6, no. 2, pp. 97-115, 1954.
6. Long, R. R.: Tellus, vol. 8, no. 4, pp. 460-471, 1956.
7. Huey, C. T., and R. A. A. Bryant: AS.ME paper 65-WA/FE-S, 1965.
8. Karplus, H. B.: Rept. no. C00-248, Armour Res. Found., June, 1958.
9. Brooke Benjamin, T.: J. Fluid Mech., vol. 2, p. 554, 1957.
10. Eddington, R. B.: AIAA J., Investigation of Shock Phenomena in a Supersonic
Two-phase Tunnel, AIAA paper 66-87, 1966.
ll. Van Rossum 1 J. J.: Appl. Sci. Res., sec. A, vol. 7, pp. 121-144, 1958.
12. Henry, R. E., and H. K. Fauske: Trans. Am.. Nucl. Soc., vol. 11, no. 1, p. 364,
JuneJ 1968.
 7
lnterfacial Phenomena

7,1 INTRODUCTION
The behavior of surfaces and interfaces is quite fascinating, involving
numerous physical and chemical effects. For instance the presence of
minute quantities of impurities can dramatically alter the appearance of
condensation and boiling, the stability of foams, the waviness of lakes,
and the tendency of mirrors and spectacles to "mist up."
For the purpose of. analyzing two-phase :flows, the majar importance
of interfacial phenomena is the way in which they affect the boundary
conditions tha t the various equations must satisfy, In single-phase flow
the usual requirements are that the stress and velocity fields should be
continuous. In two-phase flows finite discontinuities in certain compo-
nents of both velocity and stress are possible at interfaces,

7,2 1/ELOCITY BOUNllAR'I CONDITIONS

If there is no phase change or mass transfer, the velocity compatibility

condition is the same in one-component and two-component flows. With
161
m ~NE-DIMENSIONAL TWO-PHASE FLOW

Phose 2
r2N ,,,
Phase 1
'" Fig. 7.1 Velocity boundary conditions at
L,N an interface.

phase change, however, there is a possibility of a finite velocity across

the interface. Flow across the interface must satisfy continuity; there-
fore if there is a density change, there will also be a velocity change.
Re!ative to the interface shown in Fig. 7.1, therefore, the following com-
patibility conditions must be satisfied in the directions along and per-
pendicular to the interface.

l. Continuous tangential velocities:

V1t = V2t (7.1)

2. Continuity across the interface:

(7.2)

where m represents a flux of mass crossing the interface normally. It

is usual to assume that the average continuum velocity during phase
change is perpendicular to the interface.
If the lnterface is moving, its velocity is simply superimposed on
the above velocities.

7.3 STRESS BOUNDARY CONDITIONS

SURFACE-TIENSiON EFFECTS

Continuity of the stress field across an interface is modified by the effect

of surface tension. If the interface is curved, the pressure on one side
differs from the pressure on the other by an amount

p, - p, = u (2-R. + R,_l_) (7.3)

where Ra and Rb are radii of curvature of the interface in a pair of per-

pendicular directions (Fig. 7.2).
If the surface tension is uniform1 the shear stress is continuous across
an interface. However, gradients of surface tension can be set up by
the presence of impurities, dust 1 or surface-active agents, or by tempera-
INTERFACIAL PHENOMENA
'"

IFig. 7.2 Pressure boundary condition atan interface.·

Ra and Rb are radii of curvature in two perpendicular
directions.

ture gradients along the interface. In this case the shear stress jumps
by an amount equal to the surface-tension gradient, thus
(7.4)

Equation (7.4), which is a vector expression, can be readily provee! by

referring to Fig. 7.3 which shows an element of surface in which the x
direction is chosen in the direction of Va. Theforce fromfluid 1 is r 1 dx dy,
the force from fluid 2 is -r2 dx dy, and the surface tension force is
[ -u dy + (u + ílu dx) dy].
In the x direction, therefore, a force balance yields
(71), + ílu - (,,), = O (7.5)
whereas, in the y direction at right angles to x,
(Ti)y - (72), = Ü (7.6)
Equations (7.5) and (7.6) are the scalar components of Eq. (7.4).
When an interface meets a surface or another fluid, the three result-
ing interfacial tensions must be in equilibrium at the point where they

CD

0
fig. 7.3 Shear stress boundary conditions at an interface,
 ONE-DIMENSIONAL TWO-PHASE FLOW

meet. For three soap bubbles, for instance, the three surfaces must meet
at 120° to each other. For a gas and a liquid at aplane solid surface the
contact angle is defined as the angle between the gas-liquid interface and
the solid, measured through the Iiquid. Far equilibrium of the interface
(or from the equivalent condition of mínimum energy),

COS {3 = Cfsg - rJ'sj

(7.7)
(I fo

The contact angle is an elusive parameter to measure because the

interface responds not only to the surface tensions but to small variations
in surface cleanliness and microscopio surface roughness. Local oxida-
tion or adsorbed gases on the surface also can produce significant effects.
For advancing and receding interfaces the contact angle may be quite
different, especially if motion is rapid.
Surface tension is of great importance both to the hydrodynamic
and heat-transfer characteristics of gas-liquid systems. The waviness
and stability of an interface and the formation, entrainment, and atomi-
zation of bubbles and drops are all governed by surface tension. Bubble
nucleation in boiling depends on the way in which minute bubbles grow
from defects or cavities on the heated surface. Dropwise condensation
is also governed by contact angle and wetting phenomena.

Example 7.1 What is the difference in pressure between the inside and outside of a
vapor bubble of radius 10- 4 in. if CT = 58.9 dynes/cm?
Solution Since the curvature is the same in all directions, Eq. (7.3) becomes

2.
P2 - P1 = R
Therefore

p, - p, - (Z)(
5
~-i2;2-s4 ) (2.248 X 10-") - 6.73 psi

Exampie 7.2 Consider the stability of a plane horizontal interface between fluids of
density P1 and p 2• Let p1 be the density of the upper fluid and let it be greater
than p2. Will a small sinusoidal perturbation of the interface represented by
77 = r¡ 0 sin 21rz/L tend to grow or collapse?
Solution Since r¡ is small the curvature oí the interface is given by

1 d 277 41r 2 • 2.rz

R = dz 2 = - V 77 º sm L

The pressure change across the interface is therefore, from Eq. (7.3),
rnTERFAC!AL PHENOMENA 165

The excess hydrostatic pressure due to the perturbation is

. 2,z
g (pi - p2 )r¡o sm L

The perturbation will tend to collapse if Pi - P2 exceeds the excess hydrostatic

pressure, that is 1 if

4u1r 2
V> g(p1 - pz)

or, alternatively,

(7.8)

There is therefore a critica! wavelength Le given by Eq. (7.8) that must be

exceeded if the perturbation is to grow. Analysis of the rate of growth of this
perturbation, including the effects of inertia, reveals that the fastest growing
wavelength is J., = y'3 L This phenomcnon is known as Taylor instability . 1
0•

1.4 THE EFFECT OF PHASE CHANGE ON INTERFACIAL STRESSES

When phase change occurs across an interface, there is a transfer of mass

across the interface 1 and, if the two components are not moving at the
same velocity throughout, this mass transfer has an effect on whatever
momentum transfer may be occurring simultaneously.
The evaporating or condensing vapor crossing the interface under-
goes a change in normal velocity, according to Eq. (7.2); therefore there
is a pressure exerted from· the vapor on the liquid in either evaporation
or condensation of an amount
(7.9)
This effect is usually small.
The effect of phase change on the shear stresses is more difficult to
determine. For laminar flow on a flat plate with condensation, the situa-
tion is parallel to boundary-layer suction as discussed by Schlichting. 2
In this case there is an asymptotic solution which has the effect of remov-
ing ali viscous drag on the surface and retaining only the momentum
transfer term; i.e., for a difference in velocity between the streams of v12,
the interfacial stress is simply
(7.10)
This shear stress apparently acts on the Iiquid during condensation and
on the gas during vaporization so it is not possible at this stage to regard
condensation as merely negative evaporation. Qualitatively it appears
that condensation has the effect of reducing viscous shear on the gas
stream1 whereas evaporation reduces the shear on the liquid stream.
 ONE-DIMENSIONAl TWO-PHASE flOW

A simple model for predicting the interfacial shear in turbulent

stratified gas-liquid flow with phase change has been developed by Silver
and Wallisª using the Reynolds flux concept which was introduced in
Chap. 3. According to this model the shear stress on an interface without
phase change is due to fluid from the main stream striking the wall and
bouncing back again alter sharing its momentum with the wall. If a
mass e0 strikes per unit area per unit time, the shear stress is
To = Eo(Vg - V¡) (7.11}
In a very elementary model it can be assumed that the mass flux
is made up of two streams, one moving toward the wall and one moving
away from it, each occupying one-half of the flow area and traveling with
average speed uo (Fig. 7.4). Focusing attention on the vapor stream,
we have
(7.12)
Let there be a mass flux of m due to phase change in the direction
from liquid to vapor. This flux can be brought about by superimposing
the velocity in Eq. (7.2) on the existing turbulence pattern. The stream
with velocity u 0 toward the interface now acquires a velocity (Uo - V0n),
whereas the returning stream acquires a velocity (Uo + V11 n). The
amount of vapor which transfers momentum to the gas stream per unit
area is therefore _7~p 11 (uo + v0 n), whereas the net mass flux which transfers
momentum to the liquid is only ),-ip, (u 0 - v,.).
Assuming that the material which is transferred is initially in
equilibrium with the stream from which it carne, the "drag" forces on
the vapor and liquid streams per unit volume of the duct are
F, = -P,(v, - v¡)Hp,(uo + v,N) (7.13)
F¡ = Pi(v, - V¡)7fp 0 (uo - v,N) (7.14)

where Pi is the interfacial perimeter.

Gas
Liquid

Fig. 7.4 Reynolds flux with phase change, according

to Silver and Wallis. 3
 !NTERFACIAL PHENOMENA 167

Substituting in terms of e0 and m 1 we have

F, ~ -P,(v, - v1) (, 0 + ;) (7.15)

F1 ~ P;(v, - v¡) (, 0 - 'ii') (7.16)

If condensation instead of boiling were occurring, the sign of m would

simply be reversed in these equations. Therefore we may replace m by
mfrl which denotes a vaporization flux and can be either positive or
negative.
Identifying the vapor as phase 2, making use of Eqs. (3.42) and
(3.43) in Eqs. (3.45) and (3.46), and adding forces Fw 1 and Fw, to account
for wall shear stresses, we obtain, eventually, for steady flow the following
equations of motion:

p¡V¡
dv 1
dz - dp -
dz p¡g cos
e- Fw¡
1 - a
+ P;(v, - V¡)
A (l - a)
(
'º
- m¡,)
2
(7.17)

PuVu
dv, __ dp _
dz - dz Pug cos
e_ F wo
a
_ P,(v, - v1 )
Aa
(
Eo
+ m1o)
2
(7.18)

These results are identical with Eqs. (3.106) and (3.107).

It is sometimes convenient to separate the mass transfer into com-
11
ponents of flow" and "reCirculation," since the flux m may not neces-
sarily be in equilibrium with the wall (if it is injected through slots, for
instance). The recirculating or mixing flux for phase change in either
direction is

<m ~ €o - 21ml (7.19)

and is decreased by either boiling or condensation. Equations (7.15)

and (7.16) then result by adding the flow rn to the appropriate stream.
A similar modification of the wall shear stress as a result of mass
transfer or gas injection occurs in developing turbulent boundary layers.
In their book, Kutateladze and Leontev 4 report an equation which is
valid for injection of gas of the same molecular weight as the mainstream
and, with the present notation. 1 amounts to

€m_
Eo - ( m) '( m)-¡,
1--
4eo
1+-
4eo (7.20)

This result is shown in Fig. 7.5 compared with the results of severa!
authors who studied flow over a permeable flat plate. Equation (7.19)
168 ONE-OIMENSIONAL TWO-PHASE FLOW

0.8

0.6
Em
Eo
0.4

Eq. (7.23)
0.2

o
o 1.0 2.0 4.0
m
Eo

Fig. 7.5 Gas injection into a turbulent boundary layer. Data rcproduced from
Kut:üeladze and Leontev 4 compared with various theories.

in the following form is shown for comparison:

(7.21)

Also shown is an equation which was deduced from a more sophisticated

analysis by Si!ver and Wallis,"

(7.22)

and an equation from Spalding, 5

(7.23)

Rather surprisingly, tbe Reynolds flux model also gives reasonable

predictions in the case of laminar boundary layers. 7 It is also useful for
predicting the effect o! phase change on interlacial heat- and mass-
transfer coeflicients. For example, evaporating drops tend to shield
themselves from heat transfer from the surrounding gas. For Prandtl
and Schmidt numbers close to unity, the ratio em/<o is the same thing
as the ratio o! the heat- and mass-transfer coefficients to the value with
no evaporation. Many applications o! Reynolds flux theory are dis-
cussed by Spalding 5 in his book.
INTERFACIAL PHENOMENA
'"
The value of to is usually determined from a knowledge of singlc-
phase flow under the same conditions. Far example, if the shear stress
is given by a friction factor correlation for flow in a pipe 1

(7.24)

Equations (7.11) and (7.12) then show that

•o = 72C1p(v 0 - v1) (7.25)

u 0 = C¡(v 0 - v1 ) (7.26)

Therefore C1 could be interpreted as being the ratio between the mixing

velocity and the relative veloeity.
An alternative representation of turbulent mixing in single-phase
flows is based on the friction velocity,

u*=(~)" (7.27)

Evidently u* and u, are related by the expression

(7.28)

In general! u* js representative of velocity fluctuations in the main

stream, whereas u 0 is a suitable average which characterizes the mixing
process across the entire boundary layer.
If interfacial waves occur on the liquid film, the value of t=o is
increased in proportion to the interface friction factor.

7.5 FURTHER EFFECTS

The numerous effects of interfaces on two-phase thermodynamics and

transport phenomena are beyond the scope of this text. The interested
reader is referred to the book by Davies and Rideal. 6 Surface electrical
and adhesive effects c·an be included, if necessary, in the general one-
dimensional flow theory by making suitable additions to the forces Ji
and f 2 in the momentum equations (3.40). Sorne additional phenomena
will be discussed in later chapters.

l'ROlllEMS
7.1. Calculate m, V¡,. 1 and Vun in Eq. (7.2) for stcam evaporating from a surface at 1, 10,
100, and 1000 psia corresponding to surface heat fluxes of 104, 10 5, and 10 6 Btu/
(hr)(ft').
7.2. Calculate the pressure difference in Eq. (7.9) for the conditions of Prob. 7.1.
170 ONE-DIMENSIONAL TWO-PHASE FLOW

7.3. Liquid is sucked by the influence of surfacc tension into a long straight horizontal
capillary tube which is exposed to the same pressure at both ends. Show that the
distance which the interface has pcnetrated after time t is given by

7.4. Dynamic waves in annular vertical gas-liquid flow are influenced by surface
tension. Using Eq. (6.158), assuming a thin film, and accounting far the pressure
differences dueto surface tension, show that Eq. (6.162) is unchanged but Eq. (6.163)
becomes

Show that a spectrum of wavelengths is unstable. What is the shortest wavelength

which can grow? Which wavclength grows the fastcst?
7.5. When a droplet evaporates in a highly superheated vapor, its surface is clase to
the saturation temperature and the evaporation ratc is governed by heat transfer.
Show from an energy balance that

Em h111 m
EO = Cp !::.T ;;;-

where !::.T is thc vapor supcrheat and Cp thc vapor specific heat. By solving this
equation simultaneously with Eq. (7.21), show that the evaporation ratc is

m Cp t:.T Eo 1
= ------,;¡;- . 1 + 0.5 Cp t:.T /h¡g

and tha.t the second factor represcnts the effect of mass transfer in reducing the
effective heat transfer. How big is this effect far
(a) water at 1000 psi in steam supcrheated by 200ºF?
(b) gasoline at 14.7 psi in a 2000ºF flame.?
7.6. Use the Rcynolds flux model to relate the molal concentration of noncondensable
gases, (ca)w, near a surface at which condensation occurs ata rate m to the concentra-
tion in thefreestream, (ca) •. ShowthatanyofEqs. (7.21), (7.22), and (7.23) prcdict7

(e,).
-(-)- = l
Ca 8
+ -m
Eo
+ -2¡ (m)'
-
!cO
+ h'1gh er-or d er terms
7.7. Air flows over a flat plate which is oriented parallel to the original flow direction.
If the· air is at 70ºF, 14. 7 psia, and has a velocity of 2 fps, plot the shear stress versus
distance from the leading edge if air is blown uniformly from the surface of the plate
with a velocity of 0.01 fps.
7.8. Paper passing overa heated salid cylinder is dried ata rate of 20 lb water/ (ft 2) (hr).
It is suggested that blowing at the paper with air jets will increase the rate of drying.
What mínimum mass-transfer coefficient must exist under the jets if they are to be
effectivc? (Hint: First relate the mass-transfer cocfficient to uo.)
7.9. In the derivation of Eqs. (7.15) and (7.16) the interface velocity was assumed
equal to the liquid velocity. In general, however, the situation shown in Fig. 7.4
will apply to both sides of the interface, with Eou and Eo¡ being the Reynolds flux in the
1NTERFAC!AL PHENOMENA 171

T"/ú'('.//((flfl/1/ú'.ú'./4'.///(lfl//J'.

b _.,....__ Steam, Wg, 16/hr-ft

l /Jj~////77)/;;;;:~,
~,
Fig. 7.6 Sketch for Prob. 7.13.

vapor and liquid streams, respectively. Show that in this case

Fu = -Pi (v 0 - V¡)(fou + m/2)(€oJ + m/2)

€Og + €OJ

F _ p. (vu - v¡)(fou - m/2)(t 0¡ - m/2)

J - ' léOg + €OJ

and that Eqs. (7.15) and (7.16) will be correct to :first order in m/to if to is chosen so
that

What is the value of the interface velocity?

7.10. Estimate the -flow rate of air required for the elimination of friction drag on a
flat plate mounted parallel to an air stream with velocity 200 fps if p = 15 psia,
T ~ 80ºF.
7.U.. At what rate of vaporization will the interfacial shear vanish on a smooth water
film on the insidc wall of a 1-in.-diam pipe through which saturatcd steam is flowing
at 500 psi and 100 fps.
7.12. VVhat is the heat-transfer coefficient across a horizontal laminar film of condensate
carrying a liquid flow of r (pounds per hour per foot width) with velocity vu » v1
if the rate of condensation is such that m » €o?
7.13. Use the result of Prob. 7.12 to deduce the variation of film thickness and heat-
transfer coefficient as a function of position for tho situation shown in Fig. 7.6. Assume
uniform properties and a constant value of m. How <loes the plate separation b
influence the heat-transfer coefficient for a given steam flow rate?
7.14. Explain why infants have more difficulty breathing when they catch colds than
do adults.
7.15. Solve Prob. 7.3 if the tube is vertical.
7.16. Use the Reynolds flux model to show that an approximate expression for the
mínimum transport velocity which will prevent settling of particles fl.owing in sus-
pension in a horizontal pipe is

If c1 = 0.005, show that at the condition of mínimum transport

172 ONE-DIMENSIONAL TWO-PHASE FLOW

7.17. (a) An air bubblc is stationary in water in which thcre is a temperature gradient.
This causes surface-tension variations around the bubble which set up convective
currents in thc liquid. In which dircction is the resulting force on the bubble?
(b) Docs a bubble swirn up or down the temperaturc gradient as a result of the
force expcrienced in part a? Is the direction of motion compatible with "mínimum
energy" requiremcnts for cquilibrium?
(e) Explain how a gas bubble can be held do1vn against an upward facing
surface as a result of tcmpera.turc gradients.
7.18. Example 7 .1 refers to a steam bubble in ,vater at atmospheric prcssure. Use the
steam tables to determine the liquid superheat which is needed for the bubble to be
in equilibrium.

REFERENCES
l. Taylor, G. I.: Proc. Roy. Soc. (London), vol. A201, p. 192, 1950.
2. Schlichting, H.: "Boundary Layer Theory," 4th ed., pp. 487, 502-533, McGraw-
Hill Book Company, New York, 1960.
3. Silver, R. S., and G. B. VVallis: Proc. Inst. 111ech. Engrs., vol. 180, part I, pp. 36-40,
1965-1966.
4. Kutateladze, S. S., and A. I. Leontev: "Turbulent Boundary Layers in Compressi-
ble Gases," pp. 70-71, transl. by D. B. Spalding, Academic Press Inc., New York,
1964.
5. Spaldinp; 1 D. B.: "Convective Mass Transfer," ]\1cGraw-Hill Book Company, New
York, 1963.
6. Davies, J. T., and E. K. Rideal: "Interfacial Phenomena," 2d ed., Academic Press
Inc., New York, 1963.
7. Wallis, G. B.: Intern. J. Heat Mass Transfer, vol. 11, pp. 445-472, 1968.
two

Practical Applications
 8
Suspensions of Particles
in Fluids

U INTl!()DUC:TION

One example of a two-phase system is a suspension of particles in a fluid.

Because of the absence of surface tension, subdivision, and distortion,
this system is usually more simple than the equivalent gas-liquid or liquid-
liquid dispersion. N evertheless, many complications arise in practice
due to factors such as the wide variety of sizes and shapes of the particles,
nonuniform flow patterns, agglomeration, and interparticle forces.
Engineering applications include: fluidized beds for reduction of
uranium ore, building shell molds, and plastic coating; settling tanks,
filtration beds, ground-water flows to wells, soil compaction; pneumatic
and hydraulic conveying; plasma spraying, sand blasting, rocket exhausts
containing ash or unburnt metal powders; flow of paints, slurries, printing
inks, soups, and paper fibers in suspension.
In this chapter we shall examine a series of regimes of fluid-particle
flows each of which is governed by a balance between a limited number
of forces. For example, the dynamics of sedimentation are determined
175
 ON'E-01MENSIONAL TWO-PHASE FLOW
"'
by the interaction between buoyancy and drag forces, while the impor-
tant forces in high-speed nozzle flows are pressure, inertia, and interphase
drag. General equations containing numcrous terms but not leading to
any practica.l solutions will be avoided. The reader who is interested
in more complicated systems which are not discussed in this chapter will
need to modify the general techniques of Part One to suit his par-
ticular problem.

8.2 ONE-OIMENSIONAL VERTICAL FLOW OF A IJNIFOF!M

!NCOMPF!ESSIBLE DISPERSION WITH NO WALL FRICTION
GENERAL THEORY OF UNIFORM STEADY FLOW

Consider the steady flow of a system of particles and fluid in a vertical

duct of uniform cross section. If velocities are low enough and the duct
is large enough, wa.ll friction can be lleglected and overa.ll motion relative
to the container can have no influence on the relativo motion bctween
the particles and the fluid. For a given system 1 therefore, the rolative
motion is only a function of the local concentration, properties, and the
gravitational field and <loes not depend on the net flow rates of the com-
ponents. Adequate confirmation of thís assumptíon is provided in the
detailed experiments of Lapidus et al.'·'
The arguments which were presented in Chap. 4 lead to the
definition of a drift filL"'l: 1 j¡s, which depends only on concentration.
Denoting the volumetric concentration of the fluid (the void fraction)
by €, we obtain as befo re
(8.1)

wbere the subscript f refers to the fluid and s to the particles, and the
positive direction is chosen to be upward. It is more usual to use e to
describe fluid-salid systems, although an equally satisfactory description
can be obtained in terms of a = 1 - E.
Remembering that e refers to the continuous component, the
empírica! correlation (4.14) is seen to correspond to
(8.2)

where v is the terminal speed of a single particle in an infinite stationary

00

liquid. Al! the analytical techniques derived in Chap. 4 can then be

applied if one can predict the quantities v. and n from first principies.

TERMINAL VELOC!TY OF A SINGLE PARTICLE

The terminal velocity of a single particle is determined by balancing the

gravitational and drag forces. For spherical particles of diameter d,
SUSPENS!ONS OF PARTICLES IN FLUIDS m

for example,

1r dª ( P, ) -- CD• d' 3j1 p¡V. 2 (8.3)

6 g Pi - 1r
4
where CD• is the drag coeffícient and is a function of the Reynolds num-
ber, Re., defined by
Re = voodPr (8.4)
• µ.¡

Simplifying Eq. (8.3) and using Eq. (8.4) to eliminate v., we obtain
C R 2 -
4 d'p¡g(p, - p¡) (8.5)
Doo ec,¡ ~ 3 µ,¡2

The quantity Cn.Re. 2 is therefore independent of v. and can be

evaluated solely from known quantities.

10 6 l------+-----l-----1-----1

Ng
Q) 10 4 '---------+-----~<--------!
~

<-5''
10 3

10º'---''---------+---------+---------l

10-1'--L-'--'-'----'---'---'-~-~-~~
10- 2 10- 1 10º 10 1 10 2 10 3 10 4
Re.

Fig. 8.1 Variation of Gn,,,Re"' 2 with Re"' for spheres.

(Rowe. 3 )
 ONE-DIMENSIONAL TWO-PHASE FLOW
"'
Sine e CDoo is a function of Re"", e DooReoo 2 can be expressed analyti-
cally or graphically 3 (Fig. 8.1) as a lunction of Re.. Using this relation-
ship and Eq. (8.5) the quantity Re. and hence v. can he derived.
Most particle-fluid systems o! practica! interest have Reynolds num-
hers less than 1000 for which the drag coeflicicnt may he related to the
Reynolds numher by the equation 3• 4
24
Cn. = -R (1
e.
+ 0.15Re. 0 687
- ) (8.6)

For Reynolds numhers greater than ahout 1000, the drag coeflicient
is constant and is approximately
(8.7)

Modifications to these equations for particles which are not spheri-

cal are described by Heywood. 5
The right-hand side of Eq. (8.5) is readily expressed in terms of the
dimensionless inverse viscosity N ¡ (sometimes called the "Grashof num-
ber"), and the result is
(8.8)

E.VAL.UATION OF THE INDEX n

The index nin Eq. (8.2) has been shown by Richardson and Zaki 6 to be
primarily a lunction of the Reynolds number which was dcfined in Eq.
(8.4). A correction factor can also be introduced in terms of the ratio
of the particle diameter d to the tube diameter D. The correlation of
Richardson and Zaki over the whole range of Reynolds numbers is

Re.< 0.2 n = 4.65 + 19.5Dd (8.9a)

0.2 <Re.< 1 n = (4.35 + 17.si)Re.- 00

ª (8.9b)

l<Re.<200 n = ( 4.45 + 18 i) Re. - 0 1

- (8.9c)

200 < Re. < 500 n = 4.45 Re. - 0 -1 (8.9d)

500 < Re. n = 2.39 (8.9e)

Much larger values of n can be obtained with particles which

flocculate. 7 This usually occurs with particles with dimensions in the
micron range and particularly those particles of odd shapes which can
group themselves into spongelike flocs which remain intact. In these
cases it is advisable to deduce the dependence of drift flux on concen-
tration experimentally.
SUSPENSIONS OF PARTICLES IN FLUIDS .,,
Ir-ORCES ON THE PARTICLES AND THIE FLUID

Consider an experiment in which an array of particles is maintained

stationary in a hórizontal fluid stream and the force on a typical particle
is measured in terms of the fluid veloclty, properties, and concentration
(Fig. 8.2). The total force necessary to restrain a particle with volume
'l.lv will be

F' - 'l.lp (t, - :~) (8.10)

where f, is defined as in Chap. 3. F' can be correlated in terms of the

fluid flux and properties by defining a drag coefficient, far a given E,
as follows:
F'
(8.11)

where AP is a characteristic cross section far the particle. The drag

coefficient contains both the effect of the net pressure gradient and the
effect of forces contained in f,. j¡ 0 is the fluid flux relative to stationary
particles.
The entire volume between lines aa and bb in Fig. 8.2 must be in
equilibrium. There is no fluid momentum change and "end effects" can
be made vanishingly small by choosing a s-:fficiently largo volume.
Therefore, since the external force on the partid2s balances the pressure
gradient,

(8.12)

The subscript F denotes that the pressure gradient is entirely "frictional."

The drag force on the particles must be equal and opposite to the drag

a h
_J_ _ _ _ _ _ _ _ ¡__
lo..L'.o o o 1
~ 1 o o o 1
Fluid 1
flux,Jio O O O O 1

~I o o o
Pressure p -+--- - --~--
f < - - - - 8z _ _ _..,
1

Pressure p+Sp

a h

F'ig. 8.2 Forces on particles in an array. The particles are

held against the fluid drag by a force F' per particle.
180 ONE-DIMENSIONAL TWO-PHASE FLOW

force on the fluid. Therefore (as in Example 3.4),

(8.13)

A further useful parameter is

(8.14)

Combining Eqs. (8.10) to (8.14) ali of the various forces can be expressed
in terms of each other as follows:

- ~1 (ªp)
1 - , dz,
= - _}_!_
1-,
f, f F ¡, (8.15)
= -¡ = - f, = - ,(1 - ,)

If the drag coefficient (Cvl, can be determined as a function of

void fraction and fluid properties, ali of the forces can be evaluated.
For example, for ·a stationary fluidized bed in which particles are
supported against gravity by the upward flow of fluid around them,
Eqs. (3.45) and (3.46) reduce to
dp
dz = -gp¡ + f¡ = -gp, + f, (8.16)

Combining this with Eq. (8.15) we find that

g(p, - p¡) 'O,

(Cv), = A, (8.17)
~~p¡j¡o2

Now, Rowe's experiments 8 suggest that (Cv), can be expressed as the

product of a function of Reynolds number and a function of void fraction.
Thus
(8.18)

The quantity ,¡,(,) is the ratio between the drag force on a particle
in an assembly and the same particle alone in a fluid stream with the
same volumetric flux relative to the particles. CDs is given as a function
of the "superficial" Reynolds number in the usual way [i.e., Eqs. (8.6)
and (8.7)], where the Reynolds number to be used is defined as

Re, = p¡ÍJod (8.19)

µ.¡

For spherical particles the equilibrium condition equivalent to Eq.

(8.8) is then
(8.20)
SUSPENSIONS OF PARTICLES lN FLU!DS m

Equations (8.20) and (8.5) therefore give, for particles in equilibrium,

(8.21)
Using Eqs. (8.6) and (8.7) to express the drag coefficient and sub-
stituting Eq. (8.18), Eq. (8.21) becomes
Re < 1000:
Re~(l + 0.15Re~º·ºº 7
) = \1-(,)Re,(1 + 0.15Re, 0 · 687 )
(8.22)
Re> 1000:

(8.23)
For given fluid properties, Eqs. (8.22) and (8.23) are relationships
between Voo, E=, and j¡ 0, and may be used to derive the va.lue of hs in the
usual way.
Equations (8.22) and (8.23) can be compared with Richardson and
Zaki's correlations [Eq. (8.9)] by using Eqs. (8.1) and (8.2) to deduce that
(8.24)
and multiplying Eq. (8.24) by p1d/ µ.1 to gct
(8.25)
For Reynolds numbers greater than 1000, Eqs. (8.23) and (8.25)
show that
\1-(,) = ,-2n (8.26)
i.e., from Eq. (8.9e)
\1-(,) = ,-ns (8.27)
For very low Reynolds numbers on the other hand, Eqs. (8.22) and
(8.25) lead to the result
f(,) = ,-• (8.28)
Therefore, from Eq. (8.9a)
IV(,) = ,-4.65
(8.29)
A compromise between Eqs. (8.27) and (8.29) isº
\1-(,) = ,-4.7 (8.30)
The validity of Eq. (8.30) may be tested over the whole range of
Reynolds numbers by raising Eq. (8.25) to the power 4.7 /n and multiply-
ing the result by Eq. (8.22). This gives
Re~<I-4.7/nl(l + 0.15Re~º· 687) = Re,<1- 4 •7 l•l(l + 0.15Re,º· 687) (8.31)
182 ONE-DlMENSIONAL TWO-PHASE FLOW

Evidently Eq. (8.31) cannot be satisfíed exactly over a rangc of

Reynolds numbers. However, one can require that the function of
Rcynalds number given by Eq. (8.31) should be approximately constant
near the particular value under consideration. Differentiating Eq.
(8.31) therefore leads to
4 4
(1 - : ) Re.- 41 /n + 0.15 ( 1.687 - : ) Rc.< 0 · 687- 4 7
/nl = O (8.:J2)

whence
(1 + 0.15Re.º·' 87 )
47 (8.33)
· (1 + 0.253Re.º·"")
The values of n calculated this way are campared with the values
of Richardson and Zaki in Table 8.1. The value for Re = 1000 is taken
from Eqs. (8.26) and (8.30).
The results are sufficiently compatible for Eq. (8.30) to be regarded
as a good approximation.
Far spherical particles Eq. (8.15) can therefore be expressed as
-ffs -- -4.7c Ds43 -d-
E
p¡j¡o' (8.34)

and all the various forces can be calculated from known quantities.
Far a particle in equilibrium under gravity, Eq. (8.20) becomes an
explicit equation for the void fraction
4.7 - 3 CDsRes 2
1: - 4 N¡2 (8.35)

In a system in which the particles. are moving, the quantity j¡ 0 is

related to the relative velocity or the drift flux by
, J¡s
]fo= - - = V¡sE (8.36)
1 - e
In terms of the relative velocity, for example, Eq. (8.34) becames
-2. 7C 3 p¡V¡s2
f ,, = -, v, -d-
4
(8.37)

Table 8.1 Comparison between the va!ues of n from Eqs. (8.33) and (8.9)

Re. o 0.2 1 10 100 1000

Richardson and Zaki,

Eqs. (8.9) 4.65 4.65 4.35 3 ..\3 2.8 2.39
Eq. (8.33) 4.7 4.65 4.31 3.65 3.05 2.35
SUSPENSIONS OF PARTICLES !N FLU!DS
"'
This is comparable to a similar expression, in which E is raised to the
power -2.5, derived by Zuber 52 for the laminar regime of dispersed
two-phase flow.

8.3 PARTICULATE FLUIDIZATION

If soille particles of a given size are put in a vertical vessel and a fluid
of lower density is caused to flow upward through them with a sufficicntly
high velocity, the particles beco me fluidized; in other words, they no
longer rest on one another but are supported by the fluid and are free
to move about. Further increase in velocity causes the bed to expand
and the void fraction to increase.
In particulate fluidization the particles are uniformly dispersed in
the expanded mixture. Under sorne circumstances particulate fluidiza-
tion cannot be achieved and the fluid is either channeled through regions
of low resistance or forms bubbles that rise through the bed rather like
gas bubbles in liquids.

TH E MINl!VIU M FLUIDIZATION VELOCiTY

The fluid volumetric flux at which the bed first beco mes fluidized is known
as the minimum fiuidization velocüy. At the minimum fluidfaation veloc-
ity the drag and pressure forces on the particles just equal their weight.
Equation (8.35) is therefore valid at incipient fluidization and throughout
the expansion of the bed as long as the particles remain uniformly
dispersed.
The value of , for spherical particles which are randomly packed is
about 0.4. Therefore the value of e'· 7 is 0.0135 or ½ 4 . The mínimum
fluidization velocity is then given implicitly by Eq. (8.35) as
(Gn,Re, 2)mr = 0.018N¡2 (8.38)
Knowing the value of CDsRes2, Res is found from Fig. 8.1, or Eqs.
(8.6) and (8.7), and the mínimum fluidization velocity is then calculated
using Eq. (8.19).
The coefficient "in Eq. (8.38) compares well with the value (0.0195)
given by Rowe, 8 and also with the results of Pinchbeck and Popper'°
and van Herdens, et al. 11
The value of t: at which nonspherical particles are in contact may be
as high as 0.8. If this information is known, it can be used in Eq. (8.35)
to estimate the minimum fluidjzation velocity.

PRE.SSURE DROP THROUGH A FLUiDIZED BED

Fíxed bed The frictional pressure drop through a static pile of particles
can be correlated in the usual way by defining a friction factor e;
by the
 ONE-DIMENSIONAL TWO-PHASE FLOW
"'
equation

- (dE_)
dz ,
(8.39)

Ar is the total surface area of the particles and 1J the available flow
0

volume. For spherical particles of diameter d the ratio 'Da/ Ar is

Va € d (8.40)
Ar = 1- ,6
For particles which are not spherical, a shape factor <Pis often defined
by the equation
1J" _ 1 , d (8.41)
Ar - ;¡; 1 - , ¡¡

The average velocity through the bed is related to the volumetric

flux as follows:
j¡o
V¡= -
'
Substituting Eqs. (8.40) and (8.42) into Eq. (8.39) and redefining a

friction factor C1 =
3C'
T
we obtain

_(ªP)dz F
= 2 c1 PrÍto'
d
1- ,
é
(8.43)

For nonspherical particles d is replaced by d/<P.

The friction factor defined by Eq. (8.43) is usually correlated in
terms of a Reynolds number which is defined in terms of the average fluid
velocity [from Eq. (8.42)] and the hydraulic mean depth [from Eq. (8.40)].
Thus

(8.44)

Note that Re, is not the same as the "superficial" Reynolds number
in Eq. (8.19).
For low Reynolds numbers (Re 1 < 10) the Carman-Kozeny 12 •13
equation for laminar flow gives
90
C1= ~ (8.45)
Re¡
The coefficient is intermediate between the value of 75 in the Blake-
Kozeny14 equation, and the value of 100 recommended by Leva. 15
The dependence of C1 on higher values of Re¡ is shown in Fig. 8.3,
SUSPENSIONS OF PARTICLES IN FLUIDS 185

x Burke ond Plummer }

o Ergun Ref. 17
i:, M □ rcom
6 □ Ornan and Watson
"'II 4 Ergun

INº;;::
Cl -~
"" Simpson ~ Rodger

-'I ,
"{}º 10
1
6

0-
N
4

1 ~-~~~~~~-~~~~~~-~-~~~~--~--
! 2 4 6 810 4 6 100 6 1000 4

Prirod
Re¡= - - -
[i-elµ¡

Fig. 8.3 Friction factors for flow in porous media compared with Ergun. 17 [Eq.
(8.46)] and Simpson-Rodgcr corrcla.tion. 16 (Adapted from Ergun.17)

compared with the correlation of Simpson and Rodger 16 and an equation

from Ergun 17 as follows:
75
C¡ = ~R
e¡
+ 0.875 (8.46)

lnci¡,ienl fluidlzalion At incipient fluidization the particles 11re just sup-

ported by the fluid. Therefore from Eqs. (8.14), (8.15), and (8.16)

- (!~), = (1 ,-- ,)g(p, - p¡) (8.47)

Substituting Eqs. (8.44) and (8.45) into Eq. (8.43) we obtain

2 - é 2
(8.48)
CrRer - 2 (1 _ ,) 2 Nr

Equation (8.48) resembles Eq. (8.38) but has been arrived at by a

completely different route. Using , = 0.4 for the value at incipient
fluidization, Eq. (8.48) can be arranged to give

Nr' = 11.25CrRe¡2 (8.49)

 ONE-DIMENS!ONAL TWO-PHASE FLOW
"'
whereas Eq. (8.38) becomes

N/ = 55.5CD,Re, 2 (8.50)

In view of Eqs. (8.19) and (8.44) for, = 0.4,

Re, (8.51)
R e¡= ü.6

For laminar flow, CD, = 24/Re,, C1 = 90/Rc1 and the right side of Eq.
(8.49) is equal to 1690Re,, compared with 1330Re, for Eq. (8.50). Com-
parison between the two methods over a wide range of Reynolds numbers
is shown in Table 8.2.
The two methods are seen to be compatible within the range of
accuracy with which the quantities C1 and Cn are usually known.

The lluidized state In the fluidized state the condition of cquilibrium

for a slice of the bed parallel to the horizontal is

dp
- dz = g[,p¡ + (l - ,)p,] (8.52)

and therefore· the pressure gradient can be calculated if lf is known.

For a fluidized bed in a straight vertical duct the total frictional
pressure drop is constant and is simply equal to the submerged weight
of ali the particles in the bcd, divided by the total cross-sectional area.
If the total frictional pressure drop is plotted versus the fluid flux, curves
such as those shown in Fig. 8.4 are obtained. The point where the curves
flatten out corresponds to incipient fluidization.

Table 8.2 Comparison between values of N 1 2 at

incipient fluidization predicted by Eqs. (8.49) and (8.50)

Re, Eq. (8.49) Eq. (8 ..\0)

Simpson-Rodger 16 Ergun 17

10-2 16.9 14.1 13.3

1 1690 1410 1.\30
10 1.83 X 10' l. 73 X 10' 2.30 X 10'
50 l. 7 X J05 1.62 X 10' 2.14Xl0 5
200 1.35 X 10 6 1.38 X 10 6 l. 72 X 10 6
500 6.75Xl0' 7.23 X 10' 7. 78 X 10'
1000 2.20 X 10' 2.8 X 101 2.44 X 10 7
2000 7.7 X 101 10.9 X 101 9. 76 X 10'
SIJSPENSIONS OF PARTJCLES IN FLUIDS m

Fixed
bed Fluidized bed

Fig. 8.4 Total frictional pressure

drop versus fluid flux for a fluid-
ized bed.

SUMMARY OF CALCULATION PIROCEDURIES FOR PARTICULATE FLUIDIZED BIEDS

Slalionary bed (no net particle motion) If the fluid and particle prop-
erties are known and the fluid volumetric flux is specifiedJ the simplest
calculation procedure for spherical particles is as follows.

l. Calculate Re, from Eq. (8.19).

2. Substitute Re, for Re in Eq. (8.6) or (8.7) to calculate Cn,.
00

3. Calculate Cn,Re.'.
4. Using the value found in (3) calculate , from Eq. (8.35). If , Iies
between 0.4 and 1, then the bed is fluidized if the particles are
spherical, otherwise additional information about the value of 1:
in the packed bed is required.
5. If the bed is fluidized, calculate the pressure gradient using Eq. (8.52).
If the bed is not fluidized calculate the Reynolds number Re 1 from
Eq. (8.44) and use Fig. 8.3 to find C¡. Calculate (dp/dz)F from
Eq. (8.43) and add the hydrostatic gradient p¡g to find the net
pressure gradient.

If the particles are not spherical, the form of the equations remains
the same but correction factors should be applied as indicated by
Heywood. 5

Moving bed (net particle motion) The bed must be fluidized. It is

easiest to work with Richardson and Zaki's correlation.

l. Evaluate eDooReoo 2 from Eq. (8. 5).

2. Calculate Re. from Eqs. (8.6), (8.7), or Fig. 8.1.
3. Calculate v00 from Eq. (8.4).
 ONE-DIMENSIONAL TWO-PHASE FLOW
'"
4. Calculate n from Eq. (8.9).
5. Calculate j¡, lrom Eq. (8.2) and use Eq. (8.1) and the techniques of
Chap. 4 to calculate ,. Calculate the pressure gradient using Eq.
(8.52). lf wall friction is thought to be important, estímate it lrom
homogeneous flow theory and add to the gravitational pressure drop.

Example 8.1 '\Vater a.t 20ºC flows vertically upward in a tube of cross-sectional area
1 in. 2 at arate of 6.5 X 10- 5 ft3/sec. The tube contains 0.01-in.-diam cop-
per shot. Is the bed fiuidized? Calculate the pressure gra.dient and void
fraction e.
Soluti·on Vv~ orking in cgs units for colivenience we have: D = 2.87 cm, A = 6.45
cm2, µ¡ = 10- 2 g/(cm)(sec), p 8 = 8.92 g/cm 3 , p¡ = 1.00 g/cm2, d = 0.0254
cm, and Q 1 = I.84 cm 3 /sec.
The liquid flux in the tube is

Q¡ = 1.s 4
.
J¡o = A .4
6 5
= O.28'0 cm / sec (a)

From Eq. (8.19)

Re - dj¡op¡ - (2.54 X 10- 0)(0.285)(1.0) - 0 _725 (b)
• µ,¡ 0.01

From Eq. (8.6)

Cn .• - 37.1 (e)

Using Eqs. (b) and (e),

CnsRe,' - (37.1)(0.725)' - 19.5 (d)

From Eq. (8.35) and thc definition of N 1

e4.7 = ~ Cn.Re.2µ,2
4p 1 dªg(p, - p¡)
(0.75) (19.5) (O.O!)' (e)
0 0115
(10 ')(2.54)'(981)(7.92) - ·

Therefore the predicted value of e is

E = (0.0115)1 14 · 7 = 0.385 (f)

Since this value is less than the usual random packing value for spheres
(Eo = 0.4) the bed is just not fluidized.
Alternatively we can use Eq. (8.38) to calculate the value of Cn.Re. 2
which will just fiuidizc the bed. Thus

(Cn.Re/)mr = 0.018 p¡dªg(p, - pj)

µ¡ 2
(0.018) (1) (2.54 X 10- 0 )'(981) (7.92) (g)
10-4 - 22.9

Since the value given by Eq. (d) is less than this value the bed is not fluidized.
The void fraction is therefore the unfluidized void fraction and is about
E = Ü.40.
SUSPENSIONS OF PARTICLES IN FLUIDS
"'
The pressure drop is calculated from Eq. (8.43). First we find thc
Reynolds number using Eq. (8.44):
Re,~ dpfi,o ~ (2.54 X 10- 0)(1)(0.285) ~
121 (h)
(1-,)µ, 0.6X10" .
From Eq. (8.45)
· 90
C¡= - (i)
Re,
Using Eq. (8.43)

_ (dp) ~ 2CJPJÍJo'(l - ,) ~ (2)(90)(1)(0.0813)(0.6) ~ 4460 /( ')( ') (J.)

dz F d~ª (1.21) (0.0254)0.064 g cm sec
Adding the hydrostatic gradient p¡g we get the total pressure gradient

- :~ = 4400 + 981 = 5441 dynes/cm 3 (0.202 psi/in.) (k)

As a check we calculate what the pressure gradient would be if the bed were
just fl.uidized. It is given by Eq. (8.52)
dp
- d, ~ g[w + (1 - ,)p,] ~ 981[(0.4)(1) + (0.6)(8.92)] ~ 5640 dynes/cm'
(l)
Since the pressure gradient in Eq. (k) is less than the pressure gradient in Eq.
(l), the bed is predicted to be just not fluidized 1 in agreement with the conclusions
based on Eqs. (f) and (g).

8.4 UNSTEADY FLOW IN PARTICLE lllSPERSIONS

PROIPAGATION OF CONTINUITV WAVIES

From Eq. (6.40) the continuity wave velocity in a dispersion of particles is

(8.53)

In the case of a fluidized bed obeying Eq. (8.2), Eq. (8.53) becomes

Vw - j¡ + :, [v 00
,n(l - ,)] (8.54)

This result has been confirmed experimentally by Slis, Willemse,

and Kramers. 18
In steady state the f!uidization flux is related to the voidage by
Eq. (8.24). Therefore, small variations about the steady state propagate
with a velocity given by eliminating JJ or v from Eqs. (8.24) and (8.54),
00

(1 - ,) .
Vw = n~-~110 (8.55)

or,
'
(8.56)
 ONE-DIMENSIONAL TWO-PHASE FLOW
""
The continuity shock velocity is, from Eq. (6.41),
- .
V, - J
+ (j¡,)i - (j¡,), (8.57)
E:1 - €2

As long as inertia and compressibility effects are negligible, unsteady

one-dimensional vertical flow is readily described in terms of the motion
of continuity waves and shocks. However 1 an additional complication
arises when the propagation of these waves causes a density inversion due
to values of , decreasing with height. This situation is unstable and
tends to produce three-dimensional disturbances which break up the
orderly flow structure.

8.5 BATCH SEDIMENTATION

Sedimentation provides an important example of the application of con-
tinuity wave theory. If there is no net flow the process is called batch
sedimentation) whereas, if the solids are steadily removed, the operation
is called continuous thickening. In this section the case of batch sedi-
mentation will be considered in detail, but the method can be modified
to suit processes in which the net salid or fluid flows are controlled in a
specified way. 19 •2 º
Sínce the overall volumetric flux j is zero in batch sedimentation1
Eqs. (8.53) and (8.57) reduce to

Vw = f (j¡,) (8.58)
V, = (.j¡,)i - (j¡,), (8.59)
(;1 - E2

It ls usual to work in terms of the particle concentration rather than

the void fraction. Since " = 1 - ,, Eqs. (8. 58) and (8.59) become

Vw = - :" (j¡,) (8.60)

(j¡,)i - (.i1,)2 (8.61)

Vw = a1 - a2

Therefore, on a graph of j 1 s versus a, the continuity wave velocity down-

ward is represented by the slope of the tangent and the shock velocity
downward by the slope of the chord.
A typical history of a suspension which is initially uniform is as
follows: Initially the column contains a uniform two-phase mixture B,
as shown in Fig. 8.5. As settling takes place, clear liquid A begins to
appear at the top of the column and a dense sediment D at the bottom.
In between B and D there is often a region C, in which the concentration
SUSPENS!ONS OF PARTICLES IN FLU!DS
"'
A
A
A

(a)

t
AB

BC AD
Fig, 8.5 Behavior in a typical
batch sedimentation test. ( a)
Physical appearance. (b) Height
Time, t -
of interfaces as a function of time.
(h)

is not uniform. If the particles are of fairly uniform size, a sharp dis-:
continuity is formed between the layers A and B and this moves with the
velocity of the settling particles. There may or may not be a distinct
discontinuity between the regions B and C. Eventually the upper dis-
continuity meets the lower and the region B disappears altogether.
Thereafter a slow compression or compaction of the regions C and D
occurs until finaJly the sediment reaches its maximum density throughout.
Kynch 21 was apparently the first to formulate ·a comprehensive
mathematical theory of sedimentation by using a plot of total solids
flow rate against concentration which, in this case, is identical with the
J,s-Versus-a relationship.
For the usual shape of this curve, three types of batch sedimenta-
tion can be identified as follows:
Type I If a direct shock is possible from the initial val u e of a ( a 0)
to the final, fully settled value "~ (Fig. 8.6), only one stage of settling is
observed. The interface AB moves with a velocity given by the slope
of the chord joining the points a = O, a = a 0 , and the interface BD moves
with a veloclty given by the slope of the chord joining a = a 0 to a = a 00 •
Settling is complete when the two shocks meet and no compression occurs.
Far the curves of J1s versus a shown in Fig. 8.6 1 type I sedimentation
 ONE-DIMENSIONAL TWO-PHASE FLOW

Case i Case ii

(o)

AB A

h
8 AD

8D D

t
(h)
Fig. 8.6 Type I sedimentation. (a) Drift flux-concentration p1ane.
(b) Wavelines in height-time plane.

occurs for all initial voidages in case i; in case ii it is only possible for
a 0 < a1 or ao > az.
When it is not possible far a direct shock to be propagated from "'º
to the fully packed state, more complicated processes occur.
Type JI If the curve of j¡, is concave upward at <>o, the initial
settling velocity V AB is given by the slope of the chord "' ~ O, "' ~ <>o-
At the bottom of the column the fully packed state ª• is propagated
upwards with the velocity of the shock V cv which is given by the tangent
from the point a = ª""' j¡$ = O to the curve. If this point of contact is
at a2 1 the various regions in the column contain voidages in the following
ranges:

Region A: pure liquid, "' ~ O

Region B: initial voidage, a = a 0
Region C: voidages from ªº to a2
Regían D: final voidage, "'•

Since there is no abrupt change m voidage at the boundary of

SUSPENSIONS OF PARTICLES IN FLUIDS
"'
regions B and C, the interface BC may tend to be overlooked ín practice.
When the top of regían C, which moves with the velocíty of the continuity
wave corresponding to the voidage a 0 , reaches the interface AB, the
regían B disappears and the shock begíns to strengthen and slow clown.
At any subsequent time the value of a just below the shock is given by
the continuíty wave whích has just reached the interface. Thís com-
pression of region C con_tinues until the AC and CD interfaces meet and
settling is complete. The interface histories for thís type of sedimenta-
tion are shown in Fig. 8. 7 as well as the fan of continuity wave paths in
region C.
Type III If the curve of j¡, is eoncave downward at "º (Fig. 8.8),
a finíte shock occurs between the regions B and C and three dístinct
interfaces are obtained. The voidage at the top of region C is determined
by the point of contact of the tangent from the point ao, since in this case
the fastest eontinuity wave in region C is just able to keep pace with the
shock BC, which is therefore neither strengthened nor eroded. When

(o)

AB A
h

AO

co o
t
lb)

Fig. 8.7 Type II sedimentation. (a) Drift flux-

concentration plane. (b) Wavelines.
'" ONE-D1MENSlONAL TWO-PHASE FLOW

(a}

AB
A
h
B

AD
CD D
o t
(b}

Fig. 8.8 Type III sedimentation. (a) Drift flux-concen-

tration plane. (b) Wavelines.

the AB and BC interfaces meet, region B disappears and compression of

region C continues until settllng is complete. The interface history in
this case is shown in Fig. 8. 8.

A "GIENERAlJZED" REPRESENTATION OF BATCH SEDIMIENTATION 7

Let V, denote thc general velocity of the shock between clear liquid and
the upper surface of the settling particles. It is represented on the curve
of j 1, by the chord a - O, a - a and moves with the velocity }Je/a. At
time t the position of a continuity wave moving up from the bottom of
the sediment is given by

(8.62)

which is represented on the ht plane by a waveline of slope V w from the

ong1n. Since Vs and V w are both functions of a only 1 for a given fluid-
particle combination, they may be regarded as functions of each other.
SUSPENSIONS OF PARTICLES IN FLUIDS
"'
The slope of the curved part of the sedimentation curves therefore always
has a particular slope Vs when it crosses a line of slope V w from the origin.
Hence all curves for various initial heights -and concentrations have the
same shape and may be made to coincide by multiplying both ordinate
and abscissa by a suitablefactor. The factor 1/h 0a 0 (ho being the original
height) induces ali curves to pass through the point 1/ª~ when sedimen-
tation is complete, and is_the one required. Ali the results for the com-
pression stage will lie on this curve. The initial stage, in which the set-
tling velocity is constant 1 is represented by a constant value of Vs, i.e.,
by a tangent to the curve that intersects the line t = O at the point 1/a 0•
I.f the curve of j Js contains an inflexion, V w passes through a maximum
and the generalized sedimentation curve contains two branches joined
by a cusp. If a 0 is less than the value corresponding to the maximum
value of V., the initial stage is represented by tangents to the lower
branch and sedimentation of type III occurs.
If the line representing the initial stage cuts the horizontal line
1/a~, which represents the final state, befare intersecting the upper

\
\'
Type I
Type TI
Type ill
\
\
\
"'·"'· \

Of------+-----------
x--t~
hoªo

Fig. 8.9 Generalized representation of batch sedimenta-

tion. 7
 ONE-D1MENS10NAL TWO-PHASE FLOW
"'
branch at ali, type I sedimentation occurs. The three basie types of
sedimentation are shown in Fig. 8.9 related to the general curve of
Y - h/ h 0 a 0 against X - t/h 0 a 0 •
The complementary nature of the graph of j 1, against a and the
graph of X against Y is shown by the following identities:

(8.63)

(8.64)

The intercept of a tangent to the graph of X against Y with the

Y axis is given by

y - X dY - ! (8.65)
dX a

and henee from Eq. (8.63) the intercept with the X axis is l/j1 ,.
Substitution of Eqs. (8.63) and (8.64) into Eq. (8.65) yields
X(j¡,a - dj¡,)
da
1
a
(8.66)

1
X

fig. 8.10 Complementary relationships between

.graphs (_)Í ~ .ag~inst Y, and j¡, against a.
SUSPENSIONS OF PARTICLES IN FLUIDS

20

Data from Ref. 22

16 µ silica 40º(

l5x
o ªº = 0.0495
x ªº = O. 0675
y

ID
\ X
l;a 0 =0.099
o a. 0 =0.190

X
X

o
o 2 4 6 8
X. min/cm
Fig. 8.11 Sedimentation for various initia1 concen-
trations and the same initial height represented
in the XY plane. 1

whence

(8.67)

and therefore 1/X is given by the intercept of the tangent to the graph
of j¡, with the line a = O. Equation (8.64) shows that the intercept
with the a axis is 1/Y.
These interesting complementary geometrical relationships are
shown in Fig. 8.10 and make the task of deducing one curve from the
other very simple. Each curve provides more than enough information
to draw the other since both the coordinates of a point and the tangent
at the point are defined.
If j¡, is given in the standard form of Eq. (8.2), the coordinates for
the general sedimentation curve are, parametrically 1
X = [nv~a2(l - a)n-1¡-1 (8.68)
y= (n + l)a - 1 (8.69)
na 2
 ONE-DIMENSIONAL TWD-PHASE FLOW

o a¡
ª·
ª' a-
ª'

1
ª'
t
y
m
1
a;

J_
ª'
1
ª·

x-

Fig. 8.12 The various types of sedimentation repre-

sented as a function of initial concentration on the
j 1 ,a and XY planes.

For type III sedimentation it is often more accurate to draw the

initial process as a straight line joining the points 1/ao and l/j 1, on the
axes if the point of contact of tangents to the lower branch is well below
the X axis.
The ahove theory has heen verified by W allis 7 who plotted the
results of Egolf and McCabe 22 on the basis o! Y versus X. Typical
results of this treatment are shown in Fig. 8.11. For the small irregular
particles used in these experiments the value of n lay between 20 and 30,
showing a considerable effect of flocculation.
The ranges of initial a values which promote the three types of
sedimentation are shown on the aJ1s and XY curves in Fig. 8.12. If
SUSPENSIONS OF PARTICLES JN FLUIDS
"'
Eq, (8.2) is obeyed, the inflection on the aj1, cnrve and the cusp in the
XY plot occur where
2
ai=n+l (8,70)

whereas a2 is given by
a~(l + n) + [a~'(l + n) 2
- 4na~Jl'
(8, 71)
a2 = 2n

Exarnple 8.2 Consider the settling of 0.01-in.-diamcter spherical glass beads in 90%
glycerol solution at 20ºC in a 10-cm-diameter vertical tube. Describe the
subsequcnt behavior for an initial height of 100 cm and initial concentrations of
a 0 = 0.1, 0.2, 0.3, 0.4, and 0.5. How long <loes settling take in each case
before completion?

10~-----~------~---~

8 f-------\--------,--------+--------1

y _L =5
"o

4
ªº1 =::3.33

\
\
\

o 0,5
"
Qc__ _ _ _ __,__ _ _- - ' -\ ' - - ' ' - - ' - - - - - '
LO
X, sec/cm X 10- 3

Fig. 8.13 XY plot for Example 8.2.

 ONE-DIMENSlONAL TWO-PHASE FLOW

Solution To follow thc standard procedure 1ve first determine the }¡,a and XY
relationships using the parameters v.., and n.
The following data are obtained from the "Handbook of Chemistry and
Physics" (The Chcmical Rubber Company, Cleveland, Ohio).

µ¡ ~ 2.19 g/(cm)(sec) P., = 2.32 g/cm 3 p¡ = 1.23 g/cm 3

Using Eq. (8.5)
~ 4 (2.,54 X 10-')'(1.23)(981)(1.09) ~ O
e n. ¡',e. , 3 (2.19)' ·006 (a)

From Eq. (8.6) and the results of Eq. (a) we find

R
eoo
= Cn..,Re..,2
24
= 2 r:
_;)
X 10-4 (b)

From Eq. (8.4) 1 thcrefore,

v = Re..,µ.J = (2 -5 X rn-~H 2 -19 ) = 0.01753 cm/sec (e)

" dp 1 (2.,54 X I0-')(1.23)
From Eq. (8.8a) n ~ 4.65 + 19.,5 d/D ~ 4.70 (d)

Combining the results of Eqs. (e) and (d) with Eq. (8.2) we get
j 1 , = 0.01753a(l - a)º cm/sec (e)

X and Y are now found from Eqs. (8.68) and (8.69) and are uscd to draw
the graph shown in Fig. 8.13.
The final settled value of a. is about 0.6 for spheres. The value of 1/aoo
is therefore 1/0.6 = 1.667.

E
u

"'.i
-?, 50r'-------\-----''sc---+----7"';,;,S.L--+-'°':::!Co~__::Oc_.3::_ ___-----1
.Ji.
o
u
o
t
j'
~

2x10 4
Time, sec

IFig. B.14 Interface heights versus time for Example 8.2.

SUSPENSIONS OF PARTICLES !N FLUIDS m
From Eqs. (8.70) and (8.71) er:; is found to be 0.35 and er:2is 0.43. There-
fore type II sedimenta.tion occurs when 0.43 > er: 0 > 0.35.
Thc lower bound of type III scdimentation is found by drawing the
tangent to the lower branch of the XY curve which will just pass through the
intersection bctween the upper branch and the 1/er:.,, line. The range of type
III sedimentation is then 0.22 < er: < 0.35.
Type I sedimentation will therefore occur for er: 0 = 0.1, 0.2, and 0.5;
type II for er:o = 0.4; and type III for er:o = 0.3.
To get the h-versus-t graphs, and hence the history of the interfaces, the
axes in Fig. 8.13 are multiplied by thc appropriatc valucs of hoer: 0 . The resulting
curves are shown in Fig. 8.14. The overall settling times are, in hours,

"º º _1 l_º_-_2__ º_-_3__ º_A__ º_.r_,_

I __
1 1
__ __

~ 3.06 4.55 6.07 7.00

1

t, hr

8.6 PARTICLE-PARTICLE FORCES

In previous pages it has been assumed that sedimentation stops abruptly

at the point where the particles rcst on one another (a = « In fact,
00
).

if the particles (or the lattice in wbich they are arrayed) are a ble to distort
under compressive stresses, a further compaction of the bed is possible.
The force which resists this compression depcnds on thc stress-strain
relationship for the particle lattice. Thc resultant force on a particle in
the lattice due to this compression will depend on the concentration
gradient and perhaps other variables.
In order to characterize the particle-particle forces, consider a
static experiment in which an array of particles is squeezed in a horizontal
direction by a force F' acting over an area A. Assume that a suitable
mean horizontal component of stress, <T,., can be defined far the particles.
Then, from a force balance across vertical planes we have
F'
p, + u,(1 - ,) = A (8.72)

which can be rearranged to the form

F'
(u, - p)(l - ,) +p= A (8.73)

The first term in this equation can be used to define a particle pressure
p which is added to the fluid pressure in arder to determine the overall
8,

force per unit area. Thus

(8.74)
In a concentration gradient, the difference in the force per unit area
which is carried by the particles is -dp over a distance dz. However,
8
202 ONE-DIMENSIONAL TWO-PHASE FLOW

the partieles only occupy a fraction (1 - ,) o! the volume. Therefore

1
the contribution to f, from the interparticle forces is - - ddp,_ In a
1 - ' z
horizontal, static test, if F 1 , denotes the drag caused by fluid motion in
the usual way, Eqs. (3.45) and (3.46) then become

0 = _ dp + F¡, (8.75)
dz ,
O __ dp _ F 1 , _ 1 dp,
(8.76)
dz 1-, 1-,dz

Eliminating F1 , we find

(8.77)

which could have been derived lrom Eqs. (8.73) and (8.74), sinceF'/A
is constant.
Usually Ps is a function of €, which can be determined by a com-
pression test, and is independent of the fluid pressure_for most particulate
systems because on1y point contact occurs between the particles and the
hydrostatic pressure surrounds them entirely. However, if the particles
themselves are significantly compressible (gas bubbles, lor example),
p, will depend on both , and p. 87 For fibrous matcrials an empírica!
equation is often valid in the form

p,N = B(l - ,) (8.78)

The last term in Eq. (8.76) can be conveniently rewritten as

1 ap, d, d,
- - - - - =E- (8.79)
l - , a, dz dz
where

E (8.80)

is equivalent to Ísv~' which was introduced in Chap. 6, and expresses the

influence of concentration gradient on the force experienced by the
particles.

E.xample 8.3 During filtration of a fibrous material, liquid flows through a "mat"
which is held against a porous surface. How do the void fraction and pressure
vary through the mat? Assume that the Carman-Kozeny equation applies,
that 1 - 1c is small, and that the mat is built up so slowly that steady fl.ow
theory can be used.
SUSPENSIONS OF PARTICLES IN FLUIDS

Solution In terms of the fluid flux relative to stationary particles, j 10 , ,ve havc, from
Example 3.5,

F = _ 180 µ1j10 (1 - ')' (8.81)

Js d2 1:

Combining Eqs. (8.75) to (8.78) and (8.81) yields

(8.82)

Since 1 - 1: is small, we can let 1: = 1 in Eq. (8.82) and integratc with Ps = O

at z = O to get
p/1-2NJ = (1 - 2N)Cz (8.83)
where

(8.84)

The variation of void fraction with distance is obtained by using Eq. (S.78) in
Eq. (8.83). Thus

(1 - )0-2NJIN = (1 - 2N)Cz (8.85)

1: BO 2NJIN

Since N ;:::;, >:i for fi.bers, the resulting pressure and void distribution will be as
shown in Fig. 8.15.
From Eq. (8.85) thc average particle conccntration in the mat is readily
found to be related to the concentration at the wall by the equation
-- l-2N
(1 - ,) - _ N (1 - ,)w (8.86)
1

Fluid flux
Jfo

-z
Pure fluid Fiber mat Porous wall

fig. 8.15 Variation of density and pressure through a fiber

mat during filtration from a dilute suspension.
 ONE-DIMENSIONAL TWO-PHASE FLOW

1.o~~~m~-~~~~=-~~~~=-~~~~=-~~~T<
8
Uniform mat density at
s'. 6
mechanlcal compacting
-¡;; 4
pressure Ps
1

Calculated overall mat

densíty at pressure
drop D.p

Eq. {8. 88)

O.O]L_LJ___l_'Llll_L_.J_.J_l_Ll__ljj_L_.J__j__Ll_Ll_Lli__ _j__j__Ll_ll.Lll_ _l__L_j_LJ

3
4 6 2 4 106 2 4 6 10 2 4 6 10 2 4 6
Overall pressure drop D.p orcompacting pressure p 5 , cm water

Fig. 8.16 Filtration through a fiber mat. (Han and I ngmanson. 23 )

Therefore, in view of Eqs. (8.77) and (8.78) the overall pressure drop t::.p
through thc mat is related to the mean density by the equation

--- 1-2N N
(8.87)
B(I - ,) ~ l _ N óP"

The concentration (1 - 1,) is proportional to the mat density. Thus the

pressure-drop-dcnsity curve for such a mat is displaced from thc compression
stress-density curve by the factor (1 - 2N) /(1 - N). Figure 8.16 shows some
typical data for which N = 0.225. The value of (1 - 2N)/(1 - N) is 0.71.
An alternative derivation by Han and Ingmanson 23 is based on Davis' 24
modification of the Carman-Kozeny equation for fiber mats which has 1 - E
raised to the powcr ½ in Eq. (8.81). Thc result in this case is found by the
same route to be

B-(1--) _ 1 - L5N N (8.88)

-f - l-0.5Nt::.p

For N = 0.225 ,ve have (1 - 1.5N)/(l - 0.5N) = 0.75, and the difference
between the two theories can hardly be distinguished in Fig. 8.16.

8.7 UNSTEADY FLOW IN THE PF!ESENCE OF PARTICLE-PAF!TICLE FORCES

The behavior of a sediment which is being "compacted" to values of a

greater than ª• can be predicted from Eqs. (3.45) and (3.46) by omitting
the inertia terms but including the particle-particle forces. Thus

- ap - p¡g+~ F¡, (8.89)

o az
- ap - Ps9
'F¡, +Ea,
o az - 1 - , az
(8.90)
SUSPENS!ONS OF PARTICLES IN FLUIDS 205

Eliminating op/az from these equations and making use of Eq. (8.15),
we have
a,
(p, - p¡)g = E az - f1, (8.91)

f1, is a lunction o! the flow rates and thc void lraction. U sually the
value of j can be specified (for example it is zero in batch sedimentation
or a soil compaction test) and a convenient form for the f1s relationship is
f¡, = f¡,(j,j¡,e) (8.92)
The continuity equations for the fluid and for the combincd mixture
(assumed to be incompressible) are

(8.93)
and
j = const (8. 94)
Equations (8.91) to (8.94) can now be solved simultaneously.
Differentiating Eq. (8.91) with respect to z and using Eqs. (8.92)
and (8.94) we have

Eª'' + aE (ª')' - [(ªJi,) aj1 + (ªJi,) ª'] = o

az 2 0E i)z a]¡ j az ªE j az
(8.95)

Eliminating oj1/az between Eqs. (8.93) and (8.95) yields

Eª'' + aE (ª')' + (ªf¡,) [ª' - (af¡./a_e); <le] = o (8.96)

az' a, az ºJi ; at (af¡,/aJ az 1) 1

If E is zero or there is no particle-particle contact, the final term in Eq.

(8.96) is easily shown to reduce to the equation governing continuity
wave propagation during sedimentation.
If E is constant, Eq. (8.96) resembles a formo! the diffusion equa-
tion, since aJ1 Jaj 1 is negative, and can be compared with Terzaghi's
well-known equation 59 for soil consolidation,
a2E ªE
(8.97)
C az 2 = at
where c is a coeflicient of consolidation. Evidently the Terzaghi theory
is a reasonable approximation only if the diffusion of changes in E domi-
nates the motion of continuity waves.

8.8 STABILITY or FLUIDIZED SYSTEMS

The stability o! fluidized systems can be investigated by using the theory

described in Sec. 6.4. The continuity wave velocity is given by Eq.
 ONE-DIMENS!ONAL TWO-PHASE FLOW
'"
(8.53) and the dynamic wave velocity by Eq. (6.76). Note that unless
the quantity ( -f,a), which could be thought of as inverse compressibility,
is at least as great as the quantity

p¡p, (v 1 - v,) 2
a(l - a) p¡/(1 - a) + p,/a

dynamic waves are imaginary and the system is always unstable.

Unfortunately, insuffi.cient data are available at present to enable
the virtual compressibility of a fluidized bed 26 to be calculated with any
confidence.
The usual method of analysis has been to neglect J,. completely.
This leads to the conclusion that all fluidized systcms are unstable, in
thc absence of mitigating circumstances. 26- 29 However 1 the growth rate
of instabilities is very depcndent on the properties of thc system.
The growth rate of instabi1ities can be determined as a function of
wavelength by using Eqs. (6.162) and (6.163). The details of this cal-
culation are reserved for Prob. 8.24. It is found that the shortest wave-
lengths grow most rapidly. Since waves must presumably be somewhat
longer than individual particles 1 the shortest reasonable wavelength is
chosen as 20 times the particle diameter and the distance L, which this
wave must travel in order to grow by a factor of e, is computed. The
results are shown in Table 8.3 and are comparable with the values com-
puted by J ackson. 29 Also shown in the table is the valne of the Froude

Table 8.3 Stability calculation for the data of Wilhelm and Kwauk.~º L is the
distance which a disturbance i;ravels before growing by a factor of e {see Prob. 8.24)

p, L
Fluid Particle d, cm - Fe ~

L, cm
Pi 10d

Water Glass beads 0.029 2.5 0.00052 342 99

Glass beads 0.051 2.5 0.00067 266 136
Socony beads 0.336 1.6 0.00088 12.0 40
Socony beads 0.4()8 1.6 0.0099 10.6 49
Glass bcads 0.518 2.5 0.036 4.92 26
Lcad shot 0.128 10.8 0.13 2.28 29
Air Glass beads 0.029 2000 1.1 0.66 0.2
Glass beads 0.051 2000 2.03 0.54 0.3
Socony beads 0.336 1300 10.0 0.40 1. 3
Socony bcads 0.458 1300 13.0 0.39 1.8
Glass beads O. GIS 2000 40.0 o. 36 l. 9
Lead shot 0.128 9000 85.0 0.34 0.4
SUSPENSIONS OF PART!CLES IN FLUIDS 207

number 1
• 2
Fr = }_j_r,__ (8.98)
gd

which was suggested as a stability criterion by Wilhelm and Kwauk. 30

If L is short, waves grow rapidly and soon generate three-dimen-
sional disturbances as a result of density inversions. Bubbles of fluid are
formed and rise through the bed rather like gas bubbles in a liquid.

8.9 COMPRESSIBLE FLOW OF PARTICLE SUSPENSIONS

The compressible flow of gases containing dispersed particles has received

a grcat deal of attention as a result of the development of rocket tech-
nology. The high velocities which occur in rocket exhausts and the
necessity for good accuracy in thrust prcdiction require theoretical
methods for dealing with rapidly accelerating flows at high Mach number.
Thus, by far the greatcr proportion of the pu blished literature is con-
cerned with nozzle flows in which supersonic flow is achicved.
The interested reader will find the review articles of Kliegel, 31
Hoglund 1 32 and Soo 33 useful for obtaining a perspective on this topic.

ONE-DiMENSIONAL STEADY FLOW

One-dimensional steady flow is governed by the usual pairs of equa-

tions describing mass 1 momentum 1 and energy conservation. A com-
pletely general analysis is extremely complex, but it is often valid to
assume that the following effects can be neglected:

l. Heat 1 mass, and momentum transfer with the nozzle walls

2. Particle-particle forces and particle thermal (brownian) motion
3. The effect of neighboring particles on drag forces
4. Virtual or apparent mass effects
fi. Particlo interna! temperature gradients
6. Phase change and chemical reaction
7. Radiation
8. Dissociation, ionizationJ and "nonperfect gasu effects
9. The volume occupied by the particles
10. The contribution o! the mean pressure gradient to the force on the
particles (compared with the viscous drag)
11. T-v1ro- or three-dimensional effects
12. Variations in particle size and shape
13. Gravity
 ONE-OIMENS!ONAL TWO-PHASE FLOW

If any of the above assumptions prove to be invalid, the analysis

should be modified accordingly.
The continuity equations are

PaVaA = Wa = const (8.99)

Ws = const (8.100)

The momentum equation for both components together is

(8.101)

whereas the momentum equation for the particles alone takes the form

Vs
dv,
dz = 43 edps
DP, (
V¡¡- - Vs
)1
Vg - Vs
1
(8.102)

For the gas and the particles together the energy equation is

(8.103)

in which the subscript zero refers to stagnation conditions.

If the particles are incompressible, their thermal-energy equation
is, from Example 3.10,

dh, = 6h (T' _ T) (8.104)

v, cz
l Ps d ' '

The temperature r;
represents the equivalent gas temperature for heat
transfer to the particles and is not necessarily equal to the gas tempera-
ture because of the viscous dissipation in the gas layers surrounding the
particle. If the relative velocity is large the value of may approach r;
the gas stagnation temperature relative to the particles. In many cases
of interest, however, this effect is negligible, and it can be assumed that
r; = T(}.
For low particle Reynolds and Mach numbers relative to the sur-
rounding gas the usual laminar-flow equations apply to the drag coeffi-
cient and heat-transfer coefficient. Equations (8.102) and (8.104) then
become
dv, 18µ, ( )
V, -d = -d
2 Vg - Vs (8.105)
z p,

v,c, ddT, = dl;k, (T, - T,) (8.106)

z p,

e, is the particle specific heat.

SUSPENSIONS OF PART!CLES IN FLUIDS
"'
Substituting for enthalpies in terms of temperatures and assuming
constant properties, Eq. (8.103) becomes

W, c.,T, ( "")
+ 2 + W, (c,T, + 2""') = W,c.,T,o + W,c,T, 0 (8.107)

In addition, the equation of state for the gas relates changes in

pressure 1 temperature, and density. Thus,

p = p R,T,0 (8.108)

Equations (8.99), (8.101) to (8.104), and (8.108) constitute six

equations for the six unknowns, Trn Ts, Vs, Vg, p, and pg (assuming that the
flow rates and geometry are given), and they can be sol ved by a variety
of methods.

HOMOGENEOUS EQUlllBRIUM FlOW

One of the simplest results is obtained if it can be assumed that momen-

tum and energy are exchanged so rapidly that the component tempera-
tures and velocities are everywhere equal. Equations (8.105) and (8.106)
then reduce to the form
(8.109)
(8.110)
Substitution in the other equations reveals the conclusions which
were already reached in the solution to Example 2.2. The mixture
behaves as a pseudo gas with the following properties in terms of the
relative mass fraction of the particles:

R' = ~ (8.111)
1 + m,
, Cpg + m.c.
c.= 1 + m, (8.112)

'Y
,
=
Cp(J + msCs
(8.11:J)
Cv 0 + m.c.
All of the conventional one-dimensional compressible flow equations
now apply with the above parameters inserted in place of the gas prop-
erties. The sanie velocity is
e = ~ (8.114)
and is less than the sound velocity in the gas alone. The pseudo Mach
number is
V
iVI = - (8.115)
e
210 ONE-D1 MENS\ONAL TWO-PHASE FLOW

llMJTING GASES O'f NONEQUlllBRIUM FLOW

Threc further simple results are obtained by assuming that heat and
momentum transfer are either complete or nonexistent. 34 The mixture
again behaves as a pseudo gas 1vith appropriatc propertiesJ as outlined
below.

Veiocity equilibrium, füermai insulafü:.u1 Since the particles expcnence

no heat transfer, their temperature stays constant. Thus

T, = const (8.116)

Because the virtual gas density is increased by the presence of the par-
ticles1 bu t there are no other effects, the gas constant and the specific
heat at constant pressure become

(8.117)

e' =~ (8.118)
1 + 1ns
The isentropic exponent is unchanged 1 and has the value for the gas alone

-y' = 'Y, (8.119)

The sanie velocity is dctermined by the gas temperature. Thus

e= v·¡'R'T, (8.120)

Thermai equmbrlum, veiocity insulation In this case the particles move

at a low velocity and act as a heat source for the gas. To make the solu-
tion reasonable, assumptions 9 aud 10 should really be modified; however,
the Iimiting condition which is approached is

Vs = const (8.121)

The gas alone expands as if it had the propcrties

R' = R, (8.122)
e~ = CpfJ + msCs (8.123)

'Y
1
=
CpfJ + msCs
(8.124)
Cv(J + 111-sCs
Thermal and velocity insulatíon The gas <loes not interact with the
particles and expands as if it were "clean.n Aga.in, assumptions 9 and 10
should be modified to get a reasonable solution for the particle motion.
Evaluation of these bmiting cases reveals that for ma.ny practica!
cases the predicted behavior is relatively insensitive to the assumptions
SUSPENSIONS OF PARTICLES !N FLUID$ m

which are chosen. Hmvever, when greater accuracy is required, more

sophisticated methods of analysis are called for.

SIMILAR SOLUTlONS FOR CONSTANT FRACT!ONAL LAG

A family of exact solutions to flows in the laminar regime of relative

motion can be found by assuming that the ratios of the velocity and
temperature changes of the components are constants. It is also assumed
that both phases have a common stagnation temperature To 1 where they
are at rest. Following Kliegel3 5 wc define thc parameters

(8.125)

(8.126)

Equations (8.105) and (8.106) become

dv, 18µ, l - K
dz ~ d 2p, ~ (8.127)
dT, 12/c,(T 0 - T,) 1 - L
·az 2
Vad psCs KL (8.128)

The energy equation (8.107) is readily converted to the form

To - T - Vo2 1 + msK2 (8.129)

g - 2 Cpg + 1n 8
C8 L

Differentiating Eq. (8.129) we find that

1 dT, 2 dv 0
To - T, dz ~ - v, dz (8.130)

Substituting Eq. (8.130) into Eq. (8.128) the result is

dv, 6k, 1 - L
dz = d 2psCs KL (8.131)

Both Eqs. (8.127)'and (8.131) show that the gas velocity, and hence
also the particle velocity, increases linearly with distance down the nozzle.
Furthermore, since both expressions must be compatible
e,µ, 1 1/L - 1
k, 31/K - 1 (8.132)

The left-hand side of Eq. (8.132) is a kind of pseudo Prandtl number in

which the specific heat of the salid replaces the specific heat of thc gas.
This quantity determines the relationship between the thermal and
veloeity lags.
m ONE-DI MENSIONAL TWO-PHASE FLOW

Further developments are possible by using the momentum and

continuity equations. Using Eqs. (8.99), (8.12:i), and (8.126), Eq. (8.101)
becomes
dv,
p,v, dz (1 + m,K) + dp
dz = O (8.133)

which is just the one-dimensional momentum equation for a pseudo gas

with density
p' = (1 + m,K)p, (8.134)

This pseudo gas will obey a perfect-gas law equivalent to Eq. (8.108) if
it has the temperature of the gas and the gas constant

R' = R, (8.135)
1 + m,K
Equation (8.129) will be the energy equation for the pseudo gas if it has a
specific heat given by
' Cp(J + mscsL (8.136)
cp = l + msK 2

From Eqs. (8.135) and (8.136) the specific heat at constant volume is
I
e =
Cpa + msCsL
- Rr, (8.137)
' 1 + m,K 2
1 + m,K
Therefore the ratio of specific heats is

'Y' = S, =
e~
(l ~ 1++ m,K
1 msK
2

Cpa
R,
+ .msCsL
)-
1
(8.138)

The two-component flow can now be treated as a pseudo-gas flow

with the above properties and a total mass flow rate W, + W,. Ali of
the standard formulas of one-dimensional gas dynamics can be applied
in the usual way,
If a nozzle is to be designed on the above basis, the procedure might
be as follows. First the value of K is chosen arbitrarily. The value of
L follows from Eq. (8.134). Ali the pseudo-gas properties can then be
determined and the area variation and velocities as a function of pressure
determined, for given upstream conditions and flow rates. The nozzle
shape, however, is not arbitrary since Eq. (8.129) enables the distance z
down the nozzle to be calculated in terms of the other parameters. Thus
a nozzle designed for specific upstream conditions, flow rates, and frac-
tional lag will have a determínate geometry. A different choice of K
will lead to a different geometry with smaller values of K leading to
shorter nozzles for a given gas exit velocity.
SUSPENS10NS OF PART!CLES IN FLUIDS m

PERTURIBATION TECHNlQUIES

A method far analyzing flows with small departures from equilibrium

can be developed by formulating equations in terms of tbc deviations
from homogeneous flow. 36 , 37
The analysis starts from Eqs. (8.105) and (8.106) which govern
the velocity and thermal lags. If a "characteristic" velocity, length,
and temperature are defined as Ve, Le, and Te, respectively, \Ve may write
d'p,v, v, d(v,/v,) v, v,
(8.139)
18µ,L, ¡;; d(z/L,) V, v,
an:d
d'p,c,v, v, d(T,/T,) T,
(8.140)
12k,L, ¡;; d(z/ L,) T,
The quantities
d2psVc
Ev = 18µ,L,
(8.141)

and
d2CsPsVc
ET= - - - (8.142)
12/c,L,

are dimensionless parameters which characterize thc deviations of the

velocity and temperature from the homogeneous equilibrium values. If
these quantities are sma11, they can be used as the basis for a perturbation
expansion. :Furthermore, their ratio

(8.143)

is usually clase to unity. The sizes of f.11 and Er can a1so be used to esti-
mate when the assumption of homogeneous equilibrium flow is reasonable.

SHOCK WAVES

Both normal and oblique shock waves occur in particle-gas mixtures.

Moreover, the shock waves can be of (at least) two types depending on
whether the velocity exceeds the homogeneous compressible wave veloc-
ity or the gas sanie velocity. The velocity of a homogeneous compressi-
ble wave is
e'~ -v'·-/R'T (8.144)
where 'Y' and R' are given by Eqs. (8.111) and (8.113). On the other
hand the velocity of sound in the gas alone is

e,~ -v'1,R,T, (8.145)

 ONE-DIMENSIONAL TWO-PHASE FLOW

We may define Mach numbers based on either wave velocity as follows:

V
M' (8.146)
e'
(8.147)

"\Veak" shock waves are characterized by the condition JJ![' > l,

JJ1a <
1 befare the wave. In this case a rather gradual transition occurs
through the wave. "Strong" shock waves can only exist if 1\ll (/ > 1
befare the wave front. The gas undergoes a normal shock of the usual
single-phase type followed by a relaxation region in which the particles
and the gas come to equilibrium. In each case the final state is governed
by homogeneous theory.
Figure 8.17 shows a sketch of a strong normal shock wave. The
condition before the wave is denoted by the subscript 1, the state after
the gas shock by 2, and the final state by 3.
Across the gas shock the usual equations apply to the gas flow,
whereas the particles are unchanged in either their velocity or tempera-
ture. In particular
M
2
, _ M,,' +
2/(-y, - 1) (8.148)
' - [2-y,/(-y, - l)](iVl,1' - 1)
p, - 1 + ~ (M,1 2 - 1) (8.149)
p, 'Y,+1

p,
_ _ 2(p,)iv 1 2
p, - 'Y, + 1
(l __1vl,,' 1_) (8.150)

_ 2-y,R,T, + 'Y, - l (8.151)

(Vg ) 2V1 - 'Y íl +l 'Y íl + l V¡
2

Gas
shock
Upstreom
Relaxation zone

M 9 >1 M/<1

P¡

Fig. B.17 Pressure and velocity variations through a "strong" two-phase

shock wave.
SUSPENSIONS OF PARTlCLES IN FLUIDS
"'
--------------------

M;>l
M/< l
r,

Fig. 8.18 Tcmperature and vclocity variations through a "weak'' two-phase

shock wave.

or

(8.152)

Exactly the same equations apply bctween states 1 and 3 as long

as the Mach nnmber is interpreted according to Eq. (8.146) and the gas
constant and specific heat ratio are taken from Eqs. (8.111) and (8.113).
Since -y, > -y' and M; > M,1, it is clear from Eqs. (8.150) and
(8.152) that the final velocity at state 3 is less than the velocity at 2 and
that the final pressure exceeds the pressure at 2. The decelcration of
both the gas and the particles is brought about by a continuous rise in
pressure across the relaxation zone. However, dueto the drag from the
faster moving particles the gas velocity ma.y increase over the first part
of the relaxation zone. A weak shock obeys the same equations between
states 1 and 3 but is without the initial discontinuity in the gas properties
(Fig. 8.18).

THE RELAXATION ZONE

In the relaxation zone of a shock wave the properties vary continuously 1 3

whereas drag and heat transfer between the gas and thc particles govern
the rate of approach to equilibrium. 38 - 40
Carrier's analysis 38 derives from taking the ratio of Eqs. (8.105) and
(8.106) to eliminate the space coordinate and obtain

dT, 2/c, T, - T,
(8.153)
dvs = 3csµg Vg - Vs

The coefficient '3-'S (k(}/ esµ(}) is close to unity in many practica! cases.
Indeed, a relation of the same form as Eq. (8.153) is approximately
l 1'
L)
216 ONE-DIMENSIONAL TWO-PHASE FLOW

valid ovcr a wide range of Reynolds numbers, dueto the interrelationship

bchveen heat and momentum transfer.
Eliminating the pressure and the gas density between Eqs. (8.99),
(8.108), and (8.101), we get, since A is constant,
W 0 dv, + W, dv, + R W O O
,J,_ T, ~ O (8.154)
dz dz dz v0
It follows by integration that
(8.155)

Equation (8.155) enables T, to be expressed in terms of v, and v,.

Use of the result in Eq. (8.107) gives T, in terms of these velocities.
Finally, substitution in Eq. (8.153) gives a differential equation, relating
Vs and Vg which can be solved numerically.
1
Using eithcr velocity as the
independent variable, Eq. (8.105) is then used to find z, and ali the other
parameters can then be determined as functions of the position. For

5 l. l

4 l.0
t t8
~ 3 "' 0.9
' '-.

'S' 2 -s 0.8

0.7

o 0.6
o 0.8 l. 6 2.4 o 0.8 l.6 2.4
z, in. z, In.

lol lb)

l. 2 1 1 1

-
l.O ~

~
J.--

V
~ -

0.4 ' ' '

o 0.8 1.6 2.4
z, in.
le)

Fig. 8.19 Variation of gas and particle velocities and temperatures behind a strong
shock. (a) Velocity versus z. (b) Temperature versus z. (e) Pressure versus z.
Subscripts 1, 2, and 3 denote particles with radii 0.5, 1.5, and 2.5 µ. 1vlg 1 = 2.
(Kriebel. 39 )
SUSPENSIONS OF PART!CLES IN FLUIDS 217

a l. 12
"
'

1.04

1.00 l--_ __J_ __ L_ __L__ _L __ __J_ _J __ __¡__ __J

o 2 4 6 8 10 12 14 16
z, in.
la1

, 0.92
"
"-
';;: 0.88 1-------1--+-----+-------+-----------I

0.84

lb}

Fig. 8.20 Velocity and prcssure variations through a <(weak" shock.

Ma1 = 0.91, lvl1 = 1.1. (a) Velocity versus z. (b) Pressure versus z.
Subscripts 1, 2, and3 denote particlcradii 0.5, 1.5, and 2.fi µ. (Kriebel. 39 )

example, Kriebel 39 gave a detailed description of his method of integration

and solved several examples numericaIIy. He also considered three
groups of particle sizes such that 70 percent had the mean diametcr,
12 percent were ~;3 of the mean diameter, and the remaining 15 pcrcent
were % of the mean diameter. These groups are denoted by subscripts
1, 2, and 3 in Figs. 8.19 and 8.20, which show cxamples of strong and
218 ONE-DIMENS10NAL TWO-PHASE FLOW

,veak shock relaxation zones. The particles were "typical of rocket

propellants" with a mean diamcter of 3 µ. Note that the relaxation
zones are several inches thick. li'or particles 1O times as large the overall
shock thickness could be severa! feet.

OIBUQUIE SHOCKS

The initial and final states following an oblique shock can be derived in
the usual way by resolving thc motion along and perpendicular to the
shock front. In the relaxation zone both the gas and particle streamlines
are curved. In flow past a wedge, for example, there is no room for
equilibrium to be cstablished near the tip and the shock angle is deter-
mined by the gas flow alone. Farther away the equilibrium sbock angle
is approached. Just how the transition between these regions is achieved
remains rather mysterious.

TWO- AND THREIE-DIMENSIONAL IEFFECTS

The conscrvation equations can be rewritten in vectorial form to describe

multidimensional two-phase flow. Solution of these equations is possi-
ble numerically using the "characteristics" which provide a mesh for the
numerical calculations. 41 , 42
Qualitatively the particle thermal and velocity lags behave much
the same as in the arre-dimensional flow case, except that now the relative
velocity must be interpreted vectorially. If the gas flow is turned, for
example, the particles will tend to keep going straight until enough side-
ways relative velocity has been developed to produce a drag tending to
bring them into line with the gas flow. Thus, in a converging-diverging
nozzle the particles tend to leave the wall and move toward the axis,

Increasing
Center line particle diameter

Fig. 8.21 Particle trajectories in a convergent-divergent nozzle,

showing two-dimensional effects on limiting particle streamlines
(for particles starting near the wall).
SUSPENSIONS OF PARTiCLES IN FLUIDS 219

with the effect being more pronounced far larger particles (Fig. 8.21).
Similar effects occur in two-dimensional shock relaxation, impinging jet
flows, flows around bends and so on.
1

OTHER IE.FFECTS

A vast number of additional effects occur if the assumptions outlined on

page 207 are not ali valid. Sorne useful refercnces to these tapies are
given in the bibliography. 43

8.10 ADDITIONAL FORCE COMPONENTS IN RAPIDLY ACCElERATING HOWS

APPARENT MASS

When a particle is accelerated relative to a surrounding fluid it sets up a

two-dimensional flow around it which possesses kinetic energy. There-
fare work must be supplied to move the particle in addition to that which
is required to accelerate it alone. This extra energy requirement shows
up as an additional force on the particle. Thus, the acceleration a of a
sphere in a stationary inviscid fluid of large extent requires a force

(8.156)

The sphere behaves as if it possessed an additional uapparent mass"

equal to one-half of tbe fluid which it displaces.
Bodies of other shapes have different values of the apparent mass.
In general we may say that the mass is increased in the ratio 1 + Cp¡/p,
where C ~ 7f for a sphere and is% far an ellipsoid oriented with the flow
with axes in the ratio l: 2 and 0.045 far axes in the ratio 1: 6, for exam-
ple." One would also expect the presence of neighboring particles to
influence the value of C which will therefore depend on the concentration
a. For suspensions of spheres, Zuber 52 suggests that the apparent mass
should be multiplied by the factor (1 +
2a)/(l - a).
The effect of apparent mass is that it introduces an additional com-
ponent in the forces f that were introduced in Chap. 3. This force is
proportional to the relative acceleration as seen by a coordinate system
moving with the particles. Therefore

f, = -f¡ '
1 _, = [ª
-Cp1 ai (v, - v1) + v, aza (v, - v1) ] (8.157)

This force acts to reduce the velocity lag. The equation of motion of a
particle rnoving upward in steady flow, for example, is, from Eqs. (3.45),
220 ONE-DIMENSIONAL TWO-PHASE FLOW

(8.15), (8.36), and (8.157),

dv, d¡, 3, 3 (Cv),p¡(v, - v,)lv, - v1I
v,p, dz = - gp, - dz - 4d

- ep¡Vs d(v, dz- v¡)

(8.158)

Under conditions of rapid acceleration, the final term will be significant

compared with the term on the left-hand side if

Cp¡ (1 - dv¡) ~ 0.1 (8.159)

Ps dvs
For particles suspended in gases the density ratio usually makes the
effect ncgligible unless V is passing through a maximum or minimum.
8

However 1 for liquid suspensions flowing through nozzles the effect sho11ld
not be ignored.
Suppose that the inertia and pressure-drop terms domínate Eq.
(8.158) entirely. Thcn we have
dv, __ d¡, _ C d(v, - v¡)
VsPs dz - dz p¡Vs dz (8.160)

Using Eq. (8.157) the equivalent equation lor the fluid is lound to be
dv¡ d¡, + 1 - , C d(v, - v¡)
(8.161)
V¡p¡ dz = - dz - , - p¡Vs dz

Subtracting Eq. (8.161) lrom Eq. (8.160) to removc the pressure gradient
we get
dv, dv C d(v, - v1)
VsPs dz - V¡p¡ dZ1 = - 7 p¡Vs dz (8.162)

Collecting the terms in dv./dz and dv¡/dz and dividing by p 1v,, we obtain

!" + _(;') dv, =

( Pi E dz
('l!I + _(;') dv¡dz
Vs E
(8,163)

whence
dv, = (v¡/v, + C/,) (8.164)
dv¡ p,fp1 + Cj,
If C and E are constants, as in an incompressible flow with no dis-
tortion of the particles, Eq. (8.164) has the lorm

(8,165)

Letting
V,
n=- (8.166)
v,
SUSPENS!ONS OF PARTICLES IN FLUIDS m

Eq. (8.165) is transformed to

1 dv¡ -n
V¡ dn = n 2 - bn - a (8.167)

If the roots of the denominator of the right-hand side in Eq. (8.167) are
n 1 and -n 2, expansion in partial fractions and integration eventually
yields
V¡~ (v¡)o(n - n1)-n,/(n,+n,l(n + n,)-n,/(n,+n,J (8.168)

where (v 1 )o is a constant of integration.

Now, as v1 is increased during acceleration, Eq. (8.168) shows that
n will be driven toward the value n1, i.e.J

Equation (8.169) gives the maximum value of the velocity ratio during
rapidly accelerating flows.

Example 8.4 What is the limiting velocity ratio during the accelerating nozzle flow
of a dilute suspension of spherical gas bubblcs of 10\v density?
Solution The problem statement implics that

C=H E = 1

Therefore, comparing Eqs. (8.164) and (8.165),

a= 2 b ~ ¡
Using these values in Eq. (8.169) we obtain

which is the lirniting slip ratio vp/v¡. The drag term in Eq. (8.158) will act to
keep the slip ratio below this lirniting valuc.

THIE BASSET FORCE

When a particle is accelerated relative to a fluid it establishes not only

a potential flow field but also a viscous flow field. Since viscous effects,
such as boundary layer growth, are governed by diffusion equations, the
instantaneous flow field is a function of the entire previous history of the
particle motion and is rather awkward to calculate. For laminar flow 1
Basset 45 obtained the result

f ~ - ~~ /,' (t - t')-¡, d(v, - V¡) dt' (8.170)

º ..¡;;: d ,, dt'

where t' is a dummy variable which allows integration over the history of
the particle between t 0 and the present time t. In particular, if Vs - v1
varíes linearly with time from t 0 to t, it is found (cf. Prob. 8.34) that the
222 ONE-0\MENSIONAL TWO-PHASE FLOW

ratio of thc Basset force to thc steady flow drag force is d/ y ,rv1t. When
the particle is part of a suspension 1 Zuber 52 suggests that this force should
be increased by the factor (1 - a)- 2· 5, which probably overcstimatcs
the effect of neighboring particles.

8.11 FRICTION CllAl!ACTERISTICS Of PAl!TICLE SUSPENSIONS

The friction characteristics of a fluid are governed by the way in which
the shear stresses are related to an imposed rate of shearing strain. In
the case of a newtonian fluid a simple linear relationship is valid in
larrúnar parallel flow and this relationship defines the viscosity. Thus

(8.171)

JVIany suspensions of particles do not have a linear relationship

bet1veen shear stress and rate of strain and have to be treated by rheologi-
cal methods which are applicable to nonnewtonian fluids. For example,
a typical shear diagram far a thorium oxide slurry is shown in Fig. 8.22.
The larrúnar flow velocity profile can then be predicted by using a lmown
shear stress distribution to deduce the velocity gradient and integrating
across the duct, following the methods described in Chap. 5.
For low values of particle volumetric concentration, suspensions
of spherical particles are approximately newtonian, and the ratio between
the viscosity of the mixture and the viscosity of the fluid can be expressed
as a function of a. For example, Einstein's equation 46 (far a < 0.05) is

µ = µ¡(l + 2.5a) (8.172)

which is equal to the lirrút of an equation proposed by Roscoe 83 and

Brinkman, 86
(8,173)

or, equivalently,
(8.174)

A rival equation whicb has a different coeflicient was developed by

Happel 49 as follows:

µ = µ 1 (1 + 5.5a) (8.175)

and there is experimental evidence that the Einstein equation does tend
to underestimate the effective viscosity.
Attempts have been madc to obtain analogous results for particles
which are not spherical 50 and to perform expansions in higher powers
 SUSPENSIONS OF PARTICLES 1N FLUIDS m

O>
e
·e
o 4
~ • D
~
o
2 10 2 l------+-----;1-------+--------j
2 ~
{ 4

{
10' f-------+---+-t------t-------1
8
6

Fig. 8.22 Shear diagram for thorium oxide suspen-

s10n. (Thomas. 47)

of a. There is a great variation in the value of the coefficients which

are quoted by different authors.
lVIost nonnewtonian suspensions have a limiting viscosity r¡ at high
rates of shear, which can be simply related to the viscosity of the suspend-
ing fluid and the particle concentration.
The limiting viscosity of nonnewtonian suspensions is often cor-
related 51 by the equation

(8.176)

An example of this dependence is shown in Fig. 8.23. The coeffi-

cient k is determined by the particle characteristics and is generally
larger far smaller particles which flocculate and display an increased
viscosity as a result of interparticle forces. Thomas 51 has found, for
example, that k for suspensions in water can be expressed quite well by
224 ONE-DIMENSIONAL TWO-PHASE FLOW

20

10
8

~
6
µ¡
4 Particle
Material diarneter
d, microns
0.40
2 l----/a0-7'F---:,,L'-+----+--j "'Titonium dioxide
□ Graphite 2.35
o Thorium oxide 0.74
• Thorium oxide 1.35
t:,, Kaolin 2.85
1
o 0.05 0.10 0.15 0.20 0.25 0.30
Volume fraction solids, a

fig. 8.23 Etfect of volume fraction solids on limiting viscosity at high ra.tes
of shear. (Thomas. 51 )

the equation

k = 2.5 + ~<!> (8.177)

where d is the particle diameter in microns and 4> is a shape factor which
is equal to unity for spherical particles. For large particles and low
concentrations, k becomes equal to 2.5 and Eqs. (8.172), (8.173), and
(8.176) are approximately identical.
Sorne other models for the viscosiÍ,y of suspensions are discussed
by Zuber 52 anda comprehensive literature survey is given by Rutgers. 53 • 54

LA.MINAR FLOW

For laminar flow of a newtonian fluid in a round pipe the wall shear stress
is related to the viscosity by the equation
8V (8.178)
Tw = µD
and the pressure gradient is
dp) _ 32µV
- (dz ,-152 (8.179)

The equivalent relationships for a nonnewtonian fluid are dependent

upon its particular rheological properties. However, it is possible to
 SUSPENSIONS OF PARTICLES IN FLUIDS 225

define an ejfective viscosity as

Tw
µ, = 8V/D (8.180)

Equation (8.171) is then replaced by a "pseudoshear" relationship which

expresses the wall shear stress in terms of the rate of strain which would
exist at the wall for the- same overall flow rate in a newtonian fluid.
Sorne typical pseudoshear results are shown in Fig. 8.24. When one
changes the pipe diameter or velocity, the relationship between wall shear
stress and effective rate of strain is unique for a given fluid-particle com-
bination as long as the flow is laminar. Therefore the pseudoshear dia-
gram can be used to predict laminar flow pressure drop.
Another way to proceed is to use a friction factor versus Reynolds
number relationship in terms of the effective viscosity. The friction fac-

103
1
- 1 1 ' '

6 -

4-

,~
10 2
~

:a1~ 6~
e
·2 4- -
t;
L
D
ID
~
~
3 2
~
o
~
~

10'
s 1- co

f
Tube Suspension
6 ~ dio., in, properties -

,~ Th02 D 0.124 a""' 0.10 -

• t,, 0.318

j
D l.030 77=5.7cp
Fig. 8.24 Pseudoshear diagram
for concentrated suspensions
,~ Th02 • 0.124 a= 0.16
-

showing agreement of laminar .&..0.318

and turbulent data, respec- 10º
• 1, l.030 77
1
= 10.1 cp
' ' '
tively, as tube diameter was 0.6 2 4 6 8 10 20
varied. (Thomas. 48 ) Wall shear stress, Tw
 ONE-DIMENSIONAL TWO-PHASE FLOW
'"
tor is defined as

e¡ = ½2Pm V'
Tw
(8.181)

Therefore, from Eq. (8.180),

16
C¡=~ (8.182)
Rem
where R.em is defined in terms of the effective viscosity
Rem = PmVD (8.183)
µ,
In using this technique a pseudoshear diagram is still necessary in arder
to calcula te /J,e• In cases where a quasi-newtonian approximation is valid,
µ, and µ are identical.
0 Transition to turbulence occurs in round pipes
when Rem is greater than ahout 2500.
In principle 1 if the rheological relationships between stress and
strain rate can be established, an accurate prediction of the pressure drop
can be madc as a function of the flow rate in laminar flow. The shear
profile is used to deduce the velocity gradient profile and a doublc inte-
gration yields the flow rate. However 1 in practice, nonnewtonian char-
acteristics display a whimsical and bewildering variety which defies any
rational treatment from first principies. 50 One is always reduced to an
attempt to fit experimental data with a suitablc mathematical model.
Since the engineering use of any theory is usually restricted to a limited
range of shear ratesJ almost any reasonable curve fit will do.
The commonest models are:

(a) Power law:

T = k(- dv)" dr
(8.184)

(b) Bingham: 56

7 = Tv +µ ( - !,") for T > Ty

(8.185)
dv = O for T < Ty
dr

r
(e) Casson: 56

7l' = 7y)i + [µ (- ~~) (8.186)

(d) Ree and Eyring: 57

, = C(- dv}
dr,
+ _1, sinh-, ( - _l, dv)
B A dr
(8.187)
SUSPENSIONS OF PARTICLES IN FLUIDS m

1.5x10 3 10'
DVp
Reynolds number, 7J

Tube Suspension
día., in. properties
Th02 o 0.124 a - 0.10
6 0.318
o 1.030 17 = 5. 7 cp
Th02 • 0.124 a- 0.16
• 0.31 8
• 1.030 r¡=lü.lcp

fig. 8.25 Friction factor plot for concentrated suspen-

sions showing agrcement of turbulent data as tube
diameter was varied. (Thomas. 48 )
m ONE-DIMENSIONAL TWO-PHASE FLOW

The Bingham model is appropriate for many plastic fluids. Cas-

son's equation works well far blood, 58 ]nks, 56 and coal slurries; 59 it even
works better on Thomas' thoriurn oxide suspensions than the models
which Thomas originally tested. 47 The Ree-Eyring model often corre-
lates data well (because it contains three adjustable constants), but is
hard to deal with for analytical purposes.
An equation with three arbitrary constants which reduces to models
a, b, and e as special cases is

Tm=Tm+k
u ( - -dv)n
dr
(8.188)

and this equation will fit most data satisfactorily.

TURBULENT HOW

Because o! the high local rates of strain in turbulent flow, the effective
viscosity <loes not give an accurate description of the force distribution in
the flow field. A better approximation is usually obtained by using the
limiting viscosity at high rates of shear in the Reynolds number. Thus

(8.189)

A friction factor-Reynolds number plot on this basis is shown in Fig. 8.25.

Since the effective viscosity in laminar flow can be many times the limit-
ing viscosity, the laminar flow lines are displaced on such a graph.
An interesting and important effect with suspensions of fibers or long
chain molecules is the reduction of drag in turbulent flow. 48 , 61- 64 Appar-
ently the turbulence leve! is reduced by the damping action of the fibers
on the eddy pattern. In practice it is sometimes economical to add fibers
or polymers to fluids in order to reduce the pressure drop in pipelines.

PNEUMATIC TRANSPORT

The flow of particles of any significant size above a few microns which
are suspended in gases is much less amenable to analysis than the flow of
a liquid suspension. The particles "rattle around" in the pipe, there are
many different flow regimes, and factors such as the coefficient of restitu-
tion for the particles are important. Numerous empirical correlations
have been developed but few have any claim to general validity.

8.12 NONUNIFORM PARTICLE DISTRIBUTION

In previous sections of this chapter it was assumed that the particles were
uniformly distributed over the flow field. Certain cases in which this
SUSPENSIONS OF PARTICLES IN FLUIDS
"'
assumption is no longer valid will now be discussed briefly. For con-
venience it is possible to define three general classes of nonuniform flows.

l. Stratijied fiows in which the particles concentrate in layers or annuli

causing nonuniform distributions across the pipe.
2. Periodic fiows in which the concentration varjes along the pipe.
3. Aggregative fiows in which the particles or the fluid tend to form many
discrete areas of high concentration throughout thc flow.

Often there exists a close analogy between these phenomena and

similar phenomena which occur in gas-liquid flow.

STRATIFICATION

Gravilalional eflects One of the simplest ways in which stratification

can occur is in horizontal flow whcre gravitational effects cause the heavier
component to concentra te toward the bottom of the duct. The amount
of stratification which occurs is governed by the balance between the
buoyancy forces on the particle and the forces which are caused by its
motion relative to the fluid. This balance is conveniently represented
by taking the ratio of the terminal settling velocity v. to the wall shear
velocity u* defined by Eq. (7.27). The concentration distribution is then
often represented by expressing a/(a) as a function of vJu*.
According to Thomas, 6 ~ the value v /u* = 0.2 gives a convenient
00

measure of the boundary between the regimes in which the solids are con-
centrated at the bottom of the pipe or flow predominantly in suspcnsion.
Corresponding values from the Reynolds flux theory (Prob. 7.16) are
between 0.1 and 0.15. Thcsc simple expressions do not, howcver, give
an adequate picturc of the interaction between the particles and the
velocity profile. Very small particles líe deep in the boundary !ayer and
are exposed to velocities much less than u *i whereas very Iarge particles
experience lift and drag forces which are scaled by the mean stream
velocity. In general 1 particles which are small enough to lie inside the
boundary layer at the condition of the minimum transport velocity obey
the Stokes law (Re~ < 1) and the critica! value of u* for dilute suspcn-
sions (u;) is given by 65

V:
Uo
= 0.01 (d *
UoP! )' . 71
µ¡
(8.190)

For larger particles (Re~ > 1) a correlation is

":
u0
~ 4.9 (du;';p¡)º·' (_!,_)º·'
µ¡ D
(p, - Pi)º·"
p¡
(8.191)

which is remarkable for the large effect of the diameter and density ratios.
 ONE-DIMENSIONAL TWO-PHASE FLOW

The above equations refer to the condition of infinite dilution, or

a: = O. When a significant concentration of particles is present, the
wall shear velocity far minimum transport
65
u:
can be faund from the
correlation

(8.192)

The effect of the density ratio in Eq. (8.191) is to make vju.¿ larger
far ga.s-solids systems. This means that a small concenÚation of par-
ticles has a large effect in Eq. (8.192). In addition, the mínimum trans,
port velocity of the fluid becomes comparable with the terminal velocity
of the particles, instead of being many times larger as in the case of liquid-
particle systems. This is only possible if the particles "rattlé around"
in the pipe and acquire velocities comparable with the free stream.
Once they settle out, they are difficult to dislodge, tend to fill up the pipe,
and increase the pressure drop by an order of magnitude. This "point
of saltation" is an important regime boundary in pneumatic conveying. 66
Similar techniques are applicable to problems of erosion and mud
and sand transportation in alluvial rivers. 67 , 68

Symmetric:al radial conc:ent.ration variations Even when gravitational

effects are absent or negligible it is possible for quite large variation.s to
occur in particle concentration across a duct. For example, in viscous
Poiseuille flow in a small diameter tube the particles tend to concentrate
at about 0.6 of the tube radius from the axis and there is a fluid core that
is relatively free of particles. 69

Charrneling or spouting in fluiclized bi:!ds Nonuniform effects can be

produced in fluidized beds by introducing the fluid in a suitable way.
Far example, if the fluid is introduced as a high-velocity jet in the center
of a vertical column, a kind of annular countercurrent fl.ow can be set up
in which a dilute fluid-solids mixture flows up the middle while a denser
concentration of solids eirculates down near the wall. The behavior of
such a system is described by Leva. 15

PERIODICnows
Siugging Large axial variations in particle concentration can occur in a
duct as a result of the growth of the instabilities which were discussed
in Sec. 8.8 and also by the tendency of turbulence and other three-
dimensional instabilities to break up the simple one-dimensional pattern.
Since many wavelengths are unstable, no simple regular pattern is
observed in slugging. Individual slugs form, grow, collapse, and over-
take each other continuously (Fig. 8.26). If the slugs develop a two-
SUSPENSIONS OF PARTICLES lN FLUIDS m

• +
'
:' Particles
''1,

1 raining
down : '
:__,,,
1

~: .·
t :1

l---
t
movmg
Slu_gs
upword

---..
r• .

+
''
¡'

t'

Fig. 8.26 Slugging in a fluidized bed.

spheres in a 1-in.-diam tube.
Glass
(a}
1 lb}

dimensional bubble shape, they can be treated by the methods to be

described in Chap. 10. 70

Wave lormation in slra!ified flow In horizontal pipes in which dense

particles lie on the bottom while fluid flows over them, it is possible to
observe the periodic formation of waves or dunes which travel slowly in
the direction of fluid motion (Fig. 8.27). These waves appear to be the
result of a Helmholtz instability (due to the relative motion) and have
been correlated by Thomas 71 using the equation
v,
2
Pr
g),. p, - p¡
~ _!_ [_1!r2_
2.- 4gH H
(P.);s (P.)%]
d
o.m (8.193)

far islands (particle groups) moving with velocity vr, wavelength J\,
height H, in a tube of diameter D, and composed of particles of density
p, and diameter d, beneath a fluid with velocity v1 and density p¡. Simi-
lar phenomena occur with alluvial rivers, sand dunes, and snowdrifts.

AG.GIREGATIVE FLOWS

Flocculation Very small particles (about 100 µ or less in diameter) or

particles with irregular dendritic shapes tend to form clusters or fiocs
which retain their identity as they flow in the suspending fluid. The
flocs are held together by interparticle forces. Because of the open
m ONE-DIMENSIONAL TWO-PHASE FLOW

IFig. 8.27 Periodic phenomena observed with spher-

ical particles in horizontal pipes. (Thomas. 11 )

spatial structure of a floc, the suspending fluid can flow through it as well
as around it. Flocculation is one cause of nonnewtonian flow charac-
teristics of suspensions.
For flocculated suspensions the approximate value of nin Eq. (8.2)
is often very high (e.g., a value of 20 to 30 in Ref. 7). Furthermore, n
is not constant over the entire range of concentration s]nce the floc size
is not independent of concentration.
One effect of flocculation is to termínate settling at low values of ,.
The sediment contains a spongy mass of particles held together by inter-
particle forces in a loase open. structure. Further compaction is only
possible by the breaking of the floc bonds and the squeezing out of the
liquid through the surrounding particles.

Bubbling Usually, when the suspending fluid is a gas and the particles
are fairly large or dense, it is possible to observe the inverse of floc for-
mation-namely, the formation of bubbles of fluid containing no particles.
 SUSPENSIONS OF PARTlCLES IN FLUIDS
"'
The actual way in which these bubbles arise is a matter of debate. They
may be the eventual three-dimensional result o! the one--dimensional
instability discussed on page 206.
A great deal of study has been devoted to the fascinating phenome-
non of bubbling in large gas-fluidized beds. 72- 18 The shape of such a
bubble is shown in Fig. 8.28. The bubbles rise almost exactly like gas
bubbles in a fluid with negligible viscosity and surface tension. In fact
the rise velocity of a bubble in a bed which is just fluidized by gas of low
density compared with the particle density is given by an equation
similar to the equation of Davies and Taylor 79 for gas bubbles (the con-
stant is slightly different), namely, 80
Vb = o. 71l 2'lJb7É (8.194)
If the fluid density is significant, this equation is modified to the form

v, - O. 7lg¼'llé' [ l - 'º ]¼ (8.195)

p,/(p, - p¡) - 'º
The hubble surface is not impermeable to fluid and, in fact, there
·,jsa continua! circulation, relative to the hubble, up through the bubble
~se and round through the bubble wake. This circulation "holds up
the roof" and prevents the bubble from collapsing.
Although a gas-fluidized bed usually bubhles, while a liquid-fluidized

fig. 8.28 Photograph of a two-dimen-

sional bubble in a fluidized bed.
(Rowe, Partridge, and Lyall. 18 ) 550µ Ballotini L....,___) 2 em
m ONE-DIMENSIONAL TWO-PHASE FLOW

bed does not, there is a continuous transition from one form of behavior
to the other. 16 , 81 Buhbling can be obtained with lead shot in water, and,
on the other hand, gases under pressure will fluidize sufliciently light
spheres quite uniformly,
A simple, but approximate, criterion for bubbling has been given
by Wilhelm and K wauk 30 in terms of the particle Fraude number defined
by Eq, (8,98), Uniform fluidization is predicted far a Fraude number
less than unity and bubbling at higher Fraude numbers, Alternatively,
Jackson 77 has suggested that the rate of growth of one-dimensional dis-
turbances (Table 8,3) can be used as a measure of the tendency toward
bubhle formation, Similar analyses have also been carried out by
Pigford 82 and Smith, 27
Harrison, et al, 81 approach the houndary from the other side by
considering the conditions under which a bubhle can be destroyed.
They reason that, if the circulation of fluid through the hottom of the
hubble is rapid enough, it will entrain particles and fil! the bubble from
below. The criterion far this to happen is, approximately,
(8.196)
Since the bubble velocity is given by Eq. (8.194), Eq, (8.196) can
be rearranged to give for unstable hubhles

(g:)¼ < 0.71 [ps/(p,1--p:; - ,,r Cº:)¼

The Fraude numher is therefore the majar factor determining the
(8.197)

ratio of the diameter of the largest stable bubble to the particle diameter.

8.13 PERCOlATION THEORY

The flow of a fluid through a three-dimensional porous medium occurs
in :filtration equipment, leaching, settling tanks, and in many civil engi-
neering applications such as the flow of ground water to wellsJ seepage
through dams, and soil consolidation.
As usual, the formulation of the theory comprises the two continuity
equations and equations of motion. In the simplest case, flow is steady,
the fluid is incompressible, the porous matrix is homogeneous, and inertia
terms are negligible. The continuity equation for the fluid is
V· j¡ = O (8.198)
which j 1 is now a vector.
Equation (3.45) in vector form becomes
-Vp - p¡gk - f¡ = O (8.199)
k is the unit vector in the upward direction.
SUSPENSIONS OF PARTICLES IN FLUIDS 235

The value of f 1 will be determined by the value of j1 and the fluid

and particle properties (since the matrix is stationary). In the case of
laminar flow we have, from Example 8.3, far js = O,
180¡,¡ .
f1 = - d'e' (1 - ,)lJ (8.200)

The Darcy permeability factor is defined as

K = gp¡d2E2
180¡,¡(l - ,) (8,201)

If we also define the nonhydrostatic pressure p' by the equation

p' =p + gz (8.202)
the use of Eqs. (8.200) and (8.201) in Eq. (8.202) finally gives
Vp' 1 .
gp¡ = - K 11 (8.203)

Combining this result with Eq. (8.198) we obtain

(8.204)
Therefore the pressure obeys Laplace's equation, wbereas the fluid flux
everywhere is proportional to the negative pressure gradient. Ali of
the standard methods of potential flow theory can then be applied to
solve specific problems. A good reference for tbe European literature
is the book by J aeger, 84 which also gives examples of many approximate
metbods.
For turbulent flow the best approach is usually a numerical one.
An approximate solution for unsteady flow was obtained by
Terzaghi 25 who modified Eq. (8.204) to the form

(8.205)

where e was defined as the coefficient of consolidation. This theory is

still repeated in most soil mechanics textbooks although, as pointed out
in discussing Eq. (8.97), it is not entirely correct.

l'l!OBLEMS
IU.. Compare the terminal velocities of "spent" duck shot in air and water if d = 0.08
in. Consider both lead shot (p8 = 700 lb/ft 3 ) and plastic shot (p 8 = 110 lb/ft3).
8.2. What are the values of n in Eq. (8.2) for the particle-fluid combinations in
Prob. 8.1?
8.3. Rather than differentiating Eq. (8.31) use Eqs. (8.22) and (8.30) to find the best
value of n for use in Richardson and Zaki's correlatíon as a functíon of Reynolds
number,
236 ONE-DIMENSIONAL TWO-PHASE FLOW

8.4. Derive a general dimensionless drift-fl.ux-versus-concentration relationship in

terms of Nh e, and the drift-flux Reynolds number, Re¡s = p¡j¡,d/µ¡.
8.5. Wen and Yu 9 derived a general correlation for the minimum fl.uidization velocity
as follows:
(Re,)m, ~ [(33.7)' + 0.0408N1 '1" - 33.7
Compare this result with the theories in the text.
8.6. Frantz 85 gives an equation for the fluid mass flux at minimum fluidization, as
follows:
2
Gmi = 4.45 X 10 5 d p¡(p, - p¡)
µ¡

The various units are G, in pounds per hour per square foot; d, in feet; p, in pounds per
cubic foot; µ, in pounds per hour per foot. Compare with the equations in the text.
8.7. Calculate the' minimum-fl.uidization velocities for the following sizes of glass
beads (p, = 2.5 g/cm 3 ) in air or water and compare with Rowe's 8 measured values.
Assume atmospheric conditions of 20ºC and 1 atm.

M easured (j¡o)rnr,
Fluid d, cm cm/sec

Air 0.0083 0.823

Air 0.0273 7.01
Air 0.065 30 5
Water 0.06,5 o 597
Water 1.2 10. 12

8.8. As air rises through a fluidized bed it expands. Show how a tapered bed can he
designed so that the particles are just fluidized at all depths. Design such a bed for
"stucco" with diameter 0.6 mm and density 2.6 g/cm 3 if the top is to be open to the
atmosphere with an area 2 ft 2 and the depth is 3 ft.
8.9. Air at 70ºF fl.ows through a constant cross-section bed of copper shot with diameter
0.065 cm. The air velocity leaving the top of the bed is measured to be 1 m/sec at
14.7 psi. Estimate the depth to which the bed is fluidized.
8.10. Seeds }'4 in. in diameter are to be transported ·pneumatically at 50 psi and l00ºF
in a vertical duct 2 ft in diameter. It is determined cmpirically that the volume
fraction of seeds in the duct must not exceed 0.05 if flow instabilities are to be avoided.
Estimate the air velocity required to transport 100 tons of seeds per hour. One
cubic foot of packed seeds weighs 50 lb.
8.11. A cylindrical fi.lter in a compressed air line consists of sintered spheres each with a
diameter 0.02 in. and having a void fraction of 0.3. It is }i in. thick, 6 in. in diameter,
and 8 in. long. Wbat is the pressure drop for an air flo-1,v of 40 scfm at 100 psi, 90ºF
inlet conditions? What is the variation in this pressure drop if the manufacturing
toleran ce in void fraction is ± 10 % ?
8.12. Compare the predictions of Eqs. (8.48) and (8.35) for hoth low and high Reynolds
numbers over the range of void fraction 0.3 < e < 0.6. lf the methods of Chap. 4 are
used to make predictions in this range, how sensitive are the results to the particular
equation which is chosen?
SUSPENSIONS OF PARTICLES IN FLUIDS m

t
Const □ nt
salid
flux,;~

Fluid flux, jf ~

Fig. 8.29 Particle concentration characteristics for a fluid-

ized system.

8.13. Ore with density Ps = 140 lb/ft3 and mean particle diameter 0.1 cm is washed
by countercurrent flow of an acid [p¡ = 65 lb/ft.3, µ 1 = 2 lb/(hr)(ft)] in a vessel 6 ft
in diameter. Evaluate the limiting relationship between the flow rates which is set
by flooding.
8.14. Show that the void fraction of a fluidized system can be plotted versus the liquid
flow rate for various upward or downward solids flow ratcs as shown in Fig. 8.29.
Identify the direction of solids flow, the points corresponding to flooding, and the
terminal particle velocity on the sketch.
8.15. Combine the results of Prob. 8.14 with packed-bed prcssure-drop theory and
homogeneous flow estimates for wall friction to generate the pressure-gradient-versus-
:flow-rate curves shown in Fig. 8.30. Identify the various features of the curves.
8.16. Thoria with a density of 9.5 g/cm 3 is mixed into a slurry containing 500 g/liter
with a mean sizc of 10 µ in water at 25ºC. A liter of this mixture is put into a vertical
cylinder 5 cm in diameter and allowed to settle. Determine the subsequent behavior.

Fig. 8.30 Pressure gradient characteristics for a fluidized system.

 ONE-DIMENS!ONAL TWO-PHASE FLOW

8.17. A :flocculated suspension is found to obey the equation

Ús = 0.0la(l - a) 10 fps
and settling is complete at a 00 = 0.5. Predict the batch sedimentation behavior of
this suspension for a 0 = 0.01, 0.05, 0.1, 0.lfi, and 0.2.
8.18. Derive Eq. (8.71).
8.19. A fluidized bed containing catalyst beads (d = 0.4 cm, Ps = 1.6 g/cm 3) is
operated steadily in a 10-cm-diam tube with 1: = 0.8 using a flow of water at 60ºF.
If the flow of water is suddenly cut in halfi describe the subsequent behavior.
8.20. Integrate Eq. (8.82) if 1 - t is not small. If N = 0.25, what errors are intro-
duced by assuming (1 - <E) to be small for the following values of 1 - 1;: 0.002, 0.01,
0.05, 0.1?
8.21. A mat of fi.bers is formed on a filter, and the pressure drop across the mat is 1 psi.
If the stress-strain relationship for the fi.bers is p. = 8000 (1 - <E) 4 psi, what is the
mean volumetric density of fibers in the mat?
8.22. Show that, if there is no particle contact, Eq. (8.96) represents continuity waves.
If the Carman-Kozeny equation appliesi show that the continuity wave velocity is
three times the liquid velocity.
8.23. If E is constant in Eq. (8.96), what conditions are necessary for Eq. (8.97) to be
reasonably valid?
8.24. Deduce the results shown in Table 8.3 assuming that fva = O, n = 3,
Ífs o: (v¡ - v,)1. 5, and 1: = 0.4.
8.25. Design a sandblast nozzle to provide a flow of 1 lb/min of sand (d = 0.02 in.,
p. = 150 lb/ft 3 ) with a velocity of 400 fps if a compressed air supply is available with
Pou = 100 psi, Toª = 70ºF.
8.26. One of the sand particles mentioned in Prob. 8.25 is dropped into an air stream
with velocity 1000 fps at 200ºF and 20 psia. How far does it go befare attaining a
velocity of 800 fps.
8.27. A plasma spray gun is to use nitrogen at 5000ºK (µ,ª ~ 10- 3 g/(cm)(sec), Pu =
3 X 10- 5 g/cm 3 ) and nickel particles (p. = 8 g/cm 3 ). If it is designed for constant
fractional velocity lag with K = H, what distance is required to accelerate 10-µ.-diam
particles to 2000 fps? What is the time spent in the nozzle by the particles in acceler-
ating from 20 to 2000 fps? Estímate their change in enthalpy in this period if ku =
10- 3 cal/(cm)(sec)(ºK)? Do they evaporate?
8.28. If the particles in Prob. 8.27 are large so that their velocity is low and the plasma
velocity is constant, show that their enthalpy change after going a distance z is

where t::.T is an average temperature difference between the particles and the plasma.
8-.29. A 1-ft-long nozzle is to operate using air at 500ºF and velocities of about 1000 fps
and is to carry a flow of aluminum particles with density 170 lb/ft3 [and specific heat
0.2 Btu/(lb)(ºF)]. What must the particle size be if "v and IET are to be nsmall" in
Eqs. (8.141) and (8.142)?
8.30. A stationary shock wave occurs in dusty air. The upstream conditions are
v - 2000 fps, m, - 1, T - 400ºF, p, - 200 lb/fl', ,, - 0.2 Btu/(lb)(ºF), and d - 10
µ.. Discuss the shock structur-e.
8.31. Solve Prob. 8.30 if v = 1200 fps.
 SUSPENSIONS OF PARTICLES IN FLUIDS
"'
8.32. What is the effect of apparent mass on dynamic waves in two-phase flow?
Estímate thc maximum possible departure of the comprcssibility wave velocity from
homogeneous theory for an air-water mixture in which a = H and C = H.
8.33. Under what conditions- can the final term in Eq. (8.158) be larger than the drag
term? Evaluate typical cases numerically.
8.34. Prove that, if the relative velocity increases linearly with time for a coordinate
system moving with a particle, the ratio of the Basset force to the steady-flow drag
force in 18,minar flow is a;y:;;;;¡,. Explain this result physically, using simple ideas oí
diffusion. For what size of particles in atmospheric air will the Basset force be signifi-
cant if t = 0.01 sec?
8.35. A coal slurry consisting of 40% coal, with density 84.5 lb/ft3 in water a.t 70ºF, is
found to obey Casson's equation with ry = 9 X 10- 2 lb/ft 2 and 71 = 5 X 10-3
lb /(ft) (sec). Derive the pseudoshear diagram for laminar flow. What is the frictional
pressure drop for fl.ow with a mean velocity of 1 fps in a pipe 1 in. in diameter and 100
ft long?
8.36. The coal slurry in Prob. 8.35 flows in a pipe for which single-phase turbulent-flow
data were correlated by the equation C1 = 0.026 Re- 0-12. What is thc pressure drop
for a velocity of 10 fps in a 1-in. pipe which is 100 ft long?
3.37. A fl.occulated suspension is transported in water at 200ºF in a chemical treatmcnt
plant. If the particle diameter is 1-µ, the density p8 is 400 lb/ft 3, and the slurry
contains 0.3 lb of solids per pound of water, what is the pressure drop for an upward
fl.ow of 2 X 10 4 lb/hr in a 100-ft length of vertical 1-in. pipe?
8.38. What is the minimum transport velocity for pneumatically conveying grain
(p = 75 lb/ft3, d = },;i in.) in a 12-in. pipe at 200 psia, 70ºF, (a) at low concentrations,
(b) ifa ~ 0.1?
8.39. An alluvial river flows over a bed consisting of mud particles for which Ps = 2 g/
cm 3 andd = 100µ. What velocity of the river will cause erosion? If the river is lined
with pebbles 1 cm in diameter, what velocity is tolerable during the spring floods?
8.40. A suggested correlation for the minimum transport velocity of large particles
in large pipes is

V= 4 [gd(p.p; p¡)T!i
Compare this with the theory in the text.
S.41. The circulation pattern of gas in a bubble rising through a fluidized bed depends
on the ratio vb/j10 . What is this ratio for lead shot, d = 0.01 in., fl.uidized by atmo-
spheric air if 'Ub = 100 in. 3 ?
8.42. How do apparent mas·S effects alter the theory of entropy generation which was
presented in Chap. 3? Show how Eq. (3.87) can be modified to account for the kinetic
energy which is associated with the three-dimensional flow around particles which are
moving relative to a fluid. This kinetic energy is not dissipated and can be recovered.
What are the correct forms of Eqs. (3.106), (3.107), and (3.108) ih this case?

REFERENCES
l. Lapidus, L., and J. C. Elgin: A. l. Ch. E. J., vol. 3, p. 63, 1957.
2. Lapidus, L., J. A. Quinn, and J. C. Elgin: A. l. Ch. E. J., vol. 7, p. 260, 1961.
3. Rowe, P. N.: Trans. Inst. Chem. Engrs., vol. 39, p. 175, 1961.
4. Schiller, L., and A. Neumann: Z. Ver. Deutsch. lng., vol. 77, p. 318, 1935.
 ONE-DIMENSIONAL TWO-PHASE FLOW

5. Heywood, H.: Symp. Interaction Flu.ids Particles, Inst. Chem. Engrs. 1 London,
pp. 1-8, 1962.
6. Richardson, J. F., and W. N. Zaki: Trans. Inst. Chem. Engrs. 1 vol. 32, pp. 35-53,
1954.
7. Wallis, G. B.: Symp. Interaction Flu.ids Particles, Inst. Chem. Engrs., London,
pp. 9-16, 1962.
8. Rowe, P. N., and G. A. Hcnwood: Trans. Inst. Chem. Engrs., vol. 39, pp. 43-54,
1961.
9. Wen, C. Y., and Y. H. Yu: Chem. Eng. Progr. Symp., ser. 62, vol. 62 1 pp. 100-111,
1966.
10. Pinchbeck, P. H., and F. Popper: Chem. Eng. Sci., vol. 1, p. 57, 1956.
ll. van Heerden, D., A. P. P. Nobel, and D. W. van Krevelen: Chem. Eng. Sci.,
vol. 1, p. 63, 1951.
12. Carman, P. C.: "Flow of Gases through Porous Media," Butterworth & Co.
(Publishers), Ltd., London, 1956.
13. Carman, P. C.: Fluid Flow through Granular Beds, Trans. Inst. Chem. Engrs.,
vol. 15, pp. 1.50-166, 1937.
14. Blake, F. E.: Trans. Am. Inst. Chem. Engrs., vol. 14, p. 415, 1922.
15. Leva, M.: "Fluidization," pp. 42-77, McGraw-Hill Book Company, 1959.
16. Simpson, H. C., and B. W. Rodger: Chem. Eng. Sci., vol. 16, p. 179, 1962.
17. Ergun 1 S.: Chem. Eng. Progr., vol. 48, p. 93, 1952.
18. Slis, P. L., Th. H. Willemse, and H. Kramers: Appl. Sci. Res., vol. AS, p. 209, 1959.
19. Shannon, P. T., and E. ~1. Tory: lnd. Eng. Chem., vol. 57, pp. 18-25, 1965.
20. lVIontcrieff, A. G.: Trans. Instn. J.11ining J.11etal, vol. 73, part 10, pp. 729-759,
1963-1964.
21. Kynch, G. J.: Trans. Faraday Soc., vol. 48, pp. 166-176, 1952.
22. Egolf, C. B., and W. L. McCabe: Trans. Am. Inst. Chem. Engrs., vol. 33, p. 630,
1937.
23. Han, S. T., and W. L. Ingmanson: TAPPI, 21st Eng. Conf., Boston, Mass.,
vol. 50, no. 4, pp. 176-180, April, 1967.
24. Davis, C. N.: Proc. Inst. Mech. Engrs., vol. El, p. 185, 1952.
25. Terzaghi, K.: "Theoretical Soil Mechanics," chap. 13, John VViley & Sons, Inc. 1
New York, 1943.
26. ·wallis, G. B.: One-dimensional Waves in Two-componcnt Flow, UKAEA Rept.
AEEW-R162, 1962.
27. Smith, J. L.: Massachusctts Institute of Technology, 1963, unpublished v,rork.
28. ]\/[urray, J.: J. Flui"d J.11ech., vol. 21, pp. 465-4931 1965.
29. Jackson, R.: Trans. Inst. Chem. Engrs., vol. 41, pp. 13-21, 1Q63.
30. Wilhelm, R. H., and M. Kwauk: Chem. Eng. Progr., vol. 44, p. 201, 1948.
31. Kliegel, J. R.: Intern. Symp. Combust., 9th, pp. 811-826, Academic Press, Inc.,
N ew York, 1963.
32. Hoglund, R. F.: Recent Advances in Gas-particle Nozzle Flows, ARS J., pp.
662-672, May, 1962.
33. Soo, S. L.: A. l. Ch. E. J., vol. 7, no. 3, pp. 384---391, 1961.
34. Altman, D., and J. M. Carter: "Combustion Proccsses," sec. E, edited by B.
Lewis, R. N. Pease, and H. S. Taylor, Princeton University Press, Princeton,
N.J., 1952.
35. Kliegel, J. R.: IAS paper no. 60-65, 1960.
36. Rannie, W. D.: Detonation and Two-phase Flow, in "Progress in Astronautics
and Rocketry," vol. 6, S. S. Penner and F. A. Williams (eds.), Academic.Press,
Inc., New York, 1962.
SUSPENSIONS OF PARTICLES IN FLUIDS m
37. Marb!e, F. E.: AIAA J., vol. 1, no. 12, pp. 2793-2801, 1963.
38. Carrier, G. F.: J. Fluid JJ.1ech., vol. 4, pp. 376-382, 1958.
39. Kriebel, A. R.: ASME Trans. J. Basic Eng., vol. 86, no. 4, pp. 655-665, December,
1964.
40. Rudinger, G.: Phys. Fluids, vol. 7, pp. 658-663, 1964.
41. Kliegel, J. T., and G. R. Nickerson: Detonation and Two-phase Flow, "Progress
ir;i Astronautics and Rocketry," vol. 6, S. S. Penner and F. A. Williams (cds.)¡
Academic Press, Inc., New York, 1962.
42. Nickerson, G. R., and J. R. Kliegel: TRW Rept. 6120-8345-MUOOO, lVlay, 1962.
43. Boyer, M. H., and R. Grandey: Theoretical Treatment of Detonation Behavior
of Composite Propellants, Detonation and Two-phase Flow, in "Progress in
Astronautics and Rocketry," vol. 6, S. S. Penner and F. A. Williams (eds.),
Academic Press, Inc., New York, pp. 75-98, 1962.
Also Williams, F. A.: Detonations in Dilute Sprays, ibid., pp. 99-114.
Morgenthaler, J. H.: Analysis of Two-phase Flow in Supersonic Exhausts,
Ibid., pp. 145-172.
l\iarble, F. E.: Phys. Fluids, vol. 7, no. 8, pp. 1270-1282, 1964.
Bailey, VV. S., E. N. Nilson, R. A. Serra, and T. F. Zupnik: Am. Rocket Soc. J.,
vol. 31, no. 6, pp. 793-798, June, 1961.
Rudinger, G.: AIAA J., vol. 3, pp. 1217-1222, 1965.
Rudinger, G., and A. Chang: Phys. Fluids, vol. 7, pp. 1747-1754, 1964.
Gilbert, M., J. Allport, and R. Dunlap: Am. Rocket Soc. J., pp. 1929-1930,
December, 1962.
44. Prandtl, L.: uEssentials of Fluid Dynamics," p. 342, Blackie & Son, Ltd., Glasgow,
1952.
45. Basset, A. B.: "Hydrodynamics," p. 270, Dover Publications, Inc., New York,
1961.
46. Einstein, A.: Ann. Phys., vol. 4, p. 289, 1906.
47. Thomas, D.G.: A. I. Ch. E. J., vol. 6, no. 4, pp. 631-639, 1960.
48. Thomas, D.G.: A. J. Ch. E. J., vol. 8, pp. 266-278, 1962.
49. Happel, J.: J. Appl. Phys., vol. 28, pp. 1288-1292, 1957.
50. Eirich, F. R.: "Rheology," 3 vols., Academic Press, Inc., New York, 1956.
51. Thomas, D.G.: A. I. Ch. E. J., vol. 7, no. 3, pp. 431-437, 1961.
52. Zuber, N.: Chem. Eng. Sci., vol. 19, pp. 897-917, 1964.
53. Rutgers, R.: Rheol. Acta, vol. 2, no. 41 pp. 305-348, 1962.
54. Rutgers, R.: Rheol. Acta, vol. 2, no. 31 pp. 202-210, 1962.
55. Bingham, E. C.: "Fluidity and Plasticity," McGraw-Hill Book Company, New
York, 1922.
56. Casson, N.: ºRheology of Disperse Systems," C. C. Mill (ed.) 1 Pergamon Press,
New York, 1959.
57. Ree, T., and H. Eyring: chap. 3 in Ref. 50.
58. Charm, S. E., G. S. Kurland, and S. L. Brown: ASME Biomed. Fluid Mech.
Symp., Denver, Colo., pp. 89-93, 1966.
59. Huff, W. R., J. H. Holden, and J. A. Phillips: U.S. Bureau of :Mines Rept. RI6706,
1965.
60. Thomas, D.G.: Paper no. 64, Progr. Intern. Res. Thermodyn. Transport Properties,
ASME, 1962.
61. Daily, J. W., and G. Bugliarello: TAPPI, vol. 44, pp. 497-512, 1961.
62. Bugliarello, G., and J. W. Daily: TAPPI, vol. 44, pp. 881-893, 1961.
63. Mih, W., and J. Parker: TAPP I, 21st Eng. Conf., vol. 50, no. Ei, p. 237-246, Boston,
Mass., May, 1967.
 ONE-DIMENSIONAL TWO-PHASE FLOW

64. Savins, J. G.: Soc. Pet. Eng. J., vol. 4, no. 3, pp. 203-214, 1964.
65. Thomas, D.G.: Transport Characteristics of Suspcnsions, part VI, A. I. Ch. E. J.,
vol. 8, pp. 373-378, 1962.
66. Zenz, F. A.: Ind. Eng. Chem., vol. 41, pp. 2801-2806, 1949.
67. Chow, V. T.: "Open Channel Hydraulics," pp. 164-179, McGraw-Hill Book
Company, New York, 1959.
68. Kennedy, J. F.: J. Boston Soc. Civil Engrs., vol. ,52, pp. 247-266, 1965.
69. Segre, J., and A. Silbcrberg: J. Fluid "Jl!lech., vol. 14, pp. 136-157, 1962.
70. Ormiston, R.M., F. R. G. Mitchell, and J. F. Davidson: Trans. Inst. Chem. Engrs.,
vol. 43, pp. 209-216, 1965.
71. Thomas, D.G.: Science, vol. 144, pp. 534-536, 1964.
72. Rowe 1 P. N.: Chem. Eng. Progr. Symp., vol. 58 1 p. 42, 1962.
73. Rowe, P. N., and B. A. Partridgc: UKAEA Rcpt. AERE-R4660, 1964.
74. Harrison, D., and L. S. Leung: Trans. Inst. Chem. Engrs. 1 vol. 40, p. 146, 1962.
75. Davidson, J. F., R. C. Paul, M. J. S. Smith, and H. A. Duxbury: Trans. Inst.
Chem. Engrs., vol. 37, p. 323, 1959.
76. 1\.-Iurray, J. D.: Harvard University, Cambridge, Mass., NSF Grant GP-2226, 1963.
77. ,Jackson, R.: Trans. Inst. Chem. Engrs., vol. 41, pp. 22-28, 1963.
78. Rowe, P. N., B. A. Partridge, and E. Lyall: UKAEA Rcpt. AERE-R4.543, 1964.
79. Davies, R.M., and G. I. Taylor: Proc. Roy. Soc. (London), vol. 200, p. 375, 1950.
80. Davidson, J. F., and D. Harrison: "Fluidised Particles," pp. 33 and 37, Cambridge
University Pross, London, 1963.
81. Harrison, D., J. R. Davidson, and J. VV. DeKock: Trans. Inst. Chem. Engrs.,
vol. 39, p. 202, 1961.
82. Pigford, R L.: Massachusetts Institute of Technology, unpublished work 1 1959.
83. Roscoc, R.: Brit. J. Appl. Phys., vol. 3, p. 267, 1952.
84. Jaeger, C.: "Engineering Fluid Mechanics," St Martin's Press, Inc., New York,
1957.
85. Frantz, J. F.: Chem. Eng. Proc. Symp., ser. 62, vol. 62, p. 29, 1966.
86. Brinkman, H. C.: J. Chem. Phys., vol. 20, p. 571, 1952.
87. Taub, P. A.: J. Fluid Mech., vol. 27 1 pp. 561-580, 1967.
 9
Bubbly Flow

9.1 lNTl!Ol>IJCTION

The bubbly flow pattern is characterized by a suspension of discrete

bubbles in a continuous liquid. There are numerous regimes of bubbly
:flow. Void fractions range from the extreme case of a single isolated
bubble in a large container to the quasi-continuum flow of a foam, con-
taining less than 1 percent of liquid by volume. Interactions between
the forces that are dueto surface tension, viscosity, inertia, and buoyancy
produce a variety of effects which are quite often evidenced by different
bubble shapes and trajectories. The regime in which bubbles are so
large that they assume a cylindrical shape and almost fill the duct in
which they are flowing is important enough to warrant a separate name,
slug fiow, and will be discussed in the next chapter.
Engineering applications include bubble columns for promoting
mass transfer, high-pressure evaporators, flash distillation, fire-fighting
foams, pumping of beer, champagne, and ice cream, cryogenics, foam
icstripping 11 of impurities, sewers to handle the effiux from washing
Z4l
 ONE-DIMENSIONAL TWO-PHASE FLOW

machines, instant lather for shaving, underwater breathing, and control

of wave action in harbors.

9.2 BUBBLE FORMATION

It is only rarely that bubbly flow is the final stable equilibrium flow
regime in a given duct. Gas bubbles suspended in fluids usually tend to
agglomerate and lose their identity, and when evaporation or condensa-
tion occur the existence of small bubbles is only transitory. For these
reasons it is particularly important to understand the ways in which
bubbles can be formed. Moreover, the bubble size has an influence on
the dynamics of a bubbly mixture and must often be specified in terms
of the mechanism of bubble generation. In spite of the wealth of avail-
able literature only a few equations for predicting bubble size have
received general acceptance. Sorne of the simplest of these will now be
described.

BUBBllE FORMATION ATAN ORIFICE

One of the simplest ways to form a bubble is at a circular orífice facing

upward in a stationary fluid. Assuming an approximately spherical
bubble of radius Rb attached to an orífice of radius Ro by a cylindrical
neck we have, for the largest bubble which can be in static equilibrium,
(9.1)

The radius of a bubble which will be formed by blowing through a small

orifice at low flow rates is therefore given approximately by

R ~ [
b ~
3uRo
2g(p¡ - p,)
]¼ (9.2)

According to Kutateladze and Styrikovich 1 a more accurate version of

Eq. (9.2) which is derived from experimental data is

uRo
Rb - LO [ g(P! - )
]¼ (9.3)
p(J

Equation (9.3) ceases to be valid when the orífice diameter is comparable

with the bubble radius, that is, when

Ro> 0.5 [ g(P! u- ) ]'' (9.4)

p(J

When a bubble is formed at a finite rate many other factors are

important including all the liquid and gas properties and the details of
the orífice design and gas supply. The Iiterature in this field has been
BUBBLY FLOW

reviewed by J ackson,2 and tbe general phenomenon of bubble formation

is shown to be surprisingly complicated.
Wben the gas flow rate through the orífice is increased, the bubble
size first increases due to the fact that the bubble takes a finite time to
break from the orífice alter reaching the size given by Eq. (9.3). Far a
system in which the flow rate through the orífice is carefully maintained
constant, the bubble size at departure can be predicted by knowing the
time for which the bubble remains attached to the orifice. 3 Davidson
and Amick 4 have calculated this time from the equation of motion of
the rising bubble and deduced the following result for the volume o! a
bubble at detachment in an inviscid liquid:
Q%
'I.J, = 1.138 7r (9.5)
g

where Q, is the volumetric gas flow rate through the orífice. This equa-
tion is compared with data for a ,vide variety of orifice sjzes in water in
Fig. 9.1. The deviations from the theory at lower gas flow rates are due
to approaching the quasi-static limit given by Eq. (9.3). A more general
solution for bubbles growing as a power of time is considered in Probs.
9.3 to 9.6. Davidson and Bchüler 5 have also made similar calculations
for viscous liquids and obtained the result

'I.J, = (4,,-)¼
3
[2g(p¡
l5µ¡Q, ]"
- p,) (9.6)

When the gas velocity is large, the momentum flux from the orífice
becomes significant and the bubble volume is given by solving

(36,r)¼ = Zg(p¡ - p,) 'U,;, + Q,p, 'I.Jé' (9.7)

5µ¡Q, A 0µ¡

A o is the area of the orífice.

If the gas velocity through the orífice is made sufficiently high,
bubbles are no longer.formed individually, but the gas leaves the orífice
in the form of a jet whieh eventually breaks into individual bubbles.
According to Kutateladze and Btyrikovich, 1 the condition for the forma-
tion of a gas jet is

v, ...;;,
[g<r(p¡ - p,)]¼
> 1 25 [
. '
"
g(p¡ - p,)Ro 2
]¼ (9.8)

where v, is the gas veloeity through the orífice. Bubbles which are
formed in this way have a radius which is about twice the radius of the
orífice.
In commercial applications bubbles are not usually formed at a
 ONE-D!MENSIONAL TWO-PHASE fLOW

1 1 1 7 TTTTT1 1 1 1 1 111
1 1 1 1111 1 1 1 1111
'

f /
1000 -
--
t:
f- r;f -
r
Volume from: - -

•ºº"1
r
-
r
f-
f-ín.-dio.
tube .o Counting
Meosurement
ºoíl/
o/ -

-
f- Wolters
¾-in.~dio. o Counting
oº/

1/
tube Meosurement
100
o ir-water
'
::
~

t: 1- in.-dio.
"• Counting /

"'"' "'¡
f-
f-
tube -
-
f-
f-
-
-
f-
-
f- Eq. 9.5
/,
=
r7
-
- -
-- -
- -
-
- 1

f/
+ Orifice -
f-

'
+ Colderbonk Capillory, 0.265 cm + -
f-
~
+
oir-woter Slots, ~ - ¼
in. wide
'
'"+ van Krevelen Copillory, 1 air-woter T
1.0 , --
t:
f-
r /+
,,.
,A ond Hoftij2er

David san
0.23 cm ] H2-water 1 -
-
-
f-
Hale dio., 0.48 cm V
and Amick -
f-
/

,Z
f- -
Hale dio., O. 15 cm -
r
Schüler 0.25 cm -- -

1 1 1 1 111 " I I I L 1 11
0.1 ' ' '
' ' ' ' '
1.0 100 1000 10,000
Gas flow rote, Og, ml/sec

Fig. 9.1 Volume of bubbles produced by blowing from submerged orifices and nozzles
in in vise id fluids. (Davidson and H arrison. 4)

single orífice but by a group of orífices or a porous plate. In this case

the single-orifice theory is useful only as a first approximation.

FORMATION OF BUBBLES IBY TAYLOR INSTAIBIUTY

Under sorne circumstances bubbles can be formed by detachment from

a blanket of gas or vapor which is maintained over a porous or heated
surface. The continua! formation of these bubbles is not identical with
the classical case of "Taylor instability" 6 of a fluid below a denser fluid,
but the bubble size is scaled by the same dimensionless parameter.
Therefore

Ro --
u
[ g(p¡ - p,)
]¡, (9.9)
 BUBIBLY FLOW

The theory of Taylor instability has been extended to the case of

continuous bubble production by Zuber 7 and is particularly important
for describing film boiling.

FORMATION OF BUBBl!ES BY EVAPORATION OR MASS TRANSFER

Bubbles can also be produced by the evaporation of the surrounding

liquid and by the release of gases which are dissolved in the liquid (e.g.,
beer, soda, champagne). These bubbles almost invariably form around
nucleation centers which are either impurities suspended in the fluid or
pits, scratches, and cavities on the walls of the containing vessel. Of
particular interest is the formula derived by Fritz 8 for the equivalent
diameter (diameter of a sphere having the same total volume) of a bubble
which is just large enough to break away from a horizontal surface. If
(3 is the contact angle in degrees, the equivalent diameter is

d - 0.0208(3 [( ~
g Pi - Pu
) ]" (9.10)

This equation is valid muy in the quasi-static case and does not adequately
describe bubbles which are formed rapidly during boiling, far example.

THE INFLUENCIE OF SHEA.IR STRESSES ON BUBBLE Si:ZE

In forced convection or mechanically agitated systems the bubble size

is determined by shear stresses. These stresses influence both the size
of bubbles which are torn away from their point of formation and also
the maximum bubble size which is stable in the flow field. One would
expect the critica! bubble size in both situations to be governed by the
balance between surface-tension forces and fluid stresses, i.e., by a suit-
ably chosen Weber number. Little authoritative work seems to have
been done in this field but a formula from Hinze' may be useful far esti-
mating bubble size, namely,

d - 0.725 ("---)¾ (p)-¾

P!. M
(9.11)

The quantity P / M represents the mechanical power dissipated per unit

mass.

9.3 ONE-OIMENSIONAL VERTICAL FLOW Of A

BIJBBLY MIXTURE WITHOUT WALl SHEAR

As long as the wall shear stresses are small and the concentration and
velocity profiles are approximately unifarm, the techniques described
in Chap. 4 can be applied to analyze bubbly flow in a vertical pipe.
The key to a successful understanding is to find an expression far
 ONE-DIMENSIONAL TWO-PHASE FLOW

the drift ftuxj, 1 in terms o! basic quantities. For most practica! purposes
the empírica! equation (4.14) is a good approximation. Thus
(9.12)

and it remains to express v and n in terms of the fluid properties and

00

bubble size.

THIE RISE. VELOCffY OF SINGLE SUIBBLIES

The dependence of the terminal rise velocity of a single bubble, v~, upon
fluid properties has heen determined experimentally by Peebles and
Garber, 10 Haberman and lVIorton, 11 and by numerous other investigators.
The dependence of rise velocity on bubble volume for air bubbles in
water is shown in Fig. 9.2. For the smallest hubbles, which are approxi-
mately perfect spheres hecause of the dominant effect of surlace tension on
their shape, Stokes' solution 12 provides a reasonably accurate description,
1 d'g(p 1 - p,) (9.13)
"~ = 18 µ¡

The equation is valid lor salid spheres and it is assumed that the liquid
velocity goes to zero at the bubble surface. For fluid spheres contain-

60~-~-~~~~~~--~--~-~-~----~
50
40 e
30
u
•
....".:20 D
E
u

J. I0f----~--m,--------t----------+-----~
:¡:-- Bt.
·u 8 n.
o
1! Ó X ◊
o 5
-~ 4
E ◊

"' 3

2 A

1L __ _ _ ._ _L__.__.L_L_L_l_LJ__ __L_..J...___[___[__LJ....Ll.L_ _..J...__L_j

0.01 0.02 0.04 0.06 0.1 0.2 0.4 0.6 0.8 1.0 2.0 4.0
Equivalent radius, Rb, cm

F'ig. 9.2 Terminal velocity of air bubbles in filtered or distilled water as function of
bubble size. (Haberman and Morton. 11 )
BUBBLY FLOW

ing liquid with viscosity µ 0 and having a completely nonrigid surface 1

Hadamard 13 and Rybczynski 14 obtained the equation

v - d'g(p¡ - p,) 3µ, + 3µ¡ (9.14)

" 18µ¡ 3µ, + 2µ¡
If µ, « µ¡ this reduces to
d'g(p¡ - P,)
v00=--1~ (9.15)

In the complete absence of impurities, which tend to collect at the bubble

surface and give it a certain resistance to shear stress, it is possible to
obtain the result predicted by Eq. (9.15). However, in most practica!
cases sorne contamination is present and the bubble rise velocity is found
to Iie between the values which are given by Eqs. (9.13) and (9.15).
At the otber extreme, when the bubbles are very large, the effects
of surface tension and viscosity are negligible and the rise velocity is
given by the equation of Davies and Taylor, 15

(9.16)

where Re is the radius of curvature in the region of the bubble's nose.

The shape of the bubble is approximately a spherical cap with an included
angle of about 100° and a relatively flat tail.
Equation (9.16) can be expressed in terms of the volume of gas in
the bubble, u,. Thus

(9.17)

Alternatively, we can define an equivalent radius Rb, which is the

radius which the bubble would have if it were spherical

u, - Y3.,,-R," (9.18)

The bubble rise velocity is then

(9.19)

For bubbles of intermediate size, both the effects of liquid inertia,

surface tension, viscosity, and cleanliness are important 1 as well as whether
the bubbles rise in straight Iines, oscillate, or describe a spiral path.
Many correlations exist in the Iiterature of which perhaps the most com-
prehensive are from Peebles and Garber, 1º who suggest the equations
shown in Table 9.1 (for a gas density negligible compared with liquid
density). The range of applicability of each equation is determined in
250 ONE-DIMENSlONAL TWO-PHASE FLOW

Table 9.1 Terminal velocity of single gas bubbles in llquids (according to Peebles
and Garber) 10

Terminal velocity Range of applicability

2Rb 2 (p¡ - pg)g

Region 1 v,,, = Reb < 2
9µ¡

Region 2 v.,, = 0.33g0.76 (PI)""

~ Rbl.28 2 < Reb < 4.02G 1-2.2u

4.02G1-o.214 < Reo < 3.10G1-o.2s

Region 3 v..,= L35 ( " )"'"
~-
p¡Ro
or
16.32G1º· 144 < G2 < 5.75

Region 4 v.,, = 1.18 -

e•Y'
PI '
3.10G1- 0 · 25
5.75 < G2
< Re 6

terms of the following dimensionless groups:

Re, = 2p1v.R, (9.20)
µ¡
gµ,¡4
G, p¡u3
(9.21)
gRb4Voo 4p¡3
G, (9.22)
a'
It is noteworthy that in region 4 the bubble rise velocity is independ-
ent of size. Harmathy 16 suggests that a better value for the constant
(1.18) is 1.53 in this region. The upper !imit of region 4 is reached when
the rise velocity is comparable with the value given by Eq. (9.19), i.e., for

R, ~
( a)"
2 -
gp¡
(9.23)

which defines a further region 5 in which Eq. (9.19) is valid.

Example !U Calculate the rise velocity of air bubbles of equivalent radii 0.2, 0.5,
and 2 cm in water if u- = 70 dynes/cm, p¡ = I g/cm3, µ¡ = 0.01 poise, and
g = 981 cm/sec 2.
Solulion From Eq. (9.21) and the given property values

G ~ (981)(0.01)' ~ 2 86 X 10-n
2
(1)(70)' .
Therefore 4.02G1- 0 · 214 = 715 and 3.1G1-H = 1340.
From Eq. (9.23), moreover,

2 (-"-) ¼ ~ 0.51 cm
gp¡
BUBBlY FLOW 251

Rather than iterating to determine the various rcgions in Table 9.1, we use
Fig. 9.2 to obtain a first approximation to thc bubblc velocity and hence the
Reynolds number given by Eq. (9.20). The results are

Rb, cm 0.2 0.5

vb, cm/seo 25 25
Re, 1000 2500
Region 3 4

The predicted bubble velocities, using Table 9.1 and Eq. (9.19), are then

Rb = 0.2 cm Voo = 1.35 ( 70)¼

_ = 25.2 cm/sec
02
Rb = 0.5 cm V = 1.18}
or [(981)(70)]%
19.1 cm/sec

1.53 24.8 cm/sec

v. = l.00[(981)(2)]fi = 44.3 cm/sec
Comparison with Fig. 9.2 indicates agrcement with experiment and favors the
value of the constant given by Harmathy in region 4.

THE INFLUENCE OF CONTAINING WALLS

When a bubble rises in a finite vessel, its velocity is generally lower than
the value predicted by Table 9.1.
In a tube of diameter D, the ratio of
the bubble velocity v, to the velocity in an infinite medium can be
expressed as a function of d/D, where d = ZR,.
Collins 17 studied large inviscid bubbles corresponding to region 5.
A good approximation to his results is given by
d
D < 0.125 1 (9.24)
d
0.125 < D < 0.6 (9.25)
d
0.6 < D (9.26)

Equation (9.26) is equivalent to the equation governing the velocity of

rise of slug flow bubbles in an inviscid fluid [Eq. (10.5)].
Similar corrections can be made for bubbles in viscous fluids. If the
bubbles behave as salid spheres, Ladenburg 18 derived the result

(9.27)

whereas far fluid spheres with µ, « µ,¡, Edgar 19 obtained

v, = (1 + 1.6d)-l (9.28)
"~ D
 ONE-DIMENSIONAL TWO-PHASE FLOW

If d/D exceeds about 0.6, the bubbles behave as slug ·flow bubbles and
obey the equation

0.6
d
< TI v
00
(i)-
v, = 0.12
D
2
(9.29)

A reasonable fit to Eq. (9.28) which is tangential to Eq. (9.29) at

d/D = 0.6 is
v, _ l _ d/D (9.30)
V
00
-0.9

and this may be used to estímate bubble velocities for d/ D < 0.6.
INFLUENCE OF VIBRATIONS

If a bubble is placed in a vibrating vertical column of liquid, it experiences

a downward force which opposes gravity. Under suitable circumstances
it can be made to oscillate stably about a stationary mean position or
even move downward. 20 •21 ,50

THIE JNFLUIENCE OF VOlD FRACTION

The influence of void fraction is conveniently represented by Eq. (9.12).

It is necessary merely to obtain suitable values far the index n.
Wallis" made the simple assumption that the relative velocity
varied linearly with concentration and recommended a value of n equal
to 2. Using the value of bubble rise velocity given by Peebles and
Garber's region 4 (Table 9.1) led to the result
(9.31)
If Harmathy's recommendation is followed the value of the constant in
Eq. (9.31) should be changed to give
(9.32)
A comparison between the predictions of Eqs. (9.31) and (9.32)
and experimental results of Shulman and l\!Iolstad 23 for countercurrent
flow of air and water is shown in Fig. 9.3. Again Harmathy 1s value of
the constant is superior to Peebles and Garber' s and is recommended for
use with gas-liquid systems. For liquid-liquid systems Eq. (9.31) is
superior.
The method of plotting is the one indicated in Fig. 4.1 and enables
steady-state operating values to be indicated by points, whereas the
observed flooding points are indicated by tangents. The departure of
the points from the theoretical bubbly flow curve at approximately
a = 0.3 is dueto flooding and a consequent change in flow pattern. The
new flow pattern contains very large bubhles and the total gas-liquid
BUBBLY FLOW m

0.3~--------------
x Jr=0,fps
o -;¡ = 0.045
D. -;¡ = 0.089
X □ -Jr=0.133
'v -;/=0.178
o -¡,-0.222
ro - - Eq.(9.31)
t 0.2 1--------,-"- - - - Eq.(9.32)

Fig. 9.3 Results of Shulman and

J\folstad 23 for an air-water sys-
tem in vertical bubbly fiow
plotted according to the meth-
ods of .Fig. 4.1. º'---------------"'-"
o
Void fraction, a. -

interfacial area is markedly reduced above the flooding point. Thus

Shulman and l\1olstad 23 found a sudden reduction in mass-transfer
coefficients at the flooding points which are shown in tho figure.
In many applications a gas (or light fluid) is bubbled through a
stagnant liquid. In this case Eqs. (9.31) and (4.7) lead to the result

(9.33)

Comparison between this equation and the results of Kutateladze"

bubbling water through mercury are shown in Fig. 9.4. Flooding and a
change in flow reglme occur at a = 0.4 near the maximum in the curve.
There are indications that the value of n is not always exactly 21
although at low values of a in the main region of application of bubbly
flow theory this variation is not particularly significant. Gaylor, Roberts,
and Pratt 25 working with liquid-liquid systems obtained a value n ~ 2
in the region 1 given in Table 9.L Miles, Shedlovsky, and Ross 26 work-
ing with stable foams at high values of a obtained results in the range
n ~ L6 to L9. Zuber and Hench 27 present further analytical and
experimental results and recommend that the following values of n are
 •. o G)
6 ' • o 0 .,
Q)
t '"□~~
~ 5 - - - Eq. {9.33) ¡-- --- □ >----..
o

<E 1
Flow regime
m9'.°6
._ ff
/''
u e----- chonge e' Je □"' e, ., o-
º 4 to flooding -..............-----i---.--~,a'------j---,
-~

2
X

iu 3 .
o
o

;:¡ 21---=-./-+---- Orifice diameter d0

G) ~ 10 mm
1 A o o (J) ~ 5 mm
Q) 3 mm
º"-----"---~-~-~--~----c"-c----:c
O 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Water vol u me froction, a _..,,,.
(a)

•. º CD
6 18 • O@ ------+------~.~º
• . • Q) "oº'll'
u 5 - - - Eq. {9.33) □ºm/¡
(l.)

~E
~4e------~-------t-~----,------j
Dl',OI
~· 1110

6
A

o ,•
-~
X Flow regime
~ 3 chonge due
u
2' to flooding
¿t<¡o
;:¡ 2 f----•"'~--- plateFraction of
perforated
,I' •
• o o
G) ~ 44.5 %
• 0 ~ 32.3 %
G) 12.5%
º~-~--~--~--~--~-~--~
o 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Water vol u me fraction, et _ .
(6)

Fig. 9.4 Results of Kutate1adze and Moskvicheva 24

bubbling water through mercury from a variety of ori-
fices. a Fraction of plate perforated = 33 percenti
h 0 = 55 mm. b Orifice diameter d 0 = 5 mm, initial
height ho = 155 cm.
BUBBLY FlOW

applicable in the various regimes defined in Table 9.1:

Region 1 n=2
Region 2 n = 1.75
Region 4 n = L5
When the bubble size is larger than the value given by Eq. (9.23),
three-dimensional effects become important and there is significant
entrainment of bubbles in each other's wakes. The result of this "stream-
ing" or "channeling" is an increase in the relative velocity with an increase
in the number of bubbles present. The index nin Eq. (9.12) therefore
becomes less than unity. The flow pattern becomes agitated and
unsteady. This regime, which is called churn-turbulent by Zuber, appears
to represent a transition region between "ideal bubbly flow," in which
bubbles rise uniformly and steadily, and slug flow in which large bubbles
fill the tube and flow entirely in each other' s wakes. Agglomeration is
particularly significant in the churn-turbulent regime since a bubble which
is flowing in the wake of another tends to rise faster than its predecessor
and eventually coalesces with it. Steady state is not reached until
coalescence is complete, and this may require a considerable length of duct.
According to Zuber, 27 in the churn-turbulent regime bubbles rise
with the velocity characteristic of region 4 in Table 9.1 relative to the
volumetric average velocity of the mixture. The drift velocity of the

r
bubbles is therefore

v,; = L53 [ ug(p:;;- P,) (9.34)

The corresponding value of the drift flux is

· = 1 · 53 a
Jo!
[ug(p, - p,)]¼
2 (9.35)
Pi

and the value of n is zero.

MOD!FICATIONS TO THE SIMPLE THEORY TO TAKE ACCOUNT

OIF VARIATIONS IN CONCENTRATION AND VELOClTY

Variations in bubble concentration and velocity can be taken into account

by the method of Zuber and Findlay 28 which was described in Chap. 4.
Making use of the flow distribution parameter C 0, Eq. (4.24), forexample,

r
becomes

v, = C,j + L53 [ ug(p:;;- p,) (9.36)

for the churn-turbulent bubbly regime. For this regime the value of C 0
sufficiently far from the point of bubble injection usually lies between
256 ONE-DIMENSIONAL TWO-PHASE FLOW

12.0r--,----¡---,--.------,--------,--~--~-~--

u
• 8.0f-------f---------+----
-.::
E

•
'-~
◊

11
:,,.t>i 6.0r-----f----------t---r---+----+-----

Air-woter mixtures
ir,
o ft / sec cm/sec
0 o A o o
B 0.1 3.04
◊ e o. 2 6. 1 O
º D o. 3 9.15
◊ E 0.4 12.20
º F 0.6 18.30
G 0.8 24.40
º H 1.0 30.45
D == 2" == 5.08 cm

6.0 B.O
Overol! flux,j, m/sec

Fig. 9.5 Comparison between Eq. (9.36) and experimental data. (Zuber
and Findlay. 28 )

1.0 and 1.5 with a most probable value of about 1.2. Comparison
between this theory and some data is shown in Fig. 9.5. The method of
plotting shown in the figure gives C0 as the slope of the curve and v,;
as the intercept on the Va axis.

9.4 UNSTEADY flDW

Unsteady fl.ows are conveniently analyzed by combining continuity wave
theory with the appropriate expression for the drift flux. The methods
are exactly parallel to those which were used in the analysis of sedimen-
tation in the previous chapter.
For example, considcr the drainage of liquid from an initially uni-
BUBBLV FLOW

form foam in a vertical tube. Because a is clase to unity it ls convenient

to work in terms of the liquid fraction ,.
Let the initial value of E be Eo- As the foam starts to drain, waves
will begin to propagate from the top and the bottom where the end
condjtions are specified.
Figure 9.6 shows the situation on the drift flux-concentration graph
and on the time-distance plane.
The conditions for a stable shock are satisfied at the bottom of the
foam, and an interface is formed which rises with a veloclty Vs, the fluid
below thc interface being pure liquid. No shock can occur at the top of
the foam because of the upward curvature of the j, 1 curve. Continuity
waves corresponding to the spectrum of voidages E = O, to € = Eo will
therefore propagate down into the foam, each with its own particular
velocity. The foam drainage proceeds in two stages as shown in the
figure.
Stage 1 Initially a shock betwecn , = 'º and , = 1 propagates up
from the bottom while continuity waves propagatc down through the

e=I

i
h

'º
e= l, pure liquid

Fig. 9.6 Dra.inage of an initially uniform foam repre-

sented on the drift flux-concentration and time-distance
planes.
 ONE-DlMENS!ONAL TWO-PHASE FLOW
"'

Fig. 9.7 Foam drainage represented on the XY plot. For

convenient comparison with Fig. 9.6 the direction of Y is
shown downward.

foam from the top. Throughout this stage the rate of drainage of liquid
from the foam is given by the shock velocity and its constant.
8tage 2 VVhen the first continuity wave corresponding to a liquid
concentration Eo - ó1; reaches the interface, the voidage above the shock
begins to increase and the shock strengthens and slows clown. At any
subsequent time the liquid concentration above the interface is given by
the value which has just had sufficient time to propagate down from the
top.
This series of events resembles an upside-down version of type III
sedimentation and can be represented by the universal XY plot as shown
in Fig. 9.7 and described by Eqs. (8.68) and (8.69). The value of °'~ is
essentially unity when drainage is complete.
If the height of the shock at any time during stagc 2 is h and the
tube has a cross-sectional area A, then the amount of liquid which has
been drained is A (ho - h) and the amount of liquid remaining in the foam
lS

Ahoeo - A(ho - h) = Ahoao (h~ao - 1) = Ah 0 ao(Y - 1) (9.37)

The quantity A h 0a 0 is the initial gas content of the entire foam. Rewrit-
ing Eq. (8.69) in terms of , we have
(Y - 1) = n - (n + 1), - 1 = ,(n - 1) - n,' (9.38)
n(l - ,) 2 n(l - ,) 2
If , is small, an adequate approximation for the amount of liquid lelt in
the foam when e is the concentration just above the shock is therefore,
8UBBLY FLOW

from Eqs. (9.37) and (8.38),

(Ah 0a 0 ),(n - 1)
'U¡=------- (9.39)
n
This occurs after a time t which is derived fron1 the parameter X, as
follows:
aoho aoho
t ~ Xaoho
n(l _ 1:)21:n-lV"° = n1:n-1V"° (9.40)

Stage 2 starts when the value of e in these equations is cqual to 1: 0 .

Stage 1 is represented by a straight line between the initial conditions and
the start of stage 2 on the ht plane.

IExample 9.2 Miles, Shedlovsky, and Ross 26 performed experiments in a vertical

column of diameter 3.2 cm with a uniform foam which was produced from a
0.24% sodium lauryl sulfate solution. Stcady flow experiments for liquid
dr:iining through stationary foam gave results which correlated with the equation

y = 0.24xl.8 (9.41)

in which y was the rate of liquid flow through the foa.m in cubic centimeters per
minute and x was thc volume of liquid in a total 295 cm 2 of foam.
Thcy also studied unsteady flow by suddenly closing the valve in the
liquid supply line and plotting the amount of liquid remaining in the foam as a
function of time. Predict their results if at the beginning of an experiment
6.9 cm 3 of liquid were uniformly dispcrscd in a total 29,5 cmª of foam.
Solution With the present notation Eq. (9.41) becomes

-j¡o = 13.9t1 · 8 (9.42)

if velocities are measured in centimeters per second. Substituting Eq. (9.42)

into Eq. (4.7) with jg = O gives

Jgl = }3.9tL 8 (1 - t) (9.43)

Therefore v"' = 13.9 cm/sec and n = l.8. A 0 a 0 h 0 is equal to

295 - 6.9 = 288.1 cm 3 and C1.oho is 288.1/8.05 = 35.8 cm. Using these values
in Eqs. (9.39) and (9.40) we get

'01 = 12fü cm 3 (9.44)

t = 1.432t- 0 •8 sec (9.45)

Substituting "'º = 6.9/295 = 0.0234 into Eqs. (9.44) and (9.45) thc end of stage
1 is predicted to occur when

'0 1 = 3.0 cm 3, t = 28.8 sec

Stage 1 is represented by a straight line joining the initial condition to this point.
The predictions for both stages in the process are comparcd with the
experimental data in Fig. 9.8.
260 ONE-DlMENSIONAl TWO-PHASE FLOW

o Experimental
points 26
150
--Theory

©100>1-+-----+------
E
"'
50

º
º'---'--'--"-----'--'--'--"'o'--
2 4 6 8 Flg. !U Foam drainage results (for
Liquid in foam, ce Example 9.2).

9.5 SPECIAL PROBLEMS ASSOCIATED WITH THE BUBBLY FLOW REGIME

BUBEILE SiZE

A considerable problem in practicc is the determination of bubble size

since the simple equations given in 9.2 do not apply to the complex geom-
etries of most engineering equipment. However, the bubble size is an
important variable for determining bubble rise velocity from Table 9.1
and the corresponding value of n in Eq. (9.12) as well as influencing
whether operation is in the ideal bubbling or churn-turbulent regime.
Figure 9.9, for example, shows a striking example of the way in which
completely different results can be obtained for the same values o! overall
flow rates when air is blown into a column of water from various per-
forated plates.

AGGUJMERATJON AND FRACTURE OF BUBBLES

Unless special precautions are taken 1 such as the addition of surface-active

agents to the fluid, bubbles tend to coalesce as they touch each other.
If the bubbles which are introduced into a system are smaller than the
maximum stable size for the prevailing shear fi.eld, which can be estimated
from Eq. (9.11), they will eventually agglomerate. The bubble size, and
hence dependent variables such as void fraction, is then a function not
only o! the way in which bubbles are produced but also of the distance
which they travel from the point of injection. Far example, il air is bub-
bled through stagnant tap water the mean void fraction depends on the
depth of water in the vessel (Fig. 9.10). Similarly, quite different results
BUBBLY FLOW m

may be obtained in the same apparatus if the fluid purity, and hence
resistance to agglomeration, is altered (Fig. 9.11). An approach to the
agglomeration problem has been made by Radovcich and Moissis. 29

BUBBLE GROWTH AND COLLAPSE

Bubble size is time dependent when bailing) flashing, ar candensatian

occur, when gas bubbles are released fram solution ar dissalved, and when
bubbles move through significant pressure differences and expand ar cal-
lapse. Sorne effects which result from these phcnomena will be intro-
duced in the problems at the end of this chapter.

Perforated p/ates
No.of Diameter Square orray
orifices (cm) spacing (cm)
• 4.06 X 10-l
30 o 49 4.06 X 10-l 6.25 xl0- 1
o IDO 1.52 X 10- 1 9.50xl0-l
X
289 0.41 X 10-I 6.25 xl0- 1

25
o, o

<•E 20 o, o

·1
X
e º
a
ce
o 15
ü Churn- ci Bl 0
°
tocbule,t
regime o □
X
0
10 1--------1--'h--0~ + - - - - - - ~ x x x x ~
eo □ Xxxx
□ X XX
C0 □ XXX
□ X XX
@O □ X X
0 □ XX
• O X
0 O X
00 o
5 o
lf:¡ rP xx x
"toºº 0
X X xX
<1t,
0
o x x x Ideal b_ubb/ing
reg1me
~
□ x x
~ X X
ox
º"""'--=--'-------,l-,-----:'c----:cL.._---::'
O 0.1 0.2 0.3 0.4 0.5
Void fraction, a

Fig. 9.9 The cffect of inlet conditions on void fraction in bubbly

flow. (Zuber and Hench.27)
 ONE-DIMENSIONAL TWO-PHASE FLOW
"'
Initial liquid height, 12 in.
o Initial liquid height, 24 in.
"' Initial liquid height, 36 in.
- - - Ideal bubbly flow
[Eq. [9.33)]
0.3 - · · - - Fully developed
slug flow

/
·E 0.2
m
E
o
1e
/
<(

º""----____ L __ _ _ __¡__ _ _ __L__ _ _ _..J.__ __J

O 0.1 0.2 0.3 0.4

Mean void fraction, a -

Fig. 9.10 Agglomeration effects on mean void fraction when bubbling through differ-
ent dcpths of tap water_ (Wallis. 22 )

9.6 FRICTION AND MOMENTIJM FLUX IN BUBlllY flOW

Although the simple theory gives an adequatc description of vertical bub-
bly flow at low velocities, it is obvious that in general one cannot neglect
the effects of wall shear stress and momentum changes. Fortunately it
turns out that homogeneous theory is a good approximation in most cases
where friction and momentum effects are important 1 whereas at low
velocities, where homogeneous theory is inaccurate, the effects are usually
negligible.
The usual method of correlating frictional pressure drop is to define
a friction factor C¡ which is the ratio of the wall shear stress rw to the
dynarnic pressure o! the stream, Gj /2.

(9.46)
 BUBBLY FLOW
"'
This friction factor is thcn plotted versus the Reynolds number
GD
Re~- (9.47)
¡,

Bubbly mixtures in laminar flow are ncwtonian at Iow values of a and the
effective viscosity is given by Eq. (2.58)
µ ~ ¡,¡(l + a) (9.48)
If the liquid is contaminated the influence of a is cnhanced due to the
tendency of the bubbles to behave as solid spheres.
Unfortunately, Eq. (9.48) is valid only for values of a below abo_ut
0.05. At higher bubble concentrations the mixture rapidly becomes non-
newtonian, exhibiting a yield stress, 30 • 31 a decreasing apparent viscosity
with increasing shear rate, 32 and even electrical effccts. 33 Foams exhibit
considerable rigidity at high void fractions, and the bubbles behave like
the atoms in a crystal. Observations of the flow of frothy beer in plastic

0.9,---,----i---,-------,--,-------,---.--------,--r--,

t>Distilled water, flow increasing

+ Distilled water, flow decreosing
Impure distilled water, flow increasing
v Impure distilled water, f!ow decreasing
0 Top water, flow increasing
0.7
Top water, flow decreasing
6
Soap solution, flow increasing
◊ Soap solution, flow decreasing
J O.ó 1-D-e-,-e-lo_p_;,-g~---H-f--------ic_+-_ -
/

- Bubbly f low , Eq. (9.33)

slug flow / - · · · - Fully developed slug f/ow
-~
;o.s
"'u
'i
o
/
0.4 t - - - - - - - - 1 c - - - t - f t h - - - - - : - - - - - j - - - - - - j - - - - - - j - - - - - - j
o> region
L

<i' 0.3

0.2 ~-----t"f---t--;,L----------;:;,,f"-------:,=a~,.,__t-----+----------j

0.1
ly flow
--- -----
---¡-------+.:-'a.,.__
--------~
º~---::1c------::c1::-------:c½---::1-:----fc---f-:----:1::--f:----:1cc---~
O 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Mean void fraction, ot. -+-

fig. 9.11 Effect of liquid purity on void fraction in bubbly flow. (Wallis. 22)
 ONE-DIMENSIONAL TWO-PHASE FLOW

pipelines at college picnics, for example, show ali of the bubbles moving
at the same velocity, while shear strain is confined to a thin liquid film
on the tube wall.
In turbulent flow it is usually adequate to use the liquid viscosíty
in the Reynolds number and employ single-phase flow correlatíons. Up
to Reynolds numbers of 10 5 a good approximatíon is C1 ~ 0.005. For
example, Meyer 30 found that the data of Rose and Gríffith 34 scattered
randomly around this value with maximum deviations of 25 percent.
lVIomentum fluxes can also be estimated within about 20 percent accu-
racy by using homogeneous flow theory. If the component fluxes are
not very large compared wíth the dríft flux, a better estímate for ver-
tical flow is given by the solution to Prob. 4.1 as

.( G, + G, )
J 1 + fo/j, 1 - j,i/j,

Example 9.3 ·water and air flow upward in a vertical duct. The various values of
fl.uxes and densities are: Pi = 1 g/cm3, Pu = 0.002 g/cm 2, j¡ = 20 cm/sec,
Ju = 50 cm/sec. The drift flux is given by
cm/sec

What correction factor should multiply the homogeneous theory momentum

flux to get the correct value?
Solution The value of momentum flux predicted by homogcneous theory is jG.
Therefore the required correction factor is

Using Eq. (4.2) and the given values of j¡ and j 0 , the value of a is found to be
0.25, and the corresponding value of j 0 ¡ is 2.8 crn/sec. Thc values of the mass
fluxes are

G¡ = p¡j¡ = 20 g/(cm 2)(sec)

G/l = p¡¡Jo = 0.1 g/(cm 2)(sec)

The required correction factor is therefore

20 0.1
(20.1)(1 + 0.14) + (20.1)(1 - 0.056) ~ 0 878
-

9.7 THE 1/ELOCITY OF SOUND IN BUBBLY Mlll:TUllES

The velocíty of sound in a bubbly inixture can be calculated from the

equations derived in Chap. 6. For small bubbles with a density much
lower than the liquid, the homogeneous assumption is approximately valíd
and the appropríate equatíon is (6.110). Thís equatíon is the same as
SUBBLY FLOW

the equation given by \:Vood 35 and is

• 1
=~ (9.49)
e~ [ap -+-(1- -- -
0
-~---
a)p¡l[a/p,c,2 +-(1---a)/p¡c¡2]
--~
For an air-water mixture at 60ºF and an adiabatic process, c0 = 1117
fps, c1 - 4800 fps, p,/ p¡ - 0.0012, and for values of a greater than 10-a
a good approxima.tion is

(9.50)

The minimum value of e occurs at a = 0.5.

Because of transient heat-transfer effects which occur during the pas-
sage of a wave, it is not obvious which path one should use to calculate
thc quantity ap/op for thc gas. For a rapid compression and expansion
one would expect the gas to behave adiabatically, in which case

(9.51)

whereas, for slow changes one would expect approximately isothermal

behavior, i.e.,

(9.52)

The adiabatic result should be approached at high frequencies and the

isothermal prediction at low frequencies. 36 , 37
The predictions of Eqs. (9.51) and (9.52) for atmospheric pressure
and rnoderate frequencies are cornpared with sorne data in Fig. 9.12.
If the bubbles are small, their virtual compressibility is altered by
the cffects of surface tension. The sonic velocity is then obtained 38 by
multiplying Eq. (9.50) by the factor
2 -¡,
(1 + "
3pRb + 4cr)

Ne is the bubble radius and p the pressure. For very small bubbles this
factor approaches a limiting value of 0.82. Sorne resonance effects occur
at high lrequencies (above about 1 kc) which are comparable with the
natural frequency o! pulsation of the bubbles 39 given 50 by the expression
(3'"/p / p¡Rb 2) '/2_

9.8 THE LIMITS or THE BUBBLY FLOW REGIME

There is no lower limit to the void fraction at which bubbly flow can
occur in cocurrent two-phase flow. The bubbly pattern breaks down in
practice far one of two reasons.
 ONE-D!MENSIONAL TWO-PHASE flOW
"'
l. Coalescence of the bubbles either as they are being formed or as they
collide while flowing along the duct.
2. The characteristics of the process of injection or production of the gas
or vapor and its interaction with the flow dynamics of the channel.

Both of the above processes are rather whimsical. The speed with
which coalescence occurs is particularly sensitive to impurities, even in
minute quantities (see Fig. 9.11). V elocity gradients and turbulence tend
to increase the rate at which small bubbles collide, thus promoting agglom-
eration, but also have the effect of tearing apart the bigger bubbles. A
very approximate ''rule of thumbn which is sometimes used by engineers

300
• 1 kc
1 0.5 kc
(Ref. 36 acd 37)

1 o Extropolated to zero frequency

250 - - -Isothermal
1 Adiobotic
1
"
-2-
-á 200 JI
e
o
1
- ~
a 1 1
1

~,
;e
·¡¡ 150 1 1
.2
:!; º\ 1
,!\
I
l00 t\ I
/
o
' ~ ........

o ~~~.E:---'ó
/
/

o
50

oL _ _ L_ _j__j__ _ L _ _ L _ _ j _ _ L _ _ L_ _j__

O ~ Q4 Q6 0.8 1.0
Void fraction, a -

Fig. 9.12 Velocity of sound in a bubbly mixture of air and

water under atmospheric conditioru. 36 , 37
 BUBBLY FLOW 267

places the transition from bubbly to slug flow at 10 percent void fraction
for "pure" liquids although Rose 34 has reported bubbly flow at 60 percent
void fraction in tap water. At the other extreme there exist foamíng
agents which allow bubbly flow to persist up to virtually 100 pcrcent void
fraction.
The second mode of breakdown of bubbly flow often cannot be ana-
Jyzed without reference to the entire system charactcristics. For exam-
ple, in a boiler tube bubbly flow may be converted to slug flow by sup-
pressíng a sufficient number of nucleation sites so that the vapor is formed
as several large bubbles instead of numerous small ones. SimilarlyJ if
gas is introduced into a liquid stream through a porous surfaceJ there
exists a transition at the injector surface from the formation of small bub-
bles to the formation of a gas "blanket" which breaks off to form indi-
vidual slug-flow bubbles. 22 Only in certain special cases where the limit-
ing process can be identified (it might perhaps be a flooding phenomenon)
has rationalization of this transition been possible.
From an academic standpoint it should be possible to determine
exactly which bubble size distribution (if any) will be ultimately stable
in a very long channel far away from all inlets. However1 in the majority
of practica! cases the bubbly flow pattern never becomes "lully devel-
opedn and entrance effects predominate.

IE.xample 9.4 Air and water flow upward in a 10-ft-long, 2-in.-diameter vertical pipe
and discharge into an environment at 14.7 psia. Assuming bubbly flow anda
temperature of 70ºF, calculate the inlet pressure for the following volumetric
fluxes measured at atmospheric pressure and temperature,

j 1 , fps 0.5 1 10 15 30 32 34 36 38 40

Ju, fps 1 2 2 30 32 34 36 38 40

Solution The bubble size being unspecified we have a choice of the regimes in Table
9.1 and the corresponding values of nin Eq. (9.12). Assuming that no particular
precautions are taken to produce small bubbles, the most likely regirne is
probably churn-turbulent and the appropriate equation is (9.35). Substituting
values for air and water at 70ºF we find 1

V 11 ¡ = 0.82 fps (a)

Therefore j 11¡ = 0.82a (b)

Using this value in Eq. (4.5) and solving for ct we have

j,
ct = j¡ + Ju + VuJ (e)

Sin ce the pressure changes down the pipe, the val u e of j 11 will change. Assuming
 ONE-OIMENSIONAL TWO-PHASE FLOW
"'
isothermal expansion we obtain

. p, r. )
Ju = p Vil pa (d)

where pa, is the atmospheric pressure and (Ju)Pa the given gas flux.
A further complication will arise if the flow becomes sonic at the duct
exit. This corresponds to choking, and in this case the exit pressure is not
necessarily the same as the pressure of the surroundings.
The three components of pressure gradient are as follows.

Acceleration

dp) ~ G de, + G1 d,1 (e)

( dzA udz dz

·vu and v1 are found by using Eq. (e) in Eqs. (1.22) and (1.23)

V¡¡ = j¡ + Ju + Vgj (f)

• j¡ +Ju +vui (g)
V¡ = JI j¡ + Vuj

Only j¡¡ changes down the duct. Therefore,

dvu dju (h)
dz dz
dv 1 = dj 0 ~_j_
1~ (i)
dz dz j¡ + Vuf
Equation (e) therefore becomes, with the use of Eqs. (h) and (i),

- (dp)
dz A
~ (a, + G1 j¡_j_J)
+
dj,
dz Vui
(j)

If the interval is taken small enough, dj 0 /dz can be found by-differentiating

Eq. (d). Thus
dj, __ C ) p, dp (k)
dz - Ju pa p 2 dz

Combining Eqs. (j) and (k),

dp) =
( ~d
z A
(a U + G f Ji~+.
Íf
V 11 1
)
Jupa 2P"
(. ) dp
p d-Z
(l)

Friction

(m)

The appropriate values are C 1 = 0.005, D = 3-,i ft,

G = Péu + PtÚ (n)

j = j¡ }u + (o)

Gravity

- (dp)
dz a
~ g[ap, + (] - a)p1] (p)
BUBBLY FLOW
"'
íable 9.2 The components of pressure drop and in[et and
exit pressures predicted by the solution to Example 9.4

j 1, fps ir11 fps dPaccei, psi dpfric, psi dpgrav 1 psi Pe, psi Pi, psi

0.5 1 0.0004 0.003 2 546 14.7 17.25

1 2 O. 0019 0.016 2. 137 14. 7 16 85
10 2 0.061 0.937 ;J 736 14.7 19.43
15 10 0.568 2.78 2.84 14 7 20.89
30 30 6.96 11.32 2.81 14.7 35. 79
32 32 8. 32 12.61 2.87 14 .7 38.49
34 34 9.30 13. 92 2.93 15 2 41.36
36 36 9.83 15 29 2.99 16.2 44.31
38 38 10. 36 16. 72 3 05 17.2 47.33
40 40 11.39 18.21 3.10 17.7 50.40

Combining Eqs. (l), (m), and (p) and solving far dp/dz givcs the usual form of
result
2C¡Gj/D + g[ap, + (1 - a)p¡]
(q)

The second term in the denominator of Eq. (q) has the same significancc
as the square of the J\.fach nurnbcr in Eq. (2.4.5). If this term exceeds unity,
the acceleration pressure drop becomes negative and the fio,v must be supcr-
sonic. Normally this will not be permissible and the exit pressure will adjust
until choking is reached with a Mach number of unity. The condition for
choking is therefore

(r)

where P• is the exit pressure.

The method of solution is now as follows. The pipe is divided into
intervals and the pressure is calculated by working back from the pipe exit.
First Eq. (r) is used to calculate the exit pressure which would produce choking.
If this is less than Pa, the exit pressure is atmosphcric, whereas, if it exceeds
Pa, the exit is choked and the exit prcssure follows from Eq. (r). a is calculated
from Eq. (e) and -dp/dz from Eq. (q). Stepping back up the pipe new values
of p, a, j 111 andj are fÜund and the procedure is repcated until the inlet is reached.
This problem was solved on the Dartmouth College computer by Gary
Grulich and Andrew Porteous. The results are shown in Table 9.2. Note
how the gravitational pressure drop dominates at low flow rates while accelera-
tion and friction take over at the higher flow rates. Pi is the inlet pressure.

9.9 ISOTHERMAL HOMOGENEOIJS FLOW OF GAS-

UQIJID MIXTURES IN STRAIGHT PIPES

Rathcr than resorting to numerical methods, as in Example 9.4) it is often

convenient to have explicit analytical expressions for the dependent var-
270 ONE-DIMENSIONAL TWO-PHASE FLOW

iables. These equations can be derived by using homogeneous theory and

treating the bubbly mixture as a pseudo gas with appropriate properties.
Fortunately, since bubbly flow <loes not exist at high void fractions
and the liquid density is usually much higher than the gas density, the
liquid makes up almost ali of the mass flow rate and the resulting flow is
essentially isothermal. Thus, although it is interesting academically to
develop exact equations for the general case, the assumption of constant
temperature is usually quite adequate.
At this stage we shall neglect the effects of phase change, remember-
ing that this will complicatc matters under sorne circumstances.
In the usual way, the pressure gradient in a straight pipe atan angle
0 to the vertical is then

dp - 2C¡jG/D + (G/j)g cose (9.53)

dz - 1 - j,G/p
The isothermal lVIach number is
M2 = j,G (9.54)
p
Denote conditions at a Mach number of unity by tbe * superscript.
From Eq. (9.54), therefore,
(9.55)

The isothermal expansion law for the gas if therc is no area change, is
1

(9.56)

Combining Eqs. (9.54) to (9.56) we find

p* (9.57)
P= M
j, = Mj; (9.58)
Differentiation o! Eq. (9.57) yields
1 dp 1 dJlI (9.59)
pdz=-iVIdz
Substituting for ali the variables j, p, j,, dp/dz in Eq. (9.53) in terms of
the Mach number we eventually find
dM 1 - M 2 = 2C¡ ó*M +1+ acose (9.60)
dz M' D o* j¡ 2ó*(l +Mó*)
where ó is defined as j,/j¡ (Example 6.7) and
·•
ó* = J_r; (9.61)
]¡
BJBBLY FLOW 271

Equation (9.60) shows that friction and gravity in upf!ow act to drive the
Mach number toward unity since dJlf / dz is positivo wben M < 1 and
negative when 111 > 1. In downflow, however, there is apparently an
interaction between friction and gravity which can lead to a smooth tran-
sition through M = 1 when

2C¡ (l + o*)' = -g _cos 0 (9.62)

D Jt'
Since, under these conditions, the right-hand side of Eq. (9.60) is positivo
for M > 1 and negative for M < 1, thc flow must go from supcrsonic to
subsonic and the l\,íach number decreases continuously.
Equation (9.60) can be integrated to give a rather long explicit equa-
tion for the variation of lVIach number with distance. The pressure and
gas flow rate everywhere are then found from Eqs. (9.57) and (9.58).
In horizontal flow the integration is simpler and results in the "Fauno
line" equation for a bubbly mixture. Arranging in partial fractions we
get
2C¡ [ o* 0* 2 o*(l - o*')]
D dz = dM 111 2 - M - 1 + Mo* (9.63)

which integrates to give

2C¡ z*;; z= o* (}1 - 1) +0* ln M - 2

(1 - 0* 2 ) ln
1
++¡¡;;:.
1

(9.64)
and is compatible with the results of Huey. 'º• 41

9.!0 ISOTHERlll!AL HOMOGENEOUS FlOW WITH AREA CHANGE ONLY

If only the momentum terms are significant in the equation of motion

we have
_ dp = Gdj (9.65)
dz dz
N ow, in most cases in which bubbly flow occurs almost al! of the
mass flux is dueto the liquid and we may write

G = p¡a = 1 P~ o (9.56)

Furthermore, for isothermal flow from stagnation conditions 00, Po,

(9.67)
m ONE-DI MENSIONAL TWO-PHASE FLOW

Substituting Eq. (9.67) into Eq. (9.66) and using the rcsult in Eq. (9.55),
we obtain

dp (
-dz 1 + ºº -Po)
p = . dj
p¡J -
dz
(9.68)

which may be integrated from the stagnation conditions to give an explicit

expression for the velocity in terms of the pressU:re ratio, as shown by
Tangren, Dodge 1 and Seifert, 42

p¡j'
2po = l -
(p) Po
p
- 00 ln Po
(9.69)

Equations (9.66) and (9.67) can also be used in Eq. (9.54) to show that

M' = p¡J·2 ( 1 + _J)__ ) - 2 (9.70)

PoÓo PoÓo

Eliminating j2 between Eqs. (9.69) and (9.70) gives an equation for the
l\tlach number in terms of the pressure ratio, as follows:

M' = -2 ( 1 - -p - ó0 ln-
Óo
p) ( 1
Po Po
+ -PoÓo
p )-' (9. 71)

When the Mach number is unity, p = p* and the ratio p* /p, can be
found from Eq. (9.71). Sorne results are given in Table 9.3.
Knowing the pressure ratio as a function of l\-fach number, other
useful variables follow from relationships which are easily derived from
the condition of isothermal flow, such as
j, ó A* p* A* (9.72)
M=-=-~=-~
ó* A p Aji
Figure 9.13 shows the resulting dependence of Mach number, pressure
ratio, velocity ratio, and density ratio as a function of nozzle geometry.
The throat area is found in terms of the upstream conditions from
Eq. (9.54) with M = l.

1 = j,G* Q*W (9. 73)

p = A !2p*

Table 9.3 Critica! pressure ratio for bubbly fiow in a nozzle

as a function of stagnatlon gas-liquid ratio and void fraction 42

0.67 0.5 0.333 0.2 0.091 0.05

2 1 0.5 0.25 0.1 0.05
0.55 0.52 0.46 0.39 0.30 0.25
BUBBLY FLOW m
3.6 3.6

3.2 3.2

2.8 2.8

2.4 2.4
80 = O.OS
80 = 0.10
¼ 2.0 2.0
80 = 0.25 •-}
~
:¡;
80 =0.50
~
1.6 1. 6

1.2 1.2

0.8 0.8

0.4

o -=----'-----'------'-1-_l__J__j__j__J
O 0.2 0.6 1.0 0.6 0.2 O
f..-- Convergent -------4,,--- Divergent --------1
Throot_l A*/A
lal

1.2
•-}
"'0.8 0.8

0.4 0.4

o L__j____l__j___L_L__j__j__j__j----"' o "'--_j__j__j___l_~_j__j__j__j~
O 0.2 0.6 1.0 0.6 0.2 O O Q2 Q6 1.0 Q6 0.2 O
r-- Convergent -----4----- Divergent ~ 1-,--------Convergent·-----4- Divergent------..¡
fhroot_J A*/A Throot_J A*/A
(e) (d)

Fig. 9.13 Nozzle characterist.ics for isothermal frictionless bubbly flow. (Tangren,
Dodge, and Seiferl. 42 )
 m ONE-DIMENSIONAL TWO-PHASE FLOW

whence
A* = ]'_o
p*
[(Q,),W]"
Po
(9.74)

The qualitative behavior o! bubbly nozzle flows and the rnethods for
solving practica! problems are very similar to the analogous single-phase
flow cases discussed by Shapiro. 43

USE OF THE EQUATIONS OF MOTiON FOR BOTH COMPONENTS

Although the assumption o! hornogeneous flow is adequate for rnost pur-

poses it can be erroneous when a body force or accelerational field causes
significant relative velocity or u slip" between the components. We have
already seen how the drift velocity rnodel can account for slip in low veloc-
ity vertical :flow. When acceleration is significant (it can be very man y
g's in nozzles), it is necessary to consider the separate equations of motion
of the components, Eqs. (3.45) and (3.46).
Unless the J1s are substantial it is evident that a given pressure gra-
dient will accelerate the gas far more rapidly than the liquid and cause a
violation of the homogeneous fl.ow assumption. The f' s are composed of
at least three parts: drag forces, apparent mass effects, and the Basset
force. The way in which these forces interact has yet to be resolved in
detail. The apparent mass effects irnpose an upper limit of about 2 on
the slip ratio vr;/v1 , as shown in Example 8.5. Far example, practical
maximum slip ratios in high-speed nozzle flows measured by l\1uir and
Eichhorn 44 were in the range 1.1 to 1.8.

9.ll SHOCK WAVES

Eddington 45 found that both normal and oblique shock waves in bubbly
flow could be described by the isothermal homogeneous theory presented
in Sec. 6.6. The normal shock relations are simply, from Eq. (6.175),

Jl11' = _1_ = (j,)i = p, = ó, (9.75)

M, 2 (j,), p, ó,
Shock-wave thicknesses are governed by the relaxation phenomena
discussed in Chap. 8. Equilibration proceeds much rnore rapidly than
in gas-particle fl.ows. Typical measured thicknesses are about 1 in .

... PROBLEMS
9.1. An air bubble is formed very slowly in water at 20ºC. What is its diameter
when it detaches if (a) it is formed on upward-facing nonwettable orífices with radii
0.1, 0.5, 1, and 2 mm; (b) the orífice is a hale in a wettable surface with contact angle
f3 and f3 can be 30, 60, 90, or 120º?
 BU BBLY FLOW 275

~.2. Estimate the equilibrium bubble size in a milk-shake machine ,vhich contains
one liter of fluid and is run b::,r a ;!,,S:th-horsepower motor.
!U. By considering apparent mass effects show that the equation of motion of a
spherical bubble growing freely in an inviscid liquid is 4

Vb(p¡ - Pu)g = ~ [ 'Db(P¡¡ + :,]p¡) ~J

Explain why the fact that the bubble is growing introduces an additional
component into the force f¡¡•
9.4. If the bubble in Prob. 9.3 grows according to the equation 'th = Atª where A
anda are constants, and p¡¡ << PI, show that the distance its ccnter has gone after time
t is
l
z - ---gt'
a+l

9.5. If the bubble in Prob. 9.4 is grown from a nozzle, which exerts no force on it,
and breaks away when z = Rb, shO\v that the bubblc volume at departure is

- [(4~ª)" (ª- +- 1)'ºAº ]'/(o-o)

'l.\ - ~
g ~

and that this agrees with Eq. (9.5) when a = 1. 'iiVhat happens if a = 6?
9.6. When bubbles grow as a result of transient heat or mass transfer, their radii
increase in proportion to the square root of the time. A bubble is nucleated in
a superheated liquid by passing a current pulse through a hot wire. The bubble
grows according to the equation Rb = 2t½ where Rb is in ccntimeters and t in seconds.
What is the bubble volume when it breaks from the wire?
9.7. Solve Probs. 9.3 and 9.4 for a bubble growing in a very viscous liquid. Assume
that the bubble always moves with its terminal velocity. Show that Eq. (9.6) results
ifa=l.
!U!. What is the maximum bubble volume 1vhich can be bfown from a given orifice
in a viscous liquid if Eq. (9.7) is va.lid? Evaluate this volumc for air blowing through
a ¼-in. nozzle into molasses (p¡ ::::: 1.2 g/cm 3, µ 1 = 1000 poise).
9.9. Solve Example 9.1 for Rb = 0.005, 0.05, 0.1, 0.3, and 1.0 cm.
9.10. Solve Example 9.1 and Prob. 9.9 if the fluid is aniline for ,vhich Pt = 63.7 lb/fV,
µ 1 = 2.93 cp, and cr = 41.7 dynes/cm.
9.11. A bubble is nucleated in a 100-ft-high column of water which is supersaturated
with carbon dioxide and open at the top. According to Caldcrbank and J\foo-Young, 49
the mass-transfer coefficiént for rising bubbles is independent of diameter so that the
rate of gas evolution per unit surface areu is constant. If this ratc of evolution is
10- 4 g/(cm 2 ) (sec) estimate the bubble volume and rise velocity at different points in
the column if it starts from the bottom ,vith a diameter of 10 µ. Ignore evaporation
into the bubble, but do not neglect hydrostatic pressure changes or surface-tension
effects.
!U2. Nicklin 46 measured the drift velocity, vah of air bubbles as a function of gas flux
through stagnant water. His results are shown in Fig. 9.14. Derive thc relationship
between j 0 ¡ anda, the values of v00 and n, and explain the cxistence of the minimum
value of Vai shown in the figure.
9.13. Marrucci and Gioia 47 studied cocurrent air-water flow in a vertical pipe of 5.3-cm
diam. Their results are shown in Fig. 9.1.5a. Evaluate n, v..,, v11i, and j 01 as a func-
 ONE~DIMENSIONAL TWO-PHASE FLOW
"' 0.9 o
o o
' ' '
1 1 1

o
o. o
o
-~ 0.8
.• o o
o
oo
o

ru ,o o
::o O. 7 r o R -
_o
o
o
'lo e¿, o
_o
0ºº
ru
{'
0.6
o '!,
.:~i o
- o
o
o~ Li
ºe o
o
r -

0.4 ' '

r
' '
0.01 0.02 0.05 0.1 0.2 0.4
Gas flux,J~o, fps
(o}
0.9 i , I 1

o
' ' ' ' '

-
0.8 r o o
o. o o o
o
~ 0.7 ºSo
o
o
.~u 0.6 r o .
o o
-¡¡ o
> o
' o o
o o
o
o o
0.4 r .
ºo o
o
0.3 '
1

' 1 1 1 .

0.01 0.02 0.05 0.1 0.2 0.4

Gas flux,;~ 0 , fps
(b}

Fig. !U4 (a) Bubble velocity and (b) drift velocity for bubbling
of gas through stagnant liquid. (,.Vicklin. 46 )

tion of a from these data. VVhat regime of bubbly flow do you diagnose? The same
authors plotted Vy; versus a as shown in Fig. 9.15b. What values of v"' and n are
predicted from this method of plotting? Do the conclusions from the two figures
agree?
9.14. Air and benzene (a = 28.8 dynes/cm, Pf = 54.7 lb/ft3, µ,¡ = 0.647 cp) flow
vertically upward in a l-in.-diam pipe at flow ratcs of 1 and 1000 lb/hr, respectively.
The bubble diameter is 2 mm at the bottom of the pipe where the pressure is .50 psia
and the temperature lSºC. Evaluate the void fraction and prcssure as a function of
height. I-Iow long a pipe is needed to drop the prcssurc to atmospheric?
9.15. Solve Prob. 9.14 if the flow rates are each increased by a factor of 10.
9.16. A series of air bubbles with Ro = 0.5 in. flow upward in stationary fluid in a
vertical pipo of 2 in. diam. What is their velocity, (a) in glycerin at 20ºC and (b) in
SUBBLY FLOW 277

water at 20ºC? What is the gas flow rate in each case if the bubbles are 3 in. apart
from nose to nose?
iU7. An air flux of 3 cm/sec is suddenly supplied to the bottom of a 50-cm depth of
clear water containing a strong foaming agent in a long vertical tube. The bubbles
which are produced ali have a volume of 0.002 cm 3. Show that Fig. 9.16 describes
the qualitative behavior. Identify thc various features of thc sketehes and evaluate
the key parameters in detail. \Vhich parts of the process, if any, are likely to be
physically impossible or unstable.
9.18. What is the velocity of sound in a hydrogen-water mixture at 1000 psia, 70ºF,
and with mean density 40 lb /ftª?
9.19. Salve Prob. 9.18 if the pressure is 5 psia and the radius of the bubblcs is 1 mm.
9.20. Salve Example 9.4 if flow is horizontal.
9.21. 3 scfm of air are to be bubbled through a tank of 1 ft diam containing 20 gal of
water at 20ºC. The designer has a choice of the number of nozzles or orífices to use
as well as diameters. Show how the following parameters depcnd on the design of
the injection system: (a) void fraction, (b) bubble residence time in the liquid, (e)
total interfacial area in the tank, and (d) operating limits.
9.22. In a 4-in.-diam countercurrent-flow bubble column, Shulman and Molstad 23
found that flooding occurred for the following mass fluxes of carbon dioxide and water

t
~11 0.2 0 9 =1.68 liters/min

09 ""'- 0.52 liters/min

o c__ _ ___,__ _ _ _ _L__ _ _~

O 5 lO 15
Qf, liters/min
lal

<E~ 2s~----------~-----7
u o Measured values

¡:-
·e
il 1sl------l-----"'--...c--cc--l---------1
>
~

""Q 10'--------"------~-----"
o 0.1 0.2 0.3

lb)
Fig. 9.15 Sorne results of Marrucci and Gioia for
vertical bubbly fl.ow. 4 1
 ONE-D!MENSIONAL TWO-PHASE FLOW

i
h

Fig. 9.16 Transient bubbling behavior of the system described

in Prob. 9.17.

(in pounds per hour per square foot):

- w, 1-1_5_,o_o_o-+_1_0_,o_o_o~¡ sooo _
w, 38 42 ~

Compare these results with the theoretical values.

9.23. Derive Eq. (9.72).
9.24. Show that, in isothermal frictionless homogeneous nozzle flow,

_ _!e dA
A.dz
[___!!'
1-M
_ (l + -1)]
2 ó
1 dM' M'(l + 1/o)
- M' d, 2(1 + M'/o)
and that these results may be combined ,vith Eq. (9.72) to forma basis for numerical
prediction of all the variables in terms of ó* and M.
9.25. In a choked nozzle of given geometry, with given upstream stagnation pressure,
how <loes the liquid flow rate depend on the val u e of 00 ? Deduce the relationship
beteeen W 1 and Wu for the valve described in Prob. 2.21. Design such a valve if it is
to control oil flow rates between 1 and 10 tons/hr of oil with density 60 lb/ft3 at 60ºF
and 100 psia using compressed air.
BUBBLY FLOW
"'
9.26. By assuming that the ratio (1 - 01.)/C is constant through a nozzle, ,vhere C is
the coefficient of apparent mass, sho,v that the limiting value of the slip ratio in an
accelerating bubbly flow is
2(1 - a)
VC' +4C(l - a) - C
If C has the usual value of H, show how the velocity ratio depends on the
average void fraction.
If the bubbles distort to form ellipsoids with axes in the ratio 1 : 6 1 the value of
C drops to 0.045. What effect does this have on the velocity ratio?
9.27. Show that the velocity ratio across a normal shock wave in a bubbly mixture is
given 1 in tcrms of the upstream conditions, by

9.28. 100 cfm of water and 200 scfm of air at 100 psi and 70"F are supplied toan experi-
mental convergent-divergent nozzle. What is the throat area when the flow is
choked? If the nozzle discharges into the atmosphere at 14.7 psia, what values of
exit area will cause shocks to form in thc nozzle? Plot the pressure variation down
the nozzle as a function of area for various values of the cxit area. What is the
maximum thrust which the nozzle can exert on its mountings?
9.29. Discuss the various forces which act on the bubbles in one-dimensional flow.
Under what conditions is the flow governed by a balance between (a) gravity and
drag, (b) inertia and pressure forces, (e) pressure and the Basset force 1 and (d) surface
tension and buoyancy?
9.30, Show that an error is introduced by assuming ~ to be small in Eqs. (9.39) and
(9.40), which tends to underestimate the duration of stage l. Show that stage 1 in
Example 9.2 actually ends when t = 30 sec when an: amount 3.87 cm 3 of liquid has
drained from the foam.
9.31. A bottle of beer is shaken vigorously and poured into a long straight vertical
vessel (a "yard"). If the initial void fraction is 80% and no bubbles form or burst
after the first few seconds, describe what happens.
9.32. Salve Prob. 9.31 if the initial void fraction is 30%.
9.33. According to Harrison and Leung 1 48 bubbles formed at orífices in fluidized beds
abey Eq. (9.5). Estímate the void fraction as a function of the number of nozzles
when a bed of glass balls 1 0.01 cm diam 1 is fluidized with air at a volumetric flux of
3 cm/sec, measured at atmospheric pressure and temperature. The bed is 10 ft deep.
Assume that all the air flow in excess of that needed to fluidize the bed forms bubbles.
State other assumptions ·made.

REFERENCES

l. Kutateladze 1 S. S., and M. A. Styrikovich: uHydraulics of Gas-Liquid Systems/ 1

Moscow, \Vright Field trans. F-TS-9814/V, 1958.
2. Jackson, R.: Chem. Eng., vol. 42 1 pp. 107-118, May 1 1964.
3. Siemes, W., andJ. F. Kaufmann: Chem. Eng. &i., vol. 5, pp. 127-139, June, 1956.
4. Davidson, J. K., and D. Harrison: "Fluidised Particles," Cambridge Univcrsity
Press, London, 1963.
5. Davidson1 J. F., and B. O. G. Schüler: 1'rans. Inst. Chem. Engrs. 1 vol. 38, pp.
144-154 and 335-3421 1960.
 ONE-DIMENSlONAl TWO-PHASE FLOW

6. Taylor, G. l.: Proc. Roy. Soc. (London), vol. A210, p. 192, 1950.
7. Zubcr, N.: AEC Rept. U-4439, 1959.
8. Fritz, W.: Physik. Z., vol. 36, p. 623, 1933.
9. Hinze, J. O.: A. l. Ch. E. J., vol. 1, p. 289, 1955.
10. Pecbles, F. N., and H. J. Garbcr: Chem. Eng. Progr., vol. 49, pp. 88-97, 1953.
11. Haberman, W. L. 1 and R. K. l\!Iorton: David 1iV. Taylor ?v1odel Basin Rcpt. 802, 1953.
12. Stokes, G. G.: "Mathematical and Physical Papers," vol. 1, Cambridge Univcr-
sity Press, London, 1880.
13. Hadamard, J.: Compt. Rend. Acad. Sci. Paris, vol. 152, pp. 1735-1738, 1911.
14. Rybczynski, VV.: Bull. Acad. Sci. Cracovie 1 vol. A, pp. 40-46, 1911.
15. Davies, R. M., and G. I. Taylor: Proc. Roy. Soc. (London), vol. 200, ser. A, pp.
375-390, 1950.
16. Harmathy, T. Z.: A. l. Ch. E. J., vol. 6, p. 28li 1960.
17. Collins, R.: J. Fluid 111ech., vol. 28, part 1, pp. 97-112 1 1967.
18. Ladenburg, R: Ann. Physik, vol. 23, p. 447, 1907.
19. Edgar, C. B., Jr.: AEC Rept. No. NYO-3114-14 by G. B. Vilallis, pp. 19-21, 1966.
20. Jameson, G. J.: Chem. Eng. Sci., vol. 21, pp. 35-48, 1966.
21. Jameson, G. J., and J. F. Davidson: Chem. Eng. Sci., vol. 21, pp. 29-34, 1966.
22. Wallis, G. B.: Paper no. 38, Intern. Heat Transfer Conf., Boulder, Colo., ASME,
1961.
23. Shulman, H. L., and M.C. Molstad: Ind. Eng. Chem., vol. 42, p. 1058, 19.50.
24. Kutateladze, S. S., and V. N. l\foskvicheva: Zh. Tech. Fiz., vol. 29, no. 9, pp.
1135-1139, 1959.
2.5. Gaylor, R., N. W. Roberts, and H. R. C. Pratt: Trans. Inst. Chem. Engrs., vol.
31, p. 57, 1953.
26. 1\1.iles, G. D., L. Shedlovsky, and J. Ross: J. Phys. Chem., vol. 49, p. 93, 1943.
27. Zuber, H., and J. Hench: Rept. no. 62GL100, General Electric Company,
Schenectady, N.Y., 1962.
28. Zuber, N., and J. A. Findlay: 1'rans. ASME J. Heat 1'ransfer, vol. 87, ser. C,
p. 453, 1965.
29. Radovcich, N.A., and R. Moissis: Rept. no. 7-7673-22, Department of Mechanical
Engineering, Massachusetts Institute of Technology, 1962.
30. l\Ieyer, P. K, and G. B. Wallis: AEC Rept. No. NYO-3114-12 (EURAEC 1530),
1965.
31. Penny, W. G., and 1\1. Blackman: Note 282, Ministry of Home Security, Great
Britain, 1943. See also J. Herma.ns, "Flow Properties of Disperse Systems,"
North-Holland Publishing Company, Amsterdam, 1953.
32. Sibree, T. O.: Tcons. Facoday Soc., vol. 31, p. 325, 1943.
33. Raza, S. H., and S. S. Marsden: Soc. Pet. Eng,,s. J., pp. 359-368, December, 1967.
34. Rose, S. C., Jr., and P. Griffith: Rept. /5003-30, Massachusctts Institutc of
Technology, 1964.
35. Wood, A. B.: "A Textbook of Sound," Thc Macmillan Company, New York,
p. 327, 1930.
36. Karplus, H. B.: Rept. C00-248, Armour Res. Found., June, 1958.
37. Gouse, S. W., Jr., and G. A. Brown: E.P.L. Rept. DSR 8040-1, Massachusetts
Institute of Technology, April, 1963.
38. Marcha!, R.: Compt. Rend. Acad. Sci., vol. 254, pp. 2524-2526, 1962.
39. Silberman, E.: J. Acoust. Soc. Am., vol. 29, no. 8, pp. 925-933, 19;'57.
40. Huey, C. T.i and R. A. A. Bryant: A. l. Ch. E. J., vol. 13, no. 1, pp. 70-76, 1967.
41. Huey, C. T.: Can. J. Chem. Eng., vol. 44, no. 6, pp. 313-321, December, 1966.
42. Tangren, R. F., C. H. Dodge, and H. S. Seifert: J. Appl. Phys., vol. 20, no. 7,
pp. 637-645, 1942.
BUBBLY FLOW

43. Shapiro, A. H.: "Dynamics and Thermodyna.mics of Compressible Fluid Flow,"

The Ronald Press Company, New York, 1953.
44. Muir, J. F., and R. Eichhorn: JSME Semi-intern. Syrnp., Tokyo, pp. 81-92,
September, 1967.
45. Eddington, R. B.: AIAA paper 66-87, 1966. Also Rept. 32-1096, .Jet Propulsion
Laboratories, California Institute of Technology, 1967.
46. Nicklin, D. J.: Chem. Eng. Sci., vol. 17, pp. 693-702, 1962.
47. Marrucci, G., and F. Gioia: Chim. Ind., vol. 45, no. 10, pp. 1205-1211, ]\tlilan,
Italy, 1963.
48. Harrison, D., and L. S. Leung: Trans. Inst. Chern. Engrs., vol. 39, p. 409, 1961.
49. Calderbank, P. H., and J'vI. B. Moo-Young: Chem. Eng. Sci., vol. 16, pp. :-39-54,
1961.
50. Foster, J. M., J. A. Botts, A. R. Barbin, and R. I. Vachon: Trans. A.S}VJE J. Basic
Eng., ser. D, vol. 90, pp. 125-132, March, 1968.
 10
Slug Flow

10.1 INTIIOlll.lCTION
The slug-fiow regime is characterized by a series of individual large bub-
bles wbich almost fill the available flow cross section. Sorne familiar
examples of slug flow are
l. Flow in a drinking straw when the glass is almost empty
2. Flow in the "riser" section of a coffee percolator
3. Flow in the neck of a bottle which is being emptied too rapidly
A typical example of a slug-flow bubble in a vertical pipe under
laboratory conditions is shown in Fig. 10.1.

10.2 GENERAL THEOII\'

BUBBLE DYNAMICS

Severa! of the overall properties of the slug-flow regime can be predicted

if one can describe the dynamics of a typical individual bubble. This
282
SLUG FLOW

Fig. 10.1 A slug-flow bubble rising in glycerin in a vertical

tube.

is perhaps best done by considering a unit cell consisting of one bubble

and part of the liquid slugs on each side of it as shown in Fig. 10.2. For
a specified total volumetric flux j the mean velocity of the liquid in the
slug is then simply

j = Q, + Q¡ (10.1)
A
The bubble dynamics are then determined by this velocity, the cor-
responding velocity profile, the bubble Iength, pipe geometry, and fluid
properties. Apart from the effects of the wake from the preceding bub-
ble, the velocity profile in the liquid slug is a function of the pipe rough-
ness and the Reynolds number,

Re;= jDp¡ (10.2)

µ¡

Therefore the bubble dynamics are dependent on j but not on the indi-
vidual fluxes j 1 and j, of the liquid and the gas. Furthermore, as long
as each unit cell corresponding to Figure 10.2 is independent, the bubble
dynamics are not a function of the void fraction a.

BUBBLE VELOCITY

In accordance with the above paragraph the bubble velocity is a function

of the average volumetric flux j, the pipe geometry, the fluid properties,
and the body force field. In almost every case the bubble length is not
284 ONE-DIMENSIONAL TWO-PHASE FLOW

fig. 10.2 "Unit cell" for slug-flow analysis.

found to be an important varjable since the dynamics of the nose and

tail of the bubble govern the motion entirely.
The drift velocity of the gas is the bubble velocity minus the overall
volumetric fhuc Therefore
(10.3)
Hence the drift velocity is also independent of void fraction and depends
on .i and not on j¡ and .i, independently.

VOJD FRACTION

The void fraction-or mean volumetric gas concentration-can be derived

from Eq. (1.10) if the bubble velocity and the gas flux are known. Thus

a=-
J, (10.4)
v,

PRESSURE DROP

The pressure drop is conveniently divided into three parts:

l. Pressure drop in the liquid slug

2. Pressure drop around the ends of the bubble
3. Pressure drop along the body of the bubble
SLUG FLOW
"'
Usually the gas viscosity and density are much lower than the liqnid
viscosity and density. In this case the gas in the bubble is substantially
at constant pressure. The body o! the bubble is approximately cylindri-
cal and the gas-liquid interface has a constant curvature. Therefore
there is no pressure drop down the body o! the bubble and item 3 above
is zero.
The pressure drop, item 1, in the liquid slug can be calculated by
single-component flow techniques.
The pressnre drop, item 2, around the ends o! the bubble remains
to be predicted in terms o! fundamental quantities.

10.3 VERTICAL SLUG FLOW

RISE VELOCITY OF SINGLE BUBBLES IN STAGNANT LIQUID

A bubble rises through a denser liquid because o! its buoyancy. The

velocity "~ with which a single bubble rises through stagnant liquid is
governed by the interaction between buoyancy and the other forces acting
on the bubble as a result o! its shape and motion. If the viscosity o!
the gas or vapor in the bubble is negligible, the only three forces besides
buoyancy which are important are those from liquid inertia, liquid viscos-
ity, and surface tension. The balance between buoyancy and these three
forces may be expressed in terms of three dimensionless groups

Dg(p¡ - P,) D'g(p¡ - P,)

where D is a characteristic dimension of the duct cross section. The
general solution to the problem is a function of these three parameters
which may be combined to genera te new dimensionless quantities if this
is convenient. As long as the bubble length is greater than the tube
diameter it does not influence the rise velocity. Alternatively, from Eqs.
(9.26) and (9.29), the bubble equivalent diameter must exceed 60 percent
of the tube diameter.
The simplest sOlutions are obtained when only one dimensionless
group governs the motion. These limiting cases are as follows:

lnertia dorninant VVhen viscosity and surface tension can be neglected

the bubble rise velocity is given solely in terms of the first dimensionless
group. Thus

(10.5)

Approximate analytical solutions to this problem for a cylindrical duct

have been obtained by Durnitrescu 1 and by Davies and Taylor. 2 Values
 ONE-DlMENSIONAL TWO-PHASE FLOW

of the constant which were obtained for round tubes by these authors
were:
Dumitrescu 1 k, = 0.351 (10.6)
Davies and Taylor' k, = 0.328 (10. 7)

Dumitrescu' s experiments gave a slightly different result

k, = 0.346
and this is close to the value (0.345) obtained in a series of experiments
by White and Beardmore. 8 The preferred value of le, is therefore
k, = 0.345 (10.8)
If one measures the rise velocity of a bubble in a tube with an open
top there is an apparent dependence on bubble length due to the expan-
sion of the gas as it rises in the hydrostatic pressure gradient. This
expansion leads to a value of j ahead of the bubble which is not zero, and
the rise velocity is augmented by the velocity of this liquid. By timing
bubbles of different lengths in tubes which were closed at the top and
did not allow any motion of the liquid ahead of the bubble, Nicklin 4 has
shown that the bubble rise velocity relative to stationary liquid is indeed
given by Eqs. (10.5) and (10.8) and is independent of the bubble length.
The case of tubes of noncircular cross section was investigated by
Griffith. 5 For rectangular channels of dimensions D, by D, the larger
dimension Db was found to be the more significant one for insertion into
Eq. (10.5). The constant k, can then be expressed as a function of
D,/D, anda good approximation to the final result is

le, = 0.23 + 0.13 ~: (10.9)

The limiting case in which D,/D, goes to zero agrees with Birkhoff and
Carter's' theoretical predictions for plane two-dimensional bubbles.
Griffith's paper also gives results for bubbles rising in annuli and
tube bundles. For annuli the outer diameter D, is the significant dimen-
sion for insertion into Eq. (10.5). The variation of k, as a function of
the ratio of the interna! to externa! diameters (DJD,) is shown in Fig.
10.3. A remarkable conclusion is that the smaller the annular spacing
is made, the !aster the bubble will rise.
In the limit D, = D, the result is approximately
(10.10)
which would be expected from Eq. (10.9) for a plane bubble which is
wrapped around the annulus so that the liquid streams coalesce on one
side. Symmetrical bubbles are usually not observed in annuli.
SLUG FLOW m
0.6 ~---r~-,---,1-,----,----<>,-,----,-~

Tube bundles---.._

Annuli

I
0.2 f - - - - - + - - - - - f - - - - - - j

o~~~-~~~-~~~-~~
o 0.2 0.4 0.6 0.8 1.0
Dimension ratio D¡"/Do far annuli
Fig. 10.3 Dependence of the fac- [l -Dh/D 0] fortube bundles
tor k1 on duct geometry. Ds/Dh for rectangles
( Gciffith. ') (Dh = hydraulic diameter)

In tube bundles the characteristic dimension is more likely to be the

overall housing diameter D 0 rather than the dimension of individual tubes.
Experimental evidence for slug-flow bubbles in tube bundles is contained
in Griffith's paper. 5
The reader is reminded that the above equations are valid only
when surface tension and viscous forces are negligible. The limits of
validity of this regime will be discussed later.

Viscosity dominant When viscosity dominates 1 the form of the equation

for rise velocity is readily obtained from the relevant dimensionless group.
Thus
v~ = k, gD'(Pt - p,) (10.11)
/Lf

Experimental observations confirm this equation far vertical round

tubes with the following values o! k,

k, = 0.010 Wallis 7 (10.12)

k 2 = 0.0096 White and Beardmore 3 (10.13)

Surlace tension dominant Surface tension dominates when the bubble

does not move at ali. The static interface adopts a particular shape so
 ONE-DIMENSIONAL TWO-PHASE FLOW
"'
of the constant which were obtained for round tubes by these authors
were:
Dumitrescu 1 k, = 0.351 (10.6)
Davies &nd Taylor' k, = 0.328 (10.7)

Dumitrescu's experiments gave a slightly different result

k 1 = 0.346
and this is clase to the value (0.345) obtained in a series of experiments
by White and Beardmore. 3 The preferred value of k, is therefore
k, = 0.345 (10.8)
If one measures the rise velocity of a bubble in a tube with an open
top there is an apparent dependence on bubble length due to the expan-
sion of the gas as it rises in the hydrostatic pressure gradient. This
expansion leads to a value of j ahead of the bubble which is not zero, and
the rise velocity is augmented by the velocity of this liquid. By timing
bubbles of different lengths in tubes which were closed at the top and
did not allow any motion of the liquid ahead of the bubble, Nicklin 4 has
shown that the bubble rise velocity relative to stationary liquid is indeed
given by Eqs. (10.5) and (10.8) and is independent of the bubble length.
The case of tubes of noncircular cross section was investigated by
Griffith. 5 For rectangular channels of dimensions Db by D, the larger
dimension Db was found to be the more significant one for insertion into
Eq. (10.5). The constant k, can then be expressed as a function of
D,/D, anda good approximation to the final result is

k, = 0.23 + 0.13 ~: (10.9)

The limit1ng case in which D,/D, goes to zero agrees with Birkhoff and
Carter's 6 theoretical predictions for plane two-dimensional bubbles.
Griffith's paper also gives results for bubbles rising in annuli and
tube bundles. For annuli the outer diameter D, is the significant dimen-
sion for insertion into Eq. (10.5). The variation of k, as a function of
the ratio of the interna! to externa! diameters (DJD,) is shown in Fig.
10.3. A remarkable conclusion is that the smaller the annular spacing
is made, the faster the bubble will rise.
In the limit Di = D, the result is approximately
k 1 = 0.23 y;;; (10.10)
which would be expected from Eq. (10.9) for a plane bubble which is
wrapped around the annulus so that the liquid streams coalesce on one
side. Symmetrical bubbles are usually not observed in annuli.
SLUG FLOW

0.6 r--r---r-;--;---1-,-,--<,r---,---,~

Tube bun les

'
0.2 f-------+-------+-------1

o~~~-~~~-~~~-~
O O. 2 0.4 0.6 0.8 1.0
Dimension ratio D¡/Do for annuli
fig. 10.3 Dependence of the fac- [l-Dh/D 0] fortube bundles
tor k1 on duct geometry. Ds/Dh for rectangles
( GrijJith. ') (Dh = hydraulic diameter)

In tube bundles the characteristic dimension is more likely to be the

overall housing diameter D, rather than the dimension of individual tu bes.
Experimental evidence far slug-flow bubbles intube bundles is contained
in Griffith's paper. 5
The reader is reminded that the above equations are valid only
when surface tension and viscous forces are negligible. The limits of
validity of this regime will be discussed later.

Viscosity dominant When viscosity dominates 1 the form of the equation

for rise velocity is readily obtained from the rclevant dimensionless group.
Thus
v = k 2 gD'(p; - p,) (10.11)
~ µ¡

Experimental observations confirm this equation for vertical round

tubes with the following values of k,

k, = 0.010 Wallis' (10.12)

k, = 0.0096 White and Beardmore 3 (10.13)

Surface !ension dominan! Surface tension dominates when the bubble

does not move at all. The static interface adopts a particular shape so
288 ONE-DIMENSIONAL TWO-PHASE FLOW

that hydrostatic forces are completely balanced by surlace forces. Far

vertical round tubes, it has been shown analytically by Bretherton 8 and
by Hattori 9 that this will occur when

N Eü = gD'(p¡ - p,) < 3.37 (10.14)

<I

The above dimensionless number was called the Eotvos number by

Harmathy. 10
An alternati ve variable which is sometimes used is the Bond number,
defined as follows:

(10.15)

The general case Since the general solution is governed by three

parameters it can be presented as a two-dimensional plot of any two
chosen dimensionless groups with a third independent dimensionless
group as a para.meter. The actual manner in which these groups are
chosen is simply a matter of convenience. For example, thc dimension-
less bubble velocity k1, defined by Eq. (10.8), may be plotted versus the
dimensionless inverse viscosity N I given by

N¡ = [D'g(p¡ - P,)p¡]¼ (10.16)

I'!

which is obtained by eliminating ero lrom the first two dimensionless

groups. A convenient third independent parameter is obtained by elimi-
nating both "~ and D entirely to obtain the Archimedes number which
depends only on fluid properties and gravitational acceleration and is a
constant for a given fluid at a particular temperaturc. Thus

(10.17)

The resulto! plotting experimental data in this way is shown in Fig. 10.4.
The three asymptotic solutions equivalent to Eqs. (10.8), (10.12), and
(10.14) are clearly satisfied. They are

Inertia dominant N > 300 NEó > 100

k1 = 0.345
Viscosity dominant N¡ < 2 NEó > 100 (10.18)
k1 = 0.0lN¡
Surface tension dominant NE6 = 3.37
(10.19)
N¡' = 6.2NAc
SLUG FLOW

10 1 ~---~------,-------,-------,----r----r---,
o Water, NA,=2Xl0 5 ◊ 5000 viscasil, NA,= 1.23 X 10- 3
4
• 50% glycerine, NA,= 8 X 10 3 v 10,000 viscasil, NA,= 3.08 X 10-
Present . 2
1,;. 30,000 viscasil, NA,=3.42 X 10-
5
k v 70% glycerine, NA,=6.9x10
wor { <D 89% glycerine, NA,= 14 o 60,000 viscasil, NA,= 8.55 X 10- 6
10º
◊ 95% g!ycerine, NA,=2.2

Previous work using water

t rm Dumitrescu
♦ Griffith ond Wallis

Eq.[10.19)

Dimensionless inverse viscasity, Nf-----tJ-

Fig. 10.4 General dimensionless representation of bubble rise velocity in slug flow.
(Wallis.')

Alternative methods of plotting the result have also been used by

combining the dimensionless quantities in various ways. A parameter
used by White and Beardmore 3 is the property group

y~ gµ.¡' (10.20)
u3p¡

Y is simply equal to 1/NA, 2 when the gas density is low compared with the
liquid density. A plot of k1 versus NE, as a function of Y is shown in
Fig. 10.5.
An equation which has the property that it reduces to Eqs. (10.8)
and (10.11) in their appropriate range of applicability is

(10.21)
Equation (10.21) also gives a good approximation in the intermediate
range of N I shown in Fig. 10.4, when surface-tension effects are negligible.
2'0 ONE-D!MENSIONAL TWO-PHASE FLOW

0.40

0.30
~¾'
{!
'
~'

k,

0.20

soo 1000

Fig. 10.5 An alternative representation of Fig. 10.4. (White and

Beardmore. 3 )

Surface-tension effects can be approximated algebraically by a

further modification

(10.22)

where mis a function of N 1 and takes on the following values:

N¡ > 250 m = 10
18 < N1 < 250 m = 69N1 - 0· 35
(10.23)
N, < 18 m = 25
Equation (10.22) is a general correlation for bubble rise velocity in terms
of ali the relevant variables.
In the inviscid region when N 1 is large, Eqs. (10.23) and (10.22)
yield
k 1 = 0.345(1 - e'8- 37-N,aJl 10 ) (10.24)
SlUG FLOW 291

This result correlates the data whích were used by Masica and Petrash 11
equally as well as their equation, which was written in terms of the
Bond number,
4
k1 ~ 0.34 1 - - Q.84)N,,/
- .']
(10.25)
[ ( N B,

The bubble shape ís dífferent for the various regimes. In a highly

víscous fluid (N < 2) both the bubble nose and tail are rounded and the
wake is laminar, whereas in a fluid of low viscosity (N > 300) the bubble
taíl ís flat and the wake ís turbulent.

Example 10.1 What is the rise velocity for a slug-flow bubble composed of gas with
density 10- 3 g/cm 3 in a vertical round tube 2 cm in diarnetcr? The liquid
properties are p¡ = 1 g/cm 3, (j" = 100 dynes/cm, µ¡ = 1 poise.
Solution First 1 evaluate the following dimensionless groups, neglecting the gas
density in comparison with the liquid density:

,:¡ .. ~ gD'ri ~ (981)(2)'(1) ~

1 Ea (J 100 39 .2
gt,n,<p¡ (98H>)(2%)(1)
N 1 = --µ¡- = 1 = 88
• (j"J'2p1¼ (100%) (I¼)
h M = µlg¼ = (!2)(981¼) = 32
y~ _I_ ~ 10-a
N1u2

Using Eq. (10.22) we have

69
- NJ·ss - 15
m - -- -

k1 = 0.345(1 _ e-o.ss10.W) (1 - eC3.37-3Y.2Jl15) = 0.29

Therefore, from Eq. (10.5),

v.~ 0.29(1-l')[(981)(2)(1)]h
= 12.7 cm/sec

Alternatively, usi.µg Fig. 10.5 ,ve find k1 = 0.26, v"' = 11.4 cm/sec.

USE OF THE IBUBBLE VELOC!TY IN THE DRJFT-FLUX MODEL

The techniques of Chap. 4 can be applied to the slug-flow regime as long

as the effects of wall shear stress on the bubble dynamícs are small. Thís
is approxímately true for values of N 1 greater than 300 and when the
frictional pressure drop, estimated from homogeneous theory, is small
compared with the gravítatíonal pressure drop.
In Sec. 10.2 ít was found that the dríft velocíty v,1 was índependent
of a but was a function off However, for one-dimensional flow with no
wall shear effects the dríft velocity is independent of j but ís a functíon
m ONE-DlMENSIONAL TWO-PHASE FLOW

of a. The only way in which these conditions can both be satisfied is

for Vfli to be a constant. This constant can be evaluated by choosing the
special case of a single bubble in stagnant liquid for which
(10.26)
Therefore, for all values of a,
(10.27)
The bubble velocity when there is net flow is, from Eq. (10.3),
v, ~ j + v. (10.28)
Therefore the mean void fraction, from Eq. (10.4), is

(a)~~ (10.29)
J + v.
In terms of the volumetric flow rates, Eq. (10.29) can be written 12

(a)
Q,
(10.30)
Q1 + Q, + Av 00

In view of the assumption of negligible wall shear stress the pressure

gradient in nonaccelerating flow is
dp
- dz ~ g[p¡(l - a) + p,a] (10.31)

and this may be calculated by using the value of a from Eq. (10.29).

IMPROVEMENTS TO THE SIMPUFiED THEORY

In reality the bubble drift velocity is not strictly constant since it is

influenced by the velocity profile in the liquid slug. This profile is a
funetion of j-or more strictly a function of the Reynolds number defined
by Eq. (10.2)-and is also influenced by the wake of a preceding bubble.
These effects may be taken into account by applying correction factors
to Eq. (10.28). Thus
(10.32)
The coefficient C1 is a measure of the fact that the bubble does not simply
move relatively to the average liquid velocity but relative to a weighted
average value. C2 is a measure of the change in relative velocity dueto
the approaching velocity profile. C, has the same overall effect on the
equations as the coefficient C, in Eq. (4.24) although the physical reason-
ing behind the derivation of these coefficients is not identical.
A simple result is obtained in the case of fully developed turbulent
flow in the liquid slug (Reynolds number greater than approximately
8000). Nicklin 4 has found that in this case the bubble moves with a
 SLUG FLOW m

velocity "~ relative to the approximately uniform velocity in the turbulent

core. For a circular pipe and fully developed flow the result is
Re¡> 8000 (10.33)
Griffith 5 has suggested appropriate values of C, far rectangular channels
and annuli at high Reynolds numbers and these are shown in Fig. 10.6.
Accepted corrclations for C 1 and C2 in the laminar flow region are
not yet available. Griffith and Wallis 12 plotted C, versus Reynolds num-
ber with sorne success but failed to separate the effects of the two correc-
tion factors. When velocities are high, C, probably approaches the limit-
ing value of 2.27 which was predicted by Taylor 13 for horizontal flow.
The effect of the wake behind one bubble on the rise velocity of the
next bubble has been rationalized by Moissis 14 who expressed the factor
C2 as a function o! the ratio between bubblc separation (liquid slug length)
L, and pipe diameter D far circular tubes:
C2 = 1 + se-l.06L,/D (10.34)
This means that the following bubble is continually tending to eatch up
the preceding bubble and coalesce with it. Slug flow is therefore never
strictly stable and fully developed, although eventually the slug length
beeomes sufficiently long for the second term in Eq. (10.34) to be negligi-
b1e. In the entrance region in a pipe or when slug flow is developing
from bubbly flow, or when bubbles are continually being farmed as a
result of boiling, the interaction between bubbles is significant and should
be taken into account. It is rather awkward to use Eq. (10.34) in prac-

Square channel~

1.26 ,---,-----=::::=ccj=-4
Circular pipe
asymptote

1.22 /
C1

1.14
"""f----...-. Para 11 el piones ~~t-----_¡

F"ig. 10.6 Variation of the factor D;/Do for annuli

C1 with duct geometry. (Grif- (1- Dh/Do) for tube bundles
fith.S) DslD.b for rectangles
 ONE-DIMENS!ONAL TWO-PHASE FLOW

tice since the slug length L, is indeterminate solely in terms of the overall
flow rates of thc components. Griffith 5 suggests that in a channel in
which boiling is occurring it is sufficiently accuratc to assume that
e, = 1.6 (10.35)
The coefficients C, and C, modify Eq. (10.30) to give an improved
equation for the mean void fraction

(a) = C,(Qr + Q:) + C,Av~ (10. 35)

The correction to Eq. (10.31), which is necessary when wall shear

stress is significant, is difficult to estimate. The average shear stress can
be either positive or negative since sorne liquid is actually running down
the wall around the bubble. A possible procedure is to calculate the
shear stress in the liquid slug from the single-phase friction factor based
on j, i.e.,
(10.37)
where C1 is the usual funetion of Re¡. Approximately a fraction (1 - a)
of the pipe length is occupied by slugs, therefore, if the wall shear stresses
around the bubble are neglected, Eq. (10.31) becomes, with the addition
of the drag on the liquid slugs,
dp 2p1j 2
- dz = g[p¡(l - a) + p,a] + (1 - a)C1 D (10.38)

If the gas density is rnuch srnaller than the liquid density we have
Pm = (1 - a)p¡ (10.39)
and the last terrn in Eq. (10.38) can be rewritten as
- dp) = 2C1Pm.i2 (10.40)
( dz , D
which could be regarded as a homogeneous flow frictional pressure drop
in which the mean density is calculated from the void fraction in Eq.
(10.30). The techniques which were used in the second part of Example
9.4 should therefore be approximately applicable to slug flow as well as
to bubbly flow.
CORRECTION FOR LONG BUBBLES

If the slug-flow bubbles are long (greater than 15 diameters for example)
a substantial amount of liquid is held up in the film surrounding the
bubble. For purely potential flow the liquid downward velocity relative
to the bubble is V2gh at a distance h below the nose. Eventually, how-
ever, a terminal velocity is reached at which the weight of the film is
 SLUG FlOW
'"
completely balanced by the wall shear stress. The film now !alis at a
steady speed and has a uniform thiekness which can be ealculated from
falling film theory. Since its weight is completely balanced by the wall
shear stress, the liquid in the film does not contribute to the pressure drop.
Furthermore the liquid slug length is also decreased by the amount of
water. which is held up in the film. Both of these effects can be accounted
far by treating the liquidin the film as if it were gas and modifying the
void fraction in Eq. (10.38) .. For a long cylindrical bubble surrounded
by a film of thickness á (Fig. 10. 7) the ratio of the bubble volume to the
tube volume is

Therelore the effective void fraction far pressure-drop prediction in Eq.

(10.38) should be modified to

ci=a ( 2•)-2
D
1- (10.41)

The film thickness is calculated as follows. Consider a section

across the tube in the cylindrical part of thc bubble shown in Fig. 10. 7.

¼ v = /2gh relative to bubble

800 asymptotic thickness

lf'ig. 10.7 Falling film flow around

a long bubble in a vertical tube.
296 ONE-DIMENSIONAL TWO-PHASE FLOW

The gas velocity upward in the center is the same as the bubble velocity.
Therefore

(10.42)

For inviscid bubbles and turbulent flow (Re; > 8000) Eqs. (10.32) and
(10.33) give

(10.43)

The gas flow rate across the section is

(10.44)

Let the liquid flow downward across the section be Q;.

Since the total
volumetric flow across any cross section of the pipe is constant from
continuity,
Q", - Q'f -- Q -- 1rD'
4 J
. (10.45)

Combining Eqs. (10.43) to (10.45) we find eventually that the liquid flux
downward is

., = "D'
Jt
4Qí = 2.
(1 . ' + v. ) (1 -
2•)'
D - . J (10.46)

Now, Q;is related toó by the falling film theory of Chap. 11. Therefore
ó can be found by solving Eq. (10.46) and the falling film equations
simultaneously.
The theory is only valid if the value of N¡ is greater than 300 so
that the bubble velocity is given by Eq. (10.5) and is

(10.47)

[Note: Similar techniques can be applied to bubbles in viscous fluids as

soon as an equation can be developed to replace Eq. (10.43) for
Re;< 8000.]
1'*'
Using the relationship j;/v =
0 45
00
f
, falling film theory from Eqs.
(11.68) and (11.76) is now put into the convenient form
.,
Rer = j7'N, = lt (0.345)N¡ (10.48)
v.

Rer < 3500 !~ = 3.85N¡ (~)" (10.49)

3500 < Rer < 30,000 Íl = 190

v.
(-°-)¾
D
(10.50)
 SLUG FLOW
"'
Dividing Eq. (10.46) by v we have00

Íf. = (1.2 _i_ +

voo V"'
1) (1 - 2Dó)' - _J_
V"'
(10.51)

Equations (10.49) to (10.51) are now to be solved simultaneously.

Figure 10.8 shows a graphical solution. Equation (10.51) is used to
plotj;/v versus ó/D for various values of j/v
00 The intersections with
00
•

Eqs. (10.49) and (10.50) give the value of ó/D, which is substituted into
Eq. (10.41) to give a modified value of a for use in Eq. (10.38). In
general, Eq. (10.38) will tend to overestimate the pressure drop, whereas
the modification using Eq. (10.41) will underestimate it.
Acceleration pressure drop can be treated as in the bubbly flow
regime and as shown in Example 9.4. However, since the slug-flow pat-
tern is not as homogeneous as bubbly flow, the choking condition given
by Eq. (r) of Example 9.4 is likely to be in error. Furthermore, since
sorne of the liquid is moving with a velocity j in the slugs, whereas the
rest is moving downward in the fa.lling film 1 the assumption of uniform
liquid velocity is incorrect. Further studies are worthwhile in this a.rea .

.Example 10.2 Calculate the mean pressure gradicnt in a vertical oil well for thc follow-
ing conditions 1 Q1 = Qª = 35.4 liters/sec, D = 15 cm, µ 1 = I poise, u = 25
dynes/cm, P! = 0.85 g/cm 3, P/1 = 0.0025 g/cm 3 • The local pressure is 60 psia.

Eq. [10.49)

O.

0.03 0.04 0.05 0.06 0.07

'ID
fig. 10.8 Graphical solution for film thickness around a slug-flow bubble.
 ONE-DIMENSIONAL TWO-PHASE FLOW

Solution The long length of vertical pipe in an oil well is conducive to the formation
of slug-flow bubbles. Furthermore, if sufiicient length is available for agglom-
eration to occur, these bubbles may be vcry long. Depending on the particular
details of the well design the pressure gradient should thcreforc lie between the
predictions oí Eq. (10.38) which are obtained using Eq. (10.36) or (10.41), as
long as the slug flow is fully developed and no correction is introduced by Eq.
(10.34).
First we calculate the appropriate dimensionless groups as follows:
N¡ (981H)(l5¾)(0.85) ~
1440
1
N Eo ~ (981) c1g;) (0.85) ~ 7250
(25¾)(0.85H)
N'" ~ (1')(981H) ~ 3 · 2

From Fig. 10.4 it is found that k 1 = 0.345. Therefore

v. ~ 0.345(981 X 15)" ~ 42 cm/sec

The cross-sectional area is

A = ~ (15) 2 = 177 cm 2
4

The volumetric fluxes of the oil and gas are

.
Jt = A =
Qf 35,400
~ =
200 /
cm sec
. Q,
Jg = A = 35,400
~ =
200 /
cm sec

The Reynolds number for the liquid slug is

Re;~ pJÍD ~ (0.85)(400)(15) ~ 5100
µ¡ 1
There is now a problem in decidiri.g what to do since the theory is not yet
adequate for predicting C1 and C2 for Reynolds numbers below 8000. The
only procedure which is available is to assume that the error in applying the
turbulent flow theory with a Reynolds number of 5100 is not significant. Then
the value of (a/ from Eq. (10.36) with C1 = 1.2 and C2 = 1 is
35,400
(a)~ (1.2)(70,800) + (42)(177) ~ 0. 334
For a smooth pipe with a Reynolds number of 5100 the friction factor is 0.009.
Therefore the sum of the gravitatiollal and frictional pressure drops in Eq.
(10.38) is

- :; ~ 981[(0.85)(0.616) + 0.0025(0.384)] + (0.61 5 ~(o.oo 9) (0.15)(400)' {5

= 513 + 100 = 613 dynes/cm 3
The correction for acceleration pressure drop is found using Eq. (q) of
Example 9.4.- The denominator of that equation is, with sufficient accuracy,
l _ p¡j¡'j, ~ l _ (0.85)(200)ª
(j¡ +
v.)p (242) (60) (69,000) ~ 0 993
·
SLUG FlOW
"'
The pressure gradicnt is therefore

dp _ 613 _
- dz - 0. 993 - 618 dynes/cm 3

To make the correction for long bubbles we use Fig. 10.8 with the values
j/v<r, = 400/42
= 9.5 and N = 1440. The value of ó/D is found to be 0.0485.
From Eq. (10.41) the modified void fra.ction is
0.384
a' ~ [1 - 2(0.0485)]' ~ 0. 47!

Substituting this value into Eq. (10.41) the predicted pressure drop is decreased
in the ratio (1 - a')/(1 - a) to the value

_ dp _ (618)(0.529) _ d / ,
dz - 0.616 - 531 ynes cm

The pressure gradient therefore should lic bctween 531 and 618 dynes/cm 3 and
is probably closer to the lo,ver valuc.

ViSCOSiTY EFFECTS

At low values of N¡( < 300), or far values of Re; less than 8000, viscous
effects become important and the above equations are incorrect. A com-
plete study of these phenomena is not availablc. However, as long as
buoyancy effects are small compared with viscous and surface-tension
effects, the results to be described in Sec. 10.4 under the heading
"Horizontal Slug Flow" are approximately valid. It is suggested that
these results should be used when v., as derived from Eq. (10.22), is
much less than j I for example 1 when

"j• < 0.1 (10.52)

10.4 HORIZONTAL SLUG FlOW

IBUIBBLE VELOCITY

There is no drift flux in horizontal flow dueto buoyancy effccts. There-

fore v<r, loses its significance. The bubbles, however, do not move with
the same average velocity as the liquid. This is apparent from Fig. 10.9.

~L_____:~A
8 ,~ , , ~~ j

f
Fig. 10.9 Horizontal slug flow.
300 ONE-DIMENSIONAL TWO-PHASE FLOW

Since there is no pressure drop along the length of the bubble, the liquid
film on the wall is substantially stationary. If this film has a mean
thickness o, the available flow area for gas in the bubble is

(10.53)

Therefore, for continuity of volumetric flux at this cross section

(10.54)

or, if the film is thin

(10.55)

Hence, Vb is greater than j.

A useful parameter which represents the fraction of the bubble cross
section whieh is occupied by the liquid is

Ab j
m=l-~=1-- (10.56)
A Vb

In the absenee of effects due to the gas viscosity and inertia and for
bubbles which move independently, the factorsinfluencing bubble velocity
can be combined into the following dimensionless groups:

J jD J µ.¡ (p¡ - P,)gD'

Vb V¡ u u

The first group represents the ratio between the liquid velocity in
the slug and the bubble velocity; the second group is the Reynolds num-
ber for the Iiquid in the slug; the third group represents the relative
importance of viscous and surface-tension effects¡ and the final group is
the ratio between buoyancy and surface-tension forces. In fact the last
three groups are directly analogous to the groups which describe the
balance between inertia, viscosity, surface tension, and buoyancy in the
case of vertical flow. The first group is the inverse of C1.
The second and third groups can be combined to yield a parameter
which is independent of vclocity and is a constant for a particular fluid
in a particular pipe, i.e.,

(10.57)
Sl-UG FLOW
'º'
15
Suo performed experiments in the range

(p¡ - P,)gD' < 0.88 (10.58)

"
for which stratification effects were small and the buoyancy forces could
be neglected. His results are shown in Fig. 10.10 plotted in terms of the
dimensionless quantities nientioned above. It can be seen that at very
low velocities the bubbles move at the slug velocity and the liquid film
around the bubbles is therefore very thin in view of Eq. (10.55). At
high velocities and high Reynolds numbers the ratio j/vb tends to ahout
0.84, that is,

v, = l.19j (10.59)
The value of the coefficient in Eq. (10.59) is very clase to the equivalent
value (C, = 1.2) for vertical flow at high Reynolds numbers. Tbe
Reynolds numbers Re, corresponding to the approach to the asymptotic
value in Fig. 10.8 are approximately 3000. Therefore a simple expression
for bubble velocity in terms of the overall flow rates for Re; > 3000 is

- 1• 2 Q¡ A Q,
Vb -
+ (10.60)

The line corresponding to large values of A in Fig. 10.10 represents

the regime in which inertia effects are negligible and the motion is gov-
erned by the balance betWeen viscosity and surface tcnsion. In this
case the results of Taylor 13 can be represented by the empirical equation

. (10.61)

1.0 A =1.5 X 10- 5

A= 2.1 X 10- 5
1 A very large
0.6 Slug flow
Bubbly slug flow

Jµf X 10 3
cr

Fig. 10.10 Velocity ratio as a function of jµ¡/u. (Suo and Griffith. 15 )

 ONE-DIMENSIONAL TWO-PHASE FLOW

which red u ces to

m ~ 1.48 /J,Vb)º· 567

(~ u
(10,62)

at low values of µ,v,/u, Equation (10.62) 1s intermediate between

Bretherton's 8 equation

(10,63)

and one derived by Fairbrother and Stubbs 16

m =LO-;- (µ.v,)" (10,64)

and represents Bretherton's data better than either Eq, (10.53) or (10.64),
The para.meter m is very useful if one is interested in predicting thc
film thickness around a bubble, but tbe form of Eqs, (10.61) to (10.64)
forces an iteration procedure since vb is not usually known. An alterna-
tive expression which gives C 1 to an accuracy of 2 percent is

(10.65)
If an accurate value of m is required, Eq. (10.61) should be used after
employing Eq. (10,65) to estímate v,,
When C, exceeds 2, the bubble actually moves faster than the center
streamline in the liquid slug, The character of the liquid flow relative
to the bubble is therefore quite different 21 , 26 for the two cases C 1 < 2
and C1 > 2.

VOID FRACTION

The void fraction a is calculated directly from Eq, (10.4) by using the
bubble velocity which was found as described above. Far the asymptotic
limit at Reynolds numbers above 3000, Eqs, (10.60) and (10,4) give the
simple result

a = 0,84 Q¡ ~ Q, (10,66)

This equation can be compared with Eq. (4.31) and agrees well with the
experimental results of Armand. 17

IPRESSURE DIROP

The pressure drop in the liquid slug can be calculated by single-component

flow techniques. The pressure drop along the cylindrical part of the
SLUG FLOW ,.,
bubble is zero. Therefore one is left only with an additional pressure
drop per bubble due to effects at the nose and tail. Since different num-
bers of bubbles can make up the same overall flow rates in the same pipe,
the pressure drop is not determínate unless the bubble length is specified
separately.
The pressure drop per bubble was found by Suo 15 to be correlated
by the following equations far low values of Re;:

Re;< 270 Ápb = 9ܵ¡Vb

(10.67)
D
270 < Re; < 630 "'-p, = 0.163p¡v, 2 (10.68)

The sudden change at Re = 270 from purely viscous terms in Eq.

(10.67) to purely inertíal terms in Eq. (10.68) is unusual, since no sudden
change in flow dynamics occurs at that point, and it is likely that equa-
tions of the form of Eqs. (10.67) and (10.68) are valid at low and high
Reynolds numbers, respectively. The pressure drop per bubble is then
approximately equal to the pressure drop in a length of about four pipe
diameters. An approximate correlation is then for all Reynolds numbers
for the typical unít cell comprising one bubble and one slug,

' -- 4Cf ½zp¡J·2 L, D 4D

up + (10.69)

where L, is the slug length.

The mean pressure gradient is then

dp 2C¡p¡j2 L, + 4D (1 O. 70)
- dz = D L, + L,

The evaluation of the last fraction ín Eq. (10. 70) depends on the condi-
tions of the problem. Perhaps the volume u, of each bubble is known.
Them the length of a unit cell follows from a knowledge of the void
fraction,

(10.71)

Since l/C1 represents the fraction of the cross section which is

occupied by the bubble, in view of Eq. (10.54), we have a good approxi-
mation for long bubbles

(10. 72)
304 ONE-DIMENSIONAL TWO-PHASE FLOW

Substituting Eqs. (10.71) and (10.72) into Eq. (10.70) gives

_ dp
dz
= 2C1 p1j2
D
(! _C, + 4DA)
a u,
°' (10.73)
Since
J, (10.74)
°' = C,j

Eq. (10.73) can be expressed entirely in terms of the fluxes and C, as

follows:

dp_2CrPÚ(· +4DA ·) (10.75)

- dz - --y¡- J¡ u,C 1 J,

10.5 SLUG FLOW IN INCLINED PIPES

Slug flow in inclined pipes was studied by Runge. 18 The parameters

which describe this regime are precisely those which were used to describe
vertical flow together with the angle e made between the axis o! the tube
and the vertical. For the bubble rise velocity one obtains curves similar
to those in Fig. 10.4 at each value of e, and a book o! graphs would be
necessary to represent such a four-dimensional surface.
Suppose, for example, that the bubble velocity is rendered dimen-
sionless by the use of the paramcter k,. In the general case we should
have
(10. 76)

A simpler and more convenient method of presentation is to divide

Eq. (10. 76) by the value o! k 1 which would occur lor the same fluid in a
vertical pipe and which can be calculated lrom the results o! Sec. 10.3.
Canceling the common parameters wc then find that
v, k1(N1,N E,,0)
(10.77)
k1(N1,NE,,O)

v, is the bubble velocity in the inclined tube. Equation (10.77) then

states that the ratio o! tbe bubble velocity in the inclined tube to the
bubble velocity in a vertical tube can be expressed as a lunction of N¡,
NEo, and 0.
In Refs. 18 and 19 the four-dimensional surface represented by
Eq. (10. 77) was approximated over selected ranges o! the E6tv6s number
by the curves shown in Fig. 10. lla through d. These figures can be used
in conjunction with the results of Sec. 10.3 to predict the rise velocity
o! slug-flow bubbles in stationary liquid in inclined tubes to within about
10 percent for a wide range of practica! conditions.
SLUG FLOW 305

Nrn > 100

1.6

0.8 □ Nr = 23, 130

◊ Nr = 24
"'Nr=15.7
o N 1 = 0.36
0.4

º~----~---~----~----~----~----
º 15 30 45
Degrees from vertical, 0
60 75 90

[a)

67<NE 0 <87
1.6

1.2

0.8 o Nr=12,625
◊ N,~10,480
□ Nr= 900
"' N 1 = 224
0.4 v Nf = 45

º~----~-~-~~---~----~----~----~
O 15 30 45 60 75 90
Degrees from vertical, B
[b)

Fig. 10.11 Rise velocity of slug-flow bubbles in inclined pipes. (Runge and Wallis. 18 )

A remarkable feature of these results is that the bubble velocity in

inclined pipes generally exceeds the value in the same pipe placed ver-
tically. In an inviscid liquid thc bubble rises faster than in a vertical
pipe even when the ~xis is as low as 2 or 3º from the horizontal. Care
should therefore be taken in certain design situations to ensure that
"horizontal)) pipes are accurately oriented.
 ONE-DIMENSIONAL TWO-PHASE FLOW
"'
2.0

40 < Nrn < 50

1.6

,, 1.2
'~
0.8 ◊ Nr = 7250
V Nr = 1230
□ Nr = 700
A Nr = 175
0.4 □ Nr = 35

o
o 15 30 45 60 75 90
Degrees from vertical, B
(e}

2.0

20 < NEO< 30
1.6

1.2
et_
'~
0.8 □ Nr = 4460
◊ N,=410;100
□ Nr = 20
d N¡ = 12
0.4 V Nr = 8.2

o
o 15 30 45 60 75 90
Degrees from vertical, 0
(d}
Fig. 10.11 (Continued)

Figure 10.12 shows how the shape of air bubbles rising in glycerin
changes as the tube inclination is varied. When the tube is nearly hori-
zontal the bubble appears to slide along the underside of the upper part
of the tube. The liquid film along the top of the bubble can be very thin
and dry spots will occur if the contact angle is large. This phenomenon
is of particular importance in boiler design since evaporation of the thin
SUJG FLOW 307

0~ 75º
(b)
Fig. 10.12 Slug-flow bubbles in inclined tubes. (a)
o - 60º; (b) 8 - 75º.

film on the top of a bubble can lead to local overheating and blistering
of the material of the tube.

Examp!e 10.3 Solve Example 10.1 for values of 0 of 30, 60, and 80º.
Solut-ion The value of NEo is closest to the range of Fig. 10.llc. For N 1 = 88 the
approximate values of ve/v., for the given inclinations are

30º 60º ~
1.2 1.2 1 1.0-

Therefore, taking a mean of the predictions of Example 10.3, the required

velocities are 14.4, 14.4, and 12.0 cm/sec.

10.6 THE llMITS OF TH.E SUJG-FLOW i!EGIME

It is practicaIIy certain that there is no lower limit to mean void fraction 1

quality, or flow rates at which slug flow can be obtained. There is always
a tendency for the bubbly flow regime to change to the slug-flow regime
at low velocities as a result of agglomeration. Below approximately 10
percent void fraction this process is relatively slow and definite slugs may
be observed only in a particularly long pipe. If impurities are present the
bubbly regime may be prolonged.
If long slug-flow bubbles are brought to rest by stopping the flow
in a smaII horizontal pipe-ar in a zero g environment-they will break
up into smaller bubbles about two or three pipe diameters in length. 20 · 21
 ONE-DIMENSIONAL TWO-PHASE FLOW

Presumably this instability can also occur at extremely low values of the
volumetric flux j. However, this mechanism does not lead to a change in
flow regime but merely imposes an upper limit on the length of a stable
bubble. S!ug-flow bubbles can also break up when they flow through
constrictions. 21
The upper limit of slug flow occurs as a result of shear stresses and
drag forces between the liquid and the gas in the slug. One regime bound-
ary is determined by the condition that the slug-flow bubble shall be
broken up into smaller bubbles as a result of its motion relative to the
surrounding liquid. In a very viscous fluid the bubble breaks up by the
formation of long filaments which stream from the bubble tail and even-
tually split up to form small bubbles. At high Reynolds numbers, on
the other hand, the liquid flow separates at the bubble tail to form a very
agitated wake in which small bubbles are entrained. A condition far this
latter phenomenon in horizontal flow is suggested by Suo 15 ; it is
V 3P 2D2
1
'
µ.¡CI
> 1.1 X 10 6 (10.78)

This flow regime transition is not an abrupt one and slug-flow bubbles
with groups of small bubbles following them can be observed overa range
of flow rates.
In vertical flow, similar conditions are obtained. At high Reynolds
numbers the relative velocity between the gas and the liquid steadily
increases down the bubble length until eventually the liquid film reaehes
a terminal velocity due to shear stresses at the wall. The condition of
bubble destruction dueto the effects of relative velocity between the liquid
film and the gas in the bubble therefore imposes a condition of maximum
stable bubble length and not a condition far the complete destruction o!
slug flow. For practica! purposes a slug-flow bubble with many small
bubbles in its wake behaves in much the same way as the equivalent large
bubble since the dynamics of the bubble nose determine the drift velocity
and the large bubble and its wake move together.
The countercurrent flow of the gas core of the bubble and the liquid
film have been recognized by Nicklin and Davidson 22 as a potential cause
of flooding and the consequent development of large unstable waves at
the interface. Until flooding occurs the drag forces between the gas and
the liquid film are essentially .zero. Let the liquid film thickness be ó;
then the available flow area far the gas is

(10. 79)

The gas flow rate in the core is then

Q; = Abvb (10.80)
SLUG FLOW
'"'

Onset of film
insto bi Iity ~
O/ ond Qg' lie on this líne

Flooding ~ decreosing
in e ~
1

- - - -Q¡

Fig. 10.13 Determination of maximum stable bub-

ble length from Nicklin and Davidson's theory.

where vb is calculated as in Eq. (10.3) and is a function of j but not of

j¡ and j,. The downward liquid flow rate in the film is, from continuity

(10.81)

The relationship between Q; and Q; at which flooding occurs can be

determined by a separate experiment or from the equations which will be
given in Sec. 11.4. Suppose this flooding locus is plotted as in Fig. 10.13.
For a given value of j, Eqs. (10.80) and (10.81) with ó as parameter repre-
sent lines of slope -1 through the point Q¡, Q, which is determined by the
imposed flow rates.
The value of the asymptotic film thickness 0 is found, as befare,
00

from Eqs. (10.41), (10.50), and (10.51). If the value of ó at which the
line cuts the flooding curve is less than óc,o, then no instability is reached.
On the other hand, if the line cuts the flooding curve ata value of ó greater
than 0 flooding should occur when the film thickness rcaches this value.
00 ,

For example, consider the case N > 300, Re; > 8000. The upward
gas flux in the bubble for this limiting film thickness is from continuity
1 1

(10.82)

J;
since is measured downward (we have dispensed with the sign conven-
tion for convenicnce). These values ofJ; andj~ can now be tested against
the appropriate falling film flooding correlation. For N > 300 anda con-
dition in which the limiting process occurs in the film alone, the suggested
equation is Eq. (11.159), i.e.,

(10.83)
310 ONE-DIMENSIONAL TWO-PHASE FLOW

U sing the definitions of the dimensionless groups we find that

.• 0.35j¡
J¡ = -- (10.84)
voo
j: = 0.345 //>e_ j; (10.85)
'\J p¡Vc,')
Furthermore, Eq. (10.82) can be put into the dimensionless form

(10.86)

Now suppose that one starts with the parameters j/v p1 /p,, and N 1 . 00
,

First, Eqs. (10.49) to (10.51) are solved for j;/v Using this result, j;!v
00 • 00

is found from Eq. (10.86). Thenj; andj: are calculated from Eqs. (10.84)
and (10.85). The left-hand side of Eq. (10.83) is evaluated: if it is larger
than unity, flooding occurs; otherwise the falling film is stable. The con-
dition of the onset of flooding at the limiting film thickness is reached when
Eq. (10.83) is satisfied. This then specifies a relationship between the
originally chosen parametersj /v p¡/ Pu and N 1i which can be determined.
00
,

The calculations indicated above have been carried out by Porte-

ous,23 and the result is shown in Fig. 10.14. This graph has been arranged
to show the maximum allowable value of j/v for specified values of 00

VPJ/p, and N¡.

Turbulent

Rer = 3500

1~-~-~~~~~--~~~~~~~-~~~~~~~
100 1000 10,000 100,000
Dimensionless inverse viscosity, N

Fig. 10.14 Prediction of the critical value of j /v"" for flooding in slug flow.
(Porteous. 23 )
SLUG FLOW
'"
20
' '
-
1
1 ' 1
.. N=14,000 __ 1

- D = 1:025
--'·---¡-·- -

¡_...---·;ru;•
/
¿,,. .... -- . . L.- -
___ N = 2:!.,_0_QQ__'---
----,
----J:25" N=60,000
·...:.--,--
Turbulent theory
~
~
-

o ' ' 1
' '
o 0.2 0.4 0.6 0.8 LO
Jf, fps
IFig. 10.15 Comparison between the predictions of Fig. 10.14 and
Govier and Short's 24 observed slug-froth transition for low liquid
rates.

The above theory predicts when flooding will occur once the limiting
film thickness ó* is reached. If the bubbles are not very long then a
greater value of j/v~ will be necessary before noticeable flooding can occur
at values of ó greater than the asymptotic value.
Figure 10.15 shows a comparison between these predictions and the
regime boundary between slug flow and ufroth flown observed by Govier
and Short2 4 for liquid rates below j 1 = 1 fps. The substantial agreement
suggests that the froth was due to the break up of the lower end of the
long bubbles as a result of flooding. It is unlikely that this froth will
change the void fraction or pressure-drop equations for slug flow. How-
ever, there may be a significant effect on mass-transfer phenomena dueto
the increase of interfacial area which is obtained.
If the total volumetric flux is increased beyond the limiting value
given by Fig. 10.14, the effect is to move the onset of flooding up the
bubble until eventually no downward velocity is allowed in the liquid
film at ali.
This limiting value of j equals the volumetric flux of gas at the point
Q~ = O on the flooding line. This is the basis of Wallis' criterion 25 far
the transition to pure annular flow with no downward flow of the liquid,
namely,
j,p,¼ = 0,9 [gD(p¡ - P,W' (10,87)
Below the velocity predicted by Eq. (10.87) the flooding criterion predicts
a continua! bridging and unbridging of the gas coreas flooding and bubble
agglomeration occur in sequence. There is thus a region of ¡¡slug-annular
flow" between the regimes of simple slug flow and continuous film flow.
This topic will be discussed further in the following chapter.
Example 10.4 Check whether flooding occurs in the falling film around the bubbles
for the conditions of Example 10.2.
312 ONE-DIMENSIONAL TWO-PHASE FLOW

Solution From Fig. 10.14 for ...jp¡/pª = 18.,5 and N = 1440 we find that the
limiting value of j /v<» is approximately 16. For the conditions of Example
10.2, j/v""' is 400/42 = 9.5. Since this is less than the limiting value, fl.ooding
<loes not occur.

Example 10.5 Salve Example 2.5 using slug-flow theory.

Solution From thc given da.ta we obtain v"' = 0.56 fps, C 1 = 1.2, and C 2 l.
Therefore, from Eq. (10.36),

j, 47 ·2 O 773
ª ~ 1.2j + "• ~ (1.2) (50.3) + 0.56 ~ ·
The liquid fraction is
1 - a = 0.227

which agrces ,vell with the observed value of 0.23.

The gravitational and frictional prcssure drops are found from Eq.
(10.38) to be 8.2 X 10-s and 8.07 X 10- 3 psi/in., respectively. Adding these
togethcr and multiplying by 18 in. gives l::.p = 0.293 psi. lf the accelerational
effects are overestimated by using homogeneous theory (as in Example 2.5),
we have !::.p = 0.293/0.864 = 0.34 psi. These two estimates bracket the
observed value.
If the correction for long bubbles is applied, the_ predicted pressure drop
is decreased to about 40 percent of the above value. This is plainly in error
and the reason is clear from Fig. 10.14. Sincej/v = 90, flooding is prevalent
00

and only short bubbles survive.

10.7 PRESSURE OSCILlATIONS IN SLUG FLOW

Equations (10.38) and (10.75) predict the average pressure gradient.

However, the bubble is essentially at a constant pressure along its length.
Therefore the actual pressure variation over several unit cells appears as

~ 1
)
1
) C=:J Ró
1 1
1 1
1 1
l 1 1

-
1
Q
\"
""- 1
1
1
~Mean pressure

"
~
~
~ -l._
- -- -
1
Instantoneous
pressure

-- ✓
1
1
1

- --
1
1
1

Position, z -+-

Fig. 10.16 1\1ean and alternating components of pressure drop in

slug flow.
 SLUG FLOW m

in Fig. 10.16. An approximately triangular or sawtooth alternating com-

ponent of pressure drop is superimposed on the average pressure gradient
and can excite oscillations in two-phase circults in which slug flow occurs.

l'ROBLEMS

10.1. A certain silicone fluid has a viscosity of 5000 cp, a surface tcnsion of 21 dynes/cm,
anda density of 1 g/cm 3• What is the rise velocity of a slug-flow bubble in stationary
liquid in vertical pipes with diameters of 0.1, 0.5, .5, and 24 in.?
10.2. When a long bubble rises in a tube closed at the bottom, the value of j ahead of
the bubblc -is not zero because of expansion of the gas in the hydrostatic prcssure
gradient. A bubble 10 in. 3 in volume is injected into a column of water 100 ft high
in a 1-in. pipe. If the tempcrature is 70ºF and the pipe is closed at the bottom and
open to the atmosphere at the top, how long does it take after releasc befare the
bubble brcaks the surface?
10.3. Use Eq. (5.32) to show that the film thickness around a slug-flow bubble rising
in stagnant viscous fluid is about 40 % of the radius. What error would have bcen
introduced by using Eq. (11.69") which <loes not allow for the curvature of the tube?
10.4. What is the minimum tube size in which large bubbles of air will rise in stationary
water at 70ºF (a) on earth, (b) in a spaceship for which "g" = 10- 4 ft/sec 2 ?
10.5. Water at 1000 psia is evaporated slowly in a long vertical H-in.-diam tube. If
G = l0 5 lb/(hr)(ft 2), the heat flux is 1000 Btu/(hr)(ft2), and thc water is saturated
at inlet, estímate the void fraction as a function of height. How far up the tube does
slug flow persist?
10.6. Salve Probs. 2.30 and 2.31, assuming that the prevailing regime is slug fl.ow.
10.7. Solve Prob. 4.2 if the flow regime is slug fl.ow.
10.8. Estimate the rms value of pressure fluctuations when air and water fl.ow in a
vertical pipe of 1-in. diam at 30 psia and SOºF if the flow rates are W¡ = 10 4 lb/hr,
W 0 = 20 lb/hr, and the volume of each bubble is 10 in. 3 •
10.9. A large bubble is injected into the lower end of a hydraulic pipeline with 3 ft
diam, 2-mi length, and a vertical drop of 1000 ft. How long does the bubble take to
reach a closed valve at the top of the pipeline?
10.10. A liquid metal (u = 300 dynes/cm, µ¡ = 0.02 poise, p¡ = 5 g/cm 3) fills a
%-in.-diam horizontal pipe. It is desired to blow gas through the pipe to cool the
metal and solidify itas a·uniform film 0.00,5 in. thick on the walls. What gas flow
rate should be used?
10.11. A series of bubbles, each 1 cm 3 in volume, is injected at the rate of 5 per sec
into a flow of 10 cm 3 /sec of glycerol (p¡ = 1.26 g/cm 3, µ¡ = 700 cp, and u = 63
dynes/cm) in a 5-mm-diam horizontal pipe. What arethe bubble length and velocity,
the film thickness around the bubbles, and the pressure drop?
10.12. Estimate the rate of fl.ow of Vermont maple sirup (p¡ = 1.42 g/cm3, µ 1 =
10,000 cp, and u = 77 dynes/cm) from a gallan can through a 1-in.-diam pouring spout
tilted at 45º to the horizontal.
10.13. Predict the circumferential film thickness variation along a long slug-flow bubble
in a horizontal tube as a function of pipe diameter, distance from the bubble nose j,
and fluid properties. Use falling film theory.
10.14. Use Eq. (10.51) to explain why the lines of constantj/v"' in Fig. 10.8 are approxi-
mately concurrent. Show that the intersections are near the point (5/D = H.1..
314 ONE-DIMENSIONAL TWO-PHASE FLOW

j¡/v"" = %). For what value of N 1 will the film thickness around a bubble be
independent of j?
10.15. Draw the streamlines in the liquid slug for the cases C 1 < 2 and C1 > 2 in
laminar flow. Discuss the qualitative effects which can be observed in the two cases.
10.16. Assume that long slug-flow bubbles are flowing in a vertical boiler tube. Figure
10.8 can be used to calculate the falling film thickness around a bubble. As the film
fl.ows clown the wall it evaporates. Show that the maximum bubble length which is
tolerable if the film is not to "dry out" before it reaches the bottom of the bubble is

Evaluate Lb for steam-water flow at 300 psi in a tube of 0.5 in. diam if j 1, 10, or
50 fps and ef, ~ 10' Btu/(hr)(ft').

REFERENCES

l. Dumitrescu, D. T.: Z. Angew Math. JJ1ech., vol. 23, no. 3i p. 139, 1943.
2. Davies, R. M., and G. I. Taylor: Proc. Roy. Soc. (London), vol. 200, ser. A 1 pp.
375-390, 1950.
3. White, E. T., and R. H. Beardmore: Chem. Eng. Sci. 1 vol. 17, pp. 351-361, 1962.
4. Nicklin, D. J., J. O. \Vilkes, and J. F. Davidson: Trans. Inst. Chem. Engrs., vol.
40, pp. 61-68, 1962.
5. Griffith, P.: ASME paper no. 63-HT-20, Nat. Heat Transfer Conf., Boston, Mass. 1
1963.
6. Birkhoff, G., and D. Carter: J. Rat. J.11ech. Anal., vol. 61 no. 6, pp. 769-780 1 1957.
7. Wallis, G. B.: General Electric Company, Schenectady, N.Y., Rept. 62GL130,
1962.
8. Bretherton, F. P.: J. Fluid Mech. 1 vol. 10, p. 166, 1961.
9. Hattori, S.: Rept. Aeronaut. Res. Inst. Tokyo Imp. Univ., no. 115 1 1935.
10. Harmathy, T. Z.: A. I. Ch. E. J., vol. 6, p. 281, 1960.
11. Masica, W. J., and D. A. Petrash: NASA Rept. TN D-30051 1965.
12. Griffith 1 P., and G. B. ·wallis: Trans. AS1l1E J. Heat Transfer 1 vol. 83, Series C,
no. 3, pp. 307-320, 1961.
13. Taylor, G. l.: J. Fluid Mech., vol. 10, pp. 161-16.5, 1961.
14. Moissis 1 R., and P. Griffith: Trans. ASME J. Heat Transfer, Ser. C, vol. 841
p. 29, 1962.
15. Suo, M., and P. Griffith: paper no. 63-WA-96, ASJ\.IE 1 1963.
16. Fairbrother, F. 1 and A. E. Stubbs: J. Chem. Soc., vol. 1, pp. 527-529, 1935.
17. Armand, A.: AERE trans. 828 1 Harwell, England 1 1959.
18. Runge, D. E., and G. B. Wallis: AEC Rept. NY0-3114-8 (EURAEC-1416), 1965.
19. Wallis, G. B.: AEC Rept. NY0-3114-14 (EURAEC), 1966.
20. Griffith, P., and K. S. Lee: ASME paper no. 63-WA-97, 1963.
21. Goldsmith, H. L. 1 and S. G. Mason: J. Colloid. Sci. 1 vol. 181 pp. 237-261, 1963.
22. Nicklin 1 D. J., and J. F. Davidson: paper no. 4, Two-phase Flow Symp., Institution
of Mechanical Engineers, London, 1962.
23. Porteous, A.: Dartmouth College1 Hanover, N.H., unpublished work, 1966.
24. Govier, G. VV., and W. L. Short: Can. J. Chem. Eng., vol. 36, no. 1, p. 195, 1958.
25. Wallis, G. B.: Rept. AEEW-R 142, Atomic Energy Establishment, Winfrith
Heath, England 1 1962.
26. Cox, B. G.: J. Fluid JJfech. 1 vol. 20, part 2, pp. 193-200, 1964.
 11
Annular Flow

lU INTRODIJCTION

In the annular flow pattern a continuous liquid film flows along the wal!
of a pipe while the gas flows in a central ucore." If the core contains a
significant number of entrained droplets, the flow is described as annular
mist, which could be regarded as a transition between ideal annular flow
and a fully dispersed drop flow pattern.
Stratined flow occurs in horizontal or inclined pipes and is topolog-
icaliy similar to annular flow because the two components flow side by
side without mixing. The lack of symmetry in stratified flow makes anal-
ysis difficult although sorne o! the same basic techniques can be applied
to either annular or stratified flow.
Annular flow is the predominant flow pattern in evaporators, nat-
ural gas pipelines, and steam heating systems.
Theories of annular :flow provide an excellent example of the pyra-
mid of analytical techniques that was presented in Chap. l. Correla-
tions, simple models, and integral and differential methods can ali be
315
 ONE-DIMENSIONAL TWO-PHASE FLOW
"'
developed in a hierarchy of complexity. Since each method of analysis
can be applied to either the gas or the liquid, numerous combinations are
possible.
In the following developments we shall be particularly concerned
with showing how the successive levels of sophistication are interrelated.
It will usually be found that the more complex analysis provides a method
far predicting empirical factors in the simpler models, or for evaluating
correction factors which increase the level of accuracy.
Horizontal and vertical flow will be treated separately. However,
general techniques which apply to any flow orientation will be derived
wherever possible.

11.2 HORIZONTAL FLOW

THE BOUNDARIES O'F THE ANNUlAR AND STRATfflED REGIMIES IN HORIZONTAL FLOW

The limits of the various horizontal flow regimes are not yet well under-
stood. A simple plot by Baker 1 (Fig. 11.1) may give an indication of the
general trend but certainly <loes not contain ali the relevant parameters.
Qualitatively, however, high gas flow rates tend to c.ause entrainment of
droplets and high liquid rates to cause the formation of bubbles or slugs
as indicated in the figure.

Bubble ar froth

Strotified

Fig. 11.1 Flow-regime corrclations for adiabatic horizontal two-phase two-

component flow by Baker 1 (Pu and p¡ in pounds per cubic ft, G1 and G0 in pounds
per hour per square foot, a- in dynes/cm). A = [(p0 /0.075)(p¡/62.3)]H;
f - [(73/u)µ¡(62.3/PJ)']fi_
ANNULAR FLOW m

The ordinate in Baker's plot is not dimensionless and cannot be

interpreted physically in its present form. However, it <loes contain the
group (p/'j,) which represents the inertia of the gas. One would expect
the balance between this force and gravity to scale stratification effects,
probably in the form o! the dimensionless group

(11.1)

J;
For the pipe sizes considered by Baker the values o! at the bound-
aries between stratified and wavy flow and between slug and annular flow
are in the range 0.25 to 1.0.
The dispersed flow boundary is governed by the mechanism o! drop-
let entrainment. This has been lound to be virtually insensitive to pipe
orientation and is governed primarily by the drag forces that the gas
exerts on irregularities at the interface. Several regimes have been iden-
tified depending on the relative importance of surface-tension, viscous,
and inertia effects.
For an air-water system at atmospheric pressure the critical gas
velocity lar the onset of entrainment is about 70 lps except at very low
liquid rates when the viscous forces in the thin liquid film inhibit the
formation of large waves. Steen 2 suggests that as long as viscous forces
in the liquid can be ignored the critica! gas velocity is given by the equation

j,µ, (r,_)½ ~ 2.5 X 10-• (11.2)

" Pi

which is compatible with Baker's plot over the range of variables which
are represented. Further discussion of entrainment will be given in
Chap. 12.

CORRELATIONS

The correlations for separated flow which were introduced in Chap. 3

apply to horizontal annular and stratified flows. In particular, lVIar-
tinelli's correlation provides a very rapid method for estimating frictional
pressure drop and void lraction by using Eq. (3.30) with n ~ 3.5, and
Eq. (3.32). Alternatively Figs. 3.4 to 3.8 may be used. The Mar-
tinelli parameters <P1i <Pai and X are also useful for expressing the result~
of more complex methods of analysis.

SIEPARATIED FLOW, ANNULAR GIEOMIETRY MODEL

As in Example 3.3 1 the one-dimensional equations of motion in annular

flow can be developed by using average values of the interfacial and wall
shear stresses. In steady horizontal flow, with no acceleration 1 force
ha.lances for the gas core and the combincd flow relate these shear stresses
"' DNE-DIMENSJONAL TWO-PHASE FLOW

to the pressure gradient. Thus

dp 4ri
-dz=Dy-;;, (11.3)
dp 4'-w
- dz = D (11.4)

THE INTIERFACIAL SHEAR STRESS

The interfacial shear stress will depend, presumably, on the difference

between the gas velocity and sorne characteristic interface velocity. If
the gas velocity is much larger than the liquid velocity, then one may
neglect the liquid velocity and assume in the usual way that

(11.5)

If the same gas flow filled the pipe, the wall shear stress would be
given, in terms of the friction factor (GJ),, by
Twg = (C¡)g7f PuJu 2 (11.6)
and the pressure drop would be
dp) 4Twtl 2(C¡)gp0 j,,2
- ( dz , = D = D (11. 7)

Combining Eqs. (11.3), (11.5), and (11.7) with the definition of ef,,2
in Eq. (3.24) we have
(Cr)i = cé'1>?(Cr), (11.8)
which provides a physical interpretation for the Martinclli parameter ef,/.
An alternative parameter which is sometimes used is the superficial
gas jriction factor. It ís defined as the friction factor which would give
the observed pressure drop if the same gas flow rate fillcd the whole pipe.
Denoting this parameter by (C1),, we have
dp
dz (11.9)

Tbe relationship between (Gr)" and (C1 ), is simply

(C1 ),, = (~;¡, = cp/(C1 ), (11.10)

An explicit expression for (C 1 )i in terms of measurable quantities

follows from Eqs. (11.3) and (11.5). Thus

(11.11)

Figure 11.2 shows a qualitative plot of the interfacial friction factor

versus liquid fraction for both upflow, horizontal flow, and downflow and
ANNULAR FLOW
"'

Increasing
Horizontal liquid flow
ar downflow
Wave inception
"Srnooth"
annular flow
(Cr)og
o~-------o~_-1-------~o~.2-------~oL.3--"
(1-a) liquid fraction

Fig. 11.2 Qualitative aspects of the pressure-drop characteristics of various flow

regimes represented in terms of the apparent interfacial friction factor. 3

serves to indicate the various important regimes of operation. 3 In hori-

zontal or downflow at low gas velocities the liquid film is smooth and the
friction factor is approximately the smooth pipe value. Above a critica!
gas velocity for wave inception the friction factor rises rapidly to a maxi-
mum after which it tends to follow the line marked "rough annular flow."
At higher gas velocities the tops of the waves are sheared off to form
entrained droplets and the density of thc core increases. This leads to
an increase in the friction factor, in about the ratio pjPo where Pe is now
the average homogeneous density of the core.
At very high flow rates of both phases almost ali of the liquid flow is
entrained and the prcssure drop and void fraction are given by homo-
geneous flow theory as

(11.12)

In homogeneous flow the liquid fraction is

1-a (11.13)
320 ONE-DIMENSIONAL TWO-PHASE FLOW

In annular flow the volumetric fl.ow rate of the gas is usually much greater
than the liquid flow rate and these equations can be approximated by

- dp
dz v .
= C172PeJ, '(1 + G1) 4
G, D (11.14)

1 - a = Íf-
Jg
= Po_ G1
Pi Gu
(11.15)

Using Eq. (11.15) in Eq. (11.14) we find that

- : = C172p,j,' [ 1 + ~ (1 - a)] ~ (11.16)

Since the friction factor in homogeneous flow is relatively unchanged

from the value for gas alone, (C1 ),, we can use Eq. (11.16) to show that
the apparent value of the interfacial friction factor in homogeneous flo\v is

_ dp Da_%, =
dz 2puJri
(C 1), [1 + PiPu (1 - a)] a% (11.17)

For air and water at atmospheric pressure p¡/ Pu ~ 800, and for homo-
geneous flow the value of Eq. (11.17) increases very rapidly as a function
of 1 - a, as shown in Fig. 11.2. The curves tend to move toward
the homogeneous flow line as the liquid rate is increased and the
percent entrained goes up.
In vertical flow the interfacial shear has to support the film against
gravity. Furthermore any slugs of liquid or entraincd droplets add a
gravitational component to the core pressure drop. Thus, at low gas
rates in the slug-f!ow region the friction factor defined by Eq. (11.11)
increases well above the annular flow value. At lower liquid rates this is
usually at values of liquid fraction above about 0.2. At high liquid rates
the transition between slug flow and annular mist flow becomes obscure
and there is a general motion toward the homogeneous flow line.
The key to a simple analysis of "rough" or uwavy" annular flow
is a plot of the interfacial friction factor versus the dimensionless film
thickness as is shown in Fig. 11.3. The points cluster pretty well around
a line with the equation

0
(C1)i = 0.005 ( 1 + 300 ~) (11.18)

For thin films in pipes this is approximately the same as

(C1), = 0.005 [l + 75(1 - a)] (11.19)
We note at this point that the rough pipe correlations of Nikuradse 4
and lVIoody 5 can be approximated by the equation

C1 = 0.005 ( 1 + 75 ~) (11.20)
ANNUlAR FLOW m

o Martinelli, 1"-dia. horizontal

o
0.0625 • Dukler, 1"-dia. vertical
{j, Sze Foo Chien, 2"-dio. vertical
□ Charvonia, 3"-dio. vertical
- - Eq. (11.18)
:: 0.0500 - - - Eq. (11.19)
¡;
-2

•
0.0125
••
0.005

ºo 0.01 0.02 0.03

Dimensionless film thickness, 8/D
0.04

Fig. 11.3 Comparison between Eqs. (11.18) and (11.19) and various air-water
data. 3

over the range 0.001 < k,/D < 0.03, where k, is the grain size of a "sand
roughness." Equation (11.18) therefore shows that a wavy annular film
is about equivalent to a sand roughness of four times the film thiclmess.
Using Eqs. (11.19) and (11.11) we obtain for horizontal flow

- dp) = 10_, p,j,' 1 + 75(1 - a) (ll.Zl)

( dz F D a%

Equation (11.19) may also be used to give an analytical expression

for the Martinelli parameter </,,2. Assuming that (C¡), = 0.005, we have
from Eqs. (11.8) and (11.19)

cf:,y
= [1 + 75(1 -
a%
a,)]l' (11.22)

The alternative representation in terms of the superficial gas friction

factor is

( e) = O 00" 1
/su,¡)
+ 75(1
a%
- a) (11.23)

A simpler equation which gives much the same result up to 1 - a = 0.1 is

(C1 ),, = 0.005[1 + 90(1 - a)] (11.24)
 ONE-D1MENSIONAL TWO-PHASE FLOW

0.05
\
-~
(:-
1

-:: 0 . 0 4 f - - - - - - - + - - - - ~ , - ; : , 4 - - - - - + - - - - - + - + - - - - -
t
-E \
e e, = 2_541
·i 0.03 lb/l,ecllttl 2
:.E e, =1.814
O Gf =1.181
~ o.02f------,""----r'--+~c------~+---------+-----+
w
~
o
~

O.DI

0.005
ºo'----~~--,J~-~~---,--L~-----,--,'-c-c---~----c-¡-=-'
0.005 0.01 0.015 0.02 0.025 0.03 8 D
O 0.02 0.04 0.06 0.08 O.ID 0.12 (i-a)

Fig. 11.4 Comparison between Eq. (11.24) and results of Chien and lbele 6 for
vertical downflow.

and is easy to remember by the rule of thumb that a líquid fraction of

),{ 0increases the pressure drop by a factor of 10. The above equation
is compared in Fig. 11.4 with experimental curves of Chien and Ibele. 6
Exampie 11.1 Air and water flow in a 1-in.-diameter horizontal pipe. The gas mass
flow rate is 300 lb/hr and p 0 = 0.1 lb/f"t 3 • lf thc void fraction is 90 percent,
estirnatethepressuregradientusing (a) Eq. (11.21), (b) Eqs. (11.23) and (11.9).
Solution The area of the cross section is

(
4
)(1 44 ) ~ 5.45 X 10-s ft'
Therefore
. 300 1 1
Jq = 3600 0.1 5.45 X 10- 3 = 153 fps
(a) Using Eq. (11.21) -.,ye have
dp) _ _, (0.1)(153') 8.5 _ ·¡¡
( - dz F - 10 (32.2)(144)(H 2 ) 76S - 0 ·6 7 psi t
(b) The superficial friction factor, from Eq. (11.24), is

(C1 ),. ~ 0.05

whence, using Eq. (11.9),

- dp ~ O O"0 2 (0.1)(153') ~ 0.61 psi/ft

dz . >f 2 (144) (32.2)
ANNULAR FLOW m

These expressions for the interfaci.al friction factor can also be com-
pared with Levy's 7 correlation which is expressed in terms of a function
F' as follows:
7
F'
_ -- [ v,(v, - v1)(p¡
'
- p,) ] " R (11.25)

R is an empirical function-of density ratio which can be represented quite

well3 by the equation

lV[aking thc following approximations, which are consistent with

the leve! of sopbistication o! the present theory,

PJ » P,
Va>> VJ

Equations (11.2.5) and (11.26) can be combined to give

(11.27)

whence, using Eq. (11.5),

F' = (C;')" (11.28)

Levy correlated F' versus the ratio o! the average film thickness
to the pipe radius. In view o! Eq. (11.28) this is just what was done in
Fig. 11.3 apart from a constant factor. The two theories can therefore be
compared directly by substituting Eq. (11.18) into Eq. (11.28). Com-
parison with the results o! Wicks and Dukler 8 is shown in Fig. 11.5.

nn: WALL SH EAR STRESS

Perlorming a similar analysis lor the liquid we find that the equivalent
o! Eq. (11.8) is

(11.29)

where (C1)w is the wall lriction factor for the film and (C1)1 the lriction
factor for thc liquid alone in the pipe. One might expect that (C 1 )w =
(C 1) 1 , since the wall roughness is the same in both cases, and Eq. (11.29)
then gives an expression for 1>1 as follows:

1
</,¡=-- (11.30)
1 - a
 ONE-DIMENSIONAL TWO-PHASE FLOW

Wicks-Dukler 8 l" tube

o 521 16/hr water flow
~ 0.2 ◊ I042
V 1563
t
\j
ó.
a 2605
2084

11 0.1 C - - - ◊ 3126 - - - - - - - - - - - l . - ' " ' - - - ~

"--
e
o
-~ 0.06
e°
0.04
- Levy 7
Present theory,
Eq. (11.18)
0.02'-~~~~---~-~~-~~~~--~
0.004 0.01 0.02 0.04 0.1 0.2
Ratio of liquid film thickness to pipe radius, 28/D

Fig. 11.5 Comparison between Levy's correlation, Eq. (11.18)

and data for horizontal fl.ow.

Equation (11.30) has been found to be approximately valid by numerous

investigators; it is also in agreement with the results of the integral
analysis which will be presented later. Moreover, it will be found that
the wall friction factor is almost the same function of the liquid Reynolds
number as it is for single-phase flow, apart from the transition region.
The liquid Reynolds number is defined as

Re1 = p¡j¡D (11.31)

µ.¡

In general, from Eq. (11.4)

dp) 2p¡j¡' (11.32)
( - dz F = (C¡)w (1 - a)'D
where, for laminar flow,
16
(C1)w = Re¡ (11.33)

andas a simple first approximation in turbulent flow, (C1 )w can be taken

to be 0.005, as usual.

EVALUATION OF PR.ESSURE OROIP AND VOiD FRACTION

The above relationships permit the evaluation of the interfacial and wall
shear stresses in Eqs. (11.3) and (11.4). These two equations can then
be solved simultaneously for the pressure drop and void fraction. For
 ANNULAR FLOW

"
1

e
-i 0.11-------+cc..~+,-"'-;c;-----+-------+-------__j
1
70
·s /
rr /
::, /
'/ - - - Martinelli 9
/
/ - - - - Eq. (11.35)
/
/

/
/
0.01 ~~'~-~~~~~-~~~~~~-~~~~~~-~-~~~=
0.01 0.1 10 l00
Mortinelli porometer, X

Fig. 11.6 Comparison of liquid fraction predicted by Eq. (11.35) with empirica
results of Martinelli.

the void fraction in turbulent-turbulent flow 1 for example, we eliminate

(dp/dz)F between Eqs. (11.32) and (11.21) to gct
(1 - a) 2[1 + 75(1 - a)] p1j¡2
= -.-, (11.34)
Pu)r;

Alternatively, in terms of the Martinelli parameter,

(1 - a) 2[1 + 75(1 - a)] = X' (11.35)
al'

for either laminar or turbulent flow of the liquid. Equation (11.35) is

plotted in Fig. 11.6 and compares well with Martinelli's empírica! curve.
In practice, X 2 is readily calculated from known quantities; a can
then be determined using Eq. (11.35). The pressure drop follows most
rapidly by using this value of a in Eq. (11.30).

IExample 11.2 Salve Example 3.3 using Eqs. (11.35) and (11.30) if flow is horizontal
and the liquid flow rate is reduced by a factor of 3.
Solution The liquid Reynolds number is 23,000 /3 = 7700; therefore flow is turbulent
and the friction factor from single-phase .fl.ow charts is 0.009. The single-phase
liquid frictional pressure drop is

(- dp) _ (2)(0.009)(31.5
2
)(.1) _ d / ,
dz 2.48
f - - 7 . 2 ynes cm

The gas phase pressure drop, as befare, is 9.4 dynes/cm 3

Therefore
 ONE-D!MENSIONAL TWO-PHASE FlOW

Equation (11.3.5) is now solved by iteration to give l - a = 0.18. From Eq.

(11.30) 1 therefore,
1
~f ~ 0.18 ~ 5.55

whence the pressure gradient is

EXTENSION TO THE GENERAL CASE

These expressions for the wall and interfacial shear stresses can be used
in the general equations of Example 3.3 to form the basis far an analysis
of annular flow in the presence of body forces, area change, phase change,
and compressibility effects. Sorne special cases of this technique will
be developed later.
IMPROVEMENTS TO THE THEORY

Up to now we have considered only the effect of the major variables on

the dynamics of annular flow. Balances have been made between the
dominant forces, whereas secondary effects have been neglected. In gen-
eral, however, there are at least seven different interacting forces which
are due to the pressure gradient, buoyancy, surface tension, and the
inertia and viscosity of each phase. A complete representation of the
interplay between ali of these forces is not available; however, the develop-
ments which will be considered in the next few pages provide an example
of successive methods for modifying the theory in order to develop a more
accurate 1 but more complex, theoretical structure. Further modifica-
tions will be considered in the discussion of vertical annular flow.
INTEGRAL ANALYSJS

The basis of an integral analysis of annular flow is the assumption of a

velocity profile. This profile is integrated over the portien of the pipe
which is occupied by each phase in order to give the flow rates. The
wall and interfacial shear stresses are related to the profile by means of
single-phaseflow correlations which are already well established. Because
of the complications which are introduced by the boundary conditions for
the gas at the wavy interface, this analysis has been most successfully
developed for the liquid film in symmetrical annular flow.
THE LIQUID FILM

Suppose that there were a single-phase liquid flow with mean velocity ii
which occupied the whole pipe. The frictional pressure drop would be

( - dp)
dz F
= (11.36)
ANNULAR FLOW

where the friction factor C1 is a function of the Reynolds number

(11.37)

N ow suppose that only a part of this single-phase flow is present, namely,

from the wall inward to a radius ya ro, and that the rest of the pipe is
occupied by gas, a droplet-laden core, or even a plug of froth ar foam.
If the velocity profile is u(r), then the liquid flux in this annulus is

.
J¡ = - 1 2 /'"_ 2-n-ru dr
7rTo y'ar~
= f
~
i 2r *u dr*. (11.38)

For laminar :flow we have

u = 2v(l - r*') (11.39)

and Eq. (11.38) gives eventually

j1 = (1 - a)'v (11.40)

On the other hand, far turbulent flow with a one-seventh power velocity
profile we have
u = B¾ 9 v(l - r*)" (11.41)
and, from Eq. (11.38),

j¡ = (1 - y;;)!l(l + ½ y;;)v (11.42)

Both Eqs. (ll.40) and (11.42) are of the general form
j 1 = J(a)v (11.43)
where J(a) is sorne function of a which could be obtained by any assump-
tion about the velocity profile.
In a two-phase flow, if j 1 and J are known, then ii can be calculated
and the result used in Eqs. (11.36) and (11.37) to calculate the pressure
drop.
These results are readily related to the Martinelli parameter ,¡, 1 far
the liquid. Far laminar flow we have
dp) _ 32vµ 1
( - dz , - D' (11.44)
and
- dp) = 32j1µ 1 (11.45)
( dz I D'
 ONE-D1MENSIONAL TWO-PHASE FLOW

Comparing Eqs. (11.40), (11.44), and (11.45) we find

,_(-t),_ 1
(11.46)
</>¡ - ( - dp) - (1 - a) 2
dz 1
whence
(11.47)

For turbulent flow, on the other hand, the Blasius equatíon 10 (which is
compatible with the one-seventh power law) gives
C1 = 0.079 Re- 0 • 2 ' (11.48)
and, from Eq. (11.36),

( - dp) (.)"!___)¼
2
= 0.158p¡v (ll.49)
dz , D vDp1
Using Eq. (11.42) to find (dp/dz)¡ by puttingj1 for v in Eq. (11.49) and
again evaluating q:,1 we find that 11 , 12
1
q,¡ = ---=----=- (11.50)
(1 - ,Va)(l +%ya)%
Table 11.1 sbows that the difference between Eqs. (11.47) and (11.50) is
negligible for most practica! purposes and that one might as well use Eq.

Table 11.1 Comparison between the predictions of Eqs. (ll;li7) and (11.50)
and Martineili's relation 1,¡u for turbulenM:urbuient flow 11

1 % 1 %
a 1'111 f¡ ~ - - f¡ ~
1 - a error (1 - Va)(! +%Va)" error

0.96 24.4 25.0 2.5 25.7 5.1

0.95 18.5 20.0 8.1 20.5 9.8
0.91 11.2 11.1 -0.9 11.4 1.8
0.86 7.05 7.14 1.0 7.32 3.7
0.81 5.04 5.26 4.4 5.39 6.5
O. 77 4.20 4.35 3.6 4.45 5.6
0.69 3.10 3.23 4.2 3.29 5.8
0.60 2.38 2.5 5.0 2.55 6.7
0.52 1. 96 2.08 6.1 2.12 7.6
0.47 l. 75 1.89 8.0 1.92 8.9
0.34 1.48 1.52 2.7 1.53 3.3
0.24 1.29 1.32 2.3 1.33 3.0
0.16 1.17 1.19 l. 7 1.20 2.5
O.JO 1.11 1.11 o 1.12 0.9
ANNULAR FLOW

(11.47) far turbulent as well as for laminar flows. Also shown for com-
parison are Martinelli's correlated values for ,¡, 1 in the turbulent-turbulent
regime (,J,m).
The result of this analysis is to confirm Eq. (11.30) for both laminar
and turbulent horizontal annular flow. Furthermore, Eq. (11.49) can be
rearranged with the help of Eq. (11.43) to give an expression for the wall
friction factor, as follows:

(C1).
_
- 0.079 Re,
-%
,.
(1 - a) 2
J'" (11.51)

Since J¼ is almost identical with 1 - a for the one-seventh power profile

in view of Table 11.1, this equation reduces to

(11.52)

which shows that the wall friction factor can be correlated by the Blasius
equation in terms of the Reynolds number based on the overall liquid
flux j, [Eq. (11.31)]. This represents an improvement over the simple
assumption that (C1). is always equal to 0.005.

THE GAS CORE

The case of a viscous gas core is of little practica! interest. In fact the
integral analysis yields exactly the same result as the more fundamental
differential analysis which was presented in Example 5.1.
A solution for a turbulent core which is not far from the predictions
of Eq. (11.22) at a higher value of film thickness was derived by Tumer 11
who assumed a one-seventh power law based on the ratio of the radius to
the overall pipe radius. In addition, he required that the gas velocity
should equal the liquid velocity at the interface. The liquid interface
velocity was, in turn, derived from a further one-seventh power law in
the film.
The result equivalent to Eq. (11.43) was

j, - v, ( 1 - (1 - ya)!l(l + 1/7 ya)

_ «(1 - ya)ll [1 _
0.817
(P');, (µ.')li])
P! µ.,
(11.53)

giving a value of ,¡,, as follows:

,¡,, - ( 1 - (1 - ya)"(l + '½ ya)

- a(l -
0.817
ya)li [1 -(P');' (µ.')H] ¡-¼
Pi µ.,
(11.54)
 ONE-DIMENSIONAL TWO-PHASE FLOW

[)IFFERENT!AL ANALYSIS

A differential analysis for horizontal laminar symmetrica.l annular flow

proceeds as in Example 5.1 except that cos 0 = O. Alter sorne algebra
the Martinelli parameters are lound to be

(11.55)

(11.56)

Simple relationships between 'Pi and ,t,, result in the limiting cases where
µ, « l'i and µ, = µ¡. In the lormer case Eq. (11.56) reduces to

(11.57)

and this coupled with Eq. (11.55) gives

_!_ + _l_ = 1 (11.58)

'Pi 'P,
In the latter case Eq. (11.56) becomes
1
<t,,' = a(2 - a) (11.59)

Taking Eqs. (11.59) and (11.55) together we find that

1 1
<t,,' + <t,¡2 =1 (11.60)

Equations (11.58) and (11.60) are special cases o! Eq. (3.30).

A differcntial analysis o! the liquid film which covers the whole range
o! Reynolds numbcrs was perlormed by Hewitt, 13 using Eq. (5.4) up to
y+ = 5, Eq. (5.10) for 5 < y+ < 20, and Eq. (11.78) in the turbulent core.
The result can be expressed as tbe relationship between wall lriction factor
and liquid Reynolds number which is shown in Fig. 11.7. The differen-
tial analysis agrees well with the integral analysis in both laminar and
turbulent flow and has the additional advantage o! including the "transi-
tion" region.

11.3 COUNTERCUllllENT VERTICAL ANNULAll FLOW

Vertical annular flow is free from the stratification effects which can occur
with horizontal flow but is complicated by the action o! gravity, which
produces a body force in the direction of flow, and also by the possibility
o! upward or downward flow o! either component. Although the tech-
niques which apply to horizontal flow also give reasonably accurate pre-
 ANNULAR FLOW 331

e 16
f..,= Re¡

~
~ 0.1 f---------C----'~-------+--------+-------"-------l

Prediction of Hewitt's ana!ysis 13

0.,."" 0.079 Re¡ 0 •25 (Blasius)

O.OOl c_-'-....L.L.LLilu...__¡__J__LJ_L.l_lil-__J_J_--Ll...l..lJ_J_J_~L_l....LLilliL-j__LJ_Ul.llJ
10 10 2 10 3 104
Liquid Reynolds number, Re¡

Fig. 11.7 Wall friction factor in horizontal annular flow predicted from integral
analysis [Eq. (11.52)] and Hewitt's differential analysis. 13

dictions in vertical flow at high values of" and sufficiently high velocities,
when gravitational effects are relatively small, it is misleading to neglect
gravity at lower gas flow rates close to the slug-flow regime boundary and
in countercurrent flow.

F'Al.LING FiU\11 FLOW

Falling film flow is a simple case of vertical annular flow. II the gas veloc-
ity is sufliciently low, interfacial shear stresses and pressure drop are both
negligible and the governing equation for thin films in which wall curva-
ture can be neglected is

7 = g(p¡ - p,)(ó - y) (11.61)

where ó is the film thickness and y the distance fro m the wall.
For laminar flow Eq. (11.61) becomes

dv (11.62)
µ¡ dy = g(p¡ - p,)(ó - y)

Integration gives

µ¡v = g(p¡ - P,) (•Y - t') (11.63)

 ONE-DIMENSIONAL TWO-PHASE FLOW

Integrating again over the film, the volumetric flow per unit width 14 is

(11.64)

The mass flow per unit width is usually denoted by r where

r = P1q1 (11.65)
The corresponding total volumetric flow rate of a thin film in a
tube of diameter D is
Q¡ = ,r gD(p 1 - p,) ó8 (11.66)
3 µ¡

and the liquid volumetric flux is

. 4 g(p¡ - p,)ó'
Jt =3 µ D (11.67)
1

Equation (11.67) can be converted to the equivalent form

·• = 34N(º)'
J¡ 15 1 (11.68)

Altematively, a film Reynolds number can be defined as

4r
Rer = - (11.69)
µ¡

and a dimensionless film thickness as

¡¡* = ógll(p¡ - p,)llp¡ll
(11.70)
µ./3

In a circular tube Eqs. (11.69) and (11.70) are equivalent to the following:
j¡p¡D
R er = - - = J¡'*N 1 (11.71)
µ¡

o* =·iN¡l' (11. 72)

Equation (11.64) can then be rewritten as

ó* = 0.909 Rer" (11.73)
A plot of experimental data on a basis of ó* versus Rer shows agreement
with Eq. (11.73) up to a value of Rer = 1000 (Fig. 11.8). At higher
Reynolds numbers the film becomes turbulent and a new correlation is
necessary. A line through most of the data in Fig. 11.8 has the equation
ó* = 0.115 Rerº·' (11.74)
ANNULAR FLOW m

•00 Eq. (11.74)

Reynolds number

Fig. 11.8 Effect of Reynolds number on falling film thickness with zero
shear at the liquid interface. (Belkin. 15)

A suggested alternative correlation proposed by Belkin 15 has the form

o* = 0.315 (Rer ,.,;e;)% (11.75)

in which C1 is the appropriate friction factor. For a value of C1 = 0.008

this equation becomes

o* = 0.063 Rerh (11.76)

which is also shown in the figure and gives a good fit to Belkin's own data.
An advantage of Eq. (11.76) is that when it is combined with Eqs.
(11.71) and (11.72) the viscosity dependence disappears and we find

~= 0.063j¡"% (11. 77)

On the other hand, Eq. (11.74) reflects the usual dependence of the fric-
tion factor on the -0.2 power of the Reynolds number.

E.xampie 11,3 Water at 70ºF flows down a vertical surface at a rate of 10 lb /min per
foot width. What is the film thickness?
Solution At 70ºF the viscosity of water is 2.5 lb/(hr)(ft). The fl.ow rato per unit
width is 600 lb/(hr)(ft). Therefore, from Eq. (11.69),
Rer = (4) (600) = 960
2.5
 ONE-DlMENSlDNAL TWO-PHASE FLOW

Flow is therefore just laminar and Eq. (11.73) applies. The value of ó* is
,. = (0.909)(960)¼ = 8.96
Therefore, from Eq. (11.71),
(2.5')
' = (8.96) [ (32.2)(62.5')(3600')
]¼ = 3.02 X 10_'ft
A differential analysis of falling films in cocurrent flow was per-
formed by Dukler" using Eqs. (5.9) and (5.10) up to y+ = 20 and von
Kármán's expression for the eddy viscosity at larger values of y+. Thus

y+> 20 • = 0.13 (::y (~~,r, (11.78)

U sing these relationships and following the overall plan outlined in

Chap. 5, the velocity profile and film thickness were derived. The results
shown in Fig. 11.9 give ó* versus Re1 for various values of a dimension-
less interfacial shear parameter ¡3. For downflow in tubes the value of
/3 is
(11.79)

The relative importance of interfacial shear and gravitational effects is

scaled by the ratio 4/3/ ó* which, far thin films, is equivalent to - llp* /
(1 - a) and determines the shear stress variation across the film.

10,000 100,000

Fig. 11.9 Dimensionless film thickness ó* versus Reynolds number forvarious

valu es of dimensionless pressure drop ¡3. (Dukler. 16)
ANNULAR FLOW

STABJUTY OF FAlUNG ffflll..MS

On a vertical surface there is no restoring force for long wavelcngths, and

the dynamic wave velocity becomes zero. However, the continuity wave
velocity is found as in Example 6.1 to be

(11.80)
a result which was deduced by Benjamin 18 and compares with the value
2.5v1 derived by Kapitsa 19 and Portalski. 20
Since the continuity wave velocity is finite while the dynamic wave
velocity is zero, except for surface-tension effects on short wavelengths, a
falling film is always unstable in view of Eq. (6.132). The conclusion
was reached by Benjamin 18 who also showed that the rate of growth of
instabilities was strongly dependent on the Reynolds number. Severa!
authors have claimed that there is a "critica! Reynolds N umber" for the
onset of instability but this is probably due to the fact that the rate of
growth of disturbances is so small at low Reynolds numbers that insig-
nificant amplification is noticed in a finite apparatus.
Hewitt and Wallis 21 conducted experiments in which the amplitude
of waves on a falling film was measured near the top and bottom of a
vertical tube 13,,:;i: in. in diameter and 3 ft long. The wavc amplification
whieh occurred in this distance is clearly shown in Fig. 11.10 which shows
a trace of film thickness versus time ata given Iocation derived from meas-
urements of the local longitudinal electrical conductivity. The water was
made conducting by adding a small amount of salt. Even though the

f+---2 secs -------J

_i_
0.01"
T
Top of tube
(o)

Bottom of tube
(6)
Fig. 11.10 Conductance probe traces showing growth of
falling film instability in a L25-in.-diam tube 3 ft long.
(a) Top probes. (b) Bottom probes. Water rate =
300 lb/hr, wave velocity = 1.196 fps, air rate = 5 scfm.
(Hewitt and Wallis. 21 )
336 ONE-DIMENS!ONAL TWO-PHASE FLOW

film was noticeably wavy at the bottom of the tube, Eq. (11.73) was still
found to be valid for describing the average film thickness.
In practice 1 as shown in Prob. 7.4 1 a whole variety of wavelengths
and frequencies are unstable and they ali move with velocities less than
Vw in view of Eq. (6.126). These waves interact in a rather chaotic way.
Hewitt and Wallis" found that many wave velocities could be distin-
guished if individual disturbances were followed on a falling film. The
trace shown in Fig. 11.10 is also evidence that a single wavelength does
not predominate.

11.4 FLOODING

If gas is blown upward through the center of a vertical tube in which

there is a falling film, a shear stress which retards the film is set up at
the interface. As long as the film remains fairly smooth and stable, this
shear stress is usually small and the film thickness, and consequently the
void fraction, is also virtually unchanged from the value which is obtained
with no gas flow. 21 , 28 However, for a given liquid rate there is a certain
gas f!ow at which very large waves appear on the interface, the whole
flow becomes chaotic, the gas pressure drop increases markedly (Fig.
11.11), and liquid is expelled from the top of the tube. This condition is
known as flooding. U nlike the phenomenon which occurs in dispersions
of drops or bubbles, the flooding point is not approached as the limit of
a continuous process but is the result of a sudden and dramatic instability
which increases the pressure gradient by an order of magnitude.

EMPHUCAL FLOOD!NG CORIRELATIONS

Turl>ulenl llow in bofü componenls Let the gas flux upward and the
liquid flux downward be j, and jr, respectively. An empírica! flooding
correlation can be sought in the following way. Countercurrent flow is
maintained by buoyancy forces dueto the density difference between the
gas and the liquid. The flow rates are related to the film thickness by
dynamic processes that balance the driving force of buoyancy with dis-
sipative effects in the fluids. By analogy with single-phase f!ow turbulent
systems it can be assumed that the average turbulent stresses are related
to the average momentum fluxes of the components, i.e., to the quanti-
ties p,j,'/a and p¡j12/(l - a). Dimensionless groups which relate these
momentum fluxes to the hydrostatic forces 1 apart from any dependence
on the dependent varia ble a 1 are

j: = j,p/'[gD(p¡ - p,)J-¼ (11.81)

i7 = j¡p¡¼[gD(p¡ - P,)J-¼ (11.82)
ANNULAR FlOW m

300

Total pressure, 20 psia

Tube diometer, 1.25 in.
Locus of flow
reversol ---'?é?X
1
Water flows
both upword
ond downword ~
' ;¡¡ 10-21--------+-----~~4-'9'------- /

l,
•
e
'o •
ern

Water rote,
16/h,
\ -----:r-7"-;ry';''--f-----------¡---------1

10 l00 1000
Gas flow rote, 16/hr

Fig. 11.11 Pressure-drop characteristics near the boundary between

countercurrent. and cocurrent flow. (Hewitt, Lacey, and Nicholls. 22 )

A general correlation in terms of these parameters can then be sought

in cases where the effects of other forces, such as those due to viscosity or
surface tension, can be neglected. This hypothesis receives considerable
support from comprehensive studies of flooding in packed towers. For
example, Lobo et al. 23 and Sherwood et al. 24 present their results as a plot
of the variables
. 2
Jo a Po o.z
---µ¡
g F' Pi versus Íj_
Jo
(Pi)"
Po
(11.83)
 ONE-DIMENSIONAL TWO-PHA.SE FLOW

where a/Fª is a characteristic inverse length which is equal to the total

packing surface area divided by the total available flow volume.
The first parameter can be identified with 2 and the second with J;
irh: if, as is usually the case, the assumption (p¡ - p,) = p1 is made and
a/Fª is assumed to be the inverse of a characteristic length D. Alter
rearranging the coordinatesJ the curves of both Lobo and Sherwood can
be drawn on a graph of versus i7 i;
as shown in Fig. 11.12. A good fit to
both curves is given by the equation

(11.84)

The form of this equation is comparable with the results of Del! and
Pratt 25 who measured fl.ooding rates for various liquid-liquid combina-
tions in packed columns. Flooding data plotted on a basis o! the square
roots of the volumetric fluxes were found to lie on straight lines with
approximately equal intercepts on the axes. The final correlation was

1 + 0.835 ( Pd
p, )" (.~J, )¼ = C [g(p, -. p,) Fa3
74 Pc]c 2 cr¾ ] ½4 (11.85)

where the subscripts referred to continuous and discontinuous compo-

nents. If the surface-tension factor is absorbed into the coefficient C,

o
o 0 Points on Sherwood 's line
x Points on lobo's line
-Eq. (11.84)

X O

/J;- 0.41------+--"-"'-;x',x,--I-----I------
o
X

o
0.21-------1-----1----'-.--ol-----

o L.__ _[___ _l___ __[__ __j__ _j__ __L_ __j_...:__

o 0.2 0.4 0.6 0.8
,¡;;,
Fig. 11.12 The floodíng correlations of Sherwood 24 and
Lobo 23 plotted on a basis of the square roots of the param-
eters jj and j:.
ANNULAR FLOW

Eq. (11.85) may be multiplied throughout by j;½ to yield

j;" + o.835 j;½ = e (11.86)

and the similarity with Eq. (11.84) is evident.

The correction factor far viscosity which was used by Sherwood and
Lobo is insignificant for most liquids and only amounts to a viscosity
ratio raised to the one-tenth powermultiplyingj;. Similarly the surface-
1

tension factor in Eq. (11.85) is raised to the one-sixteenth power and has
little importance.
Correlations far flooding in vertical tubes resemble Eqs. (11.84) and
(11.86) and may be expressed in the general form 26

(11.87)

For turbulent flow mis equal to unity. The value of C is found to

depend on the design of the ends of the tubes and the way in which the
liquid and gas are added and extracted. For tubes with sharp-edged
flanges 1 C = 0.725 1 whereas when end effects are minimized 1 C lies
between 0.88 and 1. 27 • 20 In inclined tubes the flow rates at the flooding
points can be much higher.
Sorne hysteresis is observed in many flooding experiments. For
example, if a smooth film is set up in a vertical tube, then the gas flow
rate necessary to cause flooding coincides with the flow rate which will
cause the growth of a large wave. Once the wave has formed it disrupts
the whole film which becomes very rough and is covered with waves sev-
era! times its own thickness in amplitude. The flow rate has to he reduced
to a considerably lower leve! befare the tube will return to smooth operat-
ing conditions. The range of hysteresis seems to lie between C = 0.88
and C = l, as shown in Fig. 11.13.
Once the tube is dried out, it is found that the gas flow has to be
reduced below tbe value characteristic of sharp-edged tubes

(11.88)

befare tbe liquid will flow back down the tube."

Viscous llow in the liquld If the liquid is very viscous the term due to
liquid inertia in Eq. (11.83) should be replaced by a term which is propor-
tional to viscous forces. The ratio between viscous and buoyancy forces
is given by the expression

jj - µ,¡J¡
(11.89)
N¡ - gD 2 (p¡ - p,)
 ONE-DI MENS!ONAL TWO-PHASE FLOW

□
'- o
0
0.41-----____j----____j-"-''-'-_.CC:-'?=-------l
"
C= 0.88 _.,/';'
"'-
'-
0.2 f--------l--------l-----+--'-c------'1

0L-_L__L__L__L__~-~~-~
O o. 2 0.4 0.6 0.8
/Jl
o l "-dio. ( Nicklin and Dovidson 28 ) "short column"
well-rounded air inlet design.
x ]"-dio. (Nicklin and Dovidson 28 } "moin column"
well-rounded air inlet design.
21
6. l¼"-dia. (Hewitt and Wallis ) water injection
ond extroction through porous walls. Points
where flooding storts with increosing flow.
□ 11//-dio. (Hewitt and Wollis 21 ) points where
flooding stops with decreosing flow.
+ ¾"-dio. (Wallis, Steen, ond Brenner 29 )
flooding storts.
v 3/ "-dio. ( Wallis, Steen, ond Brenner 29 )
4
flooding stops.

Fig. 11.13 Flooding velocities for air and water in vertical

tubes designed to minimize end effects. All data at
atmospheric pressure.

Far N 1 < 2 Wallis" suggests that the appropriate flooding relation-

ship for the rough wavy film condition in tubes with smooth flanges is

j:¼ ( '*)" = 0.725

+ 5.6 J.;f (11.90)

In general, flooding for any fluid viscosity is correlated by Eq. (11.87)

where m and C are functions o! N 1 as shown in Figs. 11.14 and 11.15.
ANNULAR FLOW
"'
I0r-c--.-----,----,------r-----,
5
3

0.5
Eq.(11.90)
0.3

0.1 L__ _j__ _ _ __j__ _ _ _ _c__ _ _ _.J__ _ __

100 l000 10,000

Fig. 11.14 Variation of the coeffi.cient m in Eq. (11.87) as a function of the

dimensionless group N¡. (Wallis. 26 )

The reduction of the coefficient C far viscous fluids to the same value
which is characteristic of sharp flanges in the inviscid fluid case is prob-
ably due to end effects. 34
Figure 11.16 compares sorne flooding data with this correlation
scheme.

Example 11.ll Air and water are in countercurrent flow in a vertical pipe of diameter
2 in. If Pu = 0.1 lb/ft3 and p¡ = 62.5 lb/ft3, what is the maximum allowable
liquid flow rate when Wa = 200 lb/hr? The value of µ¡ is 2.5 lb/(hr)(ft).
Solution The cross-sectional area is

(4J¡i¡ 4 ) - 2.18 X 10-, ft'

0.9 1 1
' '

,/
,,,,-- --·
0.8 '
/
e
'

0.7
'
3
,--,- - ,
30
I
'
300
'
l000 3000 IO, 000

Fig. 11.15 Variation of the coeffi.cient C in Eq. (11.83) as a function of

the dimensionless group N 1 . (Wallis. 26 )
 ONE-DlMENS!ONAL TWO-PHASE FLOW

º~---~---~----~--~
O 0.1 0.2 0.3 0.4
0T
Fluid Symbol Viscosíty at N
10ºC, cp
Glycerol 99%±1% o 3000 3.4
" 95% □ 1270 8.2
" 90% o 498 21
" 80% ◊ 116 90
" 75% o 60 160
" 70% V 39 250
" 60% • 17 560
" 50% • 9 1000
" 33% • 4 2200
Water • 1 8200
Ethylene glycol • -30 300

Fig. U..16 Flooding velocities for aqueous glycerol solutions

and for ethylene glycol in countercurrent flow with air at
atmospheric pressure in a 0.75-in. vertical pipe (Walli.s 26 )
compared with Eq. (11.87) using m and C from Figs. 11.14
and 11.15.

The valu e of ju is
200 1 1
3600 0.1 2.18 X 10-, ~ 25 ·5 fps
From Eq. (11.82),

j: - 25.5 [ (32.2) (~:) (62.5) r- 0.43

ANNULAR FLOW
"'
Assume that end effects are minimizcd a.nd that tho past history of the
flow allows the maximum va.lue of C equal to unity.
The value of j:¼ = 0.655. Thcrefore, from Eq. (11.87),
mj}½ = e - J;½ = 1 - 0.655 = 0.345
N 1 is found from Eq. (10.16) as follows:
N - [(,{ 2 )ª(32.2)(62.5')]¼ - 3 5 X JO'
1 2.5/3600 .
Then, from Figure 11.14, m = 1 and therefore J;= (0.345) 2 = 0.119.
The maximum value of J¡ is then, from Eq. (11.83),

iJ - [ ( \ t lr
3 2 2
(0.119) - 0.276 fps

The maximum possible liquid flow rate is therefore

imA - (0.276)(62.5)(2.18 X 10-')(3600) - 1350 lb/hr

PRIEDICTION OF FLOODING FROM THE SEPARA.TE CYLINDERS MODEL 11

The separate cylinders model was introduced in Chap. 3. It obviously

provides a poor representation of the actual geometry in pipe flow, and
yet it is the usual theoretical basis for the very successful Martinelli cor-
relation of horizontal annular and stratified flow. Despite its shortcom-
ings it is remarkable as the only analytical derivation of Eq. (11.87).
The model is perhaps more appropriate for the case of packed towers
where it represents a possible flow configuration. A similar analysis for
this latter case has been presented by Dell and Pratt. 25

Turbulent flow If the separate cylinders model is applied to vertical

flow, the equations have to be adjusted to allow for the effects of gravity.
The force per unit volume on the gas changes from ( -dp/dz) to
(-dp/dz - p,g) and similarly the force per unit volume on the liquid is
(-dp/dz - p¡g). z is measured upward.
Using Eq. (5.6) and applying the results of Prob. 5.12 with the
assumption of a constant mixing length in each cylinder, we have

(11.91)

(11.92)

where a = r//r 0 2, 1 - a = r¡2/r-o2, and r0 and r1 are the radii of the sep-
arate cylinders. We can either assume that l 1 and l, are scaled by the
dimensions of each cylinder or by the overall pipe diameter. In the for-
mer case we have

l¡ = l' (11.93)
(1 - a)¼
 ONE-DIMENSIONAL TWO-PHASE FLOW

and in the latter

l, = l¡ = l' (11.94)

At the inception of flooding the turbulence leve! in the whole flow field
increases immensely. In this case it might be assumed that the mixing
length which is usually characteristic of the flow core extends over the
whole pipe. This core mixing length is given by Nikuradse 30 as 0.14ro
and can be reasonably approximated by

l' = ~ (11.95)
7

Using Eq. (11.95) in Eqs. (11.91) and (11.92) we obtain

j: = ..1p*%anf2 (11.96)
j; = (1 - Lip*)½(l - a)n/2 (11.97)

where n has the value 3.5 or 2.5, depending upon whether Eq. (11.93) or
(11.94) is valid, and has the same significance as nin Eq. (3,30).
Eliminating Lip* from Eqs. (11.96) and (11.97) we have
·*2 ·*2
J_g_
an
+ J¡
(1 - a)n
= 1 (11.98)

which resembles Eq. (6.90). The envelope of Eq. (11.98) as a is varied is

j:2/(n+ll + j7'/Cn+ll = 1 (11.99)

which is consistent with Eq. (11.87) if m = 1 and n takes on the inter-

mediate value of 3.

Viscous llow in !he fü¡uid When the liquid flow is viscous, the appro-
priate equation to replace Eq. (11.97) is

j¡ = f; (1 - a)'(l - Lip*) (11.100)

Eliminating a from Eqs. (11.96) and (11.100) we obtain

j;' )1/n
( Lip* +
[N¡(l32j¡
- Lip*)
]½ = l (11.101)

Differentiating to obtain the envelope yields

j* )" 1 j*2/n
2 ·8 ( N¡ (1 - Lip*)½ = n(Lip~)Ci+l/n)
(11.102)
ANNULAR FLOW
"'
Table 11.2 Values of C predicted by the separate cylinders modei
fer flooding with viscous fü¡uid and turbulent gas flow

Ap' o 0.1 0.2 0.3 0.4 O. 5 0.6 0.7 0.8 0.9 1

C (n = 2.5) 1 0.987 O. 956 0.929 O. 91 0.899 0.893 0.900 0.920 0.951 l

e (n = 3.5) 1 0.876 0.853 0.826 0.816 0.821 0.837 0.863 0.897 0.941 1

Equations (11.101) and (11.102) do not have a simple solution; how-

ever,J; andJ; can be represented parametrically in terms of ó.p* as follows:
56
. N1
(j';.)l' -- 2
Z(l - t,p*)¾
+ (n - 2) t,p*
(11,103)

º*~i = [ n _6.p*(Hlfn) ]n/4

J, 2 + (n - 2) t,p* (11.104)

For different values of t,p* Eqs. (11.103) and (11.104) define values
of J7 and J; which trace out the flooding curve. These values can then
be used to derive the value of C which will agree with the left-hand side
of Eq. (11.91). Results shown in Table 11.2 for n = 2.5 and 3.5 show
that C is remarkably constant.

11.5 VERTICAL Ul'WARD COCURRENT ANNULAR FLOW

THE BOUNDARiES OF THE VERTICAL ANNULAR FLOW REGIME

There are two mechanisms which determine the boundaries of vertical

annular flow,

l. "Bridging" of the gas core by liquid from the film and a consequent
transition to slug flow
2. Entraimnent of droplets from the film and a consequent transition to
annular mist flow

For sorne combinations of fluid properties and pipe size these two
criteria can overlap and no ideal annular flow occurs at all.

The slug-annular transition The various methods for defining the

transition between slug flow and annular flow provide an excellent .icase
study" in flow regime boundary predictions.
The following approaches have been adopted:

l. Definition of criteria for upward or downward flow in a liquid film as

a function of gas flow rate
2. Studies of the conditions nnder which liquid "bridges" across the gas
core disappear
'" ONE-D!MENSIONAL TWO-PHASE FLOW

3. Measurement of the fractional liquid flow rate at the axis of the tube
as a function of flow rates by means of a sampling probe
4. Comparison of void fraction data with theoretical predictions for the
slug- and annular-flow regimes
5. Pressure-drop measurements

These techniques will now be discussed in arder and related to quan-

titative theoretical criteria.
Criteria for upward or downward fiow in a liquid film If a liquid
film is set up on the wall o! a vertical pipe, for example, by inj ecting
liquid through a porous sintered metal plug, it can flow upward or down-
ward, or divide and flow partially upward and partially downward,
depending on the gas and liquid flow rates involved. The downflow in
the film is limited by flooding. For a given upward gas flow rate there
is a maximum possible value of downward liquid flow rate and vice versa.
The gas flow rate at which no liquid can flow downward gives a possible
criterion for the upper limit of slug flow at Iow liquid rates.
For tbe observed range of the constant C in Eq. (11.88) tbe limiting
value of gas flow rate at which no liquid will run clown tbe wall is given
by the relationships

(11.105)

The upper value in Eq. (11.105) is approached when end effects are
minimized.
Figure 11.11 shows typical results which are obtained when water
is supplied to a vertical tube through a porous section of the wall and the
upward air flow is continuously increased. At low air rates the liquid all
flows downward. After flooding occurs sorne water begins to flow
upward and eventually a point is reached at which all of the water flow
is upward. The condition that ali the water should flow upward gives
a way of defining the lower boundary of cocurrent flow.
When there is a very low liquid flow rate the sl:¡aded region in Fig.
11.11 becomes narrow, since if one injects liquid at the end of the flooding
curve where no liquid will flow downward at ali, the only way for it to
go is up. Experimental data for both viscous and inviscid liquids 31 cor-
relate with the equation

J·*"'og
' . (11.106)

A gas rate lower than Eq. (11.106) allows sorne downward flow in
the film, whereas a flow rate greater than this is needed to caúse a finite
amount of water to flow upward continuously. If the liquid flows clown-
ANNULAR FLOW ,.,
ward in the film and cannot escape at the bottom of the tube, then a
liquid slug will eventually be formed.
Equation (11.106) is also approximately consistent with the bound-
ary of the "forbidden region" shown in Fig. 11.24 overa wide range of
liquid flow rates.
"Bridging" of the gas core A liquid bridge across the flow cross
section can be detected using an electrical probe of the type dcveloped
by Griflith 32 and used by Haberstroh and Griffith 33 and Brenner. 35 The
probe consists of a conducting tip on the end of an insulated rod which is
mounted in the core of a two-phase flow. If the liquid conductivity is
very different from the gas conductivity, then a measurement of the
electrical resistance between the tip o! this probe and the tube wall will
indicate whether the tube is bridged by liquid or not.
The method of defining transition is rather arbitrary since there is a
rather broad region in which slugs become few and far between but do
not disappear altogether. The results of Griffith, Haberstroh, and
Brenner all correlate with the equations

j; < 1.5 (11.107)

j; > 1.5 (11.108)

11
Entrainment" measurements using a sampling probe Another
method of detecting a liquid bridge across the gas core is to sample the
flow by means of a probe at the tube axis. The probe detects both liquid
bridges and entrained droplets but will not collect any liquid in "ideal"
annular flow in which ali of the liquid flows on the tube wall. The pro be
measures a liquid flux (j¡) 0 r, near the centerline of the pipe which may be
compared with the average liquid flux j 1 . In slug flow (j1 )cr, can be
much larger than j 1 because the liquid flow can be up in the slugs and
down in the film around the bubbles; the liquid flux profile therefore
peaks at the eenter and is negative at the wall.
Figure 11.17 a shows entrainment data which were taken in this
way in a tube of 1-in. diameter for air and water at atmospheric pressure.
Zero entrainment was obtained only overa limited range of gas velocities
for low liquid rates. Entrainment at low gas rates is dueto liquid bridges,
whereas at the higher flows it is dueto droplets. It is possible to extrap-
olate the slug-flow portion of this graph to the abscissa to define a lower
critical gas velocity at which the slug-annular transition occurs. The
transition line is correlated by the equation

j: - 0.4 + 0.6j; (11.109)

whieh is significantly below the electrical probe transition line.

,.. ONE-DIMENStDNAL TWO-PHASE FLOW

180

160 0.975" -ID tube

0.185"-ID probe
• • J/ = 0.0243
140 o J/ = 0.097
J/ = 0.347
ó

Aj/ =Q.49
g 120 1-----11------+-- 111 j/ = O. 592
□ j/ = o. 840
X • xj/=1.21
~
~100
\•
-~
=
.:
e
o
E 80
e
·e
+e
o
a' 60

40

20

o
o
Dimensionless gas velocity, 19"'
(a)

Fig. 11.17a Entrainment as a function of air and water flow rates,

upward annular flow. 36

Comparison of void fraction data with theory A prime purpose of

defining the transition between any flow regimes is to predict when one
theory should be replaced by another. The slug-annular transition may
therefore be defined by seeing where the predictions of slug-flow and
annular-flow theories cross. For example, the slug-flow theory for void
fraction is, from Eq. (10.36) for turbulent flow,

J,
a=----~--=-======
l.2(j + j,) + 0.345 ygD(p - r,)/r,
(11.110)
1 1
 ANNULAR FLOW
"'
2.0 rrTTTTTTTTT7--.-rrrrrrTTTTTTTT,7--.-r,r,n
3 /,\'°-ID tube 1"- ID tube
+ j/= 0.026 o¡¡*= o.o 13
•j/= 0.103 mj/=0.125
LJ. j/= 0.520 ., Jl= 0.250 o
X j/= 1.283 V j/= Ü.375

1.6 ej/= l.903

o;, = 2.615 •
Cocurrent oir/woter
upflow
o
X
o
e 1.2 1------t---ff-----v------cr<----¡

1. /"
/✓·
c. o •

• •
1
X

0.8 f'l------,ci---tf-----f~+------+--c--
c
•E
o

,...-•

o
o
* ,·~
l::"v--v-

0.5 1.0 1.5 2.0

-----
+-+

2.5 3.0
+

3.5
Dimensionless gas velocity, ¡~*
{bl

Fig. 11.17b Pressure drop as a function of air velocity at con-

staut· water rates, upward annular flow. 35

This may be rewritten as follows:

j*
o \}
~
Pu
" (11.111)

A correlation for void fraction in upwards annular flow that will be

introduced later is
..
]¡
0.775 (11.112)
1 - 2.85(1 - a) 2.85(1 - a)
 ONE-DIMENSJONAL TWO-PHASE FLOW

These two equations are compared with data 35 taken in a 1-in. pipe,
using air and water, in Fig. 11.18. Data taken in the slug-flow regime
contain considerable inherent scatter because the void fraction measure-
ments were taken by using quick-closing valves to isolate a length of tube
and the results dcp mded on how many slug-flow bubbles happened to be
caught at the instan-t of closing. The annular flow points cluster around
the correlation. The theoretical lines cross at a liquid fraction of about
1 - u = 0.2 and at values of J7 and J; in approximatc agreement with
Eq. (11.109). In fact the line 1 - a~ 0.2 predicted by Eq. (11.112)
has the equation

(11.113)

Pressure-drop measurements Several workers (e.g., Govier et al.) 35- 3 a

have suggested that regime boundaries could be correlated in terms of
maxima or minima in pressure-drop readings. This hypothesis can be
compared with other techniques by using data such as are shown in
Fig. 11.17b.
The predictions usi:i;i.g this method are not very accurate since sorne
of the maxima and minima are relatively flat. However, the order of

Annulor flow, Eq. (11. 112)

- - - - Slug flow, Eq. (11.lll)
-0.5
"
1 "

5 points

0.1 t 1

4points,;'
•

Dimensionless gas velocity, J/

Fig. 11.18 Liquid-fraction data in the vicinity of the slug-annular transition 35 for a
dimensionless liquid flux,J; = 0.53.
ANNULAR FLOW
"'
magnitude predicted is correct, showing that pressure-drop behavior is
quite closely related to transitions in flow regime.
Discussion Of ali the methods for predicting the slug-annular
regime boundaries the most meaningful from an engineering standpoint
is the óne discussed in "Comparison of Void Fraction Data with Theory"
and based on the point at which one theory has to be replaced by another.
These results are in agreement with the entrainment pro be results which
are based on an extrapolation of lines on a graph and are hence an idcaliza-
tion. They give a working estímate of when the influence of slugs on
the overall fluid mechanics is small.
The electrical probe technique tends to estímate too high a gas
velocity by about 50 percent because occasional bridges are still detected
long alter they have ceased to be the goveming mechanism of the flow.
The region between Eqs. (11.109) and (11.107) could be interpreted
as a transition region of slug-annular flow in which a rather gradual
transition in behavior occurs.

The anm.1lar-mist transltion Considerable amounts of liquid entrain-

ment usually occur in high-velocity annular flow. If this entrainment is
measured by a probe which samples an area a at the center of the tube
and measures a liquid ftow rate of W¡, a parameter can be defined which
gives a rough measure of the fraction of liquid entrained. Thus
E = w¡ A = (j¡)cL (11.114)
W1 a j¡
If the flow were completely homogeneous and one-dimensional, E
would be equal to unity 1 whereas in pure annular flow E would be zero.
Referring to Fig. 11.17a, it can be seen that there is a very small
range of flow rates over which droplet entrainment is low. In fact most
annular flows contain sorne entrained droplets and can be analyzed by
treating the homogeneous mist and the liquid film as the two phases.
The prediction of entrainment will be considered in greater detail
in the next chapter.

CORRELATiONS FOR PREDICTlNG vom FRACTION AND PRIESSURE DROP

The Dartmoulh correlatlon 35 lor void lraclion The Dartmouth correla-

tion was developed from the hypothesis that there exists a regime of
annular flow in which viscous and surface-tension forces are small and
the fluid dynamics are governed by a balance between hydrostatic forces
and inertia forces in the gas and the liquid. This is partly based on the
observation that the film surface is quite markedly wavy and form drag
is likely to be far greater than friction drag at the interface. The result-
 ONE-DIMENSlONAL TWO-PHASE FLOW
'"
2.5

'
Symbo\ 1-a

o 0.200
0.175
'
+ 0.1625
◊

2.0
n 0.1500
o 0.1375 *o
V o. 125 ---.
◊ o. 1125 t-
·o
•o 0.100 o
w>
0.0875
• 0.075
< 0.0625 o
• 0.050 e
-~
e
+
o
.so

Flooding line,
Eq. (ll.87)
(m ~I, Ca0.88)

-0.5 O 0.5 1.0

Dimensionless liquid veloclty, j/'
Fig. U..19 Lines of constant liquid fraction in air-water annular fl.ow
versus the dimensionless parameters j;
and j;.(Wallis. 35 )

ing correlation parameters are and j; j;

which were used in the previous
flooding correlations. Plotting up the data far air and water in a 1-in.-
diameter vertical pipe (Fig. 11.19) the constant void fraction lines were
observed to be straight and to be tangential to the flooding line. The
equation which correlated these lines was
·+
(11,115)
1 - 2.85(1 - a) 2.85!{ - a) = 0.775
ANNULAR FLOW

Later tests with greater emphasis on accuracy 11 showed sorne

deviations from linearity in the constant void fraction lines and also
indicated that
º*
J¡ = 1 (1L116)
1 - 3,1(1 - a) 3,1(1 - a)

This correlation is v.alid only over the range of jf and J;

shown in
the figure and when droplet entrainment is below about 20 percent,

The modified Martinelli correlation 11 The Martinelli correlation was

originally derived far horizontal flow and <loes not apply to vertical flow
because it ignores gravity. However, it can be used when velocities are
high enough (approximately 1:
» 2) for shear stresses to domínate the
flow. At lower velocities it is more accurate to follow the procedure out-
lined in Example 3,7 to express the frictional forces on the two streams,
In the absence of acceleration or phase change the momentum equations
for the gas core and for the whole flow are then

dp
dz + p,g + </>, 2 (
- dp)
dz , = O (1L117)

~~ + g[ap + (1
3 - a)p¡] + q,¡2 ( - ~~\ = O (1L118)

In terms of the dimensionless single-phase frictional pressure drops

(1L119)

(1Ll20)

Eqs, (1L117) and (1L118) can be written as

!J.p* = cp/ 13.p: (1L121)

t>p* = q,¡2 t>p7 + (1 - a) (1Ll22)
Assuming that q,¡ and q, 3 are known functions of a, Eqs, (1Ll21)
and (1L122) can be solved simultaneously to give t,p* anda as functions
of t,p: and t>p7" This has been done by Turner 11035 using q,¡ = 1/(1 - a)
and an empirical expression for <p 0 • The results are shown in Fig. 11.20
which gives a rapid method for deriving 13.p* anda from known quantities.
The lines of constant void fraction are straight and envelop the flooding
locus J as usual.
 ONE-DlMENSJONAL TWO-PHASE FLOW

0.12 lh,d = 0.13

\ \
\ \ \
\ \
\
\ \
\ \
\
\

0.15

'\ 0.16
-"e
0.02
'
.2
e
•E 0.17
o
0.18

0.20

Flooding
locus

O 0.005 0.010
Dimensionless single-phose liquid pressure drop, ó.p;

Fig. 11.20 Prediction of void fraction and pressure drop from modifi.ed l\fartinelli
correlation. 11 , 35

SIMPLE FLOW MODIE.LS

Separa le cylinders model The separate cy!inders model can be developed

for vertical flow in exactly the same way as for horizontal flow. The
frictional pressure drop for the gas cylinder is now - (dp/dz + p,g),
whereas for the liquid it is -(dp/dz + p¡g). This model is quite
unrealistic for cocurrent :flow unless -dp/dz > p¡g and predicts counter-
current flow if -dp/dz < p1 g. However, it does give a reasonable pre-
diction of sorne correlations for :flooding.
ANNULAR FLOW
"'
Homogeneous model In Ref. 39 the homogeneous model was tested as a
method for correlating the frictional pressure drop in annular-mist flow.
In order to make a fair test of the theory it was first necessary to
obtain an estima te of the range of flow rates for which the flow regime was
annular-mist rather than sorne other pattern such as slug flow. When
entra~nment data v¡rere available, the flow regime boundary was deter-
mined by requiring that- the percent cntrainment, measured by the
sampling probe technique, should be to the right of the mínimum on a
graph such as Fig. 11.17 and above 20 percent. When entrainment data
were not reported by authors the annular-mist regime was determined
by requiring that the gas (ar vapor) flow rate should be above both the
slug and annular transition line given by Haberstroh and Griffith, Eqs.
(11.107) and (11.108), and the critica! gas velocity for the onset of entrain-
ment derived by Steen, 2 Eq. (11.2). In ali cases it was found that homo-
geneous th~ory gave a better estimate of frictional pressure drop than
lVIartinelli's correlation. However 1 l\1Iartinelli 1 s predictions of void frac-
tion were superior to those of homogeneous theory. The homogeneous
friction factor was remarkably clase to 0.005 for ali of the data which
were examined.
A more accurate version of the homogeneous model can be obtained
by introducing the surface tension as a further correlating factor. This
is motivated by the major influence which surface tension exerts on film
stability, interfacial roughness, and droplet entrainment. The CISE
correlation 40 for frictional pressure drop, for example, is essentially a
modified homogeneous flow equation of the form

- dp) GL4zi0.86a-0.4
( dz F = e Dl.2 (11.123)

where V is the average homogeneous mixture specific volume. The sur-

f a.ce tension was found to be a far more significant variable than the
viscosity of either component or any form of mixed mean viscosity. Sim-
ilarly, Bergelin et al. 41 found a significant effect of surface tension on the
effective friction factor in homogeneous flow.

Separaled-llow model The separated-flow model can be developed by

using the expressions for wall and interfacial shear stresses which were
developed in Sec. 11.3. Figure 11.21 shows that Eq. (11.18) applies to
vertical as well as to horizontal flows.
A simple result is obtained by assuming that the single-phase tur-
bulent friction factors are 0.005 for both the gas and liquid. 3 N eglecting
the effects of compressibility, relative velocity 1 and entrainment, the
 ONE-DIMENS!ONAL TWO-PHASE FLOW

CISE Test Data

0.2 o Water in 2,5-cm tube
□ Water in l.5-cm tube
00
--... ;::. Alcohol in l.5-cm tube
Q - - - Eq. (11.18)
"-11 o. 1 - - Levy 7

e 0.08
.Q
~ 0.06
u"
0.04

0.03

0.002 0.004 0.007 0.01 0.02 0.04 0.07 0.1 0.2

Ratio of liquid film thickness to pipe radius, 28/D

Fig. 11.21 Comparison between Eq. (1L18) and Levy's correlation7 , 3 of CISE data
for vertical flow.

equivalents of Eqs. (11.121) and (11.122) are

t,.p*
_,.. 1
10 Jg 2
+ 75(1 - a)
(11.124)
a%

t,.p* rn-'i}'lí¡I + (1 - «l (11.125)

(1 - «)'

For laminar flow, on the other hand,

.,.
D.p* = (l ~ a)' + 0.684(1 - a) (11.126)

where i? is defined as
32}¡ (11.127)
= N,

The factor (0.684) is a correction factor for velocity profile that can be
developed from a differential analysis.
Figures 11.22 to 11.24 show the prediction of Eqs. (11.124) and
(11.125) for the turbulent-turbulcnt flow regime. There are severa]
important qualitative aspects of these graphs.
It is evident from Fig. 11.22 that the character of the flow changes
quite dramatically when }; falls below about unity. No downflow is
possible unlessj; is less than 0.95 anda slight reduction below this value
ANNULAR FLOW
"'

1.0

_o,
e

·º
e
w
E
o
.05- 0.10 __.....-0.15 0.20

0.8

J;*-::::-

0.10
Liquid fraction, l - et

Fig. 11.22 J; versus (1 ~ a) far various values of J; far a turbulent film predicted from
Eqs. (11.124) and (11.125). (Wallis.')

requires high values of 1 - a before upflow can occur (there is actually

j;
a slight region of upflow near the origin which allows values only below
O.O!). Practically, a value of j; -
0.9 corresponds to a situation in
which thin liquid fi!ms flow downward while thick ones flow upward.
A net upflow of liquid then occurs as a result of '\vaves" of thick film
riding over a smoother and thinner falling film. With values of below j;
0.9 these thick films are usually sufficient to bridge the pipe temporarily
and bring about a transition to slug flow. This conclusion is consistent
with the empírica! conclusion reported in Eq. (11.106).
In the case of laminar flow of the liquid the picture is qualitatively
the same with the transition-occurring at = 0.8.J;
The pressure-drop graph for low liquid rates (Fig. 11.24) displays
the same effects in a different way. Below a gas rate in the range
0.8 < j:
< 1 the pressure drop increases immensely and there is a "for-
bidden region" in which no solutions can be obtained. Since the pres-
sure drop is often controlled by the external characteristics of a system,
this forbidden region can be very significant; an attempt to operate in
 ONE-D!MENSIONAL TWO-PHASE FLOW

l.0 f - - - - - - - - - - - - + - - - - - - c,°',L------l
••
\\

.,
Jg

fig. 11.23 Dimensionless pressure drop D.p* versus dimensionless gas

flux ia, for various values ofjJ [from Eqs. (11.124) and (11.125)].
(Wallis.")

it results in either a new flow regime or an unstable situation in which a

large change in the operating conditions occurs. Moreover, at a gas flux
which crosses this forbidden region at low liquid flow rates, there are three
possible solutions to the pressure drop for a given liquid rate. Rather
small fluctuations in any one of the parameters can then lead to a jump
from the low-pressure-drop to the high-pressure-drop value.
The pressure-drop minimum at a constant liquid rate is also of
significance for stability in a system in which the pressure drop and liquid
rate are controlled. If the gas rate is allowed to !ali below the value
corresponding to the minimum pressure drop, it will continue to fall until
the pipe fills with liquid and becomes flooded. Figures 11.23 and 11.24
ANNULAR FLOW
"'

0.15 f - - - - i l - + - - - - - - - - - + - - - - - - - - - - 1

·ºrn
~
e
u
u
•
.:e
- J'

º'-------'----------'-----------'
1.0 1.5 2.0
lg·•

Fig. 11.24 Enlargement of Fig. 11.23 for low values of il-

show that for Lip* less than 0.3 the maximum liquid rate ata given pres-
sure drop occurs at a value of j:
equal to about l. l.

Example 11.5 Salve Example 11.2 if flow is vertical. (a) Use the modified Martinelli
correlation and Fig. 11.20. (b) Use Eq. (11.115) and Fig. 11.19. (e) Use the
separated-fl.ow model.
Solution (a) From Example 11.2 we have (-dp/dz) 1 = 7.2 dynes/cm 3 and
(-dp/dz) 0 = 9.4 dynes/cm 3 • Therefore, from Eqs. (11.119) and (11.120) 1

9.4 7.2
t:.p 0* = - = 9.6 X 10-a t,p*
1 ~ - = 7.35 X I0- 3
980 980
360 ONE-DIMENSIONAL TWO-PHASE FLOW

Frorn Fig. 11.20, 1 - a = 0.176, t:,.p* = 0.42.

(b) The values of ju and J, are 47.2 and 1.03 fps. Therefore the values of
1: and J! are

·• [ (0.076) (12) ] i,
J, ~ 47 ·2 (0.98)(32.2)(62.5) ~ l.OlS

12
jf ~ 1.03 [ (0.98)(32.2) ]" ~ 0.636

From Fig. 11.19, therefore, 1 - o: = 0.175.

(e) Using the above values of J;
and jJ' in Figs. 11.22 and 11.23 we find
1 - a = 0.205, 6.p* = 0.3.
Since thc operating point is clase to the slug-annular regime boundary and
to the minimurn pressure gradient in Fig. 11.23, corresponding to zero wall
shear, it is likely that the separated-flow model prediction is less accurate than
the predictions of methods (a) and (b).

IE.xample 11.6 Salve Example 2.5 using annular flow theory.

Solution The values of J:
and J! are

'* [ 0.076 ]h
J, ~ (47 -2) (32.2)(0.0817)(62.5) ~ l.Ol.5
J! ~ (3.1)[(32.2)(0.0817)]-H ~ 1.91

These values are outside the range shown in Fig. 11.19, and the use of
Eqs. (11.108) and (11.109) shows that thc flow pattern is in the transition
region between slug and annular flow. Assuming that Eq. (11.116) isreasonably
valid in this region we have

1.015 1.91 ~ l
1 - 3.1(1 - a) 3.1(1 - a)

,vhence 1 - a = 0.232, which is very close to the obs.erved value.

From Fig. 11.23 .ó.p* :=:::: 0.66. Therefore the pressure drop is
(0.66)(62.5)( 1 ½ 728 ) = 0.43 psi, which is an overestimate. This value is too
high because allowance has not been made for thc curvature of the velocity
profile or for the probable alternating upward and downward flow in the
liquid film.

IMPIROV!EME.NTS TO THE SEPARATIED-IFLOW MODELª

The accuracy of the separated-flow model can be improved in numerous

ways by taking many further phenomena into account. In most cases
the form of the theory remains unchanged but various correction factors,
which depend on additional parameters, are included.

The liquld film A differential analysis of the time-averaged flow in the

liquid film can be perlormed using the methods of Chap. 5. The result
lor laminar flow was derived in Example 5.1. Over the range O < 1 - "
ANNULAR FLOW 361

Table 11.3 Sorne of the dimensionless parameters used by Hewitt for

his differential analysis of the liquid film in annular flow 42

1
Hewitt w+ ~ M - Re*
•'
This book
Re 1
4
t.p'
N¡%-
4
,, D'
N¡r -
t,p'
1 - a
- 1
Ni
~ [t.p' -
4
(1 - a)]¼

Dimensionless Flow Pressure Film Shear pro- VVall shear

form of rate drop thickness file
curvature

< 0.2 1 which is usually characteristic of annular flow, the equation for
the pressure gradient can be approximated by Eq. (11.126).
The analysis of a film containing both laminar, "buffer," and tur-
bulent layers was performed by Hewitt. 42 Sorne of the parameters which
he used are shown in Table 11.3 together with a translation in terms of
the present nomenclature and an indication of physical signi:ficance.
The results can he plotted in the form o! a wall friction factor versus
Reynolds number for various values of the other parameters, as shown in
Figs. 11.25 and 11.26.
The lines of constant ¡3 correspond to a situation in which the pres-
sure drop is kept constant while the liquid flow rate is varied. At low
flow rates the film is thin, the shear profile is approximately linear, and
the results agree with the horizontal flow curve. As the liquid rate is
increased, the shear profile distorts under the influence of gravity and

o Symbol
200 75 20 /3
- - - Horizontal
flow

0.001~-~~~~~-~-~~~~~~~~~~~-~~~~~
102 103 104 105
Reynolds number, Ref

Fig. 11.2:5 Wall friction factor in vertical flow far Re* oo deduced from Hewitt's
differential analysis. 42 (Wallis. 3)
362 ONE-DIMENSIONAL TWO-PHASE FLOW

v □ ◊ o □ Symbol

~·
ó.

V 5 10 20 50 75 200 f]
Horizontal flow
I.J 0.1 C - - - - - - - - - - - 4 ~ - - - - + - - - - - - - Blasius

t
-E
□

e □
-~o □

Reynolds number, Rer

fig. 11.26 Wall friction factor in vertical flow from Hewitt 42 for Re* = 1000.
(Wallis.')

eventually thc wall shear stress, and hence the friction factor, falls dra-
matically. The liquid Reynolds number at the point of zero wall shear
can be correlated approximately with /3 by the equations 3

Rer < 300 /3 = 1.2 Re¡l' (11.128)

2000 < Rer < 20,000 /3 = 0.13 Re/' (11.129)

A convenicnt way of taking account of both viscous effects and the

nonlinear shear profile is to introduce two correction factors into Eq.
(11.125), which is rewritten as

/',.p* = 2Cíwj:'2 + B(l - a) (11.130)

(1 - a)

c;w is what the friction factor would be if the shear stress profile were
linear, as it is in horizontal flow, and is given by Fig. 11.7. The param-
eter B takes acoount of the curvature of the shear stress profile and has
been found to give general agreement with Hewitt's results far the
following dependence on Reynolds number 3

Re1 < 1000 B = 0.684

1000 < Rer < 8000 B = 0.193 Re¡°· 183 (11.131)
Re1 > 8000 B ~ 1
ANNULAR FLOW
"'
These equations are merely one form of approximation to the general
function relating flp*, 1 - a 1 J; 1 and Re1 , which has the virtue of giving
the correct limiting behavior in laminar flow or horizontal flow. Errors
tend to be largest near the point of zero wall shear.

The gas core Sorne of the effects which were neglected in the analysis
of the gas core which led to Eq. (11.124) were liquid entrainment, the
velocity of the gas-liquid interface, gas viscosity, and compressibility.
The effect of entrainment can be taken into account approximately
by considering that the core is composed of a homogeneous mixture of
gas and droplets. Let the gas mass flow rate be W,, the film flow rate
be W ¡, and the entrained mass flow be We. The total mass flow rate in
the core is
W, ~ W, + W, (11.132)
The average density of the core is, approximately (if p¡ » p,),
w,
PG = W Pu (11.133)
o

Therefore the value of j:

is increased by the factor (W,/W,)l<.
For the interface velocity it is not obvious whether one should use
the mean Iiquid velocity, the velocity of waves on the surface, or a time-
averaged velocity. Since in most cases the film is thin and the shear
stress across it is fairly constant, a reasonable assessment is that the
interface velocity is twice the average liquid velocity.
The interfacial shear stress will now be, with these assumptions,
(1 l. 134)
The velocity ratio is given in terms of the flow rates by the expression
'I}_!_ = W¡ Poª
(11.135)
v, p¡(l - a) W,
Using Eq. (11.135) in Eq. (11.134) the interfacial shear stress, and
hence t,,p* in Eq. (11.124), is found to be modified by the factor
1 - 2 W¡¡,,,_a_)'
( W, Pt 1 - a
The viscosity of the gas core affects the boundary !ayer near the inter-
face and hence the interfacial shear. A reasonable approximation is to
use the value of the friction factor for the gas flowing alone in the pipe
instead of the constant value o! 0.005 in Eq. (11.19). Up to a Reynolds
number of 10 6 the Blasius equation may be used as follows:
(C¡), ~ 0.079 (Re,)-º·" (11.136)
 ONE-DIMENSIONAL TWO-PHASE FLOW
"'
The gas Reynolds number is calculated from thc overall core flow and the
gas viscosity. Thus
Re, ~ 4(W, + W,) (11.137)
7íDµ"
Combining these correction factors, Eq. (11.121) becomes

!lp*

(11.138)

Typical results of applying these correction factors in order to pre-

dict the superficial gas friction factor from the data of Gill and Hewitt 43
are shown in Fig. 11.27. Figure 11.28 shows the application of ali of
these factors to a wider variety of data taken with two different injection
systems. These results are based on the "nonaccelerational" pressure
drop which Gill and Hewitt derived by subtracting the acceleration com-
ponent from the overall pressure drop.
Cornpressibility effects alter both Eqs. (11.124) and (11.125) because
the acceleration of the gas core is felt in both momentum balances. Since
1 - a is small and fairly constant, it is usually sufficient to multiply the
predicted pressure gradient in both equations by the factor 1/(1 - NL')
where Jl;_[c is the isothermal core Mach number given in the usual way by
M, = G,~ = (W, + W,)Q, (11.139)
e a'lp a2A2p

I t may be convenient to work in terms of a modified core dimen-

sionless flux
.• _ •• (w-'
Je - w{})½
]g
(11.140)

and a dimensionless pressure

p* = p (11.141)
gD(p¡ - P,)
in which case the square of the core :~VIach number is
j*2
Mc 2 = ~ (11.142)
" p
Example 11.7 Gill and Hewitt 43 measured film thickness and pressure gradient in
upward cocurrent annular flow. A typical data point was W 0 = 300.5
lb/hr, W 1 = 466 lb/hr, Po= 0.0823 lb/ft 3 , p¡ = 63.3 lb/ft 3 , µg = 0.0423
lb/(ft)(hr), µ¡ - 2.615 lb/(ft)(hr), W, - .\34 lb/hr, , - 0.0106 in., D - 1.25
in., -dp/dz = 23.87 lb/(ft2)(ft). Compare these results with theoretical
predictions.
ANNULAR FLOW 365

x Uncorrected, l = 1
o Corrected for entrainment, l=a1
+ Corrected for entroinment and
6
relative velocity, t=a102 X

o Corrected for entrainment, 1

relative velocity and gas 1
Reynolds number, l = 07 02 03 X

~ •
·Q¡
<J
o
N
~ti,
4 X

o
o '
11

~¡ -
Q-Q
3 X
o

+
+

o
+

o
+

o
+
+

o o
o
2 !
Wg = 300 16/hr air
Multijet injector

o~_ _...1-_ __ L_ _____L_ _ _j __ __J

0.02 0.03 0.04

Liquid fraction, 1-a

Fig, 11.27 Results of applying successive correction factors to

the data of Gill and Hewitt 43 to obtain the ratio (C1 ),a/( C1 )a as
a function of liquid fraction. The three correction factors from
Eq. (11.138) are: a1 = WjW 0, a2 = [1 - 2W1 puo,)]2, and
a 3 = 0.079 Rea-º· 25 /0.005. The line is a modification of Eq.
(11.24): (C¡)'" - (01 ),[l + 90(1 - a)]. (Wallis.")

Solulion The Reynolds number far thc liquid film is

4W1 (4)(466)
Re¡ = 1rDµ¡ = 1r(L2)i2)(2.165) = 2180

Therefore, from Fig. 11.25 and Eq. (11.131), (C¡,,.,)1 = 0.015, B = 0.79.
jJ J:
The values of and are evaluated as 0.133 and 3.94; Reu is found from
Eq. (11.137) to be 2.4 X lüó; p* is 360; 2W¡p 0 /W 0 p¡ is 0.004; l\T 0 2 from Eq.
(11.142) is 0.043. Equations (11.130) and (11.138) with all correction factors
included are then
* 5.3 X 10- 4
t.p (1 - 0.043) - (l _ o)' + 0.79(1 - a)

t.p'(l - 0.043) - 0.112 l +7 :;! - o) ( 1 -

4
~ ~O:"º)'
 ONE-D!MENSIONAL TWO-PHASE FLOW

8
1 1 1 -
r
Symbol + o o V o o o
7 ' -

Wg(lb/hd 300 500 700 200 300 400 500 600

r
Inlet design ,~ Slot Jet o
+_
6 ~

1/
r Air-water vertical upflow
+
5 ~
D=l.25in.
o ºx
-
f-
_;:Y'

3
r

~
o
o XV" o
'

1c,1,g -
-

2
i ~? +
º
--
(C,}g
~ 1 +90 (1-"'}
-

~º•~ + -

-
- -

o
o 0,01 0,02 0,03 0,04 0_05
liquid froction, ] -a¿

Fig. 11.28 Superficial gas friction factor ratio derived from Gill and Hewitt's data 43
using ali correction factors in Eq. (11.138). (Wa.llis. 3 )

These equations are solved simultaneously to determine 6p* and a. Thc

result is 1 - a = 0.038, i1p * = 0.40 compared with the measurcd values
1 - a = 0.034 and b.p"' = 0.38.

The gas~liquid interface Various conditions havo been recognized at the

interface in annular fl.ow. Very thin films appear smooth, whercas thick
oncs are covered with large "disturbance waves." In bet·vveen these is a
region of small waves or ripples. Droplct impingement and cntrainment
modifies the appearance at high gas velocities. These various flow
regimes should affect the interfacial friction factor and perhaps also the
character of the flow in both phascs.
These effccts are illustrated in Fig_ 11-29 whieh compares Eq. (lL 18)
with sorne results of Shearer and Nedderman 44 taken in upward cocurrcnt
annular fl.ow. Clearly the regime of 11 rough" annular flow which is
describcd by Eq_ (11- 18) is characterized by disturbance waves, whereas
the friction factor is lower in the 11 ripplen and usmooth interface" regions.
Each series of data points corresponds to a constant gas flow rate.
l<'igure 11.4 showcd the results of Chien and Ibele" for downflow
plotted on the basis of superficial friction factor versus ó/ D for various
ANNULAR FLOW
"'
valucs of liquid flow rate. The results correlate with each other and
approximately with the cquation for ''rough" annular flow in thc regime
11
which the authors describe as annular mist.n Apparently the onset of
disturbance waves coincidcd with the occurrence of cntrainment.
Similar results were reported by Dukler 4 .'í and Berge1in and Gazley, 41
who o"pserved a strong influence of surface tension.
It appears, therefore, that the uscfulncss of Eq. (11.18) is restricted
to the large-disturbance wave regime. If greater accuracy is required in
the ripplc and smooth film regions, new correlations will have to be
derived as well as criteria for detcrmining the flow regime of thc film.
The ,vork of Shearer and Nedderman 44 indicates an effect of gas and
liquid flmv ratcs, viscosity, and density 1 as well as surface tension and
pipe diameter. '!"'he interaction bctween all these parameters has yet
to be expressed in a form which has any general validity.

ADDlT!ONAl EFFECTS

Besides the phenomena which have just been dcscribed, there are two
further effects which make accurate predictions from known quantities

Disturbonce
waves

Wave regime
transition

Ripples

Qc__ __ J_ _ __j__ _ __ L_ _ ___J___ _ ___j__ __ _ J

o 2 3 4 5 6
S/D X 10 3

Fig. 11.29 Effect of interfacia.1 wave regimcs on friction factor. (Jlrom data
of Shearer and Nedderman. 44 ) Pipe iliameter = 1.2fí in.; air-water upflow.
(Wallis.ª)
 ONE-DIMENSJONAL TWO-PHASE FLOW
'"
- - - laminar film theory, Eq. (11.126)

0.5 o - - - Turbulent film theory, Eq. (11.125)

o ó. Bennett ond Thornton~ 2 ¡/ f;:l 1.9
Jet injector
0.4 o
o o Hewitt, King & Lovegrove~ 0
o
i/~ 2.0
Annular slot injector
0.3 o
Both sets of data for oir and water
o in l.25"-ID vertical tube at about
atmospheric pressure. Entrainment
0.2 level unspecified.
1
Rer ==a 3000
0.1
I"
"
2 4 6 8 10
.,, X 10 4
1,
o
o 0.1 0.2 0.3 0.4 0.5
.,
Íf

Fig. 11.30 Comparison between theory and data of two differcnt investigations at
the same fl.o-\Y rates but using differing injection techniques.

difficult. These are inlet effects and droplet entrainment. Quite differ-
ent results can be obtained in the same apparatus if thc method o! intro-
ducing the phases is altered. In extreme cases the liquid can be intro-
duced either as a film on the tube wall oras a dispersion of small droplets.
For example, pressure-drop results shown in Fig. 11.30 show a difference
o! a factor o! 2 in these two cases. Thc fraction of the liquid which is
entrained may not reach equilibrium sorne 200 diameters away from the
inlet. Even in the equilibrium case, reliable methods for predicting the
degree of entrainment are not available. The techniquc which will
be presented in the next chapter (Sec. 12.10) provides only a first
approximation.
A changing level of entrainment introduces additional mass-transfer
effects between the annular film and the gas core and alters both the
accelerational and frictional terms in the momentum equations.
In engineering design it may be worthwhile to allow the degree o!
entrainment to be a dependent variable, to assess its influence on per-
formance, and to seek to control it by suitable inlet design and by intro-
ducing devices such as swirl promoters into the flow.

PROBLEMS
11.1. 1000 lb/hr of air and 1000 lb/hr of water flow in a 1-in.-diam horizontal pipe
at 70ºF and 100 psia. What is the flow pattern? Estimate the pressure drop as a
ANNULAR FLOW

function of the amount of liquid which is entrained and compare the results with Eq.
(3.30) with n ~ 3.5.
11.2. Evaluate the ratio (C1)d(C1)a from Eq. (11.8) and Martinclli's correlations.
Compare the result with the value predicted by Eq. (11.23) with (C¡)g replacing the
coefficient 0.005.
11.3. Compare Eq. (11.20) with single-phase friction factor charts.
11.4. Oil '(p¡ = 6.5 lb /gal, µ¡ = 0.6 cp) and natural gas (Pa = 3.5 lb/fts, µg = 0.014 cp)
flow at 1000 psia in a horizontal pipeline of 8 in. diam. The pressure gradient is
controlled at 3 psi/1000 ft although the flow rate ratio of the components may vary
dueto upstream conditions. Estimate the relationship bctween the flow rates under
these conditions.
11.5. Far a (l/n)th-power velocity profile in turbulent :fl.ow the equivalent of Eq.
(11.48) is C1 = A (Rem)- 21 <n+i¡. What is the value of <p1 as a function of n? Compare
the result with Table 11.1 far n = .5, 9, and 19. What is the limiting expression far
<p as n tends to infinity?
11.6. Compare Eq. (11.54) with Martinelli's correlation and Eq. (11.22) far typical
valucs of the viscosity and density ratios.
11.7. Deduce Eqs. (11.55) and (11.56) and compare (11.56) with alternative equations
relating <bo and a.
11.8. Discuss the various levels of sophistication in the analysis of the liquid film in
horizontal annular flow. Under what conditions are the more elaborate theories
worthwhile? What is the difference incurred in predicting film thickness by thc
simplest and by the most complicated theory if Re 1 = 106, -(dp/dz)F = 10-1 psi/ft,
-(dp/d,) 1 ~ 10-, psi/lt?
U..9. What is the film thickness on a vertical wall down which water at 20ºC flows
ata rate of (a) 1 cfm/ft and (b) 10- 3 cfm/ft?
11.10. What are the velocities of continuity waves for the conditions of Prob. 11.9?
11.11. Compare Eq. (11.68) with Eq. (5.32) for t.p* = O. For what value of 5/D is
the film thickness prediction in error by 10% if the tubc curvature is ignored, for a
given value of liquid flow rate?
11.12. Compare the flooding curves predicted by Eqs. (11.116), (11.124), and (11.125)
with the modified Martinelli correlation shown in Fig. 11.20.
11.13. 2 cm 3 /sec of 99 % glycerol at I0ºC flow downward inside a vertical pipe of
¾ in. diam. What is the maximum allowable countcrcurrent flow of air at 14.7 psi?
11.14. Compare the limit of the annular flow flooding correlations at very low gas-flow
rates with the condition that a slug-flow bubble should be brought to rest in a vertical
pipe.
11.15. Shearer and Davidson 46 analyzed the formation of flooding waves on the outside
of a tubein terms of the Weber number, pgju2ó/u, Rer, and Yfrom Eq. (10.20). Show
that this result and the equations presented in Sec. 11.4 are special cases of a function
of j:, jf, N 1 , and NEO•
11.16. A vertical tube, diameter D, length L, is closed at the bottom and opens at the
top into a pool of liquid of low viscosity at saturation temperaturc. Show that the
maximum uniform heat flux which can be supplied to thc tube without overheating is
,.;. _ e h1uD%[pgg(p1 - pg)]¼
'f'max - L[l + (po/p¡)H]2

where C = 78 for a sharp-edged flange at the top and C = % for a well-rounded

connection. 47
 ONE-DI MENSIONAL TWO-PHASE FLOW

11.17. Suppose that a very large 'Nave forms on a falling film in a pipe as a result of a
two-phase hydraulic jump. Show that the solution to the countcrcurrcnt potential
inviscid fiow below the wave can be formulated in terms of J; and J;.
11.18. Determine the relationship between mass flux G, quality x, and pressure p at
the boundary between slug and annular flow for water in a 1.5-in.-diam vertical pipe.
J:
11.19. Compare the relationships between jJ and for constant values of void fraction
which are predicted for air and water in a 1.25-in.-diam vertical pipe at 20 psia and
70ºF by
(a) Eq. (11.116);
(b) Eqs. (11.121) and (11.122) and Fig. 11.20;
(e) Eqs. (11.124) and (11.125) or (11.126), depending on whethcr thc flow is
laminar or turbulent;
(el) Eqs. (11.124) and (11.12.i:i) corrected for relative velocity, Reynolds
numbers of both gas and liquid, and compressibility;
(e) method (el) if there is 20% cntrainment of the liquid.
11.20. Show that the minimurn value of pressure drop predictcd by Eq. (11.126) at a
given liquid rate is almost coincident with the point of zero wall shear. Compare
with Eq. (11.128).
11.21. Compare the point of mínimum pressure drop predicted by Eq. (11.125) with
Eq. (11.129) and show tha.t the values will be cqual if the wall friction factor is 0.01
rather than 0.005.
11.22. Solov'ev 48 et al. found that a plot of !::,,p*/(!::,,p*)min versU:s J;J(j;)min gave a
unique curve for all liquid rates in upward annular flow with a laminar film. Use
Eqs. (11.123) and (11.125) to check this far values of j~* of 10-ó, 10-4, 1Q-B, and 10- 2.
11.23. Show that the transition from downflow to upflow in a viscous liquid film occurs
a.t about j: = 0.8.
11.24. Hartley and Roberts 49 correlated the interfacial friction factor in annular flow
by the equation

(C¡), - (C1), - 1.5 [i - ie, ✓(¿iJ

Compare this ·with thc correlations in the text.
11.25. Sorne typical da.ta points obtained by Gill and Hewitt 43 for upward annular-
mist flow in a 1.25-in.-diam pipe are given in Table ll.4.

Table 11.4

11p, w.,We 1 µ¡, W1, ,,

lb /ft' lb/hr lb/hr
Po,
lb/ft'
Pi,
lb/ft' lb/(ft)(hr) lb/hr in. ""'
lb/(ft)(hr)
--- --- ---
5.56 306. 1 o 0.076 63.3 2 56 20 0.0045 0.043
56.03 269.9 1110 0.086 63.3 2 53 1890 0.0175 0.043
43.7 511. 7 379 0.088 63.3 2 68 371 0.0089 0.043
23.43 669.7 1 .6 0.088 63.3 2. 50 18.4 O. 002.5 0.043
15.12 300.9 153 O.OSO 63.3 2.62 197 0.0080 0.042

Compare these results ·with the predicted values making ,vhichever corrections and
assumptions are neccssary. W. is the entrained liquid flow rate and W1 thc film flow
rate.
ANNULAR FLOW 371

11.26. Some air-water data taken by Hewitt, King, and Lovegrove 5 º in annular
upward flow in a vertical 1.25-in.-diam pipe are shown in Table 11.5. Compare
these results with
(a) Eqs. (11.115) and (11.116);
(b) Fig. 11.20;
(e) Eqs. (11.124), (11.125), and (11.126);
(d) the separated flow modcl including all effects;
(e) the "unmodified" Martinelli correlation presented in Chap. 3;
(f) homogeneous flow theory.
The mean values of the properties which are not given in the table were
µ, ~ 0.0433 lb/(ft)(hr), p¡ ~ 62.33 lb/ft".
11.27. In a vertical air-lift pump the maximum possible liquid flow rate is required
for a given pressuro gradient. If the fluid is water at 70ºF in a 3-in.-diarn pipe and
the availablo pressure gradient is 20 psf /ft, what air-flow rato should be used at a
pressure of 20 psia? VVhat is the pump "efficiencyn in terms of potential energy
acqüired by the water per unit of energy required to pump the air?

Table 11.5

µ¡,
mean water Pressure
viscosity, gradient,
W.,,lb/hr W1, lb/hr p.,, lb/ft3 lb/(lt)(h,) 1 - a 1 % lb /ft'

199.7 77.89 0.0774 2.36 2.76 4.189

199. 7 95 9 0.0775 2. 54 2.91 4 680
200.1 150 5 0.0778 2..58 3.39 5 926
200.6 189.0 0.0794 2.70 3.70 6.281
206.4 295.7 O. 0818 2.68 4.21 9.029
206.4 393.4 0.0827 2.60 4.89 10.89
208.3 489.3 0.0844 2 64 5 38 12.36
210.1 718.6 0.0862 2.69 6.66 16. 84
211. 7 841.6 0.0870 2.69 8.06 18.20
211.6 971.0 0.0878 2.72 7.37 19. 78
205.5 639.7 0.0849 2. .56 5.95 15.09
207.7 1156 o 0.0891 2.59 8.48 21.80
248.3 39.6 0.0790 2.31 l. 73 3.396
254.2 96.1 0.0803 2.43 2.13 6.568
254.4 149. 8 0.0803 2.50 2.46 8.255
257.4 196. 5 0.0811 2.58 2.78 9.580
255 8 302.2 0.0824 2.55 3.18 12.59
259.8 404.3 0.0846 2.52 3.58 15. 74
260.6 499.8 0.0860 2 57 4.09 19.09
262.6 628.3 0.0877 2 ..58 4.47 20.90
265. 7 755.6 0.0901 2.62 4.94 23.62
265.6 855.4 0.0904 2 65 ,5."7 24.69
263.7 1024. O 0.0920 2.66 6.14 28.07
268.6 1287.0 0.0961 2.68 6.86 30.89
372 ONE-DIMENSIONAL TWO-PHASE FLOW

11.28. A complete integral analysis of horizontal annular flow with a smooth gas-
liquid interface may be performed 61 by combining Eqs. (11.46) and (11.50) with
expressions for e/>¡¡ which are derived by assuming that the gas fiows in a smooth pipe
with diameter D Va. What are the expressions for ef., 0 in thc cases of laminar fl.ow
and turbulent fiow obeying the Blasius equation? Do not neglect the velocity of the
interface. Derive expressions for r/>¡¡ and a in terms of X for thc four combinations
of laminar and turbulent flow of each component. Compare with Martinelli's
correlations.
This theory is not very accurate becausc it neglects the roughness of the
interface.
11.29. Compare the curve labeled "flosv reversa!" in Fig. 11.11 with the boundary of
the "forbidden region" in Fig. 11.24. What diffcrence does it make if the effects of
gas and liquid Reynolds numbers, relativo velocity, and shear stress profile are used
to give a more accuratc estimate?
11.30. In annular flow j 0 is usually much larger than j 1 • In this case the continuity
wave velocity is given approximately by

Show that a dimensionless continuity wave velocity can be deduced as

Using Fig. 11.22 compare v:

with the dimensionless liquid velocity
v; =jj/(1- a). Under what circumstances will the ,vaves move faster than the
liquid velocity?

REFERENCES

l. Bakor, O.: 0-il Gas J., vol. 53, no. 12, pp. 185-190, 192, 195, July 26, 1954.
2. Steen, D. A.: M.S. thesis, Dartmouth College, Hanover, N.H., 1964; also D. A.
Steen and G. B. Wallis, AEC Rept. NYO-3114--2, 1964.
3. Wallis, G. B.: Papers no. 69-FE-4.5, 69-FE-46, ASME Applied Mech.-Fluids
Engg. Conf., Northwestern University, June, 1969.
4. Nikuradse, J.: Forschungsheft, p. 301, 1933.
5. Moody, L. F.: Trans. ASME, vol. 66, p. 671, 1944.
6. Chien, S. F., and W. Ibole: Trans. ASME J. Heat Transfer, ser. C, vol. 86, no. 1,
p. 89, 1964.
7. Levy, S.: Intern. J. Heat .ivlass Transfer, vol. 9, pp. 171-188, 1966.
8. Wicks, M., and A. E. Dukler: A. J. Ch. E. J., vol. 6, pp. 463-468, 1960.
9. Lockhart, R. W., and R. C. Martinelli: Chem. Eng. Progr., vol. 45, p. 39, 1949.
10. Blasius, H.: Forschungsheft, p. 131, 1913.
11. Turner, J. 1-1.: Ph.D. thesis, Dartmouth College, Hanover, N.H., 1966.
12. Armand, A.: UKAEA, AERE transl. 828, 1959, Izv. Vses. Teplotekhn. Inst.,
p. 1, 1946.
13. Hewitt, G. F.: unpublished work, 1967.
14. Nusselt, VV.: Z. Ver. Deutsch. Ing., vol. 60, p. 541, 1916.
15. Belkin, II. H., A. A. MacLeod, C. C. Monrad, and R. R. Rothfos: A. I. Ch. E. J.,
vol. 5, pp. 245-248, 1959.
ANNULAR FLOW 373

16. Dukler, A. E.: Chern. Eng. Progr. Symp. Ser., vol. 56, no. 30, p. 1-10, 1960.
17. von Kármán, Th.: NACA Rept. TJ\tl 611, 1931.
18. Bcnjamin, T. B.: J. Fluid Mech., vol. 2, p. 5;'54, 1957.
19. Kapitsa, P. L.: Zh. Eksperm. i 'l'eor. Fiz., vol. 18, pp. 2-18 and 19-28, 1948.
20. Tailby, S. R., and S. Portalski: 'l'rans. Inst. Chern. Engrs., vol. 38, pp. 324-330,
1960.
21. Hcwitt, G. F., and G. B. Vilallis: Multi-phase Flow Sym,p., ASlVlE, pp. 62-74,
November, 1963.
22. Hewitt, G. F., P. l\i. C. Lacey, and B. Nicholls: Symp. Two-phase Flow, Exeter,
England, vol. 2, pp. B401-B419, June, 196/5.
23. Lobo, Vi. E., L. Fricnd, F. llashmall, and F. Zenz: Trans. A. J. Ch. B,'., vol. 41,
pp. 693-710, 1945.
24. She,wood, T. K., G. H. Shipley, and F. A. L. Holloway: Ind. Eng. Chem., vol.
30, p. 765, 1938.
25. Dell, F. R., and H. R. C. Pratt: Trans. Inst. Chem. Engrs., vol. 29, pp. 89-109,
1951.
26. ·wallis, G. B : R.ept. General Electric Cornpany, 62GL132, Schenectady, N.Y.,
1962.
27. Wallis, G. B: UKAEA R.ept. AEEW-Rl23, 1961.
28. Nicklin, D. J., and J. F. Davidson: paper no. 4, Symp. Two-phase Flow, Institution
of Mechanical Engineers, London, February, 1962.
29. ·Wallis, _G. B., D. A. Stccn, and S. N. Brenner: AEC Rept. NYO-10,487, EURAEC
890, .July, 1963.
30. Nikuradsc, J.: Forschungsheft, p. 356, 19:-32.
31. Wallis, G. D.: UKAEA ll.ept. AEEW-ll.142, 1962.
32. Griffith, P.: Argonne Natl. Lab. Rept. ANL-6796, 196:-3.
:-3:-3. Gri[hth, P., and R. D. Haberstroh: Rept. 5003-28, Mcchanical Enginccring
Departmcnt, Massachusetts Institute of Technology, 1964.
34. Clift, R., C. L. Pritchard, and R ..M. Nedderman: Chem. Eng. Sci., vol. 21, pp.
87-95, 1966.
35. Wallis, G. B.: AEC Rept. NYO-3114-14, EURAEC 1605, 1966.
36. Govier, G. W., B. A. Radford, and J. S. G. Dunn: Can. J. Chem. Eng., vol. 35,
p. 58, 1957.
37. Govier, G. W., and W. L. Short: Can. J. Chem. Eng., vol. 36, p, 195, 1958.
38. Bro,vn, R. A. S., G. A. Sullivan, and G. W. Govier: Can. J. Chem. Eng., vol. 38,
p. 62, 1960.
39. ·wallis, G. B., and P. E. 1tfeyer: AEC Rept. NYO-3114-10, EURAEC 1480,
Scptcmbcr, 1.965.
40. Casagrande, I., L. C:r:.avarolo, A. Hassid, and E. Pedrocchi: CISE R-73, J\1ilan,
1963.
41. Bergelin, O. P., P. K. Kegel, F. G. Carpenter, and C. Gazley; Heat Transfer and
Fluid Mechanics Institute, Berkeley, Calif., 1949.
42. He,vitt, C. F.: UKAEA Rept. AERE-R3680, 1961.
4:3. Gill, L. E., and G. F. Hewitt: UKAEA Rept. AERE-R3935, 1962.
44. Shearer, C. J., and R. lVL Ncdderman: Chem. Eng. Sci., vol. 20, pp. 671-683, 1965.
45. Dukler 1 A. E.: Ph.D. thesis, University of Delaware, 1951.
46. Shearer, C. J., and J. F. Davidson: J. Fluid JJ1.ech., vol. 22i part 2, pp. 321-335,
1965.
47. Gambill, W. R.: personal communication, Oak Ridge Nationa.l Laboratory,
Tennessee, 1964.
48. Solov'ev, A. V., E. I. Prcobrazhcnskii, and P. A. Scmcnov: Ind. Chem. Eng.,
vol. 7, no. 1, pp. 59-64, 1967.
 ONE-DIMENSIONAL TWO-PHASE FLOW
"'
49. Hartley, D. E., and D. C. Roberts: Quecn Mary College, London, Nuclear
Research Memo Q6, :May, 1961.
50. Hewitt, G. F., I. King, and P. C. Lovegrove: Holdup and Pressure Drop Measure-
ments in the Two-phase Annular Flow of Air-water Mixtures, UKAEA Rept.
AERE-R3764, 1961.
51. Levy, S.: 2nd Midwest Conf. Fluid Mech., p. 337, 1952.
52. Bennett, J. A. R., and J. D. Thornton: UKAEA Rept. AERE-R3195, 1965.
 12
Drop flow

12.1 INTROllUCTION
The behavior of droplets suspended in a fluid is similar in many ways to
the behavior of bubbles. In fact, many of the equations which will be
deduced in this chapter are exactly analogous to the equivalent equa-
tions in Chap. 9.
Qualitative differences between the behavior of drops and bubbles
are most noticeable when the density difference between the components
is high, as in gas-liquid systems at low pressure. In bubbly flow most
of the inertia is in the continuous phase and as a result the drag forces
on bubbles are large compared with their momentum. Bubbles therefore
follow the motion of the surrounding fluid very closely in forced convec-
tion. Drops, however, take far longer to adjust to the motions of the
surrounding gas. For this reason the homogeneous flow model is usually
better for bubbly than for drop flow. Drop-annular flow also has no
analog in bubbly flow.
375
376 ONE-DIMENSIONAL TWO-PHASE FLOW

12.1 SINGLE-OROP FORMATION

Q.uasi-static drop formation at an orifice is an inversion of the problem

of bubble formation. The drop radius is therefore given by the analogy
to Eq. (9.3), namely,

Ra - [ ffRo
- g(p, - p1)
]¾ (12.1)

where thc subscript 2 refers to the discontinuous component. Similarly,

when drops forro on a vertical surfacc facing downward, as a result of
condensation on a ceiling, for example, the radius is scaled by the wave-
length for Taylor instability. Therefore, from Eq. (9.9)

Ra ~ [g(p, ':__ P1) r (12.2)

When the liquid velocity through the orífice is increased, the critica!
velocity given by Eq. (9.8) is soon exceeded, beeause of the eomparatively
high density of the fluid, and drops are now formed by the breakup of the
resulting jet. This problem is the classical one studied by Rayleigh 1 who
showed that such jets are always unstable with the most unstable wave-
length being about 4.5 times the jet diameter if the density of the sur-
rounding fluid ean be neglected. When a jet breaks up in this way, the
radius of the resulting drops is approximately
(12.3)
Further increase of the jet velocity eventua.lly leads to a more severe
instability due to the motion relative to the surrounding fluid. As a
result the jet becomes violently unstable elose to the orifiee and breaks
up into a shower of very small droplets. This regime of operation is
called atomization.

11.3 I\TOIVIIZATION
lVIost atomization processes produce a large number of small liquid drops
by shattering a eontinuous jet or sheet of liquid. The liquid is usually
broken up by aerodynamic forces due to relative motion between the
phases. J\1echanical, centrifuga!, electrical, and ultrasonic force fields
can also be used.
The most important dimensionless group for determining the sta-
bility of a single droplet is the Weber number based on the relative
velocity and the gas density 2- 8
We - p,(v, - v¡)'d (12.4)
u
DROP FLOW m

For nonviscous fl.uids the critica! value of the Weber number above
which droplets will break up is about 12.
Liquid viscosity apparently has a stabilizing cffect which is scaled
by the stability number, µ.¡2/ p, du. The results of Hinze 5- 6 and Isshiki 7
can be represented quite well by the equation

We = 12 [ 1 + ( P;~u)º·"]
2 (12.5)

far values of the stability number less than 5. The presence of the liquid
viscosity in the stability criterion implies that instability starts as
dynamic oscillation in the droplet shape. In the eventual process of
breakup, however, the drop is punched into a baglike shape by the
dynamic pressure of the gas acting at the stagnation point. The bag
finally bursts to farm a ring of smaller droplets. '-'
If a drop is introduced into a gas stream at high values of the Weber
number, several generations of droplets ,vill be produced by- successive
shatterings. An expression far the final drop size under these condi-
tions is 7

where the subscript O refers to the initial conditions.

The Nukiyama-Tanasawa equation,ro which has been widely used
for predicting the mean drop size for atomization with ambient air 1 is

(12. 7)

v0 is the initial relative velocity. A disadvantage of this equation is that

it is not dimensionless and care must be taken over the units which are
p¡, g/cmª; v0 , m/sec;_·d, microns; a- 1 dynes/cm; µ1i (dyne)(sec)/cm 2 •

Example 12.1 Compare the predicted drop sizes from Eqs. (12.5) and (12.7) for
atomization of benzcne (!T = 28.8 dynes/cm, Pi = 54.7 lb/ft3, µ¡ = 0.647 cp)
in a carburetor using air velocities of 250 fps and an air-fuel mass flow ratio of
18.
Solution A first estimatc of d is obtained by neglecting the stability number effect
in Eq. (12.5). The result is

(12) (28.8) _ -,
d ~ (0.0012)1(250)(30.5)]' - 5 X 10 cm

The stability number is found to have a valuc of 0.012, and the correction for
viscosity is therefore negligible.
 ONE-DIMENSIONAL TWO-PHASE FLOW
"'
The values of the terms on the right--hand side of (12.7) are

585 ✓ 28.8 44
76.3 0.876 - µ
(0.647 X 10-')'J"·"" [ (10') (0.076) ]'-' _
597 [ (70)(0.876) (18)(54.7) - O.SSµ
Therefore the predicted drop size is 4.5 X 10-~ cm.

In practice drop sizes cannot be estimated very accurately because

the atomization process is influenced by many uncontrolled variables
such as the degree of turbulence, roughnesses ar deposits of dirt in the
nozzle.

12.4 llROP SIZE SPECTRA

Equations (12.5) and (12.7) predict only an average drop size. The
commonest among the many possible "averages" is the Sauter rnean
diameter which is defined as follows 11 :

Jt" d p(d) dd
3

(12.8)
lodm« d2p(d) dd
p(d) is the probability of a drop having diameter between d and d ód. +
A drop with diameter equal to the Sauter mean diameter has the same
surface area to volume ratio as the entire spray.
N umerous statistical drop size distributions have been devised by
experimenters in arder to correlate data. lVIany authors have used the
normal and lag-normal distributions which are also in common use for
describing soils and crushed particles. 12- 14
The Nukiyama-Tanasawa 15 correlation has the form
(12.9)

A and b are normalization factors, whereas m and n are usually integers

which can sometimes be given physical significance. Quite often, as
reported by Ingebo, 16 the same data correlate equally satisfactorily using
any of these methods.
Equation (12.9) can be derived by using the "Jaynes formalism"
of statistical mechanics, as described by Tribus. 17 It is assumed that
the j/states" of the droplets are quantized and can be characterized by
indices 1, 2, 3, . . . , etc. Denoting a general one of these states by
the index i, and the probability of that state by p;, we have, since the
droplet is in sorne state,
(12.10)
DROP FLOW
"'
Furthermore, we assume that the mean value of the droplet diameter
raised to sorne power n is known from physical considerations. Typi-
cally, if the mean volume is known, n is 3, whereas if the mean area is
specified, the value is 2. Different methods of atomization may give
different values of n. Denoting this average value by (dn) we have

(12.11)
Applying the "maximum entropy" formalism we find that the probability
of a droplet being in the ith state, in the absence of further information, is

(12.12)
In order to solve this equation for the diameter probability distribu-
tion the quantum states i must be related to the droplet geometry. One
approach is to say that i represents the number of molecules in the drop,
an obvious quantization, in which case we get
i o: d3 (12.13)
On the other hand, it is possible that the major effect of drop formation
is the creation of surface area. If the quantum states represent different
amounts of surface energy, we have
i o: d2 (12.14)
In general, to allow for empiricism we let
(12.15)
Since there are so many states that the distribution may be treated as
continuous,
p,di - p(d) dd (12.16)
and substitution in Eq. (12.12), using Eq. (12.15) gives
p(d) - (m + l)dmAe-bd" (12.17)
which is the generalized Nukiyama-Tanasawa distribution function. We
may modify it to the nondimensional form used by Shapiro, 18 by employ-
ing the diameter d', at which p(d) is a maximum. Differentiating Eq.
(12.17) we find

(12.18)

whence

(12.19)
 ONE-DIMENSIONAL TWO-PHASE FLOW
"º
Defining the dimensionless diameter d* as

d* = el,_ (12.20)
d'
we find, using Eq. (12.19),

(12.21)

We have still to use the constraint which is implied by Eq. (12.10).

If there are many quantum states, the sum can be replaced by an integral
to give

1o• p, di= lo" p(d) dd = 1 (12.22)

This result, together with Eq. (12.18), leads to evaluation of the constltnt
A. Thus
(12.23)

For given values of m and n the values of b and A can now be deduced
and the whole distribution generated.
Since p(d) has the dimensions of a-1, in view of Eq. (12.16), a dimen-
sionless probability distribution can be found by defining

p* = d'p(d) (12.24)

The commonest exponents in the Nukiyama-Tanasawa equation

are m = 2 and n = 1, corresponding to quantization on a basis of drop
volume and a specified value of droplet mean diameter. In this case
Eq. (12.17) becomes, in dimensionless form,
(12.25)

In order to use this equation the value of d' must be known. A

simple method is dueto IVIacVean, 19 who found that a great deal of data
could be correlated by assuming that

d' = ~º (12.26)
2

where the characteristic diameter do was given by Eq. (12.5) or (12.7).

An alternative to Eq. (12.25) which is not very different numerically
and is more convenient for solving mass transfer and evaporation prob-
lems18 is
(12.27)
DROP FLOW
"'
12.5 THE TERMINAL VELOCITY Of SINGLE OROPS IN A GRAVITATIONAL FIELO

Small drops with a Reynolds number below unity obey tbe Hadamard-
Rybczynski equation (9.14) with the role o! the components revcrsed.
If subscripts 1 and 2 are used to indicate the continuous and discontinuous
components, Eq. (9.14) becomes
d 2glp2 - P1I elµ, + 3µ¡
V = -1 - - - - - e-~~- (12.28)
~ 18 µ, 3µ, + 2µ1

II surlace-active agents are present the surlace o! the drop is not

completely nonrigid and Eq. (12.28) should be modified in the way
discussed by Levich. 20
As long as surface tension is su:fficiently dominant in determining
the shape, the drops will be approximately spherical and Eq. (8.5) and
the methods o! Chap. 8 can be used to determine the terminal velocity.
For Reynolds numbers above 1000 (e.g., mereury drops in air) the drag
coeffieient is between 0.4 and 0.5 and the terminal velocity is about

(12.29)

An upper limit to the terminal velocity given by Eq. (12.29) is set

by the stability eriterion, Eq. (12.5). Comparing these equations far
inviscid fluids, it is faund that the maximum radius o! a stable drop is
approximately

(12.30)

This criterion does not apply to bubbles which can apparently be indefi-
nitely large, as in the case of bubbles released by underwater explosiona.
Even befare the condition o! Eq. (12.5) is reached, noticeable dis-
tortion of the drop shape can occur when the surface tension is insuffi-
ciently high. For distorted flat drops the terminal velocity is almost
independent o! size and is given by an equation similar to Peebles and
Garber's far bubbles in region 4 o! Table 9.1, namely,

(12.31)

Levich 21 recommends a value of CD = 1, whence

(12.32)

which is intermediate between Harmathy's and Peebles and Garber's

equations lor distorted bubbles. The range o! validity of Eq. (12.32) is
about the same as region 4 in Table 9.1.
 ONE-D!MENSIONAL TWO-PHASE FLOW

12.6 ONE-!llMENSIONAl VERTICAL now WIHIOUT WAll FRICTION

The techniques of Chap. 4 can be applied to suspensions of drops in fluids

if an expression for the drift flux can be found. This problem has not
been studied as intensively as tbe equivalent bubbly flow regime; how-
ever, there is considerable evidence in support of a direct analogy between
the two cases. For examplei the extensive work of Pratt and coworkers 22
on liquid-liquid extraetion in spray columns and packed towers showed
that the relative velocity of drops in a liquid was given as a function of
void fraction by the relation
v21 = (1 - a)v~ (12.33)

Substituting this value into Eq. (4.1) we have

(12.34)

This relationship was found to be very succcssful in interpreting

experimental data as long as the characteristic opening in the tower
packing was greater than the value

d,c = 2.42 [( P2 - u PI ) g]" (12.35)

Comparison between Eqs. (12.35) and (12.30) suggests that Pratt's

equation is valid when the drop diameter is determined by stability con-
siderations rather than the method of formation or the apparatus dimen-
sions. In this case, unless the viscous forces are large, Eq. (12.32) is
valid and can be substituted into Eq. (12.34) to give

(12.36)

The equation suggested by Wallis 23 • 24 is the same as Eq. (12.36)

except for a different value of the constant (1. 18) which was taken from
the work of Peebles and Garber.
Gaylor, Roberts, and Pratt 25 also indicate that an equation of the
form of Eq. (12.34) should be valid in the viscous region in which the
terminal velocity is given by Eq. (12.28).

12.7 FLOOl>ING IN lll!Ol' FLOW

The analysis of Chap. 4 is valid whichever component is dispersed; there-

fore one would expect similar results in the drop or bubbly regimes. This
is indicated in Fig. 12.1 which shows the flooding results of Blanding and
Elgin 26 for dispersions of naphtha drops in water and vice versa, plotted
as in Fig. 4.3.
DROP FLOW

i- el
.~ 0.05
.,¡-,
0.05 ,.._

.:
o
;g

el
0.05 ,...._
}
o
{'
~
o
o
z

o
o 0.5

Fig. 12.1 Flooding results of Blanding and Elgin 26

for drops of water in naphtha and vice versa.

The equations for predicting flooding given by Thornton and Pratt 27

are identical with Eqs. (4.16) and (4.17) with n = 2.
Other floocling correlations which are given in the literature are
sometimes based on the characteristic velocity v 0 and have the form 28 , 29
(12.37)
This equation is the consequence of assuming that the relative velocity
V21is independent of concentration, i.e.,
. ..
J2 Ji
V21 = - - - - = Vo (12.38)
a 1-a
and may be obtained by forming the envelope of lines of constant a in
the second quadrant (Prob. 1.19). This result is intermediate between
the results for ideal dispersed flow and churn-turbulent flow discussed
in Chap. 9 and may be of use as a general correlation which approximately
represents both possibilities. Vo may be set equal to "~ and calculated
from the equations in Sec. 12.5.
The comparison between Eq. (12.37) and the flooding equations in
annular flow [Eqs. (11.84) and (11.87)] is interesting. In many packed
 ONE-DIMENSIONAL TWO-PHASE FLOW

columns, far example, the flooding point is determined by Eq. (11.86)

rather than Eq. (12.37) dueto coalescence of the drops to form continuous
fluid streams. In each case the square roots of the fl.uxes are linearly
related.

12.8 DROP HUIDIZATION

If a suspension of drops is imagined to be fluidized in exactly the same
way as a fluidized bed and one uses Eqs. (12.36) and (4.2) to describe
the behavior, the result is
(12.39)
as long as the drops do not coalesce. Moreover, if the drops are prevented

9
-□
cf ºJ~ B
I
B
o o o Increasing }

•••
1
Decreasing } 1
¡º
I
I
.r / Mercury
dispersion
1
7 •
1
1
/,
6 •
+
t '
u

-z•
E
5
!JIT[
1

·,·i•
u
A ¡•
·~ 4 Bubble-drop

·~
transition

3 cp •
i/l: /
¡.
•
.1.
2 /'
ofb //m. ,1
1/
/
I e/
,,....,
º/
,¡ ¿ /

I / /
I / /
/ /
I / /
/

o "'-
o 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Mean volume fraction of water -

Fig. 12.2 Fluidization of mercury droplets with water

(Kutateladze and Moskvicheva 30 ). I: Bubbles of water in
mercury. II 1 III, IV: Various regimes of mercury droplcts in
water.
DROP FLOW
'"
from coalescing by the presence of surface impurities, Eq. (12.39) will
describe the behavior over the whole range of o:. Because of the difficul-
ties of inhibiting coalescence entirely, few data are available to verify
this point; however, the results of Kutateladze, "º who blew water through
mercury, are not far from the theoretical prediction (Fig. 12.2). After
fluidization for sorne time the droplet size apparently decreased and the
points followed a lower curve.
A change in character might be expected at the point where the
drops become fluidized and no longer rest on one another. For spherical
particles this usually occurs ata = 0.6 or, more accurately, from Fig. 4.2,
at 1 - o: - 0.38. Substituting this value into Eq. (12.39) we obtain an
expression for the minimum fluidization velocity,

(12.40)
Wallis 24 tested an equation of this form by blowing air through
various liquids from a porous surface and found that the onset of droplet
dispersion close to the surface occurred at a value close to the prediction.
Above the volumetric gas flux given by Eq. (12.40) a spray of droplets
was observed above a shallow liquid pool through which gas was blown.

12.9 PRESSURE llROP IN fORCED CONVECTION

The techniques of Sec. 9.6 are approximately applicable to drop fiow in

forced convection. However, the analogy should not be stretched too
far in the case of gas-liquid flow because of the much greater relaxation
times and mean free paths of droplets and their tendency to stick to walls.
For example, the virtual viscosity of an emulsion in laminar flow
can be determined 1 whichever phase is dispersed, in the form 31

µ - 1'1 (1 +2.50: +jí,,

µ,
µ2 µ1
1
) (12.41)

where the subscript 1 _refers as usual to the continuous phase. In tur-

bulent flow the homogeneous fiow model works reasonably well if the
two-phase friction factor is calculated from a Reynolds number based on
the overall mass flow rate and the gas viscosity. Thus

Re= GD (12.42)
µ,

The problem with gas-liquid fiow is that the liquid usually has an
affinity for the wall so that drops striking the wall stick together to form
an annulus. Thus, ideal drop fiow usually is never obtained but only the
hybrid drop-annular fiow. The liquid in the annulus has four effects:
'" ONE-DlMENSIONAL TWO-PHASE FLOW

l. It reduces the density of the homogeneous core.

2. It reduces the effective area o! flow for the core.
3. It forms waves which increase the effective wall roughness.
4. By means of a continuous exchange of liquid with the core, as a result
of drop deposition and entrainment, considerable momentum trans-
fer, which appears as shear stress, can take place.

The above effects are not yet fully understood. However, the
methods described in Sec. 11.5 may be used to calculate macroscopic
variables such as film thickness and pressure drop.

12.10 ENTRAINll'IENT
QUAUTATiVE OIBSIER\l'AT!ONS

Imagine that a liquid film is caused to flow down the wall of a vertical
duct or along the bottom of a horizontal or inclined channeL Then sup-
pose that gas is blown down the same duct over this film. What happens
as the gas velocity is increased?
At first, low gas velocities have little effect on the film. However,
with increasing relative velocity a destabilizing effect is noted 1 in accord-
ance ·with the discussion in Sec. 6.4. Specifically, c2 becomes more nega-
tive with an increase in relative velocity inEq. (6.83); therefore Vw 2/u 2 - l
is larger in Eq. (6.163) and hence a is larger in Eq. (6.162). In horizon-
tal or inclined channels 1 gravity acts as a restoring force and may delay
the onset of noticeable wave activity. The first waves to appear are
small ripples traveling in the direction of the film. Higher velocities of
the gas lead to an increase in the amplitude of these ripples and soon
three-dimensional disturbances are generated. The interface now has a
"pebbled" or "cross-hatched" appearance similar to the waves which are
obtained with light squalls on rivers or lakes. 23 , 33
At a gas velocity which is about double that which is necessary to
produce the cross-hatched wave pattern, the first roll waves appear.
These have a much larger amplitude and velocity than previous waves
and appear to ride over the top of the more uniform small amplitude
waves. Roll waves have a steep front and a long region of relatively
quiet fluid between crests. They have been the subject of severa! theo-
retical and experimental studies. 34- 37
When the gas velocity is sufficiently high, the drag forces on the
tops of rol! waves are adequate to pul! off droplets of liquid that become
entrained in the gas stream. The onset of entrainment is usually preceded
by a noticeable roughening of the tops of the rol! waves which resemble
patches of "white water" rushing over the top of the film. Further
increase of gas rate increases the entrainment and also has the effect of
DROP FLOW 387

Horizontal duct 12"wide x 5"high

80
o

o. r◊Onset of droplet entrainment

o o o o
0
~60 o o o o o
u ºº o o o o
o
-¡¡ .,(Onset of roll waves
~ 40 o
<(
\:><'-x..
., x ~Onset of cross-hotch regime
20
~;-;-"-..L_:-~--"-•======·===::::===·---•
.,~Onset of "ripple flow
O
di
oL__.J.__ _j__j__J__j__ __[__J_ _j__L__.J.__ _ j _ _ L _ j _ _ j __ __[__J
o 1.0 2.0 3.0 4.0
Liquid flow rote, gpm

Fig. 12.3 Regions of wave activity in stratified air-water flow. (Bemberis. 33 )

decreasing the film thickness as a result of both depletion and the higher
interfacial shear and consequent augmented liquid velocity.
Figure 12.3 shows a plot of the regions of various wave activity
for air flowing over water in a horizontal duct 12 in. wide by 5 in. high.
The liquid rate is seen to have little influence on the transition points
as long as it is sufficiently high.
The onset of droplet entrainment can be detected by noticing drop-
lets striking the unwetted walls of the duct or by detecting drops in the
gas stream. A more precise measurement can be obtained by sampling
a given area of the gas stream by means of an extraction pro be and meas-
uring the amount of liquid which enters the probe.
For example, Fig. 12.4 shows entrainment data taken by Steen 36
in a 4-in.-diameter vertical tube, and Fig. 12.5 shows ,vallis 188 data for a
vertical tube of 0.875 in. in diameter. The ordinate of these graphs is
the percent entrainment defined by Eq. (11.114). Thc air velocity shown
is the volumetric flux .i,, based on the total tube area. The decrease in
critica! gas velocity at which entrainment starts in Fig. 12.5 at high liquid
rates is probably dueto the reduction of available area for gas flow because
of the thickness of the liquid film.
It is noticeable that the critica! gas velocity in both figures displays
little sensitivity to duct dimensions, orientation, or liquid f:low rate.
Corresponding entrainment results in upward vertical flow have
already been shown in Fig. 11.17. Here the persistence of liquid slugs
obscures the critica! gas flow rate for film instability except at very low
liquid rates.
"' ONE-DIMENSIONAL TWO-PHASE FLOW

17.5 1
' 1
□
o
1 •1
•
15 r---- Water rote
855 ml/min
' □
o 1695 o
r • 2480
o 3232 □

□ 3980
•
e o
~ 10 o
e
·g .
•
e □
o '
ID
r o -
;;-e
•
□ o
'
5
. '
□ o
o
r
• '
~ o
~ '
o ~
' 80
'
90
'
l00
'
IIO
65 70
Air velocity, fps

Fig.12.4 Percent entrainment versus a.ir velocity for 4-in.-ID tube. Sampling
at center of tube. (Steen. 36 ) Cocurrent downfl.ow.

THE IEFFIECT OF iNLET CONDITlONS AND TUBE. LENGTH

The results shown in Fig. 12.4 were ohtained alter considerahle care had
been taken to prevent entrainment originating where the air and water
were introduced into the apparatus. Therelore, the data should give a
true measure of film stability. In man y practical cases, however, this
effect is completely masked by inlet entrainment. Because droplets
which are carried along in a gas stream have a high axial but low radial
velocity, and because o! splashing which occurs when they hit the liquid
film, a very long duct is necessary belore the effects of inlet conditions
can be considered to be negligible. 39 For example, the results of Gill,
Hewitt, and Hitchon 40 (Figs. 12.6 and 12.7) show how the total entrained
flow rate and mass velocity profiles vary up to 209 in. downstream from
the position at which the liquid film was placed on the wall of a 17;(-in.-
diameter tube.

DEFlNITiON OF A CRiTICAL GAS VE.LOCiTY

A critical gas velocity for the onset of entrainment can be specified as long
as entrainment results from film stability and from no other causes.
Accurate definition of this velocity is not quite as simple as it at first
DROP FLOW 389

appears. Since zero entrainment cannot be measured, the critical veloc-

ity will usually be chosen as the place where the first reasonable amounts
of entrainment occur. Referring to Fig. 12.4 we can see that this defini-
tion, for the case of 855 ml/min water flow rate, would give values for
the critica! velocity of 65, 75, or 87 fps depending on whether the measure-
ment Sensitivity is 0.1, 1, or 2.5 percent entrainment. Because of the
statistical nature of turbulent velocities in the gas and the liquid and also
the spectrum of waves on the interface) an occasional combination of cir-
cumstances can result in a drop being entrained at quite low gas velocities.
N evertheless, almost ali entrainment data display the characteristics
shown in Fig. 12.9, that is,

l. A region of negligible entrainment

2. A region of slowly increasing entrainment, with upward curvature
50
Vertical downflow "l
0.875"ID x 71" tube
rubber hose inlet
probe dio. 0.094"
Water flow rotes
40 o 320 cc/rnin
• 480
o 640
V 960
" 1280
V 4000

30 o 8450
~
e
o
E
e
"fe
©

'"' 20

o o
o

OL__ _L.L_--11,é.._J_ ___¡_ __J

o 100 200
Fig. 12.5 Entrainment dat.a for air-water down- Airflux,¡~, fps
flow in a 0.875-in.-diam tube. (Wallis. 38 )
 ONE-DIMENSIONAL TWO-PHASE FLOW
""
3. An approximately linear region in which entrainment increases steadily
with increasing gas velocity
4. Saturation at high gas velocities where the percent entrainment
maches a limit

The critica! velocity is then arbitrarily defined as the point where

extrapolation of the linear portian of the curve hits the gas velocity axis.

PREDICTION OF THE CR!TICAL GAS VELOCITY

Numerous competing correlations 37 • 39 , 41 , 42 , 43 , 44 exist for predicting the

onset of droplet entrainment, and a completely general synthesis of these
results is yet to be achieved. Steen 36 varied the gas pressure and found
that the critica! velocity varied approximately inversely as the square
root of the gas density (Fig. 12.8). The liquid viscosity was found to be
unimportant above a critica! liquid flow rate (Fig. 12.9), but the surface
tension was very important, being almost proportional to the critical
velocity. As a result of these experiments Steen suggested the following
criterion for the onset of entrainment:

. ( )l' =
J,µ,
<r
Pe_
Pt
' 71"2 > 2.46 X 10-• (12.43)

This equation is not universally valid and is incorrect if the liquid occu-
pies a significant proportion of the duct, or if viscous forces in the liquid
are significant.

600

500
e
~
..___
~ 400
e-
-1'o
• 300
-u
ID

·",g200
e
~

100
ª
o
o 50 100 150 200 250
Distance from inlet, in.

Fig. 12.6 Integrated flux of entrained water through center 1.05 in. of 1.25-in.-diam
tube as a function of distan.ce from inlet (AERE-R3954).
DROP FLOW
"'
Height, in. Height, in.
+ 6
, 24 ♦., 115
133 1•
o 42 6 151
• 60 ◊ 173
o 78 ? 191 +
150f------+- - - - + - - -·- - • 97 111 209 - - - - - + - - - -

e·
'o 209''
'"; 100 f-----H------t
X
191"
o 173"
c;c
151"
o 133''
E
I 97"
------tl,[],.. 78 ''
50

Fig. 12.7 Water mass velocity profiles as a function of distance from inlet (AERE-
R3954).

Steen' s parameter forros the basis of an approximate correlation of

equilibrium percent entrainment in long pipes which was derived by
Wallis" as a modification of the method of Paleev and Filippovich 44 and
is shown in Fig. 12.10. This curve is close to most published data in the
thick film regime for entrainment levels below 50 percent. Above this
value numerous secondary effects of pipe Iength, method of experimenta-
tion, etc., give rise to considerable scatter. Thin 1 viscous films are more
stable (Fig. 12.9) and give lower amounts of entrainment than are pre-
dicted by using Fig. 12.10.

DROPLET CONCENTRATION AND VIELOCITY DISTRIIBUTiONS

Droplet concentration and velocity profiles have been measured in sorne

detail at Harwell "· 45 and CISE. 46 The droplet concentration measured
392 ONE-DIMENSIONAL TWO-PHASE FLOW

60
o
o
Water flow rote, 0.25 gpm
Exit pressure
o 1 otm
40 ' 2 otm
:, □ 3 otm
w
E + 4 otm
e
·2
ew
"' 20

o
o
o 100 150 200 250
Air flux, fps

Fig. 12.8 The effect of pressure on entrainment. (Steen. 36 )

120
GE SF-96(5) silicone fluid
o
4158 rnl/min a
100 3036
8 2060 o
o
'
• 1100

// i
80
~
e
w .seo
328 1 / : "
E + +
e rr=20dynes/cm"
·g60 v=5 es + + + +
e +
w
a'
40 o

/·/. o
20

o • 8
./
. . //'
+
+

o 20 40 60 80 100 120 140 160 180 200

Aír flux,;~, fps

fig. 12.9 Entrainment as a function of air fluxfor various liquid flow rates. (Steen. 36)
DROP FLOW

1'
w
E
e
·g 40
e
w

"

o \ _ _ ¿ _ L _ _ L _ _ L _ _ L __ _¡___ _¡____J
2x10-4 4 6 8 10 12 l4xl0- 4

Fig. 12.10 Wallis 43 correlation for equilibrium

entrainment.

at Harwell was found to be virtually constant throughout the flow. The

corresponding ve!ocity profiles showed a tendency to become more and
more parabolic as entrainment increased (Fig. 5.2). The velocity profiles
fit Eq. (5.8) remarkably wel! except that the value of the constant k
decreased from 0.4 to as low as 0.1 for the greatest droplet concentration.
For a power !aw form of correlation the velocity defect with entrainment
follows approximately a square law with radius as though the flow had
become "laminarized," with the fine droplet dispersion resembling the
classic kinetic model of a perfect gas.

l'l!OBLEMS
12.1. Water condenses on the horizontal ceiling of a shower stall where the temperature
is 90ºF. What size droplets will eventually drip down from this surface?
12.2. A garden hose with a spray head discharges 4 gal/min through 30 boles with
}ú in. diam. Estímate the droplet size (a) in still air and (b) in a 60-mph gale.
12.3. Fuel leaks from a small hole in the tank of an aircraft flying at 20,000 ft above
sea level at a speed of 550 mph. Estimate the size of droplets which are formed.
12.4. vVhat is the maximum stable size of a falling raindrop? What is its terminal
velocity?
12.5. What is the maximum stable size of mercury droplets falling through water?
 ONE-DIMENSIONAl TWO-PHASE FLOW

12.6. A countercurrent flow scrubber to clean :flue gases is to be installed in a powe:r

station. If the flue-gas velocity is 10 fps at a temperature of 500ºF1 what size of
water droplet will ensure downflow of liquid at the bottom of the scrubber? What
is the maximum downward velocity with which these droplets can be sprayed from
nozzles if their temperature is constant at 50ºF and they are not to break up?
12.7. It is found by experiment that fuel droplets larger than 15-µ diam will fail to
"turn the corner'' at the point where an automobile carbUretor joins the inlet manifold
and will be deposited on the wall. This defeats the purpose of the carburetor as a
droplet dispersa} device and increases the time lag in engine response to a change in
throttle setting. Estimate the fraction of thc fuel which is supplied to the carburetor
which will be deposited on the wall as a function of the air-fuel ratio and the air
velocity at the fuel injection point in the carburetor.
12.8. Compare Eqs. (12.25) and (12.27) with the "best fit" normal and lag-normal
probability distributions of the form

p(d)

p(d)

d is an average diameter and u measures the "sharpness of the distribution.

11

12.9. Drop size spectra can be measured experimentally by taking high-speed pho-
tographs. Unfortunately1 this technique provides only an instantaneous sample of an
area of space and <loes not give the true size spectrum if the droplets are moving with
different velocities. Show how the measured spectrum is distorted (a) if drops are
being accelerated in a high...speed gas :flow and are photographed while they are still
accelerating and (b) if droplets are falling in still air at their terminal velocities.
Assume that the photograph is taken across the direction of motion.

+ Water,¡, 12 mm dia
0 Water rf, 6 mm dia
4~*o o Alcohol ,t, 12 mm dio
io~
o

Core momentum flux, PcJ/, gm/(cm)(sec 2 )

Fig. 12.11 Entrainment correlation of Minh and Huyghe. 48

DROP FLOW

Table 12.1 Entrainment and pressure-drop data of Cousins, Den ton, anti Hewitt. 49
Air and water Howing in a %~in.-diam tube. W. + W 1 = total water flow rate

Distance from injector, in.

W,, W" + W1,
lb/hr lb/hr
6 30 71H 107H 141 180 216
--- --- --- --- --- ---
40 100 0.8 3.3 4.0 5.0 4..5 5.1 5 7
230 5.7 8.6 12.8 16.0 20.9 24.7 33.9
50 100 1.2 5.2 7.2 7.8 8.7 11.1 13.7
230 10.1 15.0 23.8 29.8 37.2
60 100 1.6 8.9 14.5 14.4 18.8 20. 1 30.0
230 !O.O 23.3 44.7 52.7 70.5
70 100 2. 5 12.8 21. 7 24.3 28.1 34.0
230 12.7 37.9 67.0 85.6

(a) Entrainment data: the numbers in the table give the values of We
in pounds per hour.

DMtance from injector, in.

lVu, W, + WJ,
lb/hr lb/hr 138 177 213
9 27 68.!,-i 104},-§
---
40 100 39.59 38.62 36.62 34.58 33.40 31.63 29.72
230 39.27 37.62 33. 73 30.26 27.43 23.48 19 95
50 100 39.39 38.21 35. 76 33.35 31.42 28 25 20.15
230 38.78 36.87 32.06 27.86 24.47
60 100 39.20 37.77 34.17 31.63 28.39 25.38 21.52
230 38.55 36.01 29.78 25.57 21.03
70 100 39.00 37.14 32.83 29.74 26.55 22.19
230 38.21 35.51 27.52 22.82

(b) Pressure-drop data: the numbers in the table give the pressure in
pounds per square inch absolute.

12.10. Drop fluidization is a possible reason for the occurrence of a maximum heat
flux during natural convection boiling from a horizontal surface. Use Eq. (12.40) to
deduce the maximum heat flux, assuming that all the heat supplied at the surface
forms vapor near the surface and that the liquid is at saturation temperature.
[Kutateladze 47 found that the empirical value of the coefficient in Eq. (12.40) in this
case is 0.16.]
12.11. Gas is bubbled through a stagnant pool of water. Compare the gas flux needed
to cause flooding in bubbly flow with the flux needed to cause drop fl.uidization.
12.12. What values of droplet diameter and the index nin Eq. (4.14) will give agree-
ment with the right-hand curve in Fig. 12.2?
12.13. Compare the entrainment correlation shown in Fig. 12.10 with the data in
Figs. 12.5, 12.8, 12.9, and 11.17.
 ONE-DIMENS!ONAL TWO-PHASE FLOW
"'
12.14. A steam-water mixture at 500 psia fl.ows in a 1-in.-diam vertical pipe. Estimate
the equilibrium level of entrainment in the annular-mist regime as a function of
quality and mass flow rate. Using these values calculate the increase in apparent
density of the vapor core due to the presence of droplets.
12.15. A seaman observes that a salt spray is being blown off the tops of waves during
a storm. Estimate the minimum wind velocity.
12.16. Figure 12.11 shows the entrainment data of Minh and Huyghe 48 for air-ethyl
alcohol and air-water mixtures in annular :flow in pipes of 6 and 12 mm diam. Assum-
ing that the temperature was 20ºC, for which the property values are µª = 0.018 cp,
p¡ = 1 g/cm 3, rr = 72.8 dynes/cm forwater, and p¡ = 0.79 g/cm 3, u = 22.'3 dynes/cm
for alcohol, show that the data are consistent with the correlation scheme shown in
Fig. 12.10 but that the actual levels of entrainment are lower. Discuss possible
reasons for the disagreement. Pe and j are the density and volumetric flux for the
0

core flow assuming that it is a homogeneous mixture of droplets and· gas.

12.17. One difficulty with entrainment prediction is that equilibrium is never reached
because, due to decreasing pressure or phase change, the gas flux increases steadily
along the duct. Table 12.1 illustrates this effect using sorne data of Cousins, Denton,
and Hewitt 49 taken with air and water in a %-in.-diam tube. Compare these results
with Fig. 12.10 and show how the experimental curves move to the right as non-
equilibrium effects become more important.

IIEFEIIENCES
l. Rayleigh, Lord: Proc. London 11,;[ath. Soc., vol. 10, p. 1, 1878. 4, Proc. Roy. Soc.,
(London), vol. 29, p. 71, 1879. Phil. Mag., vol. 34, p. 177, 1892. For a discussion
summarízing the work, see Sir Horace Lamb's uHydrodynamics," 6th ed., pp.
471-473, Dover Publications, Inc., New York.
2. Lewis, H. C., D. G. Edwards, M. J. Goglia, R. I. Rice, and L. W. Smith: Ind.
Eng. Chem., vol. 40, no. 1, p. 67, 1948.
3. Haas, F. C.: A. I. Ch. E. J., vol. 10, pp. M920-924, 1964.
4. Prandtl, L.: "Essentials of Fluid Mechanics," p. 328, Blackic & Son, Ltd.,
Glasgow, 1953. ·
5. Hinze, J. O.: Appl. Sci. Res., vol. Al, pp. 263-2721 1948.
6. Hinze, J. O.: Appl. Sci. Res., vol. Al, pp. 275-288, 1948.
7. Masugi Isshiki, N.: Rept. 35, Transportation Technical Research Institute,
Tokyo, Japan, July, 1959.
8. Dickerson, R. A., and T. A. Coultas: AIAA paper no. 66-611, June, 1966.
9. Giffen, E., and A. Muraszew: "The Atomisation of Liquid Fuels," Chapman &
Hall, Ltd., London, 1953.
10. Nukiyama, S., and Y. Tanasawa: Trans, Soc. Mech. Engrs. (Japan), vol. 4, no.
14, p. 86, 1938.
11. Sauter, J.: NACA Rept. TM-.118, 1929.
12. Soo, S. L.: Ind. Eng. Chem. Fundamentals, vol. 4, pp. 426-433, 1965.
13. Kliege1, J. R.: Iniern. Symp. Combust, 9th, Academic Press, Inc., New York,
pp. 811-826, 1963.
14. Dallavalle, J. M.: "Micromeritics/' Pitman Publishing Corporation, New York,
1943.
15. Nukiyama, S., and Y. Tanasawa: Trans. Soc. Mech. Engrs. (Japan), vol. 5, no.
18, p. 63, 1939.
16. Ingebo, R. D.: NASA transl. D-290, June, 1960.
DROP FLOW 397

17. Tribus, M.: "Thermostatics and Thermodynamics," D. Van Nostrand Company,

Inc., Princeton, N.J., 1961.
18. Shapiro, A. H., and A. J. Erickson: Trans. ASJ.11E, vol. 79, p. 775, 1957.
19. MacVean, S. S.: unpublished work, Dartmouth College, Hanover, N.H., 1967.
20. Levich, V.G.: "Physicochemical Hydrodynamics," pp. 409-429, Prentice-Hall,
Inc., Englewood Cliffs, N.J., 1962.
21. Levich, V. G., Ref. 20, p. 431.
22. Pratt, H. R. C., et al.: Tr_ans. Inst. Chem. Engrs., vol. 29, pp. 89-148, 1951;
vol. 31, pp. 57-93 and 289-326, 1953; vol. 35, pp. 267-342, 1957.
23. Wallis, G. B.: discussion of paper no. 271 lntern. Heat Transfer Conf., Boulder,
Colo., pp. D-70-D-72, ASME, 1962.
24. Wallis, G. B.: paper no. 3, Two-phaseFluidFlow Symp., Institution of Mechanical
Engineers, London, 1962.
25. Gaylor, R., N. W. Roberts, and H. R. C. Pratt: Trans. Inst. Chem. Engrs., vol.
31, pp. 57-68, 1953.
26. Blanding, F. H., and J. C. Elgin: Trans. A. I. Ch. E., vol. 38, pp. 305-338, 1942.
27. Thornton, J. D., and H. R. C. Pratt: Trans. Insl. Chem. Engrs., vol. 31, pp. 289-
326, 1953.
28. Elgin, J. C., and F. l\!L Browning: Trans. A. l. Ch. E., vol. 31, p. 639, 1935;vol.
32, p. 105, 1936.
29. Crawford, J. W., and C. R. Wilke: Chem. Engr. Progr., vol. 47, p. 423, 1951.
30. Kutateladze, S. S., and V. N. Moskvicheva: Zh. Tech. Fiz., vol. 29, pp. 1135-1141,
1959.
31. Taylor 1 G. I.: Proc. Roy. Soc. (London), vol. Al48, p. 141, 1932.
32. Hanratty, T. J., and J. l\!L Engen: A. l. Ch. E. J., vol. 3, pp. 229-304, 1957.
33. Wallis, G. B., J. M. Turner, I. Bemberis, and D. Kaufman: AEC Rept. NYO-
3114-4, 1964.
34. Hanratty, T. J., and A. Hershman: A. J. Ch. E. J., vol. 7, pp. 488-497, 1961.
35. Chung, H. S., and W. Murgatroyd: Symp. Two-phase Flow, Exeter, England, vol.
2, pp. A201-A214, June, 1965.
36. Steen, D. A., and G. B. Wallis: AEC Rept. NYO-3114-2, 1964.
37. Van Rossum, J. J.: Chem. Eng. Sci., vol. 11, pp. 35-52, 1959.
38. Wallis, G. B.: General Electric Company, Rept. 62GL127, Schenectady, N.Y.,
1962.
39. Wicks, Jvl., and A. E. Dukler: A. l. Ch. E. J., vol. 6, pp. 463-468, 1960.
40. Gill, L. E., G. F. Hewitt, and J. W. Hitchon: UKAEA Rept. AERE-R3954, 1962.
41. Zhivaikin, L. I.: Khim. Mashinostr., vol. 6, p. 25, 1961.
42. Mozarov, N.A.: Teploenergetica, vol. 4, p. 60, 1961.
43. Wallis, G. B.: discussi.on of Ref. 44, Intern. J. Heat Mass Transfer, vol. 11, pp.
783-785, 1968.
44. Paleev, I. l., and B. S. Filippovich: Intern. J. Heat 11fass Transferi vol. 9, pp.
1089-1093, 1966.
45. Gil!, L. E., G. F. Hewitt, and P.M. C. Lacey: UKAEA Rept. AERE-R3955, 1963.
46. Cravarolo, L., A. Hassid, and E. Pedrocchi: CISE R-109, 1\.filan, Italy; 1964.
47. Kutateladze, S. S.; Izv. Akad. Nauk SSSR, Otd. Tekhn. Naulc, no. 4, p. 529, 1951.
48. Minh, T. Q., and J. D. Huyghe: Symp. Two-phase Flow, Exeter, England, vol. 1,
pp. C201-C212, June, 1965.
49. Cousins, L. B., W. H. Dentan, and G. F. Hewitt: Symp. Two-phase Flow, Exeter,
England, vol. 2, pp. C401-C430, June, 1965.
 appendix:

JU SOII/IE USEFUL CONIIEIISION FACTORS

Length 1 cm = 0.394 in. = 0.0328 ft

Volume 1 cm' = 3.51 X 10- 5 ft'
Mass 1 g = 2.205 X 10- 3 lbm
Density 1 g/cm 3 = 62.43 lbm/ft 3
Force 1 dyne = 2.248 X 10-, lb 1
Pressure 1 atm = 14.7 psia = 1.013 X 106 dynes/cm 2
Viscosity l lbm/(sec)(ft) = 14.88 g/(sec)(cm) (or poise)
Surface tension 1 dyne/cm = 6.85 X 10- 5 lb 1/ft
Energy 1 kwhr = 3413 Btu
1 watt/cm' = 3.71 Btu/(hr)(ft')

A.2 SOII/IE USEFUL NUI\IIBERS AND PROl'ERTIES

Gravitational constant on earth 981 cm/sec 2 , 32.2 ft/sec 2

Surface tension oi water in air at 20ºC 72.6 dynes/cm
Gas constant far air 53.3 (ft)(lb 1 )/(lbm)(ºF)

'"
 Ta'ble A.1 Properties of saturated steam and saturated watert
Volume, ft• /lbm Enthalpy, Btu/lbm Ji]ntcopy, Btu/(lbm)(ºF)
Press, Temp., Water Evap. Steam Water Evap. Steam Water Evap. Steam
psia ºF V¡ V Jo v, h¡ h¡, h, 8¡ s¡, s,
3208.2 705.47 0.05078 0.00000 0.05078 906.0 o.o 906.0 1.0612 0.0000 l. 0612
3000.0 695.33 0.03428 0.05073 0.08500 801.8 218.4 1020.3 0.9728 0.1891 l. 1619
2500.0 668. ll O. 02859 0.10209 0.13068 731. 7 361.6 1093.3 O. 9139 0.3206 1.2345
2000.0 635.80 O.02565 0.16266 O. 18831 672.1 466.2 !138. 3 0.8625 0.4256 l. 2881
1500.0 596.20 0.02346 0.25372 0.27719 6!1.7 558.4 !170.1 0.8085 0.5288 l. 3373
1000.0 544.58 0.02159 0.42436 O. 44596 542.6 650.4 !192.9 0.7434 0.6476 l. 3910
700.0 503.08 0.02050 0.63505 0.65556 491.6 710.2 1201. 8 0.6928 0.7377 l. 4304
500.0 467.01 0.01975 0.90787 O. 92762 449.5 755.1 1204.7 0.6490 ó.8148 l. 4639
400.0 444.60 O. 01934 1.14162 l. 16095 424.2 780.4 1204.6 0.6217 0.8630 l. 4847
300.0 417.35 0.01889 1.52384 l. 54274 394.0 808.9 1202.9 0.5882 O. 9223 1.5105
200.0 381.80 0.01839 2.26890 2.28728 355.5 842.8 !198. 3 0.5438 1.0016 1.5454
100.0 327.82 0.017740 4.4133 4.4310 298.5 888.6 !187. 2 0.4743 l. 1284 1.6027
80.0 312.04 0.017573 5.4536 5. 4711 282.1 900. 9 !183 .1 0.4534 l. 1675 1.6208
60.0 292.71 0.017383 7 .1562 7. 1736 262.2 915.4 !177 .6 0.4273 l. 2167 l. 6440
40.0 267.25 O. 017151 10.4794 10.4965 236.1 933.6 1169.8 0.3921 l. 2844 l. 6765
20.0 227.96 0.016834 20.070 20.087 196.27 960.1 1156. 3 0.3358 l. 3962 l. 7320
14.696 212.00 O. 016719 26.782 26.799 180.17 970.3 !150. 5 0.3121 l. 4447 l. 7568
!O.O 193.21 O.016592 38.404 38.420 161. 26 982.1 !143. 3 0.2836 1.5043 1.7879
8.0 182.86 0.016527 47.328 47.345 150.87 988.5 1139. 3 0.2676 1.5384 1.8060
6.0 170.05 0.016451 61.967 61. 984 138.03 996.2 !134. 2 0.2474 1.5820 1.8294
4.0 152.96 0.016358 90.63 90.64 120.92 1006.4 !127. 3 0.2199 1.6428 1.8626
2.0 126.07 0.016230 173.74 173.76 94.03 1022.1 1!16.2 0.1750 l. 7450 l. 9200
1.0 101. 74 O. 016136 333.59 333.60 69.73 1036.1 !105.8 O.1326 l. 8455 l. 9781
0.50 79.586 0.016071 641.5 641.5 47.623 1048.6 1096.3 0.0925 l. 9446 2.0370
0.20 53. 160 0.016025 1526.3 1526.3 21.217 1063.5 1084. 7 0.0422 2.0738 2. !160
O. !O 35.023 O. 016020 2945.5 2945.5 3.026 1073.8 1076.8 0.0061 2.1705 2 .1766
ti t From ASME Steam Tables, 1967, by permission.
Table A.2 Properties of air at low pressurest,t

k,
,,, ,,, µ X 101, =
,, ,,
Btu/ Pr
T,
ºR
T,
ºF
Btu/
(lb)(ºF)
Btu/
(lb)(ºF)
,,=-
,, fps
lbm/
(sec) (ft)
(hr)(ft)
(ºF)
cpµ/k
K

500 40.3 O .2396 O. 1710 1.401 1096 .4 118 O. 0143 0.71

550 90.3 0.2399 o 1713 1.400 1149.6 126 0.0156 0.70
600 140.3 O. 2403 O. 1718 1 399 1200. 3 135 O. 0168 o 70
650 190.3 O .2409 O. 1723 1 398 1248. 7 143 0.0180 0.69
700 240.3 0.2416 O. 1730 1 396 1295. 1 151 0.0191 0.68
750 290.3 0.2424 o. 1739 1 394 1339 6 158 0.0202 0.68
800 340.3 O .2434 O. 1748 1 392 1382.5 166 0.0213 0.68
900 440.3 O .2458 O. 1772 1 387 1463.6 179 0.0237 0.67
1000 540.3 0.2486 O. 1800 1 381 1539.4 192 0.026 0.66
1100 640.3 0.2516 O. 1830 1 .374 1610. 8 205 0.028 0.66
1200 740.3 0.2547 O. 1862 1 368 1678.6 218 0.030 0.66
1300 840.3 0.2579 O. 1894 1 .362 1743.2 230 0.032 0.66
1400 940.3 0.2611 o .1926 1 .356 1805 .0 242 0.035 0.65
1500 1040. 3 0.2642 o. 1956 1 350 1864. 5 253 0.037 0.65
1600 1140.3 0.2671 O. 1985 1 345 1922. O 264 0.039 0.65
1700 1240.3 O .2698 0.2013 1 340 1977.6 274 0.041 o 65
1800 1340.3 0.2725 O .2039 1 336 2032 284 0.043 0.65
1900 1440.3 0.2750 0.2064 1 .332 2084 293 0.045 0.65
2000 1540. 3 0.2773 O. 2088 1 328 2135 302 0.046 0.65
2100 1640.3 0.2794 0.2109 1 325 2185 311 0.048 0.65
2200 1740.3 0.2813 0.2128 1 322 2234 320 0.050 0.65
2300 1840 .3 0.2831 0.2146 1 .319 2282 '29 0.052 0.65
2400 1940.3 0.2848 0.2162 1 .317 2329

t The value of the gas constant R is 53.3 (ft)(lb¡)/(lb,,,)(ºF).

+From J. H. Keenan and J. Kaye, "Gas Tables," John Wiley & Sons, Inc., New York, 1948 (by
permission).

50

E
~ 4QL_ _ _ _ _ _ _ _ _ __¡__~::,__ _ _ _ _ _ _ _L _ _ _ _ _ __ j
ru
e
~

"e
-~ 30
e
2
ru
u
o
't: 2 0 ' - - - - - - - - - - - - - ' - - - - - - - - - - - - ' - ' - - - - - - - - - "
o
~

10

Oc__ __c__ _¡__L_L_L.L.C..Li_ _ _L___J_ _¡__¡_L-'_LLL_ __L_-'>.__J__J

10 20 30 50 100 200 300 500 1000 2000 5000

Pressure, psio

Fig. A.1 Surface tension of saturated water (1 dyne/cm = 6.85 X 10-• lb/ft).
 400 40
µ tor water 35
l.O psia 12,000 psia
30
10,000 psia
0 250 7500 psi □ 25
5000 psi □
X
'-:::::' 200 t - - - - - - + - - ~ - - - - j - - - - - - + - - LUUU ps1a------- 20 o
'l Saturated water 17_5 ~
175
~ ~
u 150 - 15.0-;:;-
. ~

:': 125 12.5~

~~ u
-.
=- 9IDO>-----+----+------'>---+-------->- +++-'k-"--e;: 10.0 :':
i 0 12 000 psi a 9.0 ::e'
i 80 B.O~
º§ 70 7.0 "-
6.0 e
1
"§
>~ 50
6ºt SaLrated water ~ IW'f steam
u. ~
5.0
>
40 4.0
3.5
Saturated steom 3.0
2.5

'---'--'---'---'---'--'-.l..<C::..._...L_ _ L_ ___[__l_ __L__L_---L-,-_L-'-'-'---- 2.0

32 40 50 60 70 80 100 150 200 300 400 600 800 1000 1500
Temperature, ºF

Fig. A.2 Viscosi~y of steam and water. (From 1967 ASME Steam Tables, by permission.)
íi
402 ONE-DIMENSIONAL TWO-PHASE FLOW

40
30
20

IO
8.0
6.0
4.0
3.0
2.0
M

2 l.Of------------l---'\------,,.-,.-::_-----7"',,__---h,L-

l x 0.80
0.60
0.40
,,;
0.30
i
·;;:; 0.20
§
·;;
._g 0.101-----------+-------
~ 0.080
~ 0.060

" 0.040
0.030
O.ü20

0.001 L__L___j___j__j__[_..J_L__ _L_ _j__ _J__ _L__j___j__l_[_..J_L__ _J

32 40 50 60 80 IOO 150 200 300 400 600 8001000 1500
Ternperoture, ºF

F'ig, A.3 Kinematic viscosity of steam and of water. (From 1967 ASME Steam
Tables, by permission.)
lndex
lndex

Aggregative flows, 229 Bubbly flow:

Air-lift pump, 41 shock \vaves, 274
Annular flow, 63, 113, 315 speed of sound, 144, 264
countercurrent, 330 unsteady, 134
differential analysis, 330 vertical, 24 7
flooding, 336-345 void fraction, 95, 252
integral analysis, 326
interface, 366
pressure drop, 324, 350-368
Choking, 40, 46, 67, 85
regime boundaries, 316, 345-351
compound, 68
void fraction, 324, 348, 351-368
Compressibility of fluidized bed,
Apparent mass, 219
206
Atomization, 376
Comprcssibility wave, 25, 141, 143
Average volumetric concentration,
Concentration profile, 97, 106, 391
11
Condensation, 77, 84, 131, 165
Contact angle, 164
Continuity equations, 61
Continuity waves (see Waves, con-
Baroczy correlation, 59
tinuity)
Basset force, 221
Coordinates, 14
Boiling liquid, 32
Correlation, 5, 45
flow of, 55
Baroczy, 59
Bubble, 7, 92
Martinelli, 29, 49, 51, 66, 328,
agglomeration, 260
353
in fluidized bed, 232
Countercurrent flow, 18, 92, 330
formation, 244-247
Critical gas velocity far entrain-
nucleation, 164
ment, 388-391
rise velocity, 248-255
Bubbly flow, 95, 243
churn turbulent, 255
ideal, 255 Density inversion, 190
momentum flux, 262 Differential analysis, 6
in nozzles, 271 Dimensionless groups, 107, 114
in pipes, 269 Distribution parameter, 97, 255
pressure drop, 262, 267 Drift flux, 6, 13, 89, 93, 176, 248,
regime boundaries 1 265 257,291
405
406 ONE-DI MENSIONAL TWO-PHASE FLOW

Drift velocity, 13 Gas, pseudo 1 21

Drop: Gas-particle flow, 21, 80, 207
evaporation, 170
fluidization, 384
formation, 376
size spcctra, 378 Homogeneous equilibrium flow, 18,
terminal ve!oeity, 381 209
Drop flow, 375 Homogeneous flovv, 5, 17, 3S5
flooding, 382 isothermal, 269-274
pressurc drop, 385 shock ·waves, 153
vertical, 382 wavcs, 143

Encrgy cquations 1 18, 45, 80

Influence coefficients, 25
Entrainment, 317, 347, 363, 386
Integral analysis, 6, 115, 326
Entropy production, 73-77, 79, 85
Interface, 161
Evaporators, 7, 33
boundary conditions, 161-165
friction factor, 320
shear stress, 165 1 317
Filtration, 202 ve!ocity, 86, 36:l
Flashing, 18, 26, 46, 68, 85
Flocculation, 178, 198, 231
Flooding, 94, 141, 253, 308, 336,
382 Liquid film:
Flow pattern, 7 annular flow, 326, 346, 360
Flmv regime, 6 condensation, 131
boundaries, 8 draining, 125
gravity-dominatcd, 90 falling, 125-127, 331-336
map 1 7 in rotating drum, 157
Fluidization, 64, 180-189, 237,384 around slug flow bubble, 295, 302
stability o!, 205 stability 1 149, 157 1 158, 335
Flux: Liquid-liquid systcms, 253, 382
drift, 6, 13, 89, 93,176,248,257,
291
volumetric, 11
Foam, 253 1 257 l\Iach numbers, 25, 26, 35, 1.59, 214,
Forces on the phases, 62, 77, 91, 270,364
166, 179 l\!fartinelli correlation, 29, 49, 51 1
Friction factor, 26, 28, 320, 361 66, 328, 353
Friction velocity, 108, 169 Mean density, 19
INDEX 407

~finimum kinetic energy, 48 Propagation equation:

~'1inimum momentum flux, 82 for quality, 36
e1Iixing lcngth, 109 for void fraction, 39, 104
l\íomentum equations, 23, 44 1 61, Property index 1 59
65 Pseudo gas, 21, 209, 212, 270
Niomentum exchange model, 85

Quality, 11
Noncondensable gas 1 170
Noncquilibrium effects 1 18
Notation, 9
Reaction frequency, 36
Nozzle flow, 38, 68-73, 87, 207-212,
Regimes of flmv, 6
221, 272
Relative velocity, 13, 91
Relaxation 1 L55, 215
Reynolds flux, 78, 166

Packed bcd, 63, 91, 183, 234

Particle:
drag coefficient, 178, 179 Saltation, 230
influid, 175 Sedimentation, 190-201
force on 1 180 Separate cylinders model, 50, 343,
terminal velocity, 176 3.54
Particle distribution, 228 Separatcd flow, 5, 43, 317, 355
Particle-gas flow, 21, 80, 207 waves, 137, 143
Particle-particle forces, 64, 94, 201 Shape factor, 184
Particle pressure, 201 Shear stress:
Periodic flows, 229 distribution, 111
Phase change, 64 interface, 165, 317
Pneumatic transport, 228 wall, 23, 45, 49, 317, 323
Porous media, 63, 91, 183, 234 Shock ,vaves (see Wavcs, shock)
Pressure drop (gradicht), 14, 24, 40, Sign convention, 101
45, 49 Slug flow, 282
accelcrational, 19, 23 bubble vclocity, 285-291, 299-
fluid bed, 183 302
frictional, 19, 23, 29, 49, 54 in inclined pipes, 304
gravitational, 19 1 24 long bubbles, 294
for particlc suspensions, 222-228 maximum bubble length, 308
(See also Annular flow; Bubbly pressure drop, 284,294,297,302,
flow; Drop flow; Slug flow) 312
... ONE-D1 MENSlONAL TWO-PHASE FLOW

Slug flow: Velocity:

regimc boundaries, 307, 345 minimum transport, 171, 229
unit cell, 283 profile, 97, 106, 109, 391
void fraction, 284, 292, 302 ratio, 46, 221
Slugging in fluidized beds, 230 relative, 13
Soil consolidation, 205, 235 o! sound, 25, 143, 144
Stratified flow, 229, 231 terminal, 93, 176, 248, 381
Superficial gas friction factor, 318, Viscosity of two-phase mixture, 27,
321 37, 222
Surface tension, 162 Void fraction, 11 45, 49, 54
1

(See also Annular flow; Bubbly

flow; Drop flow; Slug flow)
Volumetric flux, 11
Taylor instability, 165, 246
Traffic flow, 128, 148
Two-phase multipliers, 28, 31
Waves, 122
in annular flo\v 366
1

compressibility, 25, 213

Units, 15 continuity, 104, 123, 130, 133,
Unsteady flow, 35, 189, 204, 256 189, 256, 335
shock, 124, 127, 190
dynamic, 123, 135, 137, 141, 170
shock, 152
Variable density model, 115 roll, 149
Velocity: shock, 122, 152, 213-219, 274
drilt, 13 stability o!, 127, 148, 152
interface, 85, 114, 170, 363 in stratificd flow, 139
This book was set in Modern by The Maple Press Company and printcd
on permanent paper and bound by The Niaple Press Company. The
designer \Vas lVIerrill Haber; the dra,vings were done by Engineering-
Drafting Company. The editors were B. J. C!ark and Madelaine
Eichberg. William P. Weiss supervised the production.