Chapter01 - Flow Pattern Transitions in Gas-Liquid Systems - Measurement and Modeling

Chapter 1
Flow Pattern Transitions in Gas-liquid

Systems: Measurement and Modeling
A. E. Dukler
University af Haustan, Haustan, Texas
and
Y. Taitel
Tel Aviv University, Tel Aviv, Israel
1 INTRODUCTION
The need for reliable design methods has been the driving
force behind a very large research effort in two-phase gas-liquid
flow over the past 25 years. This work has been carried out at
universities, national laboratories, and at industrial research and
design organizations in many countries of the world. The result of
this effort has been an extraordinary number of publications on the
subject. Over 7500 papers, theses, and reports have appeared in
these 25 years. Furthermore, the rate of publication has been
increasing in recent years. Most predictive correlations that
have been proposed as a result of all this work have been based
largely on experiment and are valid only under conditions near that
of the experiment. It is precisely for this reason that so many
publications result. Each time a new condition is to be investigated
new experiments are necessary.
The basic dilemma is quite clear. In gas-liquid flow the two

phases can distribute in the conduit in a wide variety of ways not
under the control of the experimenter or designer. When changes
take place in flow rates, fluid properties, conduit shape, or
inclination, this distribution can vary. In addition, the velocities
and shapes of the interfaces are unknown. Consider the problem of
attempting to apply differential equations for conservation of
momentum, heat, or mass to pipe flow where two phases exist.
Because the spacial location of the phases is unknown, it is
impossible to specify which fluid properties should be used at each
point of space and time. Furthermore, the boundary conditions
necessary for solution (interfacial velocities and shapel are
unknown. Under this situation there is little wonder that models
that predict mechanics and heat transfer for two-phase flow are so
elusive.
Ever since the earliest visual observations of two-phase flow,

it. has been recognized that, although the two phases can distribute
in a wide variety of ways, there are natural groupings or patterns
that are observed. Within each pattern the spacial location of the
two phases is more or less the same. Likewise, one set of boundary
conditions would be expected to be common to that pattern. The task
This chapter is dedicated to Ovid Baker, with whom it all started.
G. F. Hewitt et al. (eds.), Multiphase Science and Technology

© Springer-Verlag Berlin Heidelberg 1986 1
2 A.E. Dukler and Y. Taitel
of modeling any one pattern is, thus, mueh simpler than that of
modeling all possible patterns. However, the model developed for
one distribution pattern would not be expeeted to be valid for
other patterns where the location of the phases or the boundary
eonditions could be different.
This approach was first suggested many years ago, shortly after
flow patterns were defined in a formal way. Baker (1954) published
the earliest flow regime map for horizontal pipes. Shortly
afterward, Baker (1958) also suggested that better prediction of
two-phase pressure drop could be obtained by developing equations
or eorrelations to be applied separately to eaeh regime. He was
able to show that even with empirical eorrelations the quality of
the pressure drop predictions using any one eorrelation varied
with the flow pattern to whieh it was applied. Empirical
correetions that were different for each flow pattern improved the
comparison between the calculated and measured pressure drop.
It seems reasonable that this should be so. Not only pressure

drop but also void fractions, heat transfer rates, propagation
rates for apressure disturbance, conditions for choking flow, mass
transfer rates, interfacial stability, residence time distribution,
rates of reaction, pressure oscillations, and all other conceivable
factors of interest must be expected to behave differently as the
flow pattern changes. Attempts to model these phenomena are
suspect unless they are restricted to flow patterns having similar
phase distributions and boundary conditions. Because it is so
likely that models for physical processes must be flow-pattern
specific, it might be a good new administrative rule that papers
be rejected if they propose models or correlations that purport to
be valid for all flow patterns unless they are accompanied by a
suitable generalizing principle. The concept of similitude is one
such principle, but at least with respect to pressure drop (Dukler
et al., 1964) and space average void fraction (Dukler 1969), the
result of applying similitude is not yet totally satisfactory.
If the models created for individual flow patterns are to be

useful, it is necessary to predict the flow pattern that would
exist given any flow conditions of interest (rates, conduit size,
properties, inclination, flow direction). The literature is full
of examples of empirical mapping without any basis in the mechanisms
that are responsible for the transitions". Each result is thus
valid only for a narrow range of pipe sizes, fluid properties, and
flow rates, near the conditions of experiment. After all, this
problem is a modeling task of its own with the validity of the
models that result dependent on the validity of the mechanisms
assumed and the rigor with wnich they are formulated.
Suitably formulated models incorporate in a natural way the

effect of such factors as conduit size and inclination, fluid
properties, and flow direction, and there would be no "sca'le up"
problems to deal with. We are far from that state. No single
model or collection of models has appeared that can predict flow-
pattern transitions for horizontal, vertical, or inclined pipes;
cocurrent or countercurrent flow; flow up or flow downward; steady
or transient; boiling or nonboiling - under flashing or nonflashing
conditions. However, progress is being made, and this chapter will
attempt to describe recent efforts along these lines. The
Gas-Liquid Flow Pattern Transitions 3
predictive methods discussed he re are based largely on the research

done at the University of Houston and at Tel Aviv University over
the past 8 years, which has been directed toward physical modeling
of flow-pattern transitions. There is an extensive literature on
this subject, largely reporting experimental results and empirical
correlations. It is not the intent of this chapter to provide a
comprehensive review of all this work.
2 CLASSIFICATION OF FLOW PATTERNS
The first problem, of course, is to decide what constitutes a

pattern of phase distribution that is spacially "sirnilar" for
different pipe sizes and fluid properties. Certain simple
possibilities suggest themselves. For example, in a horizontal
tube when the liquid flows as an axially cüntinuous layer along the
bottom of the pipe with the gas flowing similarly along the top,
this configuration is designated as the stratified flow pattern.
However, simple experiments show that two such patterns exist:
smooth stratified, where the interface is undisturbed, and wavy
stratified, where the gas-liquid interface is covered by a wave
structure. For a vertical tube the simplest configuration exists
where the gas is distributed as small bubbles reasonably uniformly
in the axial direction in a continuous flowing liquid phase. This
is described as the bubble-flow pattern. Beyond these two simple
cases, the subjectivity of the observer and the purposes in defining
the pattern are both factors in arriving at the definitions. For
example, when the liquid flows as a continuous film around the wall,
with the gas flowing as a core, the pattern is designated as
annular flow. Under some conditions of stratified flow, the liquid-
gas interface viewed in any cross section assurnes acrescent shape,
creeping up along the walls of the pipe. Some investigations have
designated this condition as the "semiannular" flow pattern. At
sufficiently high gas rates, much of the liquid in the film along
the walls during annular flow is dispersed into droplets carried by
the gas. This is sometimes designated as the "mist flow" pattern
or, when a significant liquid film remains on the wall, as "annular
mist" flow. Clearly, with sufficient adjectives, it would be
possible to describe a large series of states ranging from a .smooth
stratified pattern to mist flow. However, that would hardly be
useful - notwithstanding the fact that the literature contains a
disconcerting variety of flow-pattern descriptions, as surnrnarized
below [see, for example, Griffith and Wallis (1961), Duns and Ros
(1963), Sternling (1965), Hubbard and Dukler (1966), Wallis (1969),
Hewitt and Roberts (1969), Govier and Aziz (1972), Oshinowo and
Charles (1974), Taitel and Dukler (1976), Spedding and Nguyen
(1980)] :
Horizontal tubes: Stratified, stratified smooth, stratified

wavy, laminar-stratified, stratified-roll wave, stratified-inertia,
wave, plug, slug, elongated bubble, proto slug, wavy annular,
annular-flow through, pulsating froth, semiannular, annular mist,
spray, homogeneous, bubble, dispersed bubble.
Vertical tubes: Bubble, dispersed bubble, homogeneous, slug,

plug, slug-annular, churn, froth, frothy slug, quiet slug, dispersed
slug, piston, pulsating annular, annular, annular-mist, wispy
annular.
A new observer with an expanded vocabulary could likely arrive

at additional descriptors! However, as discussed in the Introduction,
the objective of flow-pattern classification is to characterize
situations where the phase distributions are spatially similar, so
that one can calculate where the phases are located as functions of
space and/or time. The trick is to make a suitable compromise with
detail to obtain a minimum number of categories that provide useful
starting points for the process of modeling. Only as our ability
to model two-phase flow becomes more refined would it be useful to
define additional patterns. Thus[ the matter of defining flow
patterns can in itself be expected to be time dependent!
In recent years there seems to nave evolved an understanding

of the simplest, useful set of flow patterns, and these are
described belmq.
2.1 Horizontal Pipes: Fig. 1
2.1.1 Stratified
In stratified flow, liquid flows along the bottom of the pipe

with gas along the top. Both phases are continuous in the axial
direction. Two subpatterns are defined: stratified smooth and
stratified wavy.
6"---_ _~6l
STRATIFIED SMOOTH J STRATIFIED
~~ STRATIFIED WAVY
c::
0I
o~
ELO"GATEO BUB:
INTERMITTENT
--------=--- j
/ : . ' , '•• (
~ .,00000 0
o 0 .,
SLUG
b: <: .. : . :6l
ANNULAR / ANNULAR-MIST J ANNULAR
S=7~
WAVY ANNULAR
DISPERSED
BUBBLE
DISPERSED BU8SLE
Fig. 1. Flow patterns in horizontal flow.

2.1.2 Intermittent
In the intermittent pattern the inventory of liquid in the pipe

is distributed nonuniformly in the axial direction. Plugs or slugs
of liquid, which fill the pipe, are separated by gas zones that
overlay a stratified liquid layer flowing along the bot tom of the
pipe. The liquid may be aerated with small bubbles that are
concentrated toward the front of the liquid slug and the top of the
pipe. The intermittent pattern is sometimes subdivided into slug
and elongated bubble patterns, but the distinction between them
has not been clearly defined in the past. The elongated bubb1e
pattern should be considered the limiting case of intermittent flow
when the liquid slug is free of entrained gas bubbles. The flow at
any cross-section is highly intermittent. As the liquid slug
passes, there exists a high liquid rate and a low gas rate. When
the gas bubble passes, the flow rate of gas is high while that of
liquid is small.
2.1.3 Annular
In annular flow the liquid exists as a continuous film around

the perimeter of the pipe and is also continuous axially, thus
forming an annulus. This liquid film surrounds a core of high-
velocity gas that usually contains entrained drops. The gas-liquid
interface is highly agitated with drop lets being formed from the
waves and entering the core and deposited froIn that carried by the
gas. The liquid film is usually thicker and the wave structure
more pronounced at the bottom than at the top of the pipe. The
condition where the film thickness at the top is fairly steady with
time is known as the annular or annular-mist pattern. Under certain
flow rates near transition to annular flow, large aerated waves
moving along the bot tom of the pipe are high enough to wet the top
surface. Then the film thickness at the top increases substantially
as a wave passes. The surplus liquid drains rather than moving
axially. This is designated as the wavy-annular flow pattern.
2.1.4 Dispersed Bubble
In dispersed bubble flow the gas phase is distributed as

discrete bubbles in an axially continuous liquid phase. The
concentration of bubbles is higher near the top of the pipe, but as
the liquid rate increases the bubbles are dispersed more uniformly.
2.2 Vertical Pipes: Fig. 2
2.2.1 Bubble Flow
In bubble flow the gas phase is distributed more or less

uniformly in the form of discrete bubbles in a continuous liquid
phase.
2.2.2 Slug Flow
In the slug flow pattern the two fluids redistribute axially

so that at any cross-section the flow rates of liquid and gas vary
with time. The gas flows 1argely in a "Taylor bubble", which
occupies most of the pipe's cross-sectional area and can vary in
1ength from the tube diameter to over a hundred diameters. Between
the Taylor bubble and the wall, a thin liquid film flows downward.
6 A.E •. Dukler and Y. Taitel
Churn
Fig. 2. Flow patterns in vertical flow. From Taitel et al. (1980).
Reprinted with permission.
Successive Taylor bubbles are separated in the axial direction by

liquid slugs that bridge the pipe and carry distributed liquid
bubbles.
2.2.3 Churn Flow
Churn flow is somewhat similar to slug flow. It is, however,

much more chaotic, frothy, and disordered. The bullet-shaped
Taylor bubble becomes narrow, and its shape is distorted. The
continuity of the liquid in the slug between successive Taylor
bubbles is repeatedly destroyed by a high local gas concentration
in the slug. As this happens and the liquid falls this liquid
accumulates, forms a bridge, and is aga in lifted by the gas.
Typical of churn flow is the oscillatory or alternating direction
of motion of a liquid.
2.2.4 Annular Flow
Annular flow is characterized by continuity in the axial

direction of the gas phase in the core. Liquid flows upward, both
as a thin film and as droplets dispersed in the gas. Except at
the highest flow rates, the liquid appears to also flow as large
fast-moving lumps that are intermittent in nature - either large
roll waves traveling over the film or a high local concentration
of droplets.
The flow patterns described above appear to exist when the

pipes are inclined, although the conditions at which they exist
change. For example, for pipes inclined at significant angles to
the vertical, the existence of symmetrical Taylor bubbles, as shown
for the slug flow pattern in Fig. 2, ehanges to the appearanee of

slug flow, as shown in Fig. 1 for horizontal pipes with the large
vapor bubble eontaeting the top wall of the pipe. However, the
patterns deseribed here still serve as suitable eharaeterization
for inelined systems.
There have been few systematie studies for noneireular

geometries. Venkataswararoa et al (1982) reported on flow patterns
for a vertical rod bundle consisting of 24 rods in a 20.32 cm diam.
shroud. They showed that the flow patterns observed in vertical
pipe flow were also observed for a rod bundle. However, two types
of slug flow existed. In one the Taylor bubbles oecupied almost
the entire cross-sectional area of the shroud, with the rods
penetrating the nose of the bubble. Liquid flowed down around the
Taylor bubble on the shroud. In the other a Taylor bubble existed
in the free space created by a four-rod cell, and liquid flowed
down along the rods. It is to be expected that with unusual
geometries there may be incentives for defining other flow
patterns not described here.
3 FLOW-PATTERN DETECTION
The problem of flow-pattern deteetion, or determining whieh

pattern actually exists, is complicated by two facts: (a) certain
flow-pattern deseriptions, as discussed above, contain arbitrary
elements and (bl transitions between some flow-pattern pairs is a
gradual process, and there is diffieulty in defining the boundary.
For example, while it is possible to define the difference between
well-developed slug flow and well-developed annular flow in a
horizontal tube, there is a range of flow conditions where it is
difficult to distinguish between the two. Thus, by the very nature
of the problem, there is little point in attempting to make fine
distinetions. With this reservation, however, it is possible to
describe a variety of methods that have proven useful. Most are
based in one way or another on observing the time, amplitude, and
positional variation of voids.
3.1 Visual Methods
The simplest method for detecting flow patterns is by visual

observation through transparent test-section walls. At flow rates
where the velocities are low enough to rnake observation possible,
and when dealing with transparent fluids, this is as desirable a
rnethod as any instrumental method of analysis. In fact, the eye
simply detects the variation of voids. This is most successfully
done by focusing on an element of liquid or gas and allowing the
line of sight to move with the velocity of the element and then
seanning the length of the pipe. Repeated Lagragian seans of this
type can reveal the structure.
At high flow rates photographic methods are useful, as

demonstrated by Raissan (1965), Hsu and Graham (1963), Bergles and
Suo (1966), and Hewitt and Roberts (1969). Most photographic
methods are limited by the size of the field of view, usually no
more than several diameters long, so that only instantaneous local
behavior can be observed. This restricts the observation of the
axial variation of the voids, which is important to the
characterization. Hhere the fluids are not transparent, X-ray
8 A.E. Duk1er and Y. Taite1
photography has been used by Derbyshire et al. (1969) and Hewitt and
Roberts (1969).
Moving-picture photography has proved to be less successful

because the field of view is smaller th~~ that necessary to get the
resolution needed for observation.
3.2 Methods Based on Pressure Measurement
Early attempts to use pressure to characterize flow pattern

were based on the observation that, as the flow pattern changed
with a systematic change in gas or liquid rate, the slope of the
time-averaged pressure gradient curve changed as weIl [Govier et al.
(1957), Isbin et al. (1959), Chaudry et al. (1965)J. While the
result was descriptive, it was not particu1arly useful for detection
since these slope changes could be related to flow pattern
transitions only through visual observations.
Hubbard and Dukler (1966) suggested a method based on the

spectral analysis of wall pressure fluctuations. The approach is
based on the idea that the fluctuations in wall pressure were
reflections of the manner in which the liquid and gas were
distributed in the pipe and of their velocities, and this was
precisely what characterized a flow pattern as weIl.
The work was done in a horizontal tube with the time trace of
static pressure at the wall analyzed to obtain the power spectral
density of the wall pressure fluctuation Spl (fl. The terms are
defined below.
pi (tl = p (t) - P (1 )
where P(t) is the measured time-dependent pressure, P the time

average, and p' (tl the time variation of the fluctuation. If the
autocorrelation function Rp ' (T) is defined as
R '(T) =
Um 1:.r; )0
(tP'(t)P'(t + Tldt (2)
P t-+co
then the power spectral density sp' (tl is the Fourier transform of
the autocorrelation
00
f (3)
computer program packages now exist whereby these spectra can be

obtained efficiently using fast Fourier transforms.
Measurements of this type showed that, despite a wide range of

apparent phase distributions, only three basic spectra existed, as
shown in Fig. 3, and these could serve as fingerprints for the
flow pattern. Type A, where the spectra peaked at f = 0 with decay
at increasing f, was characteristic of separated flow patterns,
such as stratified or annular flow with low entrainment rates.
Because each phase was continuous axially, the spectra were
similar to that of turbulent flow. Intermittent patterns, such
as those for slug or elongated bubble, displayed spectra similar
to B in Fig. 3. Here the peak spectra were displaced from zero
©~
O~-L~--------~=====-~--------~~
f
Fig. 3. Power spectral density of wall pressure fluctuations.
and represented the mean frequency of slug passage. Dispersed or

distributed flow patterns, such as bubbles in liquid and drops in
gas, revealed spectra sirnilar to band-limited white noise as shown
by curve C. Here the fluctuations were controlled by the passage
of successive elements of the dispersion. Because the time between
successive fluctuations is small, the correlation coefficient as
defined in Eq. (2) can be expected to approach a delta function.
The Fourier transform of the delta function is white noise, and a
spectra of type C can be expected.
In practical situations, what one usually observed is a

superposition of spectra. For example, the spectra displayed for
annular dispersed flow is shown in Fig. 4a, where the separated
nature of annular flow appears at low frequencies, and the presence
of the dispersion appears as a white-noise-type tail at higher
frequencies. For wavy annular flow, the spectra appear as a
superposition of those for annular and slug flow, as in Fig. 4b.
The original method did not discriminate clearly, however,

between the two types of separated flowSi annular and stratified,
or between the two types of dispersed flow; bubbly or mist.
Weisman et al. (1979) attempted to extend this method by introducing
amplitude analysis into the characterization. They suggested
specific number criteria on amplitude and frequency to characterize
all the patterns, but it seems likely that specific numbers depend
on the specific system being used in the experiment and are not
general.
For vertical systems, a method proposed by Tutu (1982) seems

particularly useful. Two pressure transducers are located axially
apart on the wall of the pipe, and time variation of the pressure
gradient is analyzed for probability density distribution. In
vertical flows the pressure gradient is strongly dominated by the
A+C=ANNULAR MIST
(al
",
- - - - - "-,..,- - ;;:_:-_==-=-=-------
' ....
O-T--------------~~--------~
---
f
A+B=WAVY ANNULAR
(b)
O~---==---~--------~----~
f
Fig. 4. Superposed spectra.
hydrostatic gradients or the void fractions, that is, for bubbly

flow, the pressure gradient will fluctuate about PL(l - a)g, while
for annular flow it will center about P~g. If the two fluid
densities are widely different, as theyUoften are, this can be used
as a diagnostic tool. For example, pressure drop traces are
sketched in Fig. 5 having the characteristics to be expected for the
bubble, slug, and annular flow patterns. Oscillations in pressure
gradient during bubbly flow reflect the variation in voids with
time between the two measuring stations, and this results from a
few bubbles entering or leaving this volume. Similarly, in annular
flow the fluctuations in Ap result from variations in the
concentration of drop lets in the core, again a small number.
However, in slug flow the pressure gradient will oscillate between
a maximum of that for bubble flow, when the liquid slug occupies the
space between the detectors and near zero when the Taylor bubble
passes. The corresponding probability density functions (PDF) for
each trace appear in Fig. 5. The quantity S represents the
distance between pressure transducers. One can quite clearly
distinguish between these three patterns by either the location of
the PDF or the existence of abimodal peak. Tutu pointed out that
for churn flow the PDFs display the character of slug flow, but the
peaks are broader and gradually shift from liquid to gas dominance.
Of course, this technique simply represents another way of
characterizing the spacial and time distributions of voids.
1.0
AP
PDF
P[.gs
o~
BUBBLY
______________ ~
\
TIME 0 .1P 1.0
PL9S
I
1.0
.1P
PDF
P[.gs ANNULAR
0'--_ _ _ _ _ _ _ _--'
TIME 0
\ .1P 1.0
PL9S
PDF
o~ ______________
SLUG ~
TIME
Fig. 5. Probability density function of pressure difference.
3.3 Methods Based on Photon Attenuation
While the pressure methods discussed above can be used to

imply the void fraction distribution, a more direct measurement
of void can be accomplished by the use of X-ray or gamma-ray
attenuation methods. The most complete approach of this type is
described by Jones and Zuber (1975), who made measurement of the
attenuation of a single beam from a continuous X-ray source in a
vertica1 configuration. The time-varying signal was subjected to
probability density analysis, and characteristic POF curves were
shown to exist for the bubbly, annular, and slug flow patterns.
For horizontal systems, the method could be applied by positioning
the beam vertically. Then slug flow would be characterized by a
bimodal density distribution, annular-dispersed flow by a broad
single peak with its maximum at low voids, and the stratified
pattern by a narrow peak at intermediate voids. The existence of
smooth or wavy stratified flow can be distinguished by the degree
of spread of the PDF around its peak value. The limitation of the
method is based largely on its high cost and the need for careful
installation and operation to ensure safety of personnel.
3.4 Methods Based on Electrical Conductivity
For electrically conducting liquids, such as water, conductance

measurements can be used to discern flow patterns. Early
applications of the method were by Solomon (1962), Haberstrah and
Griffith (1965), Raisson (1965), Bergles and Suo (1966), and
Fiori and Bergles (1966). These early applications used needle
probes and were, in many instances, not specific for all the flow
patterns. Barnea et al. (1980) have now suggested a multiple-probe
modification that appears to be able to discriminate in a more
diagnostic way between the various patterns.
The basic probe design for application to horizontal and

inclined tubes where asymmetries are expected to exist is shown in
Fig. 6 and consists of five electrical contacts or probes. (For
vertical tubes, one less probe is needed.) Probe A is mounted
flush with the wall and detects the presence of a film as in annular
flow by causing current flow between A and the flush-mounted
electrode E at the bottom of the pipe. Such conduction results in
a voltage reading across resistor RA of EA(t). Probe B is 0.25 mm
in diameter and is Teflon-coated except for the tip. This tip is
located 3 mm from the upper wall and detects the presence of bubbles
E o
LJ J
I o/C POWER SOURCE
-=-
Ee(t) Re
FROM FROM FROM

ELECTROOE A ELECTROOE B ELECTRODE C
Fig. 6. Electrode configuration for conductance method.

GaS-Liquid Flow Pattern Transitions 13
by interpretation of the voltage reading across resistor RB' Probe

C is an uninsulated needle projecting through the flow field from
the top and extending to within 3 rnrn from the flush bottom electrode,
D. This detects the condition for stratified flow.
Stratified flow is characterized by zero outputs from probes

A and Band nonzero reading from probe C. For smooth stratified
flow the signal is steady. With wavy stratified flow it fluctuates
with time, responding to the changing liquid level. This condition
appears in Fig. 7a. When slug flow takes place, the signals from
probes A and B show an alternating character, as seen in Fig. 7b and
7c. If the slugs contain no gas bubbIes, as in elongated bubbly
flow, the signals are very steady at two levels (Fig., 7b); if the
liquid slugs contain vapor bubbles, this too can be detected (Fig.
7c). During annular flow, probe B will be without a signal, while
probe A displays a fluctuating signal (Fig. 7d). For the wavy
annular flow pattern, probe B is occasionally wetted, and this
results in outputs as displayed in Fig. 7e, while for dispersed
bubbly flow the traces appear as in Fig. 7f.
For vertical flow it is suggested that the exposed tip of

electrode B be placed at the centerline. Then the outputs of
electrodes A and B can be used to effectively discriminate between
bubbly, slug, churn, and the annular flow pattern.
~~---------------------,
E MAX B
B
0
A
EMAX
E C (d)' ANNULAR
n
(al' STRATIFIED
0 E MAX
E MAX B
S
0
E 0 E
EMAX
E MAX
(b) ELONGATED BUBSLE

E MAX
E 0
0
EMAX E
EMAX
2
TIME (sec)
Fig. 7. Time traces from conductance e1ectrodes. From Barnea et

al. (1980). Reprinted with permission.
Hsu et al. (1963) attempted to apply hot film annemometry to

measure velocities in two-phase flow. While serious questions
exist about the suitability of the method for velocity measurement,
the technique is, in fact, a very sensitive indication of the
presence of the vapor or liquid phase. By analyzing the nature of
the signal rather than its absolute magnitude and depending on the
location of the detector, it is possible to draw some conclusions
as to flow pattern. In fact, with the positioning of several
detectors in locations similar to those of Barnea et al. (1980) for
conductance probes, it should be possible to use hot film
annemometry to detect flow patterns for nonconducting liquids. It
is, of course, more expensive to use than the conductance method.
4 MODELING FLOW-PATTERN TRANSITIONS:

SOME GENERAL IDEAS
The typical approach to predicting flow-pattern transitions

has been to make experiments over a range of flow rates, fluid
properties, and pipe sizes and directions and to visually observe
the flow pattern through a transparent test section window. A
search is then undertaken for a way to map the data in a two-
dimensional plot by locating transition boundaries between the
regimes. This requires adecision to be made about the coordinates
that are used. Because no theoretical basis for selection of
coordinates has existed in the past, this approach represents a
coordination of the data rather than a predictive correlation and
depends strongly on the particular data being used to prepare the
map. For this reason extension to other conditions of pipe size
or inclination, fluid properties, and flow rates is of uncertain
reliabi li ty •
A wide variety of coordinate systems has been used for this

mapping, including both dimensional and dimensionless groups with
their selection based largely on the intuition of the particular
investigators. The earliest, and perhaps the most durable, of
these maps was that proposed for horizontal pipes by Baker (1954)
using a coordinate system that was dimensional in one coordinate and
dimensionless in the second. Figure 8 is Baker's map, where WL and
We are mass flow rates in the dimension of lb mass/h, Pe and PL are
the densities in lb/ft 3 , ~T is the liquid viscosity in centipoise,
and a is the interfacial tension in dyn/ern. Although this was a
most creative attack for its time, the use of dimensional coordinates
clearly limits the result to conditions near those of the
experiments used to locate the transition boundaries. Maps based on
dimensionless coordinates have been suggested based on the idea that
when plotted in this way, the experimentally based transitions
might be valid at other conditions of pipe size and fluid properties.
An early example of this approach was that of Griffith and

Wallis (1961) for vert~cal flow. They used a dimensionless Froude
nurnber (ULS + UCS)/(gD) as the ordinate and the ratio of superficial
velocities ULS/UeS as the abscissa. The same coordinates were
later applied by Spedding and Nguyen (1980).
Of course, the central problem in the use of dimensionless

coordinates is the selection~ from the many such groups available.
One way to arrive at the possible dimensionless groups is through
dimensional analysis. Consider the following variables as having
10 5
"-
'- DISPERSED
I -
-...
C\J
,
' +-
I
- ----
BUBBLE OR FROTH
.c
......
:e
-<
10+ II
ANNULAR
<!
......
~ SLUG
STRATIFIED
2
\ 0' PG )(~ \ ; tJ! J3 rfLl 62.4J3
1\ \0.0.75 624' er l - PL PLUG
p in Ib/eu. ft;fL in Cp ; er in dynes/em.
Fig. 8. Baker flow-pattern map.
an influence on the process of transition: ULS' UGS, D, PL' Pe. ~Ll

~G' g, G, as weil as pipe roughness E and the tube inclination
angle S. Then the application of the Buckingham Pi theorem
indicates that the following dimensionless groups can enter:
2 3
ULS + Ues TI gP L D Pe u crPLD
~LS s 'C;
B
19D UGS ~L2 D PL ~L ? J..,
Sets of alternate groups can, of course, be developed, but it

is clear that any flow pattern map using only two of these groups
can never represent the variables entering into the transition
process. This method is based on a judicious guess of the
"important" dimensionless group, the influence of the other being
considered unimportant. However, there is no certainty that any
single set of two groups will characterize all of the transitions.
It is thus possible to understand the lack of success in the past
in correlating such flows.
In the remainder of this chapter we present an approach to

physical modeling of these transitions. The models are necessarilv
simple, but based on experiments in our laboratories, they are -
reasonably representative of the true situation. The situation is
far from completely understood, and the process of constructing a
fully general predictive method is still in progress. If such a
fully general model existed, it would be possible to specify flow
rates, pipe size or conduit geometry, inclination angle, flow
direction (up or down), steady or transient, with or without mass
transfer (boiling or condensing) and to predict the pattern. While

this is the ultimate goal, it is not yet possible. In what follows
we present aseries of self-consistent models for the prediction
of the flow pattern transitions for the following particular
situations: Sec. 5: Horizontal and Slightly Inclined Pipes: Steady
Motion without Mass Transfer between Phases; Sec. 6: Effects of
Flow Transients in Horizontal Pipes; Sec. 7: Effects of Boiling and
Condensation in Horizontal Pipes; Sec. 8: Vertical Pipes: Steady
Motion without Mass Transfer; Upflow and Downflowi Sec. 9: Effects
of Pipe Inclination; Sec. 10: Transition Mechanisms in Tubes: &~
Overviewi Sec. 11: Vertical Rod Bundles. Simple and logical
extensions can oe made for certain situations not described in these
sections. For example, the effects of boiling on flow in inclined
pipes are straightforwardi however, the analysis of transient
effects in vertical pipes is not as obvious.
5 HORIZONTAL fu,D SLIGHTLY INCLINED PIPES: STEADY

MOTION WITHOUT ~ffiSS TRANSFER BETWEEN pa~SES
In this section mechanistic models are developed for transition

between the following flow patterns: stratified-smooth (SS);
stratified-wavy (SW)i intermittent (I), which includes slug and
elongated bubblei and annular or annular-mist (A). These patterns
are shown in Fig. I, and the analysis follows closely that
presented originally by Taitel and Dukler (1976).
5.1 Equilibrium Stratified Flow
The process of analyzing the transitions between flow regimes

starts from the condition of stratified flow. The approach is to
visualize a stratified liquid and then to determine the mechanism
by which a change from stratified flow can be expected to take
place as well as the flow pattern that can be expected to result
from the change. In many cases stratified flow is seen to actually
exist in the entry region of the pipe. However, the fact that
stratified flow may not actually exist is not important, since it
is weil established that the existence of a specific flow pattern
at specified gas and liquid rates is independent of the path used
to arrive at that state. Since the condition of stratified flow
is central to this analysis, the initial step is the development
of a generalized relationship for stratified flows.
Consider smooth, equilibri~m stratified flow as shown in Fig.

9. A momentum balance on each phase yields
Fig. 9. Equilibrium stratified fluid. From Taitel and Dukler

(1976). Reprinted with permission.
A
L
rLdxJ
dpl - 'LSL + , 1.-.S.1.- + PLALg sin i3 0 (4)
- A
G
[dPl
ax_ - LeSe - , .S.
1.- 1.-
+ PeAeg sin S 0 (5 )
Equating the pressure drop in the two phases and assurning that at
transition conditions the hydraulic gradient in the liquid is
negligible gives the following results:
S Sr
e o
TC Ar:! - 'L
'-'
+ (6 )
er AL
The shear stresses are evaluated in a conventional manner with U

being the average phase velocity
Pe(U e - Ui )2
,.1.- .co
J •
1.- 2
(7)
with the liquid and gas friction factors evaluated from
f'
•L
f'
J G (8 )
where DL and De are the hydraulic diameter evaluated in the manner

as suggested by Agrawal et al. (1973)
(9)
This implies that the wall resistance of the liquid is similar to

that for open-channel flow and the wall resistance of the gas to
that of closed-duct flow. It has been established that, for smooth
stratified flow, fi=fe (Gazley, 1949). Even though many of the
transitions considered here take place in stratified flow with a
wavy interface, the error incurred by making this assurnption is
small. At flow rate conditions where transitions are observed to
take place, Ue »Ui. Thus, the gas-side interfacial shear stress
1s evaluated with the same equation as 1s the gas wall shear. In
this work the following coefficients were utilized: Ce CL 0.046, = =
n = m = 0.2 for the turbulent flow and Ce = CL = 16, n = m = 1.0
for laminar flow.
It is useful to trans form these equations to dimensionless

form. The reference variables are D for length D2 for area, and
the superficial velocities ULS and Ues for the liquid and gas
velocities, respectively. By designating the dimensionless
quantities by a tilde (~,), Eq. (6), substituting Eqs. (7) .and (8),
takes the form
s ] - r(",
A:: '" )-m
~ UGD G
- 4 Y o
(10)
where
I (dPldx)LS'
(11)
I (dP/dx) GS I
(;J L - PG) g 5 in ß
y (12)
I (dPldx) GS
The pressure drop of one phase flowing alone in the pipe is

designated by [(dPldx)s]. Thus, X is recognized as the parameter
introduced by Lockhart and Martinelli (1949), and it can be
calculated unambiguously with the knowledge of flow rates, fluid
properties, and tube diameter. Y is zero for horizontal tubes and
represents the relative forces acting on the liquid in the flow
direction due to gravity and pressure drop. It too can be
calculated dire~tly. All dimensionless variables with the symbol
depend only on n
= hlD, as can be seen from
I -
0.25 { rr - cos- (2h - 1) + (2h 1) [1 - (2 h- 1) 2 1\} (13)
1-
0.25 { cos- (2h - 1) - (2h - 1) [1 - (2h _ 1)2] \} (14)
rr - cos- 1 (2h - 1) (15)
- cos- 1 (2h - 1) (16 )
(17)
A
(18)
(19)
Thus, the quantities included in the two square brackets of

Eg. (10) depend only on the value of kiD. It follows that for
each value of hlD one can then solve Eq. (10) for the X-Y pairs
that satisfy the equation. The result of this calculation for
turbulent flow of both phases appears in Fig. 10. The case of
turbulent liquid flow with laminar gas can occur in practice for
transitions at low gas rates. The solutions can be easily developed
1.0
0.9
0.8
0.7
h 0.6
D
0.5
0.4
0.3
0.2
0.1
0
10.3 10-2 10" 10° 10'
X
Fig. 10. Equilibrium liquid level.
by setting n = 0.2, m = 1, CL = 0.046, and Ce = 16. The result is

remarkably elose to the turbulent ease shown in Fig. 10. The
deeision whether laminar or turbulent flow exists in eaeh phase
should be based on the Reynolds number ealculated by using the
aetual veloeity and hydraulic diameter of that phase, not the
superficial veloeity and diameter.
5.2 Transition from Stratified Flow
Extensive experimental and analytical studies (Dukler and

Hubbard, 1975) have shown that, for the range of flow eonditions
over which intermittent flow is observed, the flow at the inlet
of the pipe is, at first, stratified. As the liquid rate is
increased, the liquid level rises and a wave is formed that grows
rapidly, tending to block the flow. At lower gas rates the blockage
forms a competent bridge, and slug or plug flow ensues. At higher
gas rates there is insufficient liquid flowing to maintain or, in
some cases, even to form, the liquid bridge, and the liquid in the
wave is swept up and around the pipe to form an annulus usually
with some entrainment. Butterworth (1972) demonstrated this
mechanism for annular film formation. Thus, this transition can
be defined as that from stratified flow to either intermittent or
annular flow. It takes place when the conditions are such that a
finite-amplitude wave on the stratified surfaee will grow. This
transition can be expeeted to be sharply defined as observed in
practice.
Consider stratified flow with a wave existing on the surface

over which gas flows. As the gas accelerates, the oressure in the
gas phase over the wave decreases owing to the Bernoulli effect,
and this tends to cause the wave to grow. The force of the gravity
acting on thewave tends to make it decay. The Kelvin-Helmholtz
theory (Milne-Thomson, 1960) provides stability criteria for waves
of infinitesimal amplitude formed on a flat sheet of liquid flowing
between horizontal parallel plates. According to this theory, waves

will grow when
- PG)hGl~
uG
J (20)
>
Pe
where hr; is the distance between the upper plate and the equilibrium
liquid level.
This type of stability analysis is now extended in a rather

elementary manner to the ca se of a finite wave. Consider a finite
solitary wave on a flat horizontal surface, as shown in Fig. 11,
having a peak height h' and a gas gap dimension h'G. The
equilibrium values are hand hG. If the motion of the wave is
neglected, the condition for wave growth is
(21 )
with
p _ pt (22)
The criterion for instability then becomes
- PC)g cos S (h l - h)
u," > (23 )
L
CI
Ar; andA~ represent the flow area for gas over the undisturbed
film andUthe wave, respectively. For flow between horizontal
infinite parallel plates, this equation reduces to
(24)
where Cl depends on the size of the wave
(25)
- h
- !
Fig. 11. Instability of a solitary wave. Frorn Taitel and Dukler

For infinitesimal disturbance, he/h ' e + 1.0, Cl + 1.0, and Eq. (24)
reduces to Eq. (20). However, a comparison of these two equations
shows that finite disturbances are less stable than infinitesimal
ones, since for a finite disturbance Cl is less than unity. Wallis
and Dobson (1973) arrived at Eq. (24) with Cl = 0.5 from observation
of experimental data.
For flow in a round pipe, the desired result can be obtained

for small, though finite, disturbances by expanding Ae
using a Taylor series to yield
(26 )
r< 2 (27)
'-'2
C2 is unity, as is Cl for infinitesimally small disturbances.

However, it is now necessary to estimate C2 for disturbances of
finite amplitude. When the equilibrium liquid level approaches
the top of the pipe, Ae is small and the appearance of a wave even
of small amplitude tends to cause Ae'/Ae and C2 to approach zero.
Conversely, for low levels the appearance of a small finite-
amplitude wave will have little effect on the air gap size r and
C2 approaches 1.0. For this reason we speculate that C2 can be
estimated as folIows:
(28)
Note that for h/D = 0.5, C2 equals 0.5, which is consistent

with the result of Wallis and Dobson (1973). Kordyban and Ranov
(1970) analyzed the transition from stratified to slug flow for
water and air between horizontal parallel plates. Their data for
the air velocity needed to effect transition as a function of the
channel air and water gaps give reasonable agreement with Eq. (24),
using C2 given by Eg. (28). Thus, it is suggested that Eqs. (26)
and (28) describe the conditions for the transition in pipes from
stratified (3) to the intermittent (I) and/or to the annular
dispersed liquid (AD) flow pattern.
In dimensionless form the criterion [Eg. (26)J becomes
(29 )
where P is a Froude number modified by the density ratio
1 (30)
{Dg COS ß)'2
Note that all terms in the parentheses of Eq. (29) are functions
only of h/D. Thus" this transition cri teria can be represented by
a two-dimensional map of the values of h/D and F that satisfy Eq.
(29). The curve A of Fig. 12 represents the locus of such points.
Below this curve, stratified flow exists. Since h/D is a unique
function of X and Y (Fig. 10) 1 the transition is uniquely determined
by X, Y, and F. Once the inclination angle is set, the transition
curve can be mapped in F versus X coordinates. This is shown in
Fig. 13 as curve A for a horizontal tube.
5.3 Transition between Intermittent and Annular
Equation (29) presents the criteria under which finite waves

that appear on the stratified liquid can be expected to grow. Two
events can take place when such growth is observed. A stable slug
can form when the supply of liquid is large enough to provide the
liquid needed to maintain such a slug. When the level 1s inadequate,
the wave is swept up around the wall, as described by Butterworth
(1972) 1 and annular or annular mist flow takes place. This suggests
that whether intermittent or annular flow will develop depends
uniquely on the liquid level in the stratified equilibrium flow.
It is suggested that intermittent flow will develop when h/D ~ O.50q"
DISPERSED
BUBSLE
j
D~
INTERMITTENT 'llO-
I ?
I
!IO I
-------c K
STRATIFIED SMOOTH
10 - 3 '-'---'---'---'--''----'---L--.L_L-'"----.l 100
o ,I ,2 3 4 5 6 7 B 9 ~
h/D
CURVE A B 8 C o
COORDINATE F vs X K vs X T vs X
)2
X = [(dP/dXiLSj
(dP/ dx )GS
Fig. 12. Generalized flow-pattern map.

10'
---
ANNUlAR ~
.V:~ DISPERSED
ANN / MIST BUB8LE
--A~B/_D
A
/ \B D______
.......D,
STRATIFIED
WAVY
A INTERMITTENT
K
/----C-__ \A
/.C- STRATIFIED --" "\
~ SMOOTH C,,~
-2 -I o I 2 3
10 10 10 10 !O 10
x
Fig. 13. General flow-pattern map for horizontal tubes. See Fig.
12 for definitions of F, K, T and X.
where ~ is the volume fraction of liquid in the slugs that are

formed.
This choice can be explained as follows. Consider the

situation where elongated bubbles are formed and ~ = 1.0. Then the
criterion is that hjD '" 0.5. h'hen a finite-amplitude wave begins
to grow as a result of the suction over the crest of the wave,
liquid must be supplied from the fluid in the film adjacent to the
wave, and adepression or trough forms there. Picture the wave as
a sinusoid. If the level is above the centerline, the peak of the
wave will re ach the top before the trough reaches the bottom of the
pipe, and then blockage of the gas passage and slugging result.
When the liquid is below the centerline, the inverse will be true,
which will make slugging impossible. When ~ is less than 1.0, less
liquid is necessary to form a competent bridge, and the equilibrium
liquid level can be lower. Dukler and Hubbard (1975) have shown
that values of ~ range from 0.7 to 1.0. Thus, the hjD at
transition will vary between 0.35 and 0.50. This band of va lues is
shown by the two vertical lines marked B in Figs. 12 and 13. Indeed,
visual observations show this transition between the intermittent
and annular pattern to be a gradual one that is not well defined.
In general, higher values of hjD and X are associated with lower
values of F. Note that the location of the B curves define two
possible transitions as one moves across curve A. For low values
of h/D and X as F is increased, the pattern changes from stratified
to annular-dispersed, while for high values the transition is from
stratified to intermittent.
5.4 Transition between Stratified-Smooth and Stratified-Wavy

Patterns
For stable waves to form on a smooth liquid interface, a

source of energy input to the liquid is necessary. For horizontal
tubes this source is, of course, the gas. The transition takes
place when the velocity of the gas is sufficient to cause these
waves to form but is less than that needed to cause the rapid
growth needed for transition to intermittent or annular flow. When

pipes are inclined downward, this source of energy is gravity, and
wavy stratified flow can exist in the absence of gas flow; this is
discussed later.
The phenomenon of wave generation is quite complicated and not

completely understood. It is generally accepted that waves will be
initiated when pressure and shear forces working on a wave can
overcome the viscous dissipation in the waves. However, there is
considerable controversy over the mechanism by which the energy
transfer takes place. A good summary is provided by Stewart (1967)
In the 2resent work we use the ideas introduced by Jeffreys (1925,
1926) I who suggested the following condition for wave generation:
(31)
In this equation, s is a sheltering coefficient for which

Benjamin (1959) indicated values ranging from 0.01 to 0.03. In
the present work the value of s = 0.01 is used. C is the velocity
of propagation of the waves. For most conditions where transition
can be expected to take place, Ue »C. Theories concerning these
waves suggest that the ratio of the wave velocity to the mean of
the film velooity C/U T decreases with increasing Reynolds number of
the liquid, and the data of Fulford (1964), Brock (1970), and Chu
(1973) confirm this. At the high Reynolds numbers associated with
turbulent liquid flow taking place near these transitions, the ratio
approaches 1.0-1.25. For simplicity and because a precise location
of this transition boundary is usually not important, the relation
UL = C is used.
These approximations substituted into Eq. (31) give the

following criterion for this transition:
(32 )
In dimensionless form this can be expressed as
2
(33)
UGi uL 1 5
K "
where K is the product of the modified Froude number and the square
root of the superficial Reynolds number of the liquid:
[-"(-P-L ---p_~-G-;(~,. G.'-;~,. .~_2_c-o-s- : -,1(~ ~ ~ s) (34)
The quantities in the parentheses of Eq. (33) depend only on h/D

[see Eqs. (18) and (19)J. Thus, the locus of points that satisfies
this criterion can be represented as a plot of K verus h/D, and this
is shown as curve C in Fig. 12. Since h/D depends only on X and Y
(Fig. 10), the transition curve can also be mapped on a X versus K

coordinate system for fixed values of Y. Curve C is this
representation on Fig. 13, which is constructed for a horizontal
tube (Y = 0).
5.5 Transition between Intermittent and Dispersed Bubble Patterns
For values of X in Figs. 12 and 13 to the right of boundaries

A and B, waves will tend to bridge the pipe forming a liquid slug
and an adjacent gas bubble. At high liquid rates and low gas rates,
the equilibrium liquid level approaches the top of the pipe, as is
apparent from Fig. 10. With such a fast-running liquid stream the
gas tends to mix with the liquid, and it.is suggested that the
transition to dispersed bubble flow takes place when the turbulent
fluctuations are strong enough to overcome the buoyant forces, which
tends to keep the gas at the top of the pipe.
The force of buoyaney per unit length of the gas region is
(35)
In a manner used by Levich (1962), the force aeting because of

turbulence is estimated to be
F (3 6)
T
where V' is the radial veloeity fluctuation whose root mean square
is estimated to be approximately equal to the friction velocity.
Thus
(37 )
Dispersion of the gas is visualized as taking place when F T ~ FB , or
g ,:os ß
J J.Jr
(1 _ (38 )
In a dimensionless form, Eq. (38) takes the form
(39)
where
(40)
T can be considered as the ratio of turbulent forces to gravity

forces acting on the gas. The terms on the right side of Eq. (39)
depend only on kiD. The locus of T versus kiD points, which
satisfies this equation, is shown as curve D in Fig. 12 and for T
versus X in Fig. 13.
5.6 The Flow Pattern Map: Summary
Figures 12 and 13 display the resulting gener2~i2ed flow

pattern maps. The transitions shown in Fig. 12 are independent
of inclination angle. Given the gas and liquid rates, the tube
size, and inclination angle, the values of X and Y are calculated
and Fig. 10 is used to find hiD. The parameters F, T, and Kare
then determined, and the location of the coordinate pairs (F versus
kiD, T versus kiD, 11. versus hlD) is used to determine the pattern.
For any fixed angle a map such as that shown in Fig. 13 for
horizontal flow can be constructed from the equations. This can
then be used to find the flow pattern with one diagram.
The effect of pipe roughness on these transitions is not

specifically considered in the development. However, subject to
experimental demonstration, it is suggested that if the (dPldx)S
values are calculated by using known roughness parameters, the
transition boundaries of Fig. 13 will continue to apply (Taitel,
1977) •
üf course, it is not necessary to use a flow pattern map at

all. Given any one set of flow conditions (rate, pressure, line
size, and inclinationl, the flow pattern that exists for that
condition can be determined rather simply by using hand calculations
from Eqs. (29), (33), (39) and kiD ~ 0.50 ~, with the help of Fig. 10.
Once the physical properties, tube diameter, and inclination

angle are specified, the only remaining variables are the two
superficial velocities, and the transitions can be mapped in ULS
and UGS coordinates. Figure 14 is such an exa\llple. However, these
are not the basic coordinates. The location of the transition
curves when plotted in ULS - Uas space will vary with line size
and physical properties, while those plotted in the coordinates of
Figs. 12 and 13 will not.
5.6.1 Comparison with Data
Mandhane, Gregory and Aziz (1974) made a careful examination

of flow-pattern data. In the absence of a theoretical framework,
they used a map of ULS versus UGS to coordinate about 1000 data
points in horizontal pipes ranging from 1.3 to 15 cm in diameter.
Most of the data were for line sizes in the 1.3-5-cm-diameter range,
so the location of these empirically drawn boundaries was strongly
influenced by these data. In fact, the fit of the data to these
empirical curves was substantially less satisfactory for line sizes
larger than 5 cm.
A test of the new theory is possible by comparing its

predictions with the location of the boundaries suggested by
IYlandhane et al., which in effect represent data. Figures 14 and 15
are such comparisons. The points represent new data (Barnea et al.,
1982) taken in 2.5- and 5.0-cm-diameter pipes; the solid curves are
the prediction of the theory calculated to ULS - Uas coordinates
Gas-Liquid Flow Pattern Transition 27
10r---------------------~
u
'"
Ul
"-
E
o
o SS
o
o 0
o 0
o 0
0.: 10 100
o STRATIFIED SMOOTH (SS) }

STRATIFIED (S)
• STRATIFIED WAVY (SW)
o ELONGA TED BUBSLE
(ES) } INTERMITTENT (r)
• SLUG (SU
E. ANNULAR, ANN.lDISP (AD) }
(AW) ANNULAR (A)
... WAVY ANNULAR
'" DISPERSED SUBSLE (DB)
- THEORY
Fig. 14. Flow-pattern map for 2.5-cm-di~~eter horizontal tube.

Air-water at atmospheric conditions.
for these line sizes and fluid properties, and the hatched curves
represent the empirically located ~landhane curves that represent
data. Certain of the transitions depend on diameter, and those
theoretical curves are located differently in the two figures.
Since the ~landhane curves represent an average of data for allIine
sizes, they are at the same location on the two graphs. Agreement
between this new theory and data is quite satisfactory.
The triangular region between the intermittent, stratified-wavy,

and annular patterns shows the poorest agreement. It is precisely
in this region that the wavy-annular subpattern exists (see Fig. 1).
In the future it may be useful to define transition criteria for
this pattern.
Weisman et al. (1979) published new data that permit comparison

with the theory and for which fluid properties were varied.
Experiments were carried out with air-water, glycerine-water
solutions, and potassium carbonate-water solutions as weIl as with
Freon 113 liquid and vapor under pressure. Surfactants were used
to explore the effect of surface tension. Test-section diameters
varied from 1.2 to 5.1 cm and lengths from 1.5 to 6.1 m. Liquid
viscosity up to 150 cP, liquid specific gravity up to 1.4, and
~
(J
cu 0
~
.s
~ 0
:l 0
0
0.01 0
0
0001
001 0.1 10 10 100
uGS[m/seC]
- THEORY
Fig. 15. Flow-pattern map for S.l-cm-diameter horizontal tube.

Air-water at atrnospheric conditions. See Fig. 14 for
legend.
vapor specific gravity up to 0.04 were explored. The decision on

the flow pattern that existed was based partlyon visual
observation and on the analyses of wall pressure fluctuations as
discussed in Sec. 3. Arbitrary criteria were used in part in this
process of decision. Weisman et ale used essentially empirical
correlations to fit their data. However, a comparison with the
predictions of the theory presented here is in order.
Stratified to Intermittent: Generally good agreement is

observed between theory and the data except for the higher
viscosity liquids. The Weisman et ale data suggest that this
transition is independent of viscosity. The theory presented here
suggests a significant effect. In order to explore this apparent
discrepancy, new data were collected in our laboratories in a
horizontal 3.81-cm-diameter tube using glycerine-water having a
viscosity of 90 and 165 cP. The experimental results are shown as
the data points in Fig. 16. The dotted curve marked A is the
theoretical prediction as calculated from this model for 1.0 cP
and the solid curve is that for 165 cP. The data show that this
transition takes place for this high viscosity liquid at values of
ULS' almost two orders of magnitude less than that for water. This
is in contrast to the result of Weisman et al., whi~h showed no
effect of viscosity. Furtherrnore, the theory provides a satisfactory
prediction of the transition. Experiments repeated at a liquid
viscosity of 90 cP lead to the same conclusion.
Perhaps an explanation for the absence of the viscosity effect

in the Weisman et ale data is related to the use of a relatively
short test section (LID < 120) in their experiments. The study
1.0
/
S B
"~
-- .....
0
• • 'to.
-A~- Ci-
--E 0.1 z
Ul
w
(j)
f-
f- • \
0::
<I:
...I
;:)
~
0::
e.e.t ...J
;:)
0.01 w
f-
e.j Z
Z
Z • I <I:
-A-
STRATIFIED
0
.'I
0.001
0.1 1.0 10 100
UGS ' m/s
-THEORY FOR 165 Cp

- - - THEORY FOR 1.0 Cp
Fig. 16. Flow patterns for 165 eP glyeerine-water and air in a

horizontal 3.B-em-diameter pipe. See Fig. 14 for legend.
reported in Fig. 16 used a test seetion with an LID of 360. In

slug flow for short pipes, the liquid ahead of the slug tends to
drain by gravity. This prevents the formation of slugs that would
otherwise be observed, and higher liquid rates are neeessary to
initiate the slugging.
Stratified to Annul.ar: vveisman et al. s data are in reasonable

agreement with the theory presented here.
Intermittent to Annular: The Weisman et al. results suggest

that the slopes of the eurves in Figs. 14 or 15 that separate
intermittent and annular flow are the opposite of that reeornrnended
by this theory. This is in eonfliet not only with the new data
shown here but also with the analysis of over 1000 data points by
Mandhane et al. as weIl as unpublished data by others. The use of
wall pressure fluetuations may be very ambiguous for precisely
defining this boundary because it is a gradual transition.
Intermittent to Dispersed: Agreement of the Weisman et al.

data with the new theory is satisfactory.
Stratified-Smooth to Stratified-Wavy: The Weisman et al. data

are in satisfactory agreement with this transition with the
exception of the data at high viseosity. The reason for this
discrepancy may be related again to the use of a short section where
film drainage can be expected to occur.
Evaluation of this theory for inclined pipes is treated in

Sec. 9 of this chapter.
6 EFFECT OF FLOW TRANSIENTS IN HORIZONTAL

PIPES
The approach to modeling flow pattern transitions during a

flow transient parallels that for steady flow and has been
discussed by Taitel et al. (1978). As shown in Sec. 5, the
criteria for each of the four transitions involve the equilibrium
liquid level. For steady flows this level depends only on the flow
rates, fluid properties, and tube diameter. For unsteady flow it
also depends on time and position from the entry. In the course of
a transient the individual phase velocities vary with time. Thus,
under conditions of transient flow and depending on the nature of
th~ transient, f~~w patte~n t~ansi~i~~~ c~n take .~l~ce at liquid
ana gas rates d~Jleren~ tnan ror equ~&~b~LUm aond~~~on8.
Furthermore, in moving from one pair of flow rates to another,
flow patterns can appear that would not exist if the flow rate
changes along this path were carried out slowly.
6. 1 Analysis
Figure 17 shows the geometry of stratified flow. Liquid and

gas enter the pipe at x = 0, both flowing in the positive x
direction. Under transient flow conidtions the gas-liquid interface
is, in general, not parallel to the x axis; thus, the liquid level
h, the average liquid velocity UL' and the average gas velocity
UG, are all functions of x and t. The liquid and gas
cross-sectional areas AL and AG' and the contact perimeter between
the liquid and the wall SL' the gas and wall SG, and the liquid
gas interface Si are all functions of the local liquid level hand
therefore also depend on time and position.
The momentum and continuity equations for the liquid phase are
given by
3 (ULA L )
+ T.S.
3t 7.- 7.-
(41 )
aA ~r +
a (U TA L )
at
li
dX
o (42)
Since the cross-sectional area of the liquid film depends on the

level of the liquid in the pipe, AL = AL(h), Eqs. (41) and (42)
take the form
,----- -- - - - - --1----..---1
r
o
=:
•I
UL h{x,t)
----------~t~----------------------~~~X·
Fig. 17. Nonequilibrium stratified flow. From Taitel et al.
(43)
o (44 )
where AL =< dA
Next consider the equation of motion for the gas phase. The
pressure gradient in the liquid and the gas is assumed to be equal,
and Eqs. (41) and (42) with suitable subscripts are equally valid
for the flow of the gas. The gas velocity is much greater than that
of the liquid, and since gas flow rate changes are propagated down
the pipe very rapidly compared to the liquid, a quasi-steady state
is assumed with respect to any time interval in which changes in
liquid flow or level are significant:
(45)
(46 )
Substitution of Eqs. (45) and (46) into Eq. (43) yields
P.~(W~)2
er
Lr
PL P GAG
(47 )
o (48)
In these equations A'L = dAL/dh, TL' TG' and Ti can be calculated

from a Fanning-type relationship, as was done for steady flow [see
Eq. (8)J 1 and thus can be expressed in terms of hand velocity.
Equations (47) and (48) are two simultaneous partial differential

equations for h(x,t) and UL(x,t). Note that the right-hand side of
Eq. (47) is a known function of hand UL " For steady equilibrium
conditions, name1y, the case where the liquid level is constant,
the left-hand side of Eq. (47) is zero, and Eq. (47) becomes
identical to Eq. (6), which applies for steady flow. The knowledge
of h(x,t) can now be combined with the use of the mechanisms for
transition, as discussed in Sec. 5.
Equations (47) and (48) are two hyperbolic partial differential

equations provided that
Or(W e )2 A'r
g» P; PCA C Ae1.J
(49)
Unless Eq. (49) is satisfied, a unique solution to these equations

does not exi~t (Taitel and Dukler, 1977). Calculations show that,
even at flow rates closely approaching these for instability of the
stratified flow, the Bernoulli term is, in fact, small compared to
g. But neglecting this term can be justified on a physical basis
as weIl. The right-hand side of Eq. (49) describes a Bernou
force opposing gravity that acts when the gas is accelerated over
the crest of a wavelet and that tends to make the wave grow. To
avoid the mathematical problem, the approach to solving Eqs. (47 )
and (48) is to predict the smooth liquid level with time and
position in the absence of local waves. Then, at each point of
space and time this level is used to determine if a wave will grow
in accordance with the mechanism leading to Eq. (26).
Once the Bernoulli term is neglected, Eqs. (47) and (48) take
the form
au L aU Lah
+ U (50 )
at L 3X + g dX + E 0
ah ah + aU L
3t
+ Ur
1J ax H
3x
0 (5I)
where H H(h) = AL/A'L' and E = E(h, UL) is minus the right-hand

side of Eq. (47). Using standard methods (Stoker, 1957) we can
convert Eqs. (50) and (51) to
r
l ätJ'" l
" rI
I
L
(U L + C) 3x
+ aatJ UL +1% (U L
+ C)
a
ax +
. h + E 0 (52 )
r
-fi- l
e-
II (n -
Ur
1.J
C)
ax + all'
3'"");1 J "-'r H (U L
a +
- C) -3x aat ] h + E 0 (53)
L J
where C = /gH i5 the critical velocity of the liquid.
Equations (52) and (53) were solved using the finite-difference

technique described by Stoker (1957). In this method explicit
forward finite differences are used with respect to time, whereas
the spacial derivative is replaced by either forward or backward
finite differences depending on the direction of the characteristic
lines. Equation (52) is associated with the forward characteristic,
which has a slope dt/dx = 1/(U L + Cl; thus, backward finite
differences are used. Equation (53) is associated with the
characteristics having the slope dt/dx = 1/(UL - Cl. When at a
given x and t, UL > C, or the flow is supercritical, the. direction
of the characteristics associated with Eq. (53) is positive, and a
backward finite spacial differentiation is used. At subcritical
conditions, when UL < C, forward differentiation is used.
In the numerical scheme each point is checked to determine

whether the flow is sub- or supercritical, and the appropriate
forward or backward numerical scheme is selected. To ensure
stability, the time increment ßt is limited by ßt<ßx/(Ur, + Cl.
The boundary conditions needed for the solution of Eqs. (52)

and (53) depend on whether the local conditions are sub- or
supercritical. For the supercritical case h(x,O), Ur,(x,O), Ur,(O,t),
and h(O,t) are required as boundary and initial conditions. When
the flow is subcritical, only h(O,t) or Ur,(O,t) can be assigned,
since they are related through Eq. (53) and are associated with the
backward characteristics. In this case, however,the flow rate is
specified, and for the subcritical case, both Ur,(O,t) and h( ,c)
are determined by the flow rate.
The approach to determining the flow rate at which transitions

take place during a feed transient is as foliows:
• The variation of gas and/or liquid feed rate with time is

specified.
• Equations (52) and (53) are solved subject to this path, and
the values of h/D and Ur, are found as functions of x and t.
• The flow pattern transition criteria as developed for steady

flow are applied at each time and position in the course of
the transient, and the earliest time at which the criteria
are satisfied is found.
• The feed rates at that instant in the transient are

identified as the transition flow rates for that transient.
The transition equations used here were presented in Sec. 5 and are:
stratifiedto intermittent or annular, Eq. (26); intermittent to
annular, h/D = 0.35-0.50; stratified smooth to stratified wavy
Eq. (32); and intermittent to annular dispersed, Eq. (38).
In principle, this method assumes that the physical mechanism

underlying eaeh transition is unchanged as a result of a transient
and that the transient makes its influence feit as a result of its
effect on the liquid level. We will show below that for a certain
class of transients the complex process described above can be
reduced to a simple graphical procedure.
6.2 An Examp1e: Liquid Feed Transients
Consider flow of water and air at 25 0 C, 1 atm in a 3.8-em-

diameter tube. The solid lines in Fig. 18 represent the transition
curves of interest to this examp1e for steady-state conditions.
The dotted curve represents the locus of points where the liquid
velocity is at its critical value; Ur, = C = r'gH, H = Ar,/.4'[,. Ar,
and A'r, depend only on Hand D since A'L = dAL/dh. h can be found
from a knowledge of Ur,S and UGS (Fig. 10) given the diameter and
properties. It is thus possible to calculate the locus of points
in the Ur,S - UGS plane that satisfies the Ur, = C condition.
Consider a liquid feed rate that is changed along a constant

gas rate path A-B in Fig. 18. If the change takes place very
slowly, thetransition from the initially stratified condition at
U LS
(rn/sec)
01
I 10 100 500
UGS (rn/sec)
Fig. 18. Flow transients: Air-water in a 3.8-cm-diameter pipe.
A (UGS = 7.8 m/s, ULS = 0.03 m/s) to annular flow will take place
at B, where ULS = 0.1 rw/s. We define a slow process as one where
the liquid level remains uniform with x as the flow rate is
increased. Now consider this same flow rate change from A to B
taking place rapidly, say varying from 0.03 to 0.1 m/s linearly
over 1.0 s. The solid curves of Fig. 19a are the solution of Eqs.
(52) and (53) for h/D as a function of x and t. In this ca se the
path is supercritical, so the level at the entry remains constant
and the liquid accumulates downstream, forming a hump. At t = 0.84
s the maximum observed level is sufficient to satisfy the criteria
for growth of a large wave [Eq. (26)] and transition takes place.
At 0.84 s after start of the transient, the liquid rate is only
0.088 m/s. Thus, transition takes place for this unsteady-state
condition at a liquid rate 12% less than what would be observed
under steady flow conditions. Thus, the tI'ansition is advanced.
The path DEF shows a change in liquid feed rate taking place
at liquid velocities below the critical. If the change takes place
linearly in 1.0 s, the profiles of liquid level would change with
time as shown in Fig. 19b. For subcritical flow the level rises
with time at the inlet, and the level that satisfies the transition
criterion can be evaluated. In this case the transition takes
place at t = 0.6 s, at which point the liquid feed rate ULS has
reached a value of 0.2 m/s. Bad the change been made slowly; the
transition liquid rate, point E, would be ULS = 0.14 rn/so The feed
rate at which transition takes place during the transient is 43%
~igher than that for a slow change in feed rate. Thus the
transition is delayed.
6.3 An Exarnple: Fast Gas Transients
When Equations (52) and (53) are solved for liquid levels as
..J
~ ~
!=~ .7 ______ y'"TRANSITION LEVEL,ULS '"O.!4 m/sec
5E
.""
"',.. h 6
u:~
"- ~
o
2,~
;:: ~ 4 INITIAL EQUILfBRIUM LEVEL ,ULs= 0 03 m/sec
~~
2 X(m) 4 5
Fig. 19. Increasing liquid rate at constant gas rate. Total

transient time = 1.0 s. From Taitel et al. (1978).
the gas flow rate is changed, it becomes evident that the liquid
levels are slow to respond to changes in the gas rate. This is to
be expected since the gas influences the liquid flow through the
interfacial shear, and this requires a relaxation of the liquid
velocity distribution. The presence of liquid viscosity makes this
a slow process. For exarnple, consider the path DAG shown in Fig.
18. For a very slow change in gas rate, the transition from
smooth to wavy-stratified would be expected to take place at 3.4 m/s
and to annular at 16.5 m/s. However, if the gas rate is changed
linearly from D to G in 1.0 s, the first transition takes place at
2.0 m/s and the second at 3.4 m/s! These early transitions are
observed because the liquid level rernains high as the gas rate is
increased since there has been insufficient time to relax its level.
Define a fast gas transient as one in which the liquid level

remains essentially unchanged over the time it takes for the gas
rate to increase enough to cause a transition. It is possible to
predict the course of a fast transient without the need to solve
the controlling set of differential equations. For these cases a
particularly simple and elegant graphical procedure is available
using the equilibrium flow-pattern map. The method can be understood
by reference to Fig. 20. Consider the path A to B, where the gas
rate is increased. If the process is carried out very slowly, the
system will pass through aseries of quasi-equilibriurn states along
A-B, and transitions will be observed at each flow rate where the
line A-B crosses a transition curve. Superimposed on this curve is
aseries of lines of constant equilibrium h/D. That such lines are
straight with a slop of 45 0 on log-log coordinates for the horizontal
tube follows from the definition of X [Eq. (11)J and the fact that
X is constant for each h/D (Fig. 10) when the pipe inclination angle
is zero. For a fast gas transient the initial level remains
unchanged until 'the gas rate has exceeded that necessary for
transition. Thus, when the process takes place in a fast transient.
rather than moving along AA'A"B, of Fig. 20, the path becoraes '
AB'B"E''', along which hlO remains constant until the final gas rate
is reached, and then the level relaxes to its value at B along path
B'''B.
I
(I)
L
!
U LS
./ / '
./
(m/sec)
(A)
10 100 500
UGS (m/sec)
Fig. 20. Fast gas transient. From Taite1 et a1. (1978) .

Curve 2-2 is the locus of points re1ating hlD and UGS at which
transition between the smooth and wavy stratified flow takes p1ace.
For this transient, that transition will take p1ace at a gas rate
corresponding to B' rather than A'. Simi1arly, the curve 1-1
represents the locus of points re1ating hlD and UGS that resu1t in
an unstable stratified 1ayer and a transition either to the
intermittent or annular pattern. Along the path AA'A"B, this
takes place at A", where the transition results in annular flow.
However, for the fast transition it takes p1ace at B", where a sLug
appears! Thus, both the fZow rates at which transition takes pLaae
and the partiauZar fZow patterns observed as the gas rate ahanges
aan be different in the aase of a transient. In going along the
equilibrium path, the pattern observed would be smooth stratified,
wavy stratified, or annular. For the fast transient one would
observe smooth stratified, wavy stratified, slug, or annular. The
gas rates at which these changes take place along the transient
can be founa from the intersection of the path line and the
equilibriurn transition curves. The time? at which these transitions
take place can be deterrnined from a knowledge of the shape of the
flow rate-time curve.
Executing a fast transient in the.reverse direction C to D

will produce different results, as seen in Fig. 20. Here the path
will effectively follow CD ' D"D'IID rather than CC'C"D. In this
example the same flow patterns will be observed during the transient
as during an equilibrium path. However, transitions will take place

at higher gas rates: (UGS)D,>(UGS)C', (UGS)Du>(UGS)Cu.
This method provides a visual display of the true path to be

expected during a fast transient, determines when unexpected flow
patterns will appear during the transient, provides the values of
the gas rates at which each transition will appear, and makes
possible the calculations of the time at which each transition will
take place. It is of interest that the speed of the transient has
no effect on the path or the flow patterns that are observed, as
long as the transient can be characterized as fast.
Experiments have been carried out over a variety of paths,

both increasing and decreasing gas rates, that show excellent
agreement, with the predicted appearance of these spurious patterns
and the transition flow rates (Taitel et al., 1978).
7. EFFECTS OF BOILING AND CONDENSATION IN

HORIZONTAL PIPES
In many ca ses of industrial interest the flow of gas and liquid

is accompanied by mass transfer between the phases due to
condensation or boiling. When these transfer rates are small
compared to the flow rates of the phases, the criteria developed
in Sec. 5 can be expected to be applicable. But when that is not
the case, it is reasonable to expect that the mechanisms controlling
the transition or the conditions controlling the equilibrium liquid
level would vary, and as a result the flow, property, and pipe size
conditions at which transition would take place would change.
There are conflicting reports in the literature on the effect

of mass transfer. Hewitt (1978) stated the situation this way:
"The application of the adiabatic flow pattern diagrams to the case
of condensation should, therefore, be regarded as somewhat
uncertain." Tong (1965) voiced a similar caution. The results
of specific da ta are conflicting.
Flow-pattern transitions were measured for boiling systems by

Davis and David (1961), Hoogendoorn and Buitelaar (1961), Zahn
(1964), Fiori and Bergles (1966), Shah (1975), and Calder (1976).
These investigations reach conflicting conclusions as to the
suitability of using flow-pattern maps prepared for conditions of
no mass transfer for their data. Corresponding studies in
condensation have been reported by Soliman and Azer (1971), Travis
and Rohsenow (1973), Palen et al. (1977), and Breber et al. (1980).
In each of these cases the authors clearly state that flow pattern
maps for no mass transfer are not suitable to predict their results.
In this section an attempt is made to modify the mechanistic

theory presented earlier in the absence of mass transfer for
conditions of boiling and/or condensation. The modifications
presented here are very tentative ones and should be considered
first steps in this process of generalization.
7.1 Ana1ysis
When mass transfer between phases takes place, there are at

least three identifiable differences that can affect the transition
mechanism.
o Condensation or boiling and the associated mass transfer are

accompanied by momentum transfer betvieen the phases, and
thus additional forces are generated.
• The existence of bubbles due to flashing or heat transfer-

induced nucleation results in a decrease in the effective
liquid density. Thus, the equilibriurn liquid level that
depends on this density will also depend on the transfer
rate as weIl as the variables already shown to enter in the
analysis. Since certain transition conditions depend on hlD
and the effective liquid density, the transition states will
likewise depend on the mass transfer parameters.
• As mass transfer proceeds, the individual flow rates vary

along the tube and so will hiD.
Figure 21 pictures equilibriurn stratified flow with mass

transfer, in this case boiling. Flow is in the positive x direction,
and slip between the bubbles and liquid phase is assurned to be
zero. The liquid level is shown decreasing with increasing x. For
condensation the reverse is true. The analysis will proceed in
parallel with that for adiabatic flow developed earlier (Sec. 5).
First, expressions for equilibriurn level are developed. Then
criteria for transition as functions of this level are explored.
Equilibrium Level: Momenturn and continuity equations for the

liquid are given below. The subscript M designates the stratified
gas-liquid mixture.
+ mu
"" M = - TM
S M ~' Ti
Si - A
P rvß "lv1
dh
dx - AM dP
dx
(53 )
d (P/\PrvrM)
dx + m o (54 )
Here m is the mass rate per unit length of pipe designated as

positive when directed from the liquid to the gas phase. Cornbining
these equations gives
Fig. 21. Stratified flow geometry for a boiling system.

T .5. -
1.- 1.-
(55)
A similar procedure for the gas that flows in a continuous layer

above the liquid is
dU v dh
- PGgAv dx dP
+ m,U V - UM) -TVSV - - AV dx
• I
pGUv T .5.
dx 1.- 1.-
(56)
+ P Gg A sin ß
Here the subscript V designates that portion of the gas that flows
in a continuous layer above the liquid.
Let q be the heat flux at the inner pipe wall averaged over
the perimeter designated positive for input to the fluid (boiling)
and negative for condensation. Then with A as the latent heat of
vaporization
m = (57)
This simple treatment neglects energy transfer that might contribute

to sensible heating.
Eliminating the axial pressure gradient between Eqs. (55) and

(56) gives
dUM dU v TMS M TVS V

PZ,PM dx - P GUV dx + (PM
dh
+ PG)g dx +
AM
- AV -
(58)
,S'(jJ + -L) -
1 qrrD
T
1.-1.- "M AV (p 1'4 - PG)g sin ß -
AA * (Uv - UM) 0
A*, which appears in the denominator of the last term, is defined

as A* = AV for boilingj A* = AL for condensation. Now consider
situations where the liquid level varies slowly with Xi that iST
we exclude explosively rapid condensation or boiling processes.
Then derivatives with respect to X can be neglected compared to
the other terms in the equation, and
TVS V
- -~-- sin
h
V
= 0 (59 )
Under these conditions the solution for h, U 1 and UM depends only

on the local flow rates and forces acting on and independent of
the heat transfer history upstream. Now this equation can be
solved in a manner parallel to that done be fore in Sec. 5 with the
shear stresses anel friction factors related as in Eqs. (7) and (8).
In this case the mixture velocity UM is used instead of UL , the

density p~ instead of PLI and the viscosity ~M instead of ~L' As a
first app'roximation, it is suggested that these quantities be
defined as
P,/:t+ P L (I-Cl) (60 )
~GCl + ~L(l - Cl) (61)
Of course, for condensation the void fraction in the liquid is

zero. An estimate of the fraction voids in the bubbling liquid can
be obtained this way: A "superficial" velocity for the gas phase
rising through the liquid as a result of the boiling process is
defined as
(62)
But this velocity must be related to the bubble rise velocity Uo

through
Uo Cl (63)
Uo can be estimated from Eq. (77), discussed in connection with

flow-pattern transitions in vertical tubes.
qTiD
Cl = (64)
At high heat flux the bubbles become closely packed, and continuous
channeling of vapor through the liquid can take place. This
happens at Cl ~ 0.25, as is explained in Sec. 8.1.1. Thus, when Cl
exceeds 0.25, as calculated in Eq. (64), this maximum value is used.
If UM is neglected compared to Uv in the last term of Eq. (59),

the equation can be converted to dimensionless form in a manner
analogous to that done before for adiabatic systems. The same
normalizing variables are used, except for UMS' replacing ULS as
the normalizing velocity for UM' The result is
(65 )
Uv
- 4Y - 4Q -=-;: o
A
X and Y and Q are defined as

(4C M/D) [UMsDPp,/J.lMJ-n (P MU;S/2) I (dP/dx)MSI

(66 )
(4C G/D) [UvsDPGhGJ-m (PGU~S/2) i (dP/dx)VSI
( PM - P G ) g s in S
Y (67)
I (dP/dx)llS!
Q (68 )
AD(dP/dx)vS
It is important to note that UvS is the superficial velocity

of that portion of the gas that flows above the interface. In the
ca se of boiling, additional gas flows as bubbles with the vapor.
UGS is the superficial velocity calculated for all of the gas
moving past a cross section, UMS is the superficial velocity of
stratified liquid plus the vapor bubbles it carries, and ULS is
the superficial velocity of the liquid only. By a mass balance
(69)
(70 )
for condensing systems ULS = UMS' UGS = UvS' PM = PL' and 04* = AL'
The solution of Eq. (65) for h/D is straightforward once
particular values of X, Y, and Q are selected. However, the
calculation of X, Q, and Y requires a knowledge of UMS and UVS1 as
seen in Eqs. (66) and (68). This requires a knowledge of a
[Eqs. (69) and (70)J, and a depends on SilD [Eg. (64)J. This can
be determined from h/D or from the solution of Eq. (65) itself.
Thus, the process is an iterative one but not a complex one. The
simplest approach is to initially assurne SilD = 1.0 in Eq. (64) as
a first estimate. The results for turbulent flow of both phases
(n = m = 0.2) in a horizontal tube (Y = 0) for condensation
(04* = AL) and boiling (04* = AG) are shown in Fig. 22. The effect
of either boiling or condensation is to lower the equilibrium
level.
7.2 Transition Criteria
The transition criteria developed in Sec. 5 can be modified

to account for mass transfer. The transition between stratified
and interrnittent or annular is modified by the use of PM instead
of PL in the definition of the Froude nurnber
U
vS k (71 )
(DgcoSS)2
and the transition curve is defined by

1.0
.9 ----- CONDENSATION
.8 --BOILING
.7
.6 Q1TU vS
lL .5 Q=
0 }"D( dP/dx)VS
.4
.3
.2
.1
O~~~--~~--~~~~--~--~~--~--~
10-3
x
Fig. 22. Equilibrium level for condensation/boiling.
(72)
In fact, with rnass transfer taking place, there exists another

force normal to the interface in addition to the Bernoulli force
and that of gravity. The reactive force in the direction opposite
to that of the transfer is easily calculated, but it is cl~arly
negligible cornpared to the other two under all conceivable realistic
operating conditions.
As explained in Sec. 5, given that the criteria in Eq. (72)

are satisfied and stable stratified flow can no longer exist, the
transition between the interrnittent or annular pattern depends only
on hlD. Of course, the particular value of X at which this hlD
exists now depends on Q and whether the process is one of
condensation or boiling (Fig. 22).
The transition between interrnittent and dispersed bubbly flow

is assumed to be the result of a balance between dispersed forces
due to the turbulent fluctuations in the liquid and the forces of
buoyancy, as was presented in Sec. 5. A cornpletely parallel
developrnent leads to this criterion, which can be cornpared with
Eg. (39).
T["~i(!(')-"f ' 1.0 (73)

Gas-Li~Jid Flow Pattern Transitions 43
where
I (dPjdx)MS
I (PM
T (74)
- DC)g cos
i-
The nature of the transition between the smooth and wavy stratified
patterns is not clear when mass transfer takes place. For boiling
systems it does not seem useful to define a " smooth" stratified
condition, since bubble evolution makes this meaningless. For
condensation additional studies will be necessary to explore the
mechanism.
Generalized coordinate maps prepared for horizontal tubes

are presented in Fig. 23 for boiling and in Fig. 24 for condensation
for three values of the heat transfer parameter Q. The curves for
Q = 0 correspond to the results shown in Fig. 13 for an adiabatic
system and with the B curves located for hjD = 0.5. Of course,
Fig. 13 can once again be used to locate the transition boundaries
in terms of hjD providing that Fand T are defined using Eqs. (71)
and (74).
Now these generalized maps can be used along with Eqs. (64),
(69), and (70) to construct maps in ULS - UCS coordinates for any
specific case of interest. For example, Figs. 25 and 26 show the
100 ~ANNULAR- 01 SPERSEQ
f---r-$;::-<,""- ~
I
~~~D
or
F
161~ STRATIFIED ~<'\ INTERMITTENT
T "\
162~ Q=O
=10
\\ \
= 100
IÖ3L-__~____- L____- L____~__~L-~_ _~_ _~

10- 3 104
x
CURVES AßS o
COORDINATES F VS X T VS X
X =[(dP/dX)MS]
1/2
;p;- UyS
(dP/dx)vs F = J~ jOg cosß
Fig. 23. Generalized flo1tl-pattern map. Boiling in horizontal tubes.

10 1
ANNULAR DISPERSED DISPERSED
10°
B
---
-- ..... ~
" ,
BUBBLE
1I
10- 1
T STRATIFIED \
\
OR \ . INTERMITTENT
,\.
\
F 10- 2 --Q=O
------ Q = 10
-·--Q=IOO
\
\
J
10- 3
10- 3 10- 2 10-1 10° 10' 10 2 103 10 4
X
Fig. 24. Generalized flow-pattern map. Condensation in horizontal

tubes. See Fig. 23 for definitions.
20r---,------.------.------r~
10
U lS
(m !sec)
01
002L---~----~-- ____~~L-_L~
02 10 100 200
Fig. 25. Flow-pattern map for condensation of steam; 2.5-cm-diameter

horizontal tube at 1.0 atm.
maps for condensation and boiling of pure water in a 2.5-cm-diameter

horizontal pipe at near atmospheric pressure with a heat flux of
100 Wjcm 2 . For comparison the boundaries calculated for q = 0
are included. The profound effect of the heat transfer is evident,
as is the fact that the direction of transfer is also important.
Breber et al. (1980) summarized the flow-pattern data in the

literature for condensation in horizontal tubes. Included were
data from six investigations for refrigerant R-12, R-113, steam,
and n-pentane taken in six pipe sizes ranging from 0.48 to 5.1 cm
in diameter and from 0.61 to 6.1 m long. They compared observed
20,---,------,------,-----,,-.
10
UlS
(m/sec)
.1 STRATIFIED l
~:l '-__- -' -__:_:_~_0__'w_/c_m_2
02
--,I
___I'__\_.L···""_.··__->-I
10 100 200
lbs(m/sec)
Fig. 26. Flow-pattern map for boiling of water; 2.5-cm-diameter

horizontal tube at 1.0 atm.
patterns with the theory developed in Sec. 5 for the case of no

boiling or condensation and concluded that the agreement was
reasonably satisfactory, although not fully correct. However, with
the development just completed, it is now possible to make a
comparison with the correct theory as modified for condensation.
Although details of the thermal aspects of the experiments are

not complete estimates indicate that values of Q [Eg. ( 8)] ranged
from 0 to 10 2 . In Fig. 27 the theoretical transition boundaries A
and Bare plotted for these two values of Q. The region in which
10 1 ,---------.-------~,--------.
//
10-1 L -_ _ _ _ _ _ _ _L -______~L__ _ _ _ _ _~
10- 2
x
Fig. 27. Theory versus experiment: Condensation data. Crosshatcted
area represents data for which annular flow pattern was
observed.
annular flow is to be expected is above A and to the left of B.

Palen et al. (1980) plotted all data from these six studies, which
were reported as displaying the annular flow pattern, and these are
shown in Fig. 27 as the crosshatched area. The prediction is
quite satisfactory.
A similar comparison is made in Fig. 28 for stratified flow.

'rhe region below A is predicted to be stratified. The crosshatched
area represents the location of the data that were reported as
stratified. Again, agreement is excellent.
The comparison for intermittent flow is not as definitive.

The data for 0.61-m-long test sections are in paor agreement with
prediction, as would be expected since these pipes are not long
enough to establish slug flow. The remaining data, reported by
Travis and Rohsenow (1973) for an 8-mm-diameter, 4.4-m-long test
section using R-12 and Calder for steam in a 25.4-mm-diameter tube
1. 27-m-long, are in agreement ,>lith theory, as shown in Fig. 29.
For the data displaying slug flow, the value of Q was less than 5,
and the theoretical transition curves have been calculated
accordingly. The theoretical region for slug flow is above A and
to the right of B, precisely where the da ta are seen to fall.
8 VERTICAL PIPES: STEADY MOTION WITHOUT MASS

TRANSFER - UPFLOW AND DOWNFLOW
Many empirical flow-pattern maps have been suggested for gas-

liquid flow in vertical pipes. Most of these studies dealt with
10 1
Q = 102
A
l
A
10°
0=0
F A
I
I
;~~I
IO-I~
I
J
Al
~II
I
10- 2 10- 1 10° 10i
X
Fig. 28. Theary versus experiment: Condensation data. Crosshatched

area represents data for which stratified flow pattern
was observed.
10°F [I.
~I A~~t~~oc
10-1 0 CL
00-0
F
o TRAVIS e. ROHSENOW
(~73) R-Q
A,,\OCO °0
o CALDER (1976)
STEAM 1 o
x
Fig. 29. Theory versus experiment: Condensation data. Points
represent data where slug flow was observed.
upward concurrent flow, while a few were concerned with the

concurrent downflow configuration. In this section physically
based models will be developed for flow-pattern transitions for
these two cases. Another situation that can exist is countercurrent
flow, where the liquid flows down and the gas up. It 1s important
to predict the condition of flooding or flow reversal where the
liquid changes direction. This is, of course, a flow-pattern
transition as weil, but will not be discussed here.
8.1 Concurrent Upflow
The four flow patterns that can be observed have been discussed
and are illustrated in Fig. 2. Maps for predicting these
transitions based largely on experiment have been suggested by Duns
and Ros (1963), Sternling (1965), Wallis (1969), Hewitt and Roberts
(1969), Govier and Aziz (1972), and Gould (1974). Griffith and
Wallis (1961) invoked some ideas based on physical modeling to
arrive at a selection of coordinates for their map, and a similar
coordinate system was adopted by Oshinowo and Charles (1974) and
Spedding and Nguyen (1980). In this section physically based
transition criteria will be developed based largely on the ideas
presented by Taitel et al. (1980).
8.1.1 Bubble-to-Slug Transition
Low Liquid Rates: When gas is introduced at low flow rates

into a large-diameter vertical column of liquid (flowing at low
velocity), the gas phase is distributed into discrete bubbles.
Many studies of bubble motion demonstrate that if the bubbles are
very small they behave as rigid spheres rising vertically in
rectilinear motion. However, above a critical size (about 0.15 cm
for air-water at low pressure) the bubbles beg in to deform, and the
upward motion is a zig-zag path with considerable randornness. The
bubbles randomly collide and coalesce, forming a number of somewhat
larger individual bubbles with a spherical cap similar to the Taylor
bubbles of slug flow but with diameters smaller than the pipe.
Thus, bubble flow is characterized by an array of smaller bubbles

moving in zig-zag motion and the occasional appearance of larger
Taylor-type bubbles. The Taylor bubbles are not large enough to
occupy the cross seotion of the pipe so as to cause slug flow.
Instead, they behave as free-rising spherically capped voids, in
the manner originally described by Taylor. With increases in gas
flow rate, at these low liquid rates, the bubble density increases,
and a point is reached where the dispersed bubbles become so closely
packed that many collisions occur and the rate of agglomeration to
larger bubbles increases sharply. This results in a transition to
slug flow.
Experiments suggest that the bubble void fraction at which

this happens is around 0.25-0.30 (Griffith and Snyder, 1964). A
semitheoretical approach to this problem was given by Radovcich
Moissis (1962) by considering a cubic lattice in which the
individual bubble fluctuates. They postulated that the maximum
void fraction is reached when the frequency of collision is very
high, and it was shown that this happens around a void fraction of
0.30.
An alternative approach is to consider this problem from the

point of view of maximum allowable packing of the bubbles. If we
consider the bubbles to have spherical shapes and to be arranged in
a cubic lattice, the void fraction of the gas can be, at most, 0.52.
However, as a result of their deformation and random path, the rate
of collision and coalescence will increase sharply at void fractions
weil below this spacing at which they touch. Therefore , the
closest distance between the bubbles before transition must be the
one that permits some freedom of motion for each individual bubble.
If the spacing between the bubbles at which coalescence beg ins to
increase sharply is assumed to be approximately half their radius,
this corresponds to about 25% voids. Published data agree in that
the void fraction in bubbly flow rarely exceeds 0.35, whereas for
void fractions less than 0.20 coalescence is rarely observed
(Griffith and Wallis, 1961). Thus, at liquid rates low enough that
bubble breakup due to turbulence is small, the criteria for
transition from bubbly to slug flow is that the void fraction
reaches 0.25.
If the gas bubbles rise at velocity Ur., this velocity is

related to the superficial gas velocity UG~ by
U (74)
G
where a is the void fraction. Likewise, the average liquid

velocity is given in terms of the liquid superficial velocity as
(75)
Designating UQ as the rise velocity of the gas bubbles relative to

the average l~quid velocity, Eqs. (74) and (75) yield
(76)
In this simple model, radial variation of local velocity is ignored,

with UG and UL asslliued equal to the average velocity.
At lower liquid velocities the size of the bubbles are large

enough that their rise velocity is independent of size and depends
only on physical properties. For single bubbles rising in an
infinite media, Harmathy (1960) suggested this relationship
1. 53 (77)
However, when one bubble rises in the oresence of a swarm of other

bubbles, as is the case here, Zuber and Hench (1962) suggested that
the corrected expression can be estimated from
(78)
Substituting for UD in Eq. (76) and considering the transition to

take place when a = aT = 0.25 gives the equation characterizing
this transition
3.0 UGS - (79)
Note that this mechanism is valid only when the liquid rate is low
enough that turbulent dispersive forces are not dominant. Once
fluid properties are designated, the theoretical transition curve
can be plotted on ULS versus UGS coordinates and will remain
invariant with tube size. Such a curve is shown in Fig. 30 for the
water-air system at 25 0 C and 10 N/cm 2 where it is designated as
curve E. At higher gas and liquid flow rates, where the bubble
rise velocity relative to the liquid velocity is negligible, the
theoretical transition curve is linear with a slope of unity in
these log coordinates. On the other hand, at low liquid rates
where liquid velocity is negligible, the boundary of the bubble
region is controlled by the free rise velocity of the bubbles and
is essentially independent of liquid rate.
High Liquid Rates: At higher liquid flow rates, turbulent

forces act to break up and disperse the gas phase into small
bubbles even for vapor void fractions higher than 0.25. The theory
of breakup-immiscible fluid phases by turbulent forces was given by
Hinze (1955) and confirmed by Sevik and Park (1973). Hinze
determined that the characteristic size of the dispersion results
from a balance between surface tension forces and those due to
turbulent fluctuations. His study led to the following relationship
for the maximum stable diruueter of the dispersed phase, d max :
3/5 -2/5
d = k ( ~) (s) (80)
max PL
10 DISPERSED BUBBLE
TI
F- 2 -----..'3.-_...,--,; J
F-I/ E
1.0 "-
.,..,on '\ H
H \
"- \ \
E \ \ \ I ANNULAR
U1 0.1
I \SLUG OR CHURN
..J
::J \ \ \ I I
\ \ I I J
I INI I Y
0.01
E I I I I
I 590
IE/ D = er I~O I
I 290 I
I
0.1 1.0 10.0 100
U GS (rn/sec)
Fig. 30. Flow-pattern map: Upward flow in a 5.l-em-diameter tube.

Air-water at 1.0 atm.
where s is the rate of energy dissipation per unit mass. Hinze's

investigation explored dispersion under noneoaleseing eonditions
that ean be realized only at very low eoncentrations of the
dispersed phase. He applied his formula to the data of Clay (1950)
for drop let breakup at low concentration of dispersed phase and
found k to be equal to 0.725. Sevik and Park (1973) developed
theoretical values of k by eonsidering the natural frequeney of a
bubble or drop in its lowest-order mode of vibration.
The rate of energy dissipation per unit mass for turbulent

pipe flow E: is
Ij:1
U
fVJ
E = (81)
PM
where
dP li PM uM
2 (82)
dx D
and UfVJ = ULS + UCS ' Substitution of Eqs. (81) and (82) ~n~o Eq.
(80) shows that turbulent breakup takes place at all liquid rates
for whieh turbulent flow exists. However, if the bubble size
produeed by this breakup is still large enough to permit
deformation with the resulting zig-zag motion, then at a ~ 0.25
eoaleseenee will take plaee and the transition to slug flow will be
observed. Thus, the turbulent breakup proeess can prevent
agglomeration only if the bubble size produeed is small enough to
eause the bubbles to remain spherieal and therefore to move
reetilinearly upward with the liquid. The bubble size at which
this oecurs is given by Brodkey (1967) as
(83)
Onee turbulent fluetuations are vigorous enough to eause the

bubbles to break into a size such that d max S deritt the bubbles
remain dispersed even for a > 0.25, and the transition to slug flow
cannot take place. Setting d max = derit in Eq. (80) I substituting
for E from Eqs. (81) and (82) ,and using
f = 0.046 ( v:D)-O'
u 2
(84 )
gives the eriteria for this transition as foliows:
U 1.12 3.0 (85)

M
Onee the fluid properties and pipe size are set, Eq. (85) defines
the relationship between the values of UCS and ULS above whieh slug
flow eannot exist. For air-water at 25 0 C and 10 N/em 2 pressure,
this result is shown in Fig. 30 for a 5.0-em-diameter pipe and is
designated by eurve PI. However, regardless of how mueh turbulent
energy is available to disperse the mixture, bubble flow eannot
exist at paeking densities above a = 0.52. Thus, the F eurve
delimiting dispersed bubble flow must terminate at the eurve G,
whieh relates ULS and UCS for a = 0.52. This approach neglects the
effect of coaleseence on the bubble size produced by the turbulence.
Calderbank (1958) bubbled gas through a pool of liquid pr,viding the
mixing with an agitator. He showed that the group dmaxE~ 5/(a/PL)3/5,
which can be formed from Eq. (80), depends linearlyon a . Barnea
et a1. (1982b) assumed this resu1t could be app1ied to f10w systems
and arrived at the following modifieation to Eq. (85):
U 1.12
1'4
Ur,rl~ s)iJ . 5
3 • 0 + 1 7 • 0 ( ~., (85a)
For this development they suggest that d 9rit is twice that given by
Eq. (83). The effect on the prediction lS small, although Eqs. (85)
and (85a) prediet opposite slopes in the ULS versus UCS transition
curve, and these are shown as curve F-2 in Fig. 30.
Thus, the bubbly flow pattern can be seen to exist in zones I

and 11 of Fig. 30. In zone I to the 1eft of curve E and below F,
one predicts the presence of deformable bubbles that move upward
with a zig-zag motion with Taylor-type bubbles oecasionally appearing
in the liquid. In zone 11 above curve P and to the left of G, one
observes a more finely dispersed bubble system without any Taylor
bubbles. To the right of curves E and below F in zone 111, one
expects to see the slug pattern.
Tubes of Small Diameters: Still a different transition

mechanism comes into play in the special case of tubes of small
diameter. Consider zone 1 of Fig. 30, where one observes discrete
deformable bubbles rising in zig-zag paths and the occasional
appearance of a Taylor bubble. The velocity of rise of the
deformable bubbles relative to the liquid Uo is given by Eq. (78)
and depends only on the properties of the fluids. The rise velocity
of the Taylor bubbles on the other hand is given by Nicklin et al.
(1962), U c " 0.35 /g15 and is property independent. Whenever Uo < UC1
the rising bubbles approach the back of the Taylor bubble coalescing
with it and increasing its size. Under these conditions bubbly flow
cannot exist in zone I. On the other hand, when Uo < Uc ' the Taylor
bubble rises through the array of distributed bubbles, and the
relative motion of the liquid at the nose of the Taylor bubble
sweeps the small bubbles around the larger one, and coalescence does
not take place. For air-water at low pressure at a = 0.25 (so that
transition is expected), Uo " Ue at D " 4 cm. Thus, for tubes
smaller than about 4 cm in diameter, no bubbly flow can exist below
the curve F, and the entire zones land 111 exist as the slug flow
pattern. Only at high liquid rates (zone 11) can bubbly flow exist
for small tubes where dispersion occurs due to turbulence. The
flow-pattern map for 2.5-cm-diameter tubes with the system air-water
is shown in Fig. 31, where zones land 111 are combined. A system
having a small diameter is one that satisfies the criterion below:
(87)
It is of interest that the range of diameters usen ln most laboratory

air-water experiments, i.e. I 2-6 cm, spans this critical diameter
of 5.0 cm. This accounts for the apparent differences in
observations that have been reported. It also shows that for this
10
G
II
J
\
1.0 F-I H
u J. )...."
H
H
\
\
\
I
ANNULAR
....~ /
/ H \
\ SLUG OR CHURN
\ \
E 0.1
-(J) SLUG \ I
::J
..J \ I I
1all \ INI J }l
\ I I I
0.01 i 100 I 500
.fElD = 50 1200 I
, I
0.1 10 {OO
Fig. 31. Flow-pattern map: Upward flow in a 2.5-cm-diameter tube.

Air-water at 1.0 atm.
GaS-Liquid Flow Pattern Transitions 53
particular transition experimental data taken on small pipes cannot

be scaled to larger diameters. It is only through an understanding
of mechanisms as discussed here that rational size scaling can be
accomplished.
8.1.2 Slug-to-Churn Transition
As the gas rate is increased still further, a transition to the

churn flow pattern takes place. In slug flow the Taylor bubbles
and liquid slugs propagate at constant speed. In churn flow the
liquid slug is too short to support a stable liquid bridge between
two consecutive Taylor bubbles. The falling film around the bubble
penetrates deeply into the liquid slug, creating a highly agitated
aerated mixture at which point the liquid slug is seen to
disintegrate and to fall in a rather chaotic fashion. The liquid
reaccumulates at a lower level at the next slug, where liquid
continuity is restored, and the slug then resurnes its upward motion.
Thus, one observes an oscillatory motion of the liquid, which we
consider the characteristic identification of churn flow.
Observations of the slug-churn flow patterns on 2.5 and 5.0 cm
diameter test sections in our laboratories suggest that the churn
flow pattern is an entry region phenomenon associated with the
existence of slug flow farther along the pipe. That is, whenever
one observes slug flow, the condition near the entry appears to be
churning. Furthermore, the entry length, or the distance that such
churning can be observed before stable slug flow takes place,
depends on the flow rates and pipe size.
The process of developing a stable slug near the entrance

section can be described as foliows. At the inlet the gas and liquid
form short liquid slugs and Taylor bubbles. A short liquid slug is
known to be unstable, and it falls back and merges with the liquid
slug coming from below, causing it to approximately double its
length. In this process the Taylor bubble following the liquid
slug overtakes the leading Taylor bubble and coalesces with it as
the slug between the two bubbles collapses. This process repeats
itself, and the length of the liquid slugs as weIl as the length of
the Taylor bubbles increases as they move upward until the liquid
slug is long enough to be stable and form a competent bridge between
two consecutive Taylor bubbles. Between the inlet and the position
at which a stable slug is formed, the liquid slug alternatively
rises and falls, and this is precisely the condition of churn flow.
As the gas rate increases, it is evident that the length of this
entrance region increases to the extent that it can occupy the
entire length of any test section. Thus, one should think of churn
flow as an entrance phenomenon. Since in practice all pipes are of
a finite length, it would be useful to provide some estimates of the
lengths over which churn flow is the predominant mode. With this
objectives we develop a method for calculating the entry length
required to develop stable slug flow. The distance from the
entrance to that length will be observed to be in the churn flow
pattern. The development that follows will evolve a way of
estimating the length.
The rise velocity of a Taylor bubble is given quite accurately

by Collins et al. (1978)
(88)
where Ue is the centerline velocity upstream of the bubble. For

fully developed turbulent flow, Ue " 1.2 (ULS + UCS) = 1.2 UM"
Consider two consecutive Taylor bubbles, as shown in Fig. 32~ The
two bubbles will move at the same speed when the slug length ls is
long enough that the velocity profile in the liquid in front of the
second bubble is the same as that in front of the first, so that
their centerline velocities are the same. This is the situation to
be expected when the slugs are long enough that the turbulent
velocity distribution in the liquid can be fully re-established
before the next Taylor bubble appears.
Since the liquid slugs are shorter in the entry region, the
velocity distribution in the liquid can be severely distorted by
the flow reversal near the wall as a result of the falling film.
Consider the velocity distributions in the planes A-A and B-B
behind the leading Taylor bubble shown in Fig. 32. If the liquid
slug is long far enough behind the trailing edge of the bubble
(plane 8-B), the velocity becomes that typical of turbulent flow.
However, at A-,4 the flow is downward ne ar the wall as a result of
the falling film around the bubble. To maintain mass continuity,
the velocity at the centerline must increase. Since the velocity
of a Taylor bubble depends on the centerline velocity plus its rise
velocity, it is clear that, for liquid slugs too short to
re-establish the turbulent velocity distribution, the second bubble
will overtake the first (Moissis and Griffith, 1962). As a result,
the two bubbles will coalesce, the liquid bridge between them will
disintegrate and will fall to a lower level, creating churn flow.
Experimental observations for water-air systems suggest that

the length of a stable slug relative to this diameter raiD is
fairly constant and independent of gas and liquid flow rates
[Govier and Aziz (1972), Akagawa and Sakaguchi (1966)J. The minimum
Iloceool
000001
! gggS8
TAYLOR BUBBLE ~.:! '\1 !G 'li 1

, lAI
Ai. i UG I 1 A
---,1;j-;L--J -k+-+-
I J
LIQUID SLUG 0 eooo:"'-
1 -I I I
~+ULggl ~s
X'r--Iocoeo I
B 11
FALLING FILM
Uf
----+~
i~1 ~ U 'I
100000
! ggggg
100000
I
Fig. 32. Slug flow geometry. From Taitel et al. (1980).
value of leiD reported was 8. Studies in our laboratory, using very

long, 2.5- and S.O-cm-diameter tubes, showed stable slug lengths
approaching 16-20 D. The earlier observations can be considered
the result of two slugs, each not quite of stable length leiD = 8,
which approach each other so slowly that they would never coalesce
except in a long tube. By use of the very approximate argument
that folIows, it is possible to show that this stable length
leiD = 16 observed for air-water should be essentially independent
of fluid properties or pipe diameter.
The liquid film falling along the Taylor bubble has an average
velocity Uf and a velocity relative tQ the liquid at plane A-A
behind the bubble of (Ut + Ue)' Consider this liquid sheet as a
two-dimensional jet that enters a stagnant pool of liquid (the
slug) at a uniform velocity (U+, + Ue ). The local axial velocity U
in the liquid induced by the j~t will depend on the distance x in
the direction of the jet and on y, the normal distance from the jet
centerline. Both experimental and theoretical studies have shown
that the ratio of U(x,y) to Umax(x,O) varies as
~
max
= 1 - tanh 2 (Y ~.) (89 )
where y is a universal constant approximately equal to 7.67

(Schlichting, 1968). A stable slug is one that is long enough
that the jet has been absorbed by the fluid and the velocities
have slowed to that of the surroundings. In this case we explore
the distance x = ls' which at the centerline y = DI2 the velocity
is essentially flat, say UIUmax ~ 0.05, and thus the normal
turbulent distribution in the liquid slug is undistorted. Equation
(89) shows that this takes pI ace at leiD = 16. This is, of course,
only an approximate argument because the falling film is a wall jet,
not a free jet, and the fluid is confined, not of infinite extent.
However, Patel (1971) showed that the velocity distributions in a
wall jet on the side of the velocity maximum away from the wall can
be estimated as in a free jet. Since the falling film is so thin,
the approximations used above become quite reasonable.
Entry Length for Churn Flow: Designate lE as the entry length

of pipe required to establish stable slug flow and, therefore, as
the region that one would observe churning. Consider coordinate x
pointing downward from the trailing edge of the leading Taylor
bubble, as in Fig. 32. The velocity at the center of the pipe Ua
varies from Ue at x = 0 to 1.2 UM at x = le' Assume exponential
variation with x as follows
(90)
The constant ß determines the decay rate and was chosen ß = In 100 =
4.6, so that at x = le the decay will be 99%. The final results
are not sensitive to the particular choice of ß or to the particular
profile assumed, as long as Ua(x = 0) = Ue and Ua(x = le) = 1.2 UM'
Designating the leading bubble as the first and the trailing bubble
as the second and using Eq. (88) to calculate the velocities of the
two consecutive bubbles, we obtain an approach velocity between two
bubbles -:1; as
-x
In calculating the entry length or length for churn flow, it

is assumed that near the gas liquid inlet coalescence is
instantaneous and that short Taylor bubbles as weIl as short liquid
slugs are formed. The merging of the Taylor bubbles to larger gas
bubbles and large liquid slugs takes place when the second bubble
overtakes the leading bubble. When they combine, the volume and
the length of the newly created Taylor bUbble are doubled. Likewise,
the liquid slug behind the leading bubble falls and merges with the
liquid slug behind the second bubble to create a slug twice as long.
At approximately the same time, the third and fourth bubbles will
combine and again create a new bubble and new liquid slug of double
length. This process will go on, and pairs of bubbles will
coalesce as they move upward, doubling their length each time until
a stable liquid slug of length lslD ~ 16 is formed. The term ls
designates the length of a stable slug, while l; is the length of a
slug formed during the period of coalescence. Because the last
merger of two slugs of length &i = 8 D is quite slow, we consider
the entrance or churning length to exist up to the point where
li = 8 D (or li = lsl2), the region of churn flow.
Integrating Eq. (91) gives the time needed for each merger as
a function the distance li' Between two consecutive bubbles
t. (92)
1-
where i takes successive values of 0, 1, 2, 3, . . . . Letting Li take

the sequence from 0 to ls/4, namely, z'i = ls/4, lsl8, Z.s/16, .. -rü,
yields an infinite series for ti whose sum multiplied by Ue yields
the estimated entrances length
l U 00
s G
L (e ö/2n _ 1) (93)
ü.35S rgJ5 n=2
or considering ß = 4.6 and ls 16 D yields
35.5 (94)
But Ur- can be determined from Eq. (88) with Ue 1.2 UM' The
result is
42.6 (95)
where UM = Ues + ULS ' This shows that the dimensionlrss entry
length for churning depends on one parameter: UM/(gD) 2. The
solution to this equation for a low-pressure air-water system at
several va lues of LEID is shown in Figs. 30 and 31 for 5.0- and

2.5-cm-diameter tubes where these curves are designated as H.
Figure 30 shows typical trends for pipe diameters larger than

that needed to satisfy Eq. (S7), that is, pipe sizes where a bubb1y
flow pattern can be expected to exist. Note that the F curves
de1ineating the transition between slug and churn flow terminate
on the E curve. This E transition boundary represents the locus of
points in the ULS - UGS plane, where a = 0.25, and for a < 0.25
only bubb1e flow can exist.
On the other hand, for the smal1er pipe sizes represented in

Fig. 31, on1y slug flow is observed, even for a < 0.25. However,
churning can be expected on1y when a > 0.25. Thus, the locus of
ULS - UGS values for a = 0.25 is shown dotted on Fig. 31 and still
represents the terminus of the F curves.
It is now possible to predict if churn flow will be observed

at any axial position a10ng the pipe. Assume that the point of
observation is 200 diameters above the entry. OVer the flow rate
region to the left of the LEID = 200 curve of Fig. 30 and to the
right of the E curve, slug flow will be observed. At flow rate
pairs to the right of curve F, churn flow will be seen, providing
the rate is not high enough to cause the annu1ar pattern. Although
churning is visualized as an entry region phenomenon, conditions
where churn flow would be observed over the entire length of a tube
can be shown to exist. For example, consider a S.O-cm-diameter
pipe whose length is somewhat less than 200 D. Va lues of ULS and
UGS located just to the right of the LEID = 200 curve of Fig. 30
wou1d resu1t in churn flow being observed at all positions a10ng
the pipe. In other words, the entire tube consists of entry
1ength for purposes of developing slug flow.
8.1.3 Transition from Annu1ar F10w
For high gas f10w rates the flow becomes annu1ar. The liquid
film flows upward adjacent to the wall, and gas f10ws in the center
carrying entrained liquid droplets. The upward flow of the liquid
against gravity resu1ts from the forces exerted by the fast-moving
gas core. This film has a wavy interface, and the waves tend to
shatter and enter the gas core as entrained droplets. Thus, the
liquid moves upward due to both interfacial shear and form
"drag" on the waves and drag on the droplets. Based on the idea by
Turner et al. (1969) app1ied to gas lift operations, we suggest
that annu1ar f10w cannot exist unless the gas velocity in the gas
core is sufficient to lift the entrained droplets. When the gas
rate is insufficient, the drop lets fall back, accumu1ate, and form
a bridge, and then churn or slug flow takes p1ace.
The minimum gas velocity required to suspend a drop of

diameter d is determined from the balance between the gravity and
drag forces acting on the drop
(96)
or
2 r~ (p L - 0 G) dl ~
13 L PGC d J (97)
The drop size is determined by the balance between the impact force
of the gas that tends to shatter the drop and surface tension
forces that hold the drop together. Hinze (1955) showed that the
maximum stable drop size will be
Ku
d = (98)
- u2
PG' G
where K is the critical Weber number and takes a value between 20

and 30 for drops that are gradually accelerated.
Using Eqs. (97) and (98) yields
(99)
As suggested by Turner et al. (1969), values K = 30 and Cd = 0.44

were selected. Note that K and Cd appear in the power of~. Thus,
the result for UG is quite insensitive to their exact values.
The gas velocity given Eq. (99) will predict the minimum
value below which stable annular flow will not exist. While this
analysis is applied to the droplets within the gas core, the Sru~e
treatment can be used for the crests of the waves on the rising
film, which are pictured as being supported by the gas strea~ in a
manner similar to the support of the liquid droplets.
Characteristic of annular flow is that the film thickness is

quite low even for relatively high liquid flow rates. As a result
the true gas velocity UG can be replaced by the superficial value
UGS' and the final transition boundary is given by
3.1 (100 )
This simple criterion shows that the transition to the annular

pattern is independent of liquid flow rate and pipe diameter. For
water-air at 25 0 C, 10 N/cm 2 , this velocity is calculated to be
about 15 m/s, and the transition boundary is plotted as a vertical
line in Figs. 30 and 31, designated as curve J. The dimensionless
group in Eq. (100) is recognized as the Kutataledze number.
Equation (100) is almost identical to the empirical results

of Pushkina and Sorokin (1962), who determined the air velocity
necessary to lift the liquid film for flooding experiments in tubes
varying from 6 to 309 mm in diameter.
8.1.4 Discussion
The flow-pattern transitions on UCS - ULS coordinates can

be readily calculated from the algebraic equations presented here
once pipe size and fluid properties are specified. In summary:
• Criterion for existence of the bubbly pattern: Eq. (87).
• Bubble to slug: Eq. (79).
• Bubble to dispersed bubble: Eq. (85).
• Slug to churn: Eq. (95).
• Churn to annular: Eq. (100).
Fluid properties and pipe diameter enter into the transition

equations to differing degrees depending on the transition being
considered. To illustrate this effect, the location of the
transition boundaries is shown in Fig. 33 for a crude oil-natural
gas system flowing at 30 0 C and 670 Njcm 2 pressure in pipes with
diameters of 5.1 and 30.5 cm. Oil density was set at 0.65 gjcm 3
with natural gas at 0.05 gjcm 3 . Oil and gas viscosities were 0.5
and 0.015 cP respectively.
Theoretically predicted transitions for upflow have been

compared with the recommendations of many earlier investigators
whose results were based primarily on experimental measurements.
There is considerable disagreement between the results of these
various investigators. However, the theory presented here is in
satisfactory agreement with the weight of the earlier experimental
results. In addition, comparisons with new data taken in 2.5 and
5.0 cm diameter tubes show good agreement, as shown in Figs. 34 and
35.
10 DB _ _ D= 5.1 cm
- - o =30.5 cm
1.0
.,u
UI
"-
E
I/)
A
...J 0.1
::J
f E/D=50
.01 fE/D= 20
1.0 10 100
UGS (m Isec)
Fig. 33. Flow-pattern map: Upward flow in vertical tubes. Crude

oil-natural gas at 38 0 C, 670 Njcm 2 .
10
DISPERSED BUBBLE
G J
QQQ Q
•••• H
QQQ
QQ6
•••
Q
ü
<l> 6
'"
"-
E 0.1 ••••••••••• 6
ANN.
• •
INTERMITTENT
(fJ
•• ••
--I
::)
0.01 ••••••••••
•••
6
•••••••••• 6
0.00 I '---_ _ _ _--L-_-L---L.I._ __.J

~
0.01 0.1 1.0 10 100

lbs (rn/sec)
• INTERMITTENT 6 ANNULAR
Q CHURN T DISPERSED BUBSLE
Fig. 34. Theory versus data: Vertieal upward flow. Water-air in a

2.S-em-diameter pipe; LEID = 400. Solid eurves represent
theory.
10
F_lo lse BUBBLE • ~1---l
~==F==~~T~T~T~~~ 11
1.0 -2T~TT.
TTTTY
•••••••
••••••
I
T T TT. •••••
I TTTT
TTT.
••• ~
.~~
6
6
0.01
BUBELE"'.
T T T.
INTERM
• ••
...
H.~~
~. ~
"Q Q
Q
~
6
"
ANN .
.. E• • • •• n~: "~
0.0010 .01 0 .I L0 10 100
Fig. 3S. Theory versus data: Vertieal upward flow. Water-air in a

S.l-em-diameter pipe; l~/D = 200. Solid eurves represent
theory. ~
8.2 Concurrent Downflow
Few studies have appeared dealing with downflow. Golan and

Stenning (1969) observed the patterns existing in an inverted
u-tube consisting of a vertical riser followed by a vertical
downcorner. Martin (1973) studied the transition from bubble to
slug flow on the vertical down configuration, as did Spedding and
Nguyen (1980). The latter proposed a classification scheme rather
different from any discussed here. The models for the transitions
treated in this section are based on the work reported by Barnea
et al. (1982a). Figures 36 and 37 show th.at the patterns observed
by Barnea et al. in 2.5- and S.l-cm-diameter pipes were annular,
slug, and dispersed bubble.
Modeling of the flow-pattern transitions starts from the

condition of annular flow that exists at low liquid and
gas rates where the liquid flows as a wavy film with the gas in
the core. As the liquid rate increases, so does the film thickness.
Eventually the film ~nd its waves become thick enough to bridge the
pipe, and the annular pattern no longer exists. The transition can
take place either to slug flow or to the distributed bubble pattern.
When the pattern changes to slug flow at this transition, then at
higher liquid rates turbulent fluctuations cause breakup of the
large Taylor bubbles, and a transition to the dispersed bubble
pattern takes place.
8.2.1 Transition from Annular Flow
The criteria suggested for transition from the situation of a

falling film to a slug or bubble flow are similar to those discussed
in Sec. 5.2 for horizontal pipes. There it was shown that a slug
will form from a stratified layer when the conditions are such that
waves exist on the interface and the liquid supply in the film is
large enough to provide the liquid needed to bridge the pipe. But
waves exist on vertical falling films under all conditions, even
in the absence of gas flow; thus, it is necessary only to satisfy
the inventory condition. It is suggested that the same condition
be used in vertical downward flow as was used for horizontal
0.5 (101 )
As discussed previously, ~ 0.7, thus the transition criteria is

.4 LIA 2: 0.;) 5 •
The area ratio is found by analyzing the falling film shown

schematically in Fig. 38. Amomenturn balance is made on each phase
for steady flow and the pressure drop set equal from the two
equations to give:
o {l02j
This is, of course, identical to Eq. (6), developed for stratified

flows, except for the absence of the first term. Proceeding in an
analogous manner produces the following dimensionless equation
10
... ..
DISPERSED BUBBLE G v v
v v v v yv I ••
v v v v y vF-2 • INT
~..
1.0 •• •• • F-I • • ••• 8 •
~
• ·~.!!ld!66L'.6L'.
6 6 6 L:.~6..6.L\ü~~~L\
6 L'. 6 L'.
UCI) L'. L'. 6 6
C"
"- GI 6 L'. 6 6 L'. Ll. L'. 66
E
lfJ
ANNULAR
...J
1
::::l
GOI 6 L'. L'. L'. 6 6 6 66
6 6 6 L'. L'. 6 6 66
6 6 6 6 6 L'. 6 66
OOObol
I
0.1 1.0
I
10
I
100
UGS (rn/sec)
Fig. 36. Theory versus data: Vertical downward flow. Air-water in

a 2.5-cm-dia~eter pipe.
10
....
08 /E G
"I,F-2 v "' v •
INT.
.........
'L~ -!_L.'L__T"~.
vvvvvF-il.- ....." . . .8
1.0 ...
·····r·6666666L'.66
t:.6t:.t:...-tS66666 666
6 _-( 6 6 666
UCI) -c; 6 6 666
V>
"- 0.1 6 6 L'. 666
E
lfJ
...J
ANNULAR
::::l
GOI 6 6 6 666
L'. 6 6 666
o.OObOI GI 1.0 10 100

UGS (rn/sec)
Fig. 37. Theory versus data: Vertical downward flow. Air-water in

a S.l-em-diameter pipe. See Fig. 34 for legend.
equivalent to Eq. (10):
(ÄIL + I~) - 4 Y o (103)
X and Y are defined as in Eqs. (11) and (12) with sin ß = 1.0.
The definitions that differ from those given in Eqs. (13)-(19) are
I,'I'
I 1
I 1
III
I
I
1
1 111
1
UG,
11
It 1 I
l
I
I
1I
11 -I I
r
1[1 III!
I,
1
1\1'
'I
1
1'1 1I1
h
I 11 1
,li
D
I I,
Fig. 38. Annular do~mward vertical flow.
A
r
rr ['!!:.. - ( -h
D D! J
YlI AC rr -
(~ ~y
Sr = rr S.
"-
rr(l - 2 -i)
v
Thus, Eq. (103) provides a solution for film thiekness h onee X and
Y are specified. One might question using a frietion factor
equation for single-phase flow to ealeulate the interfacial shear
as was done here. However, it is readily shown that even at the
highest gas rates the force due to interfaeial shear is two orders
of magnitude less than that due to gravity. Thus, the seeond term
in Eq. (102) is small eompared to the first, and errors made in
that term will have little influenee on the result.
Now, given fluid properties and diameter and using Eq. (103),
it is possible to find the values of UCS and Urs that will give the
X-Y pair such that hlD = 0.097. This results in ArlA = 0.35 as
required by Eq. (101) for transition. This theoretieal result for
air-water in 2.5- and 5.1-em-diameter pipes is shown in Figs. 36
and 37 as eurve B, where reasonable agreement with the da ta is
evident.
8.2.2 Slug-to-Dispersed Bubble Transition
As in the case of upward f low, two mechanisms are operative':

(a) as the void fraction exceeds 0.25, dense packing of bubbles
results in a transition to slugs, and (b) at high liquid rates
turbulence causes breakup of slugs into a fine dispersed bubble
pattern. For downflow the expression corresponding to Eq. (79) for
the dispersed bubble-to-slug transition due to bubble packing is
3.0 UGS + (104 )
üf course this transition is absent when the diameter is smaller

than that necessary to satisfy the criteria of Eq. (87). As a
result, it will not be observed for a 2.5-cm-diameter pipe flowing
air and water, as displayed in Fig. 36. The theoretical curve is
shown as E in Fig. 37.
The mechanism of turbulent dispersion gives transition curves

for upflow from Eqs. (85) or (85a) with the latter considering the
effects of coalescence. These same processes are thought to apply
to downflow, and the transition curves appear in Figs. 36 and 37 as
F-1 and F-2. They must terminate on curve G, which is the locus of
points where a = 0.52, above which bubbly flow cannot exist.
For the 2.5-cm-diameter tube the comparison between theory and

experiment (Fig. 36) is straightforward and shows reasonable
agreement. Forthe larger tube (Fig. 37) the interpretation is
more difficult because two mechanisms for transition between the
bubble and slug flow patterns exist. The entire range for each
mechanism is indicated by the dotted curves. However, where
competing mechanisms exist in certain regions of the Urs - Ues
space, both cannot be operative, and the transitions thought to be
applicable are shown bythe solid curves that consist of segments
of each.
For example, consider the Urs - [fes region at low gas rate,
which falls above the E curve but below the B curve. The E curve
theory predicts bubbly flow above E, while the theory used to develop
the B curve predicts annular. Now consider a point on the E curve
located below B. Since a = 0.25 a transition to slug flow is
expected. However, the same liquid will distribute in annular flow
in such a way that the voids are considerably higher than a = 0.25
without causing transition. The fact that the annular pattern is
observed suggests that the preferred configuration is the one that
produces the larger value of a. In a similar way, consider the
region at low gas rates above the E curve and above the B curve
where bubbly flow is predicted from E curve theory and bubbly or
slug is predicted from B curve theory. The flow pattern that
exists in practice probably depends on which of these two
distributions maximizes the void counted. However, the question is
yet to be fully resolved.
The remainder of the theoretical predictions are in reasonable

agreement with the data. That is, above the B curve, to the right
of the E curve, below the F curve, and to the right of the G curve r
slug flow is predicted, and this is as displayed by the data. On
the other hand, above B but to the left of the G curve, and above
F but to the left of the G curve, the dispersed bubbly pattern
exists.
9 EFFECT OF PIPE INCLINATION
Small changes in the angle of inclination from the horizontal

can have profound effects on the flow patterns that exist. As
examples, consider the data for air-water in a 2.5-cm-diameter
tube for -0.5 0 (upward) inclination in Fig. 39 and +1 0 (downward)
inclination in Fig. 40. A comparison with Fig. 14 for a horizontal
pipe shows the drastic changes that can take place. For -0.5 0 the
stratified pattern disappears completely at low gas rates, while
for +1 0 the flow rate space for which intermittent flow exists
shrinks substantially. This effect can be understood from the
fact that even at very small inclination angles the force of
gravity acting in the flow direction can be of the order of the
wall shear stress.
On the other hand, small deviations from the vertical have

little effect, as seen by comparing the data in Fig. 41 for upward
flow in a test section inclined at 85 0 from the horizontal with
that for a vertical tube of the same size shown in Fig. 35. Only
the churn flow region appears to shrink somewhat in size. While
slight changes from the horizontal introduce a significant new
force acting in the flow direction, slight cbanges from the vertical
only modify existing forces due to gravity by the sine of the angle
Fig. 39. Flow patterns for air-water in a 2.5-cm-diameter tube at

0.5 0 upward inclination. Solid lines represent theory.
See Fig. 14 for legend.
--l
10,
~I:'SP:R~~ 1
10
~•
"
0
•
•
-:A~
~
INT
•
•
••
••
......
00 • • • • • • •
00 • • • • • • • •
•• A 6
IR 6.6.
6.
B
I
I
La ~
u STRT WAVY. • •• 6. 6. ANN
(lJ 0.1 I • • •
I ~ ~ K • • 'c' ~~
STRATIFIED SMOOTH I .~
0.00\ '-------'_ _- - I_ _'----J_ _---"
0.0.1 0.1 \.0. 10. 10.0.
UGS (rn/sec)
- - THEORY
Fig. 40. Flow-pattern map for air-water in a 2.5-em-diameter tube

at 1 0 downward inelination. Solid lines represent theory
See Fig. 14 for legend.
\0 DISPERSED BUBBLE
G
J
•
1.0
""
'0
(lJ
'"
'-
E "" ."" 1::.
0.1
"
.......... 6.
f:I. ... "
CI)
...J
::::> • INTERM " ... ANN
Q Q
Q Q
0.,0.\ "" •••• Q Q 6.
.....
" WlJ
Wl"
"" " "eH" 1::.
0..001 "------'-'--------'--'----'--'------'----'
0..0.1 0.1 1.0. 10 10.0
UGS (rn/sec)
Fig. 41. Flow pattern for air-water in a S.l-em-diameter tube at
85 0 upward inelination. Solid lines represent theory.
lEID = 200. See Fig. 34 for legend.
vf inelination. However, at large angles signifieant effects are
observed and these are diseussed below.
9.1 Concurrent Upward Flow in an Inclined Tube
At slight deviations from the horizontal or vertieal the

transition mechanisms already discussed would be expected to apply.

This is substantiated by comparing the transition boundary predicted
from theory as plotted in Figs. 39-41 with experiments. In fact, in
Fig. 39 it is seen that the theory correctly predicts the unexpected
presence of a dome-shaped area of stratified flow located between
the intermittent and annular pattern in the lower right region of
the curve. However, once the pipe is inclined at a large angle the
question arises as to which mechanism controls the transition - the
one that exists in a vertical pipe modified for deviation fram the
vertical, ar the one that controls the transition in a horizontal
pipe modified to account for inclination from the horizontal. The
theory for the effect of S ~ 0 on horizontal transition mechanisms
was presented in Sec. 5. Now we examine the effects of deviation
from the vertical on mechanisms presented for vertical tubes in
Sec. 8.
9.1.1 Bubble-to-Slug Transition and Minimum Inclination Angle for

Bubbly Flow
When the pipe is inclined from the vertical, buoyancy causes

the bubbles to migrate to the top wall of the pipe. High local
voids result, and a transition to slug flow can take place even
when the average void fraction in the pipe is below a = 0.25, at
which point transition occurs in vertical tubes. Two mechanisms
can operate to maintain the bubbly flow pattern: (a) at high
liquid velocities turbulence disperses the vapor into small bubbles
in the manner discussed in Sec. 8.1.1 and (b) lift forces near the
wall FL exceed the buoyant forces that move the bubble toward the
wall; thus, dispersion is maintained.
The controlling relations for mechanism (a) have already been

developed and suggest that the transition to dispersed bubble flow
will take place for any inclination angle when Eqs. (85) or (85a) is
satisfied. However, for lower liquid rates bubbly flow cannot
exist unless
(105 )
where the force of buoyancy on a bubble is the term in square

brackets on the right-hand side, and d is the maximum bubble
diameter. An expression for the lift force is more difficult to
arrive at and is approximated here by
(106 )
where Uo sin S is the axial component of the bubble rise velocity

relative to the liquid, AB is the projected area of the bubble in
the flow direction, and CL is the lift coefficient. Miyago (1925)
studied the shape and motion of bubbles in vertical tubes. He
indicated the existence of a zig-zag path as the bubble moved
upward. In addition, he observed that the bubble flattens in the
direction of motion, accelerating as it leaves the wall and
deaccelerating as it approaches the opposite wall. Figure 42 shows
how the shape changes with position along the path. Little is
RADIAL
COMPONENT
OF BUOYANT
FORCE -
Fig. 42. Zig-zag path of bubbles.
known about the shape or about the lift coeffieient, whieh depends
on the shape. Define a shape faetor for the bubble y sueh that 2
the projeeted area of the bubble in its long dimension is n/4(yd)
Here d is the diameter of a bubble of the same volurne having the
shape of a sphere. The angle at which the bubble traces its upward
path cannot be predieted, but as an estimate assume it is 45°. Then
(107)
Equating the lift and buoyant forces aeting in the direetion of the
wall gives
2
Uo2 PL GLy
cos S 3
cos 45 0 (108)
sin 2 ß
4" g PL-P G a
By this criterion, bubbly flow eannot exist when ß is less than

that which satisfies Eq. (108). In order to test the validity of
this very approximate eriteria, estimates are needed for y, CL'
and d. Uo ean be calculated from Eq. (77). Streeter (1961)
suggests values of CL ranging from 0.4 to 1.2, and based on
observation it appears that y ranges from 1.1 to 1.5. In these
experiments d ~ 0.5 cm. With these parameters the value of S
predieted from Eq. (108) ranges from 55 to 70 degrees. Shoham
(1983) reeorded extensive data over a range of inelination angles.
The results for 70 0 and 50 0 in a S.l-em-diameter pipe are shown in
Figs. 43 and 44. At 70 0 the system displays the bubbly flow
pattern, but at 50 0 it is no longer observed, thus the angle below
which bubbly flow cannot exist falls between these two, which is in
agreement with the predietion of Eq. (108). Clearlyadditional
. . . .,.,,,,. ···r·
lür------------------.----~
DISPERSED BUBBLE G
...............
1.0. .,.,.,.,.,
", .,,,, '"
... .,.,.,
.. .....
•
••J
/
F-2
••
] I •
. . ",":.. · ...
V., ...... • •
E 0.1
1
II
....
••"
....
':3 .....
BUSSLE!"'. I NTERM ANN.
üüll" E::!:
::J
0.0.0.1
'" ..
· ·
0..0.1 GI Lü 10. 10.0.

UGS (rn/sec)
Fig. 43. Flow-pattern Qap for air-water in a S.l-em-diameter tube

at 70 0 upward inelination. Solid lines represent theory.
10. DISPERSED BUBBLE

F-2
., '" ., v ... vi ......
• •••
•••• ••
·::r·..
00 • • • • • •
1.0 00 • • •
• ••• • •
o ••
00.
• J •• • I
u
<l> o. .,. IC. I
I
<J)
"- o.
E 0..1 co,• • • •
o. • •
Ul
o.
· ......
-l • • • IC.
::J INTERMITTENT ••• ANN
o.
II
....
• •• IC.
\.
oor~:··
o.
o.
• ••••
• ••• " I
00 • • • •
• "
IC.J
0..00.1
0.0.1 GI Lü 10. 10.0.
UGS (rn/sec)
Fig. 44. Flow-pattern map for air-water in a S.l-em-diameter tube

studies will be neeessary to test the validity of this meehanism

and to evolve methods for finding C&1 y, and d for other fluid
systems. Bubbly flow ean exist when the inelination angle is
greater than that indieated by Eg. (108) and when the diameter is
greater than the small diameter eriterion given by Eg. (87). When
both of these eonditions are satisfied, it is possible to have a
transition to slug flow when a ~ 0.25. The transition eriteria ean
be developed in a way fully analogous to that of Eq. (76), but the

axial component of the rise velocity Uo must be used. The final
result is equivalent to Eq. (79)
(l 09)
Curves E in Figs. 41 and 43 are plots of this equation and show

reasonable agreement with the data.
9.1.2 Transition from Annular Flow
The criteria for the transition from annular to slug or churn

flow can be obtained by a simple modification of the mechanism
developed in Sec. 8.1.3. Consider only the component of gravity
acting in the flow direction. The required modification to Eq.
(100) then becomes
3.1 (110 )
Curves J in Figs. 41, 43 and 44 show this prediction to be in

reasonable agreement with the data.
9.1.3 Slug-to-Churn Tr~~sition
Inclining the pipe from the vertical tends to segregate the

liquid and gas phases, thus providing separate paths that prevent
the fall back of liquid characteristic of churn flow. Thus, as seen
by comparing Figs. 35, 41, 43, and 44, as the inclination angle is
decreased from the vertical the churn flow region shrinks. No
quantitative description of this effect has been developed as yet.
9.1.4 Intermediate Angles of Inclination
Experimental studies by Shoham (1983) in 2.5- and 5.1-cm-

diameter pipes using air-water suggest that the vertical flow
mechanism modified for inclined tubes applies in an approximate
manner for angles as small as 150. For the range 00-150 the
horizontal flow mechanisms modified for inclination can be used.
However, the criterion for deciding which mechanism will apply in
an inclined tube is not fully clear. For example, Figs. 45-48
show the representative data of Shoham (1983) for angles from 20°
to 2°. Superposed on the data are curves representing theory for
applicable mechanisms from both horizontal and vertical tubes, as
modified for inclination angle. At 200 the J mechanism [Eq. (110)J
predicts the annular transition very well. The B mechanism, whose
origin rests in stratified flow theory, does not. However, for the
transition from intermittent to dispersed bubble flow the D
mechanism [Eq. (39)J developed for horizontal or inclined flow
shows somewhat better agreement. One way to explain this is to
assume all mechanisms are applicable but that the transition that
takes place at the lowest flow rate is the one that controls. For
example, the transition to annular flow, which depends strongly on
10r--;"""'==-==-,=--D-----,
...0 ...0 0 . ~: /"J

000. •
o. ••
•• .!
~ I ••
••
.L.
L.L.
~
00 •••••••••
0'1 : •
00.
•
•
"" OOI~ 0
INTERMITTENT
00. • •
• L. ANN.
I •
I
o 00 • • • • • • • • •
•
0.001 '-----'-_ _L--_----..l"--_----..l
0.01 01 1.0 10 100
U GS ( m/sec)
Fig. 45. Flow-pattern map for air-water in a S.l-em-diameter pipe

10r---------------~~~---.
u O.
Q)
Ul
'-
E 000000 •••
0'1
i
Cf)
.J INTERMITTENT
:J O.
!
000 00.
0.01 000000 • • •
000000 • • • • • •
0.001
0.01 01 LO 10 100
U GS (m/sec)
Fig. 46. Flow-pattern rnap for air-water in a 2.5-crn-diameter pipe

the gas flow rate, will take plaee with the meehanism that requires
the lowest gas rate. On the other hand, the transition to disperse
bubbly flow takes plaee by the meehanism that requires the lowest
liquid rate.
9.2 Coneurrent Downward Flow in an Inclined Tube
Barnea et al. (1982b) explored meehanisms and obtained data for

10 DISPERSED BUBBLE 0
"
I.0 r o
0
J °r
] I o
OOr 0
o
I 0 0 0 • •••• • •
0.001 L . ._ - - - - - ' . _ _ . . l . . -_ _.....L.LJL---'--'--'
0.01 0.1 1.0 10 100

UGS (rn/sec)
Fig. 47. Flow-pattern map for air-water in a S.l-em-diameter pipe

at 50 upward inelination. Solid lines represent theory.
000
INTERMITTENT
o.
0.001"------"------'-----'--.1...1.------'
001 0.1 1.0 10 100
UGS (rn/sec)
Fig. 48. Flow-pattern map for air-water in a 2.S-em-diameter pipe

downward flow in inelined tubes. Two new meehanisms eome into play
when dealing with downward flow: (al the transition from stratified
smooth to stratified wavy may result from the action of gravity
rather than the action of the wind and (bl a new mechanism for
transition to annular flow exists.
9.2.l Stratified Smooth-to-Stratified Wavy Transition
When liquid flows downward, a film under the influence of

gravity can form on the surface even in the absence of gas flow
and its resulting interfacial shear. As developed in Sec. 5.4 for
horizontal tubes, the source of energy for wave formation is
interfacial shear. For inclined tubes the source is gravity. For
turbulent flow in smooth tubes, a simple criterion for the condition
at which waves will appear has been proposed (Albertson et al.,
1966)
2. 1. 5 (lll )
or
(112 )
Given agas and liquid rate, fluid properties, inclination

angle, and pipe size, the methods of Sec. 5.1 can be used to find
the equilibrium level hand the normalized areas. In that way the
locus of ULS-UCS values that satisfies Eq. (112) can be found.
These values are plotted as the curves marked K in Figs. 40 and 49.
These curves terminate at higher gas rates on the C curve, which
represents the condition at which waves are induced by interfacial
shear. In view of the ambiguity in the value of the Froude number
for this transition, this agreement is quite satisfactory. At
larger inclination angles both theory and experiment show that waves
exist at all liquid rates.
9.2.2 Stratified-to-Annular Transition
In horizontal tubes the transition between annular and

stratified results from the presence of a Kelvin-Helmholtz-type
instability when the liquid level is so low that there is
insufficient liquid in the fiL~ for slugs to form. The result
defines the regions to the right of curve A and below B in Fig. 14.
However, at high inclination angles the liquid level is small and
the liquid velocity is high, giving rise to still another mechanism.
Under these conditions of high liquid velocity, droplets are torn
from the waves at the surface of the liquid and deposited on the
upper wall, resulting in an annular film.
A preliminary model for this process can be constructed from

the situation pictured in Fig. 50. A droplet is tarn from the wavy
liquid surface having an initial radial velocity V. It then
executes a trajectory shovm by the dotted curve. Ifthe maximum
height of this trajectory w exceeds the distance to the top wall
(D-hl, the drop lets are deposited and annular flow results. We can
calculate w from
(113 )
2g cosß
10r-:::::::::::::=:::-=-:::::-:-.::------::==='1
DISPERSED BUBBLE D
h=-=-==-=T-::-~ TTTI
TTTTTTAT • • • • • • • • 1NT.
••••••• • •••••• B
1.0 •
•
•
•
u •
~
CI)
••
E
0.1 • •
STRAT WAVY
~
::::l
0.01
•
0
•
o 0 ••••••
K-----...... .a
o 0 0 ~ •••• •• A
STRAT SMOOTH C •• \
aool ' -o- _ - - L0._ _--'--'--_....l..---''---l
0 ••••••
0.01 0.1 1.0 10 100

UGS (rn/sec)
Fig. 49. Flow-pattern map for air-water in a 2.5-cm-diameter pipe

at 50 downward inclination. Solid lines represent theory.
Fig. 50. ~rajectory of a torn-away liquid particle in downward

stratified flow.
The initial radial velocity V is assumed to be the maximum value of

radial component of the turbulent fluctuating velocity at the liquid
surface, and this is estimated to be twice the rms value. The rms
is calculated from the friction velocity U* using Eq. (37). Barnea
et al. (1982) suggested
V (114)
Setting w D-h and using Eq. (114) in Eq. (113) gives the criterion
gD[l - (h/D)JcoSB
(115)
f L
1 h ~ ~
[ f ~ cos B
J (116)
Thus for any liquid level and B, ULS can be found from Eq. (116).
Figure 10 provides a value of X from which UG$ can be calculated.
In this way the locus of points for the transltion can be found.
The solution to this equation is shown in Figs. 51 and 52 as curve
L. Indeed, the data show that a region of annular flow appears at
low gas velocities above this curve. The B curve represents the
locus of points for which h/D > 0.35, and this satisfies the
conditions for transition to intermittent flow. Thus, annular flow
would be expected above Land below B. Of course, dispersion to
bubbly flow can be expected as a result of the turbulence breakup
mechanism that results in curve F.
10r-----------------~--~
1.0
6.6. L {,6. LI:'. 6. I::. 6. 6. 6.1::. t:. t:.t:. t:.
U • • • • • • • • • • • • • • t:.t:.t:.
cu • • L. • .t:. t:.
.....
(/)
• • • llll
E 0.1 • • • a:
Ul
..J
«..J
::::> STRATIFIED WAVY ::J
z
• • «z
0.01 • II
A•
0.001 • ••
0..01 0.1 1.0 10. 10.0.
UGS (rn/seC>
Fig. 51. Flow-pattern map for air-water in a 5. I-ern-diameter pipe

at 80 0 downward inclination. Solid lines represent theory.
1.0 " ~ ~~:66ßßß • • • • B
~
, L~.:;;~",
.
<..>
<lJ ••••
'"
......
E ""
0.1 • • • • • • • ."" "
J STRAT1FIED WAVY A : ~ ~
0.01'1~: •
•
•
•
•
•
•
•
•
••
• \.
uUß
i~ i 1
1
0001 i__---'-__----'--__--'-__-'.
0.01 0.1 1.0 10 100
U GS (rn/sec)
Fig. 52. Flow-pattern map for air-water in a 2.5-em-diameter pipe

at 70 0 downward inclination. Solid lines represent theory.
At lower va lues of ß, the stratified-to-annular transition

disappears at low gas rates, and the transition takes place directly
from stratified to the dispersed bubbly pattern, as shown in Fig. 53.
10 TRANSITION MECHANISMS IN TUBES: AN OVE RV I Ei-l
Various mechanisms have been presented to explain the physical

basis for the observed transitions between flow patterns. These
10r-----------------~----~
DISPERSED BUBBLE G
U I •
•
I·
(\)
•
li-
~
E 0 D •
(f)
-1
::> I STRATIFIED WAVY
0.01~. • •
I
0.001
001 0.1 1.0 10 100
UGS (rn/sec)
Fig. 53. Flow-pattern map for air-water in a 5.l-em-diameter pipe

at 30 0 downward inelination. Solid lines represent theory.
mechanisms along with cognate relationships have been used to

arrive at equations that can be used to find the transition
boundaries in Ucs-Urs space once the properties, pipe size,
inclination angle, and flow direction are specified. Table 1
summarizes the various mechanisms discussed. In each case the
letters have been used to identify the theoretical curves on each
of the many graphs that have been presented comparing theory and
data. Aseries of sketches appear in Fig. 54 that show the
relative location of these curves and how the transition boundaries
change in shape and position as the inclination or flow direction
varies.
A few important points need to be made emphatically at this

point. The matter of predicting flow-pattern transitions is in a
developing state. In some cases the transition mechanisms are
reliably understood. In others the mechanisms suggested here are
more in the nature of speculations. Even when the mechanism is
understood, arriving at useful equations for predicting transition
boundaries is not a direct process and sometimes includes severe
approximations of its own. Thus, for example, the trajectory
mechanism for transition L seems quite correct, but the final
useful equation requires an estimate of the fluctuating velocity
ne ar the interface, which can only be approximated at this time.
Furthermore, only preliminary rules exist on how to decide which
mechanism is operative when both can apply, as in the case of an
inclined tube.
If the ability to predict these transition conditions is to

become more reliable and more generalized, it will require
considerable additional research into mechanisms and the basic
fluid mechanical factors that enter into them.
11 VERTICAL ROD BUNDLES
Very few studies have appeared dealing with flow-pattern

transitions in a geometry different from a round pipe. One of
these, flow in rod bundles, is of interest to the nuclear plant
industry and to those concerned with heat exchanger design and
performance. Bergles et al. (1968) studied two-phase flow of
steam and water in 4-rod bundles, arranged in a square array.
Williams and Peterson (1978) used the same fluids in a 4-rod
linear array. Venkateswararao et al. (1982) studied the
transitions in a 24-rod circular bundle with the air-water system.
The cross section they used is shown in Fig. 55. The flow patterns
they observed were bubbly, slug, churn, and annular. Two types of
slug flow were shown to exist, as illustrated in Fig. 56: {al flow
domina ted by large shroud-type Taylor bubbles whose cap is
penetrated by many rods and (bl flow dominated by cell-type Taylor
bubbles that occupy the space in a 4-rod cell and whose caps are not
penetrated. In general, shroud-type Taylor bubbles occur only
when there is a sudden increase in the gas flow rate. Patterns
observed over the operable range of flow rates for the equipment
used are shown in Fig. 57, where the solid curves represent the
experimentally determined boundaries. The approaches to modeling
these transitions presented below follow those of Venkateswarar
etal. (1982).
Table 1. A Summary of Mechanisms for Flow in Tubes
Transition
Boundary Patterns Mechanism Equation
A Stratified to Kelvin-Helmholtz instability 29

nonstratified
B Intermittent to h;D : 0.35 and A

annular
C Stratified smooth Jeffreys, wind-wave 33

to wavy interactions
D Intermittent to Turbulent fluctuations 39

dispersed bubble versus buoyant forces
E Bubble to Cl ?: 0.25 79 (vertical up)

Intermittent 104 (vertical down)
109 (inclined up)
F Intermittent to Turbulent fluctuations 85 or 85

dispersed bubble versus surface forces
G Dispersed bubble CI. ?: 0.52

to slug or churn
H Slug to churn Entry length to develop 95

stable slug
J Annular to slug Minimum gas velocity to 100 (vertical)

or churn lift largest drop 110 (inclined)
K Stratified smooth 112

to wavy (inclined)
L Stratified wavy to Trajectory of drops 116

annular torn from liquid film
Minimum inclination Lift versus buoyant 108

angle to show forces
bubble flow
11.1 Bubble Rise Velocities
Development of models for transitions requires knowledge of the

rise velocities of the dispersed as well as the Taylor bubbles. No
data exist in the literature and, therefore, it was necessary to make
new measurements. The result for single dispersed bubbles rising in a
stationary liquid appears in Fig. 58. It is seen that the presence
of the rod surfaces substantially alters the trend of the data from
that observed for single bubbles rising in an infinite liquid. When
the liquid velocity is nonzero the bubble rise velocity is usually
given as Uc = AULS + Uo- In these experiments A was shown to be
1.0 for low-density swarms of bubbles.
[Q[] [OO-J C6m G H [Q1j] /
/
_D/ C
[IJ
0---
A~Gill
B o:J B
7 ~
lffi:J -.... J
[ill C'@SW'1
b ~I~ [§JE
gs
I HORIZONTAL 2A. SMALL ß-UP 3A. LARGE ß-UP 4A. VERTICAL UP

A. UPWARD FLOW
[Qill [OOJ ~G f:! -:..-~

LQE3J /Go:::J
F-2[lJ
_ _B ' / f-2 E -B-
0------ _D--- r-r' /
_AL:-..J B []ill .- -I
!TI B/ ~
A~ f2lli] ;___ K~1iliJ LI [AtU

A
C A [Sw]
[]]J \'§iN ~\ \
I HORIZONTAL 28. SMALL ß-DOWN 38 LARGE ß-DOWN 48. VERTICAL DOWN
8 DOWNWARD FLOW
00 DISPERSED SUSSLE [ ] INTERMITTENT [1[] ANNULAR []I] STRATIFIED SMOOTH

~ STRATIFIED WAVY ßB]CHURN
Fig. 54. A schematic view of the relative location of transient mechanisms .

....
(j)
I. D. OF SHROUD = 8.89 cm
O. D. OF ROD I. 27 cm
PITCH = I. 75cm
Fig. 55. Cross section of rod bundle. From Venkateswararo et al.

00000
oo~oo
LARGE. "SHROUD" TAYLOR BUBBLE SMALL."CELL" TAYLOR BUBBLES

A B
Fig. 56. Two types of Taylor bubbles. From Venkateswararo et al.

Data on the rise velocity of shroud-type Taylor bubbles

appear in Fig. 59. These are compared with the rise velocity in a
tube and with the da ta of Grace and Harrison (1967) where Taylor
bubbles penetrated by one rod were studied. They showed that the
rise velocity depended on the bubble volume VB' Clearly, the
result is different here. Cell-type Taylor bubbles rose at a
remarkably constant rate of 24 ± 1 em/s.
.... . .. ....
1.0
I T
....
QQ..
..
.
0.50 T Q
T Q" QQ Iiilii Q
T QQ Q QQQ Iii
u
'"'"
'" QQ QiI Q QQ Q
"- 0.10 T 'V 'V TT QQ QiI Q QQ Q

E 'V T T 'VT QQ Q QQQ
~(/)
-' 0.05 T QQ Q QQQ

:J
'V Q Q Q QiI Q QQ
T T 'V VV
••• Iii Q Q Q Q Q QiI QQQ
.01 .05 .10 .50 1.0 5.0 10 50

UGS (rn/sec)
Fig. 57. Experimentally observed flow patterns. See Fig. 34 for

legend. From Venkateswararo et al. (1982). Reprinted
with permission.
- - - RISE VELOCITY OF SINGLE BUBBLES IN INFINITE

I
MEDIUM
~ - - RISE VELOCITY OF SINGLE BUB8LES IN A 4 ROD
40 CELL FROM EXPERIMENT
(j
QJ
'"
"-
E
30 l
I
.... "-
..... ,
_-
.....
-- --
(j
~
>-
r-
Ü 25
0
..J
w 20
>
w
(f) I
a::: 15~
I
10 1 I I I
2 3 4 5 6 7 8
DIAMETER mm
Fig. 58. Rise velocity of single bubbles in a 4-rod cello

Stationary liquid. From Venkateswararo et al. (1982).
CD DUMITRESCU/TAYLOR IN AN EMPTY TUBE" 0.35 (gDl ll2
® GRACE/HARRISON (1967) IN A TUBE WITH 1/2 1/6

70 ONE ROD PENETRATING THE BUBSLE "0.79 9 VB
0
®
(I)
EXPERIMENTS IN ROD BUNDLE: RUN # I
'"
...... 60
E
0
@ EXPERIMENTS IN ROD BUNDLE· RUN # 2
>-
I-
ü
g
50
.... -
w
> 40
w
(f)
ä: --------------------Q)
30
10 20
BUSSLE VOLUME ce
Fig. 59. Rise velocity of shroud Taylor bubbles compared with
other cases. From Venkateswararo et al. (1982).
11.2 Bubble-to-Slug Transition
The basic mechanism for transition appears to be that

described in Sec. 8. As the gas rate is increased for a given
liquid rate, the bubble density increases, many collisions occur,
cell-type Taylor bubbles are formed, and the transition to slug
flow takes place. We speculate that this transition takes place
when a = 0.25, in a manner similar to that done for vertical pipe
flow in Sec. 8. However, observation show that few bubbles move
in the space between rods but mi grate to the open area that exists
within an array of 4 rods. Thus, in Fig. 60 bubbles are seldom
observed in the space along the line A-B but instead are
concentrated in the space designated by the circle C. Therefore,
it is suggested that the transition to slug flow takes place when
the local voids ar in the enclosed circle C exceed 0.25. The
average void a is-related to aL by
[12 (pldl - 1J 2
2
(116 )
4/rr (pldl - 1
where p is the pitch and d is the rod dia~eter.
The relationship that gives the locus of ULS - Ues points for
the transition in Eq. (76). However, the value of a to be used at
transition is given by Eq. (116). With aL = 0.25, p = 1. 75 cm, and
d = 1.27 cm, as in these experiments, a = 0.16, and the transition
equation becomes
Fig. 60. A 4-rod cello From Venkateswararo et al. (1982).

(117 )
The theoretical curve resulting from this equation is marked A and

is dashed in Fig. 61, where it is seen to be in good agreement with
experiment.
5.00 ,-----,,-_ _ _, - _ , -_ _ _,-----,_ _--;_,.-_,--;-_---,
OISPERSED 8UB8LE C/ :
--8----_ - - - B // I
7'_ C I
A ...... v
1.00 / I
/
o I
0.50
BU8BLE CHURN
.,
<)
E
'"
...... 0::
<t
-l
Ü, :::>
z
:J-' 0.10 z
<t
0.05
EXPERIMENT
- - - - - THEORY
0.01
0.01 0.05 0.10 0.50 10 5.0 10 50
UGS,m/sec
Fig. 61. Comparison of theoretical and experimental transitions.
From Venkateswararo et al. (1982). Reprinted with
permission.
At high liquid rates, forees due to turbulenee are expeeted to

eause bubble and/or slug breakup similar to that already diseussed
for flow in round tubes, which resulted in Eq. (85). This result
is shown as the dashed curved C in Fig. 61 and terminates on curve
C along whieh a is eonstant at a value of 0.52. Limitation in
liquid pumping eapacity made it impossible to test this transition
in the experiments plotted in Fig. 57.
11.3 Slug-to-Churn Transition
In a vertical tube this transition is attributed to Taylor

bubbles overtaking the entry region, as discussed in Sec. 8.12.
However, in a rod bundle the mechanism appears to be quite
different.
As the gas rate is increased from slug flow, the size of cell-
type Taylor bubbles increases so that they occupy more of the free
space of each cello At the same time the number of cells occupied
by Taylor bubbles increases, and eventually the concentration of
occupied cells is great enough to cause coalescence. When this
happens the liquid being supported by the bubbles suddenly falls
downward in a lump, is mixed with the liquid below, and starts its
ascent again. This random falling of lumps of liquid is the
definition of the churn flow pattern. Thus, in a rod bundle the
transition to churn flow is a coalescenee phenomenon, and this can
be expected to take place at the lowest void fraction at which the
Taylor bubble density results in contaet between bubbles in
adjaeent eells. If one examines the loeation of bubble pairs
possible in the rod bundle, eonsidering an array of eells in the
horizontal plane and aseries of bubbles in the vertieal direetion,
the minimum voids at which contact can be made are given by the
configuration shown in Fig. 62. The average void Eraction
eorresponding to this configuration is given by
1f(p + d) (118 )
6 eos 8(2p + d)
when
e = arcsin p - d
p + d
In these experiments with p = 1.75 cm and d = 1.27 em, aT (at which
transition to churn flow should first be observed) is 0.335. Now
it is possible to use Eq. (76), with Ua being the rise veloeity of
the eell-type Taylor bubbles, to caleulate the relations hip between
UGS and ULS for this transition. The result appears in Fig. 61 as
the dashed eurve marked D and is shown to be in reasonable
agreement with the experimentally observed transition shown as solid
curve D.
11.4 Churn-to-Annular Transition
At high gas rates the flow pattern in the rod bundle becomes
one of annular film flow. The liquid flows upward along all rod
surfaces as thin annular films with gas flowing in the rest of
the free area. The liquid interface is highly wavy, and the gas
carries entrained liquid drops torn from the liquid. For flow in
a pipe the mechanism for this transition was related to the minimum
upper
bubble
+I
I
"C
Cl. +
Cl.
II
-*- i
I
I
!
Fig. 62. Configuration of cell-type Taylor bubbles for lowest

voids at transition to the churn flow pattern. From
Venkateswararo et al. (1982). Reprinted with permission.
gas velocity necessary to trru,sport the largest drop upward, as

discussed in Sec. 8.1.3. At gas velocities less than this value,
liquid begins to fall back, accumulate, and bridge the pipe, only
to be thrust upward again, and thus the alternating motion
characteristic of churn flow is observed. This mechanism can be
expected to apply when the minimum velocity necessary to lift the
film is less than that required to lift the largest drop.
Equation (100) gives the theoretical result that is shown as dashed
line E-l in Fig. 61.
At low liquid rates where the film is very thin, the waves
that are the source of the drops are suppressed. Thus, liquid
entrainment can no longer be controlling, and the mechanism must
change. At this condition of low liquid rate, another mechanism
comes into play. Fernandes (1981) has shown that in the churn flow
region a simple holdup model for voids in slug flow continues to be
valid
(119)
from which
(120 )
As the gas rate is increased, the pattern changes to annular

and the void is given by the equation
p2 _ (rrl 4) (d + 2h)2
Ci. (121 )
a p2 (rr/4)d2
where h is the film thickness. We speculate that as the gas rate

is increased, the condition at which iJ.ch first equals iJ. a is the
condition at which transition takes place. An equation for CL a can
be developed by finding the film thickness h, and this can be done
by applying a force balance on the liquid film in a 4-rod cell
similar to that done in Sec. 8.2.1 for film flow in a pipe. The
development is presented in detail by Venkateswararao et al.
(1982), and the resulting equation for iJ. a is
0.0025 2 0.0025 [1 + 2 g ( iJ. )'J

(I-CL a ) (Pr-p.r.)g + 2 p_U r , .. n
"LI Cd(l-iJ.) L "0 CdiJ. 3 --
a a
where
-"4rrJ
1
c rr [ (pld) 2
To find the condition at transition ,le find the intersection

of the two equations where ach = aa' and this defines the locus of
the (UCc:,U LS ) pairs at which transition takes place. Setting
(p,d) and the fluid properties to meet the conditions of these
experiments results in the prediction shown by the dotted curve
E-2 in Fig. 61. It is seen that the experimental data (solid
curve E) are bracketed by the two models. As expected, the
agreement is best at high liquid rates with the entrainment model
and at low rates with the void fraction matching model.
The transitions predicted by these models were compared with

the da ta of Williams and Peterson (1978) and Bergles et al. (1968).
In both cases the boiling steam-water system. Williams and
Peterson reported data at three pressure levels (2.8 x 10 6 ,
8.3 x 10 6 , and 13.8 x 10 6 N/m 2 ) in a rod linear array, while
Bergles et al. operated at apressure of 6.9 x 10 6 N/m 2 in a
square array. As shown by Venkateswararao et al. (1982), the
agreement is satisfactory.
NOMENCLl'.TURE
A pipe cross-sectional flow areai annular flow pattern

AD annular-dispersed flow pattern
C wave velocity; constant in the friction factor correlation
Cd drag coefficient
d bubble drop; rod diameter
D pipe diameter; hydraulic diameter
DB dispersed bubble pattern flow
E voltage
f friction factor; frequency
F modified Froude number [Eq. (30)J
g acceleration of gravity
h liquid level; gas height; film thickness
I intermittent (slug and plug) flow pattern
k constant [Eq. (80)J
K dimensionless wavy flow parameter [Eq. (34)J; Weber number
I length
L length of pipe
m exponent [Eq. (8)]
m mass flux
n exponent [Eq. (8)J
p pitch
P pressure
q heat flux
Q dimensionless heat flux parameter [Eq. (68)J
R autocorrelation function; electrical resistance
Re Reynolds number
s Jeffrey's sheltering coefficient
5 perimeter; spectral density; stratified flow pattern
55 stratified smooth flow pattern
sw stratified wavy flow pattern
time
T dimensionless dispersed bubble flow parameter [Eq. (40)J
U velocity in the x-direction
V velocity normal to the x-direction
w maximum radial height of a particle torn from the surface of a
stratified liquid
W mass flow rate
x coordinate in the downstream direction
X Lockhart-Martinelli parameter [Eq. (11)1

Y dimensionless inclination parameter [Eq. (12)J
Greek Letters
a g a s holdup
ß angle of inclination, positive for downward flow
y shape factor for a bubble
ß small increment
E pipe roughness; energy dissipation per unit mass
A latent heat of vaporization
~ dynamic viscosity
v kinematic viscosity
p density
o interfacial tension
T shear stress; time [Eq. (2) J
~ liquid holdup
Subscripts and Superscripts
c center
E entry length
[ film
G gas
i interface
L liquid
M mixt ure
s stable slug
S superficial
V vapor
w pipe surface
dimensionless
fluctuationi derivative
* frictional velocity
REFERENCES
Agrawal, S. S., G. A. Gregory, and G. W. Govier 1973, An Analysis

of Horizontal Stratified Two-Phase Flow in Pipes. Can. J. Chern.
Eng. vol. 51, p. 280.
Akagawa, K., and T. Sakaguchi 1966, Fluctuation of Void Ratio in

Two-Phase Flow. Bull. ASME, vol. 9, p. 104.
Albertson, M. L., J. R. Barton, and D. B. Simons 1966, Fluid
Mechanics tor Engineers. Englewood Cliffs, NJ: Prentice Hall.
Baker, O. 1954, Simultaneous Flow of Oil and Gas. air Gas J. vol.
53, p. 185.
Baker, O. 1958, Multiphase Flow in Pipelines 1958, Piperine News,

June, 23.
Barnea, 0., 0 Shoham, and Y. Taitel 1980, Flow Pattern

Characterization in Two-Phase Flow by Electrical Conductance Probe.
Int. J. MuZtiphase Flow ~ol. 6, p. 387.
Barnea, D., O. Shoham, and Y. Taitel 1982, Flow Pattern Transition

for Downward Inclined Two-Phase Flow; Horizontal to Vertical.
ehem. Eng. Sei. vol. 37, p. 735.
Barnea, D., O. Shoham, and Y. Taitel 1982b, Flow Pattern Transition

for Vertical Downward Two-Phase Flow. ehem. Eng. Sei. vol. 37,
p. 741.
Bergles, A. E., and M. Suo 1966, Investigation of Boiling-Water

Flow Regimes at High Pressure. Dynatech Report No. NYO-3304-8,
HTFS 1909.
Bergles, A. E., J. P. Roos, and J. G. Bourne 1968, Investigation of

Boiling Flow Regimes and Critical Heat Flux. AEC Report NYO-3304-13,
Dynatech Corporation, Can\bridge, tJ!a.
Breber, G. J., W. Palen, and J. Taborek 1980, Prediction of

Horizontal Tube-side Condensation of Pure Components Using F10w
Regime Criteria. J. Heat Transfer, vol. 102, p. 471.
Brock, R. R. 1970, Periodic Permanent Roll Waves. Proe. Am. Soe.

eiv. Eng. vol. 96, HYD 12, p. 2565.
Brodkey, R. S. 1967, The Phenomena of FLuid Motion. Reading, MA:

Addison-\qesley.
Butterworth, D. 1972, A Visual Study of Mechanism in Horizontal Air-

Water Flow. Atomic Energy Research Establishment Report M2556,
Harwell, England.
Calder, A. C. 1976, Flow-Regime Characterization for Horizontal

Two-Phase Steam Flow. University of California, Lawrence Livermore
Laboratory, VCRL-52186.
Calderbank, P. H. 1958, Physical Rate Processes in Industrial

Fermentation. I. The Interfacial Area in Gas Area in Gas-Liquid
Contracting with Mechanical Agitation. Trans. Inst. ehem. Eng.
vol. 36, p. 443.
Chaudry, A. B., A. C. Emerton, and R. Jackson 1965, Flow Regimes in

the Concurrent Upward Flow of Water and Air. Paper presented at
the Symposium on Two-Phase Flow, Exeter, England.
ehu, K. T. 1973, Statistical Characteristics and Modelling of Wavy

Liquid Film in Vertical Two-Phase Flow. Ph.D. thesis, University
of Houston.
Clay, P. H. 1950, The Mechanism of Emulsion Formation in Turbulent

Flow, Part I. Proe. R. Soe. London Sero A, vol. 200, p. 375.
Collins, R., F. F. DeMoraes, J. F. Davidson, and D. Harrison 1978,

The Motion of a Large Gas Bubble Rising through Liquid Flowing in
a Tube. J. Fluid Mech. vol. 89, p. 497.
Davis, E. J., and M. M. David 1961, Heat Transfer to High-Quality

Steam-Water JYlixtures Flowing in a Horizontal Rectangular Duct.
Can. J. Chem. Eng. vol. 39, p. 99.
Derbyshire, R. T. P., G. F. Hewitt, and B. Nicholls 1969,

X-Radiography of Two-Phase Gas-Liquid Flow. Atomic Energy Research
Establishment Report M-1321; Harwell, England.
Dukler, A. E. 1969, Gas Liquid Flow in Pipelines. I. Research

Results. New York: Amer-ican Gas Association.
Dukler, A. E., and M. G. Hubbard 1975, A Model for Gas-Liquid Slug

Flow in Horizontal and Near-Horizontal Tubes. Ing. Eng. Chem.
Fundamentals vol. 14, p. 337.
Dukler, A. E., M. Wicks, and R. G. Cleveland 1964, Frictional

Pressure Drop in Two-Phase Flow. A. A Comparison of Existing
Correlations for Pressure Loss and Holdup. AIChEJ. vol. 10, p. 38.
Dukler, A. E., M. Wicks, and R. G. Cleveland 1964, Friction

Pressure Drop in Two-Phase Flow. B. An Approach through Similarity
Analysis. AIChEJ. vol. 10, p. 44.
Duns, Jr,.H., and N. C. J. Ros 1963, Vertical Flow of Gas and Liquid
Mixtures from Boreholes. Proc. 6th WOY'ld Petroleum Congress,
Frankfurt.
Fernandes, R. 1981, Experimental and Theoretical Studies of

Isothermal Upwards Gas-Liquid Flow in Vertical Tubes. Ph.D. thesis,
University of Houston.
Fiori, M. P., and A. E. Bergles 1966, A Study of Boiling Water Flow

Regimes at Low Pressure. Report 5382-40, Department of Mechanical
Engineering, MIT.
Fulford, G. D. 1964, The Flow of Liquids in Thin Films. Adv. Chem.

Eng. vol. 5, p. 151.
Gazley, C. 1949, Interfacial Shear and Stability in Two-Phase Flow.

Ph.D. thesis, University of Delaware, Newark.
Golan, L. P., and A. H. Stenning 1969, Two-Phase Vertical Flow

Maps. PY'oc. Inst. Mech. Eng. vol. 184, p. 108.
Gould, T. L. 1974, Vertical Two-Phase Flow Steam-Water Flow in

Geothermal Wells. J. Pet. TechnoL vol. 26, p. 833.
Govier, G. W., and K. Aziz 1972, The Flow of Complex Mixtures in

Pipes. New York: Van Nostrand Reinhold.
Govier, G. W., B. A. Radford, and J. Sc. Dunn 1957, The Upward

Vertical Flow of Air-Water Mixtures: I. Effect of Air and Water
Rates on Flow Pattern Holdup and Pressure Drop. Can J. Chern. Eng.
vol. 35, pp. 58-70.
Grace, J. R., and D. Harrison 1967, Influence of Bubble Shape on

the Rising Velocities of Large Bubbles. Chem. Eng. Sei. vol. 22,
p. 1337.
Griffith, P., and G. B. Wallis 1961, Two-Phase Slug Flow. J. Beat

Transfer vol. 83, p. 307.
Griffith, P., and J. Snyder 1964, The Bubbly-Slug Transition in a
High-Velocity Two-Phase Flow. HIT Report 5003-29.
Haberstrah, R. E., and P. Griffith 1965, The Slug Annular Two-Phase

Flow Regime Transition. ASivJE Paper 65HT-52.
Harmathy, T. Z. 1960, Velocity of Large Drops and Bubbles in Hedia

of Infinite or Restricted Extent. AIChE J. vol. 6, p. 281.
Hewitt, G. F., and N. S. Hall-Taylor 1970, AnnuZar Two-Phase Flow.

New York: Pergamon.
Hewitt, G. F. 1978, Measurement oi Two-Phase FZow Parameters. New

York: Academic.
Hewitt, G. F., and D. N. Roberts 1969, Studies of Two-Phase Flow

Patterns by Simultaneous X-Rays and Flash Photography. Atomic
Energy Research Establishment Report M-2159; Harwell, England.
Hinze, J. O. 1955, Fund&~entals of the Hydrodynamic Hechanism of

Splitting in Dispersion Processes. AIChE J. vol. I, p. 289.
Hoogendorn, C. J., and A. A. Buitelaar 1961, The Effect of Gas

Density and Gradual Vaporization on Gas-Liquid Flow in Horizontal
Pipes. Chern. Eng. Sei. vol. 16, p. 208.
Hsu, Y. Y., and R. W. Graham 1963, A Visual Study of Two-Phase Flow

in a Vertical Tube with Heat Additi.on. NASA Technical Note D-1564.
Hsu, Y. Y., F. F. Simon, and R. W. Graham 1963, Application of Hot-

Wire Anemometry for Two-Phase Flow Measurements. Paper presented
at the ASHE Winter Meeting, Philadelphia, PA.
hubbard, H. G., and A. E. Dukler 1966, The Characterization of Flow

Regimes for Horizontal Two-Phase Flow. Proeeedings of the Beat
Transfer and Fluid Meehanies Institute~ (Saad, M. A. and J. A. Moller,
eds.), Stanford: Stanford University Press.
Isbin, H. S., R. H. Moen, R. O. Wickey, D. R. Hosher, and H. C.

Larson 1959, Two-Phase Steam-Water Pressure Drop. Chem. Eng. Symp.
Sero vol. 55, no. 23, p. 75.
Jeffreys, J. 1925, On the Formation of Water Waves by Wind. Proe.
R. Soe. London vol. AI07, p. 189.
Jeffreys, H. 1926, On the Formation of Waves by Wind. Proe. R. Soe.

London vol. AIOO, p. 241.
Jones, O,C. r and N Zuber 1975, The Interrelation between
Void Fraction Fluctuations and Flow Pattern in Two-Phase Flow.
Int. J. Multiphase FZow vol. 2, p. 273.
Kordyban, E. S., and T. Ranov 1970, Mechanism of Slug Formation in

Horizontal Two-Phase Flow. J. Basic Eng. Sero D. vol. 92, No. 4,
p. 85.
Levich, V. G. 1962, Physicochemical Bydrodynamics. Englewood Cliffs,

NJ: Prentice-Hall.
Lockhart, R. W., and R. C. Martinelli 1949, Proposed Correlation of

Data for Isothermal Two-Phase, Two-Component Flow in Pipes. Chem.
Eng. Prog. vol. 45, p. 39.
Mandhane J. M., G. A. Gregory, and K. Aziz 1974, A Flow Pattern

Map for Gas-Liquid Flow in Horizontal Pipes. Int. J. Multiphase
FloU! vol. I, p. 537.
Martin, C. S. 1973, Transition from Bubbly to Slug Flow of a

Vertically Downward Air-Water Flow. Proc . .4SME Symposium, Atlanta.
Milne-Thomson, L. M. 1960, Theoretical Bydrodynamics. New York:

Macmillan.
Miyagi, O. 1925, On Air Bubbles Rising in Water. Phil. Mag. vol.

50, No. 295, p. 112.
Moissis, R., and P. Griffith 1962, Entrance Effects in a Two-Phase

Slug Flow. J. Beat Transfer vol. 84, p. 29.
Nickiin, D. J., J. O. ~ülkes, and J. F. Davidson 1962, Two-Phase

Flow in Vertical Tubes. Trans. Inst. Chem. Eng. vol. 40, p. 61.
Oshinowo, T. and M. E. Charles 1974, Vertical Two-Phase Flow. 11.

Holdup and Pressure Drop. Can. J. Chem. Eng. vol. 56, p. 438.
Palen, J. W., G. Breber, and J. Taborek 1977, Prediction of Flow

Regimes in Horizontal Tube-Side Condensation. Paper presented at
17th National Beat Transfer Conf. AIChE/ASME, Salt Lake City.
Patel, R. P. 1971, Turbulent Jets and Wall Jets in Uniform

Streaming Flow. Aeronaut. Q. vol. 23, p. 311.
Pushkina, O. L., and Y. L. Sorokin 1962, Breakdown of Liquid Film

Motion in Vertical Tubes. Beat Trans. Sov. Res. vol. I, p. 151.
Radovcich, N. A. and R. Moissis 1962, The Transition from Two-Phase

Bubble Flow to Slug Flow. MIT Report 7-7673-22.
Raisson, C. 1965 Flow Regime Studies up to Critical Heat Flux

Conditions at 80 kg/m 2 . CEA Grenoble, Report No. TT22.
Schlichting, H. 1968, Boundary Layer Theory. New York: McGraw-Hill.
Sevik, M., and S. H. Park 1973, The Splitting of Drops and Bubbles
by Turbulent Fluid Flow. J. Fluid Eng. vol. 95, p. 53.
Shah, M. M. 1975, Visual Observations in an Ammonia Evaporator.

ASHRAE Trans. vol. 81, Part I, p. 295.
Shoham, O. 1983, Flow Pattern Transitions and Characterization in

Gas-Liquid Two-Phase Flow in Inclined Pipes. Ph.D. dissertation,
Tel Aviv University.
Soliman, H. M., and N. Z. Azer 1971, Flow Patterns during

Condensation inside a Horizontal Tube. ASHRAE Trans. vol. 77,
p. 210.
Solomon, J. V. 1962, Construction of a Two-Phase Flow Regime

Transition Detector. M. Sc. thesis, Mechanical Engineering
Departrnent, MIT.
Spedding, P. L., and V. T. Nguyen 1980, Regime Maps for Air-Water

Two-Phase Flow. Chern. Eng. Sei. vol. 35, p. 779.
Sternling, V. C., 1965, Two-Phase Flow Theory and Engineering

Decisions. Award lecture presented at AIChE annual meeting.
Stewart, R. W. 1967, Mechanics of the Air-Sea Interface; Boundary

Layers and Turbulence. Phys. F~uids vol. 10, p. 547.
Stoker, J. J. 1957, Wa~er Waves. New York: Interscience.
Streeter, V. L. 1961, Handbook of F~uid Dynamies. New York:

McGraw-Hill
Taitel, Y. 1977, Flow Pattern Transition in Rough Pipes. Int. J.

Mu~tiphase Flow vol. 3, p. 597.
Taitel, Y. and A. E. Dukler 1976, A Model for Predicting Flow

Regime Transitions in Horizontal and Near-Horizontal Gas-Liquid
Flow. AIChE J. vol. 22, p. 47.
Taitel, Y., and A. E. Dukler 1977, A Model for Slug Frequency

during Gas Liquid Flow in Horizontal and Near-Horizontal Pipes.
Int. J. Multiphase F~ow vol. 3, p. 585.
Taitel, Y., D. Barnea, and A. E. Dukler 1980, Modeling Flow Pattern

Transitions for Steady Upward Gas-Liquid Flow in Vertical Tubes.
AIChE J. vol. 26, p. 345.
Taitel, Y., N. Lee, andA. E. Dukler 1978, Transient Gas-Liquid Flow

in Horizontal Pipes-Modeling Flow Pattern Transitions. AIChE J.
vo 1. 24, p. 920.
Tong, L. S. 1965, Boi~ing Heat Transfer and Two-Phase Flow. New

York: Wiley.
Travis, D. P., and W. M. Rohsenow 1973, Flow Regimes in Horizontal

Two-Phase Flow with Condensation. ASHRAE-Trans. vol. 73, p. 31.
Turner, R. G., M. G. Hubbard, and A. E. Dukler 1969, Analysis and
Prediction of Minimum Flow Rate for the Continuous Removal of
Liquid from Gas Wells. J. Pet. Tecnol. vol. 21, p. 1475.
Tutu, N. K., 1982, Pressure Fluctuations and Flow Pattern

Recognition in Vertical Two-Phase Gas-Liquid Flows. Int. J.
Mu~tiphase Flow vol. 8, p. 443.
Venkateswararao, P., R. Semiat, and A. E. Dukler 1982, Flow Pattern

Transition for Gas-Liquid Flow in a Vertical Rod Bundle. Int. J.
Multiphase Flow vol. 8, p. 509.
Wallis, G. B. 1969, One-Dimensional Two-Phase Flow. New York:

McGraw-Hill
Wallis, G. B., and J. E. Dobson 1973, The Onset of Slugging in

Horizontal Stratified Air-Water Flow. Int. J. Multiphase Flow
vol. I, p. 173.
Weisman, J., D. Duncan, J. Gibson, and T. Crawford 1979, Effect of

Fluid Properties and Pipe Diameter on Two-Phase Flow Pattern in
Horizontal Lines. Int. J. MuZtiphase Flow vol. 5, p. 437.
Weisman, J., and S. Y. Kang 1981, Flow Pattern Transitions in

Vertical and Upwardly Inclined Lines. Int. J. Multiphase Flow
vol. 7, p. 271.
Williams, C. L., and A. C. Peterson, Jr. 1978, Two-Phase Flow

Pattern with High-Pressure Water in a Heated Four-Rod Bundle.
Nucl. Sei. Eng. vol. 68, p. 155.
Zahn, W. R. 1964, A Visual Study of Two-Phase Flow while

Evaporating in Horizontal Tubes. J. Beat Transfer vol. 86c, p. 417.
Zuber, N., and J. Hench 1962, Steady-State and Transient Void

Fractions of Bubbling Systems and their Operating Limit. Steady-
State Operation. General Electric Co. Report 62GL100.

Chapter01 - Flow Pattern Transitions in Gas-Liquid Systems - Measurement and Modeling

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter01 - Flow Pattern Transitions in Gas-Liquid Systems - Measurement and Modeling

Uploaded by

Copyright:

Available Formats

Chapter 1

Flow Pattern Transitions in Gas-liquid

The basic dilemma is quite clear. In gas-liquid flow the two

Ever since the earliest visual observations of two-phase flow,

This chapter is dedicated to Ovid Baker, with whom it all started.

G. F. Hewitt et al. (eds.), Multiphase Science and Technology

It seems reasonable that this should be so. Not only pressure

If the models created for individual flow patterns are to be

Suitably formulated models incorporate in a natural way the

predictive methods discussed he re are based largely on the research

2 CLASSIFICATION OF FLOW PATTERNS

The first problem, of course, is to decide what constitutes a

Horizontal tubes: Stratified, stratified smooth, stratified

Vertical tubes: Bubble, dispersed bubble, homogeneous, slug,

A new observer with an expanded vocabulary could likely arrive

In recent years there seems to nave evolved an understanding

2.1 Horizontal Pipes: Fig. 1

In stratified flow, liquid flows along the bottom of the pipe

Fig. 1. Flow patterns in horizontal flow.

In the intermittent pattern the inventory of liquid in the pipe

In annular flow the liquid exists as a continuous film around

2.1.4 Dispersed Bubble

In dispersed bubble flow the gas phase is distributed as

2.2 Vertical Pipes: Fig. 2

2.2.1 Bubble Flow

In bubble flow the gas phase is distributed more or less

2.2.2 Slug Flow

In the slug flow pattern the two fluids redistribute axially

Successive Taylor bubbles are separated in the axial direction by

2.2.3 Churn Flow

Churn flow is somewhat similar to slug flow. It is, however,

2.2.4 Annular Flow

Annular flow is characterized by continuity in the axial

The flow patterns described above appear to exist when the

for the slug flow pattern in Fig. 2, ehanges to the appearanee of

There have been few systematie studies for noneireular

The problem of flow-pattern deteetion, or determining whieh

3.1 Visual Methods

The simplest method for detecting flow patterns is by visual

At high flow rates photographic methods are useful, as

Moving-picture photography has proved to be less successful

3.2 Methods Based on Pressure Measurement

Early attempts to use pressure to characterize flow pattern

Hubbard and Dukler (1966) suggested a method based on the

where P(t) is the measured time-dependent pressure, P the time

computer program packages now exist whereby these spectra can be

Measurements of this type showed that, despite a wide range of

Fig. 3. Power spectral density of wall pressure fluctuations.

and represented the mean frequency of slug passage. Dispersed or

In practical situations, what one usually observed is a

The original method did not discriminate clearly, however,

For vertical systems, a method proposed by Tutu (1982) seems

Fig. 4. Superposed spectra.

hydrostatic gradients or the void fractions, that is, for bubbly

Fig. 5. Probability density function of pressure difference.

3.3 Methods Based on Photon Attenuation

While the pressure methods discussed above can be used to

3.4 Methods Based on Electrical Conductivity

For electrically conducting liquids, such as water, conductance

The basic probe design for application to horizontal and

FROM FROM FROM

Fig. 6. Electrode configuration for conductance method.

by interpretation of the voltage reading across resistor RB' Probe