Taitel Et. Al, 1990

ADVANCES IN HEAT TRANSFER.
VOLUME 20
Two-Phase Slug Flow
YEHUDA TAITEL AND DVORA BARNEA
Faculty of Engineering, Department of Fluid Mechanics and Heat Transfer,

Tel-Aviv University, Ramat-Aviv 69978, Israel
I. Introduction
Gas-liquid flow in conduits may take on a variety of configurations

related to the spatial distribution of the two phases in the pipe, termed as
flow patterns (Mandhane et al., 1974; Taitel and Dukler, 1976; Taitel
et al., 1980; Barnea, 1987; Weisman et al., 1979). One of the most com-
plex flow pattern with unsteady characteristics is the intermittent or slug
flow. Gas-liquid intermittent flow exists in the whole range of pipe inclina-
tion and over a wide range of gas and liquid flow rates. In vertical slug flow
most of the gas is located in large bullet-shaped bubbles, which occupy
most of the pipe cross section. These bubbles are usually called Taylor
bubbles. The Taylor bubbles are separated by liquid slugs containing
usually small bubbles. The liquid confined between the bubble and the pipe
wall flows around the Taylor bubble in a thin falling film. In horizontal and
inclined flow, slugs of liquid that fill the whole cross section of the pipe are
separated by a stratified zone with an elongated gas bubble in the upper
part of the pipe and the liquid film at the bottom. The intermittent pattern
is sometimes subdivided into slug and elongated bubble flow patterns.
When the flow is calm and the liquid slug is almost free of gas bubbles the
pattern is termed as elongated bubble flow. For high flow rates, when the
liquid is aerated with gas bubbles, the flow is designated as slug flow. In
spite of the distinction between slug and elongated bubble flows, the term
slugflow is still often used for the general intermittent flow.
Slug flow is a highly complex type of flow with an unsteady nature, thus
the prediction of pressure drop, heat, and mass transfer for such flow is a
difficult task. Obviously an exact solution of the continuum equations is out
of the question at this time. Therefore a variety of approximate methods
83
Copyright 01990 by Academic Press, Inc.
All rights of reproduction in any form reserved.
84 YEHUDATAITEL
AND DVORABARNEA
has been developed for calculating the slug hydrodynamic parameters. The
former methods simply used correlations of experimental results. More
recently there is a tendency to formulate approximate models that are
capable of simulating the flow behavior sufficiently accurate so that the
calculation of the pressure drop as well as other flow parameters can be
performed with a relatively high degree of confidence and generality. Such
models were introduced by Dukler and Hubbard (1975) and Nicholson
et al. (1978) for horizontal flow; Fernandes et al. (1983), Sylvester (1987),
and Orell and Rembrand (1986) for vertical flow; and Bonnecaze et al.
(1971) for the inclined flow.
All of the aforementioned models deal with steady slug flow, which is an
orderly flow with relatively short slugs (less than 1000) and a constant
average flow rate of liquid and gas over the time period of a slug cycle.
There are, however, more complex types of slugs that are of a typical
transient nature and are abnormally long. The most common are the
terrain-induced slugging, where slugs are generated in a pipeline when the
heavy liquid is accumulated in the lower section of a pipe that follows a
hilly terrain. The state of the art of transient slugging is not yet well
developed and perhaps only the simple case of a system that contains a
single riser and a single pipeline was adequately studied (Schmidt et al.,
1980; Taitel, 1986; Taitel et al., 1989).
In this Chapter we will treat first the steady slug flow. Various options of
modeling the hydrodynamic parameters and pressure drop will be intro-
duced using a unified approach that is applicable for the vertical, horizon-
tal, as well as the inclined cases. Transient phenomena in slug flow will also
be reviewed with the detailed example of the case of severe slugging in a
pipeline-riser system. This system is of major practical importance for the
offshore oil and gas industry. It is also the one that has been treated with
sufficient success.
Heat transfer in slug flow is also of major importance for practical
applications. The topic of heat transfer during evaporation and condensa-
tion is considered in the specialized literature. The treatment of the two-
phase flow is usually considered there as a two-phase mixture. Very few
studies have been performed that treat the liquid and the gas in slug flow as
different entities and that address themselves to questions such as the
temperature profiles in the liquid and the gas, the local fluctuation of the
wall temperature, the difference between the upper and lower parts of the
wall temperature in slug flow, and the heat-transfer coefficient in slug flow
with dependence on peripherial and axial positions. Since we feel that the
amount of work done on this subject is not yet at the point where a
meaningful, coherent summary can be written, this review is limited to the
hydrodynamic aspects of slug flow only. Note, however, that the hydrody-
namics of slug flow is the basis for any detailed heat-transfer analysis.
TWO-PHASE
SLUGFLOW 85
11. Steady-State Slug-Flow Modeling
Models for predicting the steady slug flow were usually treated sepa-
rately for the horizontal case and for the vertical case. Some papers also
consider the inclined case.
For the horizontal case, the most detailed models were presented by
Dukler and Hubbard (1975) and Nicholson et af. (1978). For vertical
upwards flow, the most detailed work has been done by Fernandes et af.
(1983), Sylvester (1987), and Orell and Rembrand (1986). The inclined
case was considered by Bonnecaze et al. (1971).
In this presentation, we are not going to describe previous work in any
consecutive manner. We will rather present our own approach to the
modeling of slug flow that we feel is the best combination of engineering
accuracy and ease of calculation. Previous work will be reviewed (in a
critical way when appropriate) as we move along in the development of
modeling the different mechanisms that take place in slug flow. We will try
to present an approach that is as general as possible and can handle
vertical, horizontal, and inclined slug flows in an unified fashion.
Slug flow is a very complex fluid-mechanics problem. The purpose of
modeling slug flow is to be fairly close to the true physical process that is
taking place, but it is also important for a model to be simple enough so
that practical solutions for the slug-flow parameters can be obtained with
reasonable effort and that the modeling could be used in routine engineer-
ing calculations.
The schematic geometry of slug flow is shown in Figure 1. The slug body
is subdivided into two main sections: the liquid slug zone of length f, and
the film zone of length I f . Although the liquid slug zone can be aerated by
dispersed bubbles, it forms a competent bridging and gas cannot penetrate
through the slug zone. The liquid holdup within the liquid slug zone is
designated as R,. Once the slug is incapable of forming a competent
bridging, the slugs are then termed protoslugs (Andritsos and Hanratty,
1987) or wavy annular (Barnea et al., 1980), and this is the beginning of
transition to annular flow. The average liquid velocity in the liquid slug
body is designated as uL. The average axial velocity of the dispersed
bubbles in the liquid slug is termed u b . Note that uL and ub are not
necessarily the same, even though for horizontal flow, both velocities are
considered equal.
The film zone consists of a liquid film and an elongated gas bubble. For
horizontal and inclined pipes, the bubble is in the upper part of the pipe. In
vertical and off-vertical pipes, a complete symmetry is assumed, the bubble
is in the center of the pipe and a thin film flows around it adjacent to the
pipe wall. In this case the large bubble is termed Taylor bubble, and the
film zone is usually termed the Taylor bubble zone. The translational
86 YEHUDA
TAITELAND DVORABARNEA
FIG. 1. Slug-flow geometry.
velocity of the elongated bubble ut is the velocity at which the elongated

bubble propagates downstream. If one moves at a velocity u, the slug
picture is seen as frozen in space. The liquid velocity in the film is desig-
nated as uf and that of the gas uG. Note that unlike the liquid slug region,
which is considered axially homogeneous, the liquid and gas velocities in
the film zone vary along the pipe due to the variation of the film thick-
ness, h,(z) or 6(z) (for the symmetrical-vertical case), behind the liquid
slug.
A. MASSBALANCES
Mass balances presented here consider both the liquid and gas as incom-
pressible. For long pipelines, where the density is not constant, we can still
consider it as locally constant for the purpose of calculating steady slug
flow *
A liquid mass balance over a slug unit can be performed in two ways.
One way is to integrate the fluid flow rate at a fixed cross section over the
time of the passage of a slug unit. The second one is by considering the
SLUGFLOW
TWO-PHASE 87
volume of fluid in a slug unit. Both methods obviously yield the same
results.
Using the first approach for the liquid mass balance yields
WL=L
tU (uLAR,pLtS+ ~ o ' f u f A R f p L d r ) (1)
where W L is the input liquid flow rate; uL is the liquid average velocity in
the liquid slug; uf is the liquid velocity in the film; tu ,t, , and tf are the times
for the passage of the slug unit, the liquid slug, and the film zone, respec-
tively. Since t, = ls/ut and tf = l f / u , , Eq. (1) takes the form
The second way of formulating the mass balance yields the following
relation,
The term in the parenthesis is the mass of the liquid in a slug unit. A slug
unit is propagating in the pipe at a velocity u, and the time for a slug unit to
pass through a fixed point in the pipe is tu = lu/ut. However, at this time,
part of the liquid in the film moves upstream (backward) relative to the
gas-liquid interface and is captured by the following slug. This amount of
picked up liquid X is given by the expression
X = (ut - U L ) P L A R
=~(ut - Uf)PLARf (4)
Using Eq. (4)for X , one can show that both Eqs. (1) and (2) are indeed the
same.
Equations (2) and (4)can be combined to yield
Exactly equivalent mass balances can be performed on the gas. How-

ever, it is more convenient to use a mass balance on one species only and
a mass balance on the mixture. A very simple continuity balance on both
liquid and gas states that for contant densities, the volumetric flow rate
through any cross section is constant. Applying this balance on a cross
section in the liquid slug zone yields
us = ULS + uGS = uLR, + ub(Ys (6)
where us is the mixture velocity within the liquid slug.
88 YEHUDATAITEL
AND DVORABARNEA
B. AVERAGE
VOIDFRACTION
The average void fraction of a slug unit is defined as
a , = ( a & + j--clfdx)/f, (7)
Using the mass balance [Eq. ( 5 ) ] to eliminate the integral term in Eq. (7)
yields
a, = (-uLs + ULRs + uta,)/ut (8)
Using Eq. (6), the liquid flow rate uLs can be replaced by uGS to yield
= (UGS - U b a s + U&s)/ut (9)

As pointed out before (Barnea, 1989), Eqs. (8) and (9) present indeed a
very interesting result. It shows that the average void fraction of a slug unit
depends only on the liquid and gas flow rates ( u G S uLs),
, the dispersed
bubble velocity u b , the translational velocity ut , and the void fraction
within the liquid slug as,and it is independent of the bubble shape, the
bubble length, the liquid slug length, and film thickness. This is a very
important and convenient result since it shows that the gravitational pres-
sure drop can be calculated independent of the detailed slug structure. For
the simple case when the liquid slug is not aerated, Rs = 1 and Eq. (9)
reduces to the simple result a, = u G S / u t .
OF THE LIQUID
C. HYDRODYNAMICS FILM
The length of the liquid film lf , its shape hf(z), the velocity profile along
the liquid film uf(z), and, especially, the film thickness and its velocity just
before pickup hfe and ufe , respectively, are important parameters for
calculating the pressure drop and heat and mass transfer in slug flow.
The shape of the liquid film is a very complex structure, especially near
the tail of the liquid slug. It is a three-dimensional problem, with free
surface, of turbulent flow and obviously an exact solution is out of the
question at this time. A reasonable approximation is to use the one-
dimensional approach of the channel flow theory. This method has been
used by Dukler and Hubbard (1975) and Nicholson er af. (1978).
In order to find solutions for the film velocity uf and the film holdup Rf as
a function of position from the rear of the slug, z , we will consider
momentum balances on the film zone. Refering to Figure 1, the momen-
tum equations for the liquid film and the gas above it relative to a coordi-
TWO-PHASE
SLUGFLOW 89
nate system moving with a velocity u , are

dUf
az dz
+
pLuf- = - - -- -+ pLgsin /3 - pLgcos p-
d P TfSf TjSj
Af Af
dhf
dz
(10)
where vf = u, - uf and vG = u, - uG. Note that although these equati0n.s

are written for the relative velocities vf and v G , the shear stresses are given
in terms of the real velocities as follows (see Fig. 1 for the definition
of the positive direction for the shear stresses):
? =fi[PGIuG - ufI(uG - uf)/2] (14)

where ff,fG , and fi are the 'friction factors between the liquid and the wall,
the gas and the wall, and the gas-liquid interface, respectively; uf and uG
are considered positive in the downstream ( x ) direction.
For smooth pipes, the Blasius correlation can be used,
ff = cf(Dhuf/ (15)
where Dh = 4Af/Sf. A similar expression can be used for the gas with the
exception that the gas hydraulic diameter is taken as Dh = 4AG/(SG+ Si)
(Taitel and Dukler, 1976). For laminar flow Cf = 16 and n = -1, while for
turbulent flow Cf = 0.046 and n = -0.2.
For rough pipes the roughness of the pipe should be taken into account.
An example for such an expression is the convenient explicit formula
(Hall, 1957):
f = 0.001375{1+[2 x 104(&/Dh)+ (106/Re)]'/3} (16)
Obviously many other correlations can be used.
More problematic is the determination of the interfacial friction fac-
tor fi. For the case of low liquid and gas velocities, the smooth surface fric-
tion factor can be used. When the interface is wavy, the wavy structure
determines the value of the average friction factor. Unfortunately, due to
the complexity of the wavy structure, the interfacial friction factor cannot
be predicted accurately and one has to use some crude correlations and
assumptions. The nature of the interface (smooth or wavy) can be deter-
mined on the basis of flow pattern maps considering the film zone as
stratified flow with the appropriate flow rates of liquid and gas.
90 YEHUDATAITEL
AND DVORABARNEA
For wavy stratified flow in the horizontal and inclined cases, a constant
value o f 5 = 0.014 was suggested (Cohen and Hanratty, 1968; Shoham and
Taitel, 1984). For the vertical case Wallis (1969) correlation for cocurrent
annular flow
fi = O.OOS[l + 300(8/D)] (17)
can be used, though the flow in the film zone is usually countercurrent. It
may be noted that Wallis et al. (1978, 1979) suggested modified correla-
tions for countercurrent flow. However, those correlations are applicable
near the flooding point and this is usually not the case for normal slug flow.
As can be seen, the information regarding the interfacial shear is very
limited, primarily for inclined pipes. Fortunately the accuracy of the in-
terfacial friction is generally not important since in most cases the interfa-
cial shear in the film zone is negligible.
Eliminating the pressure gradient from Eq. (10) and (11) yields
Using Eq. (4), the relative velocities vf and vG are given by
Vf = (u, - Uf) = (u, - UL)Rs/Rf (19)

Likewise
VG (ut - uG) = (Ut - u b ) a y , / a f (20)
Since Rf as well as af are functions of hf or 8, substituting these values in
Eq. (18) yields a differential equation for hf (or 8) as a function of z :
(21)
where for the case of stratified film flow
SLUGFLOW
TWO-PHASE 91
The differential equation [Eq. (21)] is solved numerically for hf (2)and the
corresponding u f (z) is found using the mass balance [Eq. (19)]. The integra-
tion is performed until the mass balance of Eq. (5) is satisfied, yielding the
length of the liquid film lf ,as well as the holdup Rfe,and the velocity ufeat the
end of the liquid film just before pickup.
For large z, the limiting value of hf, is the equilibrium liquid level hE ,which
is obtained when dhf/dz = 0, namely, the numerator of Eq. (21) equals zero.
Note that for aerated liquid slugs, the gas velocity in the elongated
bubbles usually exceeds that of the dispersed bubbles in the liquid bridge.
In addition the liquid film seems to be essentially free of small bubbles. The
physical picture that is consistent with this description is that the dispersed
bubbles in the liquid slug coalesce at the nose of the elongated bubble,
while gas bubbles are reentrained from the back of the bubble into the
liquid slug. Thus the liquid holdup in the front of the liquid film Rfi equals
the value of R, and ufi equal uL ; h, is the liquid level corresponding to R, .
Thus the integration of Eq. (21) starts normally with hf = hfi= h, at z = 0
and hf decreases (dhf/dz < 0) from h, towards the limit of hE .
However, under certain conditions, dhf/dz may be positive. It occurs
whenever the critical liquid level h, is less than h, , where h, is the level that
equates the denominator to zero. In this case, the liquid level reduces
‘‘instantaneously’’ to the critical level, and the integration of hf starts with
hfi = h, at z = 0. This procedure is similar to the analysis of liquid drainage
from a reservoir to a super critical channel flow (Henderson, 1966). We
may further note that in the event that h, or h, are less than the equilibrium
level h E , then hE is reached immediately. For the vertical case, the de-
nominator is never zero and a critical film thickness does not exist.
Equation (21) is the most detailed form of the one-dimensional channel
flow approach. This approach with several degrees of simplification has
been used by various investigators. Dukler and Hubbard (1975) and
Nicholson et aE. (1978) assumed that the pressure drop in the film zone is
negligible. Under this assumption, the liquid is treated as an uncoupled
free surface channel flow, and Eq. (18) takes the form
Equation (23) is still a differential equation that has to be integrated

numerically, and the neglect of the pressure drop along the gas bubble,
although usually justified, may be incorrect for very long film zones in
which the contribution of the pressure drop in the gas zone is not
negligible.
Further simplifications have been proposed in order to avoid the numer-
ical integration. The most common approach is to consider the liquid film
92 TAITELAND DVORA
YEHUDA BARNEA
as having a constant thickness in equilibrium. This equilibrium level is

indeed the solution of hE . For the case of the vertical pipe this was, in fact,
the only approach taken (Fernandes et al., 1983; Sylvester, 1987; Orell and
Rembrand, 1986). In this case, the solution for h (or S for vertical symmet-
rical flow) should satisfy
In the aforementioned works, however, the pressure drop in the gas zone
was also neglected. This neglect is usually justified only for relatively short
bubbles.
In summary, several approaches with various degrees of simplicity have
been presented for the hydrodynamics of the liquid film. We focus our
attention on three cases:
Case 1. This is the most general formulation for the one-dimensional
channel flow approximation. It is given by Eq. (18) or (21).
Case 2. The liquid film is treated as a free surface channel flow
[Eq. (2311.
Case 3. An uniform film is assumed along the bubble zone [Eq. (24)].
D. PRESSURE
DROP
Since the slug is not a homogeneous structure, the local axial pressure
drop is not constant. For practical purposes, we need the average pressure
drop over a slug unit, namely, APJl,,.
The pressure drop for a slug unit can be calculated using a global force
balance along a slug unit between cuts A-A and B-B (Fig. 1). The
momentum fluxes in and out are identical and the pressure drop across this
control volume is
+ T G ~ Gdz
APu = pug sin Plu + -
A A
where pu is the average density of the slug unit:
pu = aupG + (1 - a u ) p L (26)
The first term on the right-hand side of Eq. (25) is the gravitational
contribution to the pressure drop whereas the second and third terms are
the frictional term in the slug and in the film zones.
A second method, which is frequently used for calculating the pressure
drop, is to neglect the pressure drop in the film region and to calculate the
pressure drop only for the liquid slug zone. In the slug zone a control
SLUGFLOW
TWO-PHASE 93
volume between the plane cuts A-A and C-C is used. The resulting
pressure drop along a slug unit APu in this case is
r, rrD
APu= p,g sin fils + -1, + APmix
A
where p, is the average density of the liquid slug body, namely,
P s = %PG + RSPL (28)
The first term on the right-hand side of Eq. (27) is the gravitational term
of the liquid slug; the second term is the pressure loss due to friction, and
the third term is the pressure losses in the near-wake region behind the
long bubble. Dukler and Hubbard (1975), Nicholson et al. (1978), and
Stanislav et al. (1986) proposed that this pressure drop is associated only
with the acceleration of the slow moving liquid in the film to the liquid
velocity within the liquid slug, namely,
APmix = APacc = ~ , R s A ( u t - UL)(UL - Ufe) (29)
However, a careful mass balance between cuts A-A and C-C indicates
that the contribution to APmiX is not only due to the acceleration pressure
drop, but also due to the change in the liquid level between the film zone
and the liquid slug zone (Taitel and Barnea, 1989), namely,
A A p m i , = p L g c o S f i ~ ~ ~ f e ( h f e - y ) b -dpyL g c o s f i ~ ~ ~ " ( h f i - Y ) b d y
+ p~R,A(ut- uL)(ufi - ufe) (30)

As discussed in the previous section, hfiis usually h, ,and ufiis uL , in which
case the last term on the right-hand side of Eq. (30) is equal to the
acceleration pressure drop as given by Eq. (29). However, when h, < h,,
hfi equals h,, ufi = u,, and the acceleration term in Eq. (30) is different
from that in Eq. (29). The integration in Eq. (30) can be carried out and
written explicitly
(31)
Inspection of Eq. (30) shows that APmixis always less than AP,,, . The use
of APmix= AP,,, may cause a minor error for small-diameter pipes but it
can lead to a serious error when the pipes are of a large diameter.
94 YEHUDA
Two methods have been presented for the pressure-drop calculation. In

the first method, a global force balance is used [Eq. (25)], while the second
method is based on the momentum balance only on the liquid slug zone,
neglecting the pressure drop in the film zone [Eq. (27)]. It will now be
shown that the two methods are identical provided that in the first method
we also assume that the pressure along the film zone is essentially constant.
The integrated form of the momentum balance given by Eq. (23) is
xl0" R fdz - pLg cos p

6,
hfe ah,
Rf -dz
dz
(32)
The left-hand side of Eq. (32) is exactly the acceleration term APacc.
lohf
Integrating by parts, one can show that
lohfAfdhf = (hf - y)b dy (33)
Equation (32) then takes the form
lohf'
+ PLg cos p (hfi - y)b dY (34)
Substituting Eq. (34) into Eq. (30) yields another expression for AP,, :
= pLg sin p
A APmiX
i:' + 1:
A, dz qSfdz
By substituting APmixof Eq. (35) into Eq. (27), one can see that it is
(35)
identical to Eq. (25) (for the case where pressure drop in the film zone is
zero). Namely, the two methods for pressure drop calculation-(1) a
global momentum balance on the whole slug unit [Eq. (25)], and (2) a
momentum balance over the liquid slug only [Eq. (27)]-are identical.
For upward inclined and vertical flows the equilibrium level (or film
thickness) is frequently reached after a short distance from the liquid slug
tail. In this case, one can view the liquid film as composed of two sections.
A curved zone and an equilibrium zone where the wall shear stress bal-
TWO-PHASE
SLUGFLOW 95
ances gravity and the film thickness is uniform. In this case the upper limit
of the integrals in Eq. (35) can be the length of the curved zone. This
means that the mixing pressure drop APmixequals the sum of the gravita-
tional and the wall shear stress contribution in the liquid film adjacent to
the curved nose of the bubble. This observation was first pointed out by
Barnea (1989) for the vertical case and Taitel and Barnea (1989) for the
inclined case and it has some interesting consequences when the approx-
imation of uniform equilibrium liquid level is used.
When the simplified approach, which considers the liquid film in con-
stant equilibrium thickness, is used (Fernandes et al., 1983; Sylvester,
1987), the calculation of the pressure drop via Eqs. (25) and (27) is not
consistent and yields different results. Equation (25) in this case reads
T,TD TGSG
AP, = psg sin p l , + -1, + pfg sin plf + -IfTfSf +-
A A A If (36)
where pf = (YfpG + RfpL.

Note that in the film zone, gravity is balanced by the shear forces.
Assuming now that the pressure drop in the gas bubble is negligible [this is
the same assumption that was used in deriving Eq. (27)], this leaves only
the first two terms in Eq. (36). This result is clearly in contradiction to
Eq. (27), where APmixis given by the conventional acceleration term
[Eq. (29)]. As pointed out by Barnea (1989) and demonstrated by Eq. ( 3 9 ,
one has to consider a curved nose in order to obtain the same results for the
pressure drop by using either Eq. (25) or (27).
Barnea (1989) compared the results of the pressure drop for the constant
film thickness model with the more exact solutions. She showed that for the
vertical case, when a cylindrical bubble with a flat nose is assumed, the
results of the pressure drop, without the acceleration term [in Eq. (27)],
is usually closer to the more exact solution than when the acceleration term
is used. Note that in the work of Fernandes et al. (1983)) and Sylvester
(1987), the acceleration term was considered, although the constant film
thickness approximation was used.
E. AUXILIARY
RELATIONS
The formulations provided so far are not sufficient yet to obtain a
solution. In order to proceed, we should consider the following additional
variables: (1) the translational velocity ut , (2) the dispersed bubbles veloc-
ity u b , (3) the liquid holdup in the liquid slug zone R,, and (4)the liquid
slug length I, or the slug frequency v,. We will term these variables as
auxiliary variables. It is convenient for the time being to consider them as
96 YEHUDA
known variables, given in terms of known formulas, and postpone the

discussion on determining their values to the end of the chapter. In this
manner the solution procedure can be presented in a more general man-
ner. Namely, a solution can be performed for any of these values regard-
less of the details of their correlations.
The translational velocities are the velocities of the interfaces and this
applies, in principle, both to the elongated bubble in the film zone (Taylor
bubble in the vertical case) as well as the dispersed bubbles within the
liquid slug zone. As mentioned, the translational velocity of the elongated
bubble ut is larger than the gas velocity within this bubble uG. The
prediction of this translational velocity is not an easy task and, in fact, it is
the subject of current research. It is assumed, however, that the trans-
lational velocity can be expressed as a linear relation of the slug mixture
velocity as
where ud is the drift velocity, namely, the velocity of propagation of a large

bubble in stagnant liquid and the factor C is related to the contribution of
the mixture velocity. This factor is larger than unity as it is influenced by
the liquid velocity profile ahead of the bubble. This expression is very
similar to the Zuber and Findlay (1965) distribution parameter, although
here it results from an entirely different reason. C and u d are both consid-
ered constant for given operation conditions and, in fact, they are usually
taken as constants for all flow conditions. At the present time we will
assume that both C and ud are known. A special discussion will be given
later on the methods used to find these values.
The dispersed bubbles translational velocity can be expressed in a similar
manner:
u ~ = B u ~ + u ~ (38)
However, unlike the case of the elongated (Taylor) bubbles, the trans-
lational velocity and the gas velocity are the same for the small bubbles.
The coefficient B is the distribution parameter and uo is the drift velocity
for stagnant liquid. A discussion for the recommended values of B and uo
will be presented later.
The liquid holdup within the liquid slug zone R, depends on the flow
rates of the liquid and the gas and the pipe inclination. The relation for Rs
is obtained either from experimental correlations or by a mechanistic
model.
Finally information for the liquid slug length or slug frequency should be
given. The liquid slug length and the slug frequency are interrelated
variables and it is sufficient if one of these values is given.
TWO-PHASE
SLUGFLOW 97
As mentioned, we will assume that these four variables are known and
proceed towards a solution. Later, a special discussion will follow and the
state of the art for determining these variables will be presented.
F. CALCULATION
PROCEDURE
The set of equations obtained so far (including the auxiliary relations)
allows the calculation of the detailed slug structure, which is the basis for
calculating the pressure drop as well as heat and mass transfer. A solution
is sought for a given set of operating conditions, namely, liquid and gas
flow rates, pipe diameter and inclination, as well as the physical properties
of the liquid and the gas. As demonstrated in a previous section, the
calculation of the average void fraction a, can be performed immediately
using Eq. (8) or (9), which are independent of the detailed slug structure.
Unfortunately the detailed slug structure as well as the pressure drop
calculations is not that simple and it requires some numerical efforts. The
complexity of the calculations depends largely on the way the shape of the
liquid film is calculated. We have distinguished three basic cases as related
to the method used for calculating the hydrodynamic of the liquid film (see
Section 11,C): (1) accurate film profile, (2) simplified film profile ( T ~ ,
TG = 0), and (3) constant equilibrium level.
For convenience, we shall start with the simplest case, case 3, where an
equilibrium constant liquid level (or constant film thickness) is assumed.
This method was used primarily for vertical slug flow (Fernandes et a l . ,
1983; Orell and Rembrand, 1986), while for the horizontal case, both
Dukler and Hubbard (1975) and Nicholson et al. (1978) considered the
shape of the liquid film. In principle there is no reason why the constant
equilibrium level was adopted only for the vertical case and not for the
horizontal case. A partial justification for it is that in the vertical case the
liquid film reaches an equilibrium thickness in a shorter distance than in
the horizontal case. Yet this situation depends largely on the flow rates of
the liquid and gas.
The solution in this case first requires the calculation of the terminal
equilibrium level hE (or tiE). This is done via an implicit solution of
Eq. (24) for the film thickness. The calculation sequence can be visualized
as a trial-and-error procedure as follows:
1. u sis calculated using Eq. (6) (the superficial velocities are assumed to
be known).
2. The auxiliary variables u,, ub, R , , and 1, are determined first; ut is
calculated using Eq. (37); u b is calculated using Eq. (38); R , and 1, are also
evaluated by the proper methods (Sections I1,I. and 11,J.).
3. u L is determined using Eq. (6).
98 YEHUDA
4. A first guess for h f is assumed, which allows for the calculation of the
geometrical parameters Af , Rf , AG , Sf, SG , and Si. For a stratified film,
the following geometrical relations are used:
- c 0 ~ - ' [ 2 ( h f / D-) 11 + [ 2 ( h f / D -
R f = (1/7~){7~ ) 1141- [2(hf/D) - 112}
(39)
Sf = D{.n - cos-1[2(hf/D) - 11)
Si = DJ1 - [2(hf/D) - 11'
For a symmetrical annular film, simpler geometrical relations can be
readily obtained.
5. uf is extracted from Eq. (4).

6. uG can be calculated using a mass balance on a cross section in the
film zone [similar to Eq. (6)]:
U G a f -k ufRf = U s (42)
7. The friction factors ff, fG, and fi are evaluated on the basis of the
appropriate Reynolds number. For this purpose there are a few options as
discussed earlier. For example, one may use Eq. (16) for ff and fG and a
constant value for fi (see Cohen and Hanratty, 1968, for stratified flow) or
Eq. (17) for annular flow.
8. The shear stresses Tf, TG , and T~are calculated using Eqs. (12)-(14).
9. At this point, all the variables in Eq. (24) can be calculated, and one
can check whether the trial film thickness is correct. Obviously, the
approach for obtaining the correct solution h f is to use one of the standard
methods to ensure fast convergence such as the interval bisection method,
the method of false position, Newton-Raphson method, or any other
appropriate method.
10. Since slug length I, is considered known, the film length can now
be calculated using Eq. (2) for the unknown I,. This results in
It should also be mentioned that under most conditions T~ and q are

very small and can be neglected as was indeed the case in most of the
previous reported works. This simplification, however, saves only minor
computational efforts and is not recommended for the general case since it
can cause a serious error for the case where very long elongated bubbles
exist.
Once the film thickness is known, one can proceed and calculate the pres-
sure drop. For this purpose one can use either Eq. (25) or (27). Both, in
principle, should give the same results. However, due to the approximation
TWO-PHASESLUGFLOW 99
taken here and the neglect of the curved shape of the bubble nose,
Eq. (25) is not consistent with Eq. (27). This point was discussed in detail
by Barnea (1989) and Taitel and Barnea (1989). For vertical upward flow,
Barnea (1989) showed that in the case of an uniform film thickness,
Eq. (25) will somewhat underpredict the pressure drop, whereas Eq. (27)
will overpredict the pressure drop, comparing it to the case where the bub-
ble shape is taken into account. However, Eq. (25) is usually much closer
to the more exact solution. Thus, when using the constant-film-thickness
approach it is recommended to use Eq. (25) rather than Eq. (27), namely,
to ignore the acceleration term in the pressure calculation. This is contrary
to the way Fernandes et al. (1983) made their calculation (for the vertical
case) and is consistent with the Orell and Rembrand (1986) calculations.
For the horizontal case, both Dukler and Hubbard (1975) and Nicholson
et al. (1978) used Eq. (27). Since they did not neglect the curved nose their
method of calculation should have yielded the same result as Eq. (25).
However, as pointed out by Taitel and Barnea (1989), they calculated the
pressure drop in the mixing zone erroneously, neglecting the integral terms
on the right-hand side of Eq. (30), which may result in a serious error for
large-diameter pipes.
The term accurate film profile (case 1) is used when the exact Eq. (21) is
used to integrate the liquid film level hf(z). The simpliJiedfilrnprofile (case
2) is used when the pressure drop in the film zone is neglected, and hf(z) is
obtained via the integration of Eq. (23), instead of Eq. (21). Both cases,
however, require numerical integration and their solution is very similar.
The solution follows the following steps:
1. The variables u,, u s , R,, u b , and u L are calculated as in case 3.
2. h, is calculated, using Eq. (39) on the basis of the value of R, .
3. The critical level h, is calculated by finding the level at which the
denominator in Eq. (21) equals zero.
4. The value of hf at z = 0, h f i ,is set equal to the lower value of h, or h, .
5. Equation (21) or (23) (for case 2) is integrated numerically to yield
h,(z). This integration is carried out until the mass balance [Eq. ( 5 ) ] is
satisfied and, thus, the film zone length is obtained. Note that Dukler and
Hubbard (1975) as well as Nicholson et al. (1978) applied the integration
only to the approximate Eq. (23). Also their method of integration was
different. Instead of integrating hf(z) (or Rf), they integrated z(Rf).
As mentioned, cases 1 and 2 will usually yield similar results. However,
for the case of very long film zones, the pressure drop in the gas cannot be
neglected. Since the efforts in the calculation of case 2 are not much easier
than in the general case (case l ) , it is recommended to use the exact
method (case 1) for all the calculations.
100 YEHUDA
6. Since the numerical integration provides the profile of all the vari-
ables in the film zone, the pressure drop can be easily obtained using
Eq. (25) or (27) (when the pressure drop in the gas zone is ignored). If
Eq. (27) is used, the pressure drop in the mixing zone can be calculated
by either Eq. (30) or (35).
All in all, the calculation procedure describes a considerable number of
options for the user. It is not written in a “ready to use” single format and
the results are not presented in the form of a computer program or exact
flow chart. It leaves some effort to the user to choose the option appropri-
ate for his specific use and to write his own program. It does, however,
contain sufficient details to guide the reader in the use of the options avail-
able and present the advantages and drawbacks of the various possibilities.
G. TRANSLATIONAL VELOCITIES OF ELONGATED (TAYLOR) BUBBLES

As discussed earlier, Eq. (37) was assumed to apply for the translational
velocities, and the constants C and u d were assumed to be known. We will
discuss now the methods by which these variables are determined.
The idea behind the specific form of Eq. (37) is that the translational
velocity can be composed as a superposition of the velocity of bubbles in a
stagnant liquid (or liquid mixture) ud and the additional contribution of the
mixture velocity u s . Also it is assumed that the translational velocity is
linearly dependent on the mixture slug velocity. Obviously these assump-
tions are just an approximation, subject to experimental and theoretical
verification.
The motion of elongated bubbles is usually treated separately for the
vertical case (Marrucci, 1966; Bendiksen, 1985; Nicklin, 1962; Nicklin
et al., 1962; Davies and Taylor, 1949; Dumitrescu, 1943), the horizontal
case (Kouba, 1986; Nicholson et al., 1978; Dukler and Hubbard, 1975), and
the more general upward inclined case (Stanislav et al., 1986; Hasan and
Kabir, 1986; Zukoski, 1966; Singh and Griffith, 1970; Bendiksen, 1984.)
The first attempt, and the most successful one, was the treatment of the
vertical case in which reasonable agreement among different researchers
exists and both empirical correlations and theoretical approaches seem to
be satisfactory. For the horizontal case the situation is less clear. For
example, Wallis (1969), Dukler and Hubbard (1975), as well as Bonnecaze
et al. (1971) claimed that the drift velocity is zero for the horizontal case
since the buoyancy force does not act in the flow direction. Only later
Nicholson et al. (1978), Bendiksen (1984), and others showed that a drift
velocity exists also for the horizontal case and, in fact, it may even exceed
its value in the vertical case (Weber, 1981).
The drift velocity depends on the detailed two-dimensional flow at the
front of the bubble. Note that the simple one-dimensional approach based
TWO-PHASE
SLUGFLOW 101
on channel flow approximation is not suitable for analyzing the immediate

region of the bubble nose. Fortunately, approximations that are based on
potential flow yield reasonable results. Davies and Taylor (1949) were first
to perform such a calculation for the vertical case. Their first approach was
quite simple. They consider a potential flow around the nose of the bubble:
+
q2 = u2 sin2 8 (44)
where q is the tangential liquid velocity on the surface, 8 is a polar
coordinate, and U is the free-stream velocity.
Application of the Bernoulli equation and considering the pressure
within the bubble as constant yields that gz = q2/2. Substituting this rela-
tion in Eq. (44) yields for small 8:
u = ud = ( 2 / 3 f i ) m = 0.471 m (45)
In Eq. (45) D is the bubble diameter. Modification of this equation, which
takes into account the bubble rise in a confined pipe, was also performed
by Davies and Taylor (1949). They used a series expansion technique and
obtained the same form of Eq. (45) with a constant of 0.328 instead of
0.471. Dumitrescu (1943) performed somewhat more accurate calculations
and obtained the constant of 0.35, which thereafter was accepted as the
best value that also agrees very well with experimental observations (Nick-
lin et al, 1962).
For the horizontal case the situation is less clear. The most interesting
fact is that some of the papers do use the drift velocity, whereas others
consider the drift velocity as zero on the basis that gravity cannot act in the
horizontal direction. The more recent work of Nicholson et al. (1978),
Weber (1981), Bendiksen (1984), and Kouba (1986) clearly show that a
drift velocity exists also for the horizontal case owing to gravity-induced
drift that results from elevation difference in the bubble nose.
Consistent with the approach taken for the vertical case, in the horizon-
tal case also the inviscid theory is applied near the nose region. The drift
velocity in horizontal slug flow is the same as the velocity of the penetration
of a bubble when liquid is emptied from a horizontal tube (Benjamin,
1968). To predict this velocity the following relations are considered (see
Fig. 2):
Continuity:
Av, =A2~2 (46)
where A 2 is given by
A2 = ( 7-
~ y + 4sin 2y)r2 (47)
Bernoulli theorem is applied between point (1) and the stagnation
point (0). Note that the pressure at the stagnation point, which is the same as
102 YEHUDATAITEL BARNEA
A N D DVORA
1 2
FIG.2. Propagation of gas pocket in the draining horizontal pipe.
the pressure in the gas bubble, is taken as the reference pressure:

Pi= - 3 p v :
Bernoulli theorem between points (0) and (2) along the free surface yields
v; = 2gr( 1 - cos y) (49)
Finally a momentum balance yields
where the integral term in Eq. (50) can be solved explicitly, namely,
As shown in Benjamin (1968), Eqs. (46)-(50) are solved for the liquid
level h2 (equivalent to A2 or y) and the liquid velocities v1 and v 2 . The
results are
h 2 / D = 0.563 and v1 = u d = 0.542&$ (52)
The result of Eq. (52) is supported experimentally (for bubbles with a
negligible effect of surface tension) by Zukoski (1966) and Bendiksen
(1984). It is interesting to observe that the drift velocity in the horizontal
case is larger than the drift velocity in the vertical case.
For the inclined case, there is no proposed model and one relies primar-
ily on experimental data. The inclined case, as well as the vertical and the
horizontal cases, were studied by Zukoski (1966), Singh and Griffith
(1970), Bonnecaze et al. (1971), Bendiksen (1984), and Hasan and Kabir
(1986). All report a peculiar behavior, that the drift velocity increases as
the angle of inclination is declined from the vertical position. The drift
velocity then decreases again toward the horizontal position such that the
maximum drift velocity occurs at an intermediate angle of inclination
:::I
OO
, , ,
45
, , ,
90
B (degree)
FIG. 3. Dimensionless bubble velocity in stagnant liquid versus inclination angle for diffe-
rent surface tension parameters: (-) predicted; (0) H = 0.064; (0) Z = 0.042;
(A)I = 0.01; [Bendiksen, 19841 ( X ) I = 0.001; (+) Z = 0.01, 0.042, 0.064 [Zukoski, 19661
(after Bendiksen, 1984).
around 40" to 60" from the horizontal. Bonnecaze ef al. (1971) were the first
to give a qualitative explanation for this peculiar behavior, arguing that the
gravitational potential that drives the liquid velocity along the curved sur-
face at the bubble nose increases and then decreases as the angle of incli-
nation changes from the vertical position towards the horizontal position.
Figure 3 shows the results of Bendiksen (1984) and Zukoski (1966) for
the change of the dimensionless drift velocity u d / m with the angle of
inclination. The upper curve represents the case where surface tension is
negligible and, thus, the results for the horizontal and the vertical limits
very closely follow the aforementioned theoretical results that were based
on potential flow; namely, that u d / m z 0 . 3 5 for the vertical case
( p = 90") and u d / m = 0.54 for the horizontal case. The surface tension
effect is given in terms of the surface tension parameter C = 4a/g(p, -
pG)D2.Figure 4 shows the experimental data reported by Zukoski (1966).
It shows that the effect of surface tension can indeed be substantial, par-
ticularly for small-diameter pipes. For a small surface tension parameter
C-0.001, the results for the vertical case and the horizontal case are
very close to the potential flow theory, namely, u d / m = 0.35 and 0.54,
respectively. The drift velocity, however, decreases considerably with an
increase in the surface tension parameter (decreasing the pipe diameter)
and eventually reaches a zero velocity when C is of the order of unity.
104 YEHUDA
0.0001 0.001 0.01 0.1 1.o 10
E- (I
@L-& )gr2
FIG.4. Variation of normalized velocity with surface tension parameter for @ = O", 45",
and 90" (after Zukoski, 1966).
The drift velocity is expected to depend also on the liquid viscosity, or

the bubble Reynolds number. However, Zukoski (1966) shows that the
dependence of the drift velocity on viscosity is negligible for Reynolds
number Re > 300 (Re = udpLD/pL). This is clearly demonstrated by
Fig. 5.
Bendiksen proposed, as a practical suggestion, to use the following
formula for the drift velocity in the inclined case:
ud=u!cosp+u:sinp (53)
where ui and u: correspond to the drift velocity for the horizontal and the
vertical case, respectively.
Hasan and Kabir (1986) proposed the relation:
Ud =u:&ji(1+ c0sp)l.Z (54)
which they claim to well correlate experimental data in the range
90" > p > 30".
Next we will consider the additional contribution of the mixture velocity
to the elongated bubbles translational velocity, namely, the value of the
constant C on the right-hand side of Eq. (37). In developed slug flow, the
translational velocity is usually related to the value of the liquid slug
TWO-PHASE
SLUGFLOW 105
FIG.5. Bubble velocity versus surface tension parameter for ranges of Reynolds numbers.
Flagged symbols from Barr (1926), Dumitrescu (1943), and Goldsmith and Mason (1962)
(after Zukoski, 1966).
velocity at the centerline, where the velocity attains its maximum. This is
based on the assumption that the propagation velocity of the bubbles is
equal to the maximum local liquid velocity in front of the nose tip (Nicklin
et al., 1962; Nicklin, 1962; Collins et al., 1978; Bendiksen, 1984, 1985;
Shemer and Barnea, 1987). Although this is a rather simplified approach, it
has been found remarkably valid and supported by the more exact
approaches (Collins et al., 1978) and by experimental data. Thus for
turbulent flow, C = 1.2, which is the ratio of u , , , / u , ~ ~for
~ turbulent flow.
Nicklin et al. (1962) state that this value is valid for a Reynolds number
greater than 8000 but it is also a good approximation for a Reynolds
number less than that. For laminar flow the ratio u,,,/u,,,, approaches 2
and indeed there is a strong indication that C increases as the Reynolds
number decreases and reaches a value of about 2. A more precise theory
shows that C , for laminar flow, equals 2.27 (Taylor, 1961; Collins et al.,
1978) for the case where the surface tension is neglected. Experimental
results that were carried out at about 2 = 0.05 show that C is 1.87 (Collins
et al., 1978), 1.94 (Bendiksen, 1985), and 1.8 to 1.95 (Nicklin er al., 1962).
Figure 6 is an example of Bendiksen (1985) data on the effect of Reynolds
number on the coefficient C. As seen, C has the value of about 1.2 at a high
106 YEHUDA
FIG.6. A comparison of predicted and measured distribution slip parameters C (0)

predicted with H = 0.042; (-----) predicted with
H = 0.042 [data of Bendiksen (1984)l; (-)
H = 0 (after Bendiksen, 1985).
Reynolds number and it increases as the Reynolds number decreases. The

limit of about C = 2 is not shown in this figure.
The exact value of C for the turbulent case, and in particular for the
laminar case, is not conclusive. There is still a spread in the experimental
data as well as the various theories. Yet, at least to a good engineering
approximation, the values of C = 1.2 for turbulent flow and C = 2 for
laminar flow is quite a good approximation for the case where the effect of
surface tension is small. Furthermore, it is applicable both for the vertical
case, the horizontal case, and also the inclined case (Bendiksen, 1984).
It should be stressed that the theoretical consideration just presented is
consistent with the assumption that the translational velocity is a linear
function of the mixture (slug) velocity as expressed by Eq. (37). Although
this is a valid engineering approximation, it is not necessarily the exact
representation of the real situation. If one plots experimental data of ut
versus us and extrapolates to us= 0, obviously the result for u s= 0 is the
drift velocity. If the curve u, versus usis not quite a straight line (which is
the realistic case), then one has several options. One can draw a straight
line that originates at the point us= 0, u, = u d , and draw a best fit over
the range form us= 0 up to any desired value. Another option is to fit a
straight line over any desired interval where us is larger than zero. In this
TWO-PHASE
SLUGFLOW 107
case u d is the extrapolation of this straight line to the zero point and it
would not be the same as the mentioned drift velocity. Bendiksen (1984)
showed that in this case the value of ud is usually less than the one reported
here due to the fact that the ut versus us curve bends slightly upwards.
However, it seems to us that for any practical application a straight line
that originates from the point (0, ud) is adequate for the accuracy needed
for engineering calculations.
H. VELOCITIES
OF THE DISPERSED
BUBBLES
IN THE LIQUIDSLUG
A rough criterion to distinguish between the elongated (Taylor) bubbles
and the dispersed bubbles within the liquid slug is the characteristic value
of the pipe diameter. Bubbles with a chord length larger than the pipe
diameter D are considered elongated bubbles. Smaller bubbles are usually
termed as dispersed bubbles.
As in the case of the translational velocity for elongated bubbles, it is
assumed that the velocity of the bubbles in the liquid slug u b is a linear
combination of the bubble drift velocity uo and the mixture velocity in the
slug zone u s ,as it is reflected by the form of Eq. (38). In Eq. (38), B is the
distribution parameter (Zuber and Findlay, 1965) and uo is the drift
velocity.
For the vertical case the drift velocity is the free-rise velocity of a bubble
in the pipeline. This free rise can be evaluated by considering the free-rise
velocity of a single bubble in an infinite medium, the free rise of a bubble in
a swarm of bubbles (the effect of voids in the liquid slug), and finally the
free-rise velocity in a cylindrical pipe. This subject has been discussed
extensively (see e.g., Brodkey, 1967; Levich, 1962; Govier and Aziz,
1972; Wallis, 1969).
The free-rise velocity of a single bubble in an infinite media depends
largely on the size of the bubble. For very small bubbles the bubbles
behave as rigid spheres and the free rise is governed by Stokes law. For
larger bubbles, a boundary-layer solution is applicable. As the bubble
diameter increases, circulation effects take place and affects the free rise.
When the bubbles become larger, their spherical shape is distorted and
flattens. This has a drastic effect on slowing down the free rise compared to
an equivalent spherical bubble. When the bubble size exceeds some critical
value, the rise velocity of the dispersed bubble tends to be constant and
independent of the bubble diameter. This critical bubble size is (Brodkey,
1967)
108 YEHUDATAITELAND DVORABARNEA
The bubbles in the liquid slug zone are usually larger than dcrit.For a
relatively large and deformable bubble, the equation proposed by Har-
mathy (1960) for the bubble rise velocity is considered of sufficient
accuracy:
2 1/4
u- = 1 - 5 4 “ d P L - PG) / PLI (56)
Note that Eq. (56) indeed shows that the free-rise velocity is independent
of the bubble size.
The free-rise velocity of a bubble within a swarm of bubbles is lower than
the free rise of a single bubble. This can be viewed as the decrease of
buoyancy that acts on a single bubble in a gas-liquid mixture. This de-
crease is correlated in the form
uo = u,( 1 - (57)
For relatively large bubbles, Wallis (1969) as well as Govier and Aziz
(1972) suggested the use of n = 1.5. Fernandes et al. (1983) used n = 0.5. A
value of n = 0 was recommended by Wallis (1969) (after Zuber and Hench,
1962) for the region termed as churn turbulent. The later case is probably
most close to the flow of bubbles in the slug region and, thus, the value
n = 0 is suggested. It should be noted that no direct information is available
on the rise velocity of bubbles within the liquid slug. However, for model-
ing slug flow, the accuracy of the rise velocity is usually not that important
anyway.
Finite size of the pipe also acts to decrease the free-rise velocity. This
effect is discussed by Wallis (1969). In general, the effect of the pipe
diameter is negligible for d / D < 0.125, and it is suggested to ignore it.
For the case of inclined pipes, we may assume that the drift velocity uo
should be multiplied by sin j? (Barnea et al., 1985).
The value of B depends on the concentration distribution of the bubbles
in the liquid slug as demonstrated by the method of Zuber and Findlay
(1965). Wallis (1969) points out that B for vertical dispersed flow “usually
lies between 1.0 and 1.5 with a most probable value of about 1.2.” For the
horizontal case the bubble concentration is definitely not uniform since
bubbles tend to concentrate at the top of the pipe. Nevertheless B was
taken as unity since it was assumed that the dispersed bubbles have the
same average velocity as the liquid in the mixture. Kouba (1986) did
measure the distribution parameter in horizontal slug flow and got a value
close to 1.2. However, owing to the lack of supporting evidence we would
recommend the use of B = 1 for the horizontal case. As can be seen the
evaluation of B for horizontal, vertical, as well as the inclined case is still an
open question.
TWO-PHASE
SLUGFLOW 109
I. LIQUIDHOLDUPIN THE LIQUID

SLUGZONE
As indicated in Section II,E, the model calculations require as input data
the value of the liquid holdup within the liquid slug R, . This value may be
obtained experimentally or through a physical model.
Hubbard (1965) measured R, in an air-water horizontal slug system by
using an impact Pitot probe system. This technique proved to be very
difficult for realization. Considerable scatter was observed and the results
obtained showed little consistency.
Experimental values of R,, using a light refined oil-air system in a
horizontal pipe, were obtained by Gregory et al. (1978). They used capaci-
tance-type liquid-volume-fraction sensors, which provided a continuous
record of the in situ liquid volume fraction. The ranges of flow rates
investigated cover virtually the entire region of slug flow that was observed
in their flow loop. For air-oil slug flow in horizontal 2.58- and 5.12-cm I.D.
pipes, they found a modest diameter effect and suggested a correlation of
the following form: 1
1
R, =
1+ ( ~ , / 8 . 6 6 ) ' . ~ ~
where the mixture velocity u s has units of meters per second.

In spite of the fact that this correlation is limited and does not include the
effect of fluid properties and pipe diameter, it is frequently used because of
its simplicity.
Greskovich and Shrier (1971) presented a graphical correlation of R,
involving the mixture Froude number (u,/Dg) and the input liquid quality
(A). This correlation is based on data collected with air-water in a 1.5-cm
I.D. horizontal pipe. Values of R, between 1 and 0.5 were obtained.
Heywood and Richardson (1979) used the yray absorption method in
order to determine the average holdup within the liquid slug for an aii-
water system in a 4.2-cm horizontal pipeline. The results are presented by
a graphical correlation of R, versus uGS with parametric values of uLs.
These results are similar to the correlation presented by Gregory et al.
(1978).
Schmidt (1977) measured the liquid holdup in the liquid slug in vertical
risers by using a capacitance sensor. Void fraction in the liquid slug was
correlated with uLs and uGS.The values of the void fraction range from 0.2
to 0.8.
Fernandes (1981) measured R, in a vertical air-water slug flow using a
5-cm I.D. pipe. Based on his experimental results, he suggested that R, be
set equal to 0.25. It should be mentioned that Fernandes' data was
obtained for a relatively limited uGS-ULS range.
110 TAITEL
YEHUDA AND DVORA
BARNEA
Barnea and Brauner (1985) proposed a method for estimating the gas
holdup within the liquid slug, a,. It was suggested that the gas holdup on
the transition line from dispersed bubbles is the maximum holdup that the
liquid slug can accommodate as fully dispersed bubbles at a given mixture
velocity u s . Thus, curves of constant u, within the intermittent region
represent the locus where a, or R, is constant and is equal to the holdup of
the dispersed bubble pattern at the transition boundary. The transition
boundary itself may be obtained by any reliable predictive model or
experimentally. Once it is obtained, R, may be determined by the previously
mentioned concept. For example, using the theoretical transition boundary
from dispersed bubbles for the vertical case, yields (Barnea, 1987),
a, = 1- R, = 0.058{2[0.4~/(pL - p~)g]”2[(2f,/D)~~]2/5(
p ~ / ( + )-
~ 0.725)’
/~
(59)
Note that Eq. (59) usually applies also to the inclined and horizontal cases
(see Barnea, 1987, for possible exceptions).
The calculated value of R, ranges from 1to 0.48, where 0.48 is associated
with the maximum volumetric packing of the dispersed bubbles in the
liquid slug. For the special case of vertical and off-vertical pipes with
relatively large diameters (D> 0.05 m for air-water), the maximum value
of R, is 0.75 (Barnea and Brauner, 1985).
Barnea and Shemer (1989) used a conductance probe to detect the
instantaneous void fraction at the centerline of a vertical 0.05- m I.D. tube
in upward air-water slug flow. This information was further processed to
obtain the liquid slug holdup and its length. The experimental values of
voidage ranges from a,=O.25 on the transition from bubbly flow, to
a, = 0.6 near the transition to churn flow, as has been predicted by Barnea
and Brauner (1985).
J. SLUGLENGTHAND SLUGFREQUENCY
The slug frequency and the liquid slug length are interconnected prop-
erties and are very often alternatively used (Nicholson et al., 1978). Ex-
perimental observations for air-water systems in vertical upward and
horizontal slug flows suggest that the stable liquid slug length 1, is relatively
insensitive to the gas and liquid flow rates and is fairly constant for a given
tube diameter. The stable slug length has been observed to be of about
12-300 for horizontal slugs (Dukler and Hubbard 1975). Nicholson et al.
(1978) noted that the variations in the average liquid slug length are much
smaller than the corresponding slug unit length and reported an average
value of 300. For the vertical case the observed liquid slug length is about
10-200 (Moissis and Griffith, 1962; Moissis, 1963; Akagawa and Saka-
guchi, 1966; Fernandes, 1981; Barnea and Shemer, 1989).
TWO-PHASE
SLUGFLOW 111
Slug frequency has sometimes been considered as an entrance phe-

nomenon, namely, it results from bridging of the liquid at the entrance
(Taitel and Dukler, 1977). This is indeed the case in horizontal and slightly
inclined flows, near the transition from stratified flow. In this case low-
frequency slugs are generated causing relatively long liquid slugs at the
entrance which propagates downstream. However, generally short (high
frequency) slugs are formed at the entrance of the pipe. These slugs are
usually unstable. Shedding of liquid at the rear of the liquid slug seems to
be larger for short slugs. As a result, an elongated bubble behind a short
slug moves faster and overtakes the bubble ahead of it (Moissis and
Griffith, 1962). The bubble and the corresponding liquid slug merge in this
process, decreasing the slug frequency. The merging process continues
until the liquid slug is long enough to be stable, namely, the trailing bubble
is unaffected by the wake of the leading one. This occurs when the velocity
profile at the rear of the liquid slug can be considered fully developed
(Moissis and Griffith, 1962; Taitel et al., 1980; Barnea and Brauner, 1985;
Dukler et al., 1985).
Taitel et al. (1980) and Barnea and Brauner (1985) simulated the mixing
process between the film and the slug by a wall jet entering a large
reservoir. The process of establishing the stable slug length can be visual-
ized as follows. Referring to Fig. 7, two consequent elongated bubbles are
shown. The first is behind a long steady liquid slug. The velocity profiles
within this slug are shown as they develop from a mixing wall jet profile to
a fully developed pipe flow at the back of the slug. Since the average total
mixture velocity at any cross section of the slug is the same and equals u s ,it
is obvious that the maximum velocity decreases asymptotically towards the
value of 1 . 2 4 with distance from the front of the liquid slug. As the bubble
velocity is related to the local maximum velocity ahead of it, it is clear that
bubble B, which is behind a short liquid slug is faster than bubble A , which
is behind a fully developed profile with u,, = 1 . 2 4 . Thus, bubble B will
overtake the leading bubble A. This is the process by which short slugs
tend to disappear. This process however is terminated once all the slugs are
SHORT SLUG LONG SLUG
0 0
I '
-1c
I 1
I 1
- f k
Ut>1.2US ut'l.2us
FIG.7. Velocity profiles in liquid slugs.

112 TAITELAND DVORABARNEA
YEHUDA
long enough so that the velocity profile at the back of the slugs is fully
developed. Thus, this process is usually the one that controls the slug
length. Taitel et af. (1980) and Barnea and Brauner (1985) suggested that a
developed slug length is equal to a distance at which a jet has been
absorbed by the liquid. Using this approach, a value of 1 6 0 was obtained
for the minimum liquid slug length in vertical upflow and a value of 3 2 0
was obtained for the case of horizontal flow.
Dukler ef af. (1985), on the other hand, assumed that the liquid at the
front of the slug is well mixed with an uniform velocity profile. From this
point on, a boundary layer is developed at the pipe wall until a fully
developed velocity profile is achieved. They found that the minimum stable
slug length 1, is of the order of 200. Although the final results of Dukler
et af. are similar to the previous analysis, it does not explain well the
merging mechanism. In their approach, the value of the centerline velocity
increased with the distance from the liquid slug front, and the maximum
velocity, which determines the bubble velocity, is minimal behind short
slugs. Namely, elongated bubbles behind short liquid slugs will move
slower then those behind longer ones, contrary to experimental observa-
tion.
Shemer and Barnea (1987) used the hydrogen-bubble technique to re-
cord the velocity profiles behind the elongated bubbles in gas-liquid slug
flow. They distinguished between two zones in the development of the
velocity profile. The first zone is an annular jet, which terminates at a
distance of about 2-30, causing a strongly disturbed velocity profile in the
whole cross section of the pipe. At larger distances from the bubble, a
gradual decay of the fluctuations occurs until a fully developed profile is
obtained. Shemer and Barnea (1987) found that the bubble shape in the
wake region closely resembles the liquid velocity profile ahead of it. They,
thus, concluded that the propagation velocity of the elongated bubble is
related to the maximum instantaneous liquid velocity ahead of it. They
found a steep decrease in this maximum velocity in the near-wake region,
while a much more gradual decrease is observed at larger distances from
the leading bubble until a fully developed velocity profile is observed. In
this case the lowest possible value of the instantaneous maximum value is
obtained. This distance from the leading bubble determines the minimum
length of the stable liquid slugs where all the bubbles have a smooth
rounded front shape and propagate with identical velocity. The detected
velocity field in the wake of the bubble was utilized by the investigators to
estimate the minimum stable slug length. They found that I , is of the order
of 200.
Most of the reported data and correlations on slug frequency and slug
length are related to the downstream developed slugs and not to the
entrance frequency.
Gregory and Scott (1969) measured liquid slug frequencies for the car-
bon dioxide-water system in an 1.9-cm diameter tube. They correlated
their data by the following relation (in SI units):
Greskovich and Shrier (1972) used their own data, as well as Hubbard
(1965) data and presented the following dimensional correlation for slug
frequency in horizontal pipes,
[
us =0.0226 A (2;-+
where A is the input liquid volumetric quality (A = &us) and the pipe
diameter D is given in meters.
Heywood and Richardson (1979) used the ?ray absorption method and
determined the probability density function and the power spectral densi-
ties of the holdup. From these functions they have estimated the average
film and slug holdup, the average slug frequency, and the average liquid
slug length. They suggested that the slug frequency data may be summa-
rized by the following relation in SI units:
U, [
=0.0434 A (2;~ + 5)]1.
0 *
In all these reports, the slug frequency data exhibit a minimum when
plotted versus the mixture velocity u s and it is a strong function of the
liquid flow rate.
As has been mentioned, slug frequency can replace slug length in the
auxiliary relation, and the slug length will be an output of the model
calculations. We feel, however, that the input of the slug length as an
auxiliary variable is preferred to the slug frequency since the slug length is
based on a physical model while the slug frequencies are given primarily by
experimental correlations.
K. CONCLUDING
REMARKS
Steady-state slug-flow modeling was presented using a general approach
that treats the slug hydrodynamics for vertical, horizontal, and inclined
pipes in the same fashion. Critical review of previous work is also
presented.
Three methods of solution are presented: (1) the exact method that uses
the fully one-dimensional channel-flow solution for the liquid film; (2) the
same as (1) with the neglect of pressure drop in the film zone; and
114 YEHUDA
(3) considering an uniform equilibrium level for the film in the film zone.
These three cases differ with regard to the accuracy of the solution and ease
of calculations. Case 3 was used primarily for vertical flow and case 2 for
horizontal flows. Method 2 however, can be inaccurate for long film zones.
Since method 1 is not more difficult to solve than method 2 it is recom-
mended to use method 1 rather than 2.
A special discussion is devoted to the way pressure drop can be calcu-
lated. It is shown that there are basically two methods for calculating the
pressure drop, namely, (1) using a global momentum balance on a slug
unit, and (2) assuming that the pressure drop in the film zone is negligible.
In the latter case, the acceleration-pressure drop should be used. It is also
shown that when the uniform equilibrium thickness is considered (case 3),
the two methods of calculating the pressure drop are not consistent. Based
on some numerical examples it seems that when an uniform film thickness
is assumed, the calculation for the pressure drop using method (1) is
preferred.
Although the present model is probably the most up to date and consis-
tent model for the calculation of the hydrodynamic parameters of steady
slug flow and best suited for practical applications, it is still incomplete and
some of the approaches used may be regarded as unsatisfactory. Obviously
more research has to be performed for the purpose of bringing the theory
closer to reality. We would like, at this time, to point out some of the
deficiencies that the reader should be aware of.
The most controvertible treatment is the one related to the bubble
shape, especially near the bubble nose. We have used different theories to
describe the hydrodynamics of the liquid near the bubble nose and the
hydrodynamics of the film further upstream, and then we simply superim-
posed their effect. To predict the drift velocity, two-dimensional potential
flow analysis was used, which is different for the horizontal and the vertical
case and does not include the inclined case. To the drift velocity we add the
propagation velocity (Cu,) using the notion that the bubble nose follows
the highest local velocity of the velocity profile ahead of it. Then the
liquid film was analyzed by using the one-dimensional channel-flow equa-
tions. Obviously the behavior of the bubble nose and the body of the film
should follow a single formulation. Such a formulation would be too
complex and probably impractical for pragmatic calculations. There is,
however, a genuine need for detailed calculations and experiments in order
to improve the understanding of the slug-flow hydrodynamics as well as to
provide accurate analysis with which the simple approaches could be
compared.
The friction factors used for the liquid slug and the film zones are also of
uncertain accuracy. The flow in the liquid slug is not developed, especially
TWO-PHASE
SLUGFLOW 115
in the near-wake region. The use of a fully developed friction factor may be
inaccurate. We may also note that the interfacial shear stress is virtually
unknown. Accurate data on the interfacial shear may be essential for slugs
with very long film zones and high gas velocities.
The theories for determining the liquid holdup in the liquid slug zone R, ,
slug length I , , and slug frequency are also far from being perfect, and
considerable work, both experimentally as well as theoretically, should be
carried out along these lines.
In spite of the aforementioned deficiencies, of which the reader should
be aware, relative simplicity for practical applications is absolutely essen-
tial. We do believe that the present model is a reasonable compromise of
solid physics and ease of calculations. The scheme presented here is suf-
ficiently flexible to accommodate improvements once more accurate
theories or more up-to-date data is available.
111. Severe Slugging
A. TRANSIENT
PHENOMENA
IN SLUGFLOW
In the previous section steady-state slug flow was considered. In steady

slug flow, one is expected to see regular slugs propagating in the pipe.
These slugs are of relatively short length (less than 1000) and separated
by regular and evenly spaced elongated bubbles. There are many occasions
in which the nature of the slug flow is different than the steady-state flow
and has a nonsteady behavior.
An example of transient slugging was given by Taitel et al. (1978) when
considering unsteady flow of liquid and gas in horizontal pipes. It was
shown that one may get temporary slugging of a different nature than that
of steady-state slug flow. For example, when one has stratified flow in a
pipe and the flow rate of gas is increased slowly, a transition to annular flow
will occur. If, however, the increase of the gas flow rate occurs fast, a
temporary very long slug will be generated. The occurrence of this phe-
nomenon is easy to explain. In stratified flow the amount of liquid in the
pipe is much higher than for the case of annular flow. When suddenly
increasing the gas flow rate, there is too much liquid in the pipe for a
smooth transition to annular flow. The excess of the liquid is depleted out
of the pipe in the form of a very long slug that is being pushed by the gas
behind it.
Another example is given by Scott (1987) and Scott et al. (1987). In this
work, the existence of very large slugs in the Prudhoe Bay 5-km test
pipeline in Alaska was explained on the basis of transient pressure fluctua-

tions near the stratified-slug transition boundary. The flow that started as
stratified wavy established a high liquid content in the pipe. Due to interfa-
cial instability a slug is generated and starts to move downstream. The high
content of the liquid film ahead of the slug is consumed by the advancing
slug and causes a fast increase in the slug length. At the pipe entrance
stratified flow continue to take place. When a slug exits the pipe, an
increase in the pressure drop occurs, which is a cause for a sufficient
increase of the gas flow rate to trigger another new slug in the stratified
flow zone, and so on. The end result are long slugs that grow rapidly near
the entrance zone to sizes that are much longer than the slug size in normal
steady-state flow.
The most common unsteady slug flow that is very important for prac-
tical application is severe slugging or, as it is also called, terrain-induced
slugging. In the general case this type of slugging takes place when a
pipeline follows a hilly terrain and the liquid tends to accumulate at the
lower valleys blocking the gas passage. The gas upstream is compressed
while a long liquid slug grows in the valley. Eventually the pressure
upstream increases to the point that it overcomes the gravitational head of
the liquid and it pushes the liquid in the valley downstream in the form of
a long slug.
A general solution to such a problem is quite complex. The simple case,
however, that consists of a single pipeline and a riser, is a common
occurrence in offshore oil and gas production and has been analyzed quite
successfully. The result of this analysis will be detailed here.
B. SEVERESLUGGING
CYCLE
Figures 8-11 show the typical behavior of the severe slugging cycle. The
first step is the slug formation (Fig. 8). Liquid entering the pipe accumu-
lates at the bottom of the riser, blocks the gas passage and causes the gas in
the pipeline to compress. When the liquid height in the riser z reaches the
top of the riser z = h , the second step of slug movement into the separator
starts (Fig. 9). When the gas that is blocked in the pipeline reaches the
bottom of the riser the liquid slug is accelerated to high velocity owing to
rapid expansion of the gas in the pipeline. This step is termed blowout
(Fig. 10). In the last step, Fig. 11, the remaining liquid in the riser falls back
to the bottom of the riser and the process of slug formation starts again.
A model for severe slugging was first presented by Schmidt et al. (1980).
The purpose of such a model is to predict the slug length, slug cycle time,
and pressure fluctuations. The following analysis is a simplified version
based on the Schmidt et al. (1980) analysis.
TWO-PHASE
SLUGFLOW 117
FIG.8. Slug formation
Po
FIG.9. Slug movement into the separator.

118 YEHUDA
PO
FIG.10. Blowout.
FALLING FILM -
GAS
FIG.11. Liquid fallback.
Since the severe slugging phenomenon is typical of low flow rates of

liquid and gas, the pressure is dominated by gravity and the frictional
contribution is neglected. The liquid is considered incompressible and the
gas is assumed to behave as an ideal gas. Mass inlet flow rates of the liquid
and the gas is assumed to be constant. With reference to Fig. 8, x ( t ) and
TWO-PHASE
SLUG FLOW 119
z ( t ) can be calculated using the following relations:

Conservation of liquid and gas yields
mL = mLi + J1: dt (63)
1:
AULsPL
mG = mGi + AuGSOpGO df (64)
where mLi is the initial value of the liquid in the system and mGi is the
initial value of the gas in the pipeline. The determination of these initial
values will be discussed later.
The mass of liquid and gas can be written in terms of the values of x and
z as
mL = pLA(x + z ) + (1 - a)pLA(I- x) (65)
( L A ) is an additional gas volume that may exist between the gas inlet valve
and the liquid inlet. Usually L is zero for most practical applications. It is,
however, convenient to use L > 0 in experimental facilities (Taitel et al.,
1989) to simulate longer pipeline performance when the actual pipeline
length 1 is short. Note that mLiand mGiare given by Eqs. (65) and (66) for
x = xi and z = z i . Equations (63) and (64) are sufficient to solve for x , z,
and Pp as a function of time provided the initial values of x and z (xi and zi),
and the void fraction in the pipeline a are known. The determination of
these initial values will be discussed later.
Substituting mG and mGifrom Eq. (66) in Eq. (64) yields for the gas,
- + ( z - x sin p )
1
[(I - x)a + L]
+ (zi - xi sin p )
1[(I - xi)a+ L ]+
Substituting mL and mLifrom Eq. (65) into Eq. (63) yields, for the liquid,
UGSOPGO dt (67)
z = zi - ( Y ( X - xi) +
Id uLS dt
Equations (67) and (68) can be solved now for x ( t ) and z ( t ) , which
(68)
correspond to the slug formation step (Fig. 8). Once the slug reaches the
top of the riser (z = h ) the process is continued as shown in step 2 (Fig. 9).
Thus, after z = h the solution for x ( t ) is obtained directly from Eq. (67)
with z = h.
120 YEHUDA
The initial values of xi and zi depend on the amount of liquid that stays in
the riser at the end of the blowout process (Figs. 10 and 11). The blowout
process is usually a highly chaotic phenomenon and the prediction of the
liquid fallback is difficult. Schmidt et al. (1980) used an experimental
correlation to estimate the amount of fallback. Taitel (1986) assumed that
the blowout process is in the form of a single fast-moving Taylor bubble. In
this case the liquid film left in the riser can be calculated on the basis of a
slug-flow model. The result of such calculations showed that the void
fraction of such a Taylor bubble a’ is around 10%. Assuming that at the
beginning of the fallback the pressure in the pipeline is P, , that the falling
liquid blocks the air passage, and that the fallback is very fast, then one can
calculate xi, zi , and Ppusing the following relations: Hydrostatic pressure:
Pp= P, + pLg(zi- xi sin p) (69)
Liquid mass balance requires
ax1 + zi = (1 - a’)h
while the compression of the gas in the pipeline follows the relation
la + L
Pp= P,
(1 -Xi). +L
The calculation of the void fraction a in the pipeline can be calculated
using a steady-state stratified flow in an inclined pipeline (Taitel and
Dukler, 1976). Furthermore, since the flow of the liquid and gas is usually
low (this is the a priori condition for severe slugging), the void fraction can
be calculated as in an open channel flow. In this case a momentum balance
of shear stress and gravity on the liquid phase yields
(72)
where
(73)
The friction factor fL can be calculated from the Moody diagram with the
appropriate hydraulic diameter. For smooth pipes, for example, the fric-
tion factor can be calculated by
where CL= 0.046 and n = 0.2 for turbulent flow, and CL= 16 and n = 1 for
laminar flow. The cross-sectional area AL and the wetting periphery SLare
given in terms of the equilibrium liquid level hL [see Eqs. (39) and (41)].
Equation (73) can now be solved for the equilibrium level hL. Once hL is
TWO-PHASE
SLUGFLOW 121
given the void fraction a can be calculated by

= 1 - (AJA) (75)
The theory presented here provides sufficient means to calculate some of
the major parameters of severe slugging such as the fluctuation of the
pressure in the line Pp(t),the length of penetration of the liquid into the
pipeline x ( t ) , the slug length that enters the separator, and the cycle time of
the process.
C. BOE’SCRITERION
FOR SEVERE
SLUGGING
The severe slugging pattern is typical of relatively low liquid and gas flow
rates. It requires that the flow pattern in the pipeline be stratified. Thus,
one condition for the existence of severe slugging is that the flow pattern in
the inclined pipeline will be in the stratified flow pattern. For the deter-
mination of this condition, one needs to use flow pattern maps or any flow
pattern prediction methods (Taitel and Dukler, 1976; Barnea, 1987).
In addition to this condition, the existence of a severe slugging cycle
requires that the liquid will penetrate into the pipeline, namely, x > 0
(Boe, 1981). This requirement is usually satisfied for a relatively low gas
flow rate. Referring to Fig. 8, the condition for x to stay at zero is when the
increase of the pressure owing to the addition of liquid into the riser is
balanced by the increase in the pipeline pressure due to the addition of gas.
The increase of pressure owing to the addition of liquid is
dPp/dt = PLddZ/dt) = PLWLS (76)
The increase of pressure owing to the addition of gas is
dPp - ki UGSOPGO PpUGS
RT=- RT=- (77)
dt V, la + L la + L
Equating the right-hand sides of Eqs. (76) and (77) yield the transition
boundary proposed by Boe between the severe slugging pattern and a
steady flow in the riser (usually bubbly or slug flow):
Equation (78) can also be derived on the basis of our previous develop-
ment. Setting x and xi to zero in Eq. (67) leads to the same condition as
Eq. (78).
Equation (78) is shown by the boundary A on Fig. 12 for a specific
example reported by Taitel et al. (1989). Note that at low liquid flow rates,
lot 1 I I I r . t t 1 I I I 1 I I I ' I I 1 ' 1 u-~q
STEADY FLOW
/
uLs is a monotonic linear function of the gas inlet flow rate uGSO.For high
liquid flow rates a approaches 0 and the curve is bent to the left. Note,
however, that a here is calculated while neglecting the gas shear
[Eq. (72)]. Thus this upper limit is beyond the applicability of the present
calculations.
Boe claimed that outside the region bounded by A the flow will be of
steady-state nature while inside severe slugging will prevail. This claim,
however, will be shown to be not quite accurate. In fact the Boe criterion
may be violated and one may get steady-state flow within the region
designated by Boe as severe slugging and vice versa, one can get severe
slugging in the region designated by the Boe criterion as steady-state flow.
The occurrence of such anomalies will be discussed next.
D. STABILITY
CRITERION
The stability criterion addresses itself to the blowout step of the severe
cycle process. As discussed earlier the blowout process (Fig. 10) was
assumed to take place in the form of a spontaneous expansion of the gas in
the pipeline. Indeed this is usually the case. The criterion for determining
the condition under which a vigorous blowout will occur versus a quasi-
equilibrium penetration is termed here the stability criterion (Taitel, 1986).
TWO-PHASE
SLUGFLOW 123
Assume that the cycle of severe slugging reaches the point at which the
liquid slug tail has just entered the riser. Assume a small disturbance y that
carries the liquid somewhat higher (see Fig. 10) and that the disturbance is
fast enough so that the slow flow rates of liquid and gas are ignored while y
changes.
The net force (per unit area) acting on the liquid in the riser is
The first term on the right-hand side is the pipeline pressure driving force.
The pressure varies with y as a result of the expansion of the gas in the
pipeline. The second term corresponds to the back pressure force applied
by the separator pressure and the liquid column of density pL and height
( h - y ) . Note that for y = 0 the system is in equilibrium and A F = 0. In
Eq. (79) a’ is the gas holdup in the gas cap penetrating the liquid column;
a‘ can be estimated on the basis of a slug-flow model. Also, the gas is
assumed to expand isothermally following the ideal gas law.
The liquid column will be blown out of the pipe if A F increases with y .
Thus, the condition for stability is
d(AF)/dy<O at y = O (80)
This leads to the criterion for stability
P, [ ( a l +L ) / a ’ ]- h
->
Po PolPLg
where Po is the atmospheric, or reference, pressure.
Equation (81) is shown in Fig. 12 by boundary B, which divides the
region bounded by the Boe criterion into two subregions. The region
below line B is unstable and the blowout process is vigorous. The region
above B is characterized by a quasi-equilibrium penetration of the gas into
the liquid. Taitel el al. (1989) showed that this penetration can end up
either with steady flow in the riser or it can develop into a cyclic operation.
The latter is termed quasi-equilibrium severe slugging (to be discussed in
the next section).
The stability criterion [Eq. (Sl)] was applied within the region bounded
by curve A (Boe criterion). However, this criterion can also be used
outside this region where a steady-state flow is assumed to take place
(Taitel, 1986). Indeed, it can be shown that an unstable subregion exists
outside Boe’s region. In this region a severe slugging process will take
place as follows: gas in the pipeline will spontaneously expand into the riser
and a blowout will occur, followed by liquid fallback. Thereafter, gas will
124 YEHUDA
continue to penetrate into the riser and bubble through it while the liquid
(mixture) level in the riser, 2 , rises towards the top of the riser. At the time
the liquid level reaches the top of the riser, a steady state is expected to
ensue. However, because of the inherent lack of stability, blowout will
reoccur. This gives rise to a cyclic severe slugging process except that the
slugs produced into the separator are aerated and shorter than the riser
length, unlike the classic severe slugging.
The criterion for the existence of severe slugging under such conditions
is obtained using Eq. (81) in which pL is replaced by pL&, where & is the
average liquid holdup in the riser under steady-state conditions (the gas
density can be ignored).
The value of is obtained by using Eq. (9). Assuming unaerated liquid
slugs in the riser, the value of average liquid holdup under steady-state
condition is
- = uGS/[c(uGS + uLS) + ud] (82)
Note that Eq. (82) is also valid for bubbly flow in the riser in which case C
and u d are replaced by B and u o .
The gas superficial velocity in the riser, adjusting for the average pres-
sure in the riser, is given by
P,
Equations (82) and (83) yield the liquid holdup 5 in steady state. The
stability of this steady state can be evaluated by Eq. (81) (using (a pL). The
line of marginal stability is shown in Fig. 12 for the case of slug flow by
line C. As can be seen, there is a definite region in which one can obtain
unstable steady-state flow outside Boe's region. As a result the flow will be
cyclic similar to the severe slugging cycle. We term this cyclic behavior as
unstable oscillations.
E. QUASI-EQUILIBRIUM
SEVERESLUGGING
The region above line B in Fig. 12, although found to be stable accord-
ing to Eq. (81), may behave in a cyclic fashion termed quasi-equilibrium
steady state. In this case it is possible to calculate and predict the behavior
of the riser during this process, enabling also to predict whether the system
will end up in a steady state or a cyclic operation.
The analysis begins at the point when the riser is full of liquid and gas is
just entering the bottom of the riser under equilibrium conditions. We
assume that the condition is stable so that no blowout occurs as a result of
the penetration of the gas into the riser. Nevertheless, when gas enters the
SLUGFLOW
TWO-PHASE 125
riser the hydrostatic pressure at the bottom of the riser decreases. This
causes an expansion of the gas in the pipeline. As a result the mass flow
rate of gas into the riser thG increases. Assuming ideal gas behavior, the
instantaneous mass flow rate into the riser can be calculated by
The pressure in the pipeline (and at the bottom of the riser) is the
hydrostatic pressure exerted by the weight of the liquid column in the riser
(the gas weight is neglected). Designating the local liquid holdup in the
pipe as @, one obtains that
pp = ps +
i:'
@PLgdY (85)
The gas that penetrates the bottom of the riser is in the form of either
small bubbles or larger Taylor bubbles. In either case it is assumed that the
gas velocity equals the translational velocity, which is given in the form of
Eq. (37).
In Eq. (37) us is the mixture velocity in the liquid slug. Note that
this mixture velocity is equal to the total superficial velocity,
+
us(us = uLs uGS).In order to simplify the problem a constant superficial
velocity us is assumed. For this purpose we calculate the average gas
density as
As can be seen in Eq. (86) the average gas density is calculated based on
the local pressure in the riser weighted by the local gas void fraction
(1 - a). The local pressure is given by
P(Y) = ps + [@PLgdY (87)

Using Eq. (86), the superficial gas velocity in the riser is
Note that although us is assumed to be constant along the pipe, it is a

function of time.
The liquid holdup at the bottom of the riser is given by
@ b = 1 - (uGS/ut) (89)
The local liquid holdup in the riser is determined by simple propagation
of the liquid holdup at the bottom of the riser with a velocity u,. Thus, the
126 YEHUDA
local liquid holdup is calculated by
@(y)=@b On y=jdu,dt (90)
This mathematical formulation allows one to calculate the variation of

the pipeline pressure, gas mass flow rate into the riser as a function of time,
and the local instantaneous liquid holdup in the riser @(y,t). Although the
formulation is somewhat complex, it is very simple to program using an
explicit Lagrangian numerical scheme described next.
At time t = 0 the riser is full of liquid, @ = 1 and riZG = riZGin. The
average density of the gas at this time is the inlet density. The gas super-
ficial velocity is given by Eq. (88) and the translational velocity is calcu-
lated by Eq. (37). The riser is subdivided into small segments of length Ah
and the time step At is calculated using At = A h / u , .
After time At, (at the bottom of the riser, =@b) is given by Eq. (89);
the new pressure is given by Eq. (85); the new average gas density in the
riser is given by Eq. (86); and the new gas mass flow rate into the riser is
given by Eq. (84). Note that dP/dt in Eq. (84) is approximated numerically
by the difference between the new and old pressure divided by At. Once
the new ritG is known, the new gas superficial velocity uGSis calculated
from Eq. (88) along with the new translational velocity u, from Eq. (37)
and the new time step At (At = Ah/u,).
At the next time step, @ j + l is set equal to Qj and this takes care of the
propagation of the bubbles in the riser; is calculated as before.
This analysis can be used provided the penetration of the gas into the
riser hGis always positive (which leads finally to a steady-state flow).
Boundary D on Fig. 12 is the curve above which the flow will indeed reach
a steady state. As seen, a steady state can take place within the Boe region.
In the particular example of Fig. 12 boundary D is very close to the upper
Boe region. Obviously the region of steady flow within the Boe region can
be larger for different operating conditions. Below boundary D ritG is not
always positive, resulting in penetration of the liquid into the pipeline. Let
x ( t ) be the distance of the liquid interface penetrating into the pipeline.
Under hydrostatic equilibrium the pipeline pressure at any time is
pP= pLg(&t - x sin p) + P, (91)
where is the average liquid holdup in the riser. A mass balance on the
gas in the pipeline requires that [see Eq. (67)]
1
-+ ( 8 h - x sin p ) [ ( I -x)a + L]
= [2+ T i h ] ( I a+ L ) + - &in dt (92)
TWO-PHASE
SLUGFLOW 127
where i relates to the time when riZG = 0 and penetration of liquid into the
pipe starts.
Equation (92) can be solved for x as a function of time. For this purpose
the average liquid holdup & should be known as a function of time. The
variation of & with time can be calculated as before on the basis of the
translational velocity ut from Eq. (37). The calculation of the mixture
velocity us is then calculated on the basis of the liquid mass balance to
yield
U S = uLS - (Y (dxldt) (93)
At time ti, x = 0 and us = uLs ( I j l G = uGS= 0). For time step At, we then
calculate the new @ distribution in the riser and &, the new x , the new us
(approximately dx/dt numerically), the new u,, and the new step At. As in
the case of severe slugging, x increases to a maximum and than recedes
back to zero. When x = 0 the cyclic process is repeated.
This calculation is valid provided no fallback occurs. A condition of
fallback is defined when the top of the riser becomes clear of liquid (or
liquid mixture) and a visible liquid interface is propagating towards the top
of the riser. The condition of fallback is related to the net liquid velocity at
the top of the riser. Once the liquid velocity at the top of the riser is less
than zero no liquid exits the riser, resulting in fallback of the liquid in the
riser. Thus, the point at which fallback occurs is when uL is negative, where
uL at the top is given by (simple mass balance)
Once this situation occurs we calculate the liquid height in the riser by
z = zi = &h and the calculation proceeds in the exact manner described
before for the classic severe slugging. In this calculation x ( t ) as well as z(t)
are calculated on the basis of two equations, Eq. (95), which is a mass
balance on the gas [similar to Eqs. (92) and (67)],
1
-+ ( z - x sin p ) [ ( l - X ) ( Y + L]
[(l - Xi). + L ] + A-
RT
PLgI: riZGin dt (95)
and Eq. (96), which is a mass balance on the liquid
z = zi - CX(X - xi) +
Equations (95) and (96) are used to calculate x ( t ) and z(t).Once the slug
reaches the top of the riser, then z = h, and x ( t ) is calculated by Eq. (95)
only. The values of xi and zi are the values of x and z at the time of
fallback, namely when uL becomes negative. As in the previous case,
once x recedes to zero, the gas penetrates the riser and the cycle is
repeated.
F. SUMMARY
AND CONCLUSIONS
The severe slugging that consists of one riser and one pipeline is perhaps
one of the simplest examples of slug flow under nonsteady conditions. As is
evident, even this simple case is not at all trivial and presents quite a
number of possible operating conditions. A more detailed discussion and
experimental verification of the present theory is given by Taitel et al.
(1989) and Vierkandt (1988). A summary of the results is presented using
an example of a typical flow map as shown in Fig. 12. This map contains
four boundaries: A-the Boe criterion, B-the stability criterion, C-the
steady-state stability criterion, and D-the transition to steady flow inside
the Boe criterion.
The Boe criterion [Eq. (78)] differentiates quite well between steady and
cyclic operations with two exceptions. At high liquid flow rates, a steady
flow can also exist within the severe slugging region predicted by the Boe
criterion (above boundary D). Also there is a region outside the Boe
criterion that is in an unsteady state and leads to unsteady oscillations (be-
tween boundaries C and A).
The stability criterion [Eq. (8l)l is applied to the cases of severe slugging
(inside the Boe region) where the riser contains only liquid (B), and to the
case of steady flow of liquid and gas in the riser (C) (outside the Boe
region). The former is an approximate boundary dividing between classical
and quasi-equilibrium severe slugging cyclic operations. The latter indi-
cates when steady flow outside the Boe criterion is not possible and one
obtains unsteady oscillations.
Unlike boundaries A, B, and C, which are given by simple equations,
the condition for boundary D is a more complex one and is obtained during
a numerical solution of the quasi-equilibrium case as a dividing line be-
tween cases where the gas flow rate into the riser is always positive
(hG > 0) and the cases where riZG reaches zero in the cyclic process. Note
that in Fig. 12, curve D is very close to the upper boundary of Boe
criterion. This is not always the case and, in fact, this boundary can be
substantially lower and also higher than the Boe criterion (in which case
it is not applicable), depending primarily on the length of the pipeline
(1 and L ) (Taitel et al., 1989).
SLUGFLOW
TWO-PHASE 129
NOMENCLATURE
A pipe cross-sectional area V volume
b interface width in the pipe X coordinate in the downstream
B constant in Eq. (38) direction, also distance of liquid
C constant in Eq. (37) and in the penetration into the pipeline
friction factor correlation X mass flow rate relative to the
D pipe diameter translational velocity
d bubble diameter Y coordinate in the perpendicular to
f friction factor the pipe axis direction, also
F force vertical coordinate in the riser
g acceleration of gravity W mass flow rate
h liquid level and riser height z coordinate in the upstream
1 pipe length direction, also liquid height in the
L additional equivalent gas pipeline riser
length 0 void fraction
m mass P angle of inclination
n constant in the friction factor Y polar angle that defines the interface
correlation s film thickness
r pipe radius E roughness
P pressure 0 polar coordinate
4 local absolute velocity A liquid volumetric quality, uLs/us
R liquid holdup, also ideal gas P viscosity
constant V kinematic viscosity, also frequency
Re Reynolds number P density
S wetted periphery L7 surface tension
t time I surface tension parameter,
T absolute temperature dl[(PL - P&*I
U velocity in the axial direction 7 shear stress
U free-steam velocity @ liquid holdup in the riser
V relative velocity, usually relative to
the translational velocity
Subscripts and Superscripts

acc accelaration L liquid
b bubble mix related to mixing
C critical 0 free rise, also at standard
d drift atmospheric conditions
E equilibrium P pipeline
f film S slug, also separator
fe film at z = If S superficial
fi film at z = 0 t translational
G gas top top of the riser
h hydraulic, also horizontal U slug unit
i interface, also initial V vertical
in inlet 30 for unbounded liquid
i discretization index in the riser
Special Symbols
0 rate - average
130 YEHUDATAITEL
AND DVORA
BARNEA
REFERENCES
Akagawa, K., and Sakaguchi, T. Fluctuation of void ratio in two-phase flow (2nd report,
analysis of flow configuration considering the existence of small bubbles in liquid slugs).
Bull. JSME 9, 104-110 (1966).
Andritsos, N., and Hanratty, T. J. Interfacial instab es for horizontal gas-liquid flows in
pipelines. Int. J. Mulfiphase Flow 13, 583-603 (1987).
Barnea, D. A unified model for predicting flow pattern transitions in the whole range of
pipe inclination. Inf. J . Mulfiphase Flow 13, 1-12 (1987).
Barnea, D. Effect of bubble shape on pressure drop calculations in vertical slug flow. Int. J.
Multiphase Flow in press (1989).
Barnea, D., and Brauner, N. Hold-up of the liquid slug in two phase intermittent flow. Int. J .
Mulfiphase Flow 11, 43-49 (1985).
Barnea, D., and Shemer, L. Void fraction measurements in vertical slug flow: Applications to
slug characteristics and transition. Int. J. Mulfiphase Flow 15, 495-504 (1989).
Barnea, D., Shoham, O., and Taitel, Y. Flow pattern characterization for two phase flow by
electrical conductance probe. Int. J. Multiphase Flow 6 , 387-397 (1980).
Barnea, D., Shoham, O., Taitel, Y., and Dukler, A. E. Gas-liquid flow in inclined tubes:
Flow pattern transitions for upward flow. Chem. Eng. Sci. 40, 131-136 (1985).
Barr, G. The air-bubble viscometer. Philos. Mag. 1, 395-405 (1926).
Bendiksen, K. H. An experimental investigation of the motion of long bubbles in inclined
tubes. Int. J . Multiphase Flow 10, 467-483 (1984).
Bendiksen, K. H . On the motion of long bubbles in vertical tubes. Int. J . Multiphase Flow 11,
797-812 (1985).
Benjamin, T. B. Gravity currents and related phenomena. J . Fluid Mech. 31, Part 2,
209-248 (1968).
Boe, A. Severe slugging characteristics. Sel. Top. Two-Phase Flow, NTH, Trondheim, Norw.
(1981).
Bonnecaze, R. H., Eriskine, W., Jr., and Greskovich, E. J. Holdup and pressure drop for
two-phase slug flow in inclined pipelines. AIChE J . 17, 1109-1113 (1971).
Brauner, N., and Barnea, D. Slug/churn transition in upward gas-liquid flow. Chem. Eng.
Sci. 41, 159-163 (1986).
Brodkey, R. S “The Phenomena of Fluid Motions.” Addison-Wesley, Reading, Mas-
sachusetts, (1967).
Cohen, S. L., and Hanratty, T. J. Effects of waves at a gas-liquid interface on a turbulent air
flow. J . Fluid Mech. 31, 467-469 (1968).
Collins, R., De Moraes, F. F., Davidsom, J. F., and Harrison, D. The motion of a large
gas bubble rising through liquid flowing in a tube. J . Fluid Mech. 89, Part 3, 497-514
(1978).
Davies, R. M., and Sir Taylor, G. (F. R. S). The mechanics of large bubbles rising through
extended liquids and through liquids in tubes. Proc. R. SOC. London, Ser. A 200,
375-390 (1949).
Dukler, A. E., and Hubbard, M. G. A model for gas-liquid slug flow in horizontal and near
horizontal tubes. Ind. Eng. Chem. Fundam. 14, 337-347 (1975).
Dukler, A. E., Moalem-Maron, D., and Brauner, N. A physical model for predicting the
minimum stable slug length. Chem. Eng. Sci. 40, 1379-1385 (1985).
Dumitrescu, D. T. Stromung an einer luftblase im senkrechten rohr. Z . Angew. Math. Mech.
23, 139-149 (1943).
Fernandes, R. C . Experimental and theoretical studies of isothermal upward gas-liquid flows
in vertical tubes. Ph. D. Thesis, University of Houston, Houston, Texas, (1981).
SLUGFLOW
TWO-PHASE 131
Fernandes, R. C., Semiat, R., and Dukler, A. E. Hydrodynamic model for gas-liquid slug
flow in vertical tubes. AIChE J . 29, 981-989 (1983).
Goldsmith, H. L., and Mason, S. G . The movement of single large bubbles in closed vertical
tubes. J . Fluid Mech. 14, 42-58 (1962).
Govier, G. W . , and Aziz, K. “The Flow of Complex Mixtures in Pipes.” Van-Nostrand-
Reinhold, Princeton, New Jersey, 1972.
Gregory, G. A., and Scott, D. S . Correlation of Liquid slug velocity and frequency in
horizontal cocurrent gas-liquid slug flow. AIChE J . 15, 933-935 (1969).
Gregory, G. A., Nicholson, M. K., and Aziz, K. Correlation of the liquid volume fraction in
the slug for horizontal gas-liquid slug flow. Int. J . Multiphase Flow 4, 33-39 (1978).
Greskovich, E. J., and Shrier, A. L. Pressure drop and holdup in horizontal slug flow.
AIChE J . 17, 1214-1219 (1971).
Greskovich, E. J., and Shrier, A. L. Slug frequency in horizontal gas-liquid slug flow. Ind.
Eng. Chem. Process Des. Dev. 11, 317-318 (1972).
Hall, N. A. “Thermodynamics of Fluid Flow.” Longmans, Green, New York, (1957).
Harmathy, T. Z . Velocity of large drops and bubbles in media of infinite of restricted extent.
AIChE J . 6 , 281-288 (1960).
Hasan, A. R., and Kabir, C. S . Predicting multiphase flow behavior in a deviated well. 61st
Annu. Tech. Conf. New Orleans, La. SPE 15449 (1986).
Henderson, F. M. “Open Channel Flow.” Macmillan, New York, (1966).
Heywood, N. I., and Richardsom, J. F. Slug flow of air-water mixtures in a horizontal pipe:
determination of liquid holdup by gama-ray absorption. Chem. Eng. Sci. 34, 17-30
(1979).
Hubbard, M. G. An analysis of horizontal gas-liquid slug flow. Ph. D. Thesis, University of
Houston, Houston, Texas, 1965.
Kouba, G. E . Horizontal slug flow modeling and metering. Ph. D. Thesis, University of
Tulsa, Tulsa, Oklahoma, 1986.
Levich, V. G. “Physicochemical Hydrodynamics.” Prentice-Hall, Englewood Cliffs, New
Jersey, 1962.
Mandhane, J . M., Gregory, G . A,, and Aziz, K. A flow pattern map for gas-liquid flow in
horizontal pipes. Int. J . Multiphase Flow 1, 537-553 (1974).
Marrucci, G. An interpretation of slip in horizontal gas-liquid slug flow. Chem. Eng. Sci. 21,
718-719 (1966).
Moissis, R. The transition from slug to homogeneous two-phase flows. J . Heat Transfer 85,
366-370 (1963).
Moissis, R., and Griffith, P. Entrance effects in a two-phase slug flow. J . Heat Transfer 84,
29-39 (1962).
Nicholson, !:. K., Aziz, K., and Gregory, G . A. Intermittent two phase flow in horizontal
pipes: predictive models. Can. J . Chem. Eng. 56, 653-663 (1978).
Nicklin, D. J . Two-phase bubble flow. Chem. Eng. Sci. 17, 693-702 (1962).
Nicklin, D. J., Wilkes, J. O., and Davidson, J. F. Two-phase flow in vertical tubes. Trans.
Inst. Chem. Eng. 40, 61-68 (1962).
Orell, A., and Rembrand, R. A model for gas-liquid slug flow in a vertical tube. Ind. Eng.
Chem. Fundam. 25, 196-206 (1986).
Schmidt, Z . Experimental study of two-phase slug flow in a pipeline-riser pipe system. Ph. D.
Thesis, University of Tulsa, Tulsa, Oklahoma, 1977.
Schmidt, Z., Brill, J . P., and Beggs, H. D. Experimental study of severe slugging in a
two-phase flow pipeline-riser pipe system. SOC. Pet. Eng. J . 20, 407-414. (1980).
Scott, S. L. Modeling slug growth in pipelines. Ph. D. Thesis, University of Tulsa, Tulsa,
Oklahoma. 1987.
132 YEHUDATAITEL
AND DVORABARNEA
Scott, S. L., Shoham, O., and Brill, J. P. Modelling slug growth in large diameter pipes.
Proc. Int. Conf. Multi-Phase Flow, 3rd, The Hague pp. 55-63 (1987).
Shemer, L., and Barnea, D. Visualization of the instantaneous velocity profiles in gas-liquid
slug flow. Physicochem. Hydrodyn. 8, 243-253 (1987).
Shoham, O., and Taitel, Y. Stratified turbulent-turbulent gas liquid flow in horizontal and
inclined pipes. AIChE J. 30, 377-385 (1984).
Singh, G., and Griffith, P. Determination of the pressure drop optimum pipe size for a
two-phase slug flow in an inclined pipe. J. Eng. Ind. 92, 717-726 (1970).
Stanislav, J. F., Kokal, S., and Nicholson, M. K. Intermittent gas-liquid flow in upward
inclined pipes. Int. J . Multiphase Flow 12, 325-335 (1986).
Sylvester, N. D. A mechanistic model for two phase vertical slug flow in pipes. J. Energy
Resour. Technol. 109, 206-213 (1987).
Taitel, Y. Stability of severe slugging. Int. J . Multiphase Flow 2, 203-217 (1986).
Taitel, Y., and Barnea, D. A consistent approach for calculating pressure drop in inclined
slug flow. Chem. Eng. Sci. submitted (1989).
Taitel, Y., and Dukler, A. E. A model for prediction flow regime transitions in horizontal
and near horizontal gas-liquid flow. AIChE J . 22, 47-55 (1976).
Taitel, Y., and Dukler, A. E. A model for slug frequency during gas liquid flow in horizontal
and near horizontal pipes. Int. J . Multiphase Flow 3, 585-596 (1977).
Taitel, Y., Lee, N., and Dukler, A. E. Transient gas-liquid flow in horizontal pipes-
modeling flow pattern transitions. AIChE. J. 24, 920-935 (1978).
Taitel, Y., Barnea, D., and Dukler, A. E. Modelling flow pattern transitions for steady
upward gas-liquid flow in vertical tubes. AIChE J. 26, 345-354 (1980).
Taitel, Y., Vierkandt, S., Shoham, O., and Brill, J. P. Severe slugging in a pipeline-riser
system, experiments and modeling. Int. J. Multiphase Flow in press (1989).
Taylor, G. I. Deposition of a viscous fluid on the wall of a tube. J. Fluid Mech. 10, 161-165
(1961).
Vierkandt, S. Severe slugging in a pipeline-riser system, experiments and modeling. M. S.
Thesis, University of Tulsa, Tulsa, Oklahoma, 1988.
Wallis, G. B. “One Dimensional Two-Phase Flow.” McGraw-Hill, New York, 1969.
Wallis, G.B., Richter, H. J., and Bharathan, D. “Air-Water Countercurrent Annular Flow
in Vertical Tubes.” Rep. EPRI NP-786.Electr. Power Res. Inst. Palo Alto, Calif., 1978.
Wallis, G. B., Richter, H. J., and Bharathan, D. “Air-Water Countercurrent Annular Flow
in Vertical Tubes.” Rep. EPRI NP-1165.Electr. Power Res. Inst. Palo Alto, Calif.,
1979.
Weber, M: E.Drift in intermittent two-phase flow in horizontal pipes. Can. J. Chem. Eng.
59, 398-399 (1981).
Weisman, J., Duncan, D., Gibson, J., and Crawford, T. Effect of fluid properties and pipe
diameter on two-phase flow pattern in horizontal lines. Int. J . Multiphase Flow 5 ,
437-460 (1979).
Zuber, N.,and Findlay, J. A. Average volumetric concentration in two-phase flow systems.
J. Heat Transfer 87, 453-468 (1965).
Zuber, N.,and Hench, J. “Steady State and Transient Void Fraction of Bubbling Systems
and Their Operating Limit. Part I: Steady State Operation,” Rep. 62GL100. Gen.
Electr., Schenectady, N. Y., 1962.
Zukoski, E. E. Influence of viscosity, surface tension, and inclination angle on motion of long
bubbles in closed tubes. J. Fluid Mech. 25, 821-837 (1966).

Taitel Et. Al, 1990

Uploaded by

Copyright:

Available Formats

Taitel Et. Al, 1990

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Taitel Et. Al, 1990

Uploaded by

Copyright:

Available Formats

ADVANCES IN HEAT TRANSFER.

Two-Phase Slug Flow

YEHUDA TAITEL AND DVORA BARNEA

Faculty of Engineering, Department of Fluid Mechanics and Heat Transfer,

Gas-liquid flow in conduits may take on a variety of configurations

11. Steady-State Slug-Flow Modeling

FIG. 1. Slug-flow geometry.

velocity of the elongated bubble ut is the velocity at which the elongated

Exactly equivalent mass balances can be performed on the gas. How-

a , = ( a & + j--clfdx)/f, (7)

= (UGS - U b a s + U&s)/ut (9)

nate system moving with a velocity u , are

where vf = u, - uf and vG = u, - uG. Note that although these equati0n.s

? =fi[PGIuG - ufI(uG - uf)/2] (14)

Using Eq. (4), the relative velocities vf and vG are given by

Vf = (u, - Uf) = (u, - UL)Rs/Rf (19)

Equation (23) is still a differential equation that has to be integrated

as having a constant thickness in equilibrium. This equilibrium level is

+ p~R,A(ut- uL)(ufi - ufe) (30)

Two methods have been presented for the pressure-drop calculation. In

xl0" R fdz - pLg cos p

lohfAfdhf = (hf - y)b dy (33)

Equation (32) then takes the form

where pf = (YfpG + RfpL.

known variables, given in terms of known formulas, and postpone the

where ud is the drift velocity, namely, the velocity of propagation of a large

5. uf is extracted from Eq. (4).

It should also be mentioned that under most conditions T~ and q are

G. TRANSLATIONAL VELOCITIES OF ELONGATED (TAYLOR) BUBBLES

on channel flow approximation is not suitable for analyzing the immediate

FIG.2. Propagation of gas pocket in the draining horizontal pipe.

the pressure in the gas bubble, is taken as the reference pressure:

0.0001 0.001 0.01 0.1 1.o 10

The drift velocity is expected to depend also on the liquid viscosity, or

FIG.6. A comparison of predicted and measured distribution slip parameters C (0)

Reynolds number and it increases as the Reynolds number decreases. The

I. LIQUIDHOLDUPIN THE LIQUID

where the mixture velocity u s has units of meters per second.

Slug frequency has sometimes been considered as an entrance phe-

FIG.7. Velocity profiles in liquid slugs.

111. Severe Slugging

In the previous section steady-state slug flow was considered. In steady

pipeline in Alaska was explained on the basis of transient pressure fluctua-

FIG.8. Slug formation

FIG.9. Slug movement into the separator.

FIG.11. Liquid fallback.

Since the severe slugging phenomenon is typical of low flow rates of

z ( t ) can be calculated using the following relations:

mL = mLi + J1: dt (63)

mG = mGi + AuGSOpGO df (64)

given the void fraction a can be calculated by

lot 1 I I I r . t t 1 I I I 1 I I I ' I I 1 ' 1 u-~q

P(Y) = ps + [@PLgdY (87)

Note that although us is assumed to be constant along the pipe, it is a

local liquid holdup is calculated by

@(y)=@b On y=jdu,dt (90)

This mathematical formulation allows one to calculate the variation of

and Eq. (96), which is a mass balance on the liquid

Subscripts and Superscripts

You might also like