Nuclear Engineering and Design 204 (2001) 101– 115

www.elsevier.com/locate/nucengdes

On the stability of a one-dimensional two-fluid model

Jin Ho Song a,*, Mamoru Ishii b
a
Reactor Thermal Hydraulics and Safety Research, Korea Atomic Energy Research Institute, P.O. Box 105, Yusong Post Office,
Taejon, 305 -600, South Korea
b
School of Nuclear Engineering, 1290 Nuclear Engineering Building, Purdue Uni6ersity, West Lafayette, IN 47907 -1290, USA

Received 14 February 2000; accepted 14 February 2000

Abstract

An analysis on the stability of the governing differential equations for area averaged one-dimensional two-fluid
model is presented. The momentum flux parameters for gas and liquid are introduced to incorporate the effect of void
fraction profiles and velocity profiles. The stability of the governing differential equations is determined in terms of
gas and liquid momentum flux parameters. It is shown that the two-fluid model is well posed with certain restrictions
on the liquid and gas momentum flux parameters. Simplified flow configurations for bubbly flow, slug flow, and
annular flow are constructed to test the validity of proposed stability criteria. The momentum flux parameters are
calculated for these flow configurations by assuming a power-law profile for both velocity and void fraction. Existing
correlation for volumetric distribution parameter Co is used. By employing simplified velocity profiles, the void
fraction profile is determined from Co correlation. It is found that the void fraction is wall-peaked at low void fraction
and it becomes center-peaked as the void fraction increases. A simplified annular flow is also constructed. With these
flow configurations, the momentum flux parameters are determined. It is shown that the calculated momentum flux
parameters are located in the stable region above the analytically determined stability boundary. The analyses results
indicate that the use of momentum flux parameter is promising, since they reflect flow structure and help to stabilize
the governing differential equations. © 2001 Elsevier Science B.V. All rights reserved.

1. Introduction MOD3 (Ransom et al., 1995) and CATHARE

(Micaelli, 1988). The two-fluid model is robust
The use of an area averaged one-dimensional because it can treat liquid and vapor phase veloc-
two-fluid model (Ishii, 1975; Ishii and Mishima, ities separately and it allows thermodynamic non-
1984) is very useful for complicated engineering equilibrium between phases. However, the
problems. It is widely used in many computer two-fluid model is far from complete in the sense
codes for two-phase flow, such as RELAP5/ that the two-fluid model often causes numerical
instabilities and it still requires a lot of flow
regime dependent correlations.
* Corresponding author. Tel.: +82-42-8682850; fax: + 82-
42-8688990.
Recently, there were studies on the well-posed-
E-mail addresses: [email protected] (J.H. Song), ness of the two-fluid model. They investigated the
[email protected] (M. Ishii). validity of the one-dimensional two-fluid model

0029-5493/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 9 - 5 4 9 3 ( 0 0 ) 0 0 2 5 3 - 3
102 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115

by looking at the stability of governing differen- ŽŽuk =Žhkuk /Žhk  (3)

tial equations (Gidaspow, 1974; Stuhmiller, 1977;
Then, the one-dimensional area averaged conti-
Banerjee and Chan, 1980; Lahey et al., 1980;
nuity equation is presented as follows
Jones and Prosperetti, 1985; Sung and Chun,
1996). As pointed out by Gidaspow (1974) and ((Žhk zk )/(t + ((Žhk zk ŽŽuk )/(z = ŽYk  (4)
Jones and Prosperetti (1985), the governing differ-
For adiabatic air –water flow Yk = 0. Assuming
ential equations of the one-dimensional two-fluid
incompressibility of gas and liquid phase, Eq. (4)
model is ill posed as an initial value problem. To
becomes
resolve this intrinsic problem, several remedies
were suggested. Consideration of interfacial pres- (Žh/(t + ŽŽug(h/(z + Žh(ŽŽug/(z = 0
sure (Stuhmiller, 1977; Banerjee and Chan, 1980) (5)
and consideration of transient forces such as vir-
− (Žh/(t − ŽŽuf(Žh/(z
tual mass force (Lahey et al., 1980) were sug-
gested. However, the applicable range of their + (1− Žh)(ŽŽuf/(z = 0 (6)
models is limited and the way to model those
where h is the gas void fraction.
terms is still contradictory.
The momentum equation is written as
Here, we revisited the issue of the well-posed-
ness of the one-dimensional two-fluid model. A (Žhk zk ŽŽuk /(t + ((Žhk zkC6k ŽŽuk 2)/(z
characteristic analysis is performed for the gov-
= − Žhk (pk /(z − 4hkw~kw/D
erning differential equations for the one-dimen-
sional two-fluid model, where momentum flux + ((Žhk ŽŽ~kzz + ~ Tkzz)/(z + Žhk zk g cos q
parameters are incorporated to account for the
+(pki − pk )(hk /(z+ ŽŽ6kiŽYk + ŽMik z
effect of void fraction and velocity profiles of gas
(7)
and liquid phases. The stability criteria are newly
proposed in Section 3 and the proposed stability Cvk is momentum flux parameter that accounts
criteria are tested for simplified flow structures in for the variation of velocity and void fraction
a wide range of void fractions in Section 4. over a cross-section. It contains information on
the flow structure in the flow area normal to the
main flow direction.
2. One-dimensional two-fluid model Cvk = Žhku 2k/Žhk ŽŽuk 2 (8)

A one-dimensional two-fluid model can be ob- To investigate the stability of the above differ-
tained by integrating a three-dimensional two- ential equations we need to know the constitutive
fluid model over the cross-section and introducing relations for various terms on the right-hand side.
proper mean values (Ishii, 1975; Ishii and Here, we assume that various terms on the right-
Mishima, 1984). A simple area average over a hand side can be represented by algebraic expres-
cross-section is defined by sion of velocities, gas void fraction, and pressure

&
F= 1/A F dA (1)
except the pressure gradient and generalized drag.
Previous studies (Stuhmiller, 1977; Lahey et al.,
1980; Jones and Prosperetti, 1985) investigated
the role of these forces on the stability of govern-
and the void fraction weighted mean value is ing differential equations. They showed that the
given by generalized forces tend to stabilize the flow. By
ŽŽFk = ŽhkFk /Žhk  (2) neglecting the effect of the generalized forces, we
want to focus on the role of momentum flux
The density of each phase is considered to be parameter, which has information on flow struc-
uniform such that zk =ŽŽzk . The axial compo- ture in the cross-section, on the stability of two-
nent of the weighted-mean velocity of phase k is fluid model.
 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115 103

As there are two pressures, we need to have a where [A], [B], [C] are the coefficient matrices. To
closure relation for pressure. For dispersed flow investigate the behavior of this set of differential
we can approximate that two pressures are the equations, suppose that arbitrary data for x are
same. For the separated flow such as horizontal specified along a curve in the (z, t) plane. Intro-
stratified flow, the interfacial pressure is different duce tangential variable s(z, t) and normal vari-
from bulk pressure due to gravity. Banerjee and able n(z, t) along the curve, then the above
Chan (1980) showed that the two-fluid model equation can be transformed as
could be stable by considering the effect of inter-
{[A](n/(t + [B](n/(z}(x/(n
facial pressure. However, their argument is not
applicable to vertical annular flow, where gravity = [C] −{[A](s/(t + [B](s/(z](x/(s (14)
is in the flow direction. Here, we neglect the
difference between interfacial pressure and bulk Since x is given by the initial data as a function
pressure. Then our analysis results are applicable of s along the curve, the terms on the right-hand
to both dispersed flow and separated flow. side are known, and the derivative (x/(n will be
Let us simplify the terms on the right-hand side uniquely determined, if the determinant of the
including generalized force as Mig and assume coefficient matrix is non-singular.
that they are represented by algebraic terms. Let Determinant { [A](n/(t + [B](n/(z}" 0 (15)
us drop Ž, ŽŽ notations for convenience. Then
the continuity and momentum equation for one- The dependence of the solution on prescribed
dimensional two-fluid model can be presented as initial data can be reduced to an investigation of
follows the roots of equation

(h/(t +ug(h/(z+ h(ug/(z =0 (9) Determinant {[A]u − [B]}= 0 (16)

−(h/(t −uf(h/(z +(1 −h)(uf/(z =0 (10) where we have introduced the characteristic curve
u= − (n/(t/(n/(z.
hzg(ug/(t + zgug(h/(t + zgCvgugug(h/(z If we have real roots of u for satisfying Deter-
minant {[A]u − [B]}= 0, then the set of differen-
+2hzgCvgug(ug/(z = −h(p/(z +Mig (11)
tial equations is hyperbolic. If we have complex
(1 −h)zf(uf/(t − zfuf(h/(t − zfCvfufuf(h/(z conjugate root of u, then the set of differential
equations becomes elliptic. In this case, the above
+ 2(1−h)zfCvfuf(uf/(z = −(1 − h)(p/(z − Mig set of equations becomes ill posed as an initial
(12) value problem. From Eqs. (9) –(12), [A]u − [B] are
determined as
3. Change of type of the governing differential u− ug −h 0 0
equations − (u− uf) 0 −(1 −h) 0
zgug(u −Cvgug) hz(u −2Cvgug) 0 −h
The well-posedness of the governing differential
equation as an initial value problem can be ana- −zfuf(u −Cvfuf) 0 (1−h)zf(u −Cvfuf) 0

lyzed by a characteristic analysis (Gidaspow, (17)

1974; Stuhmiller, 1977; Jones and Prosperetti,
The determinant of this matrix becomes
1985). Here, we look at the stability of the two-
fluid model by performing a characteristic analy- Det (u)= − h(1− h)
sis for the governing differential equations and
× [(1−h)zg(uu− 2uCvgug + Cvgugug)
derive the stability criteria.
Let a vector x =(h, ug, uf, p), then the above + hzf(uu− 2Cvfufu+ Cvfufuf)] (18)
equation can be written as
If we do not consider the effect of void fraction
[A](x/(t +[B](x/(z =[C] (13) and velocity distribution across the pipe, the mo-
104 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115

mentum flux distribution parameters are equal to F(h, Cvg, Cvf, ug, uf, zf, zg)
1. Then above equation becomes
= [(1− h)zgCvgug + hzfCvfuf]
Det (u)= −h(1 − h) × [(1− h)zgCvgug + hzfCvfuf][(1− h)zg + hzf]
× [(1 − h)zg(u − ug) +hzf(u − uf) ]
2 2
× [(1−h)zgCvgugug + hzfCvfufuf]
(19)
=(Mg + Mf)2 − zmMgug − zmMfuf
The Det (u) = 0 can have real roots only if
= (Mg + Mf − zmug/2)2 − zmMfuf + zmMfug
u = ug = uf. It requires that gas and liquid speeds
should be equal. Otherwise, the two-fluid model − (zmug/2)2 ] 0 (22)
becomes ill posed. This was pointed out by
It requires that
Gidaspow (1974). For single-phase flow where h
equals 1 or 0, Eq. (18) becomes (Mg + Mf − zmug/2)2 ] (zmug/2)2 − zmMf(ug −uf)
(23)
Det (u) =zg(uu − 2uCvgug +Cvgugug),
If the right-hand side of Eq. (23) is negative, the
or zf(uu − 2Cvfufu +Cvfufuf) (20) above inequality is always met. Therefore, if the
following inequality is met, the flow is stable
It has real characteristics, u= Cvgug, or u=
4Mf(ug − uf)] zmu 2g (24)
Cvfuf. The governing equations are always hy-
perbolic and well posed as in the initial value Let S=ug/uf and call it the slip ratio and
problem. assume that liquid velocity is positive. Then, if S
Let us consider the effect of momentum flux is less than 1, the above inequality requires Cvf be
distribution parameters. To have real roots for u, negative. It is not physically possible. Also let
it is required from Eq. (18) that R= hzf/((1− h)zg) and call it the modified den-
sity ratio. Then Eq. (24) can be written as
F(h, Cvg, Cvf, ug, uf, zf, zg)
Cvf ] 14[1/R + 1]S 2/(S −1), S\1 (25)
= (1−h) zgugzgugCvg(Cvg −1)
2

If the right-hand side of Eq. (23) is positive,

+h 2zfufzfufCvf(Cvf −1) then the momentum fluxes should be related
+h(1− h)zgzf[2CvfufCvgug −Cvfufuf −Cvgugug] Mg ] 0.5zmug + [(zmug/2)2 − zmMf(ug − uf)]1/2
]0 (21) − Mf (26)
Eq. (21) indicates that the momentum flux dis- Mg 5 0.5zmug − [(zmug/2) − zmMf(ug − uf)]
2 1/2

tribution parameters can make the two-fluid

− Mf (27)
model have real roots without imposing the con-
dition that the liquid and gas velocity should be Let ug be positive, then
equal. It indicates that the velocity distributions
Cvg ] 0.5(1+R)
have some stabilizing effect. If we have positive
F(h), the two-fluid model equation allows two +0.5[(1+ R)2
real characteristic roots for u. The other two roots
− 4(1+ R)RCvf1/S 2(S−1)]1/2 − RCvf1/S
are related to the sound velocities of gas and
(28)
liquid flow. Since we assumed the incompressible
flow assumption for gas and liquid, they do not Cvg 5 0.5(1+R)
appear here.
−0.5[(1+ R)2
Let zm =(1 −h)zg +hzf, Mg =(1 −h)zgCvgug,
Mf = hzfCvfuf, then the above equation can be − 4(1+ R)RCvf1/S 2(S−1)]1/2 − RCvf1/S
rearranged as (29)
 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115 105

Eqs. (25), (28) and (29) define the boundary

between the stable and unstable region. It is very
interesting that the parameters R and S only
determine the stability boundary.
When S \ 1, the two-fluid model can be stable
if Cvf satisfies Eq. (25). If Cvf is less than that
specified by Eq. (25), Cvg and Cvf should satisfy
Eq. (28) or Eq. (29). When S 51, Cvg and Cvf
should satisfy Eq. (28) or Eq. (29).
Let us consider the case with S \ 1, this case is
practical for vertical air – water flow. Then Eq.
(25) indicates that if the liquid momentum flux
parameter is larger than a certain value, the flow
Fig. 3. Allowable gas momentum flux parameter depending on
liquid momentum flux parameter (Cvg(Cvf,2.0) stands for the
boundary of Cvg as a function of Cvf at R=10 and S= 2.0.
Cvg(Cvf,1.0) is at S= 1.0 and R =10. Above the curve is the
stable region).

is stable. Fig. 1 shows the variation of the liquid

momentum flux parameter for various values of
R.
Fig. 2 shows the boundaries of Eqs. (28) and
(29) at R= 100 and S= 2. The upper curve repre-
sents Eq. (28) and the lower curve represents Eq.
(29). The two-fluid model is stable when liquid
Fig. 1. Allowable liquid momentum flux parameter depending and gas momentum flux parameters are outside
on slip ratio (R= 1.0, 10, 100, above the curve is the stable the curve. The upper curve is concave and en-
region). closes the square defined by (0,0), (0,1), (1,0) and
(1,1). The lower curve does not seem to have
physical meaning. The upper curve indicates that
Cvg can be close to one only when Cvf is really
close to one.
The boundaries of momentum flux parameter
curves at R=10, and S=1, 2 are shown Fig. 3. If
we compare this curve with Fig. 2, it can be seen
that it allows more flexible changes in Cvg and Cvf.

4. Stability of simplified flow configuration

The morphology of two-phase flow, the flow

regime, can be represented by the velocity and
Fig. 2. Allowable gas momentum flux parameter depending on
liquid momentum flux parameter (Cvg(Cvf,2.0) is the
void fraction profiles. As the momentum flux
boundary as a function of Cvf at S= 2.0 and R =100 from parameters represent those changes in the one-
Eq. (30). Cvg1(Cvf,2.0) is from Eq. (31) at S= 2.0, R= 100). dimensional two-fluid model, it is interesting to
106 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115

investigate the role of the momentum flux Žh6 2g= ŽhŽ6g2 + [6ogm/(m + 2)]2/(m +1)Žh
parameters on the stability of two-phase flow. It
+Z6 2ogn/(n + 2)m/(m + 2)m/(m +1)
may be possible that the governing differential
equation for the two-fluid model can change type × 2/(2m +n+ 2)
similar to that of compressible transonic flow, or
× [1+ (2m + 2)/(m +n+ 2)]
it may not change type, if momentum flux
parameters vary in such a way that the two-fluid + D6og6wn/(n + 2)m/(m + 2)2/(m + n+2)
model becomes always stable. Here, we look at
(38)
the changes in momentum flux parameters and
stability of the two-fluid model for the simplified With a similar argument, the momentum flux
two-phase flow configurations, which include bub- parameter for liquid flow becomes
bly, slug, and annular flow.
Cvf = Žhf6 2f /ŽhfŽŽ6f2 = ŽhfŽhf6 2f /Žhf6f2
(39)
4.1. Bubbly flow and slug flow
Žhf6f= (1− Žh)Ž6f
Let us consider bubbly and slug flow first. The
− D6foq/(q + 2)n/(n + 2)2/(n +q+ 2)
void fraction and velocity profiles in bubbly and
(40)
slug flow in a pipe can be approximated by the
power-law profile (Ishii, 1977) as follows Žhf6 2f = Ž1 −hŽ6f2
h =hw +D(1 − r n) (30) + [6ofq/(q + 2)]2/(q +1)Ž1 − h
6g = 6wg +6og(1 − r m) (31) − D6 2ofn/(n + 2)q/(q +2)q/(q +1)
6f = 6wf +6of(1 − r q) (32) × 2/(2q + n+ 2)
where D= (ho −hw). ho is the void fraction at × [1+ (2q + 2)/(q +n+ 2)]
center, hw is the void fraction at the wall, r is the
− D6of6wfn/(n + 2)q/(q +2)2/(q + n+2)
non-dimensional radial position; 6w is the gas
(41)
velocity very near the wall, 6o is the gas relative
velocity at the center, 6wf is the liquid velocity very The above equations show that momentum flux
near the wall, and 6of is the liquid relative velocity parameters are functions of n, m, q, ho, hw, 6og,
at the center. The average void fraction and veloc- 6wg, 6wf, and 6of.
ity profiles are determined as There are many experimental data for gas and
liquid velocities and void fraction profiles (Ser-
Žh =hw + Dn/(n + 2) (33)
izawa et al., 1975; Herringe and Davis, 1976; van
ŽŽ6g = Žh6g/Žh (34) der Welle, 1985; Liu and Bankoff, 1993a,b). How-
ever, the data are not complete enough to deter-
ŽŽ6f = Ž6f −h6f/Ž1 −h (35)
mine the above parameters for the power-law
Then, the momentum flux distribution parameter profile. Therefore, we constructed rather sim-
for gas is defined as plified flow configurations to calculate momentum
flux parameters easily. It can be assumed that the
Cvg = Žh6 2g/ŽhŽŽ6g2 =ŽhŽh6 2g/Žh6g2
local gas and liquid velocities are the same, when
(36)
the relative velocity is negligibly small. Then, by
Assuming that the velocities of gas and liquid using the existing correlation for volumetric distri-
at the wall are negligibly small. bution parameter Co, we can construct velocity
and void fraction profiles. The effect of void
Žh6g= ŽhŽ6g
profile over a cross-section is typically presented
+ D6gon/(n + 2)m/(m + 2)2/(n +m+ 2) by the volumetric distribution parameter defined
(37) as
 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115 107

In Eq. (45), the distribution parameter Co is a

Co = Žhj/ŽhŽj (42)
function of average gas void fraction and density
where j is the volumetric flux. According to Ishii ratio. If the velocity profile is given, the power n
(1977), when the void fraction is less than 0.7, a of the void fraction profile and slip ratio are
lot of data for pipe flow can be correlated as uniquely determined by Co and average void frac-
Co = (1.2− 0.2 (zg/zf))(1 −e − 18Žh) (43) tion at a given density ratio. As slip-ratio S and
modified density ratio R determine the stability
By definition, the average slip ratio is deter- criteria described above, we can compare the mo-
mined as mentum flux parameters determined by this pro-
S = ŽŽ6g/ŽŽ6f = (1 −Žh)Co/(1 −CoŽh) cedure and those of stability criteria.
(44)
4.1.1. Center-peaked 6oid fraction profile
and assuming that liquid and gas velocity at the
Data from experimental research indicate that
wall is very close to zero, the volumetric distribu-
the void fraction has a power-law profile with
tion parameter for the power-law profile is calcu-
center-peaked shape when the void fraction is
lated as
between 0.2 and 0.7 for adiabatic air –water flow.
Co = 1−2/(m +n + 2)(hw/Žh −1) (45) To approximate this profile, let us assume that the

Fig. 4. Void fraction (h) and velocity profile (V) at average void fractions of 0.2, 0.5, and 0.7 (r is radial position).

Fig. 5. Gas and liquid momentum flux parameters as a function of average void fraction.
108 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115

Fig. 6. Comparison of stability boundary CVF and calculated Fig. 7. Comparison of stability boundary CVG and calculated
Cvf. Cvg.

wall void fraction is very close to zero, then the assumed a density ratio of 1000 as a typical
momentum flux parameters are determined as number for the air –water system.
Fig. 4 shows the void fraction profile and veloc-
Cvf −1= (m+ 2)/(m + 1)[1 + I(m, Co)]/C 2o (46) ity profile. It is shown that the void fraction does
Cvg − 1=(1 −Žh)(m +2)/(m +1) not become greater than one at the center, which
is correct in the physical sense. Also the velocity
× [1 −Žh − ŽhI(m, Co)] profile becomes parabolic as the void fraction
× /(1 − ŽhCo)2 (47) increases. These trends are in good agreement
with experimental observations. As can be seen
where from Eqs. (46) and (47), the momentum flux
parameters are determined only as a function of
I(m, Co)= 2(Co −1)[1 + (m +1)(Co− 1)] void fraction and distribution parameter Co. Fig.
5 shows calculated momentum flux parameters for
× /[m(Co −1) +2]
gas and liquid as a function of average void
fraction.
van der Welle (1985) found that the exponent
Since we know the density ratio and Co, we can
of the power-law profile for liquid changes as the
calculate the stability boundary defined above
void fraction changes. He proposed that the expo-
nent of the power-law profile decreases as void which is a function of slip-ratio and modified
fraction decreases. Though his power-law profile density ratio R. In Fig. 6, we compared the calcu-
is slightly different from ours, we can make an lated momentum flux parameters with the stabil-
approximation as below. ity criteria defined by Eq. (28). Surprisingly, it is
shown that the proposed power law profile is
m =10(1− Žh) (48) stable and located well above the stability
As the void fraction increases, it approaches a boundary. As the slip ratio is greater than 1,
parabolic profile. By employing this profile, we another stability boundary defined by Eq. (25) is
can determine the void fraction profile by equat- also shown in Fig. 7. It is shown that as the void
ing Eqs. (43) and (45). We were able to obtain a fraction becomes greater than 0.45, the calculated
reasonable void fraction profile when the average liquid momentum flux parameter is greater than
void fraction is between 0.2 and 0.7. Here, we the stability boundary.
 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115 109

Fig. 8. Void fraction profile at average void fractions of 0.01,

Fig. 9. Slip ratio as a function of average void fraction.
0.1, 0.2.
profiles as average void fraction changes. It is
4.1.2. Wall-peaked 6oid fraction profiles
shown that at low void fraction, the void fraction
The wall-peaked void fraction profile is ob-
profile becomes very peaky near the wall. This is
served experimentally at low void fraction. Ser-
in good agreement with experimental observa-
izawa et al. (1975) measured the water velocity,
tions. In Fig. 9, the slip ratio is shown as a
air velocity, and void fraction profile for air –wa-
function of average void fraction. As the void
ter flow in a vertical pipe. He indicated that for
fraction decreases, the slip ratio becomes less than
bubbly flow, the bubbles are packed near the wall
1.
region. As the flow changes to slug flow the
The stability of this profile can be examined by
bubbles are concentrated near the center. Liu and
calculating the momentum flux parameters. The
Bankoff (1993a,b) did similar measurements. He
momentum flux parameters are determined only
found that void fraction profiles show a distinct
as a function of average void fraction, since Eqs.
peak near the wall.
(37), (38), (40) and (41) are calculated as
If we follow the same procedure as in Section
4.1.1, we are not able to obtain the appropriate Žh6g= ŽhŽ6g+ DŽ6gn/(n +2)2/(n + m+ 2)
power-law void fraction profile with wall void (50)
fraction equal to zero, when the void fraction is
Žh6 2g= ŽhŽ6g2 + Ž6g2/(m + 1)Žh
less than 0.2. Therefore, it was necessary to
change the assumption that wall void fraction is +DŽ6g2n/(n + 2)(m +2)/(m + 1)
zero. Since the profile should be continuous at a
× 2/(2m + n+2)
void fraction of 0.2, we assumed that the velocity
profile is the same as that at a void fraction of 0.2. × [1+ (2m + 2)/(m +n+ 2)] (51)
As we relaxed the condition that the void fraction
Žhf6f= (1− Žh)Ž6f
near the wall is zero, we need another condition
to determine the wall void fraction. We assumed − DŽ6fn/(n +2)2/(n + m+2) (52)
the power of the void fraction profile as below
Žhf6 2f = Ž1 −hŽ6f2 + Ž6f2/(m + 1)Ž1 − h
n=0.4836/Žh (49)
− DŽ6f2n/(n + 2)(m + 2)/(m + 1)
Then Eqs. (43) and (45) determine the wall void
× 2/(2m + n+2)
fraction as a function of average void fraction and
density ratio. Fig. 8 shows the gas void fraction ×[1+ (2m + 2)/(m +n+ 2)] (53)
110 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115

Fig. 10. Calculated gas and liquid momentum flux parameters as a function of average void fraction.

Fig. 11. Comparison of stability boundary CVG and calculated Cvg and comparison of stability boundary CVF and Cvf.

Fig. 10 shows the calculated momentum flux 4.2. Annular flow

parameters for gas and liquid.
In Fig. 11, we compared the calculated momen-
tum flux parameters with the stability criteria For horizontal stratified flow, Banerjee and
defined above. It is shown that the proposed Chan (1980) showed that the difference of bulk
power-law profile is stable and located well above pressure and interfacial pressure stabilized the
the stability criteria in terms of gas momentum one-dimensional two-fluid model. However, his
flux parameter. When the void fraction is between argument is not applicable for vertical annular
0.07 and 0.2, the liquid momentum flux parameter flow, as the direction of gravity is in parallel with
also satisfies stability criteria. the gas –liquid interface. Here we considered the
 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115 111

Fig. 12. Variation of the gas void fraction in the gas core and liquid film depending on the velocity ratio s at the void fractions of
0.7 and 0.9.

Fig. 13. Variation of the liquid momentum flux parameter and gas liquid momentum flux parameter depending on the velocity ratio
s at void fractions of 0.7 and 0.9.

vertical annular flow regime to look at the stabil- h= hch+ hd(1−h) (55)
ity of the two-fluid model.
Co = [shc + (1− hc)]/[sh +(1− h)] (56)
The annular flow consists of two regions. One
is the gas core and the other is the film region where s is the ratio of velocity in the gas core and
near the wall. The core region may have liquid velocity in the liquid film, s=6c/6d. From Eqs.
droplets and the film region may have bubbles. (55) and (56), the void fractions of gas core and
For annular flow, Ishii (1977) suggested a correla- liquid film are determined as functions of average
tion for the distribution parameter as below void fraction, s and Co. By use of Eqs. (55) and
(56) and the definition of slip ratio S in Eq. (44),
Co = 1+(1−h)/(h + 4 (zg/zf)) (54)
it can be shown that s should be larger than the
We construct a very simplified flow structure. slip ratio S to obtain a gas core void fraction less
The velocity in the core and in the film is assumed than unity. If we use the correlation in Eq. (54),
to be constant at 6c and 6d, respectively and we the slip ratio S is calculated as a nearly constant
also assume that local slip is negligible. The void value of 8.9 within the void fraction of 0.7 and
fraction at the core and film is also assumed to be 1.0. Therefore, s should be greater than 8.9 when
constant at hc and hd, respectively. Then the fol- the void fraction is greater than 0.7. The momen-
lowing relations are determined tum flux parameters are determined as
112 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115

Fig. 14. Variation of s1 depending on average void fraction. Fig. 15. Comparison of stability boundary CVF and Cvf (a,
s2 =s1 + 0.0001).

Cvg = [s 2hc + (1 − hc)]/[shc +(1 −hc)]2 (57) function of void fraction and s as shown in Eq.
(58), we can determine s1 that satisfies inequality
Cvf =[s hd + (1 − hd)]/[shd +(1 −hd)]
2 2
(58)
in Eq. (25). Fig. 14 shows the velocity ratio s that
The variation of gas void fraction of gas core satisfies inequality in Eq. (25). It can be seen that
and liquid film is shown in Fig. 12. As the velocity the value is close to 11.0 and insensitive to the
ratio decreases, the void fraction of the core in- change in void fraction. In Fig. 15 it is shown that
creases. The variation of momentum flux parame- if s is greater than s1, the flow is stable. There-
ters Cvg and Cvf are shown in Fig. 13. It is shown fore, if the velocity ratio of annular flow is greater
that Cvg and Cvf approach 1.0 as the velocity ratio than 11.0 the flow is stable by satisfying the
becomes close to the slip ratio S. inequality in Eq. (25).
When the slip ratio S is greater than one, the In the range of s between 8.9 and 11.0, the flow
flow is stable if one of the inequalities of Eqs. can be stable, if either inequality Eq. (28) or Eq.
(25), (28) and (29) are met. Let us look at the (29) is met. From Eqs. (57) and (58) we can
inequality Eq. (25)). As Cvf is expressed as a calculate momentum flux parameters depending

Fig. 16. Comparison of calculated gas momentum flux parameters Cvg and CVG on the stability boundary depending on the liquid
momentum flux parameter.
 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115 113

on the average void fraction. This suggests that way that the two-fluid model remains in hyper-
the conventional two-fluid model equations with bolic type and stable.
momentum flux parameters equal to unity is a Much of the computer code for nuclear reac-
good approximation to an accurate two-fluid tor safety, typically represented by RELAP5/
model, if the velocity and void fraction profiles MOD3 (Ransom et al., 1995) is using
do not deviate significantly from those velocity one-dimensional two-fluid models without mo-
profiles discussed above with s = 8.9. mentum flux parameters. This kind of one-di-
In Fig. 16, the curves of stability defined by mensional two-fluid model is intrinsically
Eq. (28) and Eqs. (57) and (58) are compared unstable as indicated by Gidaspow (1974). The
for the case of s equals 8.91 and 8.94, respec- present analysis suggests the use of proper mo-
tively at void fractions between 0.7 and 0.99. It mentum flux parameters, which would make the
is shown that the stability of the flow is sensi- one-dimensional two-fluid model always numeri-
tive to the velocity ratio s. cally stable. Though it has some restrictions and
In summary, our analysis on annular flow in- we do not have enough experimental data to
dicates that this can be stable when the velocity determine the momentum flux parameters, the
ratio is greater than 8.9 by satisfying either Eq. momentum flux parameters determined in this
(25) or Eq. (28). paper can be used for pilot computer code. We
claim that the use of momentum flux parameters
is promising, since they reflect the flow structure
5. Physical significance of the momentum flux and help to stabilize the governing differential
parameters equations.

As the momentum flux parameters contain in-

formation on the two-phase structure such as 6. Conclusions
variation of liquid and gas velocities and void
fraction over a cross-section, the stability criteria By employing the momentum flux parameters
for the two-fluid model in terms of momentum for gas and liquid phases, the present paper
flux parameters restrict the allowable form of demonstrated that one-dimensional two-fluid
the two-phase flow structure. Simplified two- model can be stable with certain restrictions on
phase flow configurations including bubbly flow, momentum flux parameters for gas and liquid.
slug flow, and annular flow, are constructed in A simple two-phase flow configuration including
Section 4 to identify the relation between stabil- bubbly, slug, and annular flow, is constructed
ity criteria and flow structure. By using existing using correlation for distribution parameter Co
correlation on volumetric distribution parameter and appropriate velocity profiles. The flow
Co and an appropriate velocity profile, it was configuration is found to be stable, when com-
possible to determine the void fraction profile pared with stability criteria described by mo-
and momentum flux parameters for the sim- mentum flux parameters. These analyses results
plified flow. Sur-prisingly, the calculated mo- suggest that the one-dimensional two-fluid
mentum flux parameters are found to be in the model should be used with momentum flux
stable region. Therefore, it can be suggested that parameters.
the two-phase flow structure changes its shape
in such a way that the governing differential
equations remain hyperbolic. In other words, Acknowledgements
the driving forces, such as wall shear, interfacial
shear, gravity, turbulence, transient forces, This research is supported by the Korean
change the two-phase flow structure in such a Ministry of Science and Technology.
114 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115

Appendix A. Nomenclature ~k laminar shear stress of the kth phase

~ tk turbulent shear stress of the kth
Bd volume of bubbles phase
CD drag coefficient ~kw wall shear stress
CM virtual mass coefficient Žh average void fraction
Cvk momentum flux distribution parame- Žv average velocity
ter of the kth phase
Db bubble diameter
Dg/Dx material derivative with respect to
References
gas velocity
Dd/Dt material derivative with respect to Banerjee, S., Chan, A.M.C., 1980. Separated flow models —
dispersed phase velocity I. Analysis of the averaged and local instantaneous for-
FD drag force mulations. Int. J. Multiphase Flow 6, 1 – 24.
FV virtual mass force Gidaspow, D., 1974. Modeling of Two Phase Flow. Round
g gravity vector Table Discussion (RT-1-2), Proceedings of the 5th Inter-
Mik generalized drag force vector for kth national Heat Transfer Conference, VII, 163.
Herringe, R.A., Davis, M.R., 1976. Structural development
phase
of gas– liquid mixture flows. J. Fluid Mech. 73 (1), 97–
pk pressure of the kth phase 123.
pki interfacial pressure of the kth phase Ishii, M., 1975. Thermo-fluid Dynamic Theory of Two-phase
rb bubble radius Flow. Eyrolles, Paris.
r non-dimensional radial position Ishii, M., 1977. One dimensional drift-flux model and consti-
t time tutive equations for relative motion between phases in
uk velocity of the kth phase various two-phase flow regimes, ANL-77-47.
Ishii, M., Mishima, K., 1984. Two-fluid model and hydro-
6c velocity of the core region of annular
dynamic constitutive relations, Nucl. Eng. Des., 107–
flow 126.
6d velocity of liquid film of annular Jones, A.V., Prosperetti, A., 1985. On the stability of first-
flow order differential models for two-phase flow prediction.
6k velocity vector of the kth phase Int. J. Multiphase Flow 11 (2), 133– 148.
6ki interfacial velocity of the kth phase Lahey, R.T., Cheng, L.Y., Drew, D.A., Flaherty, J.E., 1980.
6wg, 6wf gas or liquid velocity very near wall The effect of virtual mass on the numerical stability of
accelerating two-phase flow. Int. J. Multiphase Flow 6,
6og, 6of maximum gas, or liquid relative ve-
281– 294.
locity at center Liu, T.J., Bankoff, S.G., 1993a. Structure of air– water bub-
z axial direction along pipe bly flow in a vertical pipe — I. Liquid mean velocity
Yk mass generation of the kth phase and turbulence measurements. Int. J. Heat Mass Transf.
hk void fraction of the kth phase 36 (4), 1049– 1060.
h void fraction of gas Liu, T.J., Bankoff, S.G., 1993b. Structure of air– water bub-
ho void fraction at center bly flow in a vertical pipe — II. Void fraction, bubble
velocity and bubble size distribution. Int. J. Heat Mass
hw void fraction at the wall
Transf. 36 (4), 1061– 1072.
hc void fraction in the core of annular Micaelli, J.C., 1988. Document De Synthese CATHARE,
flow SETh/LEML-EM/88-129, Grenoble Cedex.
hd void fraction in the liquid film of an- Ransom et al., 1995, RELAP5/MOD3 Code manual,
nular flow NUREG/CR-5535, Idaho National Engineering Labora-
vm mixture viscosity tory, ID, June 1995.
q inclination angle of pipe from the Serizawa, I., Kataoka, Michiyoshi I., 1975. Turbulence
structure of air– water bubbly flow — II. Local proper-
vertical direction
ties, Int. J. Multiphase Flow, 2, 235– 246.
zc density of continuous phase Stuhmiller, J.H., 1977. The influence of interfacial pressure
zk density of the kth phase forces on the character of two-phase flow model equa-
~I interfacial shear stress . tions. Int. J. Multiphase Flow 3, 551– 560.
 J.H. Song, M. Ishii / Nuclear Engineering and Design 204 (2001) 101–115 115

Sung, K., Chun, M.H., 1996. Onset of slugging criterion based van der Welle, R., 1985. Void fraction, bubble velocity and
on singular points and stability analyses of transient one bubble size in two-phase flow. Int. J. Multiphase Flow 11
dimensional two-phase flow equations of two-fluid model. (3), 317– 345.
J. Korean Nucl. Soc. 28 (3), 299–310.