ht. J. Heat Mass Transfer. Vol. 41, Nos 8-9, pp.

1103-I 112, 1998

0 1997 Elsevier Science Ltd. All rights reserved
Pergamon Printed 1x1Great Britm
0017-9310198 $19.00+0.00

PII : SOO17-9310(97)00167-l

One-group interfacial area transport in vertical

bubbly flow
Q. WU, S. KIM and M. ISHII
Thermal Hydraulics and Reactor Safety Laboratory, Purdue University, West Lafayette, IN 47907,
U.S.A.

and

S. G. BEUS
Bettis Atomic Power Laboratory, Westinghouse Electric Corporation, Box 79, West MiWin,
PA 15122, U.S.A.

(Received 20 February 1997 and in final form 28 May 1997)

Abstract-In the two-fluid model, interfacial concentration is one of the important parameters. The
objective oF this study is to develop an interfacial area equation with the source and sink terms being
properly mf>deled. For bubble coalescence, the random collisions between bubbles due to turbulence, and
the wake entrainment process due to the relative motions of the bubbles, were included. For bubble
breakup, the impact of turbulent eddies is considered. Compared with measured axial distributions of the
interfacial area concentration under various flow conditions, the adjustable parameters in the source/sink
terms were obtained for the simplified one-dimensional transport equation. 0 1997 Elsevier Science Ltd.

11.INTRODUCTION mustafaogullari and Ishii [8] generalized Reyes’

model, leading to the following equation :
In the analysis of two-phase flow, the formulation
using a two-fluid model is considered as the most
accurate model. With proper averaging in this model
[l-4], the two phases are considered separately in
terms of two sets of conservation equations that gov-
ern the balance of mass, momentum and energy in where f(~, Y, t) is the particle number density dis-
each phase. However, the averaged macroscopic fields tribution function, which specifies the probable num-
of the two phases are not independent of each other, ber of fluid particles at a given time t, in the spatial
and there are certain phase interaction terms in the range of d.%about a position R, with particle volumes
field equations to characterize the interfacial transfer between Y and Y +dY. Moreover, V&Y, Y, t)
of mass, momentum and energy. These terms all con- denotes the local time-average velocity of the particles,
tain a parameter tlhat specifies the geometric capability and s,,,,(R, “Y, t) refers to the fluid particle sink or
of the interfacial transfer, i.e. the interfacial area con- source rate due to phase change. If the phase change
centration, defined as the total surface area of the only causes fluid particle shrinkage or expansion, Q,
dispersed fluid particles per unit mixture volume [5]. can be expressed in the form of - a/&(f dY/dt), a
Therefore, a closure relation for the interfacial area term Shraiber [9] thought should be presented on the
concentration is indispensable in the two-fluid model left-hand side of equation (1). However, for the case
concerning the detailed treatment of the phase inter- of homogeneous nucleate boiling or condensation in
actions. a subcooled boiling flow, sph should also include the
Since the interfacial area concentration changes rate of change of the fluid particle population with
with the variation of the particle number density due specific volumes. Detailed studies on the phase change
to coalescence and breakage, analogous to Boltz- term follow the approach of Kocamustafaogullari and
mann’s transport equation, a population balance Ishii [lo].
approach (PBA) was proposed by Reyes [6] to The interaction term, s,(~, Y, t), represents the net
develop a particle number density transport equation rate of change of the particle number density dis-
for chemically non-reacting, dispersed, spherical fluid tribution function caused by particle breakup or
particles. A simil.ar method was employed in com- coalescence process. Some phenomenological models
bustion theory, known as the spray-equation [7]. For for these terms were summarized by Prince and Blanch
the purpose of interfacial area transport, Koca- [ 111, and Lafi and Reyes [ 121. These models presented
1103
1104 Q. WU et al.

NOMENCLATURE

a area concentration c1 dynamic viscosity

C constant, drag coefficient P density
bubble Sauter mean diameter CJ surface tension.
r” two-phase friction factor
j superficial velocity Subscripts
L length interfacial area concentration
n bubble number density : bubble
collision frequency for two bubbles cr critical
x collision rate effective
Re Reynolds number B liquid
s volume dependent source/sink terms gas
s source/sink terms : hydraulic
t time i interfacial
T time m mixture
u velocity max maximum
velocity n bubble number density
> bubble volume P particle/bubble
We Weber number ph phase change
x space coordinates pm bubble mean value
Z ilow direction coordinate. RC random collision
t turbulent
Greek symbols TI turbulent impact
a void fraction Y- bubble volume
A difference W bubble wake
& averaging factor WE wake entrainment.

the detailed insight of the mechanisms for coalescence equation (2) caused by particle coalescence and break-
and breakage phenomena. However, due to the depen- age, a general approach treats the bubbles in two
dence on the fluid particle volume, many adjustable groups : the spherical/distorted bubble group and the
parameters and assumptions were imposed that may cap/slug bubble group [13], resulting in two bubble
be beyond justification with the existing experimental number density transport equations that involve the
data. For most two-phase flow studies, where the pri- inner and inter group interactions, as shown in Fig. 1.
mary focus is on the average fluid particle behavior, The mechanisms of these interactions can be sum-
the detailed volume dependent particle number den- marized in five categories : the coalescence due to ran-
sity transport equation would be too tedious and com- dom collisions driven by turbulence, the coalescence
plicated for use in the field equations. Hence, the pre- due to wake entrainment, the breakage due to the
sent study starts from the following integral form of impact of turbulent eddies, the shearing-off of small
the particle number density transport equation : bubbles from cap bubbles, and the breakage of large
cap bubbles due to flow instability on the bubble
an(2,
t) surface. Some other mechanisms, such as laminar
at + v &n(% t)@, 0)
shearing induced coalescence [14] and the breakage
due to velocity gradient [15], are excluded because
= c &,(% 0 + S”,Ph@,4
I
(2) they are indirectly caused by the distributions of the
flow parameters and void fraction [ 161, and the direct
where n@, t) is the number density of particles of all mechanisms still follow the above five categories.
sizes, and v~,,,(R,t) is the average local particle velocity In practice, when the void fraction of a two-phase
weighted by the particle number, which is identical to bubbly flow is small, no cap or slug bubbles exist. The
the time-averaged bubble velocity weighted by the gas two-group transport equations are then reduced to
void fraction, Tg,,in the time-averaged two-fluid model one group without the involvement of the interactions
[3], if the statistical sample size is sufficiently large. between the two groups. As the first step of the general
On the right-hand side of equation (2), S,, and &,, approach, the focus of this study is on the first group
represent the total bubble number source or sink rate transport equation for bubbly flow without the occur-
per unit mixture volume. rence of cap or slug bubbles. In Section two, three
To model the integral source and sink terms in models are developed for binary bubble coalescence
 One-group interfacial area transport in vertical bubbly flow

random collkion random colIision wake-M_ xlt

wake-entraiumen~ wake-cntmimnult

turbulenceimp& surface instability

Fig. 1. Mechanisms of bubble coalescence and break-up in two-group model.

and breakage, including the bubble coalescence due 2.1. Random collision induced bubble coalescence
to random collisions driven by turbulence, bubble To model the bubble coalescence rate induced by
coalescence due to wake entrainment, and bubble turbulence in the continuous medium, the bubble ran-
breakage due to the impact of turbulent eddies. With dom collision rate is of primary importance. These
these models, the one-group interfacial area transport collisions are postulated to occur only between the
equation is obtained in Section 3, whereas the adjust- neighbouring bubbles, because long range inter-
able parameters are evaluated in section four with the actions are driven by large eddies that transport
existing experimental data obtained from vertical two- groups of bubbles without leading to significant rela-
phase bubbly flow [17]. The approach in this study tive motion [ 11, 181. Between two neighboring spheri-
provides a preliminary foundation for the latter phase cal bubbles of the same size as shown in Fig. 2, the
investigation of the two-group transport equations, time interval for one collision, At, is defined as
which serve as the general closure relation for inter-
At = E/u,. (4)
facial area concentration for the two-fluid model.
Here, u, is the root-mean-square approaching velocity
2. MODELING OF BUBBLE COALESCENCE AND
of the two bubbles, and t represents the mean trav-
BREAKAGE
elling distance between the two bubbles for one colli-
sion, which is approximated by :
For dispersed bubbly flow without phase change,
only three mechanisms for bubble coalescence and L-De-dDa(~-6,D)=~(l--b,.l”) (5)
breakage are considered in the following one-group
bubble number density transport equation :
where De denotes the effective diameter of the mixture
a@,t) volume that contains one bubble. Since the bubble
at + v $g.a, t)n(zi, t))
travelling length for one collision varies from De to

= &,,c @, 0 + S”,W&, 0 + S&T,(% 0 (3)

where the subscript ‘n’ stands for net bubble number
density change, ‘RC’ for random collision due to tur-
bulence in the continuous medium, ‘WE’ for wake-
entrainment, and ‘TI’ for turbulent eddies that impact
on bubbles resulting in bubble breakage. In this
section, these source and sink terms are modeled indi-
vidually by assuming binary spherical bubble inter-
actions. Throughout the study, the average bubble size
is characterized by the bubble Sauter mean diameter,
D E 6a/u,, where a and ai represent the local time-
averaged void fraction and interfacial area concen-
tration, respectively. Fig. 2. Geometric definitions of two approaching bubbles.
1106 Q. WU et al.

(De-D), a factor 6 is introduced in equation (5) to mean distance is very large, no collision should be
feature the average effect, whereas 6, is a collective counted because the range of the relative motion for
parameter in considering the sign of proportionality collisions between the neighboring bubbles is limited
between De and D/c?. For small void fraction, 6, by the eddy size comparable to the bubble size. To
plays a minor role due to the fact that De is much consider this effect, the following modification factor
larger than D. However, it is important if the travelling is suggested for equation (10) :
length is comparable to the mean bubble size. When
void fraction approaches the dense packing limit [l-exp(-C$)j (11)
(c( E c(,,,), the mean travelling length should be zero,
which leads to 6, equal to &y’. Using this asymptotic
where L, is the average size of the eddies that drive
value as the approximation of 6,) the mean travelling
length is reduced to : the neighboring bubbles together, which is assumed
to be on the same order of the mean bubble size,
because smaller eddies do not provide considerable
bulk motion to a bubble, while larger eddies transport
groups of bubbles without significant relative motion.
Accordingly, the collision frequency for two bubbles Thereafter, the final form of the bubble collision rate
moving toward each other is given by : is given by :

1
r R RC-(&D*)
113
%,x (c&fx - cc”3) I

-lx)I (12)
( 4n,x
113 II3
Since the bubbles do not always move toward each %ax~
x 1 -exp -C ,,3
other, however, a probability, PC,for a bubble to move [
l/3 .

toward a neighboring bubble is considered here to

modify the collision rate. By assuming a hexagonal For each collision, coalescence may not occur and
close-packed structure, this probability is given by : thus a collision efficiency was suggested by many
investigators [19,20]. The most popular model for the
and PC = 1, c( > c(, (8) collision efficiency is the film thinning model [20]. In
this model when the bubbles approach faster, they
tend to bounce back without coalescence due to the
where tl, is the critical void fraction when the center
limitation of the film drainage rate governed by the
bubble cannot pass through the free space among
surface tension. Mathematically, the coalescence rate
the neighboring bubbles. In reality, the neighboring
decreases exponentially with respect to the turbulent
bubbles are in constant motion, and the critical void
fluctuating velocity, which is much stronger than the
fraction can be very close to the dense packing limit.
linear dependence of the collision rate, resulting in an
This leads to :
overall decreasing trend of the coalescence rate as the
turbulent fluctuation increases. This caused serious
trouble when the model was applied to experimental
data following the procedure specified in Section 4.
Subsequently, the collision rate for a mixture with Tremendous discrepancies were obtained at different
bubble number density, n, is given by : liquid flow conditions. Hence, a constant coalescence
efficiency is employed in the present model to depict
the randomness of the coalescence phenomenon after
each collision [16], and the bubble coalescence rate
due to random collisions is given by :
1
RRC= G,(w’D’)
(agx-a”‘) 1
The functional dependence of the above collision rate

>I
agrees with that of Coulaloglou and Tavlarides [18]
proposed in 1976 for a liquid-liquid droplet flow (13)
system, analogous to the particle collision model in
an ideal gas. The difference is that the present model where CRCand C are adjustable parameters, depend-
contains an extra term in the bracket, which covers ing on the properties of the fluid. Nevertheless, the
the situation when the mean free path of a bubble is constant coalescence efficiency is only an approxi-
comparable to the mean bubble size. Nevertheless, the mation, and further efforts are needed to model the
model in the present form is still incomplete, since no efficiency mechanistically. The remaining unknowns
matter how far away the neighboring bubble is are the maximum void fraction and the mean bubble
located, the collision would occur as long as there is fluctuating velocity. By definition, @maxis the dense
a finite approaching velocity. In actuality, when the packing limit of void fraction when the coalescence
 One-group interfacial area transport in vertical bubbly flow 1107

rate approaches infinity. A rational choice of a,,,,, Here, u,(D) is the terminal velocity of a bubble of
should be approximately 0.8 at the transition point diameter D relative to the liquid motion. By inte-
from slug to annular flow [ 16, 211. The mean bubble grating equation (17) over the effective wake length,
fluctuating velocrty, ut, in equation (13) is pro- the average relative velocity in the wake region is given
portional to the root-mean-square liquid fluctuating by:
velocity difference between two points of length scale
D, and is given by E”~D”~ [ 131. Here, E is the energy
dissipation rate per unit mass of the continuous
medium. In a complete two-fluid model, E comes from
its own constitutive relation such as the two-phase K- = u,(D)F . (18)
E model [22, 231. For one-dimensional analysis,
however, this term can be approximated by a simple
algebraic equation, as suggested in Section 4. The exact form of F(D/L,) is not important since
the effective bubble wake region may not be fully
established. According to Tsuchiya [29], the wake
2.2. Wake-entraiirment induced bubble coalescence
length is roughly 5-7 times the bubble diameter in an
When bubbles enter the wake region of a leading
air-water system, and thus D/Lw as well as F(D/L,)
bubble, they will accelerate and may collide with the
are treated as constants depending on the fluid proper-
preceding one [24-261. For a spherical air bubble with
ties. As long as their values obtained from exper-
attached wake region in the liquid medium, the effec-
tive wake volume, V,, in which the following bubbles imental data fall into the range of D/Lw = 5-7, the
mechanism should be acceptable. Substituting equa-
may collide with the leading one, is defined as the
tion (18) into equation (15) yields the following simple
projected bubble area multiplied by the effective
expression of the bubble collision rate per unit mixture
length, L,. The number of bubbles inside the effective
volume is then given by volume due to the wake-entrainment mechanism :

R WE= CWED2u,(D)n2, with C,, = :F (19)

IV,,,= VW, z ;nD’(L,-D/2)n. (14)

Assuming that the average time interval for a bubble where C,, is an adjustable constant mainly deter-
in the wake region to catch up with the preceding mined by the ratio of the effective wake length to the
bubble is AT, the collision rate per unit mixture vol- bubble size and the coalescence efficiency. A proper
ume should satisfy : choice for CwE should yield an effective wake length
roughly between 5 and 7 from equations (19) and
R WE &!k
2 ATz
l
gnD2n2 nv
(18). The bubble terminal velocity, u,, is a function
of the bubble diameter and local time-average void
fraction. Based on the balance between the buoyancy
(15) force and drag force in a two-phase bubbly flow, Ishii
where anu is the average relative velocity between the and Chawla [30] applied a drag-similarity criterion
leading bubble and the bubble in the wake region. If with the mixture-viscosity concept to obtain the fol-
the transient for a bubble to reach its terminal velocity lowing expression for the relative velocity :
is assumed to be much shorter than the collision
“’
process, the average relative velocity, anu, can be ex- (20)
pressed in terms of the relative velocities of the con-
tinuous medium inside and outside the wake region :
c = 24U +O.lR@ and Re, ~ PfQ
D I”rW).
%

with Vr(z) as the local liquid velocity along the center (21)
line, z as the distance measured from the center of the
leading bubble, and V,,, as the ambient liquid velocity. 2.3. Bubble breakup due to turbulent impact
For spherical bubbles, since the external flow is almost For binary bubble breakup due to the impact of
indistinguishable from that around a solid sphere at turbulent eddies, the driving force comes from the
the same Reynolds number [27], the wake structure inertial force, F,, of the turbulent eddies in the con-
of the leading bubble can be analogous to that around tinuous medium, while the holding force is the surface
a solid sphere. In such a turbulent wake (Re, > 20), tension force, F,. To drive the daughter bubbles apart
which satisfies most of the practical bubbly flow with a characteristic length of D within time interval
regimes, the wake velocity along the center line At, a simple momentum balance approach gives the
roughly satisfies [28] : following relation :
213

u,(D) prD3D
V,(z)- I’, = (17) PK F,-F,,. (22)
AT2
1108 Q. WU et al.

Here, the inertia of the bubble is dominated by the

virtual mass because of the large density ratio of the (27)
liquid and gas. Rearranging equation (22) leads to the
following average bubble breakup time :
where $ is a factor depending on the shape of the
-I/Z bubbles, and ?Ir,, denotes the average bubble volume.
w:D > we
, We=--
0
cr For spherical bubbles $ equals 1/(36rc). Substituting
equation (27) into equation (2) yields :
(23)
The velocity, U,is assumed to be the root-mean-square
velocity difference between two points of length D,
which implies that only the eddies with sizes equivalent
to the bubble size can break the bubble. We,, is a
collective constant, designated as a critical Weber
number. The reported value of We,, for bubble
breakup varies in a wide range due to the resonance The second term on the right-hand side represents the
excitation of the turbulent fluctuation [31]. In an air- effects of the variation in bubble volume. If the gas
water system, Prince and Blanch [11] suggested that phase is assumed to be incompressible without phase
We,, equals 2.3. Accordingly, if the bubble number change, from the gas phase continuity equation, this
density is n, the bubble breakup rate should be : term should be zero. Subsequently, the one-group
interfacial area transport equation is reduced to :

!$+V*(@J =$ $ *(-RRC-RWE+RT,)
0
(24)
In a homogeneous turbulent flow, the probability = &,RC + &WE + &I. (29)
for a bubble to collide with an eddy that has sufficient
With the models developed in Section 2, the net
energy to break the bubble is approximately [18] :
rates of change of interfacial area concentration per
unit mixture volume are given below :

1
1
Sa,RC= - j$R"h"~) aiJx

where u$, is the critical mean-square fluctuation vel- (a$ - N”3)

ocity obtained from We,. Finally, the bubble breakup
rate per unit mixture volume is given by :
. (30)

R,, =CTiexp(-%)nz(l-%y2,
S&WE = - $ CWEuraf. (31)
We > We,,. (26)

Again, the adjustable parameters C,, and We,, should

S
be evaluated with experimental data. This expression
differs from the previous models [I 1, 121 because the
breakup rate equals zero when the Weber number is
, We > We,,. (32)
less than We,,. This unique feature permits the
decoupling of the bubble coalescence and breakup
processes. At a low liquid flow rate with small void Equations (29)-(32) constitute the one-group clos-
fraction, the turbulent fluctuation is small and thus ure relation of interfacial area concentration in two-
no breakup would be counted, which allows the fine- phase vertical dispersed bubbly flow. The variables in
tuning of the adjustable parameters in the coalescence these equations are coupled with the field equations
terms, independent of the bubble breakage. in the two-fluid model. For the information of the
local interfacial area concentration, the field equations
should be solved together with the closure relations.
3. ONE-GROUP INTERFACIAL AREA
However, the presented model is limited to the two-
CONCENTRATION TRANSPORT EQUATION
phase dispersed bubbly flow. At high void fraction
In the two-fluid model, the parameter of interest is when cap or slug bubbles appear, another transport
the interfacial area, a,, rather than the bubble number equation should be provided for the interfacial area
density. To obtain the transport equation for inter- concentration of cap or slug bubbles. The two groups
facial area concentration, equation (2) can be modi- are not independent and the inter-group transfer terms
fied with the following geometric relation : should also be modeled individually.
 One-group interfacial area transport in vertical bubbly flow 1109

4. EVALUATION OF THE ONE-GROUP MODEL

fTw =h(;r25 =$&&-J. (37)
The simplest form of the interfacial area transport
equation is the one-dimensional formulation obtained
To identify the adjustable parameters in the source
by applying crosssectional area averaging over equa-
and sink terms, experimental data of a steady air-
tion (29) :
water cocurrent up-flow in a 5.08 cm pipe [17] are
used. In these experiments, interfacial area con-
centration and void fraction were measured with a
double-sensor conductivity probe at three different
==a,,,) + (&,w,) + <&,d. (33) axial positions (L/D, = 2, 32, 62). In Table 1, seven
cases of tests are summarized for different flow con-
Due to the uniform bubble size assumption, the area- ditions. With the measured interfacial area con-
averaged bubble interface velocity weighted by inter- centration at L/D,, = 2 as the initial condition, equa-
facial area concentration can be given by : tion (35) is integrated numerically to predict the axial
distribution of interfacial area concentration, and the
adjustable parameters are determined if the pre-
dictions at the other two locations match the exper-
which is the same as the conventional area-averaged imental data. At low liquid flow rate with small void
gas velocity weighted by void fraction, if the internal fraction, bubble breakage can be neglected due to the
circulation in the bubble is neglected. The exact math- very small Weber number compared to the critical
ematical expressions for the area-averaged source and value. Therefore, the fitting involves only three con-
sink terms would involve many covariances that may stants : CRC, C and C,,. After these parameters are
further complica.te the one-dimensional problem. fixed, the transport equation is further applied to high
However, since these local terms were originally flow rate conditions solely for Cr, and We,,. The final
obtained from a finite volume element of the mixture, results of these adjustable parameters are summarized
the functional dependence of the area-averaged source in Table 1. For the wake-entrainment mechanism,
and sink terms OIL the averaged parameters should be C,, is 0.151. Assuming a flat cross-sectional wake
approximately the same if the hydraulic diameter of velocity profile, the effective wake length for bubble
the flow path is considered as the length scale of the coalescence as estimated from equations (18) and (19)
finite element. Therefore, equations (30)-(32) with the is about seven times the bubble diameter, which agrees
parameters averaged within the cross-sectional area with the observation of Tsuchiya [29]. For bubble
are still applicable for the area-averaged source and breakup, the critical Weber number is found to be 2.0
sink terms in equation (33). for the best fitting to the experimental data, which is
To utilize the experimental data at different axial slightly smaller than 2.3, a value suggested by Prince
locations under steady-state condition, the transient and Blanch [l 11.This discrepancy may be caused by
term in equation (33) has to be dropped, resulting in the one-group approach that assumes uniform bubble
size. In reality, bubble breakage exists as long as the
size of certain bubbles reaches the breakup limit in
spite of the fact that the mean size does not exceed
it. Therefore, the critical Weber number for bubble
(35)
breakage based on the average bubble size should be
Without phase change, under the assumption of smaller than that from the actual size of a breaking
incompressible flow, the void fraction should be con- bubble.
stant, independent of the axial positions. For a given At a very low liquid flow rate with small void frac-
flow condition, the only variable that should be speci- tion, as in the case 1, no bubble breakage is involved.
fied is the energy dissipation rate per unit mixture
mass, E. A sophisticated approach is to couple the
transport equation with the field equations and the
constitutive relation of e, such as k-s model [25]. How- Table 1. Test conditions [ 171 and adjustable parameters
ever, for the purpose of evaluating the model in the
0,) (m s-‘) 0,) (m s-l) (a>
one-dimensional form, a simple algebraic correlation
for E is employed in this study : Case 1 0.023 0.77 2.5%
Case 2 0.117 0.77 10.0%
Case 3 0.058 1.11 4.0%
E =hw &W’ (36) Case 4 0.117 1.11 7.0%
h Case 5 0.023 1.58 1.6%
Case 6 0.058 1.58 3.0%
where (v,) is the mean mixture velocity, Dh refers to Case 7 0.117 1.58 6.5%
the hydraulic diameter of the flow path, and fTwis
C C GE We,, CT,
the two-phase friction factor. With p,,, as the mixture 0.0&5 3 0.151 0.18 2.0
viscosity [30], fTylr
is given by :
1110 Q. WU et al.

g -5
V
=
1
3
5 -10 -totalchange
&
0 case 1
-random coBion
mm-. wake-edrainmznt
0 20 40 60 80
-15
LJDh
0 20 40 60 80 Fig. 5. Axial variation of (a,) for case 7.
L/Dh
Fig. 3. Axial variation of (at) for case 1.

the interfacial area concentration in the flow direction.

An extreme condition is for case 7, as shown in Fig.
Figure 3 illustrates such variation of the area-averaged
5, with a liquid superficial velocity of 1.58 m SK’,
interfacial area concentration in the flow direction.
where bubble breakage exists at the very beginning.
With the selected coefficients, no bubble breakage is
In this case, the coalescence rate seems balance by
present since the Weber number is smaller than the
the breakup rate along the flow path, resulting in a
critical value. Since the void fraction is small, the
relatively flat axial distribution of the interfacial area
change of interfacial area concentration caused by
concentration. Compared to the experimental data,
random collisions is only about one tenth of total
all seven cases are shown in Figs. 6(a) and (b). The
decrease. However, as the void fraction increases to
interfacial area concentrations predicted by the pro-
10% in case 2 with the same liquid superficial velocity
posed model are generally in good agreement with the
as in case I, the change of interfacial area con-
measurements at three axial positions. The maximum
centration due to random collisions increases to
relative difference is about 8% at very small void
roughly 20% of the total change, as shown in Fig. 4.
fraction with high liquid flow rate. Nevertheless, the
Moreover, because of the large coalescence rate, the
conclusion is based on the only set of published exper-
mean bubble size grows rapidly. At about L/D,, = 40,
imental data that have three axial measurements for
the bubble Weber number becomes greater than the
each flow condition. Fine-tuning of the adjustable
critical value and thus the bubble breakage term must
parameters is needed as more data become available.
be taken into account, resulting in a slower decrease of
Moreover, the constant coalescence efficiency is only
an approximation, and further efforts are needed to
model the efficiency mechanistically.
20
5. CONCLUSIONS
0 c; In this study, a one-group interfacial area transport
equation together with the modeling of the source and
sink terms for bubble breakup and coalescence was
presented. For bubble coalescence, two mechanisms
were considered to be dominant in the vertical two-
phase bubbly flow: the random collisions between
bubbles due to turbulence in the flow field and the
wake entrainment process due to the relative motion
of the bubbles in the wake region of a seeding bubble.
_ randomu+in
_.-_ wake-e- For bubble breakup, the impact of turbulent eddies
was included.
By area-averaging over the local one-group trans-
0 20 40 60 80 port equation, a one-dimensional form of the inter-
UDb facial area concentration transport equation was
Fig. 4. Axial variation of (a,) for case 2. obtained. Compared with experimental data for the
 One-group interfacial area transport in vertical bubbly flow 1111

120 REFERENCES
1. Vernier, P. and Delhaye, J. M., General two-phase flow
equations applied to the thermohydrodynamics of boil-
ing nuclear reactor. Energie Primaire, 1968,4,5-25.
2. Kocamustafaogullari, G., Therrnofluid dynamics of sep-
arated two-phase flow. Ph.D thesis, Georgia Institute of
Technology, Atlanta, GA, 1971.
3. Ishii, M., Thermo-Fluid Dynamic Theory of Two-Phase
60 Flow. Collection de la Direction des Etudes et Reserches
x d’Electricitt de France, Eyrolles, Paris, 22, 1975, p.45.
9 4. Boure, J. A., Mathematical modeling of two-phase flows.
40 Proceedings of CSNI Specialist Meeting, ed. S. Banerjee
-- and K. R. Weaver, Vol. 1. A.E.C.L., Toronto, Canada,
1978, p. 85.
5. Ishii, M. and Mishima, K., Study of two-fluid model and
interfacial area. Technical report. ANL-80-111, Argonne
National Laboratory, Chicago, IL, 1980.
6. Reyes, J. N., Statistically derived conservation equations
0 20 40 60 80 for fluid particle flows. Proceedings of ANS Winter Meet-
L/Db ing, San Francisco, CA, 1989, p. 12.
90 7. Williams, F. A., Combustion Theory, Addison-Wesley,
+ ease 1 Reading, MA, 1965.
80 8. Kocamustafaogullari, G. and Ishii, M., Foundation of
.case3
the interfacial area transport equation and its closure
70 *case4 relation. International Journal of Heat and Mass Trans-
60 fer, 1995,38,48 1.
9. Shraiber, A. A., Comments on ‘foundation of the inter-
2
2 50 facial area transport equation and its closure relations’.
International Journal of Heat and Mass Transfer, 1996,
;i:40 39, 1117.
d 10. Kocamustafaogullari, G. and Ishii, M., Interfacial area
30 and nucleation site density in boiling systems. Inter-
national Journal of Heat and Mass Transfer, 1983, 26,
20 1377.
11. Prince, M. J. and Blanch, H. W., Bubble coalescence
10
and break-up in air-sparged bubble columns. AIChE
0 Journal, 1990, 36, 1485.
12. Lafi, A. Y. and Reyes, Jr, J. N., Phenomenological mod-
0 20 40 60 80 els for fluid particle coalescence and breakage. Technical
mb report, OSU-NE-9120, Department of Nuclear Engin-
Fig. 6. (a) Axial variation of (a,) for case 2, 5, 6 and 7 ; (b) eering, Oregon State University, Corvallis, Oregon,
axial variation of (aJ for case 1, 3 and 4. 1991.
13. Ishii, M. and Kojasoy, G., Interfacial area transport
equation and preliminary considerations for closure
relations. Technical report, PU-NE-93/6, Nuclear
Engineering Department, Purdue University, West
Lafayette, IN, 1993.
axial distribution of the interfacial area concentration 14. Friedlander, S. K., Smoke, Dust and Haze. Wiley, New
York, 1977.
under various flow conditions, the adjustable par-
15. Taylor, G. I., The formation of emulsion in definable
ameters in the model were obtained. The results indi- field of flow. Proceedings of Royal Society, (London)
cate that the proposed models for bubble breakup Series, 1934, A146, 501.
and coalescence are appropriate. The ranges of the 16. Wu, Q., Kim, S., Ishii, M. and Beus, S. G., One-group
interfacial area concentration transport in vertical air-
adjustable parameters agree with the physical obser-
water bubbly flow. International Heat Transfer Con-
vations. However, the comparison was based on the ference, HTC-Vol. 10, Baltimore, Maryland, 1997, p. 67.
only set of published experimental data that have three 17. Kashyap, A., Ishii, M. and Revankar, S. T., An exper-
axial measurements for each flow condition. Fine- imental and numerical analysis of structural devel-
tuning of these adjustable parameters is needed as opment of two-phase flow in a pipe. Technical report,
PU-NE-94/2, Nuclear Engineering Department, Purdue
more data becomes available. When applied to three- University, West Lafayette, IN, 1994.
dimensional cases, the adjustable parameters for the 18. Coulaloglu, C. A. and Tavlarrides, L. L., Drop size
detailed localized transport equation have to be veri- distributions and coalescence frequencies of liquid-
fied through the coupling with the two-fluid model. liquid dispersion in flow vessels. AIChE Journal, 1976,
22,289.
Moreover, the constant coalescence efficiency in this
19. Oolman, T. and Blanch, H. W., Bubble coalescence in
study is only an approximation, and further efforts stagnant liquids. Chemical Engineering Communication,
are needed to model the efficiency mechanistically. 1986,43,237.
20. Kirkpatrick, R. D. and Lockett, M. J., The influence
of approach velocity on bubble coalescence. Chemical
Acknowledgements--This study was supported by West- Engineering Science, 1974, 29,2363.
inghouse Bettis Attomic Power Laboratory. Valuable com- 21. Wallis, G. B., One-dimensional Two-phase Flow.
ments of MS T. Wilmarth de Leonardi are appreciated. McGraw-Hill, New York, 1969.
1112 Q. WU et al.

22. Serizawa, A. and Kataoka, I., Turbulence suppression 27. Clift, R., Grace, J. R. and Weber, M. E., Bubbles,
in bubbly two-phase flow. Journalof Nuclear Engineering Drops and Particles. Academic Press, New York,
and Design, 1991, 122, 1. 1978.
23. Lopez de Bertodano, M., Lahey, Jr, R. T. and Jones, 0. 28. White, F. M., Viscous Fluid Flow, 2nd edn. McGraw-
C., Development of a k-e model for bubbly two-phase Hill, New York, 1991, p. 481.
flow. Journal of Fluids Engineering, 1994, 116, 128. 29. Tsuchiya, K., Miyahara, T. and Fan, L. S., Visualization
24. Otake, T., Tone, S., Nakao, K. and Mitsuhashi, Y., of bubble-wake interactions for a stream of bubble in a
Coalescence and breakup of bubbles in liquids. Chemical two-dimensional liquid solid fluidized bed. International
Engineering Science, 1977, 32, 377. Journal of Multiphase Flow, 1989, 15, 35.
25. Bilicki, Z. and Kestin, J., Transition criteria for two- 30. Ishii, M. and Chawla, T. C., Local drag laws in dispersed
phase flow patterns in vertical upward flow. International two-phase flow. Technical report, ANL-79-105,
Journal of Multiphase Flow, 1987, 13(3), 283. Argonne National Laboratory, Chicago, 1979.
26. Stewart, C. W., Bubble interaction in low-viscosity 31. Sevik, M. and Park, S. H., The splitting of drops and
liquid. International Journal of Multiphase Flow, 1995, bubbles by turbulent fluid flow. Journal of Fluids Engin-
21, 1037. eering, 1973,95, 53.