Chemical Engineering Science 56 (2001) 5945–5953

www.elsevier.com/locate/ces

A continuum model for gas–liquid %ow in packed towers

Shijie Liu ∗
Department of Chemical and Materials Engineering, University of Alberta, Edmonton AB, Canada T6G 2G6

Abstract
Gas–liquid %ow in packed towers is commonly encountered in the chemical and processing industry. A continuum model is
developed based on the volume-and-time averaging of multiphase %ows in isotropic rigid porous media=packed columns. Closures
are presented for the evaluations of the extra surface=intrinsic phase integral terms. Both inertia and inter-phase interactions are
retained in the volume averaged (Navier–Stokes) equations. These governing equations are solved for fully-developed axi-symmetric
single and gas–liquid two phase %ows in highly porous packed towers. It is found that the dispersion term is present in the
continuity equation as well as the momentum equations. Numerical simulations with the models show that the volume-and-time
averaged equations can predict the velocity, phase hold-up and pressure drop quite well for up to the loading point for gas–liquid
counter-current %ows. ? 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Packed tower; Multiphase %ow; Volume-and-time averaging; Counter-current %ow; Velocity distribution; Phase hold-up

1. Introduction For a laminar saturated single %uid %ow in a straight

There are three gas–liquid %ow arrangements useful for
separation and reaction engineering applications. There
is only one possible way for countercurrent %ow arrange-
ment with the heavy %uid (liquid) %owing downward and
the light %uid (gas) %owing upward. This countercur-
rent %ow arrangement is commonly used for distillation,
cooling tower, absorption and stripping operations (e.g.,
Kister, 1992; Billet, 1995). There are two possible ar-
rangements for cocurrent %ows: either upward or down-
ward %ow. While concurrent %ows can be applied for
mass and=or heat transfer operations, they are generally
used as multiphase reactors (e.g., Shah, 1979; Herskowitz
& Smith, 1983). In particular, the downward %ow ar-
rangement is commonly used. In this case, the liquid rate
can be small and the unique %ow system is called a trickle
bed (e.g., Saroha, Nigam, Saxena, & Kapoor, 1998;
Iliuta, Ortiz-Arroyo, Larachi, Grandjean, & Wild, 1999).
The cocurrent up %ow requires higher pressure drop and
liquid hold-up in tower than the down %ow mode. A con-
tinuum model for numerical simulation of %ows in such
systems will be developed in this paper. Closure models
will be presented suitable for gas–liquid countercurrent
%ows.
∗Tel.: +1-780-492-0981; fax: +1-780-492-2881.
E-mail address: shijie.liu@ualberta.ca (S. Liu).
duct, the pressure drop is proportional to the %ow rate.
The same relationship holds also for %ow in curved ducts
when the %ow is weak (Liu, Afacan, Nasr-El-Din, &
Masliyah, 1994). This unique relationship can be gener-
alized to %ow through porous media as well,
dp
− gx ˙ −U; (1)
dx
where U is the superEcial %uid velocity, p is pressure, x
is the axis of the %ow direction,  is the %uid density and
gx is the gravity in the direction of %ow. This limiting
%ow behavior has received great attention for centuries.
The direct expression resulted from the proportionality
for %ow in porous media is called the Darcy’s law.
Several approaches have been used for the semi-
theoretical modeling of packed towers. When the %ow
velocity is high, Eq. (1) is not strictly adequate. In this
case, all the existing models converge to a simple (and
experimentally observed) relationship, which is com-
monly expressed as
dp 1
− + G gx = CD G UG2 ; (2)
dx 2
where CD is the so-called packing factor and has a re-
ciprocal length unit. It is inversely proportional to the
particle size and is a strong function of the packing

0009-2509/01/$ - see front matter ? 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 0 1 ) 0 0 2 3 2 - 9
5946 S. Liu / Chemical Engineering Science 56 (2001) 5945–5953
porosity as well. The packing factor is normally supplied
by manufactures for each speciEc random packings. Be-
cause the expression is usually used for dry gas %ow
through packed beds, a subscript G was used in Eq. (2).
A collection of data for the packing factor is also avail-
able in textbooks, for example, Trebal (1980).
Richards (1931) and Muskat (1937) intuitively ex-
tended the Darcy’s law to multiphase %ows. In their ex-
tension, the equation remains the same for each phase but
allowing the %uid properties as well as the permeability
to diJer. That is,
ki
ṽi = − ( pi − i g̃);
i
where the subscript i denotes the ith %uid phase, k is the
permeability and  is the dynamic viscosity. Although
widely used for %ow in soils and sandstones, Eq. (3) lacks
three key features (Liu, 1999b): namely, inertial eJects;
diJusion eJects and the inter-phase interactions (Dullien,
1992). These defects make Eq. (3) inapplicable to the
gas–liquid %ows in packed towers. In this paper, we shall
apply the volume-and-time averaging principles to derive
a governing equation in packed towers.
2. Volume-and-time averaged Navier–Stokes
equations for multiphase ows in porous media
Navier–Stokes equations are partial diJerential equa-
tions governing the %ow of a %uid. In particular, the con-
tinuity equation for the ith phase is given by,
@∗i
+  · (∗i ṽi∗ ) = 0;
@t
and the ith phase momentum equation is given by
@(∗i ṽi∗ )
+  · (∗i ṽi∗ ṽi∗ ) +  p∗ − ∗i g̃
@t
− · i∗ [ ṽi∗ + ( ṽi∗ )T ] = 0:
Although there are no diJerences between the intrinsic
phase averaged value and its corresponding local value
for (continuous) %ow in free space, the superscript ∗ in
the above equations is used to distinguish them from the
averaged quantities in a porous medium.
For %ow through porous media, Eqs. (4) and (5) are
still applicable within each continuous %uid phase. How-
ever, solutions of these equations become formidable due
to the complex and often unknown conEgurations of the
free voids and the %ow straits of each %uid phase. To
add to the complexity, at a given location, the occu-
pant phase can be diJerent at diJerent times. Therefore,
volume-and-time averaging is necessary in order for a
meaningful solution to be reached in most applications.
In this case, the porous medium will be treated as a con-
tinuous medium and thus no internal porous matrix con-
Eguration and %ow patterns need to be speciEed.
The volume-and-time averaged continuity equation is
given by
@(i hi )
+  · (i hi ũ i ) −  · (i ũ di ) = 0: (6)
@t
Eq. (6) is similar to the traditional phase balance (or con-
tinuity) equation except for the dispersion %ux ũ di . For
multi-phase %ow with at least one non-continuous phase,
there exists a dispersion %ux for each phase. Because of
(3) the diJerence in mobility (or resistance to %ow) for each
component, the relative distributions of the diJerent com-
ponents will be diJerent at diJerent spatial locations. The
diJerence in the relative distribution (or concentration)
will give rise to “hydrodynamic diJusion” or dispersion.
However, if all the phases are continuous, the phase dis-
persion is diminished.
Because of the dispersion, one may get a false sense
that the mass is not conserved. One should note that the
velocities are not volumetric transport %uxes in Eq. (6).
They are convection velocities for the respective phases.
There are two components in the transport %ux: convec-
tion and dispersion. Eq. (6) is a mass balance equation.
The mass transport %ux is given by
m̃i = i (hi ũ i − ũ di ); (7)
Finally, the volume-and-time averaged momentum
equation is given by
@i hi ũ i hi
+  · (i hi ũ i ũ i ) + ( pi − i g̃)
@t 
(4) −  · i [ hi ũ i + ( hi ũ i )T ] + i Fi ũ i



− ·  K ij ·  (hi i ũ i − hj j ũ j )
j

−  · (i K i ·  hi ũ i ) + ij (ũ i − ũ j ) = 0: (8)
(5) j
The above averaged Navier–Stokes equation is similar
to the averaged heat and mass transfer equations.
3. Volume-and-time averaged Navier–Stokes
equations and closure models for gas–liquid ows in
packed towers
For gas–liquid %ows in packed towers, the following
assumptions can be made to establish a simple mathe-
matical model: (1) the packing medium is isotropic and
random, (2) liquid wets the solid surface and there are
negligible contacts between the gas and the solid, (3)
there is no surface force eJect on the %ow behavior. In
 S. Liu / Chemical Engineering Science 56 (2001) 5945–5953 5947

particular, liquid is not foaming. With this assumption, tion can still be recognized. Liu and Long (2000) gave
the applicability of the following analyses will be limited
to %ows below loading point, (4) both gas phase and  ∗ + b 1 − b
= = b + Er [(1 − 0:3pd )
liquid phase are incompressible. This assumption is valid 2 2
  
strictly if the system is isothermal and isobaric, (5) the 2 R−r
%uids are Newtonian and the viscosity remains constant × cos + 0:3p d ; (11)
1 + 1:6 Er 2 pd d
in the tower.
where ∗ is the local azimuthal averaged porosity, pd is
3.1. The governing equations the period of oscillation normalized by the characteristic
(or nominal) particle size d; Er is the exponential decay-
The continuity equation for both liquid and gas phases ing function and b is the bulk porosity or the porosity
are given by Eq. (6) and the momentum equations can in an unbounded packing. For spheres of uniform size
be simpliEed from Eq. (8). ds ; pd = 0:94 and d = ds . For cylinders of equal height
and diameter such as Pall-rings, d is the
√ normal size (di-
@L hL ũ L hL
+  · (L hL ũ L ũ L ) + ( p − L g̃) ameter or height) and pd = 0:94(2 + 2)=3. The expo-
@t  nential decaying function is given by
−  · L [ hL ũ L + ( hL ũ L )T ] + L FL ũ L
 3=4 
R−r
−  · [K LG ·  (L hL ũ L − G hG ũ G )] Er = exp −1:2pd : (12)
d
−  · (L K L ·  hL ũ L ) + G FLG (ũ L − ũ G ) = 0 (9)
One can note that Eq. (11) is inadequate for the center
and region as the convenient condition: d=dr |r=0 = 0 may not
hold. The modiEed equation that satisfy the convenient
@G hG ũ G hG
+  · (G hG ũ G ũ G ) + ( p − G g̃) condition is given by
@t 
 ∗ + b 1 − b
−  · G [ hG ũ G + ( hG ũ G )T ] = = b + Er [(1 − 0:3pd )
2 2
−  · [K LG ·  (G hG ũ G − L hL ũ L )]   
2 R−r
× cos + 0:3pd ; (13)
+ G FLG (ũ G − ũ L ) = 0; (10) a + 1:6 Er 2 pd d

where the velocities, ũ i ’s, are “interstitial” convection. where a is a constant depending on the ratio of the par-
The transport velocities can be deEned similar to Eq. (7). ticle size to column size.
The velocities, hold-ups and pressure can be obtained by
 3=4 
solving Eqs. (6), (9) and (10) simultaneously together 2R R
a = − 1:6 exp −2:4pd
with no-slip conditions for both the gas and liquid phases n pd d d
on the column wall. However, this requires the complete
description of all the parameters involved. and
 
2 R
3.2. Closure model parameters n = int : (14)
1 + 1:6 exp[ − 2:4pd (R=d)3=4 ] pd d
In total, there are six parameters in the volume-and-time
averaged Eqs. (6), (9) and (10). These parameters 3.2.2. The shear factor for saturated single phase 4ow
are porosity, shear factors, interaction shear factor or The shear factor for saturated single phase %ow has
inter-phase interaction factor, phase dispersion %ux, mo- been reviewed by Liu and Masliyah (1996). The full im-
mentum dispersion coeOcient tensor and interaction plementation with porosity variations is given by Liu and
dispersion coeOcient tensor. These parameters need to Masliyah (1999). The shear factor, or %uid–solid matrix
be deEned (modeled or measured) before one can ap- interaction term, is given by
ply the volume-and-time averaged equations to simulate  
%ows in packed towers.  + b
F(ds ; ; |ũ|) = F0 ds ;
2
3.2.1. Porosity Re2
As an averaged quantity, short range variations (on + F1 (ds ; ) (Re − 3); (15)
Re2 + 36
or shorter than the particle scale) need not be consid-
ered. The averaging approach actively smoothes out where subscript 0 denotes for creeping %ow contribution
short-range local variations. However, long range varia- and I denotes for inertial contribution. Re is the %ow
5948 S. Liu / Chemical Engineering Science 56 (2001) 5945–5953

Reynolds number which is deEned as, Assuming the liquid phase is %owing along the surface
of the porous matrix, one obtains
Kinetic energy of %ow  
Re = 24 hL 1 dp
Channel diameter ×Pressure drop FL0 = − − gx ; (25)
L |ũ L |  dx L∞
(1=2)u2
e ds |ũ|
= 24
 = 2 ; (16) where the subscript L∞ denotes “single” liquid phase
dp
de − d x − gx 
∞ Elled in the channels formed by the porous matrix. Since
the liquid phase is covering all the solid surfaces, the void
where ue is the equivalent velocity of %uid in the pas- ratio  is exactly the porosity, and the pressure drop can
sage, de is the equivalent hydraulic passage diameter and be related to the %ow velocity in a passage through
the subscript ∞ denotes unidirectional negligible inertial  
%ow in the channel. dp 16k1 ueL
− − gx = L ; (26)
dx L∞ de deL
|ũ| |ũ|
ue = = (17) therefore,
A & 1=6
and 18(1 − hL )hL1=3
FL0 = : (27)
11=3 d2s
4 2ds
de = = ; (18) Similar to the shear factor for single phase saturated %ow,
ap 3(1 − )
the liquid phase shear factor is given by
where A is the eJective area ratio of the void space, & is  
0:363 1 1=2
the tortuosity of the passage and ap is the speciEc surface FL = FL0 0:637 + + 0:048 + 0:46(eL
area of the bed. s'L s'L
Using the model of Liu and Masliyah (1999), the shear 
ReL2
factors are deEned as × (ReL − 3) ; (28)
  ReL2 + 36
18(1 − ) 0:363
F0 (ds ; ) = 10=3 2 0:637 + ; (19) where the Reynolds number is deEned similar to saturated
 ds s' single phase %ow
 
18(1 − ) 1 (1=2) u2
FI (ds ; ) = 0:048 10=3 2 + 0:46(e1=2 ; (20) ReL = 24
L eL

 ds s' Qp
deL − L − gx
L∞
where (e is the passage curvature ratio,
L ds |ũ L |
(e = 1 − 1=2 (21) = 5=3 hL1=3 : (29)
L
and s' is the shape factor which is deEned as Because of the tortuorous voids, the liquid phase tends to
move with a straight and uniform pathway when possible.
passage length
s' = : (22) We propose a compromise such that
passage diameter  1=3
1 − hL 
The shape factor s' can be found from the relationship s'L = s' ; (30)
1 −  hL
of dry pressure drop and FI () for the single gas phase
%ow. The relationship is given by, and
 1=3
1 1 −  hL
CD G UG2 = G FI (ds ; ∗ ) Re|ũ G |: (23) (eG = (1 − 1=2 ): (31)
2 1 − hL 

3.2.3. Shear factor for the liquid phase
For multi-phase %ows, only Attou, Boyer, and Fer-
schneider (1999) and Long and Liu (2000) attempted a
shear factor model. These two papers treat the shear fac-
tors in a very similar manner except diJerent model bases
were used. In this paper, an extension of the model shown
in Section 3.2.2 will be presented for its suitability to
packed beds of high performance random packings. The
eJective velocity in the passage for the liquid %ow is
|hL ũ L | |ũ L |
ueL = = hL1=3 : (24)
AL &L 1=2
This treatment should allow us to model %ows whereby
the liquid phase is not only strictly %owing on the solid
surface. Because there is negligible solid matrix–gas
phase contact, the shear factor of the gas phase is negli-
gible.
3.2.4. The interaction shear factor between gas phase
and liquid phase
Since we assume that the gas phase is the non-wetting
phase, the gas–solid interaction is negligible, that is,
FG = 0 and K G = 0. The interface between the two phases
is equivalent to the solid surface for saturated single
 S. Liu / Chemical Engineering Science 56 (2001) 5945–5953 5949

phase %ow. Similar to the liquid shear factor, with the solid matrix. The liquid completely wets the
18(1 − hG ) solid matrix and therefore the liquid dispersion can be
FG0 = (32) treated in the same manner as the single phase saturated
h23=6
G dsG
2
%ow. The liquid phase dispersion tensor for an isotropic
and medium is known to be of the following form
     
0:363 1 1=2
)L 0 0 )L 0 0
FGL = FG0 0:637 + + 0:048 + 0:46(eG    
s'G s'G K L = KLrr  0 1 0  = ds DT |ũ L |  0 1 0  ; (39)
 0 0 1 0 0 1
ReG2
× (Re G − 3) ; (33)
ReG2 + 36 where DT is the normalized transverse dispersion coef-
Ecient and )L is the normalized longitudinal dispersion
where s'G is the shape fact for the gas phase, (eG is factor. For packed beds of spheres
the passage curvature ratio for the gas phase and dsG is
the eJective porous matrix grain size for the gas phase DT = 10:711=3 (1 − )2=3 (1 − 0:75 Er): (40)
%ow
Here the exponential decay function is introduced into
(1 − hG )1=3 the dispersion coeOcient because of the insuOcient av-
dsG = ds : (34)
(1 − )1=3 eraging to be used for the porosity near the column
wall.
Therefore, the low %ow rate limit interface shear factor
is given by
3.2.6. The interaction dispersion coe5cient tensor
18(1 − hG )1=3 (1 − )2=3 The gas phase is completely surrounded by the liq-
FG0 = : (35)
h23=6
G ds
2 uid phase. The dispersion due to the interaction between
the gas and the liquid phases can therefore be treated
Using the same deEnition as Eq. (16) and noting the as the gas phase is %owing through a Ectitious porous
porous matrix grain size for the gas phase and interaction matrix, the solids coated with a layer of mobile liquid.
with the liquid phase at the boundary, the gas Reynolds That is
number is given by  
 1=3 √
)L 0 0
1 − hG ds |G hG ũ G − G L hL ũ L |  
ReG = hG : K GL = KGLrr  0 1 0 
1− G 0 0 1
(36)  
 1=3 )L 0 0
Similar to the shape factor and curvature related terms 1 − hG  
for the liquid phase shear factor, the passage for the gas = ds DT |ũ G − ũ L |  0 1 0  (41)
1−
phase %ow is not rigid and therefore the modiEcations 0 0 1
of these terms must allow for the frequent pulsiEed %ow
and
pattern at the constrictions
 h = DT = 10:7h11=3 2=3
G (1 − hG ) (1 − 0:75 Er); (42)
1 − hG  G
s'G = s' (37)
1 −  hG which is the same as that for saturated single phase %ow
with porosity replaced by the holdup.
and
 hG =
1 −  hG
(eG = (1 − 1=2 ): (38) 4. Numerical solutions
1 − hG 
The closure model is introduced based on the promise We consider here fully developed two phase %ows in
that the liquid phase %ows along the solid surface and packed columns. The column is circular cylindrical and
forming an internal connected void in the passage pro- the %ows are axi-symmetrical. Only one non-zero aver-
portional to the void in the porous matrix. ModiEca- aged velocity component exists: the axial %ow velocity.
tions are subsequently made to account for the %exi- The cylindrical coordinate system is used with the axial
ble liquid Elm and the alternating %ow pattern at the (x-) axis pointing upwards (Fig. 1). All the Eeld variables
constrictions. are dependent only on the radial coordinate.
The continuity equations for the liquid and the gas
3.2.5. The liquid dispersion tensor phase are obtained from Eq. (6). We treat the case here the
The liquid dispersion is the dispersion of liquid phase bed porosity is high and pores are large. Liquid not only
Eeld variables due to the interaction of the liquid phase %ow alongside the solid surface, but forms droplets as
5950 S. Liu / Chemical Engineering Science 56 (2001) 5945–5953

Fig. 2. Normalized superEcial velocity distribution for a saturated

single phase %ow through a packed bed of spheres.

 
hG dp G d dhG uG
− G g − r
 dx r dr dr
1 d d(G hG uG − L hL uL )
Fig. 1. A schematic of the reference system for a packed tower. − rKGLrr
r dr dr
+ G FGL (uG − uL ) = 0; (48)
well. Therefore, there exists a dispersion %ux in the radial where uL and uG are the phase intrinsic axial %ow veloc-
direction due to the gradient of stress on the liquid=gas ities or the gas and the liquid phase, respectively.
phases similar to particulate dispersion in liquids (Liu, A visual Basic program has been developed to solve
1999a). Thus, we have the volume-and-time averaged Eqs. (43) – (48) based
on central diJerence method. The boundary conditions
L hL − G hG used are no slip conditions on the packed column wall.
FI = constant; i:e:; not a function of
 The program was running under the Excel environment.
spatial locations; (43) Some of the numerical simulations are presented in
Figs. 2–5.
 R Fig. 2 shows the computed velocity distribution for
L
L= 2rhL uL dr; (44) a saturated single phase %ow through a paced bed
R2 0
of spheres. One can observe that the current predic-
 R tion agrees with the experimental data of Kufner and
G Hofmann (1990) reasonably well. The locations of the
G= 2rhG uG dr; (45)
R2 0 peaks and troughs in the velocity proEle agreed quite
well because of the treatment imposed by Eqs. (13)
hL + hG =  (46) and (14). The diJerence in absolute values is due to
the fact that the experimental data is taken a few mm
and the momentum Eqs. (9) and (10), are reduced to above the actual packings—the peak and trough values
  have been greatly dampened. The oscillation (appear-
hL dp ance of peaks and troughs) in velocity proEle is due to
− L g + L FL uL + G FGL (uL − uG )
 dx the porosity variation in the radial direction. Therefore,
L d dhL uL 1 d d(L hL uL − G hG uG ) the strength of the current model is that it is capable
− r − rKGLrr of predicting the %ow behaviors for beds packed with
r dr dr r dr dr
large particles. In this case, only two particles can be
1 d dhL uL packed on the column radius and the predictions are still
− rL KLrr = 0; (47)
r dr dr reliable.
 S. Liu / Chemical Engineering Science 56 (2001) 5945–5953
Fig. 3. Pressure drop and liquid hold-up variation with gas load for
air–water %ows in a column of 0:3 m in diameter packed with 50 mm
plastic Nor-PacJ rings.
Fig. 3 shows the variation of liquid hold-up and pres-
sure drop for air–water %ow in a typical packed tower (for
distillation operations). One can observe that the pres-
sure drop calculation agrees quite well with the experi-
mental data of Billet (1995) for %ow below the loading
point. Fig. 4 shows the variation of pressure drop and liq-
uid hold-up for air–water %ow variation with the liquid
rate. The pressure drop increases with increasing liquid
and=or gas loads. Following the traditional presentation,
√
the gas capacity factor, Fv = UG G , is also shown in
Fig. 3 alongside the average gas superEcial velocity. As
shown in Fig. 3, the pressure drop increases with gas load
quite smoothly—the proEle is nearly a straight line on
a log–log plane. In contrast, the pressure drop increases
with increasing liquid load gradually at low liquid load
and sharply at high liquid load. The model predictions for
the pressure drop agreed quite well with the experimental
data (Fig. 3). However, when the gas load exceeds the
%ooding point, the model is incapable of predicting the
5951
Fig. 4. Pressure drop and liquid hold-up variation with liquid load
for air–water %ows in a column of 0:3 m in diameter packed with
50 mm plastic Nor-PacJ rings.
sharp increase in pressure drop (near the %ooding point)
as the gas load (or gas capacity factor) is increased.
The liquid hold-up changes with the gas rate slowly
below the %ooding point (Fig. 3). Near the loading point,
the liquid hold-up increases sharply with increasing gas
load. The liquid hold-up increases with increasing liq-
uid load. The predictions for the liquid hold-up are in
agreement with the experimental data as depicted in
Fig. 4.
Fig. 5 shows that the computed radial liquid velocity
proEles agree well with the experimental data of Kouri
and Sohlo (1996) for various packings and under vari-
ous operating conditions. In general, the predicted super-
Ecial liquid velocity proEle is oscillatory in the diameter
of the column. On the wall, the no-slip condition is im-
posed and the liquid velocity is zero. The liquid velocity
increases sharply to a maximum just away from the wall
and then decreases sharply to a minimum further away
from the wall. The minimum liquid velocity is located
near the minimum porosity. Therefore, the number of
peaks in the liquid velocity proEle depends on the num-
ber of periods for the porosity variation along the column
radius. For example, the nominal size, d, of a Pall-ring
(Figs. 5a–c) is the diameter of the ring and is the same as
the length of the ring. The saddles, on the other hand, are
nearly semicircular on the base and with a height about
one third of its base diameter. The nominal size of a sad-
dle is its diameter and is the longest dimension for the
particle. Therefore, there are more peaks in the liquid ve-
locity proEle for packed beds of saddles (Fig. 5d) than
for packed beds of Pall-rings with similar nominal parti-
cle sizes. The experimental data are piecewise averaged
liquid velocity, which is a rather good representation of
5952 S. Liu / Chemical Engineering Science 56 (2001) 5945–5953

Fig. 5. Liquid velocity distributions for gas–liquid %ows in a column of 0:5 m in diameter and packed with random packings 3 m deep: (a)
and (b) 25 mm plastic Pallrings; (c) 50 mm plastics Pallrings; and (d) 38 mm ceramic Intalox saddles. The dot–dot–dashed lines are piece-wise
averaged experimental data from Kouri and Sohlo (1996).

the computed liquid velocity proEle averaged in the re- Notation

spective regions. They also show a high liquid rate near
the wall. A aeral porosity
ap solid surface area per unit bed (column) volume
a constant in porosity variation expression
5. Conclusions CD packing factor
DT constant in the dispersion coeOcient expression
The volume-and-time averaging technique has been d nominal (or characteristic) particle size
applied to multiphase %ow in porous media=packed beds. de hydraulic pore diameter
It is found that the dispersion term is present even in the ds equivalent spherical particle diameter
continuity equation. As such, the %ow velocities are con- Er radial exponential decay function for porosity
vective velocities, not the volumetric transport %uxes. The F shear factor or phase–phase momentum interac-
volume-and-time averaged Navier–Stokes equations are tion factor
highly non-linear since the interaction coeOcients are lin- F0 viscous term of the shear factor
early related to the %ow velocities when the %ow is strong. FI inertial term of the shear factor
√
The dispersion coeOcient and=or the eJective viscosity is Fv gas capacity factor, = UG G
also linearly related to the %ow velocities. Models of clo- G gas load or gas mass discharge velocity
sure parameters are presented in this paper, including the g gravitational acceleration
shear factors, porosity variation and dispersion coeOcient h phase hold-up
tensors. These closure models are specially designed for K dispersion coeOcient tensor (Kxx ; Kxy ; Kxz ; : : :)
packed towers of highly porous random packings. Incor- k permeability
porated with the porosity variations in the radial direc- k1 passage shape factor: e.g., k1 = 2 for a circular
tion, the model can predict the %ow behaviors even for duct and k1 = 3 for parallel plates
small column to particle diameter ratios. Several numeri- L liquid load or liquid mass discharge velocity
cal simulations are presented. The %ow velocities, liquid m̃ superEcial mass velocity
hold-up and pressure drop agree well with experimental n integer constant in porosity variation function
data. p pressure
 S. Liu / Chemical Engineering Science 56 (2001) 5945–5953 5953

pd period of oscillation in multiple of nominal par- Acknowledgements

ticle diameter
R column or bed radius The author is indebted to the Natural Sciences and
Re Reynolds number Engineering Research Council of Canada for Enancial
r radial coordinate support.
s' particle shape factor
T temperature
t time References
U superEcial %uid discharge velocity, i.e., superE-
Attou, A., Boyer, C., & Ferschneider, G. (1999). Modelling of the
cial %uid velocity averaged on the cross-sectional hydrodynamics of the cocurrent gas–liquid trickle %ow through a
plane trickle-bed reactor. Chemical Engineering Science, 54, 785–802.
u intrinsic %uid phase velocity (local, based on the Billet, R. (1995). Packed towers in processing and environmental
phase hold-up) technology. New York, USA: VCH publishers.
udi dispersion %ux of phase i Dullien, F. A. L. (1992). Porous media, 4uid transport and pore
structure (2nd ed.). Toronto: Academic press.
ue average velocity in a passage Herskowitz, M., & Smith, J. (1983). Trickle-bed reactors. A review.
ṽ superEcial %uid transport velocity A.I.Ch.E. Journal, 29, 118.
x x-coordinate Iliuta, I., Ortiz-Arroyo, A., Larachi, F., Grandjean, B. P. A., &
y y-coordinate Wild, G. (1999). Hydrodynamics and mass transfer in trickle-bed
z z-coordinate reactors: An overview. Chemical Engineering Science, 54(21),
5329–5337.
Kister, H. Z. (1992). Distillation design. New York: MacGraw-Hill.
Greek letters Kouri, R. J., & Sohlo, J. (1996). Liquid and gas %ow patterns in
random packings. Chemical Engineering Journal, 61, 95–105.
ij momentum interaction factor between phase i Kufner, R., & Hofmann, H. (1990). Implementation of radial
and phase j porosity and velocity distribution in a reactor model for
heterogeneous catalytic gas phase reactions (Torus-Model).
)L ratio of longitudinal dispersion coeOcient to the Chemical Engineering Science, 45, 2141–2146.
transversal dispersion coeOcient Liu, S. (1999a). Particle dispersion for suspension %ow. Chemical
 porosity Engineering Science, 54, 873–891.
∗ local porosity or azimuthal averaged porosity Liu, S. (1999b). A continuum approach to multi-phase %ow in porous
. azimuthal coordinate media. Journal of Porous Media, 2(3), 295–308.
Liu, S., Afacan, A., Nasr-El-Din, H. A., & Masliyah, J. H. (1994). An
(e eJective pore curvature ratio experimental study of pressure drop in helical pipes. Proceedings
 dynamic viscosity of the Royal Society London A, 444, 307–316.
 density Liu, S., & Long, J. (2000). Gas–liquid countercurrent %ows through
& tortuosity, or the ratio of the straight line distance packed towers. Journal of Porous Media, 3(2), 99–113.
and the length of eJective %uid pathway between Liu, S., & Masliyah, J. H. (1996). Single %uid %ow in porous media.
Chemical Engineering Communication, 148–150, 653–732.
two points Liu, S., & Masliyah, J. H. (1999). Non-linear %ows in porous media.
Journal Non-Newtonian Fluid Mechanics, 86, 229–252.
Subscripts Muskat, M. (1937). The 4ow of homogeneous 4uids through porous
media. New york: McGraw-Hill.
b bulk or far away from the wall Richards, L. A. (1931). Capillary conduction of liquids through porous
media. Physics, 1, 318–333.
G gas phase Saroha, A. K., Nigam, K. D. P., Saxena, A. K., & Kapoor, V.
i phase i K. (1998). Liquid distribution in trickle-bed reactors. A.I.Ch.E.
j phase j Journal, 44(9), 2044–2052.
L liquid phase Shah, Y. T. (1979). Gas–liquid–solid reactor design. New York:
r r-directional McGraw-Hill.
Trebal, R. E. (1980). Mass transfer operations (3rd ed.). Toronto:
x x-directional McGraw-Hill.
∞ single phase inertia-free %ow