You are on page 1of 9

Chemical Engineering Science 56 (2001) 1085}1093

In#uence of gas distribution and regime transitions on liquid velocity


and turbulence in a 3-D bubble column
Ch. Vial, R. LaineH , S. Poncin*, N. Midoux, G. Wild
Laboratoire des Sciences du Ge& nie Chimique CNRS-ENSIC-INPL, 1, rue Grandville BP 451, F-54001 Nancy Cedex, France
Abstract
Laser Doppler anemometry has been applied to a bubble column equipped successively with three di!erent gas distributors:
a single-ori"ce nozzle, a multiple-ori"ce sparger and a porous plate. Axial and tangential mean and rms liquid velocity values have
been measured up to a gas hold-up of 15}20%. A simple analytical model based on a bubble-induced turbulence is shown to be
consistent with the results obtained with the three spargers and is used to estimate the Reynolds shear stress. The evolution of the local
hydrodynamic parameters of the liquid phase with ;
%
is explained and related to the hydrodynamic regime, the uniformity of the gas
distribution and the start-up procedure. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Bubble column; Flow regime; Gas distributor; Hydrodynamics; LDA
1. Introduction
Bubble column design and scale-up still represents
a hard task because the in#uence of operating conditions,
reactor geometry and physico-chemical properties of
each phase on reactor performance are not yet fully
understood. For example, phenomena, such as in#uence
of the gas sparger, remain di$cult to quantify and the
origin of regime transitions is not yet completely clear.
Hydrodynamics of the liquid phase has a strong in#uence
on mixing, heat and mass transfer and therefore on
bubble column performance. However, a correct descrip-
tion has not been possible as robust measuring tech-
niques were lacking. Only recently new techniques such
as CARPT (Chen et al., 1999) and PIV (Delnoij, Kuipers
& van Swaiij, 1997) or improvements of older techniques
such as LDA (Mudde, Groen & van den Akker, 1997a)
have allowed a deeper insight in the local phenomena of
the liquid phase. Nevertheless, how liquid-phase behav-
iour is a!ected by gas distribution and regime transition
has not yet been studied extensively with these tech-
niques. Only the work of Chen, Reese and Fan (1994)
using PIV in a 3-D bubble column has shown the main
features of the three regimes they identi"ed: bubbly yow,
vortical-spiral yow and turbulent yow. The vortical-spiral
* Corresponding author. Tel.: #3-83-175223; fax: #3-83-322975.
E-mail address: souhila@ensic.inpl-nancy.fr (S. Poncin).
#ow structure corresponds to the commonly named
transition region, whereas turbulent #ow corresponds to
the heterogeneous regime (Vial et al., 2000). The aim of
this work is to provide a better understanding of liquid
#ow characteristics and to highlight the in#uence of the
sparger and regimes transitions using LDA.
2. Experimental set-up
The reactor is a cylindrical semibatch bubble column
made of plexiglas, with an inside diameter of 0.1 m and
a height of 2 m. It is equipped with a square box located
at a height of 0.8 m to avoid optical distortion when
optical methods are used, such as photography or LDA.
The super"cial gas velocity ;
%
is varied from 6 mm/s up
to 15 cm/s using two rotameters. A diagram of the set-up
is available elsewhere (Camarasa et al., 1999). The col-
umn can be equipped with three spargers: a single-ori"ce
nozzle (5 mm in diameter), a multiple-ori"ce nozzle (50
holes of 1 mm uniformly spaced) and a porous glass plate
(mean pore diameter: 15 m). The distributors are pre-
sented in Fig. 1. The porous plate and the multiple-ori"ce
sparger produce a uniform gas primary distribution,
whereas the single-ori"ce nozzle produces a highly non-
uniform one. In the present experiments, the air/water
system is always used at room temperature and atmo-
spheric pressure.
0009-2509/01/$- see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 0 0 ) 0 0 3 2 5 - 0
Fig. 1. Di!erent types of gas sparger used.
Fig. 2. LDA data rates at di!erent gas #ow rates and radial positions
with and without seeding.
Local time-averaged liquid velocity and rms velocity
(i.e. root mean square #uctuating velocity) are measured
in the axial and the tangential directions with a one-
component laser Doppler anemometer (LDA) from
Aerometrics operating in backscatter mode and equip-
ped with a RSA-1000L processor. Local measurements
are performed at mid-height of the column, far enough
from the gas distributor and the free surface to avoid end
e!ects. Laser light is provided by a 4 W Ar> laser. Mudde
et al. (1997a) have shown that LDA does not see the
bubbles and only measures liquid velocity in this mode
when their diameter is larger than 1 mm and when
a proper seeding is used. This is veri"ed by performing
preliminary measurements with and without seeding par-
ticles. Even with a poor seeding and tap water, the
ratio between the data rate obtained in both cases is at
least 25 (Fig. 2) and can reach a value higher than 100 in
distilled water. The wide di!erence between the results
with and without seeding con"rms that only the liquid
velocity is measured. To obtain a high data rate, the #ow
has to be seeded with small particles (diameter 5 m)
which have almost no slip and the quantity of particles
added has to be optimised. The technique is however
limited by the opacity of the dispersion at high gas
hold-up, but accurate measurements can be obtained up
to a gas hold-up value between 15 and 20%. Conse-
quently, measurements are possible with the di!erent
spargers up to a gas #ow rate of 8.5 cm/s. A duration of
180 s is generally su$cient to obtain time-averaged velo-
city values with an accuracy of $1 cm/s. Several
measurements have been performed at a same location in
order to verify this point and to improve the accuracy of
the experimental results.
3. Preliminary results (6ow regimes and gas phase)
Regime analysis can be performed using the evolution
of the average gas hold-up c
%
, with ;
%
(Vial et al.,
2000). The single-ori"ce nozzle operates always in the
heterogeneous regime. On the contrary, homogeneous
conditions prevail with the multiple-ori"ce nozzle when
;
%
is lower than 4 cm/s and fully established heterogen-
eous regime is reached at ;
%
"11}12 cm/s. With the
porous plate, transition is strongly dependent on the
start-up procedure (Prakash & Briens, 1990). With
a `drya start-up, transition points are identical to those of
the multiple-ori"ce sparger. But with a `weta start-up,
heterogeneous regime can be reached at a lower gas #ow
rate around 7}8 cm/s. This property is used to obtain the
three di!erent #ow patterns within the range of super"-
cial gas velocity that can be studied with LDA. Regime
transition points are summarised in Fig. 3.
Local gas hold-up is measured using an optical "bre
probe. A detailed description of the set-up and of the
experimental results with the three spargers is reported
elsewhere (Camarasa et al., 1999). It is well known that
the saddle-shaped pro"les measured can be represented
by the following expression:
c
%
"c
%
,
m#2
m
(1!K), (1)
where is the dimensionless radius. Eq. (1) has been
adjusted on experimental pro"les optimising c
%
, and m.
Optimised values of c
%
, have been veri"ed to be in
good accordance with experimental values. Optimised
values of m are reported in Fig. 4 for the three spargers.
Their evolution con"rms the classical assumption that
gas hold-up pro"les are #at in the homogeneous
1086 C. Vial et al. / Chemical Engineering Science 56 (2001) 1085}1093
Fig. 3. In#uence of gas distributor and gas #ow rate on #ow pattern
(Hom."Homogeneous Het."Heterogeneous).
Fig. 4. Evolution of the optimised values of the exponent m (Eq. 1).
Fig. 5. Radial evolution of the mean axial liquid velocity component
with ;
%
.
Fig. 6. Radial evolution of the rms liquid with ;
%
.
regime and present a parabolic shape when transition
and heterogeneous conditions prevail.
4. Experimental results
Local liquid mean and rms velocity values have been
measured with the three spargers in the axial and the
tangential directions. The experimental results of both
quantities are illustrated, respectively, in Figs. 5 and 6 for
the single- and multiple-ori"ce distributors. The classical
symmetrical pro"les with liquid inversion near the wall
are obtained (Joshi, 1980; Zehner, 1986).
Surprisingly, #at pro"les and low axial mean velocity
values are measured in the heterogeneous regime, where-
as the transition region is characterised by an established
circulation pattern with steeper pro"les and higher
values of the local liquid velocity. Consequently, the #ow
in heterogeneous conditions seems to be more uniform
than in the transition region. This result is in good
agreement with the work of Chen et al. (1994) which
shows a vortical-spiral structure in the transition region
disappearing at higher gas #ow rate when heterogeneous
conditions prevail. This behaviour is illustrated by
Fig. 7 which presents the evolution of the liquid
C. Vial et al. / Chemical Engineering Science 56 (2001) 1085}1093 1087
Fig. 7. Evolution of the liquid centreline velocity v
!
with ;
%
for the
three spargers.
Fig. 8. Evolution of the centreline turbulence intensity I
!
with ;
%
for
the three spargers.
Fig. 9. Radial evolution of the rms liquid with ;
%
using the multiple-
ori"ce distributor.
centreline velocity v
!
with ;
%
for the three spargers.
When the column is equipped with the single-ori"ce or
the multiple-ori"ce distributors, v
!
is shown to be pro-
portional to (;
%
)? with : between 0.33 and 0.375, as
commonly predicted by several correlations (Riquarts,
1981; Zehner, 1986). However, the correlation of Zehner
(1986) is adequate only when a sharp circulation pro"le is
established, i.e. in the transition region. For the air/water
system, this correlation is expressed as follows:
v
!
"0.73;"``
%
. (2)
On the contrary, when established heterogeneous con-
ditions prevail, the pre-exponential coe$cient is far
lower. An optimisation provides Eq. (3) for the centreline
velocity measured with the single-ori"ce nozzle:
v
!
"0.37;"``
%
. (3)
The strong in#uence of the #ow pattern is clearly
demonstrated by the curve obtained with the porous
plate which presents a behaviour similar to that obtained
with the multiple-ori"ce sparger at low gas #ow rate, but
tends to behave like the single-ori"ce nozzle when the
column operates under the heterogeneous regime. A sim-
ilar behaviour had been already noticed by Franz,
BoK rner, Kantorek and Buchholz (1984) but had not been
explained. This result is also in agreement with the work
of Camarasa et al. (1999) in which a sharp decrease is
reported in local bubble velocity at incipient heterogen-
eous regime with the porous plate.
The evolution of the rms velocity o
*
is also character-
istic. Radial pro"les of o
*
present a smooth maximum
when the column is equipped with the multiple-ori"ce
sparger and the porous plate (Fig. 6a), but they are
almost #at with the single-ori"ce nozzle (Fig. 6b). With
the three distributors, o
*
increases regularly with ;
%
,
but the values measured with the single-ori"ce nozzle are
higher. This behaviour is opposite to that of the mean
velocity which is higher with the multiple-ori"ce distribu-
tor. The high values of rms obtained in the heterogeneous
regime correspond to the chaotic aspect of the #ow in this
regime. The evolution observed with the porous plate is
illustrated by Fig. 8 which represents the turbulence
intensity at the centre of the column. This quantity is
de"ned as the ratio of o
*
to v
!
. Fig. 8 shows that the
column behaves as if it was operated with the multiple-
ori"ce sparger when homogeneous conditions prevail,
but it tends to behave as if it was equipped with the
single-ori"ce sparger when the established heterogeneous
begins and I
!
increases steeply with ;
%
Tangential mean velocity is shown to be close to 0 in
the whole range of gas #ow rate used in this study. The
rms velocity o
*
presents similar trends as its axial
values, but is lower. This is in accordance with previous
results of Franz et al. (1984) and Mudde et al. (1997a).
The ratio between o
*
and o
*
takes values 1.1}1.5. In
Fig. 9, the evolution of this ratio at the centre of the
column is presented. The highest values correspond to
the transition region whereas the lowest are obtained
when the heterogeneous regime is established. This little
1088 C. Vial et al. / Chemical Engineering Science 56 (2001) 1085}1093
anisotropy is due to the upward movement of the
bubbles, but it tends to disappear when heterogeneous
conditions prevail. This con"rms that the #ow is more
uniform and agrees with the evolution of the centreline
velocity.
Velocity #uctuations with time of such an intensity in
both directions do not exist in pure liquid #ow and can
only be attributed to bubble wakes. This is in agreement
with previous results as mixing in bubble columns, which
is of primary importance for reactor performance, has
been shown to be mainly due to bubble wakes (Fan
& Tsuchiya, 1990).
The experimental results can be explained using
both visual observation of the dispersion and previous
results of Camarasa et al. (1999) which have described
the gas phase. The transition is characterized by a local
recirculation of the liquid especially near the wall
where the mean liquid velocity is negative. The gas
phase tends to move towards the centre of the column,
which induces the parabolic gas hold-up pro"les
observed. High positive values of the liquid velocity can
be measured at the centre of the column and high nega-
tive values are observed near the wall. When established
heterogeneous regime is reached, large bubbles (1 cm or
more) are obtained. The existence of these large bubbles
is con"rmed by the decrease of the bubble frequency
measured with an optical probe (Camarasa et al., 1999).
The large bubbles move upward with a high velocity,
but their passage also induces important recirculat-
ions of liquid including smaller bubbles. This had
been observed by Camarasa et al. (1999), using an
ultrasonic Doppler device to measure local bubble
velocity distribution: a wider distribution was observed
in the heterogeneous regime, especially towards the
negative values. This behaviour corresponds to higher
#uctuations with time of the liquid velocity. However,
as high upward and downward liquid #uctuations
occur successively at each radial position in the
column, the mean circulation structure is destroyed
and more uniform velocity pro"les are obtained in the
liquid phase, which corresponds to a decrease of the
mean axial centreline velocity and explains the experi-
mental results.
As a conclusion, LDA measurements have been
shown to be able to provide a deeper insight in the
local structure of the liquid phase up to a gas hold-up
of 15}20%. The results are consistent with previous re-
sults obtained using PIV (Chen et al., 1994), optical
probes and an ultrasonic Doppler technique (Camarasa
et al., 1999). They contribute to a better understanding
of #ow regimes and of the in#uence of the gas distri-
butor on the local behaviour of the dispersion. In
our opinion, they will soon provide a better understand-
ing of mixing in the liquid phase as both mean pro"les
and the instantaneous values of the local liquid velocity
are obtained.
4. Analytical model and shear stress estimation
4.1. Description of the model
While the axial and the tangential Reynolds normal
stresses can be deduced from the rms velocity values, the
Reynolds shear stress cannot be directly measured with
a 1-D LDA system. Nevertheless, the local shear stress
can be estimated using the experimental pro"les of the
mean axial velocity and a simple model based on the
momentum balance equations:
j
G
c
G

c
ct
#v
G
) V

v
G
"!c
G
VP#V) (c
G
t
IG
)
#V) (c
G
t
R
)#j
G
c
G
g#f
G
, (i"L, G).
(4)
In Eq. (4), P is the pressure, v
G
is the velocity vector and
t
IG
the viscous stress tensor of phase i. f
G
represents the
interfacial forces of phase i (such as f
%
"!f
*
) and t
R
is
the Reynolds stress tensor. A simpli"ed relation can be
obtained by summing the general momentum balance
equation for each phase and by neglecting the viscous
and the inertial e!ects in the gas phase. If the #ow is fully
established, only the axial pressure evolution and the
radial evolution of the shear stress have to be taken into
account. The following expression is "nally obtained:
!
dP
dz
#
1
r
dt
dr
#j
*
(1!c
%
)g+0, (5)
where t is de"ned as
t"t
R
#c
*
t
I*
(with t
R
"j
*
v'
*
v'
*
). (6)
If the radial evolution of the gas hold-up can be modelled
using Eq. (1) that leads to
t"
j
*
gR
2

!dP
j
*
g dz
!1#c
%
,#
2c
%
,
m
(1!K)

. (7)
As the axial pressure drop can be simply related to the
wall shear stress t
U
using "1, t
U
is expressed as
t
U
"
j
*
gR
2
!1
j
*
g
dP
dz
!1#c
%
,

. (8)
Eqs. (6) and (7) lead to the following relation:
t"t
U
#
j
*
gR
2

2c
%
,
m
(1!K). (9)
The existence of a thin laminar layer (z))1) near the
wall is generally assumed. As a "rst approximation, the
C. Vial et al. / Chemical Engineering Science 56 (2001) 1085}1093 1089
Fig. 10. Theoretical velocity pro"le in a bubble column.
Table 1
Turbulence models compared in this work
Turbulence model (model number and equation) Eq. no.
1. v
R
"k' (12)
2. v
R
"k'c
%
(1!c
%
) (13)
3. v
R
"
k'R
A
6
t
N
j
*
(1!`)(1#2`) (14)
(Menzel, 1990)
4. v
R
"k'(g;
%
)`D"`
A
(1!`)"` (15)
(Burns & Rice, 1997)
5. l
K
"maxby

k'
cG
cG,
; R(0.14!0.08`!0.06")

(16)
(Geary & Rice, 1992)
#ow may be represented using the structure described in
Fig. 10 which assumes that the value of v
*
is minimal
when "z. Consequently, t"0 at this point and the
"nal expressions obtained are
t
U
"!
j
*
gR
2
2c
%
,
m
(1!zK), (10)
t"
j
*
gR
2

2c
%
,
m
(zK!K). (11)
The exponent m and the average void fraction c
%
, can
be estimated using the previous results. Eqs. (10) and (11)
may be used to predict the liquid velocity pro"le if an
adequate model is chosen to represent turbulence. Many
models have been proposed in the literature. Models
based on a turbulent viscosity v
R
can be distinguished
from those based on a mixing length l
K
. The turbulent
viscosity v
R
is sometimes assumed to be constant on the
whole section of the column (Ueyama & Miyauchi, 1979),
but it may be assumed to have a behaviour qualitatively
similar to that of a pure liquid (Menzel, In der Weide,
Staudacher, Wein & Onken, 1990, Eq. (14)). Another
possibility is to consider that v
R
is closely linked to the gas
hold-up. A last class of turbulence models is based on the
fact that v
R
is related to a characteristic length (most often
the bubble or the column diameter) and to the power
input brought by the gas: g;
%
(Burns and Rice, 1997,
Eq. (15)). Turbulence models based on a mixing length
are less frequent. Rice and Geary (1990) have proposed
a formulation which assumes that turbulence is mainly
induced by bubble wakes. Then, Geary and Rice (1992)
have improved this model and have considered that
classical turbulence competes with bubble-induced tur-
bulence (Eq. (16)).
Among these models, "ve belonging to the di!erent
classes have been selected and are reported in Table 1. In
this table, k' is a parameter which does not depend
directly on ;
%
, but may depend on the prevailing #ow
regime.
The theoretical velocity pro"les can then be calculated
using Eqs. (10) and (11), a model of Table 1 and the
following set of equations for models 1}4:
t"!
c
*
j
*
#j
R
R
dv
*
d
(0))z), (17)
t"!
c
*
j
*
R
dv
*
d
(z))1), (18)
or the following relations for model 5
t"!
c
*
j
*
R
dv
*
d
#j
*
l`
K
R`
dv
*
d
`
(0))z), (19)
t"!
c
*
j
*
R
dv
*
d
(z))1). (20)
Two additional conditions are necessary:
v
*
("z)"v
*
("z>), (21)
v
*
("1)"0. (22)
In the previous equations, the two unknown parameters
to be estimated are t
U
(or z) and k'. They can be obtained
using the experimental results obtained in Section 3.
The experimental velocity pro"les have to be "tted with
the theoretical pro"les, optimising the values of z and k'.
The "ve models of turbulence proposed in Table 1 have
been applied successively to the results obtained with the
three spargers. The evolution of k' is used to test the
consistency of the models, as k' should depend essentially
on the hydrodynamic regime. Once a consistent model is
determined, t
U
and the radial evolution of t can be
obtained.
4.2. Results
Model 1 is shown to be inadequate because the turbu-
lent viscosity presents a radial evolution. Indeed, nega-
tive velocity values near the wall are underestimated by
the calculations (Fig. 11). Models 2 and 3 provide a better
"tting of experiments, but the evolution of k' is not
1090 C. Vial et al. / Chemical Engineering Science 56 (2001) 1085}1093
Fig. 11. Comparison of experiments and theoretical velocity pro"les.
Fig. 12. Evolution of the optimized averaged mixing length l
K
, with
;
%
and the three spargers.
Fig. 13. Evolution of t
U
with ;
%
for the three spargers.
correct. Model 4 is consistent with the results obtained
with the single-ori"ce nozzle if k'"0.16$0.01, but not
with the other distributors. Finally, only model 5 is
consistent for the three spargers. The following values of
k' are obtained: k'"0.0059$0.0005 in the transition
region (multiple-ori"ce sparger) and k'"0.012$0.001
in the heterogeneous regime (single-ori"ce nozzle).
The results are in agreement with the theoretical analy-
sis of Geary and Rice (1992) which is based on the
assumption that k' is proportional to the size of the
bubble wake which depends on the bubble size and
morphology. Consequently, the rise in k' value is due to
an increase in bubble wake volume which comes from the
presence of larger bubbles. With our experimental set-up
of 10 cm in diameter, only bubble-induced turbulence
prevails and Eq. (16) can be simpli"ed as
l
K
"k'
c
%
c
%
,
. (23)
As a consequence, the parameter k' is equal to l
K
,,
the averaged mixing length on the cross-sectional area.
The evolution of l
K
, with ;
%
is reported in Fig. 12
for the three spargers. This behaviour is in accordance
with the experimental results (Figs. 7 and 8) as the values
of l
K
, for the porous plate are close to those of the
multiple-ori"ce sparger when ;
%
is low and near to those
of the single-ori"ce nozzle when ;
%
is high. l
K
, values
in the heterogeneous regime are at least twice as high as
in the transition region. A slight increase in l
K
, with
;
%
is also observed with the multiple-ori"ce sparger.
This is not contradictory as the bubble size may increase
with ;
%
even if no bubbles larger than 1 cm are reported,
contrary to what is observed with the single-ori"ce
nozzle.
Optimised values of z con"rm that the boundary layer
is thin and z is always higher than 0.99. The evolution of
the wall shear stress t
U
with ;
%
is deduced from that of
z using Eq. (10) and is reported in Fig. 13. Fig. 13
shows that the predicted time-averaged wall shear
stress is higher in the transition region than in the hetero-
geneous regime. This surprising result is simply due to
the lower liquid velocity values near the wall when het-
erogeneous regime prevails (Fig. 7). Consequently, the
in#uence of #ow transition with the three spargers agrees
with the previous results on liquid velocity and mixing
length.
The radial evolution of the shear stress t is reported in
Fig. 14 for the column equipped with the porous plate.
Similar trends are obtained with the other distributors:
t peaks when "
"
, which is between 0.65 and 0.8,
depending on ;
%
. It appears that
"
tends to decrease as
;
%
increases. Fig. 14 shows clearly that t
U
is of little
in#uence on t, except in the wall region. As a "rst approx-
imation, t
U
can be neglected in Eq. (9), which corre-
sponds to z+1 in Eq. (11).
The order of magnitude of the shear stress can be
compared to the axial normal stress. As this quantity
presents an important radial evolution (Fig. 14), we
C. Vial et al. / Chemical Engineering Science 56 (2001) 1085}1093 1091
Fig. 14. Radial evolution of the Reynolds shear stress with ;
%
for the
columns equipped with the porous plate.
Fig. 15. Evolution of t
U
with ;
%
for the three spargers.
suggest comparing the following criterion:

}
`
"Min[v'`
*
/v'
*
) v'
*
]. (24)
The evolution of

}
`
with ;
%
for the three spargers is
reported in Fig. 15. The shear stress values are shown to
be at least one order of magnitude lower than the axial
normal stress. This agrees with the experimental results
of Mudde, Lee, Reese and Fan (1997b) who have found
similar ratio using PIV, even if their work is limited to
low gas #ow rate in a 2-D bubble column.
Finally, these results show clearly the strong aniso-
tropy of the stress tensor in bubble columns. It explains
why turbulence models proposed for pure liquid or gas
#ows, such as the k}c model, do not represent adequately
the dispersion in bubble columns except at very low gas
input (Sokolichin & Eigenberger, 1999). As a conclusion,
a model has been proposed, which is consistent with the
experimental results obtained with the three spargers.
This model is based on the assumption of a bubble-
induced turbulence due to bubble wakes and uses a mix-
ing length proportional to the local gas hold-up. The
radial pro"les of t and the evolutions of t
U
and l
K
, have
been shown to con"rm the experimental results.
5. Summary and conclusions
In this work, the time-averaged liquid mean and rms
velocity pro"les have been measured using LDA in
a bubble column equipped with three di!erent spargers
up to gas hold-up values of 15}20%. Their evolution
with ;
%
has been related to gas distribution (uniform or
non-uniform distribution and start-up procedure for the
porous plate) and hydrodynamic regimes (homogeneous
regime, transition region and heterogeneous regime). The
well-known liquid circulation pattern has been reported
in the transition region, but this structure is destroyed
when large bubbles appear at the beginning of the estab-
lished heterogeneous regime. A simple model based on
momentum balance equations can be used to evaluate
the Reynolds shear stress if an adequate turbulence for-
mulation is chosen. Five models have been selected in the
literature and the model of Geary and Rice (1992) has
been shown to be consistent with the results obtained
with the three spargers. The evolution of t pro"les and of
t
U
with ;
%
has been analysed and the strong anisotropy
of turbulence has been outlined. Finally, a precise de-
scription of the local hydrodynamics of the liquid phase
has been obtained and its evolution with ;
%
has been
explained. We hope that this study will be useful for
further works in order to propose better models to rep-
resent turbulence in bubble columns, which can be imple-
mented in CFD codes.
Notation
g gravitational constant, m/s`
k' parameter used in Table 1
I
!
centreline turbulence intensity
l
K
mixing length, m
l
K
, cross-sectional averaged mixing length, m
m exponent de"ned in Eq. (1)
P static pressure, Pa
r radial coordinate
R column radius, m

}
`
ratio de"ned in Eq. (24)
;
%
super"cial gas velocity, m/s
v
!
centreline liquid velocity, m/s
v velocity, m/s
v velocity vector, m/s
v' #uctuating velocity component, m/s
z axial coordinate
1092 C. Vial et al. / Chemical Engineering Science 56 (2001) 1085}1093
Greek letters
c
%
local gas hold-up
c
%
, volume-averaged gas hold-up
z boundary layer thickness, m
j viscosity, Pa s
j
R
turbulent viscosity, Pa s
v molecular kinematic viscosity, m`/s
v
R
turbulent kinematic viscosity, m`/s
j density, kg/m`
o rms velocity, m/s
t shear stress, Pa
t
I
viscous stress tensor
t
R
turbulent stress tensor
t
U
wall shear stress, Pa
dimensionless radius
Subscripts and other symbols
ax axial
rad radial
G gas phase
liquid phase
References
Burns, L. F., & Rice, R. G. (1997). Circulation in bubble columns.
A.I.Ch.E. Journal, 43, 1390}1401.
Camarasa, E., Vial, Ch., Poncin, S., Wild, G., Midoux, N., & Bouillard,
J. (1999). In#uence of coalescence behaviour of the liquid and of gas
sparging on hydrodynamics and bubble characteristics in a bubble
column. Chemical Engineering Proceedings, 38, 329}344.
Chen, J., Kemoun, A., Al-Dahhan, M. H., DudukovicH , M. P., Lee, D. J.,
& Fan, L. -S. (1999). Comparative hydrodynamic study in a bubble
column using CARPT, computed tomography and PIV. Chemical
Engineering Science, 54, 2199}2207.
Chen, R. C., Reese, J., & Fan, L. S. (1994). Flow structure in a 3-D
bubble column and three-phase #uidized bed. A.I.Ch.E. Journal, 40,
1093}1104.
Delnoij, E., Kuipers, J. A. M., & van Swaiij, W. P. M. (1997). Computa-
tional #uid dynamics applied to gas}liquid contactors. Chemical
Engineering Science, 52, 3623}3638.
Fan, L. S., & Tsuchiya, T. (1990). Bubble wake dynamics in liquids and
liquid}solid suspensions. Boston, USA: Butterwoth Heinemann.
Franz, K., BoK rner, T., Kantorek, H. J., & Buchholz, R. (1984). Flow
structures in bubble columns. German Chemical Engineering, 7,
365}374.
Geary, N. W., & Rice, R. G. (1992). Circulation and scale-up in bubble
columns. A.I.Ch.E. Journal, 38, 76}82.
Joshi, J. B. (1980). Axial mixing in multiphase reactors: A uni"ed
correlation. Trans. Institution of Chemical Engineers, 58, 155}165.
Menzel, T., In der Weide, T., Staudacher, O., Wein, O., & Onken, U.
(1990). Reynolds shear stress for modeling of bubble
column reactors. Industrial Engineering Chemistry, Research, 29,
988}994.
Mudde, R. F., Groen, J. S., & van den Akker, H. E. A. (1997a). Liquid
velocity "eld in a bubble column: LDA experiments. Chemical
Engineering Science, 52, 4217}4224.
Mudde, R. F., Lee, D. J., Reese, J., & Fan, L. S. (1997b). Role of coherent
structures on Reynolds stresses in a 2-D bubble column. A.I.Ch.E.
Journal, 43, 913}926.
Prakash, A., & Briens, C. L. (1990). Porous gas distributors in bubble
columns. E!ect of liquid presence on distributor pressure drop.
E!ect of start-up procedure on Distributor Performance. Canadian
Journal of Chemical Engineering, 68, 204}210.
Rice, L. F., & Geary, N. W. (1990). Prediction of liquid circulation in
viscous bubble columns. A.I.Ch.E. Journal, 36, 1339}1347.
Riquarts, H. P. (1981). A physical model for axial mixing of the liquid
phase for heterogeneous #ow regime in bubble columns. German
Chemical Engineering, 4, 18}23.
Sokolichin, A., & Eigenberger, G. (1999). Applicability of the standard
k}c turbulence model to the dynamic simulation of bubble columns.
Chemical Engineering Science, 54, 2274}2284.
Ueyama, K., & Miyauchi, T. (1979). Properties of recirculating turbu-
lent two-phase #ow in gas bubble columns. A.I.Ch.E. Journal, 25,
258}266.
Vial, Ch., Camarasa, E., Poncin, S., Wild, G., Midoux, N., & Bouillard,
J. (2000). Study of hydrodynamic behaviour in bubble columns and
external loop airlift reactors through analysis of pressure #uctu-
ations. Chemical Engineering Science, 55, 2957}2973.
Zehner, P. (1986). Momentum, mass and heat transfer in bubble
columns. Flow model of the bubble column and liquid velocities.
Institution of Chemical Engineers, 26, 22}35.
C. Vial et al. / Chemical Engineering Science 56 (2001) 1085}1093 1093

You might also like