You are on page 1of 8

Chemical Engineering Science 54 (1999) 4913}4920

Experimental and numerical studies of the hydrodynamics


in a bubble column
S. Lain, D. BroK der, M. Sommerfeld*
Fachbereich Ingenieurwissenschaften, Martin-Luther-Universita( t Halle-Wittenberg, D-06099 Halle (Saale), Germany

Abstract

The hydrodynamics in a bubble column of 140 mm diameter and a height of 650 mm was analysed using a phase-Doppler
anemometer (PDA). In order to allow the application of PDA, the bubble column was aerated with relatively "ne bubbles with a size
spectrum between about 0.3 and 1.5 mm. The gas hold-up was varied in the range between 0.5 and 3%. The measurement of the liquid
velocities in the bubble swarm was done by adding #uorescing seed particles. Moreover, the Euler/Lagrange approach was extended
to allow time-dependent calculations of the #ow evolving in a bubble column based on the Reynolds-averaged Navier}Stokes
equations together with the k}e turbulence model which was extended to account for turbulence modi"cation by the bubbles. The
coupling between the phases was considered through momentum source terms and source terms in the k- and e-equation which also
include the e!ect of wake-generated turbulence. The bubble motion was calculated by solving an equation of motion taking into
account drag force, added mass, buoyancy and gravity, and the transverse lift force. In order to identify the relative importance of the
di!erent physical phenomena involved in the model, a detailed parametric study was performed and the numerical results were
compared with the measurements for validation.  1999 Elsevier Science Ltd. All rights reserved.

Keywords: Bubble column; Hydrodynamics; Turbulence; Phase-Doppler anemometer; Numerical calculations; Euler/Lagrange approach; Wake-
turbulence

1. Introduction predictive tools for design and optimisation of bubble


columns. In recent years, computational #uid dynamics
Bubbly #ows are found in a number of technical and (CFD) has become an attractive tool for supporting pro-
industrial processes, such as sewage water puri"cation, cess design and optimisation and hence commercial CFD
#otation devices, and bubble column reactors. The latter tools are increasingly used by industry. For the numer-
"nd applications as gas}liquid contactors in chemical ical computation of two-phase #ows two approaches are
and biochemical processes. The hydrodynamics in mainly applied, namely the Euler/Euler and the Eu-
bubble columns is determined by the bubble rise and ler/Lagrange approach. The "rst method considers both
hence bubble size distribution and gas hold-up. More- phases as interacting continua, while in the second
over, turbulence will be induced by the bubble rise due to method the discrete nature of the dispersed phase is
the evolution of large-scale #ow structures, shear produc- taken into account by tracking a large number of indi-
ed in the vicinity of the bubble, oscillations of the vidual bubbles through the #ow "eld. Recently, numer-
bubbles, and the wakes of the bubbles. Especially mass ical methods have been developed which consider the
transfer in bubble columns will be largely a!ected by this unsteady nature of the #ow in bubble columns based on
small-scale turbulence generated on the scale of the both the Euler/Euler (Sokolichin & Eigenberger, 1994)
bubble. and the Euler/Lagrange approach (Lapin & LuK bbert,
Although attempts have been undertaken for many 1994; Murai & Matsumoto, 1995). A comparison of the
years to theoretically describe the #ow structure in performance of both approaches for the prediction of
bubble columns, in order to predict industrial processes, bubbly #ows was recently conducted by Sokolichin,
there is a lack of detailed physical understanding and Eigenberger, Lapin and LuK bbert (1997). In all these cal-
culations however, turbulence of the continuous phase
was not considered, rather an e!ective viscosity was used
* Corresponding author. in order to match calculated results with measurements.

0009-2509/99/$ - see front matter  1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 9 9 ) 0 0 2 1 2 - 2
4914 S. Lain et al. / Chemical Engineering Science 54 (1999) 4913}4920

Moreover, in some cases the bubble motion was cal- cross-section of the aerator which is 100 mm in diameter.
culated in a rather crude fashion, by assuming a "xed slip The gas #ow rate is varied by increasing the supply
velocity and adding some "ctitious di!usion component pressure. As a result of the stronger bulging of the mem-
for simulating turbulent dispersion (Lapin & LuK bbert, brane at higher pressures also the bubble size is slightly
1994). increasing with gas #ow rate. Measurements were per-
The present study concerns the extension of an formed up to a gas hold-up of about 3% and the estab-
Euler/Lagrange approach to allow time-dependent lished bubble size spectrum was in the range between 0.3
calculations of bubbly #ows including the modelling of and 1.5 mm. In order to reduce bubble coalescence pro-
turbulence and the consideration of an equation of panol was added to the tap water at a volume concentra-
motion to calculate the bubble motion. The numerical tion of 0.004%. A summary of the experimental
calculations are validated based on detailed experiments. conditions is given in Table 1.

2. Experimental facility 3. Phase-doppler anemometer

A schematics of the entire experimental facility is A two-component "bre optics PDA (Dantec, Fiber-
shown in Fig. 1. The bubble column used in the experi- PDA) was used to measure bubble size and velocities,
mental investigations has a diameter of 140 mm and and also the continuous-phase velocities. In order to
a height of 650 mm (i.e. water level in the column). The reduce refraction e!ects of the laser beams at the curved
aerator is build by using a porous membrane with a pore wall of the bubble column, it is placed in a square vessel
size of 0.7 lm. In order to "x the membrane it is mounted which is also "lled with tab water. The transmitting and
between two perforated plates, which are screwed on top receiving optics modules were mounted on a computer-
of a small stagnation chamber. The aerator is connected controlled 3-d traversing system (Fig. 1). This allowed
via a #ow meter to a pressurised air supply system. Once fully automated measurements of cross-sectional pro"les
the aerator is pressurised the membrane bulges and small at di!erent heights above the aerator. The pro"les were
bubbles are produced at the holes of the perforated plate measured in the direction of the optical axis of the trans-
so that a homogeneous aeration is established over the mitting optics. As a result, the optical path length of the
laser beams in the water changes when scanning a pro"le.
This results in a shift of the measurement volume away
from the geometric beam crossing location. In order to
compensate for this e!ect, the receiving optics was auto-
matically moved in the horizontal direction in order to
ensure that the optical axis always intersects with the
centre of the measurement volume. The optical path
length of the scattered light remains constant since the
bubble column is placed in a square vessel. The receiving
optics is mounted at a scattering angle of 703, where
re#ection by the bubbles is the dominant scattering
mechanism and a linear phase-size relation is obtained
(BroK der & Sommerfeld, 1998). The selected optical con-
"guration of the PDA-system allowed for a sizing range
up to 2 mm (BroK der & Sommerfeld, 1998).
For allowing the measurement of the liquid-phase vel-
Fig. 1. Schematics of the test facility with bubble column and PDA- ocities, #uorescing tracer particles with a nominal size of
System. 8.3 lm are added. These particles are excited by the green

Table 1
Flow conditions and measured cross-sectional averages (z"480 mm) of bubble number mean diameter and bubble Reynolds number based on
average slip velocity

Air #ow rate (l/h) Super"cial gas velocity Gas hold-up (%) Bubble number mean diameter Bubble Reynolds number
(cm/s) (lm) (dimensionless)

80 0.14 0.68 750 143


160 0.29 1.24 850 187
320 0.58 2.92 900 180
S. Lain et al. / Chemical Engineering Science 54 (1999) 4913}4920 4915

laser light at a wave length of 514.5 nm. The radiated


light emission has a wave length of 585 nm. Hence, the
Doppler signals from bubbles and tracer particles can be
separated using a narrow band optical "lter around
a wave length of 575 nm. When this "lter is inserted into
the optical system the scattered light from the bubbles is
blocked and only Doppler signals from the #uorescing
tracer particles are received. Therefore, the measure-
ments of the liquid phase velocities and bubble phase
properties is done sequentially, once with the optical
"lter and once without.

4. Experimental results

The measurement of the bubble-phase properties and


the continuous-phase velocities was performed at four
cross-sections above the aerator (i.e. at 30, 100, 300, and
480 mm). Results of these measurements for di!erent gas
hold-up have been published previously (BroK der & Som-
merfeld, 1998) and will be presented here only in com-
parison with the numerical calculations. However, one
phenomenon which is quite important for the calcu-
lations will be discussed. The experimental data revealed
that the bubble rise and slip velocity which was obtained
from the di!erence of the mean velocity of the entire
bubble size spectrum and the mean velocity of the tracer
particles, are considerably larger than the terminal velo-
city of single bubbles in contaminated water (Clift, Grace
& Weber 1978). This e!ect is illustrated in Fig. 2 showing
the size distributions of the bubbles measured in the core Fig. 2. Bubble number size distribution (a) and bubble size}velocity
region of the bubble column 480 mm above the aerator correlation (b) in the core of the bubble column 480 mm above the
aerator for di!erent gas hold-ups.
and the size}velocity correlations for di!erent gas hold-
ups. It is obvious, that the bubble size distributions have
a peak around 800 lm and a long tail up to about turbulence induced by the bubbles whereby the drag
1800 lm. With increasing air #ow rate a second max- coe$cient is reduced. A drastic increase of the measured
imum appears in the size distribution around 1300 lm. rise velocity with gas hold-up is found for the smaller
The "rst reason for this e!ect is connected with the bubbles up to a diameter of 900 lm. This indicates that
increase of the aeration pressure in order to increase the the liquid #ow is driven by the large bubbles and hence
gas #ow rate. Secondly, there might be still some coales- the smaller bubbles are dragged by the larger ones due to
cence whereby also larger bubbles are produced. the hydrodynamic interaction. In order to further analyse
The measured correlations between bubble size and this e!ect detailed experiments by particle tracking
rise velocity are compared with the terminal velocity of velocimetry are presently conducted which allow to
single bubbles in Fig. 2(b). As expected, the measured rise study the bubble behaviour and liquid #ow structure in
velocity increases with size up to a value of 1000 lm and the bubble swarm.
approaches a limiting value of about 0.3 m/s, which is
close to the maximum rise velocity of single bubbles (Clift
et al., 1978). With further increasing size the rise velocity 5. Basic equations and numerical approach
again decreases slightly up to a value of 0.25 m/s which
corresponds to the behaviour of single bubbles. For the The time-dependent calculations of the #ow pattern
larger bubbles the rise velocity is independent of the evolving in the bubble column were performed using the
volume fraction. However, for bubble sizes up to about Euler/Lagrange approach. The #uid #ow was calculated
1200 lm the measured rise velocity in the bubble swarm based on the Euler approach by solving the Reynolds-
is considerably larger than the terminal velocity of single averaged conservation equations in a time-dependent
bubbles. It is supposed that this observation is the result way. The resulting conservation equations are closed
of hydrodynamic interactions between bubbles and the using the well-known k}e turbulence model (Launder
4916 S. Lain et al. / Chemical Engineering Science 54 (1999) 4913}4920

Table 2
Summary of terms in the general equation, e!ective viscosity, and constants of the turbulence model

S S C
( ( .
1 ! 0 !

 
* *; *p S k#k
; C H ! #og 3 . R
*x *x *x V
H

 
* *; *p < = S k#k
< C H ! !2C #o 4 . R
*x *r *r r r
H

<= =*(Cr) S k#k


= !o ! 5 . R
r r *r

k G !oe S k
I I . k# R
p
I
e e S k
(C G !C oe) C . k# R
k  I  p
C

               
*;  *<  <  *; *<  * =  *=  k
G "k 2 # # # # # r # , k "C o
I R *x *y r *r *x *r r *x R I e

C "0.09; C "1.44; C "1.92; p "1.0; p "1.3; C "1.1


I   I C C

& Spalding, 1974) which was extended by accounting for Crank}Nicolson method, while the di!usive and the con-
the e!ects of the dispersed phase. The dispersed phase (i.e. vective terms are discretised using central and upwind
bubbles) was simulated in a Lagrangian way by tracking di!erences, respectively.
a large number of bubbles simultaneously through the The simulation of the bubble phase by the Lagrangian
#ow "eld. The time-dependent conservation equations approach requires the solution of the equation of motion
for the #uid phase are written for a two-dimensional for the bubbles. The appropriate gas #ow rate is realised
axisymmetric #ow in the general form as by using the concept of computational bubbles where
each tracked bubble represents a number of real bubbles

 
*(o
) * 1 * * *

# (o;
)# (or<
)" C with the same properties (i.e. size and velocity). This
*t *x r *r *x *x approach allows to simulate the appropriate gas-phase

 
1 * *
volume fraction by a reasonable number of computa-
# rC #S #S , (1) tional bubbles which are simultaneously tracked through
r *r *r ( (.
the #ow "eld. The number of real bubbles represented by
where o is the liquid density, ; and < are the Reynolds- a computational bubble is determined by the mass #ow
averaged velocity components, C is the e!ective viscosity, rate of air, the bubble size sampled from the input size
and S and S are the additional source terms of the distribution, the number of the desired computational
( ( .
continuous and dispersed phase which are summarised in bubbles, and a reference time. Depending on the bubble
Table 2 for the di!erent variables
. Since at present only size typically between 4000 and 8000 bubbles are simulta-
relatively low void fractions are considered, the liquid neously present in the #ow "eld. The bubble rise after
density is assumed to be una!ected by the presence of the injection at the inlet is calculated by solving the following
bubbles. The resulting set of equations is solved by using set of ordinary di!erential equations:
a "nite-volume discretisation scheme and applying an
iterative solution procedure based on the SIMPLE algo- dx
G"u , (2)
rithm. The time derivatives are discretised using the dt G

 
du 3 o o *u du
m G" * m c (u !u ) " u!u "#0.5m * G! G
dt 4o D " G G o *t dt

 
o o
#0.5m *[(u!u ) ;(
;u)] #m g 1! * . (3)
o G G o
S. Lain et al. / Chemical Engineering Science 54 (1999) 4913}4920 4917

Here x are the coordinates of the bubble position,


G
u are the velocity components, D is the bubble dia-
G
meter, and o is the gas density which is assumed to be
constant at present. The di!erent forces in the equation
of motion are: the drag force, the added mass
force, the transverse lift force, and the force due to
buoyancy and gravity. Other forces such as pressure
force, and Basset history term are neglected at the present
stage. The drag coe$cient c is calculated by using em-
"
pirical correlations for either a #uid sphere or a rigid
bubbles:
Rigid bubble:

24.0
c " (1.0#0.15Re ), Re (500,
" Re
c "9.5 ) 10\Re , 500(Re (1500, (4)
"
c "2.61, Re *1500. Fig. 3. Timing between Eulerian and Lagrangian calculation (a) and
" averaging of the source terms along the bubble trajectory for the period
of the Eulerian time step (b).
Fluid sphere:

16
c " , Re (1.5,
" Re opted. After a converged solution for the continuous
c "14.9Re\ , 1.5(Re (80, (5) phase is obtained at a given time level, all bubbles are
" tracked through the #ow "eld for a time interval corre-
48
c " (1.0!2.21Re\ ), 80(Re (700 sponding to the Eulerian time step (i.e. the time step to
" Re
solve the #uid #ow equations). The Eulerian time step
*t is selected to be considerably larger than the Lagran-
with #
gian time step *t for calculating the bubble trajectories
*
(Fig. 3(a)). For the present computations the Eulerian
o D "u!u " time step is selected to be *t "0.1 s. During the Lagran-
Re " * . (6) #
k gian calculation the bubble-phase properties and the
source terms are evaluated for each control volume by
The instantaneous #uid velocity components at the averaging over the bubble residence time in the control
bubble location occurring in Eq. (3) are determined from volume and over all bubbles passing the control volume
the local mean #uid velocity interpolated from the neigh- (Fig. 3(b)). With the new source terms a converged solu-
bouring grid points and a #uctuating component gener- tion for the #uid #ow is then calculated at a new time
ated by the Langevin model described by Sommerfeld, level. Thereafter, the Lagrangian calculation is proceeded
Kohnen and RuK ger (1993). In this model the #uctuation and so forth.
velocity is composed of a correlated part from the pre- The Lagrangian time step is automatically adjusted
vious time step and a random component sampled from during the trajectory calculation and must be smaller
a Gaussian distribution function. The correlated part is than 10% of the following relevant time scales:
calculated using appropriate time and length scales of
turbulence obtained from the k}e turbulence model and E the time required for a bubble to cross a control
the random part is re-generated every Lagrangian time volume,
step. E the bubble response time scale:
Since the liquid #ow in the bubble column is driven by
4 o D
the bubble rise, appropriate source terms of the bubble q (c )" (7)
" 3 Re c k
phase need to be considered in the conservation equa- "
tions of the liquid phase (see Table 2). As both phases are
computed time-dependent, the evaluation of the source E and the local time scale of turbulence (Sommerfeld et
terms and the coupling between the phases requires some al., 1993):
special treatment in order to yield reasonable averages of
the source terms for each control volume where bubbles k
¹ "0.24 . (8)
are present. Therefore, the following approach is ad- # e
4918 S. Lain et al. / Chemical Engineering Science 54 (1999) 4913}4920

The averaging procedure for the momentum source In the present study a value of 1.1 has been chosen for
terms is expressed by the following equation: C .
C
1
S G ."! m N
3 < *t I I
AT # I 6. Numerical results

   
o
; (uL>!uL )!g 1! * *t , (9) The numerical calculations were performed according
I G I G G o *
L to the con"guration and the #ow conditions of the ex-
where the sum over n indicates averaging along the periments described above. As a "rst approach the #ow
bubble trajectory and the sum over k is related to the was assumed to be axisymmetric and hence only half of
number of computational bubbles passing the considered the #ow domain needed to be considered. In the calcu-
control volume with the volume < (Fig. 3(b)). In this lations the bubbles were injected 30 mm above the aer-
AT
context it should be noted, that m is the mass of an ator which corresponds to the "rst cross-section where
I
individual bubble and N is the number of real bubbles measurements are available. The calculations presented
I
being represented by a computational bubble. The modi- here were performed for an air #ow rate of 80 l/h. In
"cation of the turbulence by the bubbles is accounted for order to identify the relative importance of di!erent
by appropriate source terms in the k- and e-equation. physical phenomena included in the model equations,
Two approaches are considered for the k-equation: a base case is established. Starting from this, several
1. The source term resulting from the Reynolds-aver- e!ects are investigated, such as drag correlation (rigid
aging procedure which is related to the momentum trans- versus #uid sphere), added mass, transverse lift force, size
fer between the phases is given by distribution of the bubbles or formulation for the bubble
phase source terms in the equations of k and e. All the

S " (u S G !; S G ). (10) computational test cases are summarised in the Table 3.
I . G 3N G 3N
G It should be pointed out that the experimental data are
2. The source term proposed by Crowe (1998) is given averages over a rather long period, but the numerical
by calculations are performed time-dependent in order to
resolve the unsteady nature of the #ow. Therefore, it is
 a o
S " +"; !u "#(u u !u u ),. (11) necessary to ensure that a quasi-stationary regime is
I . q (c ) G G G G G G reached in the calculations before the bubble-phase prop-
G "
The "rst contribution which is proportional to the square erties are averaged. For the considered #ow conditions,
of the relative velocity between the bubbles and the the time-dependent calculations revealed that the #ow
liquid, takes into account the wake-induced turbulence "eld does not change very much when a quasi-steady
whereas the second term represents a redistribution of regime is reached.
the turbulent kinetic energy between the phases. In this In the following, comparisons between experimental
formula a is the volume fraction of gas and q (c ) is the data and calculations are presented for the mean axial,
" #uctuating axial and #uctuating tangential velocities of
bubble response time scale (see Eq. (7)) including the drag
coe$cient. the bubble phase at a cross-section 480 mm above the
The modelling of the source term in the e-equation is aerator. The comparison of the results shown in Figs.
performed in the standard manner (Desjonqueres, 1987), 4 and 5, reveals the problem in prescribing the appropri-
which assumes that it is proportional to S and the ate drag coe$cient. Although the experimental condi-
I . tions are such that due to contamination and the
Lagrangian turbulent characteristic time scale C k/e.
* addition of propanol the bubbles are expected to behave
Grouping the coe$cient C with the proportionality
* like rigid spheres, the predicted bubble velocity does not
constant, one obtains
agree with the measurements when using the rigid sphere
k drag coe$cient. The bubble rise velocity is considerably
S "C S . (12)
C . C e IN under-predicted. As discussed in the experimental section,

Table 3
Computational test cases considered for 80 l/h air #ow rate (gas hold up 1.24%)

Case Drag law D (mm) Size distribution S Added mass


IN
and lift force

C1 #uid 0.75 No Eq. (11) Yes


C2 rigid 0.75 No Eq. (11) Yes
C3 #uid 0.75 Yes Eq. (11) Yes
C4 #uid 0.75 No Eq. (10) Yes
S. Lain et al. / Chemical Engineering Science 54 (1999) 4913}4920 4919

Fig. 6. Bubble-rise velocity and #uctuating velocity components of the


Fig. 4. Bubble-rise velocity and #uctuating velocity components of the bubbles, comparison of measurement and calculation, Case: C3.
bubbles, comparison of measurement and calculation, Case: C1.

Fig. 5. Bubble-rise velocity and #uctuating velocity components of the Fig. 7. Bubble-rise velocity and #uctuating velocity components of the
bubbles, comparison of measurement and calculation, Case: C2. bubbles, comparison of measurement and calculation, Case: C4.

this e!ect is the result of a reduced drag in the bubble term without wake e!ects (Eq. (10)) the #uctuating com-
swarm. However, appropriate correlations for the drag ponents are considerably under-predicted as shown in
coe$cient which account for this swarm e!ect are not yet Fig. 7. Moreover, the average rise velocity of the bubbles
available. By accident the drag coe$cient of a #uid is lower than predicted with the source term including
bubble reproduces the measured bubble rise velocity for wake-generated turbulence (Eq. (11)).
the considered bubble size and void fraction reasonably
well (Fig. 4). The #uctuating velocity of the bubbles in the
vertical and tangential direction is predicted to be identi- 7. Conclusions
cal and only the tangential component agrees with the
measurements. The vertical component is considerably Measurements of liquid and bubble velocities in
under-predicted for the cases C1 and C2 (Figs. 4 and 5). a bubble column revealed that the bubble-rise velocity is
The consideration of the bubble size distribution in the considerably larger than the terminal velocity of single
calculations is of great importance as demonstrated in bubbles for a gas hold-up in the range of 0.5}3%. This is
Fig. 6. As the di!erent-sized bubbles have di!erent rise supposed to be a result of the hydrodynamic interaction
velocities the #uctuating component in the vertical direc- between bubbles. In order to establishing appropriate
tion increases considerably compared to a calculation correlations for the modi"ed drag coe$cient as a func-
with mono-sized bubbles (compare Figs. 4 and 6) and tion of gas hold-up and bubble Reynolds number further
a very good agreement with the measurements is comprehensive experimental studies are needed. Such
achieved for both #uctuating components. correlations are indispensable for a reliable numerical
Finally the e!ect of the source term in the k-equation prediction of bubbly #ows as demonstrated in the paper.
was analysed (case C4). When using the standard source Moreover, the numerical calculations revealed that the
4920 S. Lain et al. / Chemical Engineering Science 54 (1999) 4913}4920

bubble-size distribution needs to be considered for pre- Clift, R., Grace, J. R., & Weber, M. E. (1978). Bubbles, drops, and
dicting the non-isotropy of the bubbles #uctuating particles. New York: Academic Press.
motion. For the prediction of the bubble-induced turbu- Crowe, C. T. (1998). On models for turbulence in #uid-particle Flows.
Proceedings. FEDSM'98, ASME #uid engineering division summer
lence appropriate source terms which also account for meeting, Paper 5034.
the wake e!ects were incorporated in the turbulence Desjonqueres, P. (1987). Modelisation Lagrangienne du comportement de
model. The resulting predicted bubble #uctuation vel- particules en ecoulmente turbulent. Ph.D. Thesis de 3eme Cycle,
ocities agreed very well with the measurements. University de Rouen.
Lapin, A., & LuK bbert, A. (1994). Numerical simulation of the dynamics
of two-phase gas-liquid #ows in bubble columns. Chemical Engin-
eering Science, 49, 3661}3674.
Acknowledgements Launder, B. E., & Spalding, D. B. (1974). The numerical computation of
turbulent #ows. Computation Methods in Applied Mechanics and
The "nancial support of the present studies by the Engineering, 3, 269}289.
Deutsche Forschungsgemeinschaft under contract num- Murai, Y., & Matsumoto, Y. (1995). Three dimensional structure of
a bubble plume-Measurement of the three dimensional velocity.
ber So 204/13 is gratefully acknowledged. The "rst Flow Visualization and Image Processing of Multiphase Systems,
author acknowledges the support by the Commission of ASME, FED-209, 187}194.
the European Communities in the frame of a Marie Curie Sokolichin, A., & Eigenberger, G. (1994). Gas-liquid #ow in
Research Training Fellowship (Contract No. CT-97- bubble columns and loop reactors: Part 1. Detailed modelling and
2206). numerical simulation. Chemical Engineering Science, 49,
5735}5746.
Sokolichin, A., Eigenberger, G., Lapin, A., & LuK bbert, A. (1997). Dy-
namic numerical simulation of gas}liquid two-phase #ows: Euler-
References Euler versus Euler-Lagrange. Chemical Engineering Science, 52,
611}626.
BroK der, D., & Sommerfeld, M. (1998). Simultaneous measurements of Sommerfeld, M., Kohnen, G., & RuK ger, M. (1993). Some open questions
continuous and dispersed phase in bubble columns by PDA. Pro- and inconsistencies of Lagrangian particle dispersion models. Pro-
ceedings of the ninth international symposium on applications of laser ceedings ninth symposium on turbulent shear yows, Kyoto, Aug. 1993,
techniques to yuid mechanics, Lisbon, Portugal, July 1998, vol. 2, Paper 15.1.
Paper 27.2.

You might also like