Professional Documents
Culture Documents
www.elsevier.com/locate/ces
Experimental and modelling study of gas dispersion in a double
turbine stirred tank
S. S. Alves
, C. I. Maia, J. M. T. Vasconcelos
Centro de Eng. Biol ogica e Qumica, Department of Chemical Engineering, Instituto Superior T ecnico, 1049-001-Lisboa, Portugal
Received 7 February 2001; received in revised form 16 July 2001; accepted 7 August 2001
Abstract
Gas dispersion in a double turbine stirred tank is experimentally characterised by measuring local gas holdups and local bubble size
distributions throughout the tank, for three liquid media: tap water, aqueous sulphate solution and aqueous sulphate solution with PEG.
For all these media, bubble coalescence generally prevails over breakage. Where average bubble size decreases, this can be attributed to
the dierence in slip velocity between dierent sized bubbles. Most of the coalescence takes place in the turbine discharge stream.
A compartment model that takes into account the combined eect of bubble coalescence and breakage is used to simulate gas dispersion.
The model predicts spatial distribution of gas holdup and of average bubble size, with average bubble size at the turbines as an input.
Reasonable agreement between experiment and simulation is achieved with optimisation of two parameters, one aecting mainly the slip
velocity, the other related mainly to the bubble coalescence}breakage balance. Dierent sets of parameters are required for each of the
three liquid systems under study, but are independent of stirring}aeration conditions. The model only fails to simulate the smaller average
bubble diameters at the bottom of the tank. ? 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Stirred tank; Gas dispersion; Bubble size
1. Introduction
Gas dispersion in stirred tanks is very important, since it
strongly inuences gasliquid mass transfer. It is an exceed-
ingly complex phenomenon, involving not only the com-
plexity of the continuous phase owitself and its interactions
with the dispersed phase, but also the behaviour and interac-
tions within the dispersed phase, including bubble breakage
and coalescence.
Modelling eorts have increased in recent years. Some
models predict the gas holdup distribution throughout the
tank, but make no attempt at predicting bubble sizes. These
include Eulerian}Eulerian CFD two-uid models which
assume a given constant bubble size to calculate the in-
teraction between phases (Gosman, Lekakou, Politis, Issa,
& Looney, 1992; Morud & Hjertager, 1996; Friberg, 1998;
Lane, Scwartz, & Evans, 2000), the Eulerian}Lagrangian
treatment of Patterson (1991) and the two-dimensional net-
work of zones proposed by Mann (1986) and Mann and
Hackett (1988).
(: t
G
) = Q
q
, (3)
where : is the local gas holdup and Q
q
is the local aera-
tion rate per unit volume, which, in the present case, is zero
throughout the vessel, except at the aeration point (Fig. 1).
The gas velocity, t
G
, is given by the sum of the liquid ve-
locity, t
L
, and the slip velocity, t
S
:
t
G
=t
L
+t
S
. (4)
The axial slip velocity is obtained through a force balance
on the bubbles, assumed spherical:
j
L
qJ
b
= C
D
1
2
j
L
t
2
s
4
d
2
b
, (5)
490 S. S. Alves et al. / Chemical Engineering Science 57 (2002) 487496
Table 2
Normalised values of the rms turbulent velocity in a standard unaerated tank stirred by a Rushton turbine
a
Reference u
}(ND)
Exit turbine Near the wall Average for the bulk of the tank
Cutter (1966) 0.33 (r = 0.27 T) 0.21 (r = 0.45 T)
Ranade and Joshi (1990) 0.23 (r = 0.45 T) 0.13
Kusters et al. (1990) 0.30 (r = 0.27 T) 0.18 (r = 0.37 T) 0.10
Dyster et al. (1993) 0.34 (r = 0.27 T) 0.25 (r = 0.45 T)
Lee and Yianneskis (1998) 0.34 (r = 0.27 T) 0.19 (r = 0.38 T)
Deglon et al. (1998) 0.34 (r = 0.20 T) 0.24 (r = 0.34 T) 0.14
Values used in this work 0.33 (r = 0.27 T) 0.22 (r = 0.45 T) 0.12
a
Range of geometries and conditions: 0, 1 61(m) 60, 4, H = 1, D = 1}3, 1}3 6C 61}2, 0, 8 6ND(m}s) 63, 8 (r= radial coordinate).
where d
b
is the bubble diameter, J
b
is the bubble volume,
j
L
is the liquid density and q is the acceleration of gravity.
Centripetal force on bubbles is neglected. The drag coe-
cient, C
D
, depends on bubble Reynolds number and can be
calculated using appropriate correlations (Morsi & Alexan-
der, 1972). Since these correlations are valid only in a stag-
nant uid, turbulence is accounted for using a modied bub-
ble Reynolds number, Re
b
, as suggested by Bakker and van
den Akker (1994):
Re
b
=
j
L
t
s
d
b
p
, (6)
where p
= p
L
+ C
p
E
= p
L
+ C
j
L
u
(7)
assuming that the eddy viscosity is proportional to turbulent
velocity, u
(n
b
t
G
) = n
b
+
Q
q
J
b0
, (9)
where Q
q
is the local gassing rate per unit volume and J
b0
is the average bubble volume at the gas source. The change
in bubble number density due to breakage}coalescence pro-
cesses, n
b
, is calculated in the discretized version of Eq. (9)
as a fraction, F, of the density of bubbles entering the vol-
ume element:
n
b
= F(n
b
t
G
A)
in
}J, (10)
where J is the compartment volume. F is given by
F = C
:d
1
b
u
(11)
an expression inspired in the gas kinetic theory. Allowance
for bubble breakage in the tank could be introduced by a
term in Eq. (11), as in Bakker and van den Akker (1994).
However, since coalescence dominates over breakage ev-
erywhere except in the turbine swept volume, this would un-
necessarily complicate the model, adding an extra parameter
with likely identiability problems. It is deemed preferable
to keep a single lumped parameter (C
,
is required for calculations of the slip velocity, t
S
, and the
bubble breakage}coalescence parameter, n
b
.
A survey of the literature for unaerated conditions reveals
some measure of general agreement in the normalised val-
ues of the rms turbulent velocity, u
}ND used in
the simulation, also shown in the table, were average values
taken from the authors quoted. A single average value was
used for the bulk of the tank, since u
q
u
P
q
P
1}3
(13)
where c
q
, c
u
and c
b
are the turbulent energy dissipation val-
ues, respectively, for aerated liquid, unaerated liquid and the
contribution due to bubbles, the latter assumed negligible.
u
q
and u
u
are the aerated and unaerated turbulent velocities.
4. Results and discussion
4.1. Experimental results
Fig. 3 shows both experimental and simulated average
bubble sizes at several locations in the tank, for the condi-
tions presented in Table 1. Diameters are represented as a
function of a spatial coordinate which follows the longest
liquid ow circulation path from each turbine back to the
same turbine, as illustrated in the gure. Diameters are vol-
ume averaged, d
43
, which was found to be the simplest def-
inition compatible with the model equations. As the surface
mean diameter is more useful for mass transfer calculations,
the relationship between d
43
and d
32
was examined and ap-
pears in Fig. 4, which shows a simple relationship between
the two diameters for all of the operating conditions studied.
Fig. 5 attempts a comparison of d
32
bubble size diam-
eters obtained by dierent authors using dierent meth-
ods, including the suction probe (Barigou, 1987) and
video}photographic methods (Martin, 1995; Machon, Pacek,
& Nienow, 1997; Bouai & Roustan, 1998). Since the op-
tical methods only measure bubble diameters near the tank
wall, the sampling location considered in the suction probe
experiments was also chosen to be near the wall. From
this gure we may draw the following conclusions: (i) all
experimental diameters lie within 35% maximum error
from the diameter predicted by the regression; (ii) scatter
within each method is of the same order of magnitude as
the overall scatter; (iii) results of both methods overlap,
although Barigou (1987) tended to obtain larger diameters.
Fig. 6 shows both experimental and simulated local gas
holdup at several locations in the tank, for the same sets of
conditions. Integration of the experimental local gas holdup
for the tank was consistent with the measured overall gas
holdup (Table 3).
From Fig. 3, it can be seen that the general trend is for
bubble size to increase along the circulation path, which
indicates that bubble coalescence prevails over breakage.
Coalescence is usually observed to be most intense in the
turbines discharge streams, which may be explained by a
larger collision frequency due to higher turbulence. Another
point of intense coalescence is in circulation loop 3, below
the upper turbine, near the wall, where upcoming bubbles
from the lower part of the tank meet the downcoming half of
the upper turbine discharge stream. This is common to both
so-called non-coalescing media (sulphate solution, sulphate
(f) (e)
(c) (a) (b)
(d)
0
1
2
3
4
0.0 0.1 0.2 0.3 0.4
x (m)
d
b
(
m
m
)
0
1
2
3
4
d
b
(
m
m
)
0
1
2
3
4
d
b
(
m
m
)
0
1
2
3
4
0.0 0.1 0.2 0.3 0.4
x (m)
0
1
2
3
4
0.0 0.1 0.2 0.3 0.4
x (m)
0
1
2
3
4
0
1
2
3
4
0
1
2
3
4
0
1
2
3
4
0.0 0.1 0.2 0.3 0.4
x (m)
d
b
(
m
m
)
0
1
2
3
4
d
b
(
m
m
)
0
1
2
3
4
d
b
(
m
m
)
0
1
2
3
4
d
b
(
m
m
)
Loop 4 Loop 4
Loop 3 Loop 3
Loop 2 Loop 2
Loop 1 Loop 1
Loop 4
Loop 3
Loop 2
Loop 1 Loop 1
Loop 4
Loop 3
Loop 2
0
1
2
3
4
0
1
2
3
4
0
1
2
3
4
Loop 1
Loop 4
Loop 3
Loop 2
0
1
2
3
4
d
b
(
m
m
)
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
0.0 0.1 0.2 0.3 0.4
x (m)
Loop 1
Loop 2
Loop 3
Loop 4
x
x
x
x
Fig. 3. Experimental () and simulated () average bubble diam-
eters as a function of position along liquid circulation loops: (a)
Identication of liquid circulation loops. (b) Aqueous Na
2
SO
4
0.3 M
solution, N = 300 rpm, Q
q
= 10 l}min. (c) Aqueous Na
2
SO
4
0.3 M
solution, N = 450 rpm, Q
q
= 10 l}min. (d) Aqueous Na
2
SO
4
0.3 M
solution, N = 450 rpm, Q
q
= 20 l}min. (e) Aqueous Na
2
SO
4
0.3 M
with 20 ppm de PEG, N = 450 rpm, Q
q
= 10 l}min. (f) Tap water,
N = 450 rpm, Q
q
= 10 l}min.
solution+PEG) and to the coalescing medium (tap water).
A decrease in bubble size along the liquid circulation paths
only occurs at the bottom of the tank and above the two
turbines, particularly at low stirrer speeds. When this occurs,
the gas holdup also decreases. The explanation is that, at
those locations only small bubbles, with small slip velocities,
tend to follow the downwards liquid ow, particularly at
low circulation velocities.
The eect of stirrer speed on gas dispersion can be as-
sessed by comparing Figs. 3(a) with (b) (bubble sizes)
and 6(a) with (b) (gas holdups). At higher stirring speed,
492 S. S. Alves et al. / Chemical Engineering Science 57 (2002) 487496
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
d43 (mm)
d
3
2
(
m
m
)
Fig. 4. Relationship between d
43
and d
32
for all operating conditions under study.
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
2 4 6 8 9
ln(Pg/V)
l
n
(
d
3
2
)
Barigou and
Greaves, 1992
Martin, 1995
Machon et al,
1997
Bouaifi and
Roustan, 1998
This work
Correlation
7 5 3
Fig. 5. Sauter mean bubble diameter near tank wall as a function of power input per unit tank volume, with water as the liquid medium.
smaller bubble sizes are found in the turbines discharge
stream, due to higher energy dissipation. In the bulk of the
tank, after the eect of coalescence, the dierence in bubble
diameter between the two conditions is not very signicant,
while at the bottom of the tank and immediately above the
turbines bubble sizes are actually larger for higher turbine
speeds, when larger liquid circulation velocities manage to
re-circulate larger bubbles.
The eect of aeration on gas dispersion can be assessed by
comparing Figs. 3(b) with (c) and 6(b) with (c). Larger aer-
ation leads to slightly larger bubbles, due to the part played
by a larger gas holdup on coalescence.
The dierence between a so-called non-coalescing
medium, sulphate solution, and a coalescing medium, tap
water, can be examined by comparing Figs. 3(a) with (e)
(bubble sizes) and 6(a) with (e) (gas holdups). As expected
bubble sizes are larger with tap water, since coalescence
is more ecient, and, as a consequence, gas holdups also
tend to be smaller.
The eect of adding a surfactant (PEG) to an already
non-coalescing solution (Fig. 3(d) vs. (b)) leads to further
repression of coalescence and to slightly lower bubble size
diameters.
4.2. Simulation vs. experiment
The proposed model calculates local average bubble sizes
and local gas holdups throughout the tank, with experi-
mental values of average bubble size produced by each of
the turbines as an input. Two parameters are adjusted: C
model parameter
C
model parameter
d
b
bubble diameter, m
d
32
surface based mean bubble diameter, d
32
=
n
i=1
d
3
b
i
}
n
i=1
d
2
b
i
, m
d
43
volume based mean bubble diameter, d
43
=
n
i=1
d
4
b
i
}
n
i=1
d
3
b
i
, m
D turbine diameter, m
F fraction of bubble number density entering the vol-
ume element
q acceleration of gravity, m s
2
H liquid height, m
n
b
bubble number density, m
3
n
b
change in bubble number density due to
breakage}coalescence, m
3
s
1
N turbine speed, s
1
or min
1
N
q
pumping number
P turbine power, W
P
q
aerated turbine power, W
Q
C
circulation rate, m
3
s
1
Q
q
local gassing rate per unit volume, s
1
Q
P
turbine pumping rate, m
3
s
1
r radial distance from centre of symmetry, m
Re
b
modied bubble Reynolds number
1 tank diameter, m
u
turbulent velocity, m s
1
u
q
turbulent velocity under aerated conditions, m s
1
u
u
turbulent velocity under unaerated conditions,
m s
1
J compartment volume, m
3
J
b
average bubble volume, m
3
J
b0
average bubble volume at the gas source, m
3
t
G
gas velocity, m s
1
t
L
liquid velocity, m s
1
t
S
gasliquid slip velocity, m s
1
Greek letters
: gas holdup
c
q
total turbulent energy dissipation rate under aera-
tion conditions, m
2
s
3
c
b
turbulent energy dissipation rate due to bubbles,
m
2
s
3
c
u
unaerated turbulent energy dissipation rate, m
2
s
3
p