Sunteți pe pagina 1din 13

Chemical Engineering Science 61 (2006) 28952907

www.elsevier.com/locate/ces
Mixing analysis in a coaxial mixer
Christian Rivera
a
, Stephane Foucault
a
, Mourad Heniche
a
,
Teodoro Espinosa-Solares
b
, Philippe A. Tanguy
a,
a
URPEI, Department of Chemical Engineering, cole Polytechnique, P.O. Box 6079, Station Centre-Ville, Montral, Que., Canada H3C 3A7
b
Departamento de Ingeniera Agroindustrial, Universidad Autnoma Chapingo, P.O. Box 161, Chapingo 56230, Edo. de Mxico, Mxico
Available online 19 January 2006
Abstract
The performance of a coaxial mixer in the laminar-transitional ow regime was numerically investigated with Newtonian and non-Newtonian
uids. These mixers comprised two shafts: a central fast speed shaft mounted with an open turbine, and a slow speed shaft tted with a
wall scraping anchor arm. To model the complex hydrodynamics inside the vessel, the virtual nite element method (POLY3D
TM
software)
coupled with a Lagrange multiplier approach to cope with the non-linearity coming from the rheological model was employed. Co-rotation
and counter-rotation mode were compared, based on several numerical criteria, namely, mixing time, power consumption and pumping rate. It
was found that co-rotating mode is more efcient than counter-rotating mode in terms of energy, pumping rate and homogenization time.
2005 Elsevier Ltd. All rights reserved.
Keywords: Mixing; Hydrodynamics; Simulation; Fluid mechanics; Coaxial mixer; Virtual nite element method
1. Introduction
The characteristics of a mixer are particularly critical for the
process economics and the quality of the end product. Usually,
the design is based on process objectives taking into account
many variables. For example, the high viscosity of phases usu-
ally restricts the mixing to the laminar-transitional regime due
the inefcient task to generate turbulent instabilities in such
conditions.
Nowadays, the industry needs impellers that can work in
laminar, transitional or turbulent regimes with minimum mod-
ications. Standard agitators like close clearance and open im-
pellers exhibit some limitations with this aspect. On one hand,
close clearance impellers such as helical ribbons have a good
mixing performance in laminar regime (Yap et al., 1979; De
la Villeon et al., 1998). However, this situation is completely
reversed when the condition changes from laminar to transi-
tional or turbulent (Hoogendoorn and Den-Hartog, 1967). On
the other hand, open impellers like the Rushton turbine are
known to be very efcient at high Reynolds number but in lam-
inar regime, segregated zones are produced (Salomon et al.,

Corresponding author. Tel.: +1 514 340 4017; fax: +1 514 340 4105.
E-mail address: philippe.tanguy@polymtl.ca (P.A. Tanguy).
0009-2509/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2005.11.045
1981). The situation becomes critical if along the process time
the phases to be mixed develop non-Newtonian rheological
properties such as shear-thinning or thixotropy.
Recently, several innovative strategies have been proposed
to tackle this problem, based for instance on coaxial mix-
ers (Tanguy et al., 1997; Espinosa-Solares et al., 1997, 2001;
Thibault and Tanguy, 2002; Foucault et al., 2004, 2005), plan-
etary mixers (Tanguy et al., 1999) or conical mixers (Dubois
et al., 1996). The main idea is simple, association of differ-
ent agitators rotating at different speeds. In this way it is pos-
sible to create a mixer that achieves the process objectives,
blending the capabilities of several agitators. At the end, a
dynamic mixing unit that adapts with the process necessities is
obtained.
Coaxial mixers have been shown as a good alternative to gen-
erate particles suspension and Newtonian and non-Newtonian
mixing (Foucault et al., 2004, 2005). A standard conguration
consists of the combination of a dispersive turbine and a wall
scrapping anchor. The anchor blades rotate at low speed in or-
der to scrap the vessel wall and bring back into the bulk the ma-
terial to be mixed. The dispersing turbine rotates at high speed
to produce a distributive and dispersive effect throughout the
bulk. The superposition of both effects is known to produce a
very efcient mixing.
2896 C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907
From this perspective, the association of a close clear-
ance impeller with an open agitator seems reasonable but the
question of optimal operation conditions arises. In general,
dimensionless analysis on pilot rigs has been utilized to design
mixing systems (Tatterson, 1992). However, for this kind of
mixers, the presence of different impellers rotating at differ-
ent speeds makes ambiguous the choice of the characteristic
length and velocity. Consequently, the denition of dimen-
sionless parameters like the Reynolds number becomes not
obvious. In addition, there are several factors to take into ac-
count, namely; number and type of impellers, speed ratio, uid
rheology, rotation mode and geometric position. This makes
the design of this equipment a challenging task. Ideally, a
compromise must be found among all these variables so that
the mechanical energy is enough to ensure a good distributive
and dispersive action yet not too large to avoid large power
consumption. Foucault et al. (2005) in their work clearly show
using dimensionless analysis that a slight modication in the
Reynolds number denition can be meaningful to correlate
experimental data as mixing time and power consumption
with operating conditions in a generalized fashion. In this
way, these authors were able to build power consumption
and homogeneity time master curves valid for both Newto-
nian and non-Newtonian uids. Co-rotating mode was found
to be better than counter-rotating mode from a mixing view-
point. However, the hydrodynamic causes were not completely
elucidated.
Torquemeter
Motor 2.23 kW
(without gearbox)
Tank
Torquemeter
Gearbox (15 : 1)
Motor 373 W
High speed shaft
= 0.0254 m
Dc = 0.38 m
Da = 0.36 m
Wa = 0.0318m
Cw = 0.0095m
Low speed shaft
= 0.0381 m
Temperature
sensors
Computer
Wa
Cw
Da
Dc
0
.
5
8

m
0
.
6
6

m
Fig. 1. Schematic coaxial mixer conguration.
Both numerical and experimental studies have been re-
ported on the hydrodynamics of complex mixing systems
involving multiple independent impellers (Tanguy et al., 1997;
Espinosa-Solares et al., 1997, 2001). Due to the complex ge-
ometry of these systems, it is usually very difcult to analyze
in depth their hydrodynamics from an experimental stand-
point. The objective of the present investigation is to carefully
assess the hydrodynamic conditions of coaxial mixing in co-
rotating or counter-rotating operation when Newtonian and
non-Newtonian uids are employed in the laminar-transitional
regime. The methodology is based on a numerical approach
experimentally validated following the philosophy of our re-
search group.
2. Mixing system and numerical methodology
The mixing system consisted of a centered Rushton turbine
with an anchor rotating at a speed ratio equal to 10. Fig. 1
illustrates the mixing system and presents tank and impellers
dimensions. In this work, we consider a Newtonian uid with
a viscosity of 10 Pa s and a non-Newtonian uid described by
the power law model with ow index value (n) of 0.5 and a
consistency index (m) of 8.3 Pa s
n
. The uid density for the
Newtonian uid was 1350 kg/m
3
and for the non-Newtonian
liquid 1010 kg/m
3
.
Co-rotating and counter-rotating modes were investi-
gated for two speed couples, 20020 and 10010 RPM, with
C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907 2897
Fig. 2. Computational domain: (a) three-dimensional unstructured mesh for the vessel, (b) virtual impellers and (c) three-dimensional mesh with the immersed
virtual impellers.
Newtonian and non-Newtonian uids. These operating condi-
tions avoid the formation of a vortex at the free surface. Further-
more, for comparison purposes, a conguration that consists
of a rotating Rushton turbine with a static anchor was studied.
The experimental power consumption and mixing time were
obtained following the work of Foucault et al. (2004, 2005).
To predict numerically the three-dimensional ow eld in a
stirred tank, the momentum and mass conservation equations
were solved with help of standard 3D Galerkin nite element
method.
j

jv
jt
+ v v

+ p + t = f , (1)
v = 0. (2)
The boundary conditions are as follows:
No normal velocity at the uid surface (v
z
= 0).
No slip condition at the vessel wall (v = 0).
Constant angular velocity at the impeller surface (N
i
=constant).
To deal with the non-linear rheological model an augmented
Lagrangian approach was utilized as proposed by Tanguy et al.
(1984). With this method, the rheological non-linearity is re-
moved from the momentum equation and handled separately
in an auxiliary equation with help of a Lagrange multiplier.
The Lagrange multiplier can be viewed as a kernel that con-
nects both equations. Finally, the variational formulation of
the momentum equation along with the incompressibility con-
straint and Lagrange multipliers are solved in a set of Uzawa
iterations (Bertrand and Tanguy, 2002) with the help of Bi-
CGSTAB method as a linear solver (Van der vorst, 1992). This
loop is immersed into a Newton scheme to tackle the convective
term.
Since both impellers rotate at different speeds, the use of a
Lagrangian frame of reference does not help the simplication
of the problem. Then, in this work, a Eulerian frame of reference
was used resorting to the ctitious domain method (Glowinski
et al., 1994) to reproduce the unsteady rotation of the impellers.
This method has been extended to mixing problems by Bertrand
et al. (1997) and Tanguy et al. (1997). Briey, this approach is
based on the imposition of the impeller kinematics by means
of a set of control points distributed along the surface of the
impeller (this is done using Lagrange multiplier and penalty
techniques). At each time step, the velocity and position of the
control points is updated and a new problem is solved. One of
the main advantages of this method is that only a single mesh
is needed avoiding the necessity of re-meshing at each time
step. To take into account the unsteady nature of the ow, a
Gear scheme was employed (Fortin et al., 1986). A total of 120
time steps per revolution for the anchor and 12 for the Rushton
turbine were used. Computations were carried out until periodic
solutions were obtained.
2898 C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907
In VFEM, the meshing is an important issue as shown by
Rivera et al. (2004). In this work an intelligent mesh was built
composed of different partitions near the anchor and turbine
blade to allow a good denition of the virtual objects. Fig. 2
illustrates the constructed mesh with the virtual impeller. Due
to the intrinsic complexity of the geometry, tetrahedral 8 nodes
elements P
+
1
.P
0
that approximate the velocity with a super-
linear polynomial and consider constant the pressure inside
each element was employed. This type of elements ensures
a rigorous stability and convergence for complex uid ow
problems (Bertrand et al., 1992). The nal computational mesh
required approximately 90,000 elements and 200,000 nodes
producing a system of approximately 1,250,000 equations.
All the described numerical features are available in the
commercial 3D nite element software POLY3D
TM
(Rheosoft,
Inc.). The total memory requirements for the Newtonian and
non-Newtonian solutions were 1.3 and 1.6 Gbyte, respectively.
All simulations were run on an IBM computer cluster. The
intelligent meshing was generated on I-DEAS (EDS) soft-
ware and the visual post-processing was carried out with En-
sight (CEI).
3. Results and discussion
3.1. Power consumption
Following the work of Foucault et al. (2005) the Reynolds
number of a coaxial mixing system can be dened by Eq. (3).
It is worth noting here that along all this work, N is dened
employing the same denition as Foucault et al. (2004, 2005);
co-rotating N = N
t
N
a
, counter-rotating N = N
t
+ N
a
and
single Rushton turbine N = N
t
.
Re
counter-rotation
=
j(N
t
+ N
a
)D
2
t
j
;
Re
co-rotation
=
j(N
t
N
a
)D
2
t
j
. (3)
For the case of non-Newtonian uids, the generalized Reynolds
number is dened as
Re
counter-rotation
=
j(N
t
+ N
a
)
2n
D
2
t
k
;
Re
co-rotation
=
j(N
t
N
a
)
2n
D
2
t
k
. (4)
Table 1
Power consumption for the investigated scenarios in the coaxial mixer
Operating conditions Re
Mod
N
p
Exp
N
p
Num
Power
Exp
(W) Power
Num
(W) P/V (W/m
3
)
Co-rotating Newtonian uid 16.20 7.00 6.94 81.64 81.00 2041.0
Rushton impeller only Newtonian uid 18.00 5.50 5.25 88.00 84.00 2200.0
Counter-rotating Newtonian uid 19.80 5.30 5.07 112.86 108.00 2821.5
Co-rotating non-Newtonian uid 8.94 3.25 3.43 3.54 3.75 88.5
Rushton impeller only non-Newtonian uid 10.47 3.25 3.02 4.86 4.52 121.5
Counter-rotating non-Newtonian uid 12.08 3.25 3.06 6.47 6.10 161.8
Furthermore, for a given velocity eld, the power consumption
can be computed by Eq. (5) (Tanguy et al., 1997).
P =

O
t : dO. (5)
For analysis purposes, the obtained power consumption can be
translated into the power number N
p
(Nagata, 1957) which can
be determined using Eq. (6).
N
p
=
P
jN
3
D
5
t
. (6)
In the same manner as the Reynolds number, the power number
can be redened as Eq. (4) (Foucault et al., 2004, 2005).
N
p
counter-rotation
=
P
j(N
t
+ N
a
)
3
D
5
t
;
N
p
co-rotation
=
P
j(N
t
N
a
)
3
D
5
t
. (7)
Table 1 summarizes the power number with their respective
Reynolds number obtained by experimental and numerical ap-
proaches. It can be readily seen that a quite good agreement is
obtained between the data. Furthermore, the observed higher
power consumption and specic energy for the counter-rotation
mode is in accordance with the work of Foucault et al. (2004,
2005).
3.2. Flow patterns
To ease the analysis of the three-dimensional hydrodynam-
ics, the velocity was projected onto two-dimensional planes.
Fig. 3 shows the ow patterns for the Newtonian case. We
restrict ourselves to show only the Newtonian patterns since
the non-Newtonian ones exhibit the same particularities. In the
plane XZ, the ow is mainly dominated by the well-known tori
structures typical of the Rushton turbine. However, when the
operating mode changes from co-rotating to counter-rotating
there is a shift in the position of the center of such structures.
In counter-rotating mode, the center is located closer to the im-
peller. In the XY-plane, a dominant angular motion is readily
observed. A secondary circulation region can be noted around
the anchor when the impellers are in counter-rotating mode.
This irregularity is reduced by the single turbine and completely
removed in co-rotating mode..
The global effect is a contraction in the size of the well
mixed region. This can be better addressed with the velocity
C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907 2899
Fig. 3. Flow patterns (velocity, m/s) for planes XZ and XY (at Z = 0.2 m): (a) co-rotating mode, (b) counter-rotating mode and (c) single Rushton turbine.
iso-contour presented in Fig. 4 For the case of co-rotating
mode the size of the well mixed region is larger. With the non-
Newtonian uid, the contraction in the cavern for the counter-
rotating case is more evident. This is produced by the fact that
in counter-rotating mode the shear is higher than for the other
two operation modes. This generates a low viscosity zone close
to the central turbine that diminishes the pumping in the axial
direction as we will show.
To quantitatively address the impact of the operation mode
over the hydrodynamics, we present in Fig. 5 the velocity pro-
les along the tank height for the six cases considered. These
plots clearly illustrate the fact that co-rotating mode generates
an overall increase in axial and tangential velocities close to
the impeller for both Newtonian and non-Newtonian cases. In
Fig. 5a, it is interesting to note a considerable axial velocity re-
duction in the upper part of the tank (z/H =0.7.1) for the case
of non-Newtonian turbine and counter rotating scenarios. Fur-
thermore, in Fig. 5b, we can observe that the counter-rotating
mode generates a tangential ow in both positive and negative
directions. As a consequence, the ow segregation is greater in
counter-rotating mode. Finally, we plot the radial velocity pro-
le in Fig. 5c. As must be expected, the co-rotating mode in
both Newtonian and non-Newtonian cases generates a higher
radial velocity close to the impeller. However, at the upper and
lower parts of the vessel (0.4 >z/H >0.7), the situation is re-
versed due to the modication in the tori structures as was al-
ready pointed out.
To better understand the ow inside the coaxial mixer, the
shear rate norm was plotted in the radial discharge zone of
the Rushton turbine in Fig. 6a for the Newtonian uid and
Fig. 6b for non-Newtonian uid. The most interesting feature
is the peak at r/R = 0.75 that can be found for counter-
rotating mode and single Rushton turbine. This fact illustrates
the high shear region generated between the anchor and the
turbine when they operate in counter-rotating mode. However,
for single turbine the shear rate decreases near the wall. We
can also observe that for the case of the non-Newtonian uid
the peak becomes magnied. This has a strong repercussion
over the viscosity, since it is shear rate dependent. We plot-
ted the viscosity prole for the non-Newtonian scenario in
2900 C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907
Fig. 4. Velocity iso-contours at 0.3 m/s for the Newtonian case and at 0.15 m/s for the non-Newtonian case: (a) co-rotating, (b) counter-rotating and (c) single
Rushton turbine.
Fig. 7. The viscosity differences generated by the Rushton
turbine case can be readily observed, where a small cavern
of low viscosity is surrounded by a high viscosity zone. In
the counter-rotating mode this situation is less prominent
but a region of high viscosity still persists at the top of the
vessel.
3.3. Pumping rate
The computation of the total horizontal contribution in the
radial (r) and tangential (0) pumping rate was done solving
numerically the stream function (Rivera et al., 2004; Heniche
et al., 2005). The pumping rate can be computed by
Q
r
(z) =
1
H

H
0
(
r
max

r
min
) dz,
Q
0
(z) =
1
H

H
0
(
0
max

0
min
) dz. (8)
In addition, the axial pumping rate was computed by means
of
Q
+
z
(z) =
1
H

H
0

A
Vz
+
dA =
1
H

H
0

A
Vz

dA, (9)
where the + stands for the positive axial velocities. Since the
obtained velocity eld ensures mass conservation, the corre-
sponding values of Q
+
z
is of the same magnitude as Q

z
(where
the stands for the negative axial velocities).
For analysis purposes, the dimensionless ow number is
helpful which was determined with Eq. (10).
N
q
i
=
Q
i
ND
3
, i = r, 0, z. (10)
Table 2 illustrates the comparison of axial volumetric ow
rate in the radial (r), tangential (0) and axial (z) components
for the different investigated scenarios. It is noted that axial and
tangential pumping throughout the vessel is improved in the
case of co-rotating mode. In the case of the Newtonian uid,
they are 71% and 26% higher than the counter-rotating mode,
respectively. For the non-Newtonian scenario, differences of
41% and 29% were obtained. In agreement with the velocity
proles, the radial pumping is higher for the counter-rotating
mode (100% Newtonian and 22% non-Newtonian).
3.4. Pressure patterns
Even though in the mixing analysis of mechanically agi-
tated systems the pressure parameter is usually neglected, we
C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907 2901
Fig. 5. Velocity proles along a line parallel to the Z-axis at x = 0.1 m and
y = 0 m: (a) axial velocity, (b) tangential velocity and (c) radial velocity.
Fig. 6. Shear rate norm in the radial discharge zone: (a) Newtonian uid and
(b) non-Newtonian uid.
cannot forget that an impeller is a kind of pump. Fluid mo-
tion is basically powered by pressure forces. Then, based on
this analogy we can better explain why the co-rotating mode
performs better than the counter-rotating mode. In Fig. 8, the
pressure patterns are presented. It is noticed in the XZ-plane
that for co-rotating mode, a larger low-pressure zone forms
around the centered turbine. In the XY-plane a low-pressure
zone forms behind the blades of the turbine and the anchor. If
the impellers are in co-rotating mode, the pressure gradient is
in the forward direction; thus in this case, the pressure forces
2902 C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907
Fig. 7. Non-Newtonian viscosity prole (Pa s): (a) co-rotating mode, (b)
counter-rotating mode and (c) single Rushton turbine.
are additive. The turbine drags the anchor in the ow direc-
tion and as a consequence a lower power is obtained. On the
other hand, in counter-rotating mode the pressure gradient of
Table 2
Flow number for the investigated scenarios in the coaxial mixer
Operating conditions N
q
r
N
q
0
N
q
z
Co-rotating Newtonian uid 0.119 0.901 0.125
Rushton impeller only Newtonian uid 0.136 0.635 0.111
Counter-rotating Newtonian uid 0.236 0.715 0.073
Co-rotating non-Newtonian uid 0.163 0.739 0.093
Rushton impeller only non-Newtonian uid 0.073 0.441 0.068
Counter-rotating non-Newtonian uid 0.201 0.570 0.065
both impellers are in opposite direction, then, the ow com-
ing from the turbine blade suddenly faces a higher-pressure
zone generated by the anchor. This produces a repulsive ef-
fect that increases the total power consumption. These nd-
ings are in agreement with the experimental work reported by
Foucault (2005). Furthermore, the ow is segregated in two re-
gions: an enclosed volume close to the turbine and a recircula-
tion zone at the wall clearance in agreement with the previous
discussion.
To quantify the above observations, we compute the differ-
ence of pressure in front of and behind the blades of the im-
pellers (anchor and turbine). Then, we were able to determine
a representative total driven force in the vessel by

p
Total
=
p
Turbine
+
p
Anchor
. (11)
Table 3 presents the computed pressure drops. It appears
that the co-rotating mode offers the higher values which are in
agreement with the latter discussion. Based on all the facts, we
nally can represent a coaxial mixer as two centrifugal pumps
working in series (Fig. 9). In co-rotating mode the discharge of
the rst pump is connected to the input of the second one, gen-
erating a balanced ow. For counter-rotating mode, the output
of the rst pump is connected with the output of the second.
3.5. Distributive and dispersive mixing
The fact that the counter-rotating mode offers more shear-
ing is commonly related with a better mixing performance.
However, we must take into account that mixing is a compro-
mise between distributive and dispersive actions. The former
is directly related to the pumping, while the latter is related to
the shearing effects. Thus, a parameter that quanties the bal-
ance between the dispersive and distributive mixing would be
helpful.
In mixing analysis, a dimensionless number called head num-
ber can be related to the shearing action. This is dened by
N
h
=
2gH

2
N
2
D
2
, (12)
where H is a theoretical height created by the impeller that
comes from the analogy between an agitator and a centrifugal
pump
H =
P
Qjg
=
N
p
N
q
=
ND
2
g
2
. (13)
C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907 2903
Fig. 8. XZ and XY (at Z=0.2 m) planes for the pressure (Pa) patterns in the Newtonian case: (a) co-rotating, (b) counter-rotating and (c) single Rushton turbine.
Table 3
Pressure drop for the investigated scenarios in the coaxial mixer
Operating conditions P
Anchor
(Pa) P
Turbine
(Pa) P
Total
(Pa)
Co-rotating Newtonian uid 350.96 2049.568 2400.52
Counter-rotating Newtonian uid 1293.16 1645.82 352.66
Rushton impeller only Newtonian uid 580.63 2064.51 1483.88
Co-rotating non-Newtonian uid 65.20 260.12 325.32
Counter-rotating non-Newtonian uid 230.14 213.67 16.47
Rushton impeller only non-Newtonian uid 67.64 238.86 171.22
2904 C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907
Rushton turbine
P
0
- P
0
-
P
0
+ P
0
+ P
1
+
P
1
+ P
1
-
P
1
-
Counter-rotating mode
Anchor
Rushton turbine
Anchor
Co-rotating mode
Fig. 9. Pumps analogy for the coaxial mixer.
Table 4
Head number and ow number for the investigated scenarios in the coaxial
mixer
Operating conditions N
q
N
h
N
h
/N
q
Co-rotating Newtonian uid 0.917 1.546 1.685
Rushton impeller only Newtonian uid 0.658 1.474 2.477
Counter-rotating Newtonian uid 0.756 1.631 1.948
Co-rotating non-Newtonian uid 0.761 0.865 1.136
Rushton impeller only non-Newtonian uid 0.452 1.082 3.222
Counter-rotating non-Newtonian uid 0.608 1.457 1.778
Furthermore, the total pumping rate can be computed as
Q =

Q
2
r
+ Q
2
0
+ Q
2
z
. (14)
A parameter to quantify the relationship between shearing
and pumping actions can be obtained with the ratio between
the head number and the ow number. A large value indicates
a good dispersive action, while a low one corresponds to a
Fig. 10. Tracer dispersion for the Newtonian uid: (a) co-rotating mode at 15 s, (b) co-rotating mode at 150 s, (c) counter-rotating mode at 15 s and (d)
counter-rotating mode at 150 s.
distributive dominant ow. Table 4 presents the results for the
different studied scenarios. As it can be seen, the ow produced
by the co-rotating mode is more balanced since we obtain a
distributive action with not excessive shearing. On the other
hand, as could be expected from the last discussions on shear
and pumping rate, counter-rotating mode offers good dispersive
mixing properties but poor distributive ow.
3.6. Mixing time
To determine the mixing time, the most important task was
the computation of tracers trajectories. This has been usually
based on the velocity integration over time with high order
schemes (Ottino, 1989; Souvaliotis et al., 1995). The challenge
is to nd an optimal time step to predict accurately the trajec-
tories. In this work, to overcome this difculty, an element by
element tracking of massless particles as proposed by Heniche
and Tanguy (2005) was employed. In this work, it was ob-
served that the obtained velocity eld is periodic over time.
C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907 2905
Fig. 11. Intensity segregation evolution along the time: (a) Newtonian uid and (b) non-Newtonian uid.
Table 5
Mixing time and dimensionless mixing energy for the studied scenarios
Operating conditions T
m
Exp
(s) T
m
Num
(s) NT
m
Exp
NT
m
Num
N
p
(NT
m
) (Dimensionless mixing energy)
Co-rotating Newtonian uid 125 105 375 319 2631
Rushton impeller only Newtonian uid 281 225 936 777 5152
Counter-rotating Newtonian uid 1589 5828 30,888
Co-rotating non-Newtonian uid 341 210 512 389 1664
Rushton impeller only non-Newtonian uid
Counter-rotating non-Newtonian uid
For a better numerical efciency, the time periodic velocity eld
was approximated by a Fast Fourier Transform (FFT) with 20
harmonics.
The intensity of segregation was selected as the homogeneity
criterion, which is dened as
I
seg
=
1
C(1 C)
1
V
Total
M

i=1
(C
i
C)
2
V
i
. (15)
The nal computation was based on the trajectory of 6000 trac-
ers injected from the top. Fig. 10 illustrates the mixing action
of both rotation modes. After 15 s an unmixed region appears
with counter-rotating mode. At the same time, the co-rotating
mode exhibits a larger mixing zone. In Fig. 11 the evolution
of intensity of segregation is illustrated. Table 5 summarizes
the numerical and experimental mixing times. It is readily seen
that the co-rotating mode gives the shorter mixing times, the
single Rushton turbine stays as an intermediate case, and the
counter-rotating mode exhibits the longer ones. The numerical
results follow the same trend as the experimental ones. Finally,
we conclude by computing the dimensionless mixing energy
as follows:
E
mix
= N
p
N T
m
. (16)
From all the above discussion, it is clear that co-rotating
mode offers low energy consumption, making the counter-
rotating mode the most inefcient in terms of energy consump-
tion.
4. Conclusions
The objective of this work was to analyze by means of CFD
the hydrodynamics of a coaxial mixer in the case of Newtonian
and non-Newtonian shear thinning uids. It was shown that co-
rotating mode is more efcient than counter-rotating mode in
terms of energy, pumping rate and homogenization time. The
fact that the anchor and the turbine rotate in the same direction
allows a collaborative action that have a strong inuence on
the mixing performance. The co-rotating mode yields a ow
where both distributive and dispersive mixing capabilities are
balanced. Reversely, the counter-rotating mode produces a less-
balanced system where high shearing and unmixed regions co-
exist. The situation becomes worse for non-Newtonian uids
where the viscosity is shear rate dependent, resulting in smaller
well-mixed zones. Finally, we expect that the generated infor-
mation would be helpful for further investigations of this inter-
esting and versatile device for difcult mixing applications.
2906 C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907
Notation
C concentration of tracers
C average concentration of tracers
D diameter, m
E
mix
mixing energy, dimensionless
f body force, N/m
3
g gravity, m/s
2
H tank height, m
I
seg
intensity of segregation, dimensionless
k consistency index, Pa s
n
M number of nite elements
N angular speed, rev/s
N shear thinning index, dimensionless
N
h
head number, dimensionless
N
p
power number, dimensionless
N
q
pumping number, dimensionless
p pressure, Pa
P power, W
Q pumping rate, m
3
/s
Re Reynolds number, dimensionless
T
m
mixing time, s
v velocity, m/s
V
Total
total volume
Greeks letters
deformation rate, s
1
j Newtonian viscosity, Pa s
j density, kg/m
3
t viscous stress, Pa
stream function
Mathematical symbols
j partial derivative
difference
gradient
divergence
Subindices
a anchor
r radial direction
t turbine Rushton
z axial direction
0 tangential direction
Acknowledgements
The nancial assistance of NSERC is gratefully acknowl-
edged.
References
Bertrand, F., Tanguy, P.A., 2002. Krylov-based Uzawa algorithms for the
solution of the Stokes equation using discontinuous-pressure tetrahedral
nite elements. Journal of Computational Physics 181, 617638.
Bertrand, F., Gadbois, M.R., Tanguy, P.A., 1992. Tetrahedral elements for
uid ow. International Journal for Numerical Methods in Engineering
33, 12511257.
Bertrand, F., Tanguy, P.A., Thibault, F., 1997. A three-dimensional ctitious
domain method for incompressible uid ow problems. International
Journal for Numerical Methods in Fluids 25, 719736.
De la Villeon, J., Bertrand, F., Tanguy, P.A., Labrie, R., Bousquet, J.,
Lebouvier, D., 1998. Numerical investigation of mixing efciency of helical
ribbons. A.I.Ch.E. Journal 44, 972977.
Dubois, C., Thibault, F., Tanguy, P.A., Ait-Kadi, A., 1996. Characterization of
viscous mixing in a twin intermeshing conical helical mixers. Institution
of Chemical Engineers Symposium Series 140, 249258.
Espinosa-Solares, T., Brito-De La Fuente, E., Tecante, A., Tanguy, P.A., 1997.
Power consumption of a dual turbine-helical ribbon impeller mixer in
ungassed conditions. Chemical Engineering Journal 67, 215219.
Espinosa-Solares, T., Brito-De La Fuente, E., Tecante, A., Tanguy, P.A., 2001.
Flow patterns in rheologically evolving model uids produced by hybrid
mixing systems. Chemical Engineering Technology 24, 913918.
Fortin, A., Fortin, M., Tanguy, P., 1986. Numerical simulation of viscous
ows in hydraulic turbomachinery by the nite element method. Computer
Methods in Applied Mechanics and Engineering 58, 337358.
Foucault, S., Ascanio, G., Tanguy, P.A., 2004. Coaxial mixer hydrodynamics
with Newtonian and non-Newtonian uids. Chemical Engineering
Technology 27 (3), 324329.
Foucault, S., Ascanio, G., Tanguy, P.A., 2005. Power characteristics in coaxial
mixing Newtonian and non-Newtonian uids. Industrial & Engineering
Chemistry Research 44, 50365043.
Glowinski, R., Pan, T.W., Periaux, J., 1994. A ctious domain method
for Dirichlet problem and applications. Computer Methods in Applied
Mechanics and Engineering 111, 283.
Heniche, M., Tanguy, P.A., 2005. A predictorcorrector shooting scheme for
tracer trajectory calculations. In: Proceedings of the Fourth International
Conference on Computational Fluid Dynamics in the Oil and Gas,
Metallurgical Process Industries, Trondheim, Norway, 68 June.
Heniche, M., Reeder, M.F., Tanguy, P.A., Fassano, J., 2005. Numerical
investigation of blade shape in static mixing. A.I.Ch.E. Journal 51, 4458.
Hoogendoorn, C., Den-Hartog, A., 1967. Model studies in the viscous ow
region. Chemical Engineering Science 22, 16891699.
Nagata, S., 1957. Mixing Principles and Applications. Wiley, USA.
Ottino, J.M., 1989. The Kinematics of Mixing: Stretching, Chaos and
Transport. Cambridge University Press, UK.
Rivera, C., Heniche, M., Ascanio, G., Tanguy, P.A., 2004. A virtual nite
element model for centered and eccentric mixer congurations. Computers
and Chemical Engineering 28, 24592468.
Salomon, J., Elson, T.P., Nienow, A.W., Pace, G.W., 1981. Cavern sizes in
agitated uids with yield stress. Chemical Engineering Communications
11, 143164.
Souvaliotis, A., Jana, S.C., Ottino, J.M., 1995. Potentialities and limitations
of mixing simulations. A.I.Ch.E. Journal 41, 16051621.
Tanguy, P.A., Fortn, M., Choplin, L., 1984. Finite element solution of dip
coating, II: Non-Newtonian uids. International Journal for Numerical
Methods in Fluids 4, 459475.
Tanguy, P.A., Thibault, F., Brito-De La Fuente, E., Espinosa-Solares, T.,
Tecante, A., 1997. Mixing performance induced by coaxial at blade-
helical ribbon impellers rotating at different speeds. Chemical Engineering
Science 52, 17331741.
Tanguy, P.A., Bertrand, F., Dubois, C., Ait-Kadi, A., 1999. Mixing
hydrodynamics in a planetary mixers. Transactions of the Institution of
Chemical Engineers 77, 318324.
Tatterson, G., 1992. Scale Up and Design of Industrial Mixing Processes.
McGraw-Hill, New York, USA.
C. Rivera et al. / Chemical Engineering Science 61 (2006) 28952907 2907
Thibault, F., Tanguy, P.A., 2002. Power draw analysis of coaxial mixer with
Newtonian and non-Newtonian uids in the laminar regime. Chemical
Engineering Science 57, 38613872.
Van der Vorst, H.A., 1992. BI-CGSTAB: a fast and smoothly converging
variant of BI-CG for the solution of nonsymmetric linear systems. SIAM
Journal on Scientic and Statistical Computing 13, 631644.
Yap, C.Y., Patterson, W.I., Carreau, P.J., 1979. Mixing with helical ribbon
agitators, Part III: Non-Newtonian uids. A.I.Ch.E. Journal 25, 516521.

S-ar putea să vă placă și