The Flow Pattern in an Industrial Paper Pulp

Chest with a Side Entering Impeller

Andr Bakker
Julian B. Fasano

A mathematical model for the combined laminar and turbulent flow of paper pulp has been
developed. The results of the model predictions agree well with experimental data.
The model is used to analyze the flow pattern in an industrial paper pulp chest equipped with
a side entering impeller. Turbulent, laminar and stagnant regions can easily be located. The model
is an excellent tool for the optimization of agitators for large industrial paper pulp chests.

Keywords: Mixing, Paper Pulp Chest, Computational Modeling, Stirred Tank, Laminar and Turbulent Flow.
Published in The Online CFM Book at http://www.bakker.org/cfm.
(c) 1998 Andr Bakker
Updated: February 15, 2000

THE ONLINE CFM BOOK

INTRODUCTION
Recently significant progress has been made in the
application of Computational Fluid Dynamics (CFD)
to the analysis of fluid flow in stirred chests.
However, most work concentrated on the
computation of flow patterns in lab-scale tanks
equipped with top entering impellers.
Industrial problems are usually much harder
to tackle. Complicating factors are the use of fluids
with complex non-Newtonian behavior, the use of
side entering agitators instead of top entering
agitators and the fact that the agitator may be Figure 1 A typical shear stress vs. shear rate
operated in the transitional regime.
diagram for paper pulp.
A difficult mixing problem is found in the
paper industries. The paper pulp, which is a suspension of thin, flexible fibers, exhibits a very
complex rheology. As a result of this rheology multiple flow regimes are found in paper pulp storage
tanks. Part of the chest will be laminar; some parts of the chest are turbulent.
Pulp storage chests can be rectangular or cylindrical. The bottom of the chest is usually
filleted, sloped and/or curved. Although paper pulp chests are sometimes equipped with top entering
agitators, the paper industry prefers using side entering agitators.
The curved bottom shape and the side entering agitator require the use of body fitted grids
for CFD modeling of this type of stirred tanks. The fact that multiple flow regimes occur makes
modeling difficult since there are no well tested models for this type of flow available.
This paper deals with the development of a mathematical model for the flow of paper pulp
in stirred tanks. First the model will be described. The model predictions will then be compared with
experimental results for a lab-scale tank. Finally, the model will be used to predict the flow pattern
in an industrial chest.

RHEOLOGICAL PROPERTIES OF FIBER SUSPENSIONS

The rheological properties of fiber suspensions are discussed by Gullichsen (1985). Figure 1 shows
a typical shear stress vs shear rate diagram. This diagram shows that the fiber suspension initially
behaves as a non-Newtonian fluid with a shear stress y. When the shear stress exceeds a certain
value d the fiber network structure is disrupted and the suspension behaves nearly as a turbulent
Newtonian fluid.
As a result of this rheological behavior fiber suspensions are extremely difficult to agitate.
To provide motion through the whole tank the shear stress has to exceed the yield stress everywhere
in the fluid. Since gradients in shear stresses can be expected, there will be regions in the fluid where
the fiber network structure is disrupted and the flow will be turbulent. At the same time the flow may
be laminar or even be stagnant in other parts of the chest.
This combination of turbulent flow and laminar flow of a non-Newtonian fluid makes this

PAPER PULP

a difficult to model problem with CFD. As far as the present authors know, no other work has been
published about the direct computation of this type of flow in stirred tanks.

DESCRIPTION PAPER PULP MODEL

The computational model is based on solving the conservation equations which describe the flow
in the tank. For the mean flow of an incompressible liquid, the equation for the conservation of mass
reads:

Ui denotes the mean velocity in direction xi. Under steady-state laminar flow conditions the
momentum balance is given by:

The stress tensor ij is given by:

When the flow is turbulent an additional term appears in the momentum balance for the mean flow:

ui is the fluctuating, turbulent, velocity component in direction xi. The cross correlation term
denotes the Reynolds stress tensor. Although analytical equations for the Reynolds stress tensor can
be derived, these equations again contain higher order cross correlation terms, and the set of
equations is not closed. For modeling the Reynolds stresses various so-called turbulence models are
available. An often used model is the k- model which models the Reynolds stresses as being
proportional to gradients in the mean flow.

THE ONLINE CFM BOOK

The turbulent viscosity t is given by:

The model equations for the conservation of the turbulent kinetic energy density k and the turbulent
energy dissipation rate density are:

Equation (5) for the Reynolds stresses is similar to equation (3) for the laminar shear stresses. The
turbulent viscosity t plays the same role as the molecular viscosity . As a result the form of the
turbulent momentum equations is similar to the form of the laminar momentum equation except that
is replaced by an effective viscosity eff:

For modeling the fiber suspension, which is partly laminar and partly turbulent, the following
method was used. For every computational cell the computations are first performed as if the flow
is turbulent. After that it is checked if the total shear stress is indeed larger than d

The shear rate S is calculated as the square root of the scalar invariant of the shear rate tensor:

PAPER PULP

If equation (10) is not satisfied

then the calculations for that particular cell are repeated as if the flow is laminar. The local apparent
viscosity is then calculated from the experimental shear rate - shear stress curves and the local shear
rate S.
The CFD solver used in this work is Fluent V4. Fluent contains the k- model for turbulent
flow, but does not contain any models for combined laminar and turbulent flow. The additional
equations were coded in Fortran subroutines and linked to the Fluent solver. The solver was then
configured such that these additional subroutines were called every iteration.

EXPERIMENTAL VERIFICATION
Usually the results of CFD models are verified by performing flow visualizations, e.g. with the aid
of Laser Doppler velocity measurements or other techniques, and comparing the results with the
model predictions. Here this is not possible since it is not possible to see through the fiber
suspension.
To allow for easier flow visualization a model fluid was designed. The model fluid consists
of a suspension of glass fibers in a mixture of Tetrachloroethylene (TCE) and Carbondisulfide (CS2).
The concentrations of TCE and CS2 were chosen such that the refractive index of the mixture
matched the refractive index of the suspended glass fibers. The resulting mixture was clear enough
to be able to see through and to spot the motion of tracer particles, despite a 10% volume holdup of
glass fibers.
The experimental set up consisted of a cylindrical glass vessel with a diameter of 0.3 m

Figure 2 Model predictions.

black: stagnant; dark grey:
laminar; light grey: turbulent.

Figure 3 Experimental visualization.

THE ONLINE CFM BOOK

equipped with a 0.12 m Chemineer HE-3 impeller, operating at a speed of N = 10.8 1/s (= 650
RPM). The experiments were performed with a top entering impeller. The fact that the model will
later be applied to tanks with side entering agitators is not expected to be a problem.
White tracer particles were added to the tank. A video film of the experiment was made and
pictures were taken. The results of the flow visualization experiments were then compared with the
results of the model calculations.
Figure 2 shows the model predictions for this tank. The model predicts three zones in the
tank: turbulent flow, laminar flow and stagnant. The region around the impeller is characterized by
high velocities and is turbulent. This is the only region where the shear stresses are large enough to
disrupt the fiber network structure. Around this turbulent zone there is a zone where the fiber
suspension is moving but where the stresses are not large enough to disrupt the fiber network and
where the flow will be creeping laminar. Even further away from the impeller, in the corners at the
bottom and in the top of the tank the suspension is not moving at all.
These predictions compare well with the experimental results. The experimental results are
summarized in Figure 3. In the experiments it was not possible to distinguish between laminar and
turbulent flow. The experiments show larger stagnant zones than the model predictions do. However,
in most regions where the flow should be stagnant the model predicts very low velocities. The
overall comparison between the model predictions and the experiments is satisfactory.

INDUSTRIAL CHESTS
A typical industrial paper pulp chest is shown in Figure 4. The sides of the chest are 10 m each and
the liquid level is 7 m. The bottom of the chest is curved. The side entering impeller is mounted near
the draining point. This is the generally accepted procedure in the paper industries.
The agitator studied here was a 1.22 m (= 48") diameter Chemineer HE-3S impeller running
at 4.67 1/s (= 280 RPM). The computer model was used to study the flow pattern generated by this
agitation system for various pulp consistencies. The consistency is defined as the mass fraction of
bone dry paper fibers. The shear stress vs. shear rate curves used here were the curves published by
Gullichsen for Softwood Kraft pulp. Other types of paper pulp have will have different rheological
behavior depending on, amongst other things, fiber length and diameter.
The body fitted computational grid (see Figure 4) was generated with Fluent PreBFC V4 and
then exported to Fluent V4 for the computations and the post processing. The tank was divided in
30x30x21 = 18,900 internal cells.

RESULTS
Figure 5 shows the predicted velocity vectors in a plane through the impeller for turbulent flow of
water. The jet coming from the impeller impinges on the opposite chest wall and a long circulation
loop is formed.

PAPER PULP

Figure 4 Outline and grid.

Figure 5 Velocity vectors for water.

Figure 6 Velocity vectors, 1% pulp.

Figure 7 Velocity vectors, 3.5% pulp.

Figure 8 Turbulent volume, 5% pulp.

Figure 9 Volume with velocities larger than

1E-3 m/s for 5% pulp.

THE ONLINE CFM BOOK

Volume Fraction

Consistency %
Figures 6 and 7 show the velocity vectors in the same
1
plane for pulp consistencies from 1% and 3.5%. The
1
size of the circulation loop decreases with increasing 0.8
3.5
pulp consistency.
0.6
The length of the arrows is proportional to the
5
local velocity. Short arrows denote low velocities; 0.4
7
long arrows denote large velocities. No arrows at all
0.2
means that the velocities are too small to be
displayed, although they are not necessarily zero.
0
0.0001
0.001
0.01
0.1
1
Figure 8 shows a wireframe surrounding the
Minimum Velocity (m/s)
volume around the impeller where the shear stresses
are large enough to disrupt the fiber network
Figure 10 Volume fraction as a function of
structure and where the flow will be turbulent (5%
minimum velocity (m/s).
consistency). Figure 9 shows a similar wireframe, but
here the wireframe encloses the volume where the velocities are larger than 1E-3 m/s. The region
outside the wireframe can be regarded as stagnant.
Figure 10 shows a velocity distribution diagram. The volume fraction of fluid moving with
a velocity larger than a certain minimum velocity is plotted for various consistencies. This diagram
shows that this agitator provides motion control for consistencies which are lower than 5%, which
agrees well with the results of the Chemineer design procedures. The velocities of 1E-3 m/s may
seem low at first, but creeping laminar flow is usually all that is needed in this type of storage chests.

CONCLUSIONS
For the first time a model has been developed for the multi-regime flow of a fluid with a complex
rheology like paper pulp. The results of the model calculations agree satisfactory with experimental
flow visualizations.
The model can be used to predict the flow pattern in large industrial chests where zones with
turbulent mixing, zones with laminar mixing and stagnant zones can easily be located. This model
is an excellent tool for the optimization of agitators for large industrial chests.

REFERENCES
[1]

Gullichsen J. (1985)
Medium Consistency Processing - Fundamentals
Bleach Plant Operations/TAPPI Seminar Notes, 135-142

PAPER PULP

NOTATION
c1
c2
g
k
N
p
Pk
S
u
U
x

e
k

d
y

t
eff

Model Constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (-)

Model Constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (-)
Acceleration by Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (m s-2)
Turbulent Kinetic Energy Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (m2s-2)
Impeller Rotational Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (s-1)
Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (Pa)
Production of k . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (m2s-3)
Shear Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (s-1)
Fluctuating Component of Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (m s-1)
Mean Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (m s-1)
Coordinate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (m)
Density of Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (kg m-3)
Dirac Delta function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (-)
Model Constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (-)
Model Constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (-)
Shear Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (Pa)
Disruptive Shear Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (Pa)
Yield Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (Pa)
Molecular Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (Pa s)
Turbulent Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (Pa s)
Effective Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (Pa s)
/ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (m s-2)
t/ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (m s-2)