Professional Documents
Culture Documents
Mixing Dynamics of Pulp Suspensions in Cylindrical Vessels: Kwok Wai Leo Hui, 2011
Mixing Dynamics of Pulp Suspensions in Cylindrical Vessels: Kwok Wai Leo Hui, 2011
Mixing Dynamics of Pulp Suspensions in Cylindrical Vessels: Kwok Wai Leo Hui, 2011
IN CYLINDRICAL VESSELS
by
DOCTOR OF PHILOSOPHY
in
(Vancouver)
April 2011
Abstract
Cylindrical agitated chests are frequently used to facilitate manufacturing processes in pulp and
paper industry and one of their main functions is to attenuate any process disturbances. However,
owing to the inherited non-Newtonian nature of pulp suspensions, it is not easy to achieve
complete mixing and with the improper chest design, these agitated chests do not always perform
ideally or satisfactorily. The cavern formation in incomplete mixing may induce bypassing and
dead zones, which significantly affect the chest performance. A study of cavern formation in a
cylindrical agitated chest was thus carried out. Also, a dynamic model developed by Soltanzadeh
et al. (2009) was used to quantify the mixing dynamics of the cylindrical chest. In addition, using
computational fluid dynamics (CFD), the simulated results of the flow in the chest were
compared with the experimental results to verify the applicability of the CFD model on the study
tomography (ERT) and ultrasonic Doppler velocimetry (UDV) were applied to estimate the
cavern shape and size. Both methods gave satisfactory results and as expected, the cavern size
was found to increase with impeller speeds. The cavern shape was best described as a truncated
right-circular cylinder. Based on this observation, a model considering the interaction between
the cavern and chest walls was developed to calculate the cavern volume.
With the dynamic model, a series of dynamic tests were carried out to characterize the
mixing behavior of the lab-scale cylindrical chest. It was found that the proposed flow
configuration with the outlet close to the cavern could minimize the bypassing which affects
mixing quality. Also, ERT verified the presence of cavern and dead zone when the chest was not
Numerical simulations using CFD were compared with the experimental results under
different operating conditions. Pulp suspensions are a mixture of water and wood fibres that can
entangle each other to form flocs affecting the mixing flow. Owing to this complex rheology, it
is not easy to model the agitation precisely in CFD using a homogeneous fluid model. The floc
formation and air entrapment observed in experiments were difficult to be numerically taken into
account in the simulations. Although the CFD model could not exactly predict the mixing
situation of pulp suspensions, it still can be used to estimate the mixing flow patterns, e.g., flow
Preface
Ultrasonic Doppler velocimetry (UDV) were applied to estimate the cavern shape and size. I
selected the conductive tracer to identify the cavern and designed the measurement method to
determine its size. After carrying out the experiments, I analyzed the data, developed a cavern
model with the help of Chad Bennington and wrote the manuscript. Finally, my two supervisors,
Chad Bennington and Guy Dumont, revised and edited the manuscript. A version of this chapter
has been published as “Cavern formation in pulp suspensions using side-entering axial-flow
Chapter 4 illustrates the study of mixing dynamics in a lab-scale cylindrical chest. Using a
pseudo-random binary signal (PRBS) to control the addition of a conductive tracer, I carried out
a series of tests to analyze the mixing dynamics of the chest. With the input-output data and the
dynamic model, I estimated the model parameters using MATLAB and concluded the chest
performance. Finally, I wrote the manuscript which was revised and edited by my two
supervisors. A version of this chapter will be submitted for publication as “Mixing Dynamics in
Table of Contents
Abstract ........................................................................................................................................... ii
Preface............................................................................................................................................ iv
Table of Contents ............................................................................................................................ v
List of Tables ................................................................................................................................ vii
List of Figures .............................................................................................................................. viii
List of Symbols and Abbreviations............................................................................................... xii
Acknowledgements ..................................................................................................................... xv
1 INTRODUCTION .................................................................................................................. 1
1.1 Thesis background............................................................................................................ 1
1.2 Pulp suspension rheology ................................................................................................. 2
1.3 Agitated chest design ....................................................................................................... 6
1.4 Research objectives .......................................................................................................... 9
2 STUDY METHODS ............................................................................................................. 11
2.1 Electrical resistance tomography (ERT) ........................................................................ 11
2.2 Ultrasonic Doppler velocimetry (UDV) ......................................................................... 14
2.3 Computational fluid dynamics (CFD) ............................................................................ 17
3 CAVERN MEASUREMENT IN PULP SUSPENSIONS ................................................... 23
3.1 Introduction .................................................................................................................... 23
3.2 Experimental set-up and procedure ................................................................................ 26
3.3 Results and discussion.................................................................................................... 30
3.3.1 Selection of cavern measurement technique ........................................................... 30
3.3.2 Effect of impeller speed and pulp mass concentration on cavern volume .............. 37
3.3.3 Effect of pulp type on cavern volume ...................................................................... 38
3.3.4 Impeller offset from the rear wall of the mixing chest ............................................ 39
3.3.5 Comparison of cavern volumes determined by ERT and the cavern models .......... 40
3.3.6 Cavern model including interaction with vessel walls ........................................... 44
3.4 Summary ........................................................................................................................ 52
4 DYNAMIC TEST STUDY ON LAB-SCALE CHEST ....................................................... 53
4.1 Introduction .................................................................................................................... 53
4.2 Experimental set-up and procedures .............................................................................. 56
4.3 Results and discussion.................................................................................................... 61
4.4 Summary ........................................................................................................................ 75
5 CFD SIMULATION OF PULP MIXING ............................................................................ 76
5.1 Introduction .................................................................................................................... 76
5.2 Experimental study of pulp mixing ................................................................................ 79
5.3 CFD modeling of pulp mixing ....................................................................................... 80
5.3.1 Computational geometry ......................................................................................... 80
5.3.2 Modelling suspension rheology .............................................................................. 82
5.3.3 Condition setup for CFD model .............................................................................. 84
5.3.4 Cavern volume determination and dynamic test simulation ................................... 86
5.4 Results and discussion.................................................................................................... 87
5.5 Summary ...................................................................................................................... 108
vi
List of Tables
Table 3.1 Comparison of ERT/UDV cavern measurements (Cm = 2% hardwood pulp, E/D = 0.6,
C/D = 0.7 and Z/T = 1.0) ............................................................................................................... 36
Table 4.1 Time delay (Td) measured versus theoretical values for hardwood pulp Cm = 3% (N =
425, 525, 672 and 810 rpm at each flow rate) .............................................................................. 61
Table 4.2 Time constant (2) measured for complete mixing of hardwood pulp at Q = 14 L/min
(Theoretical 2 = 149s) .................................................................................................................. 62
Table 4.4 Calculation of impeller speed on scale-up using different criteria ............................... 73
Table 4.5 Power predicted for complete mixing of a hardwood pulp suspension in cylindrical
stock chests using different scale-up criteria (D/T = 0.43, Z/T = 0.8) .......................................... 73
Table 5.2 Grid independence study (N = 425rpm, 3% softwood pulp (K1 = 135, n = 0.21, y =
155Pa) ........................................................................................................................................... 91
Table 5.3 Comparison of results from CFD and dynamic tests for Cm = 2% hardwood (Q =
14L/min., Z/T = 0.8).................................................................................................................... 103
Table 5.4 Comparison of results from CFD and dynamic tests for Cm = 3% softwood (Q =
14L/min., Z/T = 0.8).................................................................................................................... 103
viii
List of Figures
Figure 1.1 A generalized stress-rate curve for a fiber suspension (Gullichen and Harkonen, 1981). 4
Figure 2.1 Operation of electrical resistance tomography (Malmivuo and Plonsey, 1995). ........ 12
Figure 2.2 Image reconstruction grid (Industrial Tomography Systems, 2007). .......................... 13
Figure 2.3 Principle of UDV showing the measurement of velocity along the measuring line
(Takeda, 1995). ............................................................................................................................. 16
Figure 2.4 Cell types used in computational grid (Ferziger and Peric, 1999). ............................. 20
Figure 2.5 2D structured and unstructured grids (Ferziger and Peric, 1999). .............................. 20
Figure 3.1 Cross-section of cylindrical stock chest showing ERT sensor planes and range of
impeller positions used. ................................................................................................................ 28
Figure 3.3 Photograph of conductive tracer particles used in tests. Individual particles range in
diameter from 4 to 6 mm (typically 5 mm) and length from 15 to 20 mm (typically 17.5mm). .. 32
Figure 3.4 ERT images obtained for a Cm = 3% hardwood pulp suspension agitated at N = 650
rpm with E/D = 0.4, C/D = 0.7 and Z/T = 1.0. Images on the left were reconstructed using data
obtained with tracer particles (Fig. 3.3) and images on the right were reconstructed by adding a
conductive fluid of 10.9 mS/cm NaCl (15mL) to a pulp suspension having a background
conductivity of 0.2 mS/cm (43.4L). The impeller is located between planes 2 and 3 and
electrodes 5 and 6 as shown for reference purposes in the upper left image. ............................... 33
Figure 3.5 Comparison of ERT images reconstructed using the tracer particles (left) and saline
solution (right). Uppermost sampling plane (P5). The visual surface view is given in the centre
image with the dotted line enclosing the region of active surface motion (determined visually).
(Hardwood pulp suspension at Cm = 3%; N = 650 rpm, E/D = 0.4, C/D = 0.7, Z/T = 1.0). ......... 34
Figure 3.6 Probe locations for UDV measurements. The probe is moved around the chest to
obtain the velocity profiles needed to map the cavern boundary. The arrows show the direction of
the sonic emissions and the numbers give the ERT electrode positions....................................... 34
Figure 3.7 (a) Velocity profile measured by UDV for a representative probe location. The
locations where the velocity reaches zero indicate the cavern boundary. (b) Cavern cross-section
formed by connecting the boundary points measured from all sample locations. ........................ 35
ix
Figure 3.8 Comparison between ERT and UDV caverns made for hardwood pulp at Cm = 2% and
N = 200 rpm (E/D = 0.6, C/D = 0.7 and Z/T = 1.0). ..................................................................... 36
Figure 3.9 Cavern shape (shaded volume) determined for a Cm = 2% hardwood at various
impeller speeds using ERT data (E/D = 0.6, C/D = 0.7, Z/T = 1.0).............................................. 37
Figure 3.10 Effect of impeller speed (N) and pulp consistency (Cm) on cavern volume (Vc ) for
hardwood and softwood pulps (E/D = 0.4, C/D = 0.7 and Z/T = 0.8). ......................................... 39
Figure 3.11 Effect of impeller position (E/D) on cavern volume (Vc) with (a) rotation speed and
(b) power (Cm = 2% hardwood, C/D = 0.7, Z/T = 1.0). ................................................................ 41
Figure 3.12 Comparison of measured cavern volumes and predictions made using the equations
of Amanullah et al. (1998) and Elson (1990). .............................................................................. 42
Figure 3.13 Axial force number (Nf) (a) and power number (NP) (b) for the D = 140mm Maxflo
impeller as a function of impeller speed and suspension mass concentration. ............................. 43
Figure 3.14 Diagram showing intersecting circles with the impeller located at the centre of a
virtual cylinder. The overlapping area between the mixed zone and mixing chest is the cavern
region. ........................................................................................................................................... 46
Figure 3.15 Rotation speed (a) and wall stress calculated (b) for complete mixing in the vessel
with hardwood pulp as a function of suspension mass concentration (E/D = 0.4, C/D = 0.7). .... 47
Figure 3.16 Comparison of wall stresses measured/estimated and suspension yield stress for
hardwood pulp as a function of mass concentration. Estimates for w based on the yield stress (y
- measured), the point at which the cavern first completely filled the mixing vessel, and friction
losses in flowing pulp suspensions. .............................................................................................. 49
Figure 3.17 Cavern volume versus impeller speed for laboratory mixer. Comparison of
experimental data (measured by ERT) with model predictions for hardwood pulp. Z/T = 0.8, E/D
= 0.4. (a) Cm = 2%: y = 9 Pa ;w =12 Pa ; a = 0; Cm = 3%: y = 38 Pa ;w = 42 Pa ; a = 0; Cm =
4%: y = 69 Pa ;w =90 Pa ; a = 0. (b) (y = w = a) Cm = 2%: y = 9 Pa; Cm = 3%: y = 38 Pa; Cm =
4%: y = 69 Pa. .............................................................................................................................. 51
Figure 4.1Dynamic model for non-ideal flow in agitated pulp stock chests (2>>1). .................. 56
Figure 4.2 Cross-section (a) and photograph (b) of cylindrical stock chest. The arrows in (b)
show the direction of suspension flow in continuous operation. .................................................. 59
Figure 4.3 Schematic of apparatus used for the dynamic tests . ................................................... 60
Figure 4.4 Typical signals measured for model parameter identification. ................................... 60
x
Figure 4.5 ERT images of a dynamic test at Cm = 3% and N = 425rpm (E/D = 0.6, C/D = 0.7, Z/T
= 0.8). The suspension inlet and outlet positions are shown in the images of P4 and P1 at time t =
0 s, respectively. ............................................................................................................................ 65
Figure 4.6 Effect of impeller speed (N) and pulp consistency (Cm) on time constant (2) (E/D =
0.6, C/D = 0.7, Z/T = 0.8). ............................................................................................................ 66
Figure 4.7 Effect of flow rate (Q) on time constant (2) with rotation speed (Cm = 3%, E/D = 0.6,
C/D = 0.7, Z/T = 0.8). ................................................................................................................... 67
Figure 4.8 Effect of flow rate (Q) on mixing volume (Vmix) with rotation speed (Cm = 3%, E/D =
0.6, C/D = 0.7, Z/T = 0.8). ............................................................................................................ 67
Figure 4.9 Effect of pulp types (HW: hardwood and SW: softwood) on cavern size obtained in
dynamic and batch (ERT) operations at Cm = 3% (E/D = 0.6, C/D = 0.7, Z/T = 0.8). ................. 69
Figure 4.10 Effect of pulp types (HW: hardwood and SW: softwood) with similar yield stress
(38Pa) on mixing volume (Vmix) with rotation speed (N) (E/D = 0.6, C/D = 0.7, Z/T = 0.8). .... 69
Figure 4.11 Comparison between cavern size obtained in dynamic and batch (ERT) operations at
various hardwood pulp consistencies (Cm) (E/D = 0.6, C/D = 0.7, Z/T = 0.8, Q = 14 L/min for the
dynamic tests). .............................................................................................................................. 70
Figure 5.1 Experimental cylindrical chest and computational domain for (a) batch and (b)
continuous-flow mixing. ............................................................................................................... 82
Figure 5.2 Locations where velocity samples were determined for comparison of three different
meshes. .......................................................................................................................................... 89
Figure 5.3 Velocity profiles calculated for softwood Cm = 3% at two locations (a) and (b) in front
of the impeller (as shown in Figure 2) for three different meshing schemes at N = 425rpm. ..... 90
Figure 5.4 Comparison of cavern size obtained from ERT and numerical simulations for a Cm =
2% hardwood (Z/T = 1.0). ............................................................................................................. 91
Figure 5.5 Comparison of cavern size obtained from ERT and numerical simulations for a Cm =
3% softwood (Z/T = 0.8). .............................................................................................................. 93
Figure 5.6 Flow fields of batch mixing for a) Cm = 2% hardwood at N = 200rpm using Bingham
plastic model and b) Cm = 3% softwood at N = 475 rpm using Herschel-bulkley model. ........... 94
Figure 5.7 a) Top view and b) side view of probe locations for UDV measurements. The arrows
show the direction of the sonic emissions at three heights (12.5mm, 92.5mm and 252.5mm
measured from the chest bottom). ................................................................................................. 96
xi
Figure 5.8 Comparison of measured and computed velocity profiles at three positions for a Cm =
2% hardwood at N = 200rpm. Positive velocity means the velocity direction is away from the
probe. ............................................................................................................................................ 97
Figure 5.9 Comparison of measured and computed velocity profiles at three positions for a Cm =
2% hardwood at N = 250rpm. Positive velocity means the velocity direction is away from the
probe. ............................................................................................................................................ 98
Figure 5.10 Comparison of measured and computed cavern volumes. (a) Cm = 2% hardwood
pulp using Bingham plastic model for cavern determination (Z/T = 1.0); (b) Cm = 3% softwood
pulp using Herschel-Bulkley model for cavern determination (Z/T = 0.8). ............................... 100
Figure 5.11 Experimental and computed dynamic responses for Cm = 2% hardwood pulp. ...... 104
Figure 5.12 Experimental and computed dynamic responses for Cm = 3% softwood pulp. ....... 105
Figure 5.13 Path lines of particles (each particle is represented by a different color) simulated by
CFD using Cm = 2% hardwood pulp at N = 250rpm and Q = 14L/min...................................... 106
Figure 5.14 Experimental and computed determination of mixed volume for (a) Cm = 2%
hardwood pulp and (b) Cm = 3% softwood pulp in continuous operation. ................................. 107
xii
a distance between the impeller location and the vessel centre, (m)
A2 surface area of cavern, (m3)
b perpendicular distance from the intersection point of the vessel wall and the
cavern to the distance a, (m)
c speed of sound, (m/s)
C height between the chest bottom and the impeller centre, (m)
C1, C2 constant, dependent on the impeller type and impeller geometry
Cm mass concentration or consistency, (%)
D, D1, D2 impeller diameter, (m)
Dc cavern diameter, (m)
Dp pipe diameter, (m)
Dr, DR rotor diameter, (m)
DT housing diameter, (m)
E distance between the chest wall and the impeller centre, (mm)
f bypassing parameter
f1 factor to correct for temperature and pipe roughness
F total force on the cavern boundary, (N)
F external body forces
Fa axial force, (N)
Fd Doppler shift frequency, (Hz)
Fe transducer frequency, (Hz)
g gravitational acceleration, (m/s2)
H rotor height, (m)
Hc cavern height, (m)
H/L friction loss, (m water/100m pipe)
I unit tensor
k constant in the Herschel-Bulkley model
K numerical coefficient, constant for a given pulp
K1 consistency index
M, M1, M2 torque, (Nm)
Mo, Mo1, Mo2 momentum number (m4/s2)
Mo level momentum (m2/s2)
n power-law index
N, N1, N2 impeller rotation speed, (rpm)
Nf dimensionless axial force number; Fa/N2D4
NP power number; P/N3D5
p static pressure, (Pa)
P, P1, P2 power, (W)
Q pulp flow rate through the chest, (L/min.)
r position vector in the rotating frame
rc cavern radius, (m)
rv vessel radius, (m)
xiii
Greek symbols
angular velocity of the rotating frame
, , indices, constant for a given pulp
F power dissipation per unit volume, (W/m3)
strain rate (s-1)
y strain rate at yield stress, (s-1)
Doppler angle, ()
1 angle for determining the cavern-suspension surface, ()
2 angle for determining the cavern-wall surface, ()
fluid density, (kg/m3)
dynamic viscosity, (kg/ms)
o yielding viscosity, (kg/ms)
plastic viscosity, (kg/ms)
shear stress, (Pa)
stress tensor
1 time constant for bypassing, (s)
2 time constant for mixing zone, (s)
a friction force per unit area at the air suspension interface, (Pa)
th theoretical time constant of the system, (s)
xiv
Abbreviations
Acknowledgements
I would first like to express my profound gratitude and appreciation to my deceased supervisor
Dr. Chad P.J. Bennington for his guidance, encouragement and support throughout the research.
I also would like to give special thanks to Dr. Guy A. Dumont for his innumerable assistance and
I gratefully acknowledge the valuable advice and helpful suggestions of the members of my
I am grateful to Mr. Ali Soltanzadeh for his assistance in dynamic model parameter estimation
I acknowledge the assistance of all the staff in the Pulp and Paper Centre and the Chemical and
Biological Engineering Department at UBC. I am especially grateful to Peter Taylor and Tim
Patterson for their valuable help in the fabrication, installation and maintenance of the
experimental setup.
Financial support from the National Sciences and Engineering Research Council of Canada
To my parents,
1 INTRODUCTION
1.1 Thesis background
Pulp suspension mixing in stock chests is an important operation in pulp and paper industry. It
prevents the pulp suspension from dewatering, enhances chemical contact in bleaching stages
and blends pulp streams before papermaking. Improper mixing could lead to non-uniform
delivery of pulp stock and increase process variability, resulting in unstable production and poor
product quality. Although regulatory control loops are employed to minimize the effects of these
process upsets and to maximize product uniformity, they are only effective at regulating out the
disturbances at low frequencies (Bialkowki, 1992). Thus, agitated chests are required to remove
However, these chests do not perform ideally because of the complex rheology of pulp
suspensions and the ineffective design of feed flow in and out of the chests. The suspension
yield stress resulted from the mechanical strength of the wood fibre networks may lead to the
formation of mixing and stagnant zones, creating undesired flows in the agitated chests. Ein-
Mozaffari et al. (2003) analyzed the output signal of an industrial pulp stock chest to a step input
change and found that the non-ideal flows, i.e., bypassing and dead zones, deteriorated the
response signal. To identify what parameters affecting the responses of the agitated pulp chests,
Ein-Mozaffari et al. (2005) studied the mixing dynamics in a rectangular pulp chest using two
different pulp suspensions: a short-fibred hardwood pulp and a long-fibred softwood pulp over a
range of fibre mass concentration and addressed the importance of inlet and outlet locations
Other studies have also been carried out to investigate the flow patterns and dynamic
behavior of pulp suspensions in agitated stock chests and most of these studies were conducted
2
in rectangular chests (Bakker and Fasano, 1993; Ein-Mozaffari et al., 2003; Ford, 2004). Very
little research (Wilstrom and Rasmuson, 1998a) about pulp agitation has been carried out in
cylindrical chests, which are widely used, particularly as controlled mixing zone in bleaching
towers and as dilution zone for high density pulp storage chests. Therefore, the purpose of this
study is to investigate the mixing performance of cylindrical chests using electrical resistance
tomography (ERT), ultrasonic Doppler velocimetry (UDV) and dynamic tracer tests. In
addition, computational fluid dynamics (CFD) was used to simulate the mixing behavior of pulp
especially rheology, is important and essential. Rheology governs fluid motion during agitation.
Pulp suspensions, as a non-Newtonian fluid, have caverns (the turbulent mixed regions) formed
around impellers when they are not in complete mixing. The cavern size would significantly
affect the mixing efficiency of the stock chests which are designed, based on past experience and
semi-empirical techniques, with their shape and size to facilitate the agitation of pulp suspension.
Fluid rheology plays an important role in determining the fluid behavior in stirred vessels. A
pulp suspension is a heterogeneous mixture of water and wood fibres. The fibres interact and
entangle with each other to form a viscoelastic network consisting of fibre aggregates (flocs)
(Gullichsen and Harkonen, 1981). The longer the fibers, the easier the floc formation. This fibre
network can be described with two levels: MACRO (inter-network) and MICRO (intra-network)
scales (Wikstrom and Rasmuson, 1998b). The MACRO scale represents the network between the
3
fibre flocs and the MICRO is described as a 3-D fibre network within a fibre flock. This complex
structure determines the rheological behavior of the pulp suspension, which initially acts as a
non-Newtonian fluid having a significant yield stress (y) (Figure 1.1). When the applied shear
exceeds the suspension yield stress, the fibre network is disrupted and fluid-like motion can be
created within the suspension. At low shear stresses, the MACRO structure of the fibre network
is first broken and the flow of pulp suspension is dominated by the break-up between the fibre
flocs. The fibre flocs flow as fibre flock spheres. When the shear becomes higher and exceeds a
certain value (d) at the fluidization point, the MICRO network is ruptured into individual fibres,
i.e., fluidization and the suspension behaves nearly as a turbulent Newtonian fluid.
The simplest rheological model for a fluid possessing a yield stress is the Bingham model,
which is a linear equation for the shear rate (Silvester, 1985; Barnes et. al, 1989; Chhabra and
Richardson, 1999):
(1.1)
where is the shear stress,y is the yield stress, is the shear rate and is the plastic viscosity or
coefficient of rigidity.
However, based on the generalized stress-rate curve of the pulp suspension in Figure 1.1,
at the low shear rates where the shear stress is between y and d (the flow is laminar), shear-
thinning behavior is observed after the yield stress point. The Herschel-Bulkley model
(Macosko, 1993; Chhabra and Richardson, 1999), which contains a yield stress and a shear-
thinning parameter, can be better suited for describing the non-Newtonian nature of pulp
suspensions.
4
Laminar Turbulent
Shear stress,
Bingham model
d
y
Newtonian fluid
Shear rate,
Figure 1.1 A generalized stress-rate curve for a fiber suspension (Gullichen and Harkonen, 1981).
y k n (1.2)
The shear-thinning behavior is obtained when the power index, n, is set to less than 1 and k is a
constant in the model. For pulp suspensions, Ford (2004) estimated k and n to be 0.001 and 0.25
respectively.
To create motion or turbulence in the pulp suspension, the yield stress must be exceeded
and the required stress has been correlated with the suspension mass concentration by the
y Cm (1.3)
where y is the yield stress, Cm is the suspension mass concentration in percent, and are fitted
constants. Ranges of and were reported to be 1.18 - 24.5 and 1.25 - 3.02 respectively. For
the mass concentration between 0.6% and 30%, the yield stress will vary from 0.3 Pa to 3104Pa.
5
Pulp suspensions in a turbulent state are often referred as “fluidized”, which implies fibre
motion leading to energy dissipation (Bennington and Kerekes, 1996). Fluidization can be
quantified through power and energy expenditure, and Wahren (1980) applied this concept to
estimate the power dissipation per unit volume of pulp suspension for the onset of fluidization,
y2
F 1.2 10 4 C m
5.3
(1.4)
where F is in W/m3 and is the viscosity of water. However, the use of the water viscosity to
estimate the power required to fluidize a pulp suspension is inappropriate because the apparent
viscosity of the suspension is larger than that of water (Bennington and Kerekes, 1996). Thus
using the criterion chosen by Gullichsen and Harkonen (1981) for the onset of fluidization,
Bennington and Kerekes (1996) developed an expression to estimate the energy dissipation
2 .3
DT
F 4.5 10 Cm
4 2.5
;
DR
(1.5)
D
. T 3110
13 . ; . Cm 12.6
DR
where DT is the diameter of the housing and DR is the diameter of the rotor. When the gap size is
Owing to the complexity of the rheological behavior of pulp suspension, it is not easy to
agitate the suspension thoroughly. The shear stress provided by the agitators must exceed the
6
yield stress everywhere in the suspension to create motion through the whole chest. In actual
mixing situations, however, the applied shear stress may not be strong enough to create thorough
mixing in the suspension, leading to the formation of mixing regions surrounded by stagnation
zones.
To facilitate the bulk motion and the cavern generated by the impeller, the chest shape and size
are important in achieving thorough agitation in the suspension. Generally, the agitated pulp
stock chests are designed based on experience. One of the methods for designing these chests
was developed by Yackel (1990). Derived from the concept of “conservation of momentum”,
the method is based on the impeller momentum number and the level momentum. Each
Mo C1 N 2 D 4 (1.7)
where N is impeller rotation speed, D is impeller diameter and C1 is a constant that depends on
the suspension rheology, impeller type and impeller geometry. By referring to sets of diagrams
and tables relating chest diameter or width, consistency, stock level and retention time to obtain a
process number which is converted to the corresponding momentum number, the required
impeller size and horsepower for agitation could be determined. However, there is no direct
reference to suspension rheology in Yackel‟s method. The non-ideal flows like bypassing and
dead zone are not considered in the chest design and the degree of motion generated in the
In terms of chest geometry, rectangular chests are widely used in pulp mixing because
they save space when they are grouped together using common walls. The optimum shape for a
rectangular chest to achieve complete motion in a pulp suspension is a cube (Yackel, 1990).
Owing to space limitation or restriction, all chests cannot be perfect cubes and so a general rule
of thumb for rectangular chests is that the length to width ratio should not be over 1.5. Besides
rectangular ones, cylindrical chests are used in pulp and paper industry. Since cylindrical chests
take advantage of hoop stress design (tension), they generally have thinner walls than rectangular
chests of the same height (Yackel, 1990; Reed, 1995). Thus, it could be economical to have a
cylindrical chest rather than a rectangular one. To achieve complete motion of a pulp suspension
with the minimum power in a cylindrical chest, the ideal ratio of stock depth (Z) to the chest
The reduced-bottom chest is another chest design that is employed to save energy (Figure
1.2). It is like a high-density storage chest placed on top of a dilution chest where controlled
zone agitation is maintained. The top of the chest operates in the range of 5% to 12% stock
consistency and the stock flows slowly down into the bottom of the chest where dilution water is
added to reduce the consistency to 3 - 4%. The lower section is a cylindrical chest with a height
equal to one-half of the diameter (Yackel, 1990). The diameter of the upper section where plug
flow occurs to prevent any stagnant zones is usually designed to be 1.6-1.8 times the diameter of
the lower section. The overall height is three or more times the lower-section diameter.
Depending on the diameter of the lower section, a fillet or a solid horseshoe structure made of
metal or concrete may be used to reduce the volume of the dilution chest. It allows a smaller
mixing zone for decreasing the agitator horsepower without the build-up of unwanted stagnant
pulp zones (dead zones). A general guideline for maximum fillet size is one-third chest diameter
8
but owing to material cost, normally fillets are 25% or less of chest diameter (Devries and Doyle,
1995).
Z
T2
Chest specifications:
- T2 = T1(1.6-1.8);
- Zm = 0.5 T1; Zm Mixing zone
- Z/T1 = or 3
T1
Even though these design rules are followed, channeling (bypassing) of pulp through the
reduced-bottom chest may still occur because no consideration for cavern formation in the
controlled mixing zone is included in the design. The flows induced by the cavern may create
channeling in the upper chest portion where the high-density pulp suspension flows downward.
Also, if the cavern is not large enough, the resultant non-mixing regions will create some
shortcuts for the downward pulp stock to flow directly to the exit, without entering the mixing
zone. This channeling can reduce the residence time and create operation problems like poor
To analyze the mixing dynamics in cylindrical vessels, the investigation was divided into two
parts: experimental study and numerical simulation. Chapter 2 first describes two methods used
(UDV), and the numerical simulation method - computational fluid dynamics (CFD). Then the
following two chapters show the experimental results of a lab-scale cylindrical chest. In chapter
3, the mixing behavior of the chest in batch operation was characterized using ERT and UDV,
and the experimental results of the chest operated in continuous mode are given in chapter 4.
After the experimental study, a CFD model was developed to simulate the mixing situations for
verifying its ability to predict the experimental data and the evaluation is described in chapter 5.
Finally, the overall conclusions of the research and recommendations for future work are given
in chapter 6.
I. To characterize the batch mixing of pulp suspensions in a cylindrical chest with a side
entering agitator by studying the cavern formation and the flow fields with the aid of
II. To carry out dynamic tests under scaled industrial operating conditions for identifying and
determining the response of an experimental chest. The data will be used to estimate the
channeling parameter and the time constant (for the mixed zone) in the dynamic model of
chest dynamics.
III. To simulate the mixing situation by means of computational fluid dynamics (CFD) and
The achievement of the objectives can enhance the understanding of mixing dynamics of
pulp suspensions in cylindrical vessels under batch and continuous operations. The study can
provide information about the properties of cavern (mixing zone) in batch mixing. The cavern
shape and size directly affect the mixing capacity of agitated pulp stock chests. In continuous
operation, the dynamic tests can estimate the chest performance in terms of some quantitative
parameters. These parameters can provide some insights whether the existing chest is operated
at its full capability. Finally, the numerically simulated results can verify the predictability of the
CFD model for pulp suspension agitation, evaluating its potential as a tool for chest design.
11
2 STUDY METHODS
With its mathematical concept introduced in the nineteenth century, tomography is a technique
for reconstructing the internal distribution of 2D and 3D objects from multiple external
viewpoints to provide cross-sectional slice images through the object (York, 2001; Hoyle et al.,
2005). Among different types of process tomography, electrical resistance tomography (ERT) is
increasingly used by many researchers to study industrial processes because of its advantages
including non-intrusive flow visualization, high speed measurement and no radiation hazard
(Mann et al., 1997; Ma et al., 2001; Kim et al., 2006). It is used to determine the resistance
distribution in the domain of interest by obtaining a set of measurements using sensors that are
distributed around the periphery without affecting the flow or movement of materials. The main
continuous phase in the domain must be slightly conductive and the other phases (components)
have different values of conductivity. The resistance distribution in a cross region is obtained by
injecting electrical currents on the domain and measuring voltages on it through a number of
spaced electrodes which are mounted on its periphery. The current is injected by using a pair of
neighboring electrodes and the voltage differences are measured by using all other pairs of
neighboring electrodes (Figure 2.1a). This process is then repeated by injecting current using
different pairs of neighboring electrodes and the spatial gradients of electrical conductivity are
measured (Figure 2.1b). Since ERT can detect local changes in resistivity/conductivity, it can be
used to study the mixing dynamics of a system if the fluids to be blended have different
conductivities. An injection of a fluid tracer with high conductivity can help to examine the flow
12
patterns of those non-conductive mixing components. Also, the use of multi-plane electrical
sensors in ERT can provide a pseudo three-dimensional description of the mixing process.
a) b)
Figure 2.1 Operation of electrical resistance tomography (Malmivuo and Plonsey, 1995).
system (DAS) and an image reconstruction system (Dickin and Wang, 1996). For a cylindrical
chest, the sensor system is usually a set of 16 electrodes arranged in rings, across which an
electrical current is applied and the resulting voltage is measured. The DAS, mainly consisting
of a signal source, voltmeters and a control circuit unit, coordinates the current injection and the
voltage measurement to obtain the required data for image reconstruction. The image
regions of different conductivities within the mixing vessel by processing the collected data and
displaying the resultant distribution image. Further information about the three ERT components
is given in Appendix A. Although the ERT process is fast for obtaining results, the image
resolution is limited because only 316 pixels (the default square grid has 20 20 = 400 pixels)
are available for representing the cylindrical chest interior cross section (Figure 2.2).
13
ERT has been be used as a measurement technique in many applications like level
monitoring in a horizontal pipe (Ma et al., 2001), flow visualization (Bolton et al., 2004; Wang,
2005) and jet mixing within pipelines (Stephenson, 2007). Ma et al. (2001) applied ERT to
monitor air/water two-phase flow in horizontal pipes and developed a Liquid Level Detection
method to check the water surface level. In the analysis of jet mixing within pipelines,
Stephenson et al. (2007) demonstrated the capability of ERT to visualize tracer plume
development exiting a coaxial or side entry jet in mains water and to allow a measure of mixing
progress under different operating conditions. ERT is also a popular analysis method in mixing
processes in agitated vessels. Owing to the slurry conductivity, William et al. (1996) applied
ERT for the three dimensional imaging of the concentration of solids in a slurry mixer as a
function of key process variables (particle size, impeller type and agitation speed) and for
information collection to formulate the improved mixing models and mixer design. Holden et al.
(1998) demonstrated the ability of the ERT system to distinguish the different flow patterns from
two types of impellers and to examine the mixing processes by monitoring the dispersion of a
14
brine pulse tracer. Since the resistivity would be varied by the void spaces due to bubbles in the
fluid, Wang et al. (2000) detected and constructed the pseudo-stationary pattern of gas-liquid
mixing in 3-D using an 8-plane 16-ring elements ERT system. Hosseini et al. (2010) correlated
the ERT measurements in solid-liquid mixing to solid concentration profiles to quantify the
degree of homogeneity and studied the effects of impeller and particle characteristics on the
mixing quality. Similarly, by introducing a saline solution into the region near the impeller, ERT
can be applied to study the cavern size formed in pulp suspension mixing.
Ultrasonic Doppler velocimetry (UDV) is a real-time and non-intrusive fluid flow measurement
method to obtain a spatial distribution of the velocity field (Takeda, 1995; Shekarriz and Sheen,
1998). Initially developed for medical applications, it has been increasingly used as a tool to
study the physics and engineering of fluid flow (Takeda, 1999). Besides providing
spatiotemporal information, UDV could be an efficient flow mapping process for comparison
with numerical simulation. Unlike laser Doppler anemometry, it is also applicable to opaque
fluids like pulp suspensions because it is unaffected by the optical properties of the fluid. UDV
is based on pulsed ultrasound echography and relies on the principle of the Doppler frequency
shift of moving particles or scatters within the flowing fluid. An ultrasonic pulse (0.45 -
10.5MHz) is emitted from the transducer along the measuring line and the same transducer
receives the echo reflected form the surface of the particles suspended in the liquid (Figure 2.3).
The velocity field information is contained in the echo. Information on the position, X, from
which the ultrasound is reflected is extracted from the time delay, td, between the start of the
ct d
X (2.1)
2
where c is the velocity of sound in the medium. At the same time, the flow velocity (v) is
Fd c
v (2.2)
2 Fe cos
where Fd and Fe are Doppler shift frequency and basic frequency of ultrasound respectively.
Thus a velocity profile can be obtained by analyzing the echo signal to derive instantaneous
frequencies at each instant. The measurement accuracy of UDV stated by Takeda (1991) is good:
5% for velocity and 1% for position. In the case of velocity measurement in pulp suspensions,
the sonic velocity in the suspension would be affected by suspension temperature and fibre
concentration. At room temperature (20C), the sound velocity in water is 1482 m/s. When there
are fibres in the suspension, the sound velocity would increase, but the change would not be
significant (Xu, 2003). In a recent study of measuring pulp fiber suspension flow in a rectangular
channel, Xu used a commercial UDV instrument (DOP 1000, Signal Processing, Switzerland)
Figure 2.3 Principle of UDV showing the measurement of velocity along the measuring line
(Takeda, 1995).
Besides the normal instantaneous velocity measurement, UDV has been successfully
applied in some special flow configurations like oscillating pipe flow and T-branching flow of
mercury (Takeda, 1995). Also, in a stirred tank, Bouillard et al. (2001) showed that ultrasonic
techniques can be used to identify fluid recirculation patterns, cavern regions and velocity
fluctuations (turbulence) although their reliability are hampered by the presence of bubbles in the
liquid phase. In addition, Wang et al. (2004) studied how the temperature affected UDV
measurements and concluded that the temperature effect may not be a significant issue for
With regard to pulp suspensions, a number of studies have been carried out in pipe flows
and stirred tanks. Dietemann and Rueff (2004) analyzed the flow of a pulp suspension at three
consistencies up to 2.1 % in a transparent 80-mm duct using UDV and different flow regimes,
i.e., plug flow, mixed flow and turbulent flow, were clearly observed. Xu and Aidun (2005)
17
employed the pulsed ultrasonic Doppler velocimetry to measure the velocity profiles of fibre
suspension flow in a rectangular channel and a correlation was derived for the velocity profile of
fibre suspension in turbulent flow. In addition, Ein-Mozaffari et al. (2006) identified the flow
characteristics, i.e., flow number and pumping rate, of a Maxflo impeller by using the velocity
profiles measured by UDV. Since the flow velocity inside the mixing cavern is obviously higher
than that in the surrounding regions, UDV can be applied to detect the location and the
dimension of the cavern by measuring the velocity profile across the mixing zone. This method
could provide an accurate and well-defined picture of the cavern but it requires more time to get
In engineering analysis or design, many flows of practical interest cannot be easily described or
solved analytically. The approximate solutions of these flows and related phenomena can only be
obtained numerically. Computational fluid dynamics (CFD) is the numerical simulation of fluid
motion by using computers. It estimates and predicts the flow field by solving a set of
conservation equations for fluid flow, turbulence, chemical species transport and heat transfer.
These equations form a system of coupled non-linear partial differential equations (PDEs) which
describe the continuous movement of the fluid in space and time. The two important governing
equations for fluid flow are the conservation of mass and momentum. For mass, which is neither
created nor destroyed in the flows of engineering interest, the conservation equation is (Ferziger
v 0 (2.3)
t
18
where for fluid density, t for time and v for velocity vector. In the case of momentum
conservation, besides the transport by convection (the left-hand side of the Eqn. (2.4)), several
momentum sources like the pressure gradient, diffusion and other forces are also involved. If
gravity is the only body force with no other forces, the equation can be written as (Ferziger and
t
v vv p g (2.4)
where p is the static pressure, is the stress tensor, and g is the gravitational body force. The
first term is the rate of change of momentum, the second is the convection of momentum and the
v v T v I
2
(2.5)
3
where is the dynamic viscosity, I is the unit tensor and the second term on the right hand side is
To solve these equations numerically, all aspects of the process must be discretized, i.e., a
change from a continuous to a discontinuous formulation. The domain where the fluid flows
needs to be described by a series of connected control volumes, or computational cells and the
governing equations need to be written in an algebraic form. The process to decompose the
domain into a set of discrete sub-domains, or computational cells is called grid generation. In
the elements can be tetrahedra, wedge-shaped prisms, pyramids or hexahedra (Figure 2.4). Also,
the way of dividing the domain into a finite number of cells can be classified into two main
groups: structured and unstructured (Figure 2.5). In structured grids, cells of a single family
occupy the whole domain and the positions of grid points are uniquely identified in a way
19
logically equivalent to a Cartesian grid. This simple and regular cell connectivity creates a
standard matrix of the algebraic equation system, which can be easily solved by a number of
methods. The unstructured grid, on the other hand, can consist of cells of any shape, without any
restriction on the number of neighbor elements or nodes. It is more flexible than the structured
grid and applicable to very complex geometries. However, the node locations and neighbor
connections need to be specified explicitly, so the matrix of the algebraic equation system is
irregular and it takes more time to solve it than that for the structured grid. Generally, the density
of cells in a computational grid needs to be fine enough to capture the flow details, but not so
For discretization of the equations, the well known processes used in CFD are the finite
difference (FD) method and the finite volume (FV) method. In the finite difference method,
derivatives in the PDEs are written in finite differences evaluated at the grid nodes using
Taylor‟s series expansion or polynomial fitting (Ferziger and Peric, 1999). The FD method can
be applied to any grid type but generally it is mostly used in structured grids. It is also simple,
effective and very easy to obtain higher-order schemes on regular grids. In the finite volume
approach, the integral form of the conservation equations is applied to the control volumes (CVs)
defined by cells (Ferziger and Peric, 1999). The variables are approximated at the centroid of
each CV. This method is suitable for complex geometries and conservative by construction
provided that surface integrals are the same for the CVs sharing the boundary.
20
Figure 2.4 Cell types used in computational grid (Ferziger and Peric, 1999).
Figure 2.5 2D structured and unstructured grids (Ferziger and Peric, 1999).
The finite set of coupled algebraic equations resulting from the discretization process
should be solved simultaneously in every cell in the solution domain. Owing to the non-linearity
of the equations that govern the fluid flow and related processes, an iterative solution procedure
is required and two methods are commonly used. A coupled solution approach is the one where
all variables, or at a minimum, momentum and continuity, are solved simultaneously in a single
cell before the solver moves to the next cell, where the process is repeated. This approach is
excellent for equations to be linear and tightly coupled. On the other hand, when the equations
are complex and non-linear, a segregated solution approach is preferred. In this method, one
21
variable at a time is solved throughout the entire domain. Each equation is solved for its
dominant variable, temporarily treating the other variables as known, using the best available
values for them. This process is iterated through the equations until all equations are satisfied.
The segregated solution approach is popular for incompressible flows with complex physics,
CFD is a useful tool for studying the flow dynamics in fluid mixing. Many literatures
show that CFD can predict velocity distribution, evaluate industrial stirred equipments like
fermenters and solid suspension vessels, and optimize impeller design (Armenante et al., 1997;
Bhattacharya and Kresta, 2002; Paul et al., 2004; Li et al., 2005; Kasat et al., 2008; Ankamma
Rao and Sivashanmugan, 2010). Ein-Mozaffari and Upreti (2009) applied CFD to study the
performance of three impellers in mixing of pseudoplastic fluids (xanthum gum solutions) which
were treated as a Herschel-Bulkley fluid. The computed velocities were found to agree
reasonably well with the measured results. The validated model was then used to simulate the
mixing time and the results showed the fluid yield stress and the clearance of the impeller had
considerable effects on the mixing time. Using an experimentally validated CFD model, Bakker
et al. (2010) numerically determined cavern shapes in a pilot-scale flotation cell for a range of
mineral slurries. The combined Herschel-Bulkley and Bingham plastic rheology models were
used to describe the slurries. A model was developed to estimate cavern height, which was found
to be inversely proportional to the slurry yield stress, in the flotation cell and this would be useful
for an engineering approximation of the cell size in the preliminary design. Gomez et al. (2010)
investigated the flow field in a rectangular vessel filled with glycerin solution and equipped with
a side-entering agitator, using CFD and particle image velocimetry (PIV). The CFD predicted
22
velocities agreed very well with the PIV measurements and this confirmed CFD as a useful tool
CFD has also been used in the study of pulp mixing. With the rheological data from
Gullichsen and Harkonen (1981) for the pulp suspension, Bakker and Fasano (1993) applied
CFD to model the flow in a rectangular pulp chest with a side-entering impeller and the
simulated results were in satisfactory agreement with the flow field visualizations. Wikstrom and
Rasmuson (1998a) assumed the pulp suspension as a Bingham fluid and studied the effect of
fibre suspension rheology on flow behavior in a laboratory tank equipped with a jet nozzle
agitator. The comparison of experimental results with theoretical CFD-calculations showed that
the applied Bingham model could not fully describe the rheology of the pulp suspension since
the calculated flow field deviated increasingly from the measurements as the distance from the
impeller increased and so a new rheology model must be developed to predict the entire flow
field correctly. In the CFD modeling of a rectangular pulp stock chest with a side-entry agitator,
Ford (2004) described the pulp rheology using a modified Bingham plastic model and showed
that the velocity vectors obtained from the CFD simulations agreed qualitatively with the flow
patterns observed in the experiment. The non-desirable flows like bypassing and recirculation
within the chest could be described by the predicted flow fields. Using a Bingham approximation
for the suspension rheology, Bhole et al. (2009) used CFD to model impeller flow and the
computed power and axial thrust numbers were found to be very close to the experimental
Bhattacharya et al. (2010) studied two industrial pulp chests by simulating the process conditions
to calculate the steady-state flows using CFD. The simulated flows were compared with the
3.1 Introduction
Agitated pulp stock chests play an important role in pulping and papermaking operations. They
are used to prevent pulp suspensions from dewatering, control pulp consistency (mass
concentration) prior to other processing steps, and for blending pulp streams ahead of
stress, which under certain conditions allows creation of caverns (regions of active mixing)
around impellers. Ein-Mozaffari et al. (2003) showed that industrial pulp chests were not ideally
mixed and developed a dynamic model which included the possibility of stock channeling to
quantify the mixing quality attained. Channeling was exacerbated by the size and location of the
cavern relative to the stock exit (Ein-Mozaffari et al., 2005) and consequently to fully understand
pulp mixing, the cavern location, size and shape should be characterized and correlated with the
as a function of fluid properties and mixing conditions (Wichterle and Wein, 1981; Solomon et
al., 1981; Silvester, 1985; Elson, 1990; Amanullah et al., 1998; Wilkens et al., 2005). Solomon et
al. (1981) showed the presence of caverns in both shear thinning and yield stress fluids (CMC,
Carbopol and Xanthan gum) agitated with top-entering radial flow impellers in conventionally
stirred baffled tanks. The caverns formed were spherical and increased in size as impeller speed
was increased. Elson (1990) examined the effect of impeller type and speed on cavern
development using 1% Xanthan gum solutions in water in a similar mixing configuration. Prior
1
A version of this chapter has been published. Leo K. Hui, C.P.J. Bennington and G.A. Dumont (2009) Cavern
formation in pulp suspensions using side-entering axial-flow impellers, Chem. Eng. Sci., 64, 509-519.
24
to the cavern reaching the vessel wall, cavern volume per unit power input was greatest for the
axial flow impeller studied when compared with a Rushton turbine (radial flow impeller).
However, once the cavern interacted with the baffles vertical expansion of the cavern was only
slightly affected by impeller type with the height of the cavern increasing with the impeller
Several mathematical models have been developed to predict cavern size as a function of
mixing conditions and fluid properties. In general, these models specify a shape for the cavern
(based on observation) and then balance the shear force transported to the cavern surface with
the yield stress of the fluid. In all cases the models were developed for isolated caverns that do
not interact with the vessel walls. Soloman et al. (1981) assumed a spherical cavern with the
torque induced by the impeller acting tangentially at the cavern boundary. This gives an equation
Dc 4 N D
3 2 2
3 NP (3.1)
D y
where , N, D and y are the fluid density, impeller rotation speed, impeller diameter and fluid
yield stress, respectively. The impeller power number, NP = P/(N3D5), must be known or
measured and the yield stress Reynolds number, Rey = N2D2/y, calculated.
Elson et al. (1986) used an x-ray technique to image caverns produced in Xanthan gum
solutions with Rushton impellers. The cavern was described by a right circular cylinder of height
Hc and diameter Dc with the ratio of Hc/Dc set between 0.35 and 0.45 for the Ruston impeller.
Dc N 2 D 2
3
1
NP (3.2)
D H c Dc 1 3 y
2
25
Amanullah et al. (1998) considered the total force transported to the cavern boundary by
an axial flow impeller as being the sum of the tangential and axial force components. The model
1 N 2 D 2
2 2
Dc 4N P
N 2
(3.3)
y 3
f
D
where Nf =Fa/(N2D4) is the dimensionless axial force number that must be measured or
determined from correlations. Amanullah et al. extended the model to shear thinning power-law
fluids, and located the cavern boundary as the surface having a specified (and low) velocity.
The model by Elson et al. (1986) predicted the cavern (right-circular cylindrical cavern)
diameter within 9% for disc turbines. The model by Amanullah et al. (1998) predicted the
cavern (spherical) diameter for an axial flow impeller with Re > 20 to within 5 – 15%. As
volume scales with the cube of diameter, the error expressed on a volume basis ranged from 16
to 53%.
First, in most industrial pulp mixer configurations the impellers are side-entering. This will
impose interaction between the cavern and the vessel floor and side walls, which has not been
considered in past work. The yield stress of pulp suspensions can be measured and characterized,
but the preferred method of Amanullah (which requires fitting the rheological curve to a power
law relationship) would be difficult as the flow curve of a fibre suspension beyond the yield
In this chapter, cavern size and shape was measured as a function of mixer operating
conditions (impeller speed, impeller offset from the wall, suspension height in the chest) for
several pulp fibre suspensions (hardwood and softwood pulp with suspension mass
26
concentrations from 1 to 5%). A scaled version of a commercial axial flow impeller designed for
pulp suspension agitation was used in the standard side-entering configuration in a cylindrical
chest. Due to the opaque nature of the suspensions it was impossible to use direct optical
techniques to measure cavern size. Two methods, electrical resistance tomography (ERT) and
ultrasonic Doppler velocimetry (UDV) were evaluated to obtain the needed data. Measured
cavern sizes and volumes were then compared against the cavern models available in the
literature, and a new model that accounts for interaction with the vessel boundaries was
developed.
A transparent 1/10 scale-model of a cylindrical stock chest (T = 38.1 cm) was built for studying
mixing in pulp suspensions. The chest was equipped with a side-entering Maxflo impeller
(Chemineer Inc., Dayton, OH) having D = 16.5 cm mounted 12.0 cm (C) from the bottom (i.e.,
C/D = 0.7. C/D is typically in the range of 0.5 to 1.5 (AGIMIX International, Uddevalla,
Sweden)). The impeller assembly could be moved horizontally to study the effect of impeller
position on cavern formation and two offsets, E = 7.0 and 9.6 cm (giving E/D = 0.4 and 0.6),
were used. The laboratory chest was fitted with 6 equally-spaced sensor planes for electrical
resistance tomography (ERT) with each plane having 16 square (2.52.5 cm2) stainless-steel
electrodes flush-mounted around the vessel periphery at 22.5o intervals (Fig. 3.1). Three stock
height to chest diameter ratios (Z/T = 0.8, 1.0 and 1.2) were examined (based on
Equations 3.1 through 3.3 demonstrate the predicted dependences of cavern diameter on
yield stress, which for a pulp suspension depends primarily on fibre (pulp) type and suspension
27
mass concentration. Two bleached kraft pulps (hardwood and softwood) from Domtar Inc.
(Windsor, ON) were used in the study. The hardwood pulp had a length-weighted mean fibre
length of 1.28 mm and the softwood pulp had a length-weighted mean fibre length of 2.96 mm.
For the hardwood pulp, detailed cavern development was studied for three different suspension
mass concentrations (Cm = 2, 3 and 4%) while for the softwood pulp, only Cm = 3% was used.
The point at which the cavern completely filled the vessel was determined over a wider
concentration range for both pulps. The suspension yield stress was measured using a Haake
RV12 Rotovisco concentric cylinder viscometer (Thermo Fisher Scientific, Waltham, US). A
four-bladed vaned rotor (with a diameter of 19 mm and a height of 38 mm) was immersed in the
suspension and the rotor accelerated slowly from rest to a maximum speed of 0.2 min-1. The peak
value of torque was measured and used to calculate the yield stress
Tm
y (3.4)
3 H 1
Dr
2 Dr 3
where Tm, Dr, H and y are maximum torque, rotor diameter and height, and suspension yield
stress, respectively (Macosko, 1994). This calculation assumes that the yielding surface is
defined by the outer surface of the rotor blades. Figure 3.2 summarizes the yield stress as a
function of suspension mass concentration for the pulps studied in this work. For the hardwood
pulp suspension, the yield stress varied to the power 3.0 0.1 of suspension mass concentration;
for the softwood suspension the dependence was to the power 2.0 0.1. These dependences are
Two methods were evaluated for measuring the cavern size in our tests: electrical
resistance tomography (ERT) and ultrasonic Doppler velocimetry (UDV). ERT is a non-invasive
technique that can reconstruct the conductivity distribution within a region of interest from
28
electrical measurements made through a series of electrodes placed uniformly around the
periphery of a vessel. Electrical current is injected through a pair of neighboring electrodes with
voltage measurements taken between all remaining pairs of neighboring electrodes. The process
is repeated by rotating current injection around all the electrodes. The aggregate data is then used
to reconstruct the original conductivity distribution within the vessel (Mann et al., 1997; York,
2000). A P2000 ERT tomography instrument (ITS, Manchester, UK) was used to obtain the data
with the image reconstructed using the linear back projection algorithm. A suitable conductive
tracer was added to the cavern, with ERT used to locate the region of higher conductivity and
thus the cavern size and shape. Data collection using the ERT process is computerized and rapid,
with about 45 minutes required to obtain cavern dimensions for a given test. However, the spatial
resolution of the cavern boundary is limited due the characteristics of the electric field and the
algorithm and grid size used for image reconstruction. The measurement error of the cavern
boundary (based on the resolution of the image grid) is typically given as 5 to 10% of the chest
w = 5.1mm
381mm
P5
50.8mm
80mm
625mm
P4
P3
P2
120mm
P1
Figure 3.1 Cross-section of cylindrical stock chest showing ERT sensor planes and range of impeller
positions used.
29
1000
100
y (Pa)
10
hardwood
softwood
0.1
0.1 1 10
Cm(%)
Ultrasonic Doppler velocimetry (UDV) was used to measure velocity profiles across the
mixer and has been used successfully on pulp suspensions in other flow configurations
(Dietemann and Rueff, 2004; Xu and Aidun, 2005; Ein-Mozaffari et. al., 2006). A single probe is
used to transmit and receive the ultrasonic signals. An ultrasonic pulse, emitted at a set frequency,
travels through the suspension and is reflected back to the probe by the fibres suspended in the
flow. The time delay (t) and frequency shift (Fd) measured by the probe allow the velocity (v) of
the particles (fibres) a distance (X) along the probe sampling direction to be determined
ct
X (3.5)
2
Fd c
v (3.6)
2 Fe cos
30
where c is the speed of sound in the suspension, Fe is the frequency emitted from the transducer
and is the Doppler angle (the angle the particle trajectory makes with the propagation direction
Switzerland) was used in our study. The probe was placed at various locations around the vessel
periphery to locate the points at which the suspension velocity fell to zero and hence the cavern
boundaries. The measurement accuracy of the UDV technique is good: 5% for velocity and 1%
for position (Takeda 1991) which is equivalent to a spatial resolution of about 4 mm in this study.
However, reconstruction of the cavern boundary for one mixing condition takes about 2 days
because data collection and analysis are time consuming. Results obtained for selected tests
made with UDV and ERT were compared to determine the most suitable technique for our study.
The point at which the cavern completely filled the vessel was readily determined by
visual inspection as the laboratory chest is fabricated from Plexiglas and is transparent. In
addition, video clips were taken of selected tests to confirm the cavern shapes/dimensions by
locating the cavern boundary at the vessel wall and suspension surface. Image orientation
between visual, ERT and UDV data was accomplished easily using the ERT electrodes as
reference points.
Electrical resistance tomography (ERT) was able to image caverns formed in the pulp
suspensions once a conductive tracer was added to the well-mixed region around the impeller
and dispersed throughout the cavern. Initially, a concentrated saline solution was used as the
31
tracer. However, the reconstructed images did not agree with other measurements (visual
inspection of the cavern surface in contact with the chest wall or with UDV measurements).
Other difficulties occurred. For example, when the saline solution was injected into the mixed
region, although the signal (the conductivity contrast between traced and non-traced regions) was
initially strong, the reconstructed image became blurred or completely disappeared as tracer
spread and was diluted through the cavern. To overcome the difficulties with suboptimum
increase the initial concentration/dosage. In both cases the imaged cavern was larger than that
observed by other techniques which may be due to convection or diffusion of the aqueous tracer
To create reliable cavern images and avoid concerns about potential tracer diffusion,
conductive tracer particles were used. These particles were made by covering small pieces of
sponge with aluminum foil, with a typical particle being cylindrical, 5 mm in diameter and 17.5
mm in length, and having a volume of approximately 0.29 cm3 – a bit larger than that of a typical
fibre floc. The tracer particle density ranged from 0.90 to 1.06 g/cm3 (they could absorb water
and change density with prolonged use) and would move inside the cavern with the fibre
suspension. The number of particles deployed in a test depended on the volume of the cavern to
be measured and was optimized to obtain a good ERT image. Typically 100 to 120 particles
were introduced into the mixing zone near the impeller – corresponding to about 0.2% of the
cavern volume. The particles are shown in Figure 3.3. Figure 3.4 compares ERT images
reconstructed for a one set of test conditions (hardwood pulp at Cm = 3% with N = 650 rpm)
made using both the saline solution and the conductive particles as tracers. The cavern shape in
each measurement plane is represented (P1 is at the base of the chest and the impeller is located
32
between electrodes 5 and 6 and imaging planes 2 and 3, see Figure 3.1). In all cases the cavern
size imaged is larger when the saline tracer solution is employed. Figure 3.5 shows that the ERT
image obtained using tracer particles (on the right) agreed best with visual inspection of the top
surface of the suspension (note that P5 is 4.85 cm below the suspension surface).
Figure 3.3 Photograph of conductive tracer particles used in tests. Individual particles range in
diameter from 4 to 6 mm (typically 5 mm) and length from 15 to 20 mm (typically 17.5mm).
UDV was used to measure fibre suspension velocity profiles at eight different positions
spaced evenly between the ERT sensors on each sample plane as illustrated in Figure 3.6. The
cavern boundary was located at the points where suspension velocity fell to zero along this
measurement chord. By rotating the probe around the vessel, an image of the cavern was created
(Figure 3.7). Repeating these measurements for the each imaging plane allowed a three
dimensional reconstruction of the cavern to be created that was directly comparable with the
ERT measurements and from which the cavern volume could be calculated.
33
Plane 2
Plane 3
Plane 4
Figure 3.4 ERT images obtained for a Cm = 3% hardwood pulp suspension agitated at N = 650 rpm
with E/D = 0.4, C/D = 0.7 and Z/T = 1.0. Images on the left were reconstructed using data obtained
with tracer particles (Fig. 3.3) and images on the right were reconstructed by adding a conductive
fluid of 10.9 mS/cm NaCl (15mL) to a pulp suspension having a background conductivity of 0.2
mS/cm (43.4L). The impeller is located between planes 2 and 3 and electrodes 5 and 6 as shown for
reference purposes in the upper left image.
34
10
1
9
2
8 3
7 4
6 5
Figure 3.5 Comparison of ERT images reconstructed using the tracer particles (left) and saline
solution (right). Uppermost sampling plane (P5). The visual surface view is given in the centre
image with the dotted line enclosing the region of active surface motion (determined visually).
(Hardwood pulp suspension at Cm = 3%; N = 650 rpm, E/D = 0.4, C/D = 0.7, Z/T = 1.0).
13
12 14
11 15
UDV
10 16 probe
9 1
8 2
measuring line
7 3
6 4
5
Figure 3.6 Probe locations for UDV measurements. The probe is moved around the chest to obtain
the velocity profiles needed to map the cavern boundary. The arrows show the direction of the sonic
emissions and the numbers give the ERT electrode positions.
35
70
a) 60 b) 13
12 14
50 11 15
40 10 16
Velocity (mm/s)
30
x1
9 1
x2
20
8 2
10
Cavern region 7 3
0 6 4
x1 x2 5
-10
0 100 200 300 400
Distance (mm)
Figure 3.7 (a) Velocity profile measured by UDV for a representative probe location. The locations
where the velocity reaches zero indicate the cavern boundary. (b) Cavern cross-section formed by
connecting the boundary points measured from all sample locations.
The cavern areas for each sample plane and the total cavern volume are summarized in
Table 3.1 for a representative set of tests where both the ERT (using tracer particles) and UDV
techniques were used. A visual comparison of the cavern images measured in each plane for one
test condition is also presented in Figure 3.8. There is no comparison of images for P3 because
the design of the chest could not allow the velocity measurement using UDV probe at that plane.
Cavern images reconstructed using both ERT and UDV were similar, although the ERT data
consistently gave a larger cavern volume which we attribute to the image reconstruction
technique used. The UDV data underestimates the location of the cavern boundary as it measures
only the suspension velocity component in the direction of the probe. For example, in planes P4
and P5 (Figure 3.8), the cavern area measured by UDV does not extend to the vessel wall
although suspension flowed vertically (observed visually) there. Despite these differences, the
measured cavern volumes are within 10% of each other, with the ERT technique predicting
36
slightly larger cavern volumes. The uncertainty in cavern volume is estimated to be 16% for
ERT and 7% for UDV measurements (based on the accuracy to which the cavern boundary can
be located by each technique). ERT, with conductive particles used as tracer particles, was used
for all subsequent tests due to the considerable reduction in time required for data acquisition and
analysis.
Table 3.1 Comparison of ERT/UDV cavern measurements (Cm = 2% hardwood pulp, E/D = 0.6,
C/D = 0.7 and Z/T = 1.0)
Impeller Speed, N (rpm)
Cavern area or
250 200 150
volume
ERT UDV ERT UDV ERT UDV
2
P1 (dm ) 6.42 6.39 5.59 4.83 5.19 4.32
P2 (dm2) 6.28 7.08 5.84 4.94 5.44 5.01
P4 (dm2) 7.11 5.62 5.37 3.88 3.67 3.48
P5 (dm2) 6.02 6.24 2.21 4.06 2.21 2.14
Volume (dm3) 24.6 24.1 18.1 16.9 15.7 14.2
12 13 14 12 13 14
11 15 11 15
10 16 10 16
9 1 9 1
8 2 8 2
7 3 7 3
6 4 6 4
5 5
P1 P2
12 13 14 12 13 14
11 15 11 15
10 16 10 16
9 1 9 1
8 2 8 2
7 3 7 3
6 4 6 4
5 5
P4 P5
Figure 3.8 Comparison between ERT and UDV caverns made for hardwood pulp at Cm = 2% and N
= 200 rpm (E/D = 0.6, C/D = 0.7 and Z/T = 1.0).
37
Figure 3.9 from ERT data at three impeller speeds. The cavern is cylindrical in shape, although a
pronounced taper is evident towards the suspension surface at lower rotational speeds. As the
impeller speed is increased, the cavern enlarges, particularly at the suspension surface, with the
line of
rotation
Figure 3.9 Cavern shape (shaded volume) determined for a Cm = 2% hardwood at various impeller
speeds using ERT data (E/D = 0.6, C/D = 0.7, Z/T = 1.0).
3.3.2 Effect of impeller speed and pulp mass concentration on cavern volume
The cavern volume increased with increasing impeller speed as shown in Figure 3.10. The
growth in cavern volume is fairly uniform until the cavern boundary begins to interact
significantly with the vessel walls (about half the suspension volume is agitated at this point).
Any further increase in impeller speed rapidly causes the cavern to fill the vessel. As suspension
mass concentration (and consequently the suspension yield stress) was increased, higher impeller
38
speeds were required to create a given cavern volume (in agreement with Eqns. 3.1 through 3.3).
However, in all cases once the cavern had grown to about half the vessel volume a more
dramatic increase in cavern volume was created for a modest change in impeller speed. This step
change is similar to the dramatic change in channeling that was observed for small changes in
impeller speed in dynamic mixing tests made previously (Ein-Mozaffari et al., 2005). Perhaps we
Figure 3.10 also includes a plot of cavern development for a softwood pulp suspension at
Cm = 3%. Here the suspension yield strength is significantly larger than that of the corresponding
hardwood pulp, which results in a reduced cavern volume for a given impeller speed. Pulp fibre
suspensions display unique flow behavior that depends on many aspects of the suspension,
primarily the suspension mass concentration but also the average fibre length and fibre length
distribution, fibre surface properties and the suspending medium. Although the suspension yield
stress is readily characterized, suspension behavior in flow has not yet been adequately described
by a rheological model.
39
0.040
3
0.035 suspension volume = 0.0347m
0.030
Vc ( m ) 0.025
3
0.020
0.015
Cm(%) y(Pa)
0.010
HW 2 9
HW 3 38
0.005 HW 4 69
SW 3 119
0.000
0 200 400 600 800 1000
N (rpm)
Figure 3.10 Effect of impeller speed (N) and pulp consistency (Cm) on cavern volume (Vc ) for
hardwood and softwood pulps (E/D = 0.4, C/D = 0.7 and Z/T = 0.8).
3.3.4 Impeller offset from the rear wall of the mixing chest
The location of the impeller in an industrial vessel is fixed, with the typical offset clearance
from the rear wall being from E/D = 0.3 to 1.0 (AGIMIX International, Uddevalla, Sweden).
Proximity to the wall can restrict suspension flow to the impeller and hence reduce the pumping
ability of the impeller. The impact of impeller clearance on cavern volume is shown in Figure
3.11 for our laboratory vessel. Here a Cm = 2% hardwood kraft pulp suspension was agitated
using two different impeller clearances. When the off-wall clearance was small (E/D = 0.4) the
cavern volume was smaller than when the off-wall clearance was larger (E/D = 0.6) for a fixed
impeller speed. As the power required by the impeller (at a fixed speed) was about the same
(Figure 3.11b) the larger offset resulted in slightly more energy-efficient mixing. However, any
40
increase in impeller offset may increase stresses on the impeller shaft and a balance between
3.3.5 Comparison of cavern volumes determined by ERT and the cavern models
The predictions for cavern volume made with the available literature correlations are
compared with our experimental data in Figure 3.12. Here cavern volume is plotted against the
yield stress Reynolds number, with predictions made using the Amanullah et al. (1998) model
for a spherical cavern and the Elson (1990) model for a right circular cylinder. In both cases the
specific power number and/or axial force number measured for the impeller under the test
conditions (identical impeller geometry, impeller speed, pulp type and suspension mass
a)
0.05
3
suspension volume = 0.0434m
0.04
0.03
Vc ( m )
3
0.02
0.01
E/D = 0.6
E/D = 0.4
0.00
100 200 300 400
N (rpm)
b)
0.05
3
suspension volume = 0.0434m
0.04
0.03
Vc ( m )
3
0.02
0.01
E/D = 0.6
E/D = 0.4
0.00
0 2 4 6 8 10
Power (W)
Figure 3.11 Effect of impeller position (E/D) on cavern volume (Vc) with (a) rotation speed and (b)
power (Cm = 2% hardwood, C/D = 0.7, Z/T = 1.0).
42
0.040
3
0.035 suspension volume = 0.0347m
0.030
Amanullah et al. (1998)
0.025
Vc ( m )
3
0.020
Measured Predicted
HW 2%
0.015 HW 3%
HW 4%
0.010 SW 3%
0.005
Elson (1990)
0.000
0 20 40 60 80 100
Rey
Figure 3.12 Comparison of measured cavern volumes and predictions made using the equations of
Amanullah et al. (1998) and Elson (1990).
The predictions of the Amanullah et al. (1998) model show the best agreement with the
experimental data, although the cavern formed is clearly cylindrical in geometry (Figure 3.9).
Amanullah‟s model predicts more rapid cavern growth with increasing Rey than found. Also, the
rotational speed required for the cavern to fill the vessel is under predicted. The model by Elson
(1990) significantly under-predicts cavern volumes for all test conditions despite modeling the
cavern as a cylinder. The main explanation for this is that axial thrust is not explicitly included in
the model. Elson also noted that once the cavern reached the baffles a further increase in impeller
speed increased cavern size at a slower rate than prior to contact with the vessel boundary,
opposite to that found here. As both models assume no interaction with vessel walls the
Consequently, we sought to develop a model where cavern interaction with the vessel was
accounted for.
a)
100
10
Hardwood Cm(%)
Nf
0
1
1 2
3
4
0.1
0 200 400 600 800 1000 1200
N (rpm)
b)
100
10
Hardwood Cm(%)
NP
0
1
1 2
3
4
0.1
0 200 400 600 800 1000 1200
N (rpm)
Figure 3.13 Axial force number (Nf) (a) and power number (NP) (b) for the D = 140mm Maxflo
impeller as a function of impeller speed and suspension mass concentration.
44
A model for cavern development with interaction between the cavern and vessel walls due
to impeller location is developed below. We begin with the equation developed by Amanullah et
al. (1998) to estimate the force transferred by an axial flow impeller to a spherical surface
2
4 Po
F N D 2 4
N
2
(3.7)
3
f
This force is assumed to be balanced with the total force resisting motion at the cavern surface
regardless of the cavern geometry. For unbounded caverns this is the suspension yield stress at
the cavern-suspension interface. For bounded caverns, as considered here, the total resistive
force can arise from interactions with three surfaces: the cavern-suspension surface, Sp, the
cavern-wall surface, Sw, and the cavern-air surface, Sa. Equating the force transferred by the
impeller (Eqn. 3.7) with the total contribution from these three surfaces gives
F y SP w Sw a Sa (3.8)
where y is the pulp suspension yield stress (measured as discussed previously), w is the friction
force at the vessel wall (which must be measured or estimated as detailed later), and a is the
friction force per unit area at the air suspension interface, taken as zero.
To proceed further we must specify the shape of the cavern. To a good approximation the
cavern is a truncated (due to interaction with the side of the mixing chest) right circular cylinder
(see Fig. 3.9) of radius rc (here the cavern “radius” is measured from the centre of the impeller to
the cavern boundary in front of the impeller) and height Z (extending from the base of the vessel
to the suspension surface). Once the cavern shape is specified the appropriate surface areas SP, Sw
Analytical equations for the area and arc lengths of two intersecting circles are given by
Martin (1874) and adapted to the cavern-vessel geometry as shown in Figure 3.14. The cavern-
45
vessel surface area, Sw, is given by two components – the interaction at the base of the vessel, Sb,
and interaction along the side walls. The surface area where motion occurs at the base of the
vessel is given by
2
r 2 a 2 rc 2 a 2 rc 2 rv 2 a 2 rc 2 rv 2
S b rv cos v
1 rc 2 cos 1 a rc 2
2
(3.9)
2arv 2arc 2a
where the impeller is located at the centre of a virtual cavern. Here rv is the vessel radius (rv =
T/2), rc is the cavern radius and a is the distance between the centre of a virtual cylindrical
cavern (the impeller location) and the centre of the mixing vessel (a = (T/2-E)). To find the
cavern-side wall and cavern-suspension surface areas, the distance, b, perpendicular to the line
running from the impeller to the centre of the vessel and the point where the cavern intersects the
vessel wall must be determined. We calculate this using Pythagoras‟ theorem, giving
rc rv a 2
2 2
b rc x 2 where x
2
(3.10)
2a
allowing the total cavern-wall surface area, Sw,
b b
S w 2 2 rv Z Sb 2 sin 1 rv Z Sb where 2 sin 1 (3.11)
rv rv
and the cavern-suspension surface area, Sp
b b
S p 21rc Z 2 sin 1 rc Z where 1 sin 1 (3.12)
rc rc
to be determined. Z is the height of suspension in the chest (the height of the cavern).
The cavern volume was calculated using an iterative procedure. The three surface areas,
Sa = Sb, Sp and Sw were determined as a function of rc for the chosen mixer geometry (given D, T,
E). The resistive force contributed by these interfaces (Eqn. 3.8) was then balanced with the
46
force generated by the impeller (Eqn. 3.7) to determine the cavern radius associated with a given
impeller speed. The cavern volume was then computed by multiplying the floor surface area Sb at
Chest wall
rc b rv
1 2
x
a
Impeller
Cavern region
Figure 3.14 Diagram showing intersecting circles with the impeller located at the centre of a virtual
cylinder. The overlapping area between the mixed zone and mixing chest is the cavern region.
To use Eqn. 3.8 we must determine the resistive force at the suspension-wall interface, w,
which we did by measuring the rotational speed at the point where the cavern first completely
filled the vessel. The impeller power number and flow number were determined using the data in
Figure 3.13 which permitted the force at the interface to be calculated using Eqn. 3.7. (The
power numbers measured during the cavern tests and those determined using the data from
Figure 3.13 agreed within 3%). At this point Sp = 0, a = 0. Sw is given by the vessel geometry
and w was obtained by dividing the total applied impeller force by Sw (shown in Figure 3.15(b)
a)
1600
N (rpm) 1200
800
400
Z/T
0.8
1.0
1.2
0
0 1 2 3 4 5
Cm(%)
b)
200
150
w (Pa)
100
50
Z/T
0.8
1.0
1.2
0
0 1 2 3 4 5
Cm(%)
Figure 3.15 Rotation speed (a) and wall stress calculated (b) for complete mixing in the vessel with
hardwood pulp as a function of suspension mass concentration (E/D = 0.4, C/D = 0.7).
48
w can also be estimated from published pipe friction data. We used the correlation
presented in the TAPPI (1998) technical information sheets to do this. Here the head loss in a
where the factors f1, K,, and are given for a range of pulp types for a hardwood pulp: f1 =
1.15, K = 236, = 0.27, = 1.78 and = -1.08 (TAPPI TIS 0410-14). The chest diameter is T =
381mm. The velocity at which the head loss is determined must be specified and we chose
values of V = 0.001, 0.01 and 0.1 m/s for our calculations although the suspension velocity along
the cavern-wall interfaces will vary with position throughout the chest and with impeller speed.
Cm is the suspension mass concentration (%). Figure 3.16 compares these „pipe flow‟ estimates
of w with those determined using the point at which the cavern first filled the vessel and the
measured values of the suspension yield stress, y. The pipe flow estimates of w were much
our mixing vessel. The axial force and power number at a given impeller speed and suspension
consistency were determined using the data in Figure 3.13. The force generated by the impeller
was iteratively balanced with the forces at the cavern interfaces, as described previously. We
examined two scenarios. In the first (Figure 3.17a) we use the suspension yield stress as
measured in the rheometer for y (data in Figure 3.2), w determined at the point where the cavern
first filled the vessel (using the procedure described above), and a = 0. Figure 3.17a compares
model predictions (solid lines) with the experimental data (ERT) for hardwood pulp at three
mass concentrations. The dashed lines show the predictions with the suspension yield stress
varied by 20% (the average coefficient of variation for the yield stress measurements was 22%,
49
with a 20% variation in y giving a confidence interval of approximately 95%). The error bars
shown for selected experimental data represent 16% uncertainty in the cavern volume
measurements made using ERT. The figure shows that the model predicts the trend in cavern
development well, particularly for the suspensions at Cm = 3 and 4%. In general the experimental
points are reasonably close to those predicted by the model (volume predicted to an average of
21%). Note that because the model was used to evaluate w each curve intersects the
200
150
Measured y (HW)
w, y (Pa)
Experimental
100
Calculated TAPPI)
v=0.001m/s
v=0.01m/s
v=0.1m/s
50
0
0 1 2 3 4 5
Cm (%)
Figure 3.16 Comparison of wall stresses measured/estimated and suspension yield stress for
hardwood pulp as a function of mass concentration. Estimates for w based on the yield stress (y -
measured), the point at which the cavern first completely filled the mixing vessel, and friction losses
in flowing pulp suspensions.
50
From Figure 3.16 we see that the values for w measured using the onset of complete
motion in the vessel are almost identical to the values of y measured by rheometry. Indeed, it
seems that this estimate of w evaluates y. In any case, an argument can be made for requiring
that the suspension yield stress must be exceeded at all cavern interfaces. This is because the
suspension yield stress must be exceeded at all suspension-suspension interfaces just prior to the
cavern reaching the vessel wall or suspension-air interface. Thus for the second scenario, given
rheometry. These predictions agree better with the experimental data, and require that only one
suspension measurement – the suspension yield stress – be made. For both scenarios, cavern
development (shape of the volume vs. impeller speed curves) is largely determined by cavern
geometry and its interaction with the vessel walls. However, the model (using a = w = y) was
found to predict the cavern volume to an average of 13%, which is better than that found for
previous models. This accuracy depends primarily on the yield stress which was determined to
22%. In the future, computational fluid dynamics may improve cavern estimation by
calculating the cavern geometry directly; however, the specification of suspension rheology is
a)
0.04
3
suspension volume = 0.0347m
0.03
Vc ( m )
3
0.02
Cm(%)
2
0.01 3
4
Model
+/-20% y
0.00
0 200 400 600 800 1000
N (rpm)
b)
0.04
3
suspension volume = 0.0347m
0.03
Vc ( m )
3
0.02
Cm(%)
2
0.01
3
4
Model
+/-20% y
0.00
0 200 400 600 800 1000
N (rpm)
Figure 3.17 Cavern volume versus impeller speed for laboratory mixer. Comparison of
experimental data (measured by ERT) with model predictions for hardwood pulp. Z/T = 0.8, E/D =
0.4. (a) Cm = 2%: y = 9 Pa ;w =12 Pa ; a = 0; Cm = 3%: y = 38 Pa ;w = 42 Pa ; a = 0; Cm = 4%: y =
69 Pa ;w =90 Pa ; a = 0. (b) (y = w = a) Cm = 2%: y = 9 Pa; Cm = 3%: y = 38 Pa; Cm = 4%: y = 69
Pa.
52
3.4 Summary
Since pulp suspensions are opaque, cavern size must be imaged using indirect techniques. We
evaluated two methods, ERT and UDV, and found that both techniques gave satisfactory
measurements from which cavern shape and volume could be determined. However, ERT was
chosen due to the marked decrease in the time required to acquire data. The shape of the cavern
was best approximated as a truncated right-circular cylinder, and as expected, increasing impeller
speed increased cavern volume. However, development of cavern volume with increasing
impeller speed was not uniform, which was attributed to interaction between the cavern and the
vessel walls.
Current models for predicting cavern development in yield stress fluids treat the cavern
as being isolated with no interaction with the vessel walls and do not describe our experimental
results well. This prompted us to develop a model that included interaction with vessel
boundaries which required specification of cavern shape (based on ERT imaging) and
development of equations to balance the force provided by the impeller with the net resistive
force at the cavern boundaries. The model proposed predicts the trend in cavern volume with
increasing impeller speed well although the absolute cavern volume is only predicted to within
13%. These predictions are sensitive to the values of suspension yield stress which was measured
to 22%.
53
4.1 Introduction
Agitated stock chests are used in conjunction with process control algorithms to maintain process
and product uniformity during the manufacture of pulp and paper products. Process control is
effective at attenuating process disturbances below the cut-off frequency of the control loop,
which is determined by process dynamics and the type of control algorithm and tuning employed
(Bialkowki, 1992). Aggressive tuning increases the cut-off frequency, although over aggressive
tuning can create oscillations which amplify variability. The agitated stock chest acts as a low-
pass filter to remove the high-frequency variability and extend the range of disturbance
attenuation.
In most cases stock chests are designed based on past experience and semi-empirical
techniques (Yackel, 1990; Reed, 1995). Two common chest geometries are employed for pulp
suspension mixing: rectangular and cylindrical. Rectangular chests save space when grouped
together using common walls (as in the basement of a papermachine) while cylindrical chests
can take advantage of hoop stress design which allows the use of thinner walls for vessels of the
same height. For cylindrical pulp stock chests, complete motion across the suspension surface is
most readily achieved when the stock height (Z) to chest diameter (T) ratio is 0.8 (Z/T = 0.8)
(Yackel, 1990). An axial flow impeller mounted in a side-entering mode is typically used in this
application. Guidelines for impeller installation, including impeller offset from the rear chest
wall (E) and height from the chest bottom (C) are typically E/D = 0.3 – 1.0 and C/D = 0.5 – 1.5
2
A version of this chapter will be submitted for publication. Leo K. Hui, C.P.J. Bennington and G.A. Dumont (2011)
Mixing Dynamics in Cylindrical Pulp Stock Chests.
54
design strategy matches the momentum required to agitate a pulp suspension in a given
application with that generated by the impeller (Yackel, 1990). Cylindrical chests are widely
used in the pulp and paper industry, particularly for creating controlled mixing zones in
Agitated stock chests often do not behave ideally (Ein-Mozaffari et al., 2003). When a
non-Newtonian fluid, such as a pulp suspension, is agitated, a cavern (a region of active motion
around the impeller) is created (Wichterle and Wein, 1981; Solomon et al., 1981; Silvester, 1985;
Elson, 1990; Amanullah et al., 1998; Wilkens et al., 2005; Hui et al., 2009). In batch operations
there is limited or no fluid flow outside the cavern; in continuous operations, suspension flow but
limited mixing occurs outside the cavern. Ein-Mozaffari et al. (2003) measured the efficiency of
several industrial pulp chests and found significantly reduced attenuation efficiency. The non-
ideal flows identified included bypassing or channeling (where part of the feed directly flowed to
the exit without passing through the impeller zone) and the formation of dead zones (regions of
stagnant suspension that were disconnected from or moved significantly slower than the bulk
flow). These phenomena were exacerbated by the non-Newtonian nature of the pulp suspensions,
reducing chest mixing capacity and the ability to attenuate process disturbances.
To identify the factors that affected mixing quality, Ein-Mozaffari et al. (2005) studied
mixing dynamics in a laboratory-scale rectangular stock chest using two pulp suspensions, a
shorter-fibred hardwood pulp and a longer-fibred softwood pulp, over a range of fibre mass
concentrations. It was found that the cavern volume, which was affected by the suspension yield
stress and the impeller rotational speed, played a significant role in determining mixing dynamics.
As the yield stress decreased or impeller speed increased, the extent of bypassing and dead zone
formation was reduced and the degree of disturbance attenuation improved. In addition, the
55
imposed flow configuration, determined by the position of the stock inlet and outlet points
relative to the cavern, affected mixing performance. Bypassing was minimized when the pulp
outlet was located within the cavern zone. Other studies with pulp suspensions (the majority in
rectangular chests) have confirmed these findings (Bakker and Fasano, 1993; Ein-Mozaffari et
In this chapter, the dynamic response of a cylindrical chest was examined as a function of
impeller speed for several pulp suspensions. The results were also compared with previous
findings for a rectangular chest configuration. A dynamic mixing model (Soltanzadeh et al., 2009)
was used to quantify mixing dynamics. Figure 4.1 shows the continuous-time dynamic model
used to represent the observed behavior of a pulp chest. The model incorporates two parallel
suspension flow paths through the vessel: one that enters the cavern created by the impeller and
one that bypasses this mixed zone and exits the chest with minimal mixing. f is the fraction of
pulp stock that channels through the chest while the remaining stock (1-f) enters the mixing zone.
1 and 2 are the time constants for each zone, respectively (with 1 2). Note that it is possible
to have transport delay, Td, in the step response signal (due primarily to flow delay in the process
piping prior to and following the chest). Using input-output data obtained when the chest is
excited with an appropriately designed signal allows the model parameters to be identified. For
the model chosen (Fig. 4.1), ideal mixing is approached when the extent of bypassing approaches
zero and the time constant for the mixing zone approaches the theoretical time constant of the
system, i.e., th = Vt/Q, where Vt and Q are the suspension volume and flow rate.
56
1
f
1+1s
+
u y
+
1-f 1
1+2s
Figure 4.1Dynamic model for non-ideal flow in agitated pulp stock chests (2>>1).
The transparent 1/10 scale-model of a cylindrical stock chest (T = 38.1 cm) with a laboratory-
scale (D = 16.5 cm) side-entering axial-flow impeller (Maxflo, Chemineer Inc., Dayton, OH)
used in this study is shown in Fig. 4.2. The impeller was located 7.0 cm from chest wall (E/D =
0.4) and 12.0 cm from the bottom of the chest (C/D = 0.7). The pulp suspension was pumped
from a feed tank, through the chest and then to a discharge tank. The suspension was added to
the pulp chest through an inlet pipe placed at the suspension surface near the wall opposite the
impeller (a thin plastic sheet was attached to the pipe opening to avoid pulp splashing as it
dropped into the chest). The pulp exit was located at the wall behind and below the impeller.
This stock entrance/exit configuration is optimal for forcing the suspension feed through the
The mixing dynamics were measured by adding a conductive tracer (saline solution) to
the pulp feed using a pseudo-random binary signal (PRBS) to control tracer addition (Fig. 4.3)
57
and excite the chest dynamics at the appropriate frequencies. The conductivity of the suspension
entering and leaving the chest was measured (Fig. 4.4 shows one typical data set) with the first
portion of the experimental data used to estimate the model parameters and these parameters
were then used to model the second part of the data. The MATLAB program used for this model
parameter estimation is shown in Appendix B. The model output for the latter portion of the test
was compared with the measured data and the extent of agreement expressed as:
1 n y computedi y measuredi
% fit 100% 1 (4.1)
n i 1 y measuredi mean( y measured )
where ycomputed is the output calculated from the model and ymeasured is the actual output. The
effectively mixed volume (Vmix) was estimated from the identified parameters using Vmix = Q2(1-
f).
Dynamic tests were performed with hardwood and softwood bleached kraft pulps from
Domtar Inc. (Windsor, QC). The hardwood pulp had a length-weighted mean fibre length of 1.28
mm and the softwood pulp had a length-weighted mean fibre length of 2.96 mm. Suspension
concentrations of Cm = 2, 3 and 4% were used for the hardwood pulp and Cm = 1.8 and 3% for
the softwood pulp. The suspension yield stress for the pulps were measured using a Haake RV12
Rotovisco concentric cylinder viscometer (Thermo Fisher Scientific, Waltham, US) and a vaned
measuring bob, as described by Hui et al. (2009). The pulp flow rates through the chest were 7,
14 and 20 L/min, with the mean pulp residence time greater than 4 times the blend time (Yackel,
1990). The batch blend time was determined for a typical test condition (Cm = 3% hardwood
pulp at N = 500 rpm (the median impeller rpm used for the tests)) by measuring the response to a
tracer addition above the impeller using a conductivity probe located on the opposite wall, giving
58
a blend time of 6 s. The residence times at the highest flow rate of 20 L/min and at the lowest
flow rate of 7 L/min were 104 s and 298 s respectively (both > 24 s). The net power required to
create complete motion in the chest was determined using P = 2NM, where the impeller speed
(N) and net torque (M) were measured using an inductive-rotary torque transducer (model
0411IE50, Staiger Mohilo, Germany) mounted on the impeller shaft, with complete motion
determined by visually inspecting the vessel through the clear walls (i.e., the entire suspension
was moving at all the vessel boundaries, not just on the suspension surface).
For selected tests, electrical resistance tomography (ERT) was used to image flow
through the chest under conditions where a cavern existed. ERT is a non-invasive imaging
technique that reconstructs the conductivity distribution inside the mixing vessel using voltage
measurements made at the chest periphery. A step change of saline concentration was introduced
to the suspension feed with the vessel imaged using a P2000 ERT tomography instrument (ITS,
Manchester, UK) as a function of time. Four sensor planes (P1 - P4), each with 16 stainless steel
electrodes installed on the chest periphery, were used to measure the conductivity changes within
the chest and allowed the temporal reconstruction of the tracer‟s progress through the vessel.
ERT was also used under batch mixing conditions to compare the agitated cavern volume with
a)
Inlet
50.8mm
625mm
(24.6”) P4
P3
70mm
P2
120mm
P1
Outlet
b)
Feed
Impeller
shaft
Exit
Figure 4.2 Cross-section (a) and photograph (b) of cylindrical stock chest. The arrows in (b) show
the direction of suspension flow in continuous operation.
60
9
Liquid
Tracer
7
1 1
1 1
5
8
5 6
Feed Discharge
Tank 3 4 tank
Stock
chest 6
7
2
7
1 : Transmitter
2 : Indicator
3 : Tachometer
4 : Torque transducer
5 : Magnetic flowmeter
6 : Conductivity sensor
7 : Pump
8 : Solenoid valve
9 : Pulsation dampeners
7
Estimation Verification
6
Conductivity (mS)
3
Input
Output
2
0 600 1200 1800 2400
Time (s)
The accuracy of the dynamic model was verified by measuring the time delay (Td) and mixing
time constant (2) under conditions of complete vessel agitation and comparing them with
theoretical values. Under conditions of ideal mixing, the volume of the suspension in the piping
between the locations of conductivity measurement and the chest inlet and outlet should account
for the delay. The time delay (and its uncertainty) at several flow rates was estimated by
averaging the results of four independent tests made using hardwood pulp at Cm = 3% under
conditions where different extents of mixing existed in the vessel. Different impeller speeds
were used, but no trend in delay time with impeller speed was observed. The delay estimated
using the model was within experimental error of that calculated, as shown in Table 4.1 (the
uncertainty in the theoretical time delay was estimated using the accuracy of pipe and flow rate
measurements). Also, it was found that the 2 estimate (for complete mixing) was always
slightly larger than the theoretical value (by 6% on average, which is close to the error (7%) in
parameter estimation) (Table 4.2). As the preparation for each test is time-consuming, only one
Table 4.1 Time delay (Td) measured versus theoretical values for hardwood pulp Cm = 3% (N = 425,
525, 672 and 810 rpm at each flow rate)
Flow rate (Q) Td estimated from Td based on flow
(L/min.) dynamic tests (s) and system geometry (s)
7 79.6 9.5 81.9 6.4
14 38.4 2.0 40.9 2.9
20 31.7 4.4 28.7 2.1
62
Table 4.2 Time constant (2) measured for complete mixing of hardwood pulp at Q = 14 L/min
(Theoretical 2 = 149s)
Percentage of fitting
Cm (%) N (rpm) 2 (s) Eqn. (4.1)
(%)
380 1557 91
2
460 14210 91
672 1634 92
3
810 16121 78
830 16311 93
4
1000 16413 91
The data from all the tests are reported in Table 4.3. On average, the identified model
parameters were found to fit the experimental data to 88%, with all cases fitted above 78% (Eqn.
4.1). The flow configuration used in the tests forces the pulp feed through the cavern before
leaving the chest and consequently, the parameters f and 1 were expected to be low or negligible.
This was found in all but six cases. In the cases where the bypassing was identified it was found
To verify the presence of a cavern and a poorly agitated zone in a partially agitated chest
under continuous operation conditions, ERT was used to image the mixing of a pulse of tracer
(220 s in duration) through the chest under typical operating conditions – agitation of a
hardwood pulp suspension at Cm = 3%, with the standard vessel geometry (E/D = 0.6, C/D = 0.7
and Z/T = 0.8) using an impeller speed that would lead to cavern formation (N = 425 rpm and Q
= 14 L/min). Under the test condition 2 was found to be 70s (Table 4.3), implying that the
cavern occupied about 47% of the vessel volume since the theoretical 2 of complete vessel
The step change in tracer was followed as it entered the vessel at the suspension surface
and flowed towards the side exit located between ERT sensors 3 and 4 above plane 1 (P1), as
shown in the t = 0 image in Fig. 4.5. Note that before the injection of tracer, the conductivity in
the chest was low (0.6 mS/cm) as indicated by the blue color in the tomographic images. As the
saline solution entered the chest the local conductivity began to increase (to green in the images).
At t = 30 s the tracer moved to the impeller-side of the chest along the suspension surface by the
upper circulation loop and began to fill the top part of the vessel. The pulse then moved
downward and spread out to the lower planes (t > 30s). As more tracer accumulated in the chest,
the conductivity increased (indicated by the red colour in the images for t = 100s and 170s) and
the truncated cylindrical shape of the cavern can be clearly seen. Tracer injection stopped at t =
220 s, and the tracer began to be washed from the chest (t = 255 to 480s). Towards the end of the
test, the presence of tracer was observed to linger in regions of the vessel (note the green regions
in the top image plane, P4, of the t = 360 s image) which indicates regions of slower mixing. No
obvious sign of direct passing of tracer to the exit could be seen, which supports the “negligible
bypassing” results obtained from the dynamic tests. In addition, the ERT images show that
outside the cavern there is a stagnant region the tracer does not enter. This dead zone was also
observed visually during the experiment (lack of flow adjacent to the vessel wall). This suggests
that a fillet (the insertion of solid material in this region) could be used to reduce this unwanted
dormant volume.
64
5 13 5 13 5 13 5 13
4 14 4 14 4 14 4 14
outlet 3 15 3 15 3 15 3 15
2 1 16 2 1 16 2 1 16 2 1 16
9 9 9 9
30 7
8 10
11 7
8 10
11 7
8 10
11 7
8 10
11
6 12 6 12 6 12 6 12
5 13 5 13 5 13 5 13
4 14 4 14 4 14 4 14
3 15 3 15 3 15 3 15
2 1 16 2 1 16 2 1 16 2 1 16
9 9
40 7
8 9 10
11 7
8 10
11 7
8 9 10
11 7
8 10
11
6 12 6 12 6 12 6 12
5 13 5 13 5 13 5 13
4 14 4 14 4 14 4 14
3 15 3 15 3 15 3 15
2 16 2 1 16 2 16 2 1 16
1 1
9 9 9
100 7
8 9 10
11 7
8 10
11 7
8 10
11 7
8 10
11
6 12 6 12 6 12 6 12
5 13 5 13 5 13 5 13
4 14 4 14 4 14 4 14
3 15 3 15 3 15 3 15
2 16 2 1 16 2 1 16 2 1 16
1
9 9 9
170 7
8 9 10
11 7
8 10
11 7
8 10
11 7
8 10
11
6 12 6 12 6 12 6 12
5 13 5 13 5 13 5 13
4 14 4 14 4 14 4 14
3 15 3 15 3 15 3 15
2 2 16 2 16 2 1 16
1 16 1 1
9 9 9 9
220 7
8 10
11 7
8 10
11 7
8 10
11 7
8 10
11
6 12 6 12 6 12 6 12
5 13 5 13 5 13 5 13
4 14 4 14 4 14 4 14
3 15 3 15 3 15 3 15
2 1 16 2 1 16 2 1 16 2 1 16
9 9 9
255 7
8 9 10
11 7
8 10
11 7
8 10
11 7
8 10
11
6 12 6 12 6 12 6 12
5 13 5 13 5 13 5 13
4 14 4 14 4 14 4 14
3 15 3 15 3 15 3 15
2 16 2 1 16 2 1 16 2 1 16
1
9 9
290 7
8 10
11 7
8 9 10
11 7
8 10
11 7
8 9 10
11
6 12 6 12 6 12 6 12
5 13 5 13 5 13 5 13
4 14 4 14 4 14 4 14
3 15 3 15 3 15 3 15
2 1 16 2 16 2 1 16 2 16
1 1
9 9 9 9
360 7
8 10
11 7
8 10
11 7
8 10
11 7
8 10
11
6 12 6 12 6 12 6 12
5 13 5 13 5 13 5 13
4 14 4 14 4 14 4 14
3 15 3 15 3 15 3 15
2 1 16 2 1 16 2 1 16 2 1 16
9 9 9
480 7
8 9 10
11 7
8 10
11 7
8 10
11 7
8 10
11
6 12 6 12 6 12 6 12
5 13 5 13 5 13 5 13
4 14 4 14 4 14 4 14
3 15 3 15 3 15 3 15
2 16 2 1 16 2 1 16 2 1 16
1
low conductivity high conductivity
Figure 4.5 ERT images of a dynamic test at Cm = 3% and N = 425rpm (E/D = 0.6, C/D = 0.7, Z/T = 0.8). The
suspension inlet and outlet positions are shown in the images of P4 and P1 at time t = 0 s, respectively.
66
Figures 4.6 and 4.7 show that 2, and hence the mixing quality, increases with increasing
impeller speed until a plateau is attained. The error in determining 2 was ~7% (the average
standard deviation calculated in model parameter estimation using MATLAB for three hardwood
increasing the active mixed volume and hence 2. As suspension mass concentration increased,
the suspension yield stress increased which decreased cavern volume at a fixed impeller speed.
Thus a higher impeller speed was needed to maintain the same 2. At a fixed impeller speed, 2
decreased as the flow rate was increased because it is inversely proportional to Q (Fig. 4.7).
When Vmix was calculated using 2 (Fig. 4.8), it was found that for a fixed suspension
concentration, changing the flow rate did not significantly affect the mixed volume which
depended mainly on the impeller speed. The uncertainty of Vmix was estimated to be 10% based
200
100
s
Hardwood Cm (%)
50
2
3
4
0
200 400 600 800 1000
N (rpm)
Figure 4.6 Effect of impeller speed (N) and pulp consistency (Cm) on time constant (2) (E/D = 0.6,
C/D = 0.7, Z/T = 0.8).
67
0
400 500 600 700 800 900
N (rpm)
Figure 4.7 Effect of flow rate (Q) on time constant (2) with rotation speed (Cm = 3%, E/D = 0.6, C/D
= 0.7, Z/T = 0.8).
0.04
Vt = 0.0347m3
0.03
Vmix (m )
3
0.02
Q (L/min)
0.01 7
14
20
0.00
400 500 600 700 800 900
N (rpm)
Figure 4.8 Effect of flow rate (Q) on mixing volume (Vmix) with rotation speed (Cm = 3%, E/D = 0.6,
C/D = 0.7, Z/T = 0.8).
68
Figure 4.9 shows the change in mixed volume (cavern size) with impeller speed for the
hardwood and softwood pulps at Cm = 3%. Vmix for the hardwood suspension is larger than that
for the softwood for all impeller speeds. However, comparison on the basis of suspension
concentration is not appropriate as cavern size is determined primarily by the yield stress of the
suspension (Hui et al., 2009). A better comparison can be made using hardwood and softwood
suspensions of the same yield stress. In Fig. 4.10, the cavern volume of a Cm = 1.8% softwood
suspension is compared with that of the Cm = 3% hardwood suspension (both have a yield stress
of 38 Pa). We see that the mixed volume is very similar, except at the lowest impeller speeds
tested where the softwood pulp had the higher mixed volume. This same trend was also reported
by Ein-Mozaffari et al. (2005) for tests in rectangular chests. Thus, in addition to the suspension
yield stress other suspension factors (which we can only categorize in the broadest sense of „pulp
type‟) also affect mixing dynamics. As impeller speed is increased to approach the fully mixed
volume, the differences in cavern volume are reduced within our ability to measure them. Other
factors may account for these observations, including the uncertainty with which the yield stress
can be determined experimentally or the effect of the fibre type on the impeller flow and power
Finally, mixing volumes calculated using the dynamic tests were compared with those
measured using ERT under batch conditions as shown in Fig. 4.11. For the batch tests, the
cavern was imaged using conductive particle tracers added to the impeller region (see Hui et al.,
2009). It was found that the mixed volume determined using the dynamic model, Vmix, was
slightly larger than that determined by ERT under batch conditions, Vc. This may be due to the
location of the feed pipe and the momentum added to the mixed suspension by the entering flow.
69
However, changes in flow rate (Q = 7, 14, 20 L/min) did not significantly change the mixed
volume.
0.045
0.040
0.035 Vt = 0.0347m3
0.030
Vmix, Vc (m )
0.025
3
0.020
0.015
Figure 4.9 Effect of pulp types (HW: hardwood and SW: softwood) on cavern size obtained in
dynamic and batch (ERT) operations at Cm = 3% (E/D = 0.6, C/D = 0.7, Z/T = 0.8).
0.04
Vt = 0.0347m3
0.03
Vmix (m )
3
0.02
0.01
HW 3%
SW 1.8%
0.00
400 500 600 700 800
N (rpm)
Figure 4.10 Effect of pulp types (HW: hardwood and SW: softwood) with similar yield stress
(38Pa) on mixing volume (Vmix) with rotation speed (N) (E/D = 0.6, C/D = 0.7, Z/T = 0.8).
70
0.04
Vt = 0.0347m3
0.03
Vmix, Vc (m )
3
0.02
Continuous Batch
0.01 (Dynamic) (ERT)
HW 2%
HW 3%
HW 4%
0.00
0 200 400 600 800 1000
N (rpm)
Figure 4.11 Comparison between cavern size obtained in dynamic and batch (ERT) operations at
various hardwood pulp consistencies (Cm) (E/D = 0.6, C/D = 0.7, Z/T = 0.8, Q = 14 L/min for the
dynamic tests).
Scaling procedures are used to design mixers based on data obtained in geometrically
identical mixers at a different, usually smaller, scale. Yackel‟s (1990) design procedure for pulp
and paper chests uses the concept of level momentum. The momentum number (Mo) and level
Mo C1 N 2 D 4 (4.2)
Mo C1 N 2 D 4
Mo 2
2
(4.3)
3 3
V V
where C1 is a constant that depends on impeller type and geometry. The V2/3 term is due to the
proportionality between chest surface area and chest volume, which also fits the experimental
criteria Yackel uses (of having complete motion across the surface) to determine good mixing.
According to Yackel, designing for equal level momentum produces equal bulk velocities in the
2 4 2 4
C1 N1 D1 C 2 N 2 D2
2
2
(4.4)
V1 3 V2 3
where N1, D1 and V1 are the impeller speed, impeller diameter and fluid volume used in the
smaller chest and V2 is the volume of the larger chest having an impeller diameter of D2.
Assuming C1 = C2 (geometric similarity), the corresponding impeller speed (N2) for a cylindrical
1
N 2 D 4V 2 2 3 2
N 2 1 41 2 (4.5)
D V 3
2 1
Since V1 and V2 are proportional to T13 and T23, respectively, and that D1/T1 = D2/T2 (geometric
1
N1 2 D1 4T2 2 2
D
N 2 4 2
N1 1 (4.6)
D2 T1 D2
Other scale-up methods are used in literature, including constant power/volume (P/V) and
constant torque/volume (M/V). Constant P/V is the most commonly recommended scaling
criteria for mixing operations (Oldshue, 1983; Wilkens et al., 2003; Paul et al., 2004). This
method correlates well with mass-transfer characteristics in a mixer and is suitable for
calculating the power needed for gas dispersion or dispersion of immiscible liquids (Zlokarnik,
2006). Equal M/V is also a practical and common scale-up criterion because it relates directly to
the overall size, torque capability and cost of the mixer (Paul et al., 2004). It is the same as
scaling for constant tip speed and is usually applied when flow velocities in the impeller region
need to be the same in both vessels. The cavern model developed by Hui et al. (2009) can also be
used to scale pulp mixers. This model takes into account the interactions between the actively
mixed cavern and the vessel walls by balancing the forces produced by the impeller with those
72
acting at the cavern boundary. To create complete mixing in the vessel, the cavern boundaries
must extend to fill the mixer volume. Table 4.4 gives the dependence of N2 on N1 for four scale-
Table 4.5 gives the power predicted by scaling three- and ten-fold the volume of our
laboratory cylindrical chest using the criteria described above. Geometric similarity was
maintained (D/T = 0.43, Z/T = 0.8) with the base conditions measured in the 34.7L vessel for
complete motion. Of the methods considered, the constant P/V criterion gives the highest power
consumption. The constant M/V and level momentum scaling criteria give identical predictions
because, for geometric similarity, they both scale as N2=N1(D1/D2) as shown in Table 4.4.
Determining N2 using the cavern model does not require knowing N1, just the impeller
characteristics (Nf and Np) and suspension yield stress (y). The model predicts power levels close
to those calculated using the constant M/V and level momentum scaling methods. This is
expected as the cavern model also scales as N2=N1(D1/D2) for geometric similarity and constant
impeller behavior.
compare which shape is more efficient, the specific power to create complete mixing in a
rectangular chest (measured by Ein-Mozzaffari, 2002) was compared with that for a cylindrical
chest. As the cylindrical chest used in our study had a lower volume than that used by Ein-
Mozaffari, the base data was scaled to that of a chest having equal volume and allowing for the
different suspension height in the rectangular chest, Z/T = 1.2, and the fact that D/T changed
between chests. The power for the scaled cylindrical chest was found to be 1.4 W/kg and 4.8
W/kg for Cm = 1.8% and 3% softwood pulp suspensions, respectively. Both these predictions are
close to the values measured for the rectangular chest (1.4 W/kg at Cm = 1.8% and 5.4 W/kg at
73
Cm = 3%). Due to the use of scaling and the corrections needed to make this comparison we can
only state that vessel geometry, cylindrical or rectangular, requires approximately the same
F N 2 D2 N f
2 4 2
y A2
Cavern model N
3 2
D2 4 N f 2 4 N p 3 2
(A2: surface area of cavern;
Nf: axial force number of impeller)
*Assumptions: D/T = constant; Np = constant, geometric similarity
Table 4.5 Power predicted for complete mixing of a hardwood pulp suspension in cylindrical stock
chests using different scale-up criteria (D/T = 0.43, Z/T = 0.8)
Power (W)
Cm Scale-up Constant Constant Cavern
(%) y (Pa) volume ratio Level
power/volume torque/volume model
momentum
(Wilkens et al., (Wilkens et al., (Leo et al.,
(Yackel, 1990)
2003) 2003) 2009)
1:3 22 18 18 12
2 9
1 : 10 76 41 41 28
1:3 142 98 98 105
3 38
1 : 10 497 227 227 243
1:3 374 264 264 258
4 69
1 : 10 1310 593 593 579
74
Other factors should be considered when selecting a pulp mixer. Rectangular chests are
more susceptible to the creation of stagnant or dead zones, especially near the chest corners
(Ford, 2004). Also, a sudden transition to non-ideal behavior has been observed in a pilot-scale
rectangular chest under certain operating conditions (e.g., a dramatic increase of bypassing could
occur when the impeller speed was decreased slightly, all other operating conditions remaining
fixed) (Ein-Mozaffari, 2005). This was attributed to the location of the suspension feed and exit
relative to the impeller. This undesirable flow bifurcation could create fluctuations and/or
instability in the performance of an agitated pulp chest but was not observed in the cylindrical
chest. It can be avoided by ensuring that entrance and exit locations are optimized.
75
4.4 Summary
Dynamic tests were carried out in a laboratory-scale cylindrical chest to identify and characterize
the mixing quality. The dynamic model chosen, which included an actively mixed region with
the possibility of bypassing, was found to accurately describe the dynamics measured. The
insignificant bypassing measured indicated that the flow configuration used (with the exit located
within the active cavern, close to and below the impeller) is an effective configuration for
avoiding stock bypassing. Imaging, using electrical resistance tomography (ERT), verified the
size and shape of the caverns created when the chest was not fully agitated. Cavern size could be
mixing quality. The power required to completely agitate a cylindrical chest was similar to that
5.1 Introduction
Agitated stock chests are important in pulp and paper processing. They keep pulp suspensions in
motion to avoid dewatering and blend different pulp stocks or blend pulp with other chemicals.
They also act buffers between processes and act as low-pass filters to remove high frequency
variability and to complement the action of control loops. However, pulp chests do not always
function as expected. Ein-Mozaffari et al. (2003) studied the mixing performance of industrial
chests and noticed the non-ideal dynamic response of rectangular pulp stock chests. This
abnormality was attributed to improper chest design and the non-Newtonian rheology of pulp
suspension, which has a yield stress promoting the formation of caverns (regions of active
motion) around impellers. Under certain flow configurations, these caverns can induce
undesirable flow, like bypassing (part of the feed flows directly to the exit without entering the
cavern) and dead zones (stagnant regions outside the cavern) which degrade mixing and reduce
the ability of the chest to attenuate variability. Poor chest performance can lead to unstable
downstream processes and low product quality, resulting in high production costs.
Pulp stock chests are typically rectangular or cylindrical. Their design is based on the use
of semi-empirical methods and experience with a design strategy of matching the momentum
generated by an impeller with that required to provide complete suspension motion across the
surface of the chest (Yackel, 1990). This method does not specifically take the locations of pulp
inlet and outlets into consideration, although heuristics are provided. Laboratory studies using
tracer analyses, ultrasonic Doppler velocimetry (UDV) and electrical resistance tomography
(ERT) have investigated mixing and suspension flow through agitated stock chests. Ein-
77
Mozaffari et al. (2005) used dynamic test results to show that cavern size and its position relative
to inlet and outlet locations played an important role in determining the amount of feed entering
the cavern which affected the mixing dynamics of the chests. Bypassing was minimized when
the stock outlet was located within the cavern zone. In addition, Ein-Mozaffari et al. (2007)
measured the flow characteristics (impeller flow number and pumping rate) of a Maxflo impeller
and located the cavern using the velocity profiles measured by UDV. Hui et al. (2009) applied
ERT and UDV to measure the cavern volume in a cylindrical chest. A model was developed to
estimate the cavern size using a force balance with consideration of interaction between the
Although experimental studies provide some information about the mixing dynamics
inside stock chests, obtaining this data is very difficult due to the opaqueness of the suspensions.
Computational fluid dynamics (CFD) has been used to obtain information about the velocity
fields inside these and other vessels, although there are difficulties associated with modeling the
suspension and validating the simulations made. Many studies have shown that CFD can predict
the velocity distribution, evaluate industrial stirred equipment, and optimize impeller design
(Armenante et al., 1997; Bhattacharya and Kresta, 2002; Paul et. al., 2004; Li et al., 2005;
Ankamma Rao and Sivashanmugan, 2010). For pulp suspensions, a number of studies using
CFD have been carried out to investigate the flow in agitated pulp chests. Bakker and Fasano
(1993) applied CFD and used the rheological data from Gullichsen and Harkonen (1981) for pulp
suspension to model the flow in a rectangular pulp chest with a side-entering impeller. The
numerically simulated results were found to agree with visual observation of the suspension
surface. In the study of the impact of pulp suspension rheology on flow field in an agitated pilot
tank with a jet nozzle mixer, Wikstrom and Rasmuson (1998a) compared the CFD-calculations
78
with the measured results and concluded that a more sophisticated rheology model should be
used instead of the Bingham model because of the increasing deviation between the computed
and measured flow field as the distance from the impeller increased. Ford et al. (2006) developed
a CFD model of a rectangular pulp mixing chest and used the Bingham plastic model as the pulp
suspension rheology. The model captured the mixing dynamics of the chest fairly well but it
overestimated the extent of mixing in the bypassing flow, especially in flow situations with
significant bypassing. This deviation was attributed to the inability of the Bingham model to
fully describe the suspension rheology. In the CFD modeling of a cylindrical tank agitated with
an axial-flow impeller, Saeed et al. (2007) treated pulp suspension as a Herschel-Bulkley fluid.
The CFD simulations picked up the features of the flow field and the computed velocities agreed
satisfactorily with the measured results. The discrepancy between the computed and measured
velocities was believed to be due to the suspension rheology. Bhattacharya et al. (2010) modeled
two industrial pulp stock chests using CFD and described the pulp rheology as a modified
Herschel-Bulkley model. The computed dynamic response of the chests was found to agree
reasonably well with the experimental measurements and the simulated flow fields provided
insight about cavern formation, stagnant regions and bypassing zones created in the chests.
However, these studies have not provided a detailed examination of the cavern flow, which is
important in determining the non-ideal flow in pulp mixing chests. The extent of cavern growth
could determine the degree of bypassing and the size of dead zone in agitated chests.
In this chapter, the commercial CFD software (FLUENT 6.2, Fluent Inc, Lebanon, NH)
was used to model a cylindrical chest equipped with a scaled version of a commercial axial-flow
Maxflo impeller (Chemineer Inc., Dayton, OH) in the standard side-entering configuration for
simulating the mixing of two pulp suspensions (hardwood and softwood) over a range of
79
impeller rotational speeds (N) and suspension concentrations (Cm). Ford et al. (2006) have
equipped with a side-entering agitator for a range of operating conditions. This similar approach
was used to simulate the pulp mixing in the cylindrical chest. Two non-Newtonian fluid models
were used to describe suspension rheology (based on rheological measurements). The simulation
results were then used to calculate the cavern size in batch mixing mode and to estimate the
dynamic response of the chest in continuous operation. The computed results were then
compared with the experimental results obtained in the studies of Hui et al. (2010). By
determining the degree to which the CFD model can predict experimental data, its usefulness as
a tool for chest design can be evaluated. If the evaluation is positive, the CFD model can be used
for optimization and economical design of pulp chests for a range of objectives, including
A number of experimental methods were used to study the mixing behavior of pulp stock
chests. For batch operation, electrical resistance tomography (ERT) and ultrasonic Doppler
velocimetry (UDV) were applied to study the cavern shape and size. ERT is a non-invasive
imaging technique that can image the conductivity distribution inside a region of interest using
voltage measurements made through a series of electrodes placed at the chest periphery. With
the addition of a conductive tracer to the cavern, a P2000 ERT tomography instrument (ITS,
Manchester, UK) was used to locate the region of higher conductivity and thus the cavern shape
and size (Hui et al., 2009). The measurement error of the cavern boundary is typically about 5-
80
10% of the chest diameter. Also, the cavern results obtained from ERT were validated by visual
observations of motion at the vessel walls and the suspension surface. Another method used to
study caverns was UDV, a real-time fluid flow measurement based on the principle of Doppler
frequency shift. An ultrasonic pulse is emitted from the UDV probe (DOP2000, Signal
Processing, Switzerland) and the echoes reflected from the particles in the fluid are used to
determine the velocity profile along the measurement line with the velocity direction parallel to
the pulse path. The locations where the suspension velocity falls to zero are the cavern
boundaries. The measurement accuracy of UDV is about 2.4% for velocity (Xu, 2003).
For continuous operation, a dynamic mixing model (Fig. 4.1) was used to quantify the
mixing dynamics in the cylindrical chest. f is the fraction of pulp stock that bypasses the mixed
zone with minimal mixing and (1-f) is the remaining stock entering the mixed zone. 1 and 2 are
the time constants for each portion, respectively and Td is the transport delay due to the flow
delay in the process piping prior to and following the chest. The model parameters were
determined by adding a conductive tracer (saline solution) to the pulp feed using a pseudo-
random binary signal (PRBS) to control tracer addition and excite the chest dynamics. The
conductivity of the suspension entering and leaving the chest was measured and this data set was
The configuration of the mixing domain, identical to the lab-scale cylindrical chest was modeled
using GAMBIT, a geometry and grid generation software bundled with FLUENT (Figure 5.1(a)).
81
It was used for both batch and continuous-flow systems; except that for the dynamic simulations,
an inlet was created at the top of the surface and an outlet at 45 with respect to the impeller as
shown in Fig. 5.1(b). The flow in the mixing chest is not steady in an inertial frame of reference
because the impeller blades sweep the domain periodically. However, since no baffles or stators
were involved in the mixing chest, the flow could be viewed as steady relative to the rotating
(non-inertial) frame (Ford, 2004). Thus, a multiple reference frame approach, including a
rotating frame and a stationary frame, was used to model the mixing domain. The rotating frame
is a cylindrical region comprising the impeller with the impeller shaft acting as its axis. The rest
of the vessel is in the stationary frame. To obtain reliable computational results, it is desirable to
balance a fine mesh structure with the need of having a reasonable computational time. A mesh
of 700,000 cells took 2 to 3 days to converge (on a Pentium(R) 4 CPU 3.00GHz) while meshes
of 200,000-300,000 cells took about one day. To compromise between the running time and
computational reliability, a finer mesh was only used in the rotating domain and a check of mesh
For the batch simulation, the mixing domain was divided into six blocks for meshing.
Unstructured mesh, using tetrahedral elements connected one other, was applied to the rotating
domain and its neighbouring block because of their irregular geometries. To capture flow details
near the impeller, a finer mesh was created in the vicinity of the impeller blades. A conformal
mesh (having element nodes at identical locations at the block interfaces) was created using a
size function in the rotating domain to allow mesh elements to grow slowly as a function of the
distance from the impeller hub and blades so that the fine mesh in the rotating domain were
logically linked to the surrounding stationary domain which was not highly meshed (Ford, 2004).
The other four regular blocks were discretized using a structured mesh, comprising hexahedral
82
elements logically connected. To ensure the simulated results were not dependent on the grid
density, three different mesh sizes (198,084, 407,592 and 633,013 cells) were used to check grid
independence under the same simulation conditions (i.e., N = 425 rpm and Cm = 3% softwood).
Stationary frame
a)
Rotating frame
Impeller
Impeller
b) Inlet
Inlet Stationary frame
Impeller
Outlet
Outlet
Figure 5.1 Experimental cylindrical chest and computational domain for (a) batch and (b)
continuous-flow mixing.
A pulp suspension is a heterogeneous fluid (fibres and water) that is not easily modeled. As it
possesses a yield stress, the first comparable model that can be used to describe its rheology is
the Bingham model. Wikstrom and Rasmuson (1998) used this approach to model the agitation
83
of pulp suspensions with a jet nozzle agitator and found that the flow field far away from the
boundary conditions was underestimated due to the shear-thinning property of pulp suspension.
stress, can also be used to model the pulp suspension rheology. Both models are expressed in
Fluent, as
n y n
y K1
o
(5.1)
where y, K1, n, o and are yield stress threshold, consistency index, power-law index, yielding
viscosity and strain rate respectively. All these parameters depend on the mass concentrations
(Cm) of pulp suspensions and can be determined from rheological measurements of suspensions.
The rheological parameters for the Herschel-Bulkley model were determined by fitting the flow
curves of pulp suspensions in the study of Gomez et al. (2010) to the model equation. For the
Bingham approximation, the pulp suspension is described with K1 = 0.001kg/ms (equal to the
viscosity of water) and n = 1 since the pulp and paper industry often uses the water viscosity to
design equipment where the suspension is exposed to high shear rates (Ford, 2004). Both
rheological models were applied in the CFD simulations of pulp agitation. The rheological
parameters for hardwood pulp (Cm = 2%) and softwood pulp (Cm = 3%) are summarized in Table
5.1.
84
In the rotating domain, an angular velocity was specified equal to the impeller rotation rate used
in the experimental test. The rotation axis was set at the centre of the impeller hub, with the
positive Y direction extending into the vessel (Figure 5.1). The impeller hub and blades were
modeled as rotating walls and moved with zero relative velocity with respect to the rotating
frame. The continuity and momentum equations governing the rotating domain are given by:
vr S m (5.2)
t
where is the fluid density, t is time, vr is the relative velocity and Sm is the mass added to the
continuous phase from the dispersed second phase and any user-defined sources.
t
v vr v v p g F (5.3)
where v is the absolute velocity, p is the static pressure, is the stress tensor, g is the
gravitational body force and F is the external body forces. The relative velocity, vr , is related to
the absolute velocity, v , using the following equation:
85
vr v r (5.4)
where is the angular velocity of the rotating frame, r is the position vector in the rotating
frame.
In the stationary frames, the chest walls were set with a velocity of zero in the absolute
reference frame. The upper surface of the suspension was treated as a stationary wall with zero
shear stress. Since no rotation is involved in the stationary domain, the continuity and
v S m (5.5)
t
t
v vv p g F (5.6)
At the boundaries between the two domains, the continuity of absolute velocity is enforced to
provide the correct neighbor values of velocity for the corresponding domain.
In the pulp mixing experiments, the observed suspension flow was generally laminar, so
the continuity and momentum equations were solved in laminar regime using a second-order
upwind scheme. Near the impeller, the flow might be turbulent, but the fluctuation velocities
would fade out quickly in the regions outside of the impeller zone because of the fibrous
structure of the heterogeneous suspensions. Also, the non-Newtonian Reynolds number used by
Gibbon et al. (1962) and Blansinski et al. (1972) for mixing pulp suspensions was calculated and
found to be below 103 (i.e., in the laminar regime), so treating the suspension flow as laminar in
When continuous operation was simulated, inlet and outlet pipes were added to the chest
geometry. The lengths (L) and diameters (Dp) of the pipes in the simulation followed the
dimensions of the experimental set-up, i.e., L = 215 cm and Dp = 5cm for the inlet and L =
257cm and Dp = 5cm for the outlet. The inlet was defined in the absolute frame as having a flat
profile with a velocity magnitude ( v 4Q D p ). Since L/Dp for the outlet was greater than 10,
2
the exit was defined as an outflow boundary condition under the fully-developed flow
assumption. The boundary conditions at the outflow boundary are a zero diffusion flux for all
flow variables and an overall mass balance correction. The zero diffusion flux condition means
that the conditions of the outflow plane are extrapolated from within the domain and have no
Several researchers have determined the cavern boundary by assuming that the shear stress at the
cavern surface is equal to the yield stress of the suspension (y) (Solomon et al., 1981; Elson,
1990; Amanullah et al., 1998; Hui et al., 2009). Thus, this criterion was used to determine the
cavern boundary in the CFD simulations. Since the shear stress was not given in Fluent, the
strain rate at the yield stress (y) was used to define the cavern boundary
y
y (5.7)
o
A User Defined Function (UDF) was used to determine those cells included in the cavern criteria
The flow results calculated from the CFD simulations were also used to determine virtual
dynamic responses to the experiments conducted by Hui et al. (2010). Since the properties of the
87
tracer and the modeled pulp suspension are similar, the converged results for the flow field of the
suspension were used to perform transient species calculation for the tracer (Ford, 2004). Here
the species conservation equation was solved in time-dependent form, discretized in both space
and time. A UDF (Appendix D) specifying the tracer injection sequence at the inlet was linked to
the FLUENT solver for the continuous-flow system simulation and the concentration of the
tracer „measured‟ at the outlet. The inlet-outlet concentration profiles were then analyzed using
the dynamic model of Soltanzadeh et al. (2009) to quantify the mixing and compare them
directly with the experimental results. The model parameters identified included the transport
delay (Td) arising mainly due to flow in the process piping before and after the chest, the
bypassing (channeling) fraction (f) of suspension which flows through the chest with minimal
mixing, and the time constant for the mixing zone (2). The data were divided into two parts: one
for model parameter estimation and one for parameter validation. The extent of agreement
between the model output and the simulated data was expressed using
1 n y computedi y simulatedi
% fit 1 100% (5.8)
n i 1 y simulatedi mean( y simulated )
The time constant (2) was then used to estimate the effectively mixed volume (Vmix) in the chest
The effect of the meshing scheme used in the CFD calculations was first examined. The velocity
profiles calculated at two locations in front of the impeller (as shown in Figure 5.2) for the three
88
meshing schemes are given in Figure 5.3. There is no significant difference among the three
schemes. Thus, the lowest density mesh was used for the simulations because of the shorter
computer running time required. Table 5.2 shows cavern volumes determined using the three
meshes described earlier, with cavern size decreasing slightly as the cell density increases (and
consequently the average cell size decreases). The volume difference is small, less than 2.5%,
and is attributed to the finer resolution of the cavern as cell density increases.
CFD simulations were made for a range of operating conditions (two pulp types, a range
of suspension concentrations and a number of impeller speeds). Figures 5.4 and 5.5 compare
cavern shape and volume measured using ERT and computed using CFD for Cm = 2% hardwood
and Cm = 3% softwood at three different impeller speeds. For the hardwood pulp (Figure 5.4),
when the stock height (Z) was equal to the chest diameter (T), all the simulations were unable to
predict the suspension surface motion observed visually during the experiments. The possible
reason for this deviation could be that the yield stress may not be consistent throughout the pulp
suspension during agitation. Since pulp suspension is a two-phase fluid, the pulp fibres would be
redistributed in the suspension by the vigorous action of the impeller in actual situation. Fewer
fibres would be in the region around the impeller and so the yield stress in the cavern could be
slightly lower than the surrounding, promoting the cavern growth. However, in CFD simulations,
the pulp suspension is assumed to be homogenous, with no localized difference in yield stress
and so the predicted cavern volume could be smaller. The best way to justify this explanation is
to repeat the experiments using a homogeneous fluid with known rheology (possibly similar to
the properties of pulp suspension) and then compare with the results of heterogeneous pulp
suspensions. Also, in actual mixing situation, the top surface of the agitated chest was in motion
and open to the air, allowing air to enter the chest and promoting the cavern growth which
89
enhances the surface motion. However, in CFD simulations, the top surface was modeled as a
flat stationary wall with no air entering and so the actual situation may not be easily simulated.
The two rheology models examined did not make a significant difference in the cavern
shape determined for hardwood pulp. Compared with the ERT results, the average percentage
differences in cavern volumes for both models are also similar: 53% for the Herschel-Bulkley
model and 49% for the Bingham model. However, the Bingham model predicted the cavern
volume better at higher impeller speeds. When the impeller speed was increased to the situation
close to complete mixing (i.e., 397rpm - experimental value for 2% hardwood pulp), the
Bingham plastic model using the viscosity of water for K1 and n = 1 gave the cavern volume
more comparable to the experimental result. This observation agrees with the study of Gullichsen
and Harkonen (1981), showing that pulp suspensions behave more like water at high rotational
speeds.
Chest wall
Impeller
(a) (b)
Locations where velocity profiles are calculated
at z = 12cm (measured from the chest bottom)
Figure 5.2 Locations where velocity samples were determined for comparison of three
different meshes.
90
a)
low mesh
1.0 medium mesh
high mesh
0.6
0.4
0.2
0.0
-0.20 -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20
Distance from the shaft centre line (m)
b)
0.00035
0.00030
low mesh
medium mesh
Velocity magnitude (m/s)
0.00020
0.00015
0.00010
0.00005
0.00000
-0.20 -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20
Distance from the shaft centre line (m)
Figure 5.3 Velocity profiles calculated for softwood Cm = 3% at two locations (a) and (b) in
front of the impeller (as shown in Figure 2) for three different meshing schemes at N =
425rpm.
91
Table 5.2 Grid independence study (N = 425rpm, 3% softwood pulp (K1 = 135, n = 0.21, y =
155Pa)
No. of element cells Vc (dm3)
198,084 9.2
407,592 9.0
633,013 8.9
Cavern
shape
(150rpm)
Cavern
shape
(200rpm)
Cavern
shape
(250rpm)
Figure 5.4 Comparison of cavern size obtained from ERT and numerical simulations for a
Cm = 2% hardwood (Z/T = 1.0).
92
The CFD simulations for softwood pulp (Cm = 3%) are similar to those made for the
hardwood pulp. The two rheology models did not make significant differences in cavern shape
although mild surface motion was predicted in the cases of Herschel-Bulkley model, which also
provided cavern volumes closer to the experimental results, with an average of 13% lower than
the ERT results (Figure 5.5). Since the fibres of softwood pulp are longer than those of hardwood
pulp, they easily tangle together, especially at low impeller speeds, but they gradually align more
substantially than shorter hardwood fibres in the direction of increasing shear, producing less
resistance to flow. This makes softwood pulp behave more like a shear-thinning fluid, which can
be further supported by the value of the power-law index (n = 0.21 for Cm = 3% softwood
whereas 0.328 for Cm = 2% hardwood with n < 1 for shear-thinning fluid and n = 1 for
Newtonian fluid). Also, the impeller speeds used in the comparison are not close to the point of
complete mixing (i.e., 1020rpm for 3% softwood pulp) and so it is unlikely that the suspension
would behave like water (Bingham model with n = 1). With favorably tangled long-fibred
structure, the 3% softwood pulp would be better described by Herschel-Bulkley model in mixing
Figure 5.6 shows the flow field computed at the measurement location along the cross-
section of the impeller hub (Fig. 5.2) for the simulations of two pulp types using their most
comparable rheology models mentioned above. Both simulated flow fields of hardwood and
softwood pulps were found to match the flows expected for an axial-flow impeller. The impeller
discharges the suspension along the rotation axis and up to the surface for return to the suction
side of the impeller, creating a large upper circulation loop which was observed during
experiments. The smaller loop generated under the impeller is also shown in the two simulations.
93
Cavern
shape
(425rpm)
Cavern
shape
(475rpm)
Cavern
shape
(525rpm)
Figure 5.5 Comparison of cavern size obtained from ERT and numerical simulations for a
Cm = 3% softwood (Z/T = 0.8).
94
a)
b)
Figure 5.6 Flow fields of batch mixing for a) Cm = 2% hardwood at N = 200rpm using
Bingham plastic model and b) Cm = 3% softwood at N = 475 rpm using Herschel-bulkley
model.
95
In addition to the observed flow from the sides of the vessel, the velocity profiles
calculated from CFD simulations were compared with the ones measured using UDV at the
specified location shown in Figure 5.2 at three different heights (Figure 5.7). The velocity profile
along the measurement line (the dotted arrow lines) was measured with the velocity direction
parallel to the pulse path. The positive velocity means the velocity direction is away from the
UDV probe. Figures 5.8 and 5.9 compare the measured velocities with the ones calculated using
two rheology models for Cm = 2% hardwood at two impeller speeds. The sudden irregular ups
and downs shown on the UDV-measured velocity curves are possibly due to the interference of
bubbles introduced by air entrapment at the free suspension surface during agitation. The bubble
movement and coalescence-breakup behavior induce turbulent fluctuations to the flow field in
both direction and amplitude, and this would distort the ultrasonic signals and affect the velocity
measurement (Bouillard et al., 2001; Wang et al. 2003). Since UDV measured the velocity of
pulp fibres whereas CFD computed the fluid velocity, it would be expected that the computed
values would be slightly different from the measured values. At N = 200 rpm, neither model can
exactly predict the measured velocities but the velocity profiles obtained from the Bingham
plastic model are closer to the measurements, e.g., at position 2, the velocity profile computed
using the Bingham model follows the direction change of the measured profile (from positive to
negative) at about +0.05m from the chest centre. At N = 250rpm, the velocity profile shape of
the Bingham model is very close to the measured one. The peaks and dips of both profiles occur
almost at the same locations and so it is obvious that the Bingham model would be preferred for
a)
Impeller
UDV probe
Line of the impeller
shaft
b)
UDV probe
Position 3
Position 2
Position 1
Figure 5.7 a) Top view and b) side view of probe locations for UDV measurements. The
arrows show the direction of the sonic emissions at three heights (12.5mm, 92.5mm and
252.5mm measured from the chest bottom).
97
a) Position 1 0.04
0.03 UDV
Herschel-Bulkley
0.02 Bingham plastic
0.01
Velocity (m/s)
0.00
-0.01
-0.02
-0.03
-0.04
-0.2 -0.1 0.0 0.1 0.2
Distance from the chest centre (m)
b) Position 2
0.05
0.04 UDV
Herschel -Bulkley
0.03 Bingham plastic
0.02
Velocity (m/s)
0.01
0.00
-0.01
-0.02
-0.03
-0.2 -0.1 0.0 0.1 0.2
Distance from the chest centre (m)
c) Position 3
0.00
-0.02
UDV
Herschel-Bulkley
Velocity (m/s)
Bingham plastic
-0.04
-0.06
-0.08
Figure 5.8 Comparison of measured and computed velocity profiles at three positions for a
Cm = 2% hardwood at N = 200rpm. Positive velocity means the velocity direction is away
from the probe.
98
a) Position 1
0.10
0.08 UDV
Herschel-Bulkley
0.06 Bingham plastic
Velocity (m/s)
0.04
0.02
0.00
-0.02
-0.04
-0.2 -0.1 0.0 0.1 0.2
Distance from the chest centre (m)
b) Position 2
0.20
0.15
UDV
Herschel-Bulkley
0.10 Bingham plastic
Velocity (m/s)
0.05
0.00
-0.05
-0.10
-0.15
-0.2 -0.1 0.0 0.1 0.2
Distance from the chest centre (m)
c) Position 3 0.01
0.00
-0.01
-0.02
Velocity (m/s)
-0.03
-0.04
UDV
-0.05
Herschel-Bulkley
Bingham plastic
-0.06
-0.07
Figure 5.9 Comparison of measured and computed velocity profiles at three positions for a
Cm = 2% hardwood at N = 250rpm. Positive velocity means the velocity direction is away
from the probe.
99
CFD was also used to predict the point where the suspension was in complete motion
throughout the mixing chest (i.e., where the cavern volume equaled the suspension volume). The
impeller speeds for complete mixing predicted by CFD for both pulps are about twice the
experimental one. This deviation may be due to the complex pulp suspension rheology, which is
not easily described by CFD, e.g., floc formation easing the flow of pulp suspension. Also, for
both the hardwood and softwood pulps, the cavern volumes determined using CFD and ERT
increased with impeller speed (Fig. 5.10). The increasing trend of cavern volume observed in
ERT results was predicted satisfactorily for the hardwood pulp but not for the softwood pulp,
which has CFD results increasing more steadily and slowly than those of ERT. The possible
reason for this difference is the difficulty of using CFD to describe the complex behavior of pulp
fibres in agitation, i.e., softwood pulp forms flocs more readily than hardwood pulp in actual
situation. With the same amount of fibres, the total surface area of flocs is smaller than the sum
of individual fibre surface area. Thus, the resistance to flow in softwood pulp suspension
becomes lower than the expected because flocs would move more easily than long pulp fibres. In
CFD, no simulation of floc formation was involved and so the resistance to flow would not be
changed, leading to relatively smaller cavern volume when compared with the experimental
results. In addition, during experiments, it was observed that air entrapment was significant in
agitation of softwood pulp. This would enhance the cavern growth, which cannot be easily
simulated in CFD.
100
a) 50
45 3
Suspension volume = 43.4dm
40
35
30
Vc ( dm )
3
25
20
ERT
15
CFD
10
0
100 200 300 400 500 600
N (rpm)
b)
40
3
Suspension volume = 34.7dm
35
30
25
Vc ( dm )
3
20
15
ERT
CFD
10
Figures 5.11 and 5.12 compare the responses simulated from CFD with those obtained
from experiments for dynamic tests. It seems that both the CFD and experimental results are
reasonably similar and increasingly close to each other with increasing impeller speed. Both
simulations and experiments show no bypassing was detected in the studied mixing situations.
Figure 5.13 shows the simulated path lines of particles released into the chest using CFD. It is
101
clear that the particles enter the region (cavern) around the impeller before exiting the chest and
this agrees with the experimental result of no channeling. Tables 5.3 and 5.4 shows the dynamic
parameters (Td – time delay and 2 – time constant for mixing zone) obtained from CFD and
experimental responses. For both pulps, Td predicted by CFD is larger than that of dynamic test,
especially for softwood pulp. The possible explanation for this is the air entrapped in the pipe
flow in actual situation. During experiments, the feed tank was open and agitated by a top-entry
mixer. Air could then be drawn into the pulp suspension and flowed with the suspension. Thus,
besides pulp suspension, there was air in the piping. This air occupancy would create “shortcut”
in the pipes and shorten the time delay. However, no such phenomenon was considered in CFD
simulations and so the CFD Td was found to be larger than the experimental one. Since air
entrapment is more significant for softwood pulp than for hardwood pulp as observed during
For the 2 of 2% hardwood pulp, excluding the value at low impeller speed (i.e., 125rpm),
the average difference between simulated and experimental values is about 16%. This result is in
accord with that of batch mixing in Figure 5.10a where CFD predicted fairly well at high speeds.
For 3% softwood pulp, the difference in 2 becomes higher as the impeller speed increases. The
predicted time constants of the last two impeller speeds (i.e., 1020 and 1100rpm), which show
complete mixing in experiments, are much lower than the experimental ones. The reason is that
under CFD simulations, these two speeds are not at complete mixing, as shown in Figure 5.10.
The impeller speed calculated from CFD for complete mixing of 3% softwood pulp is over 2300
rpm and so that‟s why the predicted time constants are lower than the experimental ones. Figure
5.14 also shows the effectively mixed volume calculated from 2 for Cm = 2% hardwood and Cm
= 3% softwood. It is clear that CFD predicted better for hardwood pulp than softwood pulp and
102
the reason would possibly be the higher tendency of long-fibred softwood pulp to form flocs,
which offer less resistance to flow and this phenomenon could not be easily simulated by CFD.
This also agrees with the CFD results of complete mixing in batch operation (Figure 5.10),
showing that CFD approximated better for hardwood than softwood. In addition, the percentages
of fitting for CFD are generally higher than those for dynamic tests because unlike the
experimental outputs, the simulated responses generally do not contain any noises that would
To conclude the study, the possible reasons for the difference between experimental
results and numerical simulations are the heterogeneous nature of pulp suspensions and the air
entrapment during experiments. To verify these explanations, this study should be repeated with
a homogeneous fluid with known rheology (possibly similar to the properties of pulp
suspensions) in a closed system. The approach using glycerin solution in the study of the flow
field in a rectangular vessel by Gomez et al. (2010) can be used and then the outcome can be
compared with the results of this study. In addition, their study showed very good agreement
between the experimental and computational results. Since the CFD modeling in this study is
similar to that of Gomez‟s study, the deviation between the modeling and experimental results is
most likely due to the special rheology of pulp suspension, which is not easy to be formulated
Table 5.3 Comparison of results from CFD and dynamic tests for Cm = 2% hardwood (Q =
14L/min., Z/T = 0.8)
Td (s) 2 (s) Percentage of fitting (%)
N (rpm) CFD Dynamic CFD Dynamic CFD Dynamic test
test test (Eqn (5.8)) (Eqn. (4.1))
125 65.0 58.8 124 47 94 89
250 55.0 40.0 116 106 95 85
380 49.0 34.8 178 155 90 91
460 41.6 38.7 177 142 89 91
Table 5.4 Comparison of results from CFD and dynamic tests for Cm = 3% softwood (Q =
14L/min., Z/T = 0.8)
Td (s) 2 (s) Percentage of fitting (%)
N (rpm) CFD Dynamic CFD Dynamic CFD Dynamic test
test test (Eqn.(5.8)) (Eqn.(4.1))
425 60.0 41.4 70 69 88 86
525 56.0 40.0 72 97 85 82
1020 50.0 36.2 68 160 90 91
1100 55.0 40.3 75 156 83 80
104
12
12.0
11.5
11
11.0
Conductivity (mS)
Conductivity (mS)
10.5
10
10.0
9 9.5
9.0 CFD
CFD Experimenta
Experimental
8 8.5
0 240 480 720 960 1200 1440 1680 1920 2160 2400 0 240 480 720 960 1200 1440 1680 1920 2160 2400
Flow time (s) Flow time (s)
125rpm 250rpm
7.5 10
7.0
9
6.5
Conductivity (mS)
Conductivity (mS)
6.0
8
5.5
5.0
7
CFD CFD
4.5
Experimental Experimental
4.0 6
0 240 480 720 960 1200 1440 1680 1920 2160 2400 0 240 480 720 960 1200 1440 1680 1920 2160 2400
Flow time (s) Flow time (s)
380rpm 460rpm
Figure 5.11 Experimental and computed dynamic responses for Cm = 2% hardwood pulp.
105
3.5 5.0
3.0 4.5
2.5 4.0
Conductivity (mS)
Conductivity (mS)
2.0 3.5
1.5 3.0
1.0 2.5
CFD CFD
0.5 Experimental 2.0 Experimental
0 240 480 720 960 1200 1440 1680 1920 2160 2400 0 240 480 720 960 1200 1440 1680 1920 2160 2400
Flow time (s) Flow time (s)
425rpm 525rpm
5.5
2.5
5.0
2.0
4.5
Conductivity (mS)
Conductivity (mS)
1.5
4.0
1.0
3.5
0.5 3.0
CFD CFD
Experimental Experimental
0.0 2.5
0 240 480 720 960 1200 1440 1680 1920 2160 2400 0 240 480 720 960 1200 1440 1680 1920 2160 2400
Flow time (s) Flow time (s)
1020rpm 1100rpm
Figure 5.12 Experimental and computed dynamic responses for Cm = 3% softwood pulp.
106
Inlet
Outlet
Figure 5.13 Path lines of particles (each particle is represented by a different color)
simulated by CFD using Cm = 2% hardwood pulp at N = 250rpm and Q = 14L/min.
107
a) 45
40
35
30
25
Vmix ( dm )
3
20
15
10
Dynamic test
CFD
5
0
100 200 300 400 500
N (rpm)
b) 40
35
30
25
Vmix ( dm )
3
20
15
10
Dynamic test
CFD
5
0
400 600 800 1000 1200
N (rpm)
Figure 5.14 Experimental and computed determination of mixed volume for (a) Cm = 2%
hardwood pulp and (b) Cm = 3% softwood pulp in continuous operation.
108
5.5 Summary
A CFD model of a lab-scale cylindrical mixing chest was developed using FLUENT 6.2. To
verify the ability of the CFD model for flow estimation in agitated pulp stock chests, a number of
numerical simulations were carried out to compare with the experimental results in terms of
cavern volumes and dynamic model parameters. Although the CFD model cannot precisely
depict the cavern shape and the surface motion, it can predict the increasing trend of cavern size
with impeller speed. Based on the cavern comparison, it seemed that Bingham plastic model
would be preferred for the hardwood pulp at impeller speeds close to complete mixing whereas
Herschel-Bulkley model predicted better for the softwood pulp at low impeller speeds. The
suitability of the Bingham model for the hardwood pulp was further supported by the UDV
measurement. The reason why the CFD model cannot fully predict the actual mixing situations is
that the actual behavior of pulp fibres, like floc formation of softwood pulp in agitation, and the
air entrapment during agitation cannot be easily described by CFD. The best way to justify the
explanation is to repeat the mixing experiments using a homogeneous fluid with no known
rheology similar to that of pulp suspensions in a closed system and then compare the outcome
with the results of this study. Actually, a major difficulty in using CFD for simulation of pulp
mixing is the lack of good description of rheology in CFD code. However, the agreement of
simulated results of particle path lines with the experiment result of no passing suggests that the
CFD model can provide valuable information about the flow patterns in agitated pulp chests.
109
Since pulp suspensions possess a complex rheology, it is not easy to achieve an effective
agitation in cylindrical chests. Caverns form when the chests are not completely agitated and this
may lead to bypassing and dead zone formation. In order to avoid undesirable flows affecting
chest performance, a study of pulp mixing in batch and continuous operations was carried out.
Electrical resistance tomography (ERT) and Ultrasonic Doppler velocimetry (UDV) were used to
study the cavern shape and size in batch operation. Both methods gave comparable results but
ERT was preferred due to less time required to acquire data. The cavern shape was best
described as a truncated right-circular cylinder and its volume was found as expected to increase
with impeller speed but the increase was non-uniform due to interaction between the cavern and
There are models for estimating cavern volumes in fluids with yield stresses but without
including the interaction between the cavern and the vessel walls. Like the situations in the study,
they do not describe the experimental results well. Thus, based on the observed cavern shape in
ERT imaging, a model was developed to estimate the cavern volume by balancing the force
provided by the impeller with the resistive force at the cavern boundaries. Although the proposed
model could not accurately calculate the cavern volume, it could predict the increasing trend in
In continuous operation, dynamic tests were used to characterize the mixing quality in a
cylindrical chest. A model including a mixed region (cavern) and the possibility of bypassing
110
was shown to precisely describe the mixing dynamics of the chest because the time delay
estimated using the model agreed closely with the calculated result based on flow and system
geometry. The measured bypassing was found to be insignificant, indicating that the flow
configuration used in the study (with the exit within the cavern, close to and below the impeller)
is effective for avoiding feed bypassing. Also, ERT further confirmed the presence of cavern
when the chest was not fully agitated in continuous operation with feed and the cavern volume
increased with impeller speed. In addition, the power for complete agitation in a cylindrical chest
Besides the experimental study of pulp mixing in a cylindrical chest, the ability of
computational fluid dynamics (CFD) to estimate the flow in agitated pulp chests was examined.
Numerical simulations were carried out to compare with the experimental results. Owing to the
complex properties of pulp suspension, it is not easy to describe its unique behavior like floc
formation in agitated chests using CFD. The application of both Bingham and Herschel-Bulkley
models could not exactly portray the cavern, but the increasing trend of cavern volume with the
impeller speed was still predicted. The CFD simulations also picked up the features of the flow
field measured by UDV at the high impeller speed. The calculated path lines of the simulated
particle feed matched with the experimental results of no bypassing. Therefore, in spite of its
limitation to model pulp suspensions, the CFD model can still provide useful flow information in
The outcome of this research can aid the design of industrial cylindrical chests. The
cavern model can estimate the mixing volume in proposed agitated pulp chests, aiding the sizing
of chests. The dynamic test study provides insight about the positions of in/out piping of new
chests, e.g., the outlet should be placed close to the cavern, to avoid non-ideal flows which affect
111
the chest performance. Besides new chests, this test method can also be used to estimate the
performance of existing pulp chests, trying to improve or enhance their mixing efficiency. In
addition, the flow configuration of the proposed chest design can be checked using CFD. Saving
the need of building a lab-scale chest and doing experiments, the CFD model can provide
information about the flow pattern in the proposed chest, helping chest design and optimization.
The experimental results about the cavern geometry in cylindrical vessels can provide
insights for the design of chest geometry to facilitate pulp mixing, e.g., installation of
fillets is needed to minimize dead zone. The proposed cavern model could be used to
estimate the mixing volume in existing and proposed pulp chests to see whether they meet
the requirement.
The results of dynamic tests show that the flow configuration with the exit installed near
the cavern can avoid bypassing. This information confirms the importance of exit location
The results of the CFD model can provide valuable information about the flow pattern,
e.g., the presence of bypassing, inside the existing and proposed chests. This also helps
the design of flow configuration of agitated pulp chests, like the locations of the impeller,
The results of this study provided some insights for future consideration as follows:
The presence of dead zone under incomplete mixing implies the possibility of disturbance
to normal operation (e.g., the oozing of undesirable materials from the dead zone into the
cavern). The installation of fillets can avoid this hazard and its effect on the cavern
Based on the experimental results of ERT and UDV, a model for determining the cavern
volume in a cylindrical chest was developed. Similar modeling should be carried out in
rectangular chests because they are also widely used in the industry. This model can be
For the agitation of softwood pulp, entrapment of air into the suspension is significant at
high impeller speeds and this will affect the cavern size. Thus, the effect of air entrapment
Owing to the poor knowledge of pulp suspension rheology, it was proposed that the
heterogeneous structure of pulp suspensions could cause the discrepancies between the
experimental results and CFD simulations. To validate this proposition, a study of the
same mixing system using a homogeneous fluid with known rheology should be carried
out.
In this study, the simulated results of CFD in cavern size are not in complete agreement
with the experimental results and one of the possible reasons is the incomplete quantitative
description of pulp suspension rheology, e.g., the interactions of flocs. A better numerical
description of pulp suspension rheology, possibly in two stages (MACRO and MICRO),
Besides cylindrical chests, reduced bottom chests are frequently used in pulp and paper
industry. They do not always perform satisfactorily because of the undesirable flows such
as channeling induced by the cavern. Thus, cavern formation and dynamic behavior of
BIBLIOGRAPHY
Amanullah, A., S.A. Hjorth and A.W. Nienow, “A New Mathematical Model to Predict Cavern
Diameters in Highly Shear Thinning, Power Law Liquids Using Axial Flow Impellers”,
Chemical Engineering Science, 53(3), 455-469 (1998).
Armenante, P.M., C. Luo, C. Chou, I. Fort and J. Medek, “Velocity Profiles in a Closed,
Unbaffled Vessel: Comparison between Experimental LDV Data and Numerical CFD
Predictions”, Chemical Engineering Science, 52(20), 3483-3492 (1997).
Bakker, A. and J.B. Fasano, “A Computational Study of the Flow Pattern in an Industrial Paper
Stock Chest with a Side-entering Impeller”, AIChE Symposium Series, 89(293), 118-124 (1993).
Bakker, C.W., C.J. Meyer and D.A. Deglon, “The Development of a Cavern Model for
Mechanical Flotation Cells”, Minerals Engineering, in press (2010).
Bennington, C.P.J., R.J. Kerekes and J.R. Grace, “The Yield Stress of Fibre Suspensions”, The
Canadian Journal of Chemical Engineering, 68(10), 748-757 (1990).
Bennington, C.P.J. and R.J. Kerekes, “Power Requirements for Pulp Suspension fluidization”,
TAPPI Journal, 79(2), 253-258 (1996).
Bhattacharya, S. and S.M. Kresta, “CFD Simulations of Three-dimensional Wall Jets in a Stirred
Tank”, The Canadian Journal of Chemical Engineering, 80, 1-15 (2002).
Bhattacharya, S., C. Gomez, A. Soltanzadeh, F. Taghipour, C.P.J. Bennington and G.A. Dumont,
“Computational Modelling of Industrial Pulp Stock Chests”, The Canadian Journal of Chemical
Engineering, 88, 295-305 (2010).
Bhole, M., C. Ford and C.P.J. Bennington, “Characterization of Axial Flow Impellers in Pulp
Fibre Suspensions”, Chemical Engineering Research and Design, 87, 648-653 (2009).
Bialkowski, W.L., “Newsprint Variability and Its Impact on Competitive Position”, Pulp and
Paper Canada, 93(11), T299-T306 (1992).
Bird, R.B., W.E. Stewart and E.N. Lightfoot, Transport Phenomena, John Wiley & Sons, Inc.
(2002).
115
Bolton, G.T., C.W. Hooper, R. Mann and E.H. Stitt, “Flow Distribution and Velocity
Measurement in a Radial Flow Fixed Bed Reactor Using Electrical Resistance Tomography”,
Chemical Engineering Science, 59, 1989-1997 (2004).
Bouillard, J., B. Alban, P. Jacques and C. Xuereb, “Liquid Flow Velocity Measurements in
Stirred Tanks by Ultra-sound Doppler Velocimetry”, Chemical Engineering Science, 56, 747-
754 (2001).
Chhabra, R.P. and J.F. Richardson, Non-Newtonian Flow in the Process Industries, Butterworth-
Heinemann (1999).
Devries, T.C. and M.D. Doyle, “Dilution Zone Agitation Critical in Blowtanks, High-density
Towers”, Pulp and Paper, 87-95 (1995).
Dickin, F. and M. Wang, “Electrical Resistance Tomography for Process Applications”, Meas.
Sci. Technol., 7, 247-260 (1996).
Ein-Mozaffari, F., “Macroscale Mixing and Dynamic Behavior of Agitated Pulp Stock Chests”,
PhD Thesis, University of British Columbia, Vancouver, Canada (2002).
Ein-Mozaffari, F., L.C. Kammer, G.A. Dumont and C.P.J. Bennington, “Dynamic Modeling of
Agitated Pulp Stock Chests”, TAPPI Journal, 2(9), 13-17 (2003).
Ein-Mozaffari, F., C.P.J. Bennington and G.A. Dumont, “Suspension Yield Stress and the
Dynamic Response of Agitated Pulp Chests”, Chemical Engineering Science, 60, 2399-2408
(2005).
Ein-Mozaffari, F., C.P.J. Bennington, G.A. Dumont and D. Buckingham, “Using Ultrasonic
Doppler Velocimetry to Measure Flow Velocity in Pulp Suspension Mixing”, 12th European
Conference on Mixing, Bologna, 27-30 June 2006.
Ein-Mozaffari, F., C.P.J. Bennington, G.A. Dumont and D. Buckingham, “Measuring Flow
Velocity in Pulp Suspension Mixing Using Ultrasonic Doppler Velocimetry”, Chemical
Engineering Research and Design, Trans IChemE, Part A, 85(A5), 591-597 (2007).
Ein-Mozaffari and S.R. Upreti, “Using Ultrasonic Doppler Velocimetry and CFD Modeling to
Investigate the Mixing of non-Newtonian Fluids Possessing Yield Stress”, Chemical Engineering
Research and Design, 87, 515-523 (2009).
116
Elson, T.P., D.J. Cheesman, A.W. Nienow, “X-ray Studies of Cavern Sizes and Mixing
Performance with Fluids Possessing a Yield Stress”, Chemical Engineering Science, 41(10),
2555-2562 (1986).
Elson, T.P., “The Growth of Caverns Formed Around Rotating Impellers During the Mixing of a
Yield Stress Fluid”, Chemical Engineering Communication, 96, 303-319 (1990).
Ferziger, J.H. and M. Peric, Computational Methods for Fluid Dynamics, Springer-Verlag Berlin
Heidelberg (1999).
Ford, C., “CFD Simulation of Mixing Dynamics in Agitated Pulp Stock Chests”, MASc Thesis,
The University of British Columbia, Vancouver, Canada (2004).
Gibbon, J.D. and D. Attwood, “Prediction of Power Requirements in the Agitation of Fibre
Suspensions”, Trans. Inst. Chem. Engrs., 40, 75-82 (1962).
Gomez, C., C.P.J. Bennington and F. Taghipour, “Investigation of the Flow Field in a
Rectangular Vessel Equipped with a Sider-entering Agitator”, Journal of Fluids Engineering,
132(5), 051106-1 – 051106-13 (2010).
Gomez, C., B. Derakhshandeh, S.G. Hatzikiriakos and C.P.J. Bennington, “Carbopol as a model
fluid for studying mixing of pulp fibre suspensions”, Chemical Engineering Science, 65, 1288-
1295 (2010).
Holden, P.J., M. Wang and R. Mann, “Imaging Stirred-vessel Macromixing Using Electrical
Resistance Tomography”, AIChE Journal, 44(4), 780-790 (1998).
Hoyle, B.S., H. McCann and D.M. Scott, “Process Tomography”, Process Imaging for
Automatic Control, Taylor & Francis Group, LLC (2005).
Hui, L.K., C.P.J. Bennington and G.A. Dumont, “Cavern Formation in Pulp Suspensions Using
Side-entering Axial-flow Impellers”, Chemical Engineering Science, 64, 509-519 (2009).
Hui, L.K., C.P.J. Bennington and G.A. Dumont, “Mixing Dynamics in Cylindrical Pulp Stock
Chests”, 2010 EXFOR & PAPTAC Annual Meeting, Feb. 2-4, 2010, Montreal, QC (2010).
117
Kasat, G.R., A.R. Khopkar, V.V. Ranade and A.B. Pandit, “CFD Simulation of Liquid-phase
Mixing in Solid-liquid Stirred Reactor”, Chemical Engineering Science, 63, 3877-3885 (2010).
Kim, S., A.N. Nkaya and T. Dyakowski, “Measurement of Mixing of Two Miscible Liquids in a
Stirred Vessel with Electrical Resistance Tomography”, International Communications in Heat
and Mass Transfer, 33, 1088-1095 (2006).
Li, M., G. White, D. Wilkinson and K.J. Roberts, “Scale Up Study of Retreat Curve Impeller
Stirred Tanks using LDA Measurements and CFD Simulation”, Chemical Engineering Journal,
108, 81-90 (2005).
Ma, Y., Z. Zheng, L. Xu, X. Liu and Y. Wu, “Application of Electrical Resistance Tomography
System to Monitor Gas/Liquid Two-phase Flow in a Horizontal Pipe”, Flow Measurement and
Instrumentation, 12, 259-265 (2001).
Macosko, C.W., Rheology, Principles, Measurement and Applications, VCH Publisher, New
York (1994).
Mann, R., F.J. Dickin, M. Wang, T. Dyakowski, R.A. Williams, R.B. Edwards, A.E. Forrest and
P.J. Holden, “Application of Electrical Resistance Tomography to Interrogate Mixing Processes
at Plant Scale”, Chemical Engineering Science, 52, 2087-2097 (1997).
Martin, A., “To Find the Area Common to Two Intersecting Circles”, The Analyst, 1(2), 33-34
(1874).
Oldshue, J.Y., Fluid Mixing Technology, McGraw-Hill Publications Co., New York (1983).
Paul, E.L., V.A. Atiemo-Obeng and S.M. Kresta, Handbook of Industrial Mixing: Science and
Practice, John Wiley & Sons, Inc. (2004).
Reed, C.S., “Selecting the Right Equipment for Agitation and Blending, Part 2”, TAPPI Journal,
78(7), 241-244 (1995).
Saeed, S., F. Ein-Mozaffari and S. R. Upreti, “Using Computational Fluid Dynamics Modelling
and Ultrasonic Doppler Velocimetry to Study Pulp Suspension Mixing”, Industrial &
Engineering Chemistry Research, 46, 2172-2179 (2007).
Shekarriz, A. and D.M. Sheen, “Slurry Pipe Flow Measurements using Tomographic Ultrasonic
Velocimetry and Densitometry”, Proceedings of FEDSM, June 21-25, 1998, Washington, D.C.,
USA.
118
Solomon, J., T.P. Elson and A.W. Nienow, “Cavern Sizes in Agitated Fluids with a Yield Stress”,
Chemical Engineering Communication, 11, 143-164 (1981).
Stephenson, D.R., M. Cooke, A. Kowalski and T.A. York, “Determining Jet Mixing
Characteristics using Electrical Resistance Tomography”, Flow Measurement and
Instrumentation, 18, 204-210 (2007).
Takeda, Y., “Development of an Ultrasound Velocity Profile Monitor”, Nuclear Engineering and
Design, 126, 277-284 (1991).
Takeda Y., “Ultrasonic Doppler Method for Velocity Profile Measurement in Fluid Dynamics
and Fluid Engineering”, Experiments in Fluids, 26, 177-178 (1999).
TAPPI, Generalized Method for Determining the Pipe Friction Loss of Flowing Pulp
Suspensions, TIS 0410-14 (1998).
Wahren, D., “Fibre Network Structures in Paper Making Operations”, Conference Paper Science
and Technology, The Cutting Edge, Institute of Paper Chemistry, Appleton, WI, 112-132 (1980).
Wang, L., K.L. McCarthy and M.J. McCarthy, “Effect of Temperature Gradient on Ultrasonic
Doppler Velocimetry Measurement During Pipe Flow”, Food Research International, 37, 633-
642 (2004).
Wang, M., “Impedance Mapping of Particulate Multiphase Flows”, Flow Measurement and
Instrumentation, 16, 183-189 (2005).
Wang, T., J. Wang, F. Ren and Y. Jin, “Application of Doppler Ultrasonic Velocimetry in
Multiphase Flow”, Chem. Eng. J., 92, 111-122 (2003).
Wilkens, R.J., C. Henry and L.E. Gates, “How to Scale-Up Mixing Processes in Non-Newtonian
Fluids”, Chem. Eng. Prog., 99(5), 44-52 (2003).
Wilkens, R.J., J.D. Miller, J.R. Plummer, D.C. Dietz and K.J. Myers, “New Techniques for
Measuring and Modeling Cavern Dimensions in a Bingham Plastic Fluid”, Chemical
Engineering Science 60, 5269-5275 (2005).
William, R.A., X. Jia and S.L. McKee, “Development of Slurry Mixing Models using Resistance
Tomography”, Powder Technology, 87, 21-27 (1996).
Wilstrom, T. and Rasmuson A., “The Agitation of Pulp Suspensions with a Jet Nozzle Agitator”,
Nordic Pulp Pap. Res. J., 13(2), 88-94 (1998a).
Wikstrom, T. and Rasmuson, A., “Yield Stress of Pulp Suspensions. The Influence of Fibre
Properties and Processing Conditions”, Nord. Pulp Pap. Res. J., 13(3), 243-246 (1998b).
Xu, H., “Measurement of Fiber Suspension Flow and Forming Jet Velocity Profile by Pulsed
Ultrasonic Doppler Velocimetry”, PhD thesis (2003).
Xu, H. and Aidun, C.K., “Characteristics of Fiber Suspension Flow in a Rectangular Channel”,
International Journal of Multiphase Flow, 31, 318-336 (2005).
Yackel, D.C., Pulp and Paper Agitation: The History, Mechanics, and Process, TAPPI Press,
Atlanta (1990).
York, T.A., “Status of Electrical Tomography in Industrial Applications”, Process Imaging for
Automatic Control, Proceedings of SPIE 4188, 175-190 (2001).
A. Sensor system
The size and the material of the electrodes are both important in producing and sensitively
measuring the electrical field distribution (Mann et al., 1997). Wang et al. (1995) found that the
smaller the size of electrodes, the higher the sensitivity of the voltage measurement electrodes
and the higher common voltage at current driven electrodes, when measurement and current
excitation use the same electrode system. However, the higher common voltage will produce a
poor signal-to-noise ratio. Also, current-injecting electrodes should have large surface areas in
order to generate an even current density (Dickin and Wang, 1996). Thus, the electrode size
should not be too large or too small. In addition, the electrodes must be made of a material more
electrically conductive than the process fluid (Ricard, 2005). Metallic electrodes are thus often
used for process applications, e.g., stainless steel, silver, gold or platinum.
In order to track the small changes of resistivity in real-time, the data collection has to be quick
and accurate, enabling the reconstruction algorithm to provide a precise indication of the true
resistivity distribution. The DAS is composed of a signal source, an electrode multiplexer array,
voltmeters, signal demodulators and a system controller (Mann et al., 1997). This complexity is
required because of the low amplitude of measurements at the boundary, the small responses of
dynamic change, the large number of electrode channel operations, the high common voltages
and the large stray capacitance of coaxial cable. The signal source consists of a master oscillator
frequency and amplitude reference for all of the current sources channels and as a switching
121
function for the demodulator stage, generates a harmonically pure sine-shaped waveform signal
in order to “probe” the material under investigation. The sine-wave voltage output from the
the electrical “probe” due to the variation of contact impedance between the electrode and the
fluid inside the vessel. The multiplexers are used to “share” the current source and voltage
measurement stages between any numbers of electrodes. They must exhibit a number of
properties: low on-resistance, fast switching speed, low inter-switch cross talk and low power
consumption. The signal demodulators are used to “decode” the voltage signal and to optimize
the signal-to-noise ratio by recovering the amplitude attenuation and phase shift of the sine wave
The image reconstruction algorithm can be thought of simply as a series of procedures performed
resistivities (e.g., component concentrations) within the process vessel. There are two different
images depicting a change in resistivity relative to an initially acquired set of “reference” data
and the quantitative one creates images depicting values of resistivity or conductivity for each
equipotential lines. The potential difference, calculated by the forward solver, between two
equipotential lines on the boundary is back-projected to a resistivity value in the area enclosed by
the two lines for all possible injection/measurement combinations. The main advantage of this
122
algorithm is that it can be performed in a single step using a pre-calculated pixel sensitivity
developed for nonlinear problems. It is intended to quantify the variation of the conductivity in
the region of interest during the process. The reconstruction process is initiated when a set of
resistivities for the region of interest is fed into the forward problem solver. When the least-
squares error between the calculated boundary voltages and the data acquisition voltages is less
than the pre-defined one, the reconstruction process is halted and the final updated set of
References
Dickin, F. and M. Wang, “Electrical Resistance Tomography for Process Applications”, Meas.
Sci. Technol., 7, 247-260 (1996).
Mann, R., F.J. Dickin, M. Wang, T. Dyakowski, R.A. Williams, R.B. Edwards, A.E. Forrest and
P.J. Holden, “Application of Electrical Resistance Tomography to Interrogate Mixing Processes
at Plant Scale”, Chemical Engineering Science, 52, 2087-2097 (1997).
Plaskowski, A., M.S. Beck, R. Thorn and T. Dyakowski, Imaging Industrial Flows, Applications
of Electrical Process Tomography, Institute of Physics Publishing (1995).
Wang, M., F.J. Dickin and R.A. Williams, “The Grouped-node Technique as a Means of
Handling Large Electrode Surfaces in Electrical Impedance Tomography”, Physiological
Measurement, 16, 219-226 (1995).
124
clc
clear
for i=1:1
filen=strcat('lab','1')
load(filen);
Ts=ts;
Tres=3;
datac=iddata(cout',cin',Ts)
% datac=resample(datac,1,Tres)
% datac=datac(10:end);
datac.int='foh';
% ze =datac(1:floor(length(datac.u)/2));
% defining data set
ze=datac(1:floor(length(datac.u)/2));
%removing mean
ze=detrend(ze,'constant')
advice(datac)
% identification command
m =
pem(ze,'P1D','kp',{'max',2},'kp',{'min',0.5},'Td',{'max',100},'Td',{'min',2},
'Tz',{'min',0},'dist','arma2')
m = pem(ze,m)
present (m)
% zv=datac(floor(length(datac.u)/2):length(datac.u));
zv=datac(floor(length(datac.u)/2):length(datac.u));
zv=detrend(zv,'constant');
Kp(i)=m.kp.value;
Td(i)=m.td.value%*Tres*Ts;
Tz(i)=m.tz.value%*Tres*Ts;
Tp1(i)=m.tp1.value%*Tres*Ts;
Tp2(i)=m.tp2.value%*Tres*Ts;
tu1(i)=min(Tp1(i),Tp2(i));
tu2(i)=max(Tp1(i),Tp2(i));
f(i)=(Tz(i)-tu1(i))/(tu2(i)-tu1(i));
figure(i)
compare(zv,m);
resid(zv,m);
advice(m);
% figure(i+1)
[yo,fit(i)]=compare(zv,m);
ou=yo{1,1};
subplot(3,1,1)
plot(ou.sa,ou.y,'b',zv.sa,zv.y,'k')
title(['exp no' int2str(i) ': ''Td=' num2str(Td(i)) ', ' 'K='
num2str(Kp(i)) ', ' 'f=' num2str(f(i)) ',' 'Tz=' num2str(Tz(i)) ', ' 'Tp1='
num2str(Tp1(i)) ', ' 'Tp2=' num2str(Tp2(i)) ', ' 'Ts=' num2str(Ts)])
legend('m out',['real ' '%Fit= ' num2str(fit(i))])
125
subplot(3,1,2)
plot(ze.sa,ze.u,'b.',ze.sa,ze.y,'k')
title(['exp no' int2str(i) ': ' 'Estimation Data'])
legend('input data','output data')
subplot(3,1,3)
plot(zv.sa,zv.u,'b.',zv.sa,zv.y,'k')
title(['exp no' int2str(i) ': ' 'Validation Data'])
legend('input data','output data')
end
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'A1','
P2DZ,Arma1');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'A2',f
it');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'B1','
Kp');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'B2',K
p');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'C1','
Td');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'C2',T
d');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'D1','
Tz');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'D2',T
z');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'E1','
Tp1');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'E2',T
p1');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'F1','
Tp2');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'F2',T
p2');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'G1','
tu1');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'G2',t
u1');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'H1','
tu2');
126
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'H2',t
u2');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'I1','
f');
%
write2excel('C:\MATLAB7\work\Ali\Identifyingonedelaymodel\fitresult',0,'I2',f
');
Program “write2excel.m”
function write2excel(fileloc,promptforsave,varargin)
% write2excel(fileloc,promtforsave,range1,data1,range2,data2,...)
%
%
% Uses ActiveX commands to write data_n into range_n in an
existing Excel
% spreadsheet. Inputs (excluding fileloc and promptforsave) must
be paired.
% As of 10/04 update, you may provide the target range (upper
left
% cell to lower right cell) OR just the upper right cell. If
% the range is specified, the function will verify that the
% corresponding data block is the correct size, and give an
% error if not. (This may be useful for error checking, for
instance.)
% If only the upper left cell is provided, write2excel will
% compute the target range.
%
% (Please use caution, as you can now overwrite data pretty
easily.)
%
% Additionally, you may now specify cells by address (eg., 'H3')
OR row, column
% (eg, '[3,8]').
%
% FILELOC: Enter a string representing the location of an Excel
file.
% Example: 'c:\brett\my archives\test1.xls'
% PROMPTFORSAVE: binary variable. 1 (DEFAULT) = Prompt before saving
% 0 = No prompt required
% RANGE SPECIFIER(S): Enter the range(s) to read. You may use Excel
% cell-references, as in:
% 'B1:P5'
% 'B1:B1' (or simply 'B1')
% OR, alternatively, you may specify the row, column values, as
% in:
% '[8,3]:[12,4]' (to write from row 8, column
% 3 to row 12, column 4);
% '[8,3]' to write from row 8, column 3
% TO WHATEVER RANGE IS REQUIRED FOR THE
DATA BLOCK.
%
% DATA: NOTE: To enter multiple strings, use cell arrays. Size
compatibility is verified
127
if nargin < 4
msgstr = sprintf('At a minimum, you must specify three input
arguments.\nThe first is a string indicating the location of the excel
file,\nthe second is a range to be written, and the third contains the data
to write.');
error(msgstr);
elseif ~iseven(nargin-2)
msgstr = sprintf('Please enter input variables in pairs...\n''write
range'',data,''write range'',data')
error(msgstr)
end
tmp = varargin;
sheetchanges = [];counter = 1;
for ii = 1:length(tmp)
if ischar(tmp{ii}) & (strcmp(tmp{ii},'sheet') |
strcmp(tmp{ii},'sheetname'))
sheetchanges(counter) = ii;
counter = counter + 1;
end
end
if ~isempty(sheetchanges)
[sheetnames{1:length(sheetchanges)}] = deal(varargin{sheetchanges+1});
end
[pathstr,name,ext] = fileparts(fileloc);
if isempty(ext)
fileloc = [fileloc,'.xls'];
end
if isempty(pathstr)
fileloc = which(fileloc,'-all');
if size(fileloc,1) ~= 1
error('File was either not located, or multiple locations were found.
Please reissue readfromexcel command, providing absolute path to the file of
interest.');
end
end
if ismember(ii,sheetchanges) | ismember(ii,sheetchanges + 1)
continue
end
% How are cells specified?
if any(ismember(double(varargin{ii}),[65:90,97:122]))
addrtype = 'letternumber';
else
addrtype = 'rowcol';
end
% Is range provided, or should it be auto-calculated?
autorange = isempty(findstr(varargin{ii},':'));
switch addrtype
case 'letternumber'
if autorange
r1{ii} = varargin{ii};
[rx1,cx1] = an2nn(r1{ii});
rx2 = rx1 + size(varargin{ii+1},1)-1;
cx2 = cx1 + size(varargin{ii+1},2)-1;
r2{ii} = nn2an(rx2,cx2);
else
tmp = findstr(varargin{ii},':');
r1{ii} = varargin{ii}(1:tmp-1);
r2{ii} = varargin{ii}(tmp+1:end);
[rx1,cx1] = an2nn(r1{ii});
[rx2,cx2] = an2nn(r2{ii});
end
case 'rowcol'
if autorange
r1{ii} = varargin{ii};
[t,r]=strtok(r1{ii},',');
rx1 = str2num(t(2:end));
cx1 = str2num(r(2:end-1));
r1{ii} = nn2an(rx1,cx1);
rx2 = rx1 + size(varargin{ii+1},1)-1;
cx2 = cx1 + size(varargin{ii+1},2)-1;
r2{ii} = nn2an(rx2,cx2);
else
tmp = findstr(varargin{ii},':');
r1{ii} = varargin{ii}(1:tmp-1);
[t,r]=strtok(r1{ii},',');
rx1 = str2num(t(2:end));
cx1 = str2num(r(2:end-1));
r2{ii} = varargin{ii}(tmp+1:end);
[t,r]=strtok(r2{ii},',');
rx2 = str2num(t(2:end));
cx2 = str2num(r(2:end-1));
r1{ii} = nn2an(rx1,cx1);
r2{ii} = nn2an(rx2,cx2);
end
end
if ~autorange % Validate size match for target range, data block
sz = [rx2 - rx1 + 1, cx2 - cx1 + 1];
switch class(varargin{ii+1})
case {'double','cell'}
sz2 = size(varargin{ii+1});
case 'char'
sz2 = [size(varargin{ii+1},1),1];
129
end
if ~isequal(sz,sz2)
error(sprintf('Mismatched range/data size for input pair %d.
Specified range is %d x %d, data block is %d x
%d.',(ii+1)/2,sz(1),sz(2),sz2(1),sz2(2)));
end
end
end
Excel = actxserver('Excel.Application');
Excel.Visible = 0;
w = Excel.Workbooks;
try
excelarchive = invoke(w, 'open', fileloc);
catch
invoke(Excel, 'quit');
release(w);
delete(Excel);
error(sprintf('Sorry...unable to open file %s',fileloc));
end
Sheets = Excel.ActiveWorkBook.Sheets;
archive = Excel.Activesheet;
initval = get(archive,'Index');
archive.Unprotect;
sheet = get(Sheets,'Item',initval);
130
invoke(sheet,'Activate');
if ~promptforsave
invoke(excelarchive,'save');
end
invoke(Excel, 'quit');
release(excelarchive);
release(w);
delete(Excel);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%
%SUBFUNCTIONS
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%
function k=iseven(x)
k = x/2==floor(x/2);
return
function cr = nn2an(r, c)
% convert number, number format to alpha, number format
%t = [floor(c/27) + 64 floor((c - 1)/26) - 2 + rem(c - 1, 26) + 65];
t = [floor((c - 1)/26) + 64 rem(c - 1, 26) + 65];
if(t(1)<65), t(1) = []; end
cr = [char(t) num2str(r)];
Program “testexcel.m”; this program converts the raw data in excel file to M-file format
for processing
Program “readfromexcel.m”
if nargin < 2
msgstr = sprintf('\nAt a minimum, you must specify three input
arguments.\nThe first is a string indicating the location of the excel
file,\nand the second is a range to be read.');
error(msgstr);
end
sheetchanges =
[strmatch('sheet',varargin,'exact');strmatch('sheetname',varargin,'exact')];
if ~isempty(sheetchanges)
[sheetnames{1:length(sheetchanges)}] = deal(varargin{sheetchanges+1});
end
[pathstr,name,ext] = fileparts(fileloc);
if isempty(ext)
fileloc = [fileloc,'.xls'];
end
133
if isempty(pathstr)
fileloc = which(fileloc,'-all');
if size(fileloc,1) ~= 1
error('File was either not located, or multiple locations were found.
Please reissue readfromexcel command, providing absolute path to the file of
interest.');
end
end
Excel = actxserver('Excel.Application');
Excel.Visible = 0;
w = Excel.Workbooks;
try
excelarchive = invoke(w, 'open', fileloc);
catch
invoke(Excel, 'quit');
release(w);
delete(Excel);
error(sprintf('Sorry...unable to open file %s',fileloc));
end
Sheets = Excel.ActiveWorkBook.Sheets;
archive = Excel.Activesheet;
initval = get(archive,'Index');
%Parse range
rangespec = 0;
if strcmp(lower(varargin{ii}),'all') %Range of the form 'ALL'
rangespec = 1;
else
tmp = findstr(varargin{ii},':');
if isempty(tmp) %Range of the form 'A' or '2' or 'A2'
r1 = varargin{ii};
134
r2 = r1;
if ~any(ismember(r1,num2str([1:9]))) %Range of the form 'A'
rangespec = 2;
elseif all(ismember(r1,num2str([1:9]))) %Range of the form '2'
rangespec = 3;
else %Range of the form 'A2'
rangespec = 4;
end
else % Range of the form 'A2:B3', '2:2', or 'A:A'
r1 = varargin{ii}(1:tmp-1);
if all(ismember(r1,num2str([1:9]))) %Range of the form '2:2'
r2 = r1;
rangespec = 5;
elseif ~any(ismember(r1,num2str([1:9]))) %Range of the form 'A:A'
r2 = r1;
rangespec = 6;
else
%Range of the form 'A1:B2'
r2 = varargin{ii}(tmp+1:end);
rangespec = 7;
end
end
end
try
switch rangespec
case 1
readinfo = get(archive,'UsedRange');
case {2,6}
readinfo = get(archive,'UsedRange');
[r,c] = an2nn(r1);
r1 = nn2an(readinfo.row,c);
[m,n] = size(readinfo.value);
r2 = nn2an(readinfo.row+m,c);
readinfo = get(archive, 'Range', r1, r2);
case {3,5}
readinfo = get(archive,'UsedRange');
[m,n] = size(readinfo.value);
r2 = nn2an(r1,readinfo.row+n);
r1 = nn2an(r1,readinfo.column);
readinfo = get(archive, 'Range', r1, r2);
case {4,7}
readinfo = get(archive, 'Range', r1, r2);
otherwise
readinfo.value = {};
fprintf('Error parsing input argument %d.',ii+1);
end
catch
fprintf('Error reading range specified by input argument %d.',ii+1);
invoke(Excel, 'quit');
release(excelarchive);
release(w);
delete(Excel);
return
end
varargout{argcount} = readinfo.value;
argcount = argcount + 1;
135
end
% Reset to initial active sheet
sheet = get(Sheets,'Item',initval);
invoke(sheet,'Activate');
try
release(readinfo);
end
invoke(excelarchive,'close'); %This closes without saving, so changing the
active sheet is temporary
%invoke(excelarchive,'save'); %Note: Instead of invoke(excelarchive,'close'),
I use the save option after
% switching back to the initially active sheet.
This stops Excel from showing
% "previously saved versions" when the file is
% next opened.
invoke(Excel, 'quit');
release(excelarchive);
release(w);
delete(Excel);
return
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%
%SUBFUNCTIONS
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%
function cr = nn2an(r, c)
% convert number, number format to alpha, number format
%t = [floor(c/27) + 64 floor((c - 1)/26) - 2 + rem(c - 1, 26) + 65];
t = [floor((c - 1)/26) + 64 rem(c - 1, 26) + 65];
if(t(1)<65), t(1) = []; end
cr = [char(t) num2str(r)];
DEFINE_ON_DEMAND(cavern_calc_s)
{
Domain *d;
real strain, volume, vol_tot;
Thread *t;
cell_t c;
d = Get_Domain(1);
thread_loop_c(t,d)
{
begin_c_loop(c,t)
{
volume = C_VOLUME(c,t);
strain = C_STRAIN_RATE_MAG(c,t);
DEFINE_PROFILE(hwtracer_mf_profile, t, i)
{
real flow_time = CURRENT_TIME;
face_t f;
begin_f_loop(f,t)
{
if (flow_time > 3411)
F_PROFILE(f, t, i) = 0.0847;
else if (flow_time > 3284)
F_PROFILE(f, t, i) = 0.1096;
else if (flow_time > 3192)
F_PROFILE(f, t, i) = 0.0847;
else if (flow_time > 3070)
F_PROFILE(f, t, i) = 0.1096;
else if (flow_time > 2986)
F_PROFILE(f, t, i) = 0.0847;
else if (flow_time > 2531)
F_PROFILE(f, t, i) = 0.1096;
else if (flow_time > 2453)
F_PROFILE(f, t, i) = 0.0847;
else if (flow_time > 2208)
F_PROFILE(f, t, i) = 0.1096;
else if (flow_time > 2130)
F_PROFILE(f, t, i) = 0.0847;
else if (flow_time > 1992)
F_PROFILE(f, t, i) = 0.1096;
else if (flow_time > 1803)
F_PROFILE(f, t, i) = 0.0847;
else if (flow_time > 1566)
F_PROFILE(f, t, i) = 0.1096;
else
F_PROFILE(f, t, i) = 0.0847;
}
end_f_loop(f,t)
}
138
P4 P5
P1 P2 P3
150
P4 P5
P1 P2 P3
200
P4 P5
Pulp type: hardwood; Mass concentration: 2%; E/D = 0.6; Z/T = 1.0
P4 P5
P1 P2 P3
250
P4 P5
Conductivity
low high
scale
140
Pulp type: hardwood; Mass concentration: 2%; E/D = 0.4; Z/T = 1.0
P4 P5
P1 P2 P3
150
P4 P5
P1 P2 P3
200
P4 P5
Pulp type: hardwood; Mass concentration: 2%, E/D = 0.4; Z/T = 1.0
P4 P5
P1 P2 P3
250
P4 P5
Conductivity
low high
scale
142
Pulp type: hardwood; Mass concentration: 2%; E/D = 0.4; Z/T = 0.8
125 P1 P2
P3 P4
150 P1 P2
P3 P4
200 P1 P2
P3 P4
Pulp type: hardwood; Mass concentration: 2%; E/D = 0.4; Z/T = 0.8
225 P1 P2
P3 P4
250 P1 P2
P3 P4
Conductivity
low high
scale
144
Pulp type: hardwood; Mass concentration: 3%; E/D = 0.4; Z/T = 0.8
425 P1 P2
P3 P4
450 P1 P2
P3 P4
475 P1 P2
P3 P4
Pulp type: hardwood; Mass concentration: 3%; E/D = 0.4; Z/T = 0.8
500 P1 P2
P3 P4
525 P1 P2
P3 P4
Conductivity
low high
scale
146
Pulp type: hardwood; Mass concentration: 4%; E/D = 0.4; Z/T = 0.8
550 P1 P2
P3 P4
575 P1 P2
P3 P4
600 P1 P2
P3 P4
Pulp type: hardwood; Mass concentration: 4%; E/D = 0.4; Z/T = 0.8
650 P1 P2
P3 P4
675 P1 P2
P3 P4
Conductivity
low high
scale
148
Pulp type: softwood; Mass concentration: 3%; E/D = 0.4; Z/T = 0.8
425 P1 P2
P3 P4
450 P1 P2
P3 P4
475 P1 P2
P3 P4
Pulp type: softwood; Mass concentration: 3%; E/D = 0.4; Z/T = 0.8
500 P1 P2
P3 P4
525 P1 P2
P3 P4
Conductivity
low high
scale