Professional Documents
Culture Documents
A Thesis
Presented to
In Partial Fulfillment
Master of Science
October, 2016
INVESTIGATION OF DESIGN AND OPERATING PARAMETERS IN
Thesis
Approved: Accepted:
Department Chair
Dr. Sergio D. Felicelli
ii
ABSTRACT
The modern rubber industry is always in pursuit of improvements in the properties of the
final product resulting from the mixing of the rubber compounds with different fillers and
additives. Depending on the functional characteristics of the final product and thus the
compounding ingredients, different types of mixers can be used for the rubber mixing
process. Hence, the choice of an appropriate mixer is critical in achieving the proper dis-
tribution and dispersion of fillers in rubber, and a consistent product quality, as well as is
the attainment of high productivity. Besides rotor design, operational parameters such as
speed ratio and the orientation of the mixing rotors with respect to each other also play
significant role in the mixing performance. With the availability of high-performance com-
puting resources and high-fidelity computational fluid dynamics tools, understanding the
flow field and mixing characteristics associated with rotor orientations, speed ratios and
complex rotor geometries, has become more feasible over the last two decades. As part of
this effort, all the simulations here are carried out in a 75% fill chamber with two counter-
rotating rotors using a CFD code. In the phase angle and rotor design studies conducted
here, the rotors rotate at 20 rpm even speed, whereas for speed ratio study, only the left
rotor rotates at 20 rpm and the right rotor rotates at a speed, which is a multiple of 20 rpm
by the speed ratio specified. The computational models used in this research are based on a
iii
finite volume method to simulate a partially filled mixer equipped with different tangential
rotor types. The model solves for transient, isothermal and incompressible set of govern-
ing fluid equations for the mixing of non-Newtonian high-viscosity rubber. The research
here considers phase angles of 45◦ , 90◦ and 180◦ , speed ratios of 1.0, 1.125 and 1.5, and
rotor designs including 2-wing, 4-wing A and the 4-wing B rotors. Investigation of each
The flow field is analyzed via pressure and velocity contours, mass flow patterns,
velocity vectors and particle trajectories. Dispersive mixing is evaluated through his-
tograms of mixing index, joint probability density functions of mixing index and shear rate,
the particles. Distributive mixing is quantified statistically using cluster distribution index,
axial distribution, inter-chamber particles transfer, segregation scale and length of stretch.
The results helped in understanding the mixing process and material movement, thereby
generating information that could potentially improve the productivity and efficiency in tire
manufacturing process.
iv
ACKNOWLEDGEMENTS
First of all, the author would like to thank her adviser Dr. Abhilash J. Chandy for his
guidance, support, supervision, and insights throughout the research, as well as for showing
her right direction not just in research, but in other aspects of graduate life as well.
The author is grateful to Dr. Alex Povitsky and Dr. Jae-Won Choi for being in the
The author would also like to thank The Goodyear Tire & Rubber Company,
Finally, the author would like to express her gratitude towards her parents, for
v
TABLE OF CONTENTS
Page
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
CHAPTER
I. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.1 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.2 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
V. RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
vi
5.5 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
vii
LIST OF TABLES
Table Page
5.1 Percentage of Particles transferred at least once from one chamber to the other 95
viii
LIST OF FIGURES
Figure Page
1.3 Banbury’s machine for treating rubber or heavy plastic materials [3]. . . . . . 8
1.5 Weidmann and Schmid’s mixing apparatus (a) Rotor geometry, (b) Ma-
terial motion [5]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6 Nortey’s 2-wing non-intermeshing rotors (a) Rotor geometry, (b) Mate-
rial motion [6]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.7 Nortey’s 4-wing non-intermeshing rotors (a) Rotor geometry, (b) Mate-
rial motion [5]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.8 Takakura et al.’s 6-wing non-intermeshing rotors (a) Cross section of
rotor, (b) Material motion [7]. . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.9 Inoue et al.’s kneading apparatus (a) Rotor geometry, (b) Material motion [8]. 14
1.10 Inoue et al.’s batch mixer (a) Rotor geometry, (b) Material motion [9]. . . . . 14
1.11 Valmis et al.’s batch mixer (a) Rotor geometry, (b) Material motion [10]. . . . 16
1.12 Limper et al.’s batch mixer (a) Cross section of rotor geometry, (b) Ma-
terial motion [11]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.13 Yoshida et al.’s kneading rotor (a) Rotor geometry, (b) Material motion [12]. . 36
ix
1.15 Wiedmann and Schmid’s intermeshing rotors [5]. . . . . . . . . . . . . . . . . 38
1.18 (a) GK-N series tangential mixers, (b) GK-E series intermeshing mixers,
(Courtesy of HF mixing group). . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.19 Cooling systems used in the rotors (Courtesy of HF mixing group). . . . . . . 40
2.1 2-wing rotor geometry: (a) Top view, (b) Cross-section at the middle. . . . . . 43
2.2 Orientation of the rotors for phase angle 45◦ (a), 90◦ (b) and 180◦ (c). . . . . . 45
2.3 Cross-section of the rotor along a plane in the middle at different time
instants during a rotor revolution, represented by different colors, for a
speed ratio of 1.125. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.4 Differential phase angles Vs. revolutions of the left (slower rotating) rotor . . 47
2.6 Comparison of average velocity of rubber between 100% and 75% fill
factors for 5 revolutions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.1 Change of non-Newtonian viscosity of rubber with shear rate based on
Carreau-Yasuda model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.1 2D unstructured quadrilateral mesh for phase angle 45◦ (a), 90◦ (b) and
180◦ (c). (d) Closer view of mesh at the tips. . . . . . . . . . . . . . . . . . . 55
4.2 Polyhedral mesh on the rotors. . . . . . . . . . . . . . . . . . . . . . . . . . . 59
x
5.1 Pressure contours for phase angles 45◦ (left), 90◦ (middle) and 180◦
(right) at 84◦ (top) and 360◦ (bottom) rotations of the rotors after 19
revolutions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.2 Velocity contours for phase angles 45◦ (left), 90◦ (middle) and 180◦
(right) at 84◦ (top) and 360◦ (bottom) rotations of the rotors after 19
revolutions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.3 Histogram of mixing index calculated over last 15 revolutions for phase
angles 45◦ , 90◦ and 180◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.4 Cumulative probability distribution function of maximum shear stress
experienced by particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.5 (a) Initial position of the particle-clusters (magenta circular-middle, cyan
diamond-rotor tip gap, black triangle-left center near rotor wall), (b)
Particle-cluster dimension. . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.6 Particle distribution for phase angles 45◦ (left), 90◦ (middle) and 180◦
(right) after 5, 15 and 20 revolutions (from top to bottom). . . . . . . . . . . . 75
5.7 Evolution of cluster distribution index for the three phase angles and
three different initial cluster positions: (a) cluster-1 (b) cluster-2, and
(c) cluster-3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.8 Evolution of cluster distribution index for all the three clusters combined. . . . 80
5.9 Transfer of particles from the (a) left to right chambers, and (b) right to
left chambers for three phase angles: 45◦ (green), 90◦ (green), and 180◦ (red). 81
5.10 Joint probability density function (jPDF) of the mixing index and shear
rate calculated over a period of 8 rotor revolutions of the left rotor after
12 revolutions for speed ratios of (a) 1, (b) 1.125 and (c) 1.5. . . . . . . . . . 85
5.11 Cumulative distribution of maximum shear stress experienced by the
particles over a period of one revolution . . . . . . . . . . . . . . . . . . . . 86
5.12 (a) Initial position of the particles (b) Probability distribution of the par-
ticle pair distances at the initial condition (red line); Also shown here
is the probability distribution of the particle pair distances of the ideal
distribution (black line with symbols). . . . . . . . . . . . . . . . . . . . . . 88
5.13 Particle distribution for speed ratios 1 (left), 1.125 (middle) and 1.5
(right) after 5, 15 and 20 revolutions (from top to bottom). . . . . . . . . . . . 91
5.14 Evolution of the cluster distribution index with time. . . . . . . . . . . . . . . 92
xi
5.15 Evolution of axial distribution with time. . . . . . . . . . . . . . . . . . . . . 93
5.16 Two types of particles with concentration 1 (green) and 0 (black) defined
at the end of 10 revolutions for scale of segregation calculation . . . . . . . . 94
5.17 Front view of the distribution of particles for speed ratios 1 (left), 1.125
(middle) and 1.5 (right) at 12 (top) and 20 (bottom) revolutions; The
green and black particles represent concentrations of 1 and 0, respectively. . . 94
5.18 Top view of the distribution of particles for speed ratios 1 (left), 1.125
(middle) and 1.5 (right) at 12 (top) and 20 (bottom) revolutions; The
green and black particles represent concentrations of 1 and 0, respectively. . . 97
5.19 Evolution of scale of segregation with time. . . . . . . . . . . . . . . . . . . . 98
5.20 Summary of (a) dispersive mixing and (b) distributive mixing capabili-
ties for three speed ratios: 1 (blue), 1.125 (red), and 1.5 (green). . . . . . . . . 99
5.21 Cross-sectional planes along the (a) axial, and (b) and (c) transverse di-
rections for analysis of flow rates (using (a) and (b)) and velocity vectors
(using (c)) at time, t=0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.22 Instantaneous velocity vectors colored by velocity magnitude for 2-wing
(left), 4-wing A (middle) and 4-wing B (right) for cross-sections at 25%,
50% and 75% (from top to bottom) of the rotor length at 1.2 revolutions;
cross-sections shown in Figure 5.21c. . . . . . . . . . . . . . . . . . . . . . . 102
5.23 Instantaneous velocity vectors colored by velocity magnitude for 2-wing
(left), 4-wing A (middle) and 4-wing B (right) for cross-sections at 25%,
50% and 75% (from top to bottom) of the rotor length at 1.67 revolu-
tions; cross-sections shown in Figure 5.21c. . . . . . . . . . . . . . . . . . . 103
5.24 Instantaneous velocity vectors of 2-wing (left), 4-wing A (middle) and
4-wing B (right) rotor at 1.2 (top) and 1.67 revolutions (bottom), for the
cross section through the both rotor along their axis. . . . . . . . . . . . . . . 106
5.25 Evolution of normalized flow rate of 2-wing (left), 4-wing A (middle)
and 4-wing B (right) rotors for 2 revolutions from 15 to 17, through
the cross sections shown in Figure 5.21a; Lines are colored by cross-
sections shown in Figure 5.21a. . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.26 Evolution of normalized flow rate of 2-wing (left), 4-wing A (middle)
and 4-wing B (right) rotors for 2 revolutions from 15 to 17, through
the cross sections shown in Figure 5.21b; Lines are colored by cross-
sections shown in Figure 5.21b. . . . . . . . . . . . . . . . . . . . . . . . . . 108
xii
5.27 Joint probability density of the mixing Index and shear rate calculated
over a period of one rotor revolution after 17 revolutions for the (a) 2-
wing (b) 4-wing A and (c) 4-wing B rotors. . . . . . . . . . . . . . . . . . . 110
5.28 Particle distribution for 2-wing (left), 4-wing A (middle) and 4-wing B
(right) at the initial condition, and after 1, 10 and 17 revolutions (from
top to bottom). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.29 Probability distribution of particles (a) at the initial time, and after 17
revolutions for (b) 2-wing, (c) 4-wing A and (d) 4-wing B rotors (solid
lines); Also shown are the corresponding ideal distribution of particles
for the three different rotors in (b), (c) and (d) (lines with symbols). . . . . . . 115
5.30 Evolution of the cluster distribution index with time. . . . . . . . . . . . . . . 116
5.32 Velocity profile along x-axis at the mid-cross section of the left half rotor
for 600k, 1.3M & 2M Cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.33 Comparison of pressure profile obtained from Freakley and Patel’s ex-
periment at 50 rpm at the speed ratio of 1:1.125 [17] and simulation at
20 rpm even speed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
xiii
CHAPTER I
INTRODUCTION
Mixing is a basic and important step in the manufacturing process of polymer industries,
such as rubber compounding and plastics processing. The process of enhancing mechan-
ical, chemical and physical properties by reducing the non-uniformity of the ingredients,
known as mixing, affects the throughput as well as the final product quality [18]. Different
physical properties are attainable, depending on the range of concentration and type of the
ingredients. Tensile strength, fracture resistance, wearing and tearing resistance are some
examples of physical attributes in such a process. Besides polymer industries, high viscous
mixing has applications also in the field of food processing, paint mixing and so on.
Achieving uniformity is the primary goal of any mixing process. However, mixing
ingredients with highly viscous fluids can be very challenging and can affect productivity.
The way to obtain good mixing is to have a better understanding of the mixing dynamics,
material characteristics, effect of different process parameters on mixing and the choice of
an appropriate mixer. In addition, the efficiency of mixing depends on flow pattern, which
in turn depends on the geometry, as well as different operational parameters such as rotor
1
1.1 Background
Devices used for polymer mixing include single screw extruders, twin screw extruders, in-
ternal mixers, two-roll mills and continuous mixers. The purpose of any mixer is to reduce
non-uniformity while mixing one or more polymers with additives and fillers. Extruders
and continuous mixers are widely popular for plastic compounding [20]. In the rubber
industry, extruders and continuous mixers serve the purpose of finishing the mixing, re-
moving gaseous material, and defining shape after receiving the mixed compound from
internal mixer [20–22]. Two-roll mills are sometimes used for enhancing the dispersion of
fillers [18, 23]. In two-roll mills, it is possible to adjust the gap between the rolls to obtain
required dispersion. These devices are used for different applications, depending on the
Internal batch mixers are primarily used for the sole purpose of rubber mixing
[22]. The history of using internal batch mixers for rubber mixing can be traced back
to the early 20th century [24]. With an increasing interest for rubber mixing in the tire
industry, internal batch mixers started to gain more popularity [24]. Internal batch mixers
comprise mainly of mixing rotors, a mixing chamber, a feeding hopper, a ram, a drop door,
a base plate and end plates (See Figure 1.1). Two C-shaped cylindrical halves enclose the
mixing rotors constituting the mixing chamber. The two ends are closed by the end plates.
Thus, the movements of materials are limited radially and axially. The feeding hopper
introduces the ingredients to be mixed inside the chamber. Inside the rectangular housing,
2
Figure 1.1: Cross-section of an internal mixer [1].
3
the ram can move up and down vertically. Lowering the ram pushes the materials down and
increases the pressure inside the chamber. It is possible to control the ram position further
in order to achieve the desired pressure. At its lowest position, the ram closes the mixing
chamber and the drop door is also kept closed to ensure that ingredients participate actively
in the mixing. The mixed material can be discharged through the drop door. Different
cooling and heating configurations are employed to maintain the desired temperature for
mixing. The most essential part of any mixing device is the set of mixing rotors. The
an internal mixer, rotors are generally counter rotating. The effect of mixing depends on the
mixing rotor design. Mixing rotors are of basically two types: Non-intermeshing mixers
with each other during mixing. Thus, there is always a constant gap between the two
rotors and it is possible to rotate each of them at different speeds. When the rotors run at
different speeds, there is a friction ratio, and when they run at equal speeds, the process
is referred to as even speed mixing. Usually the rotors are rotated counter-clock wise.
Various rotor types and profiles have been proposed and commercially utilized over the
years to improve mixing. The design of the rotors can vary depending on the number of
wings, wings length, shape of the wings, angles associated with wings, wing orientation
and their relative positions. Tangential rotors are well-known for its ease of feeding and
discharge [16].
In the beginning of the 20t h century, Kempeter, Pointon and Banbury filed patents
4
for tangential mixers separately, and today’s tangential mixers have similarities to those. In
1915, J. E. Pointon patented a rotor geometry particularly for mixing rubber (See Figure
1.2) [2]. The invention describes a rubber preparing machine that has a pair of intercom-
municating chambers, each containing one rotor. Each of these rotors has a spiral blade.
The spiral blade starts from one of the rotors, and ends before reaching the other end cre-
ating a passage. The rotors are arranged such that the passages from opposite rotors are
at the opposite ends. These passages help material transfer between the chambers. The
invention claims that such movement of material can be very effective in the mastication
(a) (b)
heavy plastics [3]. This invention is very similar to Pointon’s invention, but each rotor
has two blades, instead of the one in Pointon’s case. The blades on the rotors create acute
5
angles with stationary surfaces, and when rotated, presses the rubber against the surfaces.
The rotation of the blades pushes the rubber, and imparts rolling, kneading and squeezing
forces. Each rotor blade starts from the rotor ends, and terminates at the middle of the
rotor (See Figure 1.3. The blades move the materials from the rotor ends towards the
middle of the chamber applying a drag force. Banbury also suggested the use of a two-
cylinder chamber with opening at the middle for transfer of material. Material at the middle
mixes and exchanges between the rotors through the opening at the middle. According
to the patent, a weight (ram) can be applied on the upper part of the chamber to ensure
feeding and mixing of rubber between the rotors. Banbury filed another patent in 1916,
which introduced design of an automatic door and provides more detail on the ram [25].
According to the patent, lower surface of the chamber is replaced with a drop door. It is
possible to lock the door in tightly fit during the mixing period. The ram can be moved up
and down with a piston. The pressure within the chamber can be controlled using the ram
weight. A chute can be used to incorporate materials into the chamber. A water cooling
system was described as well, admitting water through hollow rotor blades and holes in the
chambers. The patent provided more details of rubber mixing machine compared to the
earlier designs.
The very first 4-wing rotor design was given by Lasch and Frei [24]. The 4-wing
rotor of Lasch and Frei was known as N-rotors, and was very popular in the 70’s [16].
Later, D. Z. Tyson and L. F. Comper from Goodyear Tire and Rubber company patented
a second 4-wing rotor design in 1966 [4]. Figure 1.4 shows the two rotors are parallel to
each other, but oriented in reversed direction. Each of these counter-rotating rotors has
6
two long wings at one end, and two short wings at the other end. The wings move the
materials back and forth along the axial direction of the rotors, as well as in transverse
direction. The patent claims to provide homogenously mixed rubber in a shorter amount
of time compared to the earlier inventions. According to the patent, the arrangement of the
wings and reduced cross section at the junction of the wings give better material intake and
Weidmann and Schmid from then Werner & Pfleiderer filed a patent for rotor
geometry of both intermeshing and non-intermeshing types [5]. The details of intermeshing
rotor have been discussed later in this section. In Weidmann et al.’s non-intermeshing
mixer, the rotors are placed parallel and mirror symmetric (See Figure 1.5). Principle
wings terminate before reaching the other end, and thus create passages. The short wings
at the two ends work as scraping tools, so that there is no stagnant material. The rotors
can have a cylindrical shaft core or conical core with aforementioned wing designs. The
conical cores will create a pressure gradient, not to mention the changing tip clearance
along the rotor axis. The spacious clearance will permit more material and give better
40, a projected wing length to rotor length ratio of at least 0.5, and a clearance between
wing tip and chamber wall to chamber diameter ratio of at least 0.02. The patent claims
that the described design will have better heat transfer, and thus rotor can be operated at
higher speed.
Later around 1988, Nortey from Farrel Corporation along with his colleagues filed
a series of patents on both two wing and four wing rotor designs [1, 6, 26]. Nortey’s first
7
(a) (b)
Figure 1.3: Banbury’s machine for treating rubber or heavy plastic materials [3].
(a) (b)
Figure 1.4: Tyson and Comper’s four wing rubber mixer [4].
8
patent is on two wing rotors. Each rotor has two long wings, each coming from opposite
ends (See Figure 1.6). The rotors can be run at both equal and unequal speeds. The patent
provides experimental results for angles of 0◦ , 90◦ , 135◦ and 180◦ and suggests that for
optimum results a 180◦ phase angle should be used. The patent claims the design will
provide better mixing for first helix angle of 25◦ to 40◦ and second helix angle of 35◦ to
55◦ . Other specifications are rotor orientation of 176◦ to 184◦ , first twist angle of 80◦ to
110◦ and second twist angle of 90◦ to 120◦ . The argument is that the described design
provokes more wing to wing interaction, whereas in the prior designs, there is sitting of
material at the middle of the two rotors due to lack of interactions. The patent also specifies
its superiority in terms of higher fill factor capacity. The second patent of Nortey is for 4-
wing and 3-wing tangential rotors. Figure 1.7 shows each four wing rotor having one long
and one short wing originating from both ends. The patent describes increased twist angles
of wing, stimulating more homogenous mixing through rolling banks of material in axial
direction. Furthermore, the larger wing length ratio increases transfer of material from the
ends of the rotors. Patent claims more dispersion and thorough mixing is possible due to
A 6-wing rotor designed by Harada et al. [27] had the issue of reduced capacity of
the rotor due to an increased number of wings. In an attempt to solve this problem, another
6-wing rotor was developed by Takakura from Kobe steel (Kabushiki Kaisha Kobe Seiko
Sho.) [7]. Each rotor has three equally spaced long wings starting from one end of the rotor
and three equally spaced short wings starting from the other end. The short wings scrape
off the material from the rotor end, preventing sticking of material (See Figure 1.8). The
9
patent argues that the 4-wing mixer does not give uniform mixing within the mixing time.
As a result mixing time gets longer, which may lead to vulcanization. Therefore, often, the
mixing is carried out in multiple stages of cooling and mixing while using 4-wing mixers.
But, Takakura’s 6-wing rotor has better heat transfer, and it is possible to reduce the number
Kimino Inoue and others from Kobe steel patented a rotor [8], which has at least
two long wings with a three-staged tip portion and three clearances in the axial direction
of each wing (See Figure 1.9). The design also consists of two short wings for scraping
off material at the rotor ends. The patent notes that, if the clearance between the cham-
ber wall and rotor tip decreases, shearing force increases and gives better dispersion. But
at higher rotor speeds, as tip clearance decreases, temperature of the mixing material in-
creases. Thus for a low temperature allowable material, the rotors speed would have to
be decreased, which leads to low productivity. As a result, for such material, rotors with
larger tip clearance is used. On the other hand, for high temperature allowable materials,
rotors with smaller tip clearance is used. Therefore, as the materials to be mixed change,
the rotors need to be changed as well, which patent describes as impractical. The objective
of this patent is to design a mixer with a plurality of the tip clearances that can be used in
various mixing conditions. Inoue et al. later provided a new mixing rotor design with plu-
rality of mixing rotors in 2010 [9]. The rotor consists of four long and four short wings (See
Figure 1.10). The long and short wings have different clearance sizes, and twists opposite
of the direction of rotation. The long wings follow a twist curve, such that twist angle be-
comes smaller towards the middle. The invention described claims to prevent diminishing
10
(a) (b)
Figure 1.5: Weidmann and Schmid’s mixing apparatus (a) Rotor geometry, (b) Material
motion [5].
(a) (b)
Figure 1.6: Nortey’s 2-wing non-intermeshing rotors (a) Rotor geometry, (b) Material mo-
tion [6].
11
of performance and productivity resulting from heat generation, and thereby increasing the
At the beginning of the 21st century, Valsamis et al. from Farrel Corporation
developed a noteworthy rotor design shown in Figure 1.11 [10]. The patent specifies that,
rise, and does not provide equally good dispersive and distributive mixing. The four wing
rotor described in the patent consists of two long wings and these long wings have different
helix and approach angles. Similarly, each rotor has two short wings, and these short wings
have different helix and approach angle. The designed rotor promotes more exchange and
interaction of materials between the two rotors and eliminates possible area of material
stagnation. A feature of this invention is that intensity of mixing can be varied by varying
Andreas Limper along with his colleagues from HF mixing group offered a new
design on mixing rotor in 2009 [11]. Limper’s rotor has four long wings, which extend at
least over half of the axial length of the chamber (See Figure 1.12). Each wing has two
sections; first section has a helix angle greater than the second section. The wings divide
material to be mixed into four portions per rotor giving intensive dispersive and distributive
mixing.
Kobe Steel [12]. Yoshida et al.’s design includes two long wings and two short wings. The
short wings are linear and have length less than half of the rotor length. Each short wing
is placed behind each long wing in the rotational direction. The first long wing has a twist
12
(a) (b)
Figure 1.7: Nortey’s 4-wing non-intermeshing rotors (a) Rotor geometry, (b) Material mo-
tion [5].
(a) (b)
Figure 1.8: Takakura et al.’s 6-wing non-intermeshing rotors (a) Cross section of rotor, (b)
13
(a) (b)
Figure 1.9: Inoue et al.’s kneading apparatus (a) Rotor geometry, (b) Material motion [8].
(a) (b)
Figure 1.10: Inoue et al.’s batch mixer (a) Rotor geometry, (b) Material motion [9].
14
angle 15◦ to 35◦ and three different clearances, ranging from large to small in the axial
direction. The second long wing, on the other hand, is nonlinear, and has twist angle of 45◦
Intermeshing mixers creates an intermeshing zone at the middle of the two rotors.
Maximum clearance between them is smaller than the sum of maximum radius of the two
rotors. So, it is necessary to rotate them always at constant and equal speed. Otherwise the
The first intermeshing rotor in the rubber industry was developed by R. T. Cooke
in 1935 [13]. Cooke designed the rotor in an attempt to prevent overheating, reduce manual
labor, and increase productivity as well as to intensify the mixing process. Cooke’s design
has two rotors placed parallel to each other (See Figure 1.14). Each rotor has one long
spiral wing with a relatively wider peripheral face, compared to non-intermeshing rotors,
concentric with the rotor axis. The ends of the spiral are slightly tapered as they meet the
rotor ends. The long wing extends the total axial length of the rotor and occupies half of
its cylindrical side. The other half has two short wings at the two ends. The long wing of
Wiedmann and Schmid designed a intermeshing rotor in 1979 [5]. The rotors were
known as PES rotors [16] and later went through a lot of improvement over the decades.
The objective behind this invention was to increase material transport and obtain uniform
temperature distribution. Figure 1.15 shows the parallel rotors in intermesh with each
other. Each has one long wing at the middle and two wings at two ends of the rotor. Later,
upgraded versions of the design, such as PES 3, PES 5 and PES 6, attracted the attention
15
(a) (b)
Figure 1.11: Valmis et al.’s batch mixer (a) Rotor geometry, (b) Material motion [10].
(a) (b)
Figure 1.12: Limper et al.’s batch mixer (a) Cross section of rotor geometry, (b) Material
motion [11].
16
of tire industry.
Later in 1986, Johnson and his colleagues designed an intermeshing rotor based
on Cookes design [14]. Shown in Figure 1.16b, each rotor has one long wing with wider
wing tip. The wing starts from one end of the rotor and terminates before reaching the
other end creating a passage. There are two small wings placed before and after the long
wings at the two ends. During rotation of the counter-rotating rotor, one rotor’s long wing
gears up in between other rotor’s short wings. This can be seen as foot print in the Figure
1.16a.
In 1987, Passoni from Pomini Farrel, made a noteworthy contribution in the field
of rubber mixer by designing the first Variable Internal Clearance (VIC) intermeshing ro-
tors [15, 24]. Figure 1.17 shows Passoni’s internal mixer. The distance between the axes of
the rotors can be varied to achieve the best possible gap suitable for mixing process. The
rotors can be locked at desired gap, or the gap can vary during a mixing cycle.
Farrel Corporation, Pomini Plastics and Rubber, and Werener & Pfleiderer devel-
oped a lot of designs for rubber mixing over the years. As of 2010 Farrel Corporation,
Pomini Plastics and Rubber, and Werner & Pfleiderer became wholly-owned subsidiaries
of HF mixing group. Some standard products of HF mixing group are the NST tangential
rotor, NR5 intermeshing rotor, GK-N series tangential rotor, and GK-E series intermeshing
rotor. According to the HF mixing group, NST rotor gives perfect combination of material
flow and shear intensity, while NR5 rotor provides high thermal efficiency. The features
of GK-N type rotors (Figure 1.18a) are high fill factor, and less feeding and discharging
time. On the other hand GK-E type intermeshing rotor (Figure 1.18b) is good for tem-
17
perature sensitive compounds and can reduce mixing time and cost tremendously. Other
related products from Kobe steel include 2-wing S tangential rotor, 4-wing H tangential
rotor, 4-wing N tangential rotor, 6-wing VCMT tangential rotor and intermeshing rotor.
ing system uses a combination of a conventional internal mixer with ram and a larger
mixer without any ram. In the conventional internal mixer filler dispersion is done under
ram pressure. This stage might in fact increase temperature. The compound mixed in this
stage is then discharged to tandem mixer for the next stage of mixing. Larger surface area
of the tandem mixer cools down the compound and continues further mixing. Reactive
substances are added to the compound in the tandem mixer. In this way, mixing happens
continuously in a tandem mixer and reduces the time of stopping and changing of mixers
During mixing modeling, while choosing rotor design, it is also essential to understand the
characteristics of the raw materials. Rubber is an elastomer that allows large deformation
and 100% recovery of the deformation without breaking. However, under high stress and
low temperature, rubber can transform into a glassy substance [28]. At high temperature,
rubber acts like viscous fluid. Mixing is done at elevated temperatures, when rubber acts
like a viscous fluid. Although for better dispersion, milling can be done at lower tempera-
ture before mixing [28]. Depending on the application, such as sealing, coating, insulating,
or tire making, different types of rubber or a combination of them can be used. Mostly
18
used rubbers are Natural Rubber (NR), Styrene Butadiene Rubber (SBR), EPDM, Buta-
diene Rubber (BR) and so on. NR is very viscous and has high molecular weight. It is
temperature sensitive, but at low temperature, it has tensile strength and good resistance
to tearing and abrasion. NR is largely used for sealing and vibration damping. SBR is
extensively used in tire industry for its lower temperature sensitivity and high wear, tear
and abrasion resistance. EPDM is mainly a polymer blend of ethylene and propylene. Due
to it’s excellent resistance to heat, chemical and water, it is used for sealing hot water and
steam in faucets. BR is sometimes used in blend with SBR in tire industry. Viscosity
of BR does not change much with temperature, and the sticky nature of BR helps absorb
fillers and oils. Other types of rubber includes Nitrile Butadiene Rubber (NBR), Butyl
Rubber, Fluor Rubber, Chloroprene Rubber (CR), Polyethylacrylate etc. NBR has high
viscosity and good resistance to abrasion; Butyl Rubber is less viscous but very reactive;
Fluor Rubber is highly viscous, resistant to chemicals even at higher temperature; and CR
is very adhesive and corrosive, but gives good dispersion. Masterbatches are another kind
of rubber, which has oil or carbon black blended with them. The ingredients such as, car-
bon black, silica, antioxidant, oil, resins etc., when mixed with rubber contributes to the
improvement of its properties. For instance, carbon black is used as reinforcing filler. De-
pending on which grade of carbon black is being used, it can improve hardness, tensile
strength, tearing and wearing resistance, electrical conductivity and so on. Introducing sil-
ica in the mixing of rubber, reduced rolling resistance by 25%, which in turn reduced fuel
consumption by 5% [16]. Oil is added to the rubber to increase plasticity and durability.
Plasticizer, oil or other liquid additives are added during mixing to prevent glass formation,
19
as well as to provide desired properties to final product. Resins are added to the rubber in
tire industry to increase tackiness. Chemicals such as sulphur, zinc oxide are added as part
of the curing process. All these ingredients can have desirable effect on rubber properties,
if mixed properly.
In order to bring out the best of ingredient properties, it is crucial to understand, observe
and control the process parameters such as: fill factor, speed, friction speed ratio, phase
Fill factor is one of the most important influential parameter on mixing quality and
productivity. Fill factor can be defined as the ratio of volume of chamber space occupied by
rubber to the total volume available to be occupied. Higher filler factor does not necessarily
lead to higher productivity. It may require more time to achieve proper mixing with a higher
fill factor. It would be quite difficult and power consuming to run the rotors at higher speed
at higher fill factor. Besides, high fill factor leads to problems in terms of controlling the
temperature, due to viscous heating. Overfilling may cause some rubber to stick to the
discharge door and chamber throat, and not take part in mixing at all. On the other hand,
too much under-filling is uneconomical. If the amount of rubber is too small, there will be
little interaction between rubber and filler, leading to poor mixing. Therefore, search for
an optimum fill factor is inevitable. With an optimum fill-factor, rubber has enough space
to move around and provide effective mixing through folding and stretching. Commonly
used fill factor falls in between 0.65-0.85, depending on the material properties and rotor
20
type [21]. Fill factor tends to be lower for highly viscous material, whereas it is possible
to use a higher fill factor for low viscous material. Optimum fill factor also gives better
heat transfer from rubber to the cooling surfaces through both conduction and convection.
Rubber is a poor conductor, so conductivity plays very little role in heat transfer. So, the
heat transfer largely depends on the convection, which is controlled by fill factor.
happens in multiple stages. It is essential to control temperature in all the stages, especially
the stages where chemical reaction happens. Besides, temperature can markedly affect
dispersive and distributive mixing of fillers. Dispersive mixing is process of breaking the
fillers into small pieces, and distributive mixing is the process of spreading the fillers homo-
geneously in the entire domain. At high temperature, the viscosity reduces and increases
mobility of material, but reduces stress on the fillers. So, higher temperature can help dis-
tributive mixing and hinder dispersive mixing. Moreover, excessive temperature can cause
damage to the rotor surfaces. Temperature inside the chamber depends on the heat transfer
rate from the rubber to the chamber and the rotor walls. Temperature can be a limiting
factor and can create many issues especially, at high speed mixing. All the surfaces avail-
able are used to cool down overheated material from mixing. Typically, water is circulated
over the chamber wall, ram, discharge door, and through rotors [21]. Water can be sprayed
over the chamber wall and rotor wall or passed through drilled passage. Figure 1.19 shows
two cooling system from Farrel Corporation. Relatively, newer inventions use steam heat-
ing to deal with the condensation on chamber and rotor surfaces at cold environment [21].
21
Optimum rotor speed depends on the design of the rotors as well as the material
to be mixed. In order to maximize the dispersive mixing, the material’s high viscosity
is maintained by running the rotors at lower rpm. At high rpm there is an increase in
temperature, which in turn, reduces the viscosity. Besides viscous heating, viscosity can
also reduce at high shear imparted by the high rotor speed. But, running the rotors at
lower rpm takes longer time to mix. Distributive mixing on the other hand, improves
with increasing rotor speed. Due to this conflicting effect of rotor speed on dispersive and
distributive mixing, the whole mixing process is carried out in multiple stages.
Both of the mixing rotors can be rotated at either same or different speeds. When
two rotors are rotated at the same speed it is known as even speed mixing. When two rotors
run at different speed, their speed ratio is known as friction speed ratio. The difference
in speed creates a velocity gradient between the two chamber lobes that contain the rotors.
This velocity gradient can be very effective in dispersive and distributive mixing depending
Furthermore, most of the times, rotors are run at phase angle with each other. The
difference in the angular position between the two rotors facilitates better movement of the
material resulting in better distribution of the fillers. Depending on the rotor design the
low ram pressure can result in very poor mixing. Moreover, the ram forces the rubber
and ingredient to engage in mixing. The pressure applied by the ram should be sufficient,
so that the pressure inside the chamber cannot cause upraising of the material. A regular
22
periodic upraising of material can happen when the rotors tips come together and push the
materials. Figure 1.20 shows the cyclic movement of ram to maintain its position.
Mixing of rubber with fillers, chemicals or liquid additives happen in four different
stages: mastication, incorporation, dispersion and distribution (See Figure 1.21). Mastica-
tion decreases viscosity of the polymer by reducing the chain length. In the incorporation
process, fillers and additives get soaked into rubber. Mastication and incorporation gener-
ally happen in the early stages of mixing. The topics of interest in this study are mostly
[29]. The complex geometry of rotors, their relative orientation and speed cause a variation
in local forward, backward and transverse flow patterns, and are critical in analyzing dis-
persive and distributive mixing indexes [30]. Fillers, such as silica and carbon black that
are typically mixed into rubber during the manufacturing process are difficult to disperse,
and thus dispersion efficiency has a significant influence on the final product [31]. So a
Dispersion is the process of breaking down the ingredients such as solid agglomer-
ate or liquid droplets, held together by cohesive forces, whereas distribution is the process
of spreading broken agglomerate throughout the fluid domain [32] (See Figure 1.22). Dis-
persive mixing is achieved by prevailing shear and elongational forces in the mixer. While
shear stress is essential in breaking up the (solid and liquid) agglomerates, elongational
23
components characterize the flow strength. Studies have shown that elongation has more
effect on agglomerate dispersion than simple shear flow [33]. In a complex geometry how-
ever, as is the case here, there is a superposition of flow regimes ranging from pure rotation
to pure elongation. This makes it necessary to determine elongation along with shear rate
for assessing dispersive mixing. There are lot of experimental techniques available such
for dispersion measurements, which have their own limitations either in terms of cost or
accuracy [34]. One way to quantify this characteristic in numerical simulations is to cal-
culate the mixing index [35]. Even though challenges still remain in the field to find a
universal statistical quantity that can fully explain the dispersive mixing, quantities such as
mixing index and maximum shear stress, which are presented here, will give a reasonable
Furthermore, while the breakup of the agglomerate into fine particles is important,
the effectiveness of the mixer can also be evaluated by the homogeneity of the distribution
of the small particles throughout the domain, including the stretching of the agglomerate,
i.e. the distributive mixing [36]. In order to characterize distributive mixing, a large set
of massless particles can be used. Qualitatively, distribution can be analyzed from a spatial
distribution of fictitious particles, but more specifically they can be quantitatively assessed
using parameters such as segregation scale, cluster distribution index, axial distribution and
using a length of stretch [36]. This study uses all these parameters in evaluating mixing
24
1.2 Investigation of rubber mixing
Several studies have been conducted both experimentally and numerically in the past, in
order to understand effect of different parameters on the flow field and mixing efficiency.
For instance, Freakley and coworkers made a noteworthy contribution in the field of rubber
and Idris proposed to replace the front and side of mixing chamber with glass, so that move-
ment of the material could be analyzed [40]. They argued that, even though measurement
lems. Along with flow visualization, pressure was also obtained from experiments, and
stress was determined using a mathematical model to compare fill factors. Higher pressure
was detected for higher fill factors. Freakley along with Patel later performed a detailed
analysis of flow and mixing characteristics for different fill factors with instrumented BR
Banbury and biconical rheometer [17]. Mathematical model was also developed in an at-
tempt to determine velocity, stress, and power distribution. The authors found out that at
high rotor speeds and low fill factors distributive mixing was poor. Later, they also stud-
ied the pressure and temperature developed in rubber mixing with intermeshing rotors [41].
Two rotor designs were studied at selected fill factors and rotor speeds. The authors found a
high shear stress distribution at the clearance between rotor tip and chamber wall, and also
in the rotor channels. The authors also discovered material to be viscous in the channels
25
ing of rubber. Tokita and White studied behavior of gum elastomer in a two-roll mill at
different temperature [23]. The study showed that at high temperature, elastomers behave
like polymer solution and at low temperatures, they can crumble into powder. Also math-
ematical analysis was carried out to compute stress and velocity. In the following years,
White and coworkers did many experimental studies to visualize and understand flow of
rubber and other elastomers mixing with fillers and additives. Freakley et al.’s experimen-
tal setup was adopted by Min and White to visualize motion of various elastomers and
molten plastics in Banbury mixer [42]. SBR and BR found to be tearing when stretched,
causing stagnant regions below the ram. NR does not tear when stretched, but rather bands
on the rotor circumference. Addition of carbon black and oil with different type of elas-
tomers such as BR, NR, and SBR using various rotor designs was investigated by Min and
White [43]. Torque, temperature and power consumption were also analyzed. From obser-
vations of flow of rubber and ingredients, authors were able to detect stretching, tearing and
stagnant regimes. Later, Min used a flow visualization experimental setup to study blend-
Phase morphology of the EPDM/PS blend was investigated using a scanning electron mi-
croscope. At low temperature, EPDM exhibits brittle nature and PS forms bands on the
chamber wall. At high temperature, EPDM forms sheets and PS exhibits liquid like flow.
Morikawa et al. studied flow visualization of gum rubber compounding with carbon black
and oil [45]. According to their research, better homogenization with carbon black is ob-
tained if rubber is pre-masticated, and oil is added at the later stage of mixing. Later, the
authors investigated compounding of blends of elastomers with carbon black and oil [46].
26
Color pigment was added for homogenization study. Authors detected improvements in
mixing, when carbon black and oil were added to blends of elastomers. Other work from
Kim and White include study of the effect of different fill factors by measuring torque,
pressure, number of circulations per fill factor, and flow field analysis [47]. Koolhiran and
White compared intermeshing and tangential rotors in dispersing carbon black, and found
Kawanishi and coworker also did some extensive experimental flow analysis on
rubber mixing, by changing different parameters of rotor design [48]. Motion of pig-
mented silicone was observed using a glass window at the front of the chamber. It was
demonstrated that, the homogenization time became shorter with the increase of rotor tip
clearance, overlap ratio of the wings and twist angle of wings, and decrease of ratio of rotor
speeds. Influence of rotor designs and mixing conditions on physical properties of rubber
compound was studied to predict the relationship that existed between them [49]. Most of
the flow analysis experiments were done using a scale-down laboratory mixer. So, Kawan-
ishi et al. investigated parameters of rotor design and mixing conditions to reproduce the
compound from the large scale industrial mixer using a laboratory scale mixer [50].
carbon blacks experimentally against different operating conditions and material proper-
ties [51]. Griffith et al. studied flow of natural rubber using a Brabender mixer with
Banbury type rotor [52]. Color was used in rubber to study the effect of process parameters
and randomness in the flow [53]. The dependence of mixing energy on different types of
fillers was studied experimentally by Leblanc and Lionnet [54]. Toh et al. experimentally
27
studied the effect of twist direction of rotors in the mixing of butadiene rubber with ZnO,
using three kinds of rotors [55]. Kaewsakul et al. experimented to optimize mixing tem-
perature and mixing interval of silica-silane-rubber [56]. Ghari et al. did experiments to
compare single phase and dual phase fillers in natural rubber as well as investigated effect
of sequencing of dual phase fillers [57]. Coming across all these researches, the importance
to study flow of rubber, as well as dispersive and distributive mixing is fittingly realized.
Although these experiments do provide invaluable real data, they are expensive
and difficult to set up. Besides, it is not possible to replicate the exact industrial situation in
the experimental setup and obtain useful data from it. It is almost impossible to determine
dispersive and distributive mixing capability from the experimental data directly. On the
other hand, numerical studies can provide statistical tools that can help understand the
effect of parameters more rigorously, without disturbing the flow field, which in turn can
assist in narrowing down the choices for experiments. Due to recent improvements in
computing techniques and resources, modeling of flow and the associated mixing processes
are capable of providing comprehensive data that can help optimize parameters and rotor
Over the last few decades there have been several studies featuring 2D and 3D
chambers including extruders and roll mills that employed both in-house and commercial
CFD codes. For instance, the need for a numerical model in roll milling was realized by
Kiparissides and Vlachopoulos in 1976 [58]. The authors used the Galerkin method to
perform finite element analysis of Newtonian and non-Newtonian flow at the nip region
28
of roll mills. Non-Newtonian behavior of fluid was defined by both power-law and hyper-
bolic tangent model. The lubrication approximation was used to determine pressure for
both symmetric and asymmetric calendering. In spite of the limitations related to the 2D
approximation, great effort was offered in the field of numerical analysis of rubber mixing.
Along with experimental studies of rubber mixing, White and coworkers con-
ducted some numerical analysis as well. Hydrodynamic lubrication theory was used to
stimulate Newtonian flow in internal mixer and modular corotating intermeshing twin-
screw extruders [59]. Pressure profile and flow field was determined in this isothermal
fully filled simulation. Kim et al. also attempted to simulate mixing of Newtonian fluid
using two counter rotating non-intermeshing rotors in order to obtain global circulation pat-
tern [60]. Kim and White later conducted non-isothermal simulations of non-Newtonian
flow using 3D internal mixer [61]. The limitation here was the use of unwound geometry,
which neglects the curvature of the rotors, which in turn might affect the conclusions on
mixing.
Yagii and Kawanishi noted that, it is not possible to measure flow pattern, shear
strain and stress, and pressure experimentally [62]. Flow visualization through transparent
walls is not perfect either. So, they decided to model mixing of rubber using finite element
method. The chamber was assumed to be fully-filled, and the material was assumed incom-
pressible and a power law liquid. The model helped in better understanding of dispersive
and distributive mixing. The model developed was 2D, and hence did have the limitation
29
ber mixing. Manas-Zloczower and Feke studied dispersive mixing of the agglomerates
in four different flow fields: simple shear flow, pure elongational flow, uniaxial flow and
biaxial flow [63]. The study found pure elongation to be more efficient in agglomerate
breakup than simple shear, biaxial extension being the most efficient. Cheng and Manas-
Zloczower studied the velocity and pressure profile of a 2D Banbury mixer using the fluid
dynamics analysis package, FIDAP, that implements a finite element method [64]. They
also studied effects of design and processing variables such as rotor tip gap, rotor speed and
speed ratio for a Banbury mixer calculating mixing index, shear rate and shear stress [65].
Mixing index, a parameter introduced by the authors, is now widely used for assessing dis-
persive mixing. The authors’ key contribution was the development of parameters such as
flow strength, mixing index and length of stretch for rubber mixing to quantify dispersion
and distribution characteristics [35, 39]. Later, the authors looked into the effects of rota-
tional speed on the distributive mixing, by calculating length of stretch and area of stretch,
and concluded that higher rotational speed increases distributive mixing efficiency [38].
Wong and Manas-Zloczower studied three different phase angles along with uneven speed
cases to determine optimum operating modes [37]. They used particle trajectory snapshots
functions for particles, and from that cluster distribution indices was introduced for a quan-
titative assessment of distributive mixing. For uneven speed cases, authors found speed
ratio close to 1 to be give the best mixing. Yang and Manas-Zloczower explored the ef-
fect of a changing gap on distributive and dispersive mixing performance of the Variable
Intermeshing Clearance (VIC) mixer by calculating cluster distribution index, mixing in-
30
dex, shear rate, and also observing pressure and velocity profiles of the mixer [33]. Yao
and Manas-Zloczower studied two different design of VIC in comparison to original de-
sign [66]. Dispersive mixing was evaluated in terms of shear stress and elongational flow
components, whereas distributive mixing was evaluated in terms of particle trajectory snap-
shots and cluster distribution index. VIC with enlarged chamber and smaller clearance
showed better dispersion and distribution. Authors also investigated distributive mixing in
the axial direction for four different designs of continuous mixer [67]. More studies from
Manas-Zloczower along with others, include modeling of the batch mixer with a simplified
double-Coutte flow geometry in the laminar and turbulent flow regimes [68], development
of mixing index to study dispersive and distributive mixing in scale-up twin-flight sin-
gle screw extruder [32], and development of method using Renyi entropy to characterize
processing conditions [69, 70]. However all of these studies were based on the assumption
that the mixing chamber was fully filled in contrast to reality where chamber is partially
filled.
Yao and Manas-Zloczower also studied the shape of the free surface at the bank
and nip region of the two roll-mill using finite element method. The shape of the in-
terface was guessed initially and then changed at each iteration until convergence was
achieved [71]. The disadvantage of using a 2D model is that axial flow and distribution
are completely neglected, which in turn might affect the conclusions with regard to dis-
A slightly different approach was adopted by Gramann and Osswald [72]. The
31
boundary element method was used to simulate polymer mixing process in Banbury type
mixer. The boundary element method eliminates the cumbersomeness associated with
meshing and re-meshing and gives more accuracy for higher order derivatives. Hutchinson,
Rios and Osswald later used the same technique to develop a new mixing index and to com-
pare performance for different rotor designs and mixing conditions [73]. Rauwendaal et al.
also used the boundary element method to investigate dispersive mixing in extruders and
internal mixer. However, these simulations ignore the axial distribution and non-Newtonian
effect of polymers.
Furthermore, the slip-stick boundary condition was analyzed in the study by Al-
steens et al. [74], and a parametric study of twin screw extruder was conducted as well [75].
Dispersive mixing of agglomerates using the mass density function was characterized by
Collin et al. [31]. Salahudeen et al. performed a comparative study of three different rotor
designs: cam, banbury and roller, through observations of velocity profiles and particle
trajectories [76]. Simulations were conducted using the FEM solver Polyflow and mesh
superposition technique was used for rotational motion of the rotors. The Carreau-Yasuda
model was used to represent the rubber viscosity, and mixing performance of the rotors
was quantified using length of stretch, instantaneous efficiency and time average efficiency.
Roller rotors were found to be more effective compared to the other two. Salahudeen and
others also studied the optimization of rotor speed using cam rotors based on stretching,
efficiency and viscous heating characteristics. Authors found that while viscous heating
and length of stretch increased with rotor speed, the efficiency decreased [77].
High viscous fluid mixing and extrusion is also of interest in the food industry.
32
Dhanasekharan and Kokini carried out numerical simulations of wheat dough extrusion
using a commercial finite element package known as Polyflow [78]. Effect of scaling was
explored by varying geometric parameters of single screw extruders. Connelly and Kokini
analyzed the effects of rheology with different viscoelastic and shear thinning models, and
evaluated the effectiveness of mixing using statistical quantities such as segregation scale,
cluster distribution index, mixing index, length of stretch and efficiency of mixing [79].
They also examined the ability of single and twin screw extruders for dough mixing using
finite element methods [36]. More recently, the effect of the change in paddle stagger
angle on the velocity distribution of the corn syrup was studied by Vyakaranam et al. [30],
and the effects of mixer speed, flow rate and paddle angle for mixing of wheat flour were
investigated using velocity vectors, shear rate and mixing index by Rathod et al. [80].
Nassehi and Salemi also developed a finite element model based on Galerkin
method to simulate non-isothermal viscometric flows in rubber mixing [81]. The Power-
low method was used for determining the non-Newtonian viscosity of rubber. Flow pattern,
temperature and pressure profiles were obtained and compared for different blade geome-
tries. Their model provided insights on necessity and validity of using numerical method.
Much of the published literature dealing with the mixing processes of polymer as-
sume a fully-filled chamber, in order to avoid the complexity associated with the calculation
of free surface, when in reality the chamber is never fully-filled. There are a few studies
on partially filled chambers, including Nassehi and Ghoreish’s free surface simulation us-
ing volume of fluid (VOF) method in Eulerian framework [82]. Fluid in the mixer was
considered as incompressible and air phase was considered compressible. They were able
33
to successfully predict the flow pattern in transient non-isothermal, non-Newtonian flow
simulation. Another partially-filled study done by Nassehi and Ghoreishy [83] was a finite
element simulation of internal mixer with long blade tips, where a Eulerian-Lagrangian
method for free surfaces was used in order to solve mixing of non-Newtonian fluid with
intermeshing type mixers in a 2D model. But these simulations ignored axial distribution,
partially-filled co-rotating twin screw extruder and compare with the fully-filled case [84].
The 3D simulation model used a finite volume and Eulerian multiphase method, while the
free surface flow was solved using the local volume fraction of a cell. One of the more
recent studies on a partially-filled internal batch mixer was a CFD study of the distributive
mixing characteristics for two different geometries and two different fill factors [85], where
a volume of fluid (VOF) method was used to calculate the free surface flow.
1.3 Objective
and analyze the effect of different rotor designs and operational conditions such as phase
angle and speed ratio through parametric studies of batch mixer in a chamber partially filled
with rubber. Phase angle is the angle at which the two rotors are oriented with respect to
each other during mixing. Since there has never been a computational investigation of
phase angle in partially filled chamber, here the effort was made to offer a better judgment
gle, is different at the different instances of a revolution can potentially improve dispersive
more randomization [20]. To the author’s knowledge, there have not been any 3D stud-
ies investigating the effect of different speed ratios under partially filled conditions. The
current study attempts to demonstrate this phenomena by considering two different speed
ratios: 1.125 and 1.5, in comparison with even speed (i.e. speed ratio of 1.0) in terms of dis-
persive and distributive mixing. Finally, it is also objective of this research to assess three
different rotor designs, the 2-wing, 4-wing A and 4-wing B rotor, in terms of its overall dis-
persive and distributive mixing characteristics for partially filled chamber considering the
fact that length, angle and number of the wings, and also their position with respect to each
other, control the material movement in the axial and transverse directions and randomizes
35
(a) (b)
Figure 1.13: Yoshida et al.’s kneading rotor (a) Rotor geometry, (b) Material motion [12].
36
(a) (b)
37
(a) (b)
(a) (b)
38
(a) (b)
39
(a) (b)
Figure 1.18: (a) GK-N series tangential mixers, (b) GK-E series intermeshing mixers,
Figure 1.19: Cooling systems used in the rotors (Courtesy of HF mixing group).
40
Figure 1.20: Ram movement [16].
41
Figure 1.22: Distributive and dispersive mixing.
42
CHAPTER II
PROBLEM DESCRIPTION
2.1 Geometry
As a part of the current research effort, phase angle and friction speed ratio were studied
for two wing rotors. Three different designs of the mixing rotors, rotating even speed, were
also studied.
(a) (b)
Figure 2.1: 2-wing rotor geometry: (a) Top view, (b) Cross-section at the middle.
The orientation of rotors with respect to each other plays a big role in the mix-
ing of rubber. Here, three different rotor orientations will be studied in a partially filled
43
chamber. The geometry used here is a two-dimensional (2D) mid-cross section of a three-
dimensional (3D) model, built following the US patent 4,714,350 [6] (See Figure 2.1).
Detailed specification of the 3D geometry of the 2-wing rotor can be found in Table 2.1.
The rotors in the figure 2.1 are oriented at 180◦ with respect to each other. The other two
orientations being studied are 45◦ and 90◦ (See figure 2.2). While in operation, the left
rotor rotates clockwise, and the right rotor rotates counter-clockwise, about their own axes.
Each of these rotors has two helical long wings originating from opposite ends. These
kinds of rotors are very common in the mixing of polymer materials such as rubber and
plastics with their reinforcements. Frequently-used phase angles have been taken under
In order to study the mixing effect of different speed ratios, a 3D geometry of a 2-wing
non-intermeshing rotor, similar to the rotors of phase angle study, has been employed here.
The rotors are shown in Figure 2.1 and are oriented at 180◦ with each other. The left and
right rotors rotate clockwise and counter-clockwise about their axes, respectively. Each of
the rotor has two helical long wings originating from opposite ends. These kinds of rotors
are widely used in batch mixing of polymer materials such as rubber and plastics. The
rotor design used here follows the US patent 4,714,350 [6]. The specifications are listed
in Table 2.1. See [6] for a description of these parameters. The difference in helix angles
of the two wings is about 7◦ , which causes the wing of one rotor to move the material
off the wing of the other rotor, as they approach each other. For uneven rotor speeds, the
44
(a) (b)
(c)
Figure 2.2: Orientation of the rotors for phase angle 45◦ (a), 90◦ (b) and 180◦ (c).
45
interaction between the opposing wings from the two rotors changes with time. The speed
ratios investigated here are 1, 1.125 and 1.5. The left rotor always rotates at 20 rpm, where
the speed of the right rotor varies with the speed ratios. The rotors’ angular positions at
in the middle, are shown in Figure 2.3 for the case of speed ratio 1.125. Here 0◦ indicates
the initial position (green) of the rotors. Since the right rotor rotates faster than the left
rotor, when left rotor rotates 90◦ , the right rotor rotates 101.25◦ (blue). Thus, a differential
phase angle of 11.25◦ , exists between the rotors. Similarly a 180◦ rotation of the left rotor
corresponds to 202.5◦ rotation of right rotor (red) and so on. Figure 2.4 shows the change
in the differential phase angles with revolutions for speed ratios 1, 1.125 and 1.5.
been conducted here to evaluate the performance of the different rotor designs. All the
geometries used here have been taken from United states patents [1, 6, 26], specifications
of which are listed in Table 2.1. The non-intermeshing rotors studied here have either a
different number or configuration of helical wings and are usually chosen based on the
rubber compounding ingredients used and final products. All these three rotors can be
fitted into the mixing chamber without any further modification. Along with the 2-wing
rotor, mentioned in the previous two sections, the other two rotors considered for the study
are 4-wing A and 4-wing B rotors. Both types of the rotors have two helical long wings and
two helical short wings. In the 4-wing A rotor, one long wing and one short wing originate
46
Figure 2.3: Cross-section of the rotor along a plane in the middle at different time instants
during a rotor revolution, represented by different colors, for a speed ratio of 1.125.
Figure 2.4: Differential phase angles Vs. revolutions of the left (slower rotating) rotor
47
(a) 2-wing (b) 4-wing A
(c) 4-wing B
48
from one end of the rotor, and another long wing and another short wing originate from the
opposite end as shown in Figure 2.5b, whereas in the 4-wing B rotor, both the long wings
originate from one end and the short wings originate from the opposite end (See Figure
2.5c).
2.2 Materials
The material properties used in the current study are presented in this section, following
previous works [31, 65, 86]. A multiphase technique is adopted to model highly viscous
rubber and air. 75% of the domain is filled with non-Newtonian rubber, with density, ρ1 =
1100 kg/m3 . The rest of the mixing chamber is filled with air, with a density, ρ2 = 1.18415
kg/m3 and a constant dynamic viscosity, µ2 = 1.85508 × 10−5 Pa-s. Rubber resides at
the bottom 75% of the chamber. 75% is a commonly used fill factor and mimics better a
practical situation in comparison to fully-filled chamber [87]. Figure 2.6 shows the average
velocity of rubber for fully-filled and 75% filled conditions. The velocity of rubber in the
75% filled case is higher, and this is expected due to the larger space available for the
material to move. Besides, due to the high viscosity of the rubber, there is higher amount
of plug flow in the 100% filled chamber. Thus, prediction of the distributive and dispersive
mixing statistics will not be a true representation of a real situation, if the chamber is
assumed to be fully filled. In reality, the rubber compound contains ingredients such as
antioxidants, filler, oil etc, which are critical to the desirable properties in the final product,
and the entire mixing process is carried out in multiple passes. However in the current
for 5 revolutions.
50
Table 2.1: Rotor Geometry Specification
51
CHAPTER III
MATHEMATICAL FORMULATION
In the numerical simulations presented here, the laws of conservation of mass and mo-
mentum are solved using a finite volume technique for incompressible transient flows. The
governing equations including continuity (Equation 3.1) and the momentum (Equation 3.2)
I
v · da = 0 (3.1)
A
Z I I I Z
∂
ρv dV + ρv ⊗ v · da = − pI · da + Tl · da + (fg + fs ) dV (3.2)
∂t V A A A V
In the above equations, V is the volume, A is the surface area, da is the differential face
area, t is the time, v is the velocity vector, p is the pressure, ρ is the density, Tl is the
viscous stress tensor, fs is the surface tension force, fg is the body force due to gravity, I
is the identity tensor and ⊗ is the convolution operator. The viscous stress tensor Tl for
Tl = 2µ(γ̇)D (3.3)
1
∇v + ∇vT and γ̇ is the shear
Here D is the rate of deformation tensor, defined as, D = 2
p
rate, defined as, γ̇ = (2D : D).
Carreau-Yasuda model was found to be best suited in depicting the rheological behavior
52
of rubber [88–90]. The viscosity of rubber, according to Carreau-Yasuda model, can be
defined as:
Here the power constant n = 0.4, the zero-shear viscosity, µ0 = 100, 000 Pa-s, the infinite-
shear viscosity, µ∞ = 1 Pa-s, and the relaxation time constant λ = 10 s. Figure 3.1 shows
Figure 3.1: Change of non-Newtonian viscosity of rubber with shear rate based on Carreau-
Yasuda model
The VOF method has been employed in the current study in order to model the
free surface existing between rubber and air. In the VOF method, both the phases are
53
defined over the whole domain, and are assumed to be immiscible. The fluid properties are
obtained from functions of properties of its constituent phases and their volume fractions,
such as,
ρ = ∑ ρi αi (3.5)
i
µ = ∑ µi αi (3.6)
i
Vi
αi = , (i = 1, 2) (3.7)
V
where, ρi , µi , and Vi are the density, dynamic viscosity and volume of the ith phase. Here
volume fraction α1 corresponds to the amount of rubber phase, while α2 corresponds to the
air phase so that, α1 + α2 = 1. So control volumes with α1 = 1 are fully filled with rubber
and control volumes with α2 = 1 are fully filled with air. The free surface is constituted of
control volumes with 0 < α1 < 1, which usually takes value of α1 = 0.4 to 0.5 [19].
The solutions of the governing equations are obtained for each phase sharing the
same basic principles of mass and momentum transport. The transport equation for the
d
Z Z
αi dV + αi v · da = 0. (3.8)
dt V S
54
CHAPTER IV
COMPUTATIONAL DETAILS
(a) (b)
(c) (d)
Figure 4.1: 2D unstructured quadrilateral mesh for phase angle 45◦ (a), 90◦ (b) and 180◦
55
conducted using a CFD code, for partially filled mixing chamber with rubber compounds.
Initially rubber is assumed to occupy the bottom 75% of the chamber and top 25% is
occupied by air. For each case, a finite volume mesh is generated for the fluid domain
surrounding each rotor and also for the general fluid domain outside those domains. For
2D phase angle study, finite volume mesh constitutes of unstructured quadrilateral cells
(See Figures 4.1). To ensure sufficient resolution of the mesh near the wall, four prism
layers are created on the surface of the rotors and chamber. A polyhedral mesh gives much
better mesh resolution with less number of cells. So for big 3D geometries, finite volume
unstructured meshes made up of polyhedral, prism and surface elements (See Figure 4.2 &
4.3) are used to reduce total number of cells, in turn reducing simulation time. Adequate
number of cells are created in the clearance between the rotor tip and the chamber wall, and
in the gap between the two rotor tips at the middle. A mesh-independence study was carried
out for three different grid resolutions to choose the appropriate resolution. The time step
was chosen prudently to resolve the motion resulting from the shortest distance traveled
by the rotor based on its rotational speed. A lower time step increases the simulation time
significantly without considerable improvement of results, while greater time step can lead
stationary chamber without remeshing the entire domain at every time step, the moving
mesh method was used. In the moving mesh method, fluid region encompassing the ro-
tors rotates, while rest of the fluid region stays steady, thereby creating a direct interface
between these fluid regions such that mass can transfer through the interface.
56
A no-slip boundary condition was used on the rotors’ surfaces and chamber walls,
such that the velocity of the fluid at the rotors surface is equal to the rotor velocity, and the
velocity of the fluid on the stationary chamber wall is zero. Since the rubber considered
here is highly viscous, the flow Reynolds number, Re, is really low and hence the flow is
the density of the fluid, µ is the dynamic viscosity, Ut is the tangential velocity of the rotor
tip, and Dr is the maximum diameter of the rotor. For the uneven rotor speed cases, the
maximum speed was used for the Re estimation. The simulations also take into account the
effect of surface tension and gravity. Compared to the high viscous force, generated from
To solve for the transport equations for pressure and velocity, a segregated flow
solver was used. The SIMPLE algorithm is used to solve for the pressure-velocity coupling.
The spatial and temporal discretization of Equation 3.2 is second order and first order
accurate respectively [91]. An Eulerian multiphase model is used to solve the multiphase
flow of air and rubber, where air and rubber have their own velocity field but share the same
pressure field. To obtain a sharp interface between two immiscible fluids such as air and
rubber, the high-resolution interface capturing (HRIC) scheme is used for convection terms
in the VOF transport equation [92]. In addition, continuum surface force (CSF) model is
massless particles were used. Massless particles are transported by the continuous phase,
and hence their velocity is equal to that of the surrounding fluid, v p = v. A Lagrangian
57
multiphase model is used to track the massless particles. These particles were inserted
in the rubber phase initially, and hence it is ensured that they stay in the rubber phase
58
(a) 2-wing (b) 4-wing A (c) 4-wing B
(c) 4-wing B
59
CHAPTER V
RESULTS
A commonly-used tool to quantify dispersive mixing is called mixing index (See for in-
stance [19, 31, 32, 36, 79, 94]), defined by Dr. Manas-Zloczower as follows [35]:
|γ̇|
λMZ = (5.1)
|γ̇| + |ω|
where |γ̇| and |ω| are the magnitude of the rate of shear and the vorticity tensors, respec-
tively. The mixing index, λMZ , ranging from 0 to 1, characterizes the elongational and rota-
tional components of a flow, with 0 for pure rotational, 0.5 for simple shear, and 1 for pure
elongational flow [20]. Studies have found that for a highly viscous fluid like the rubber
considered here, elongational flows are more effective in droplet as well as solid agglomer-
ate dispersion than simple shear flow, [33,95]. However, mixing index is not time invariant
and multiple times need to be considered within a revolution for an appropriate analysis.In
fact, mixing index is more like a global quantity that discriminate between various designs
and operational parameters, and does not provide enough information about local defects
in the flow field [79]. To assess dispersion, one has to consider the shear rate along with
the mixing index, since higher shear is also needed for effective dispersion [33, 96].
60
5.1.2 Cluster Distribution Index
One of the methods to quantify the goodness of mixing is to compare the calculated mixture
distribution with the ideal one. A statistical quantity to measure the difference between
the two distributions is known as cluster distribution index. The cluster distribution index
indicates the quality of mixing of a cluster of ingredients in a fluid domain [36]. The cluster
distribution index can be calculated using a discrete pairwise correlation function defined
as [20]:
Z r+∆r/2
2
f (r) = ∑ δ (r0i + r)δ (r0i ) = c(r)dr (5.2)
N(N − 1) i r−∆r/2
where f (r) is the coefficient of the correlation function for distance between particle pairs
and thus represents the probability of finding a neighboring particle at the range of r − ∆r/2
to r + ∆r/2 of the ith particle located at ri0 . δ (r) = 1 indicates that the particle is present,
and otherwise, δ (r) = 0. c(r) is the coefficient of the probability density function of the
correlation function, f (r), and the area under curve of c(r) is constant, irrespective of the
Here rmax is the largest dimension of the geometry. So no correlation exists between the
particles separated by distance larger than rmax , i.e. c(r > rmax ) = 0. Hence, the cluster
value of ε depends on the initial position of the particles and the number of the particles.
61
With all the particles clustered together initially (at time t = 0), as is the case in this paper,
ε will have a relatively large value. With time, as the particle distribution gets closer to the
The value of ε is in fact sensitive to the number of the particles and the initial
Since all the particles are clustered together initially in the form of a squared, ε will have a
The contribution of the axial drag flow created by the helix wing can be estimated using
axial distribution. One of the major components in the distribution process results from
the transfer of material in the middle to the front and back of the chamber, or in other
words, the axial distribution. Axial distribution represents the homogeneity of the particles
distributed on the planes perpendicular to the axial direction or axis of the rotation of the
rotors (i.e. the z-axis) [97]. In order to calculate axial distribution, pairwise correlation
The probability density function of pairwise correlation function was then com-
pared with the ideal distribution. The equations for calculating the axial distribution index
are similar to the cluster distribution index, and are given by:
Z z+∆z/2
2
f (z) = δ (z0i + z)δ (z0i ) = c(z)dz (5.5)
N(N − 1) ∑
i z−∆z/2
62
and Z ∞
[c(z) − c(z)ideal ]2 dz
0 Z
εz = ∞ (5.6)
[c(z)ideal ]2 dz
0
scale of segregation, which closely relates to the properties of mixture [98,99]. Tadmor and
Gogos used the scale of segregation to describe the texture of the mixture [18]. Connelly
and Kokini also used this parameter to quantify the distributive aspect of mixing [36]. The
Z ζ
Ls = R(r)dr (5.7)
0
where
0 00
∑N
i=1 (xi − x) (xi − x)
R(r) = (5.8)
NS2
and
M
2
∑ j=1 x j − x
S2 = . (5.9)
M
R(r) is called the coefficient of correlation and measures the degree of correlation between
that the two particles have the same correlation separated by zero distance, while R(ζ ) = 0
means that there is no correlation between the particles separated by distance ζ . xi0 and xi00
represent the concentration of the ith pair, where x̄ indicates the average concentration. M
is the total number of particles and N is the total number of pairs. S2 is the variance and
calculated from the concentration of all the particles. j indicates index of each particle. Ls
63
is a global measure of the segregated region and does not identify the local defects of the
flow. It is limited by the resolution, which represents smallest length scale within which
Another parameter that is frequently used to quantify distributive mixing and distinguish
between various mixer designs is the length of stretch [35]. Length of stretch can be defined
as the ratio of distance between a pair of particles at time t, to the distance of same particles
|x|
l= (5.10)
|x0 |
where |x| is the distance between two particles at time t, and |x0 | is their initial distance
[100]. Effective mixers shows continuous increase of the length of stretch throughout the
64
5.2 Phase Angles
Figure 5.1 shows the pressure contours for the different phase angles at two different in-
stants of a revolution after 19 revolutions of the rotors. Maximum pressure occurs always
at the front of the rotor tip, while minimum pressure occurs at the back of the rotor tip.
Pressure gradient at the rotor tip depends on the relative position of the tip with respect
to the chamber wall. The pressure gradient is maximum when the rotor tips are inside the
left chamber lobe or the right chamber lobe. Due to higher pressure gradient, dispersion of
the fillers occur at the rotor tip and around the tip corners. It is also noticeable that higher
pressure, which will favor dispersion, occurs at 84◦ rotation as compared to 360◦ . Among
the different phase angles, the materials in the case of 180◦ experience the highest pressure
as the rotor tips come together at the middle, which again is more favorable for dispersion.
The flow behavior in the entire domain, which mainly depends on the rotors’ ge-
ometry and their relative orientations, can be analyzed with the help of velocity contours.
Velocity contour lines for three phase angles are shown in Figure 5.2 at 84◦ and 360◦ ro-
tations of the rotors after 19 revolutions. The velocity contour lines are colored based on
the velocity magnitude. Contours show that the materials are moving in a circular motion
following the rotation of the rotors. The velocity is maximum along circular line created
by the rotor tip. When the rotor tips are at the center of the two chamber lobes, material
exchange occurs. This is evident from the velocity contour lines around the rotor tip when
it wipes of the material from the other rotor causing the material to transfer to the other
65
(a) (b) (c)
Figure 5.1: Pressure contours for phase angles 45◦ (left), 90◦ (middle) and 180◦ (right) at
84◦ (top) and 360◦ (bottom) rotations of the rotors after 19 revolutions.
66
chamber lobe [See figure 5.2a, 5.2b, 5.2e].
The exchange of materials between the two chambers helps in distribution. The
flow is more streamlined in the 180◦ phase angle case, and thus plug flow is dominant
here. For the 45◦ and 90◦ phase angle cases, the circular streamline pattern is interrupted
by the rotor tip and hence the flow is less streamlined in these cases compared to the 180◦
phase angle case. At 84◦ position of the rotors, for the case of 180◦ phase angle, the two
rotor tips come together at the center and squeeze the materials, thus potentially helping in
dispersion.
Figure 5.2: Velocity contours for phase angles 45◦ (left), 90◦ (middle) and 180◦ (right) at
84◦ (top) and 360◦ (bottom) rotations of the rotors after 19 revolutions.
67
5.2.2 Dispersive mixing
A proper dispersion of fillers and other ingredients, through break up of the liquid droplets
and solid agglomerates, is essential in order to obtain homogeneous mixing. This gradual
reduction of filler and agglomerate size by overcoming the cohesive forces, is a critical step
in any mixing process. Many properties of the final product such as tear, abrasion and wear
resistances, and tensile strength depend on the degree of dispersion [101]. A quantitative
study of the filler dispersion is of great importance due to the complexity and difficulty
in dispersing fine and small structured particles of fillers like carbon black and silica [31].
Even though the methods available for quantifying dispersion, have their limitations, the
quantities such as mixing index and maximum shear stress described here, together can
Figure 5.3 shows a histogram of mixing index that indicates the percentage of cells con-
taining only rubber for ranges of mixing index. The histogram was calculated over a period
of 15 rotor revolutions of the rotors. There is a high percentage plug flow in all the phase
angle cases, indicated by low mixing index values (λMZ < 0.1). This is mainly due to the
2D simulation. Since there is no axial movement of the fluid, a large amount sticks to the
rotor wall and rotates with it. There is still a high number of cells with λMZ values that
fall between 0.1 to 0.4, which indicates less dispersive mixing capability. This is especially
evident in the case of the 45◦ phase angle orientation with approximately 39.5%. Within
this range the flow is mainly a combination of plug flow and shear flow, with plug flow
68
being dominant. The fluid near the surface of the rotor primarily experiences this range of
mixing index values, resulting in poor dispersive mixing. For a more effective dispersive
mixing behavior, these fluid elements need to move to the high shear stress regions. The
180◦ phase angle case has the least number of cells (approximately 47%) in the range of
0 < λMZ < 0.3 followed by 90◦ (approximately 48%). 45◦ phase angle, on the other hand,
has highest number of cells in that range (approximately 50%). Thus, materials in 180◦
phase angle has least possibility to experience plug dominant flow and 45◦ phase angle has
highest possibility.
0.4 and 0.7 for all the three cases, which in turn characterize a shear flow behavior. More
specifically, the percentage values are 33.5, 32.5 and 32 for the 180◦ , 90◦ , and 45◦ phase
angles, respectively.
which is the most effective for dispersion [33,96]. Elongational flow is mostly experienced
by fluid elements at the gap between the rotor tip and chamber wall, and also in the middle
of the two rotors, when both the rotor tips squeeze materials. The percentage of cells
experiencing elongational flow is very low with the 90◦ phase angle case having the highest
percentage (≈ 9.5%), closely followed by 180◦ case (≈ 8%) and the 45◦ case, which has
the least value of 7.5%. However for effective dispersion, these elongation forces have to
be accompanied by strong shear rates. Therefore, to conclude the observations from Figure
5.3 related to the mixing index calculations, the phase angle of 180◦ and 90◦ perform better
than 45◦ , with 90◦ emerging as the winner by a slight margin. The 180◦ case has the highest
69
percent of cells in the mixing index range 0.4 to 0.7, simple shear, and also high percent
of cells (slightly less than that of the 90◦ case) in the elongational flow range. The 45◦
case performs the worst with the highest percentage of plug flow and lowest percent of
elongational flow.
Figure 5.3: Histogram of mixing index calculated over last 15 revolutions for phase angles
Along with elongation, primarily quantified by mixing index, shear stress evaluation is
essential as a measurement of the dispersion effectiveness in the domain [33, 96]. Shear
70
stress is generally higher in the middle of the two rotors and the gap between rotor tip and
chamber wall [33]. Reduction of fillers and other ingredient agglomerates is a function
the shear stress they experience. So the higher is the stress experienced, the greater would
be the amount of dispersed fillers. Thus, to fully understand the dispersive mechanism
of a design, it is imperative to evaluate both mixing index and shear stress magnitude. A
cluster of massless particles is injected into the domain for this kind of evaluation. These
particles are tracked by a Lagrangian multiphase model, and are transported by the con-
tinuous phase, and hence their velocity is equal to that of the surrounding fluid, v p = v.
These particles were inserted in the rubber phase initially, and hence it is ensured that they
stay in the rubber phase throughout the simulation. Figure 5.4 represents the cumulative
probability distribution function of maximum shear stress experienced by the particles for
15 revolutions. The further the curve is towards the right, the higher is the possibility for
the particles to experience larger maximum shear stress, and thus better is the dispersion.
So from the figure it is clear that the 90◦ phase angle case is the furthest towards the left,
and 180◦ is furthest towards the right, implying that the 180 ◦ phase angle case performs
the best in terms of dispersive mixing. In addition, if the critical shear stress required for
breaking a certain kind of filler is knows, it is possible to predict what percentage of that
filler would be broken down at the end of mixing [97]. For example, if 100 kPa is the
critical shear stress, then in the 180◦ phase angle case, at least 30% of the fillers will be
broken down, and 22% and 10% would be the corresponding numbers for the 45◦ and 90◦
71
Figure 5.4: Cumulative probability distribution function of maximum shear stress experi-
enced by particles.
72
5.2.3 Distributive mixing
order to obtain an unbiased assessment of distributive mixing for the three phase angles,
clusters of particles were injected at three different positions. Each cluster is of the same
size and contains 2500 particles each. The three clusters were tracked through out the
simulation and statistics such cluster distribution index and interchamber material transfer
were calculated.
(a) (b)
Figure 5.5: (a) Initial position of the particle-clusters (magenta circular-middle, cyan
diamond-rotor tip gap, black triangle-left center near rotor wall), (b) Particle-cluster di-
mension.
As mentioned above, three clusters of massless particles are injected at the beginning of the
simulation at three different locations. Figure 5.5 shows the initial positions of the three
73
particle clusters and also dimension of the cluster. Figure 5.6 shows the particle distribution
of the three clusters for the three phase angles after 5, 15 and 20 revolutions. In general,
most of the particles are homogeneously distributed over time, while some particles do
stick to the chamber wall. The cluster-1, which is initially in the middle of the chamber,
is distributed into the two chambers after 5 revolutions, with the 90◦ phase angle case
having more clustered particles near the left rotor wall. This is probably due to this cluster
being pushed to plug flow zone by the right rotor. After 20 revolutions, the cluster is more
uniformly distributed in the 45◦ and 180◦ phase angles compared to the 90◦ phase angle.
Cluster-2 was placed initially at the gap between rotor tip and chamber wall, and
due to the high shear and elongation experienced in this region, these particles were dis-
tributed more efficiently and evenly with time. Although after 5 revolutions 180◦ phase
angle had most of the particles in the left chamber, there are less differences between the
three phase angle cases after 20 revolutions. Cluster-3 on the other hand, being placed
initially at the plug flow zone, was very poorly distributed for all the three cases, with the
90◦ phase angle case performing the best. Due to the interference of the right rotor tip with
streamlines of the left chamber, cluster-3 was better distributed in the 90◦ phase angle case.
The phase angle of 45◦ performs the worst, as cluster-3 always followed the streamlines
near left rotor, without transferring any particles to the right chamber.
phase angles of 90◦ and 180◦ seem to have more uniformly distributed particles compared
to that of the 45◦ phase angle for all the three cluster positions. Further analysis of dis-
tributive mixing is presented in the next section, with a quantitative analysis of the cluster
74
distribution.
Figure 5.6: Particle distribution for phase angles 45◦ (left), 90◦ (middle) and 180◦ (right)
One of the main purposes of any mixer is to uniformly distribute the ingredients throughout
the domain; in other words obtain what is known as an optimum distribution. A frequently
75
used quantity to measure how far the obtained distribution is from the optimal distribution
is the cluster distribution index. Wong, Yang and Manas-Zloczower introduced the concept
various designs in rubber mixing [33, 37]. Also, Connely and Kokini used cluster distri-
viscous fluid domain [36, 79]. Cluster distribution index is a statistical quantity that mea-
possible to estimate ideal distribution for any design by assuming all the particles in the
cluster are uniformly distributed throughout the domain. Note that the ideal distribution is
unique, and for the most part, is not time invariant. The difference in the ideal distribution
more, in the current study for the three phase angles, even though the ideal distribution
does not vary more than 1%, obtained distribution of each phase angle case was compared
with the respective ideal distribution for the calculation of cluster distribution index.
In the current study, cluster distribution index has been calculated and plotted
with time for all the three clusters, as shown in Figure 5.7. Due to a disproportionately
large value of ε at the initial time, the first revolution is not considered in the figure. As
mentioned earlier, the cluster distribution index converges to a lower value with revolutions
until it reaches an almost constant value. Among the different cases, the one that reaches
the lowest value in the fastest time has the best distributive mixing characteristics. The
oscillations in the curves are due to the circular motion of the particles as they follow the
streamlines.
76
As mentioned earlier, depending on the position of the cluster, value of ε can be
very high or low. For cluster-1, 90◦ phase angle has a very high value at the beginning and
reduces to 0.175 at the end of 20 revolutions. Phase angle 180◦ performs the best reaching
a lowest value of 0.017 at the end of 20 revolutions, while the 45◦ phase angle case falls
in the middle. Cluster-2 on the other hand, being positioned in the gap between rotor tip
and chamber wall, has a very low value of ε for all the three cases. Although, the plots
fall on top of each other for the three phase angles initially, 90◦ phase angle levels out at a
higher value, but the phase angles of 45◦ and 180◦ keep decreasing even at the end of 20
revolutions.
With regard to cluster-3, 45◦ phase angle performs very poorly with a large value
of ε = 20 at the end of the 10 revolutions. This is due the lack of any disruption of the
streamlines by the rotor tip movement. This causes the cluster to follow the plug flow
along the rotor wall. Even after 20 revolutions, the value is as high as 4.84. With help of
an interfering rotor tip, 90◦ phase angle performs the best with a value of 0.088 followed
Cluster distribution of all the three clusters were calculated taking positions of
all the particles in each cluster at every instant, and then were plotted against revolutions.
Overall cluster distribution index also indicates the mixing between particles from different
clusters. Based on the cluster distribution index for all the three cluster positions, 180◦
45◦ phase angle performs the worst for the cluster-3 position and 90◦ phase angle
performs the worst for cluster-1 position. This can be further confirmed from Figure 5.8,
77
which depicts the cluster distribution index for the three clusters put together. Even though
ε is lower in the beginning for the 90◦ phase angle, ε becomes even lower in 180◦ case
after 13 revolutions. Note that these conclusions are based on 2D simulations and a better
A continuous transfer of materials from one chamber to the other makes the filler and other
three phase angle cases, particles (originally from cluster-2) in the left and right chambers
were identified after 10 revolutions. As was noted earlier in the cluster distribution analysis,
particles of cluster-2 became quite homogeneously distributed after 10 revolutions for all
the three cases, since the cluster itself was placed in the gap between rotor tip and chamber
wall. For cluster-2, the cluster distribution index values similar to each other for all the
phase angles considered here. The percentage of particles in one chamber (left/right) that
transfers to the other is calculated every one second for 10 revolutions and plotted in Figure
5.9. Figure 5.9a illustrates materials transferred from left to right chamber. The 180◦
phase angle transfers the highest percentage of materials, ≈ 13%, and the 90◦ phase angle
transfers least percentage of materials, ≈ 6.5%. Figure 5.9b on the other hand shows a
different trend, where the 45◦ phase angle having the highest transfer of 29%, and the 90◦
phase angle has a higher transfer percentage compared to 180◦ . Also note that, even though
the 90◦ phase angle transfers a high percentage of materials from right to left chamber, the
same is not true when transferring materials from left to right. The 180◦ phase angle
78
(a) (b)
(c)
Figure 5.7: Evolution of cluster distribution index for the three phase angles and three
different initial cluster positions: (a) cluster-1 (b) cluster-2, and (c) cluster-3.
79
Figure 5.8: Evolution of cluster distribution index for all the three clusters combined.
80
case however, has almost an equal transfer of materials from one chamber to the other.
Thus, based on these plots, it can be deduced that the 180◦ phase angle exhibits a better
(a) (b)
Figure 5.9: Transfer of particles from the (a) left to right chambers, and (b) right to left
chambers for three phase angles: 45◦ (green), 90◦ (green), and 180◦ (red).
81
5.3 Speed Ratios
This section presents results from 3D numerical simulations of rubber compound mixing in
a partially-filled chamber equipped with two counter-rotating 2-wing rotors, for three dif-
ferent speed ratios: 1, 1.125 and 1.5. A speed ratio of 1 indicates that the two rotors are at
even speed, whereas the other ratios indicate uneven speeds. A comprehensive analysis is
acteristics. Approximately 4300 particles were injected at the beginning of the simulation
In an effective mixing process, the size of the ingredient agglomerates is continuously re-
duced by breakup of the solid clumps and immiscible droplets, by overcoming the cohesive
bonds that exist. But fine and small structured particles of fillers like carbon black and silica
are difficult to disperse [31], and dispersion of the fillers in the polymer governs properties
such as tensile strength, abrasion, tear and wear resistances [101]. Therefore, a quantitative
A joint probability density function (jPDF) of mixing index and magnitude of shear rate
is shown in Figure 5.10. This quantity indicates the percentage of cells containing only
rubber for ranges of both shear rate and mixing index over a period of 8 rotor revolutions
of the left rotor, after 12 revolutions. Eight revolutions are chosen here because, only for
82
8 revolutions of the left rotor, does the right rotor have an integer number of revolutions in
all the speed ratio cases of 1, 1.125 and 1.5. From the figures, it is clear that, irrespective
of the shear rate, a higher density of the mixing index values exists between 0.5 and 0.6 for
all the three cases, which indicates that the flow is dominated by the shear flow. However,
in the 1.5 speed ratio case, these high-density mixing index regions exist for very low shear
rates (< 10 s−1 ). In the cases of speed ratios of 1 and 1.125, shear rate value ranges from
0 s−1 to 60 s−1 . In all these cases, for mixing index values ranging between 0.5 and 0.6
and shear rates ranging between 10 and 60 s−1 , the 1.125 speed ratio case constitutes the
highest percentage of cells (≈ 18%), followed by the even speed case (speed ratio = 1.0)
at 8%. Note that elongational flow (0.6 < λMZ < 1) is known to be most effective for
dispersion [33].
In the results presented here, elongational flow is more dominant for lower values
of shear rates. Specifically, the speed ratio of 1.5 has the highest percentage of pure elonga-
tional flow (≈ 4%), followed by the even speed case (≈ 3%) and then the 1.125 speed ratio
case (≈ 2%). Elongation generally occurs in the middle of the two rotors, when both the
rotor tips squeeze the agglomerate, thus helping in dispersion. However due to low shear
rates, the elongational forces may not be strong enough for effective dispersion.
In addition, there is a high percentage (≈ 44%) of low mixing index values (0 <
λMZ < 0.2) at low shear rates (< 10 s−1 ) characterizing plug flow especially in the 1.5
speed ratio case. The plug flow contribution is observed to be the least for the 1.125 speed
ratio case, which is about 16%. The low mixing index values occur near the surface of
the rotor, as the material rotates along with it, resulting in poor dispersive mixing. For
83
dispersive mixing to be more effective, these fluid elements have to undergo high shear
rates.
So in summary, the results demonstrate that the speed ratio 1.125 performs the
best, since this case gives the lowest percentage of plug flow and the highest percentage
of high shear rates (> 10s−1 ) high mixing index values. Even though, the speed ratio 1.5
has the highest elongational flow, the shear rate is low, and the percentage of high shear
rate at high mixing index values is the lowest. Furthermore, the speed ratio of 1.5 gives the
Even though mixing index provides some valuable insight with regard to dispersive mixing,
it does not indicate the flow strength for effective agglomerate breakup [97, 102]. Besides
mixing index, another important parameter that illustrates the dispersive mixing capability
is the maximum shear stress [33, 39], especially in a shear-driven flow such as the one
presented in this paper, where the rubber viscosity is of the order of 105 Pa-s. The higher the
shear stress experienced by the ingredient, the greater is the possibility for dispersion. The
cumulative distribution plot in Figure 5.11 depicts the maximum shear stress experienced
by the particles over one revolution after 20 revolutions for the three speed ratios studied
here. The further the curve is towards the right, the higher is the probability of maximum
shear stress. Ideally, one would want to consider a period of 8 revolutions for the analysis
of the maximum shear stress for reasons mentioned in the jPDF analysis above. However,
due to the unavailability of data, only results over a period of one revolution are presented
84
(a) (b)
(c)
Figure 5.10: Joint probability density function (jPDF) of the mixing index and shear rate
calculated over a period of 8 rotor revolutions of the left rotor after 12 revolutions for speed
85
here, and it is assumed that ranking of the three cases does not change whether we look at
one revolution or 8 revolutions. If the critical value required for rupture of particular filler
is known, it is possible to determine the number of particles broken up under shear stress at
least one time [97]. It is apparent from the plot that in the case of speed ratio of 1.5, most of
the particles experienced significantly lower shear stress in comparison to the speed ratio
of 1.125 and the even speed case. For instance, about 33% of the particles experienced
a shear stress higher than 60 kPa for a speed ratio of 1.125, followed by the even speed,
which was about 27%. For the speed ratio of 1.5, less than 5% particles experienced shear
Figure 5.11: Cumulative distribution of maximum shear stress experienced by the particles
86
5.3.2 Distributive mixing
A good spatial distribution of the broken agglomerates represents good distributive mixing.
The dynamics of distributive mixing is assessed in this study through the investigation of
parameters such as cluster distribution index and scale of segregation. These quantities
can be numerically calculated by tracking the motion of a set of massless particles in the
domain. In addition, the axial distribution of these particles is also investigated in order
to analyze the effectiveness of this feature in assisting towards better distributive mixing
characteristics.
In the current study, particles are injected in the form of a spherically-shaped cluster for
all the three speed ratios at the beginning of the simulation. An ideal distribution has been
used as a reference for an unbiased comparison of the three cases. The ideal distribution is
obtained by assuming that all the particles are uniformly distributed throughout the domain
and hence it is the most optimum distribution that can be obtained by the mixer. Fig-
ure 5.12a shows the initial spherical cluster distribution of particles relative to the rotors.
The corresponding probability distribution function for the pairwise correlation function
is shown in Figure 5.12b. The values are zero at the ends, which represent minimum and
maximum dimensions of the geometry. The curve has a very sharp peak because, initially
all particles are lumped together in the shape of a sphere, this representing a small particle
pair distance. Also shown on this plot is the probability distribution function of the ideal
distribution. The purpose of any mixer is to spread out the particles and in the ideal case,
87
(a)
(b)
Figure 5.12: (a) Initial position of the particles (b) Probability distribution of the particle
pair distances at the initial condition (red line); Also shown here is the probability distribu-
tion of the particle pair distances of the ideal distribution (black line with symbols).
88
the curve is more flat due to the particles being homogeneously distributed. It should be
noted that the area under both these curves is the same. Also, the differences in the ideal
distribution for a fixed geometry at different times were negligible, and the changes in the
geometry of the three cases resulted in differences < 1% in the ideal distribution. So a
fixed ideal distribution was used for all the three cases for the evaluation of ε.
Figure 5.13 shows the instantaneous particle distribution at three different times,
i.e. after 5, 15 and 20 rotor revolutions, for the three cases studied here. Qualitatively, after
5 revolutions (See Figures 5.13a, 5.13b and 5.13c), the speed ratios of 1.125 and 1.5 have
their particles more spread out in comparison with the even speed case. This is expected
due to the asymmetric nature of the flow field resulting from the uneven speeds, which
do cause some particles to move faster than others. However after 15 revolutions (See
Figures 5.13d, 5.13e and 5.13f) and 20 revolutions (See Figures 5.13g, 5.13h and 5.13i),
the differences between the cases are not as obvious. This indicates that the uneven speeds
do result in a more homogeneous distribution occurring faster than the even speed case. But
with time the distributions between the three cases tend to be similar to each other. These
Figure 5.14 shows the temporal evolution of the cluster distribution index for the
three speed ratios. Since cluster distribution index depends on the position of the cluster,
the continuous flow field change at the middle of the rotors due to the speed ratio, affects
the mixing performance. The oscillations are due to the circulation of particles in the
flow domain by the stretching and unstretching of the cluster of particles. The oscillation
89
amplitude is lower in the speed ratios of 1.125 and 1.5, since the differential phase angle
of the two rotors interrupts the mostly regular and circular pattern, except very near the
rotor walls, thereby introducing more randomization in the mixing. As mentioned earlier,
the cluster distribution index, ε, is large initially and keeps reducing with revolutions. The
plot does not include the first revolution due to the disproportionately high values resulting
from the particles being lumped together. In first few revolutions, ε reduces faster for both
the uneven speeds (speed ratio of 1.125 and 1.5) compared to the even speed. However, ε
for the speed ratio of 1.5 levels out at a higher value compared to the other two cases at the
end of 20 revolutions. In addition, the speed ratio of 1 and 1.125 decrease with time, with
1.125 converging to the lowest value in the fastest time. Based on the cluster distribution
index, a speed ratio of 1.125 exhibits the best distributive mixing characteristics among all
Figure 5.15 shows the time evolution of the axial distribution index after the 1st revolution,
again due to disproportional large values at the initial time. The lower the value is for the
axial distribution index, εz , the more uniform are the particles along the axis of rotation.
Clearly, at the start, εz for the even speed case is higher compared to the speed ratios of
1.125 and 1.5. This is because initially all the particles are at the middle and velocity
gradient is higher for speed ratios 1.125 and 1.5, since one rotor rotates faster than the
other. At the beginning of mixing, εz in the speed ratio of 1.5 decreases faster than the
others. However, it reaches a plateau at about 0.25, after 11 revolutions. The even speed
90
(a) (b) (c)
Figure 5.13: Particle distribution for speed ratios 1 (left), 1.125 (middle) and 1.5 (right)
91
Figure 5.14: Evolution of the cluster distribution index with time.
case converges at a slightly lower value of 0.24, while the speed ratio of 1.125 continues
to decrease and eventually converges to the lowest value of about 0.20 at the end of the
20 revolutions. This indicates that the speed ratio of 1.125 provides best axial distribution,
To obtain better mixing it is also necessary for the ingredients to transfer from one chamber
to the other as the rotors rotate. To evaluate the amount of transfer, particles injected in the
happens between the left and right chambers within ∆t, the transfer is counted.
Table 5.1 presents the total counts as a percentage of total particles transferred at
92
Figure 5.15: Evolution of axial distribution with time.
least once from one chamber to the other, crossing the middle region between the rotors.
From the table, we observe that the percentage of transfer increases with time for all the
speed ratios. However the rate of this increase is the highest with the speed ratio of 1.125
at 58.95% after 20 revolutions, while it is the least for the speed ratio of 1.5 at 42.8%
after 20 revolutions. These results confirm what was observed with the cluster distribution
index, which also showed that 1.125 and 1.5 had the best and worst distributive mixing
characteristics, respectively.
tributed after 10 revolutions are assigned two different concentrations, values of which are
chosen arbitrarily. The concentrations of particles in the upper half of the chamber are
93
Figure 5.16: Two types of particles with concentration 1 (green) and 0 (black) defined at
Figure 5.17: Front view of the distribution of particles for speed ratios 1 (left), 1.125
(middle) and 1.5 (right) at 12 (top) and 20 (bottom) revolutions; The green and black
94
Table 5.1: Percentage of Particles transferred at least once from one chamber to the other
0th revs. 5th revs. 10th revs. 15th revs. 20th revs.
assigned as 0, whereas those in the lower half are assigned 1, represented by the black and
green particles, respectively, as shown in Figure 5.16. From the front view of the rotors
in Figures 5.17a, 5.17b and 5.17c, it is clear that two revolutions after the particles were
assigned dual concentrations, most of particles are still segregated for speed ratios of 1 and
1.5, with 1.5 showing more. As time progresses, more integration between the particles
does occur for all the cases as seen after 20 revolutions in Figures 5.17d, 5.17e and 5.17f.
However, the particles in the speed ratio of 1.5 are still quite segregated, and more segre-
gated than both 1.125 and 1. Furthermore, Figure 5.18 shows the green and black particle
distribution in a top view for all the cases at two times: 12 and 20 revolutions. Here it
can be seen that particles in the middle are mostly mixed as a result of the interchamber
particle exchange (See Figures 5.18a, 5.18b and 5.18c), especially for speed ratios of 1.125
and 1.5. This is because the uneven speeds create a velocity gradient in the middle. Later
after 20 revolutions, there is more randomness and less segregated regions in the particles
distributed, as the materials are getting stretched and folded (See Figures 5.18d, 5.18e, and
95
5.18f). Overall, the speed ratio of 1.125 has the least segregated regions, especially in the
middle and near the rotor surface (See Figures 5.17b and 5.18b), and the speed ratio of 1.5
has the most segregated regions, thus failing to transfer material in the upper and lower
halves. It has less segregated regions at the middle due to the velocity gradient, but most of
the particles at the top and bottom are segregated (See Figures 5.17f and 5.18f). From these
figures it is expected that the speed ratio of 1.5 will have the highest segregation scale and
1.125 will have the least segregation scale at the end of the simulation. One of the reasons
that the 1.5 speed ratio case exhibits such a behavior is that one rotor rotates much faster
than the other, and the the chamber with the slower rotor is unable to take in material, since
Figure 5.19 shows the evolution of the scale of segregation with revolutions. The
initial value of scale of segregation is higher due to the initial setup of particles (Figure
5.16) in the upper and lower halves. As the particles get mixed with time, scale of segre-
gation reduces and converges to a lower value. After 1 revolution (at 11 revolutions on the
x-axis), the 1.125 speed ratio case has the lowest value and continues to further decrease,
whereas the 1.5 speed ratio has the highest value of scale of segregation. The evolution
of scale of segregation with revolutions also shows oscillations. This can be explained by
the movement of particles in the rotating flow region causing stretching and unstretching
of the material. The speed ratio of 1.125 shows the least oscillations compared to the other
two speed ratios because, among all the three cases, the 1.125 case takes the longest time
(8 revolutions of the left rotor) for the rotor-to-rotor interaction pattern to repeat. After 10
revolutions, the segregation scale of all the three speed ratios converge to an average value,
96
with the speed ratio of 1.125 having the lowest value of 0.007, followed by the speed ratio
of 1 with 0.03, and then the 1.5 speed ratio case with an average segregation scale value
of 0.14. This confirms the qualitative picture of segregated particles seen in Figures 5.17
and 5.18, where the largest regions remained segregated even after the mixing for the 1.5
speed ratio case. On the other hand, the speed ratio of 1.125 exhibited superior distributed
Figure 5.18: Top view of the distribution of particles for speed ratios 1 (left), 1.125 (mid-
dle) and 1.5 (right) at 12 (top) and 20 (bottom) revolutions; The green and black particles
97
Figure 5.19: Evolution of scale of segregation with time.
the three speed ratios, Figure 5.20 presents a bar graph of the major performance param-
eters including, elongational flow components, maximum shear stress, and plug flow for
dispersive mixing, and cluster distribution, axial distribution, inter-chamber particle trans-
fer and segregation scale for distributive mixing. For each parameter, different speed ratios
have been ranked from 1 to 3 based on performances. Since the plug flow contribution
scale. With regard to dispersive mixing the 1.125 case is superior due to its least plug flow
contribution and highest maximum shear stress ranking. While the 1.5 case does have the
low shear-rates. Due to this reason and also the fact that it has the highest plug flow con-
tribution, the 1.5 speed ratio has the worst dispersive mixing characteristics. On the other
98
hand, for distributive mixing, 1.125 is consistently the best in all performance parameters
(a) (b)
Figure 5.20: Summary of (a) dispersive mixing and (b) distributive mixing capabilities for
99
5.4 Rotor Design
ber partially-filled with rubber compounds have been conducted. The rotor types are 2-
wing, 4-wing A and the 4-wing B rotor. The mixing chamber is taken to be partially-filled
with 75% rubber and both the rotors have the same speed of 20 RPM. The primary objec-
tive of this study is to assess and compare dispersive and distributive mixing characteristics
between the rotor types via various quantities such as flow patterns, flow rates, and other
Favorable material movement enhance effective mixing and result in an improved perfor-
mance. Velocity vectors and flow rates are analyzed in the entire domain through various
cross-sections as shown in Figure 5.21, after blanking out the air. Figures 5.22 and 5.23
show the velocity vectors through cross-sections of Figure 5.21c at two different times,
corresponding to 1.2 and 1.67 revolutions. Figure 5.24 shows the velocity vectors through
the axial cross-sections at two different times, again corresponding to 1.2 and 1.67 revolu-
tions. These times were chosen for clearer visualization of the typical material movement,
which solely depend on the different rotor designs. The velocity vectors are colored by
velocity magnitude and big arrows are employed to demonstrate the directions of resulting
bulk movement. A lot of information about material movement can be further affirmed by
analyzing flow rates across sections. Studies have been previously conducted with flow
rate analyses for understanding the effect of rotor design and fill factor [85]. Flow rate
100
plots corresponding to the 15-17 revolutions are shown here, so that their magnitudes are
unaffected by the initial condition. A change of flow rate from positive to negative and vice
versa within different phases of a revolution indicates the rubber movement direction and
magnitude. Normalized flow rates across cross-sections shown in Figure 5.21 and colored
by the same cross-sections, are presented in Figures 5.25 (colored by Figure 5.21a) and
Figure 5.21: Cross-sectional planes along the (a) axial, and (b) and (c) transverse directions
for analysis of flow rates (using (a) and (b)) and velocity vectors (using (c)) at time, t=0.
For the 2-wing rotor at 1.2 revolutions and cross section at 25% (See Figure 5.22a),
there is a transverse movement of the material from the right chamber cavity to the left
chamber cavity as the wings give push, and in the opposite direction at 1.67 revolutions
(Figure 5.23a). The flow is vice-versa for the 75 % cross section as shown in Figures 5.22g
and 5.23g. So at the 50% cross-section (in Figures 5.22d and 5.23d) the transverse material
movement in the two directions results in the material being squeezed down. Axial flow
101
(a) (b) (c)
Figure 5.22: Instantaneous velocity vectors colored by velocity magnitude for 2-wing (left),
4-wing A (middle) and 4-wing B (right) for cross-sections at 25%, 50% and 75% (from top
to bottom) of the rotor length at 1.2 revolutions; cross-sections shown in Figure 5.21c.
102
(a) (b) (c)
Figure 5.23: Instantaneous velocity vectors colored by velocity magnitude for 2-wing (left),
4-wing A (middle) and 4-wing B (right) for cross-sections at 25%, 50% and 75% (from top
to bottom) of the rotor length at 1.67 revolutions; cross-sections shown in Figure 5.21c.
103
patterns in Figure 5.24 show material moving forward and backward at the left and right
sides, respectively at 1.2 revolutions (See Figure 5.24a) and vice versa at 1.67 revolutions
(See Figure 5.24d), thus keeping the material in continuous motion. These flow patterns
are further confirmed from the flow rate evolutions over two revolutions shown in Figures
5.25 and 5.26. For the 2-wing rotor, distinct peaks due to the two long wings are seen in
the flow rate evolutions across cross-sections in the front and the back of the rotor (Figures
5.25a and 5.25g, respectively), whereas in the middle (Figure 5.25d), the transverse flow in
both directions results in negligible net flow. Axial flow patterns across the cross-sections
on the left and right (Figures 5.26a and 5.26g, respectively), show the exact same behavior
as well, and again, in the middle (Figure 5.26d), flow is a lot more random.
Flow patterns of the 4-wing A rotor have some differences compared to the 2-wing
rotor. Due to the additional two small wings 4-wing A rotors has additional fluid movement
which possibly intensifies mixing action in axial and transverse direction (Figures 5.22b,
5.22e, 5.22h, 5.23b, 5.23e, 5.23h,5.24b, and 5.24e). So with regard to the flow rates across
axial cross sections, the peaks from the two long wings are no longer as evident as 2-wing
rotor in the 4-wing rotors case for any of the cross-sections, due to a more simultaneous
transfer of material in both directions (See Figures 5.25b and 5.25h). With regard to axial
flow rates on the other hand, the major flow pattern is very similar to the 2-wing rotor (See
Distributive mixing in the 4-wing B rotor is mainly achieved by the long wings
in them transferring fluid from one chamber to another (Figures 5.22c, 5.22f, 5.22i, 5.23c,
5.23f,5.23i, 5.24c, and 5.24f), but more interestingly, it can be seen that, because both the
104
long wings originate from the same end in the 4-wing B rotor, material transfers always in
one direction, rather than switching directions at different rotor phases. This can be further
confirmed from the flow rate plots, where the transverse motion of the material across
the axial cross-sections is similar to the 4-wing A rotors for the axial cross-sections (See
Figures 5.25c and 5.25i) and for the transverse cross-sections, the positive-only value at the
left side (blue) and negative-only value at right side (green) of the flow rate plot indicates
that flow almost does not change the direction on the sides of the rotors, again since both
the long wings are on the same side (See Figures 5.26c and 5.26i). Hence, it is possible
that the 4-wing B rotor will have poor dispersive and/or distributive mixing compared to the
other two rotor cases. However, that must be further proven using other more quantifiable
mixing metrics.
Dispersion of fillers and agglomerates through a process of breaking them down using
stresses and forces in order to overcome cohesive force and interfacial forces is necessary
for the resulting mixture to be homogeneous. Since elongational flow is more effective
than simple shear, mixing index is a quantity that can determine dispersive performance
of a mixer, specially for a high-viscous fluid [35]. There have been lot of studies, where
mixing index was used as a means to quantify dispersive mixing [19, 31, 32, 36, 79, 94].
However, in order to really assess how effective the shear and elongational forces are,
one has to consider the shear rate in conjunction with the mixing index [33]. Figure 5.27
shows the joint probability density of mixing index and shear rate for rubber alone, using
105
(a) (b) (c)
Figure 5.24: Instantaneous velocity vectors of 2-wing (left), 4-wing A (middle) and 4-wing
B (right) rotor at 1.2 (top) and 1.67 revolutions (bottom), for the cross section through the
106
(a) (b) (c)
Figure 5.25: Evolution of normalized flow rate of 2-wing (left), 4-wing A (middle) and
4-wing B (right) rotors for 2 revolutions from 15 to 17, through the cross sections shown
107
(a) (b) (c)
Figure 5.26: Evolution of normalized flow rate of 2-wing (left), 4-wing A (middle) and
4-wing B (right) rotors for 2 revolutions from 15 to 17, through the cross sections shown
108
calculated data collected at multiple time instants over a period of one rotor revolution after
17 revolutions.
Mixing index values close to zero and 0.5 indicate plug flow and shear flow, re-
spectively, whereas, values higher than 0.5 and tending towards 1 indicate elongational
flow. From the figures it is clear that the flow is dominated by the shear flow, as a higher
density of the mixing index values exists between 0.5 and 0.6 for varying shear rates for
all the three rotors. Also, the density of elongational flow is very low for any shear rate
for all the rotors, with the 2-wing rotor being marginally better than the other two rotors.
The elongation occurs mostly in the middle of the two rotors when both the rotor tips push
down the rubber, thereby squeezing and breaking the material. Furthermore, there is a rea-
sonably high percentage of low mixing index values (0 < λMZ < 0.2) at low shear rates
(< 10s−1 ) which indicates plug flow and thereby poor dispersive mixing. This kind of be-
havior is exhibited the most by the 4-wing B rotor about 30% and the least by the 4-wing
A rotor about 10%. Typically this is found near the surface of the rotor and along with
which the material rotates without undergoing breakdown of the fillers and agglomerates.
For effective mixing these fluid elements needs to move to a high shear and mixing index
region. So in summary, the feature to be assessed from this figure is the percentage of node
points having the high mixing index and the high shear rate and according to that metric,
the 4-wing A rotor shows the highest percentage of those points followed by the 2-wing
followed by the 4-wing B rotors. The 4-wing A rotor has about 22% of mixing index values
between 0.5 and 0.6 for the highest shear rates ranging from 10-60 s−1 , with the number
for the 2-wing rotor being at 12% for this same shear rate. The 4-wing B rotor has a max-
109
(a) (b)
(c)
Figure 5.27: Joint probability density of the mixing Index and shear rate calculated over a
period of one rotor revolution after 17 revolutions for the (a) 2-wing (b) 4-wing A and (c)
4-wing B rotors.
110
imum shear rate range of 10-40 s−1 with only about 7% of mixing index values between
0.5 and 0.6 experiencing it. The results demonstrate that, based on the metric presented
here, the 4-wing B rotor has the worst, while 4-wing A rotor has the best dispersive mixing
characteristics.
In order to study distributive mixing of the three rotors, massless particles were inserted into
the computational domain. These particles are tracked using the Lagrangian Multiphase
model and their positions are calculated at every time step using the Eulerian solution of
the velocity at every grid point. The particles were injected initially within a sphere and
were tracked through a certain time. The data and information obtained from particles are
4000 particles were inserted at the middle of the two rotors initially and the snapshots
of the evolution of particle distribution are presented here in Figure 5.28 for a qualitative
evaluation. From the figures, it can be seen that after one revolution of the rotor, some of the
particles (for all 3 rotor types) move to the left, while the rest move to the right due to the
various wing actions. As time progresses and after 10 revolutions, the 2-wing and 4-wing
Again after 17 revolutions it is clear that the 4-wing A rotor shows the most homogeneity
in particle distribution followed by the 2-wing and 4-wing B rotors. This can also be tied
to the highest and lowest shear rate distributions of the shear flow regions in the 4-wing
111
(a) (b) (c)
112
Figure 5.28: Particle distribution for 2-wing (left), 4-wing A (middle) and 4-wing B (right)
at the initial condition, and after 1, 10 and 17 revolutions (from top to bottom).
A and 4-wing B rotors, respectively. The high shear rate for instance is very effective in
breaking down the agglomerate of particles and resulting in the particles being spread out
The particle distribution in the three different rotors are compared to one another using the
the premise that the ultimate goal of any mixing equipment is to spread out all the particles
uniformly throughout the chamber. This distribution solely dependent on the geometry of
the rotor. The ideal distribution is not time invariant but within one revolution the change is
negligible in our case and thus considered time independent. Figure 5.29a shows the initial
distribution of particles, while Figure 5.29b, 5.29c & 5.29d show the simulated distribution
of particles after 17 revolutions along with the ideal distribution for the three rotors. The
the maximum dimension of the mixing chamber. The two 4-wing rotors have a slightly
different ideal distribution compared to the 2-wing rotor, due to the two additional wings,
but in spite of the short wings originating from different sides, ideal distributions are almost
indistinguishable between the 4-wing A and the 4-wing B rotors. The initial distribution
of particles, which is exactly the same for all the three rotors, has a very sharp peak very
at the very beginning of the simulation in a sphere. As time progresses the particles get
distributed throughout the domain, and the sharp peak of initial condition disappears. On
113
comparing the particle distribution after 17 revolutions with the ideal, the 4-wing B rotor
has largest deviation compared to the 2-wing and the 4-wing A rotors, which have similar
This deviation from the ideal distribution can be quantified using the cluster dis-
tribution index parameter defined earlier. Figure 5.30 plots the evolution of the cluster
distribution, ε, with revolutions for all the three rotor designs. It is expected that ε will
usually starts from a high value (depending on the initial distribution) and then decrease
and converge to a value with time. A better rotor design will reach a smaller converged
value in a shorter time. The first revolution has been excluded from the comparisons due
to the disproportionately large value at the beginning of the simulation, resulting from the
particles being packed in a sphere. Plots show that for the 4-wing A rotor, the ε drops to
its minimum (of ≈ 0.07), in the fastest time (≈ 5 revs), whereas, the ε of 4-wing B rotor
decreases to approximately 0.12, after around 10 revs. The ε of the 2-wing rotor decreases
to 0.08 after about 6 revs. These results indicate that the 4-wing A rotor approaches the
ideal distribution in the fastest possible time, thus exhibiting the best distributive mixing
The length of stretch is an indication of the ability of mixers to stretch a fluid element
over time by stretching and folding. The evolution of the mean of the length of stretch for
the three rotors have been plotted with time in Figures 5.31. The mean is calculated from
the particle data corresponding to the material. As the number of particles experiencing a
114
(a) Initial Distribution (b)
(c) (d)
Figure 5.29: Probability distribution of particles (a) at the initial time, and after 17 revolu-
tions for (b) 2-wing, (c) 4-wing A and (d) 4-wing B rotors (solid lines); Also shown are the
corresponding ideal distribution of particles for the three different rotors in (b), (c) and (d)
115
Figure 5.30: Evolution of the cluster distribution index with time.
116
continuous stretching due to the shear increases with time, the mean of the length of stretch
increases. The oscillations in these figures represent the stretching and unstretching of the
fluid element. The figure shows that the average length of stretch is the highest for the
4-wing A rotor indicating again its best distributive mixing qualities, consistent with the
117
5.5 Validation
A mesh independence study was conducted to ensure accuracy and confidence in the results
presented here. Figure 5.32 shows the velocity profile along the x-axis across the center
cross section of the left half rotor of the 2-wing geometry. The mesh independence study
was conducted for 600k, 1.3M and 2M cells at 1.2 revolutions. Assuming the simulation
with 2M cells yields the most accurate results, the error obtained for 1.3M cells case was
about 2%. Hence, all other simulations were carried out using 1.3M cells. In addition, time
step and iteration independence studies were also carried out to assure less numerical error.
Furthermore, the simulation result was qualitatively matched against previously published
Figure 5.32: Velocity profile along x-axis at the mid-cross section of the left half rotor for
experimental results. Freakley and Patel [17] performed experiments to investigate flow
and temperature properties in a internal mixer. The Farrel Bridge BR Banbury, which has
118
two wings, was used for this investigation. The rotors were rotated at 50 rpm at the speed
ratio of 1:1.125. 75% of the chamber was filled with rubber. This particular publication
was chosen for experimental validation, since the Farrel Bridge BR Banbury is very similar
to the 2-wing geometry used in the current study. The measured pressure profile from the
study of Freakley and Patel was plotted against the pressure obtained from the simulation of
the 2-wing rotors at 20 rpm even speed [See Figure 5.33]. At each revolution there are two
peaks, as two wings cross over the point of interest (pressure transducer location), twice
within a revolution. The experimental results have higher peak due to the larger dynamic
pressure head resulting from the larger speed. Other discrepancies are due the differences
Figure 5.33: Comparison of pressure profile obtained from Freakley and Patel’s experiment
at 50 rpm at the speed ratio of 1:1.125 [17] and simulation at 20 rpm even speed.
119
CHAPTER VI
CONCLUSION
A simplified 2D simulation model was developed to analyze the mixing capability of three
phase angles: 45◦ , 90◦ and 180◦ . Later, 3D simulations were conducted for speed ratios
1.0, 1.125 and 1.5. Mixing with three different rotor designs, 2-wing, 4-wing A and 4-wing
B rotors was also investigated to determine their performance. All numerical calculations
were carried out for partially-filled rubber mixing in a rotor-equipped chamber. The cham-
ber was assumed to 75% filled with rubber. The two rotors were rotating at 20 rpm for the
phase angle and rotor design investigations, and the right rotor had higher speed than the
left rotor, in the case of speed ratios of 1.125 and 1.5 for the speed ratio investigations. All
the analysis was performed for evaluating both dispersive and distributive mixing capabil-
1. Phase angles: The flow field was analyzed from velocity and pressure maps in the
domain. The velocity contours showed lot of circulation mainly because of the 2D
geometry being considered. The flow was observed to be more streamlined in the
180◦ phase angle case compared to other two. This is because of the absence of any
disruption of the circular streamlines by the rotor tips in the 180◦ phase angle case.
Pressure contours showed that pressure was maximum at the front of the rotor tip
and minimum at the back. Pressure gradient was maximum at the gap between rotor
120
tip and chamber wall. The phase angle of 180◦ had a higher possibility of dispersion,
due to the squeezing of materials, as the two rotor tips meet each other in the middle.
(a) Dispersive mixing of the three phase angle cases was quantified by calculating
the histogram of mixing index of all the cells, and cumulative distribution function
Mixing index showed that elongation was slightly better in 90◦ phase angle compared
to the 180◦ phase angle, and 45◦ phase angle performed the worst, since the latter
case showed both highest plug flow and lowest elongation. Cumulative probability
distribution function of maximum shear stress showed that fillers in the 180◦ phase
angle case had highest possibility to experience a large maximum shear stress, and
90◦ phase angle showed worst possibility of filler breakup. Hence, this analysis
concluded that 180◦ was superior to the others in terms of dispersive mixing.
tion snapshots, and quantitatively by considering cluster distribution index and in-
terchamber material transfer. Since the results very much depended on the initial
position of the cluster, three different positions were chosen for particle-cluster in-
jection in the current study in order to provide a complete and unbiased judgement
of the distributive mixing characteristics. So the 180◦ phase angle case fared best in
terms of both cluster distribution index and interchamber material transfer, and 45◦
phase angle was the worst. In addition, the distributive mixing performance of 90◦
phase angle was slightly behind the 180◦ phase angle case.
121
The computational model, developed here, has been expanded to apply in 3D sim-
ulations for the rest of the research topics, i.e., analysis of friction speed ratios and
rotor designs.
2. Speed Ratios: (a) Dispersive mixing, which is very vital in breaking up the ingredi-
ent agglomerate and droplets, has been evaluated using statistical tools such as the
jPDF of mixing index and shear rate, and maximum shear stress. The jPDF signi-
fies the plug flow and elongational flow in conjunction with shear rate. From the
jPDF of mixing index and shear rate, it can be concluded that the speed ratio of 1.5
does provide with the best elongational flow components, but they are at very low
shear rates. Hence, these conditions may not be necessarily good in breaking the
agglomerates. This case also has the highest plug flow contribution. Plug flow is un-
desirable for mixing, since, higher the plug flow, the worse is the mixing due to more
material getting attached to the rotor without undergoing any stress. The jPDF also
indicates lower elongational flow for the speed ratio of 1.125, but also lower plug
flow. In addition, the speed ratio of 1.125 demonstrated higher percentage of parti-
cles undergoing maximum shear stress in comparison with other speed ratios, thus
further enhancing ingredient breakup. It should be noted that the maximum shear
stress analysis was based on the data from one revolution of the left rotor, and the
assumption in the current study is that, it is sufficient for the conclusions presented
here. Based on these results, one could summarize that 1.125 and 1.5 exhibited the
122
(b) Furthermore, distributive mixing was qualitatively evaluated using axial distri-
lower halves, each of which is critical in achieving homogeneity throughout the do-
index and segregation scale were investigated for obtaining very insightful informa-
tion about the distributive mixing capability. Cluster distribution index for instance,
evaluates the deviation of the calculated distribution of the particles from the ideal
distribution. Based on the cluster distribution index, the speed ratio of 1.125 was the
closest to the ideal distribution at the end of the 20 revolutions, whereas the speed
ratio of 1.5 was the farthest. Axial distribution analysis also predicted a similar pat-
tern. In addition, the segregation scale showed that the speed ratio of 1.5 levels out
at a much higher value compared to the other two speed ratios, indicating that there
were still a lot of segregated particles at the end of the simulation. The speed ratio
of 1.125 showed a superior performance in that aspect, while the even speed perfor-
mance fell between 1.125 and 1.5. Here again, 1.125 and 1.5 exhibited the best and
3. Rotor Designs: Flow patterns and flow rates across various cross-sections were an-
and motion.
(a) Joint probability density plot of mixing index and shear rate was used for the
quantitative evaluation of dispersive mixing since both elongation and shear are crit-
123
ical for the breakdown of agglomerates. The 4-wing A rotor showed shear flow
mixing index regions (i.e. 0.5 < λMZ < 0.6) having higher shear rates than any of
(b) Distributive mixing was investigated here qualitatively using particle distribution
snapshots, and quantitatively using different particle statistics such as cluster distri-
bution index and length of stretch. Cluster distribution index evaluates the difference
the cluster distribution index decreased with time/revolutions, the particle distribu-
tion got closer to an ideal one. Length of stretch increased with time as particle
moved further from neighboring particles and gets distributed. Oscillatory behavior
of the length of stretch curve implied stretching and unstretching or folding of the
fluid element as the fluid moved from one chamber cavity to the other and in axial di-
rection. The 4-wing A rotor not only reached the lowest value of cluster distribution
index the fastest, but also maintained a lower value compared to other two rotors.
The length of stretch of 4-wing A was also higher compared to the other two rotors.
All of these indicated better dispersive and distributive mixing capabilities of the
4-wing A rotor compared to the other two rotors. From a design perspective, this
was mainly due to two reasons: (i) Compared to the 2-wing, the 4-wing A rotor had
two additional short wings, which randomized and intensified the material movement
further. (ii) Compared to the 4-wing B rotor, the 4-wing A rotor had the two long
wings on the opposite side of each rotor, as opposed to the same side, and hence,
124
resulted in a higher shear rate and a more back-and-forth material movement among
the chambers.
of rotor design on mixing performance, several factors were not considered. For instance,
opportunity remains in the field for further improvement by extending the calculations to
non-isothermal through the additional consideration of viscous heating. Other factors such
as slip boundary conditions [74] and ram pressure also need further investigations. In
future, attempts can be taken to incorporate more of these realistic operating conditions in
order to improve the assessment and analysis of such mixing equipment, thereby providing
a better understanding of the various processes and also paving the way for more improved
designs.
125
BIBLIOGRAPHY
[1] Nortey, N. O., 1988. Internal batch mixing machines with non-intermeshing rotors
of increased performance, May 17. US Patent 4,744,668.
[2] Pointon, J. E., 1915. Machine for preparing rubber., May 4. US Patent 1,138,410.
[3] Banbury, F. H., 1916. Machine for treating rubber and other heavy plastic material.,
Oct. 3. US Patent 1,200,070.
[4] Comper, L. F., and Tyson, D. Z., 1966. Rubber mixer, Jan. 25. US Patent 3,230,581.
[5] Wiedmann, W., and Schmid, H.-M., 1980. Mixing apparatus for kneading of plastic
substances, Nov. 18. US Patent 4,234,259.
[7] Takakura, K., Ureshino, K., Yamada, N., and Kurokawa, Y., 1998. Hermetically
closed kneading apparatus, Aug. 11. US Patent 5,791,776.
[8] Inoue, K., Ureshino, K., Yamada, N., Takakura, K., and Kurokawa, Y., 1999. En-
closed kneading apparatus, Nov. 16. US Patent 5,984,516.
[9] Inoue, K., Yamada, N., Hagiwara, K., and Nakano, H., 2010. Batch mixer and
mixing rotor used in the same, Dec. 21. US Patent 7,854,542.
[10] Valsamis, L., Borzenski, F., Wagner, R., Rapetski, W., and Baurmeister, H., 2002.
Four wing, non-intermeshing rotors for synchronous drive to provide improved dis-
persive and distributive mixing in internal batch mixers, Dec. 17. US Patent 6,494,607.
[11] Limper, A., Keuter, H., Rinker, M., and Berkemeier, D., 2009. Internal mixer for
kneading plastic materials, July 7. US Patent 7,556,420.
126
[12] Yoshida, N., Uemura, M., Nakano, H., Hagiwara, K., Inoue, K., Nishida, M., and
Fukutani, K., 2015. Kneading rotor, batch kneader and method of kneading materi-
als, Jan. 6. US Patent 8,926,166.
[13] Rupert, T. C., 1935. Rubber mixing or preparing machine, Sept. 24. US Patent
2,015,618.
[14] Johnson, F., Homann, H., Rother, H., and Weckerle, G., 1986. Mixing machine,
Feb. 5. EP Patent 0,170,397A1.
[15] Passoni, G. C., 1988. Closed parallel-rotor mixer with adjustable interaxial separa-
tion, Oct. 4. US Patent 4,775,240.
[16] Limper, A., 2012. Mixing of rubber compounds. Carl Hanser Verlag GmbH Co KG.
[17] Freakley, P., and Patel, S., 1985. “Internal mixing: A practical investigation of
the flow and temperature profiles during a mixing cycle”. Rubber chemistry and
technology, 58(4), pp. 751–773.
[18] Tadmor, Z., and Gogos, C. G., 2013. Principles of polymer processing. John Wiley
& Sons.
[19] Liu, J., Li, F., Zhang, L., and Yang, H., 2015. “Numerical simulation of flow of
rubber compounds in partially filled internal mixer”. Journal of Applied Polymer
Science.
[20] Manas-Zloczower, I., 2012. Mixing and compounding of polymers: theory and
practice. Carl Hanser Verlag GmbH Co KG.
[21] Freakley, P. K., 2012. Rubber processing and production organization. Springer
Science & Business Media.
[22] Wildi, R. H., and Maier, C., 1998. Understanding compounding. Hanser Verlag.
[23] Tokita, N., and White, J. L., 1966. “Milling behavior of gum elastomers: Experiment
and theory”. Journal of Applied Polymer Science, 10(7), pp. 1011–1026.
[24] White, J. L., 1992. “Development of internal-mixer technology for the rubber in-
dustry”. Rubber chemistry and technology, 65(3), pp. 527–579.
[25] Banbury, F. H., 1917. Machine for treating rubber and other heavy plastic material.,
May 22. US Patent 1,227,522.
127
[26] Nortey, N. O., 1989. Optimized four-wing, non-intermeshing rotors for synchronous
drive at optimum phase relation in internal batch mixing machines, May 30. US
Patent 4,834,543.
[27] Harada, J., and Nishigai, K., 1989. Rubber-like material kneading apparatus, Oct. 3.
US Patent 4,871,259.
[28] Nakajima, N., 2000. Science and practice of rubber mixing. iSmithers Rapra Pub-
lishing.
[29] Manas-Zloczower, I., Nir, A., and Tadmor, Z., 1984. “Dispersive mixing in rubber
and plastics”. Rubber chemistry and technology, 57(3), pp. 583–620.
[30] Vyakaranam, K. V., Ashokan, B. K., and Kokini, J. L., 2012. “Evaluation of effect
of paddle element stagger angle on the local velocity profiles in a twin-screw con-
tinuous mixer with viscous flow using finite element method simulations”. Journal
of Food Engineering, 108(4), pp. 585–599.
[31] Collin, V., Peuvrel-Disdier, E., Alsteens, B., Legat, V., Avalosse, T., Otto, S., and
Metwally, H., 2006. “Numerical and experimental study of dispersive mixing of
agglomerates”. In Society of Plastics Engineers Annual Technical Conference 2006,
ANTEC 2006, Vol. 2, Society of Plastics Engineers, pp. Pages–908.
[32] Wang, W., and Manas-Zloczower, I., 2001. “Temporal distributions: The basis for
the development of mixing indexes for scale-up of polymer processing equipment”.
Polymer Engineering & Science, 41(6), pp. 1068–1077.
[33] Yang, H.-H., and Manas-Zloczower, I., 1994. “Analysis of mixing performance in a
vic mixer”. International Polymer Processing, 9(4), pp. 291–302.
[34] Nikiel, L., Mason, R., and Wampler, W., 2016. “Advances in filler dispersion mea-
surements”. Rubber Chemistry and Technology, 89(1), pp. 142–153.
[36] Connelly, R. K., and Kokini, J. L., 2007. “Examination of the mixing ability of
single and twin screw mixers using 2d finite element method simulation with particle
tracking”. Journal of food engineering, 79(3), pp. 956–969.
[37] Wong, T. H., and Manas-Zloczower, I., 1994. “Two-dimensional dynamic study
of the distributive mixing in an internal mixer”. International Polymer Processing,
9(1), pp. 3–10.
128
[38] Cheng, H., and Manas-Zloczower, I., 1998. “Distributive mixing in conveying ele-
ments of a zsk-53 co-rotating twin screw extruder”. Polymer Engineering & Science,
38(6), pp. 926–935.
[39] Manas-Zloczower, I., and Cheng, H., 1996. “Analysis of mixing efficiency in poly-
mer processing equipment”. In Macromolecular symposia, Vol. 112, Wiley Online
Library, pp. 77–84.
[40] Freakley, P., and Idris, W. W., 1979. “Visualization of flow during the processing of
rubber in an internal mixer”. Rubber Chemistry and Technology, 52(1), pp. 134–145.
[41] Freakley, P., and Patel, S., 1987. “Internal mixing: A practical investigation of
the influence of intermeshing rotor configuration and operating variables on mix-
ing characteristics and flow dynamics”. Polymer Engineering & Science, 27(18),
pp. 1358–1370.
[42] Min, K., and White, J. L., 1985. “Flow visualization of the motions of elastomers
and molten plastics in an internal mixer”. Rubber chemistry and technology, 58(5),
pp. 1024–1037.
[43] Min, K., and White, J. L., 1987. “Flow visualization investigations of the addition
of carbon black and oil to elastomers in an internal mixer”. Rubber chemistry and
technology, 60(2), pp. 361–380.
[44] Min, K., 1987. “Flow visualization of blending of elastomers and thermoplastics in
an internal mixer”. Advances in Polymer Technology, 7(3), pp. 243–257.
[45] Morikawa, A., Min, K., and White, J. L., 1988. “Flow visualization of internal
mixer factory cycles for rubber compounding”. Advances in Polymer Technology,
8(4), pp. 383–405.
[46] Morikawa, A., Min, K., and White, J., 1989. “Flow visualization of the rubber
compounding cycle in an internal mixer based on elastomer blends”. International
Polymer Processing, 4(1), pp. 23–31.
[47] Kim, J. K., and White, J. L., 1989. “An experimental and theoretical study of star-
vation effects on flow and mixing elastomers in an internal mixer”. Nihon Reoroji
Gakkaishi (Journal of the Society of Rheology, Japan), 17(4), pp. 203–210.
[48] Kawanishi, K., and Yagii, K., 1990. “Flow analysis in an internal mixer: Part ii: Es-
timation of mixing efficiency by batch homogenization time”. International Polymer
Processing, 5(3), pp. 173–177.
129
[49] Kawanishi, K., Yagii, K., Obata, Y., and Kimura, S., 1991. “Relationship between
rotor designs in an internal mixer and physical properties of mixed rubber com-
pounds”. International Polymer Processing, 6(2), pp. 111–120.
[50] Kawanishi, K., Yagii, K., Obata, Y., and Kimura, S., 1991. “Scale-up effect in
internal mixers”. International Polymer Processing, 6(4), pp. 279–289.
[51] Cotten, G. R., 1984. “Mixing of carbon black with rubber i. measurement of dis-
persion rate by changes in mixing torque”. Rubber chemistry and technology, 57(1),
pp. 118–133.
[52] Griffith, R., Kannabiran, R., and Tomlinson, G., 1987. “Rubber flow in an internal
mixer”. Rubber chemistry and technology, 60(1), pp. 111–124.
[53] Menges, G., and Grajewski, F., 1988. “Process analysis of a laboratory internal
mixer”. International Polymer Processing, 3(2), pp. 74–78.
[54] Leblanc, J. L., and Lionnet, R., 1992. “Determining the components of mixing en-
ergy when preparing rubber compounds in instrumented internal mixers”. Polymer
Engineering & Science, 32(15), pp. 989–997.
[55] Toh, M., Gondoh, T., Mori, T., and Mishima, M., 2005. “Mixing characteristics of
an internal mixer: Uniformity of mixed rubber”. Journal of applied polymer science,
95(1), pp. 166–172.
[56] Kaewsakul, W., Sahakaro, K., Dierkes, W., and Noordermeer, J., 2012. “Optimiza-
tion of mixing conditions for silica-reinforced natural rubber tire tread compounds”.
Rubber chemistry and technology, 85(2), pp. 277–294.
[57] Ghari, H. S., Arani, A. J., and Shakouri, Z., 2013. “Mixing sequence in natural
rubber containing organoclay and nano-calcium carbonate ternary hybrid nanocom-
posites”. Rubber Chemistry and Technology, 86(2), pp. 330–341.
[58] Kiparissides, C., and Vlachopoulos, J., 1976. “Finite element analysis of calender-
ing”. Polymer Engineering & Science, 16(10), pp. 712–719.
[59] White, J. L., Kim, J.-K., Szydlowski, W., and Min, K., 1988. “Simulation of flow
in compounding machinery: Internal mixers and modular corotating intermeshing
twin-screw extruders”. Polymer Composites, 9(5), pp. 368–377.
[60] Kim, J., White, J., Min, K., and Szydlowski, W., 1989. “Simulation of flow and
mixing in an internal mixer”. International Polymer Processing, 4(1), pp. 9–15.
130
[61] Kim, J., and White, J., 1991. “Non-newtonian and non-isothermal modeling of 3d-
flow in an internal mixer”. International Polymer Processing, 6(2), pp. 103–110.
[62] Yagii, K., and Kawanishi, K., 1990. “Flow analysis in an internal mixer: Part i:
Application of finite element analysis”. International Polymer Processing, 5(3),
pp. 164–172.
[63] Manas-Zloczower, I., and Feke, D., 1989. “Analysis of agglomerate rupture in linear
flow fields”. International polymer processing, 4(1), pp. 3–8.
[64] Cheng, J.-J., and Manas-Zloczower, I., 1989. “Hydrodynamic analysis of a banbury
mixer 2-d flow simulations for the entire mixing chamber”. Polymer Engineering &
Science, 29(15), pp. 1059–1065.
[65] Cheng, J., and Manas-Zloczower, I., 1990. “Flow field characterization in a banbury
mixer”. International Polymer Processing, 5(3), pp. 178–183.
[66] Yao, C.-H., and Manas-Zloczower, I., 1997. “Influence of design on mixing effi-
ciency in a variable intermeshing clearance mixer”. International Polymer Process-
ing, 12(2), pp. 92–103.
[67] Yao, C.-H., and Manas-Zloczower, I., 1998. “Distributive mixing in an axial dis-
charge continous mixer”. International Polymer Processing, 13(4), pp. 334–341.
[68] Teverovskiy, M., Manas-Zloczower, I., Elemans, P., and Rekers, G., 2000. “Numer-
ical simulations and experiments in a double-couette flow geometry”. International
Polymer Processing, 15(3), pp. 242–254.
[69] Wang, W., Manas-Zloczower, I., and Kaufman, M., 2001. “Characterization of dis-
tributive mixing in polymer processing equipment using renyi entropies”. Interna-
tional Polymer Processing, 16(4), pp. 315–322.
[70] Wang, W., Manas-Zloczower, I., and Kaufman, M., 2003. “Entropic characterization
of distributive mixing in polymer processing equipment”. AIChE journal, 49(7),
pp. 1637–1644.
[71] Yao, C.-H., and Manas-Zloczower, I., 1996. “Study of mixing efficiency in roll-
mills”. Polymer Engineering & Science, 36(3), pp. 305–310.
[72] Gramann, P. J., and Osswald, T., 1992. “Simulating polymer mixing processes using
the boundary element method”. International Polymer Processing, 7(4), pp. 303–
313.
131
[73] Hutchinson, B., Rios, A., and Osswald, T., 1999. “Modeling the distributive mixing
in an internal batch mixer”. International Polymer Processing, 14(4), pp. 315–321.
[74] Alsteens, B., Avalosse, T., Legat, V., Marchal, T., and Slachmuylders, E., 2003.
“Effect of the full-slip condition along rotors on the mixing efficiency of internal
mixers.”. In ANTEC 2003 Conference Proceedings, pp. 173–177.
[75] Alsteens, B., Legat, V., and Avalosse, T., 2004. “Parametric study of the mixing effi-
ciency in a kneading block section of a twin-screw extruder”. International Polymer
Processing, 19(3), pp. 207–217.
[76] Salahudeen, S. A., Elleithy, R. H., AlOthman, O., and AlZahrani, S., 2011. “Com-
parative study of internal batch mixer such as cam, banbury and roller: Numerical
simulation and experimental verification”. Chemical Engineering Science, 66(12),
pp. 2502–2511.
[77] Salahudeen, S. A., AlOthman, O., Elleithy, R. H., Al-Zahrani, S., and Rahmat, A.
R. B., 2013. “Optimization of rotor speed based on stretching, efficiency, and vis-
cous heating in nonintermeshing internal batch mixer: simulation and experimental
verification”. Journal of Applied Polymer Science, 127(4), pp. 2739–2748.
[78] Dhanasekharan, K. M., and Kokini, J. L., 2003. “Design and scaling of wheat dough
extrusion by numerical simulation of flow and heat transfer”. Journal of Food Engi-
neering, 60(4), pp. 421–430.
[79] Connelly, R. K., and Kokini, J. L., 2004. “The effect of shear thinning and differ-
ential viscoelasticity on mixing in a model 2d mixer as determined using fem with
particle tracking”. Journal of Non-Newtonian Fluid Mechanics, 123(1), pp. 1–17.
[80] Rathod, M. L., and Kokini, J. L., 2013. “Effect of mixer geometry and operating
conditions on mixing efficiency of a non-newtonian fluid in a twin screw mixer”.
Journal of Food Engineering, 118(3), pp. 256–265.
[81] Nassehi, V., and Salemi, R., 1994. “Finite element modelling of non-isothermal vis-
cometric flows in rubber mixing”. International Polymer Processing, 9(3), pp. 199–
204.
[82] Nassehi, V., and Ghoreishy, M., 1997. “Simulation of free surface flow in partially
filled internal mixers”. International Polymer Processing, 12(4), pp. 346–353.
[83] Nassehi, V., and Ghoreishy, M., 2001. “Modeling of mixing in internal mixers with
long blade tips”. Advances in Polymer Technology, 20(2), pp. 132–145.
132
[84] Pokriefke, G., 2007. “Numerical simulation of viscous flow in a partially filled co-
rotating twin screw extruder”. International Polymer Processing, 22(1), pp. 61–72.
[85] Fukutani, K., Higashi, K., Yamada, S., and Yamane, Y., 2013. “Numerical study
on distributive mixing characteristics in a partially filled internal mixer”. In 184th
Technical Meeting of Rubber Division, ACS, no. 71, American Chemical Society.
[86] Wu, S., 1971. “Calculation of interfacial tension in polymer systems”. In Journal of
Polymer Science Part C: Polymer Symposia, Vol. 34, Wiley Online Library, pp. 19–
30.
[87] Wiedmann, W., and Schmid, H., 1982. “Optimization of rubber mixing in internal
mixers”. Rubber Chemistry and Technology, 55(2), pp. 363–381.
[88] Carreau, P. J., 1972. “Rheological equations from molecular network theories”.
Transactions of The Society of Rheology (1957-1977), 16(1), pp. 99–127.
[89] Yasuda, K., 1979. “Investigation of the analogies between viscometric and linear
viscoelastic properties of polystyrene fluids”. PhD thesis, Massachusetts Institute of
Technology.
[90] Escudier, M., Poole, R., Presti, F., Dales, C., Nouar, C., Desaubry, C., Graham, L.,
and Pullum, L., 2005. “Observations of asymmetrical flow behaviour in transitional
pipe flow of yield-stress and other shear-thinning liquids”. Journal of non-newtonian
fluid mechanics, 127(2), pp. 143–155.
[91] Patankar, S., 1980. Numerical heat transfer and fluid flow. CRC press.
[92] Muzaferija, S., Peric, M., Sames, P., and Schellin, T., 1998. “A two-fluid navier-
stokes solver to simulate water entry”. In Proceedings of the 22nd symposium on
naval hydrodynamics, Washington, DC, pp. 277–289.
[93] Brackbill, J., Kothe, D. B., and Zemach, C., 1992. “A continuum method for mod-
eling surface tension”. Journal of computational physics, 100(2), pp. 335–354.
[94] Vyakaranam, K. V., and Kokini, J. L., 2012. “Prediction of air bubble dispersion in
a viscous fluid in a twin-screw continuous mixer using fem simulations of dispersive
mixing”. Chemical Engineering Science, 84, pp. 303–314.
[95] Keuter, H., and Rinker, M., 2005. “Rotor optimization to improve mixing effi-
ciency.”. In European Rubber Research-Practical Improvements of the Mixing Pro-
cess: SATPRO, ROTOR and Dust Stop.
133
[96] Vyakaranam, K., Evans, M., Ashokan, B., and Kokini, J. L., 2009. “12 evaluation
of mixing and air bubble dispersion in viscous liquids using numerical simulations”.
Food Mixing: Principles and Applications, p. 253.
[97] Avalosse, T., Alsteens, B., and Legat, V., 2005. “Rotor shape design by numeri-
cal simulation: A new way to improve dispersive and distributive mixing in batch
mixer.”. In European Rubber Research-Practical Improvements of the Mixing Pro-
cess: SATPRO, ROTOR and Dust Stop.
[98] Danckwerts, P., 1952. “The definition and measurement of some characteristics of
mixtures”. Applied Scientific Research, Section A, 3(4), pp. 279–296.
[99] Danckwerts, P., 1953. “Theory of mixtures and mixing”. Research, 6, pp. 355–361.
[100] Ottino, J. M., 1989. The kinematics of mixing: stretching, chaos, and transport,
Vol. 3. Cambridge university press.
[101] Schuster, R. H., 2005. “Dispersion of fillers- a decisive factor for the processing
of rubber compounds and the performance of elastomers”. In European Rubber
Research-Practical Improvements of the Mixing Process: SATPRO, ROTOR and
Dust Stop.
[102] Emin, M., and Schuchmann, H., 2012. “Analysis of the dispersive mixing efficiency
in a twin-screw extrusion processing of starch based matrix”. Journal of food engi-
neering, 79(3), pp. 956–969.
134