You are on page 1of 90

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/280599439

Mixing in the Farrel Continuous Mixer

Chapter · January 1994

CITATION READS
1 1,644

2 authors, including:

Eduardo Luis Canedo


Universidade Federal de Campina Grande (UFCG)
104 PUBLICATIONS   418 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Non-isothermal crystallization of PHB and PHB/CB compounds View project

PP/PBAT Blends View project

All content following this page was uploaded by Eduardo Luis Canedo on 02 October 2015.

The user has requested enhancement of the downloaded file.


MIXING IN THE FARREL CONTINUOUS MIXER

by

Eduardo L. Canedo and Lefteris N. Valsamis

Farrel Corporation
Ansonia, CT 06401

1993
Reprinted with corrections, Novenber 1994.

Published as Chapter 23 of Mixing and Compounding. Theory and Practice, edited by Professors
Ica Manas-Zloczower and Zehev Tadmor, as part of the series Progress in Polymer Processing.

Copyright © 1994 Carl Hanser Verlag, Munich – All rights reserved.


CONTENTS

1. Introduction 1
2. Structure and Principles of Operation 2
2.1. Solids Conveying 3
2.2. Melting 3
2.3. Mixing 4
2.4. Devolatilization 5
2.5. Pumping 6
3. Modeling 8
3.1. Circumferential Flow 10
Basic Equations 11
Flow Patterns 14
Shear Rate and Shear Strain 16
Extensional Flow Components 20
Shear Rate Distribution Function 20
Non-Newtonian Effects 22
Thermal Effects 24
3.2. Global Flow Models 26
Residence Time Distribution 27
Number of Passes Distribution 30
Maximum Shear Distribution 33
3.3. Scale-up Considerations 35
4. Rotor Design 38
4.1. Single Stage Rotors 38
4.2. Two-Stage Rotors 40
5. Conclusion 42
Nomenclature 43
References 45
Figures 1-36 51
Mixing in the Farrel Continuous Mixer 1

1. INTRODUCTION

Mixing, compounding, blending, alloying, and reactive extrusion are some of the terms
often used in polymer processing to describe the process functions employed to develop,
transform, or extend the range of physical and chemical properties of materials prepared
with existing, well-known raw components. These processes can, in principle, be per-
formed in various types of single- and twin-screw extruders, and batch and continuous
mixers. The selection of the type of equipment depends on the specific characteristics of
the mixing or blending process in terms of shear and thermal history, residence time and
residence time distribution, sequential addition of additives coupled to venting and de-
volatilization, and so on. Some or all of these process functions may be needed during a
given application. The underlying principle for successful materials development is a thor-
ough understanding of the detailed mechanisms associated with these process functions
and the ability of a piece of equipment to induce these process functions in a controllable
mode. Machine manufacturers have therefore devoted considerable effort toward the elu-
cidation of flow and thermal fields in processing equipment, with the goal of developing
improved machinery to meet the increasingly stringent process requirements imposed by
plastics producers.

The Farrel Continuous Mixer (FCM) was introduced in the early 1960s (Ahlefeld et al,
1964, 1966), primarily as a continuous rubber compounder, based on the well-known high
intensity batch Banbury® mixer. Since its inception, the FCM has enjoyed wide acceptance
in the plastics industry whenever a high intensity mixing was required. Although the mixer
has been very successful in post-reactor processing (Valsamis and Canedo, 1989a) and,
more recently, as a general purpose compounder in the Compact Processor (CP) series,
very little work has been done in terms of mathematical modeling of the FCM because of
the complex flow patterns prevailing in it. The recent work of Kim and White (1992) is
the first attempt to simulate the global flow patterns in the FCM. However, important
information can be extracted by a detailed analysis of selected regions of interest in the
mixer, such as the vicinity of the wing tips and the region between the two rotors. The
high stress levels developed in the former are responsible for dispersive mixing, while the
prevailing flow patterns in the latter promote efficient laminar extensive and distributive
mixing. These and related concepts are reviewed in detail in this chapter.
2 Mixing in the Farrel Continuous Mixer

2. STRUCTURE AND PRINCIPLES OF OPERATION

The Farrel Continuous Mixer (FCM) is a counterrotating nonintermeshing twin-rotor


mixer as illustrated in Figure 1. A conventional single- or double-flighted screw section,
which extends beyond the length of the hopper opening, accommodates the feeding of
material to the mixing section. In the mixing section, each rotor is made of two helically
twisted wings, approximately 180° apart, with the wings of each rotor twisted in opposite
directions. Each wing is composed of a forward pumping section followed by a reverse
pumping section, the two being of unequal length and twisted by slightly different angles.
In general the apex of a wing (i.e., the point at which the opposing sections intersect) is
offset from the apex of the complementary wing of the same rotor, and from the two wings
of the other rotor. In some rotor designs, the reverse pumping wing is followed by a short
neutral (0° helix angle) pumping section. Figure 2 shows several standard rotor styles.

The rotors are housed in cylindrical enclosures, known as chamber halves, which
communicate with each other along the entire mixing section. Feed and exit openings are
located at the intersection of the chamber halves. Both the housing and the rotors are tem-
perature controlled, although quite often both are operated in neutral (i.e., adiabatic) con-
ditions (Figure 3).

Material entering the mixing section is dragged forward and circumferentially by the
forward wings of the two rotors in a helical path, with the reverse helix wings opposing the
forward movement of material. Forward motion is obtained primarily because of the
viscosity differential between inlet and outlet, which results in higher pressure gradients in
the forward wing section of the rotor. An adjustable orifice or gate is used to provide
added resistance to the forward motion of material, and thus to control the overall fill factor
in the mixer at given operating conditions. The moderate pressure levels generated by the
rotors in both the forward and reverse helix sections ensure good exchange of material
from one chamber half into the other. Such exchange of material is a prerequisite to good
extensive mixing

Transfer of material from chamber to chamber can be further enhanced by setting and
maintaining during operation a preferred orientation of the wings of one rotor with respect
to other. To achieve this mode of operation the two rotors must operate at synchronous
speed. This is the preferred mode of operation, although in the past the rotors were oper-
ated exclusively at different speeds, maintaining what is called a friction ratio. The use of
Mixing in the Farrel Continuous Mixer 3

stationary and rotating dams can also serve to enhance interchamber material transfer.
Stationary dams are simply inserts positioned through the housing at the intersection of the
two chamber halves, in the space between the two rotors. Rotary dams or blisters are in-
corporated on the rotors as rings with a uniform clearance all around. In either case the
dam provides an added resistance to the forward flow while forcing material away from the
window of interaction of the two rotors. Multiple dams can be used in a given operation
and their position, number, and geometry are dictated by process considerations. A more
complete treatment of the FCM structure and principles of operation can be found
elsewhere (Farrel, 1965; Farrel, 1971; Hold, 1984; Kearney, 1991; Canedo and Valsamis,
1993).
Mixing is only one of the many operations a continuous mixer is required to perform.
Solids handling, melting, pumping, and often devolatilization, must accompany the mixing
process. Those operations are briefly examined next.

2.1. Solids Conveying


Solids handling is restricted to the initial section of the mixer and is performed by a short
screw section, using single- or double-flighted screw geometries. Single-flighted screws
are standard. Double-flighted screws are used predominantly in large mixers to accom-
modate high feed rates or low bulk density materials. Flow is drag induced and, since the
mixer is normally starve-fed, pressure rise in the solids conveying region is minimal and is
accompanied by a moderate rise in material temperature. Since the primary function of the
feed screw section is to transport the material to the mixing section, solids conveying is not
usually the rate limiting factor. In freezing experiments, in which the mixer is stopped
during operation and disassembled for visual examination, solid particulates are loosely
packed toward the mixing section.

2.2. Melting
Melting mechanisms have been investigated by freezing experiments similar those re-
ported in the literature (Tadmor and Klein, 1970). Figure 4a shows a typical section of
solidified polymer which has been removed from the mixer. The exit orifice of the mixer
was adjusted during operation to maintain the entire mixing section full. The material used
in the investigation was a 50:50 blend of low density polyethylene (LDPE) and polystyrene
(PS). Visual examination of the solidified material reveals that melting occurs very early
4 Mixing in the Farrel Continuous Mixer

in the mixing section. The energy dissipated at the entrance of that section is sufficient to
fuse the solid particles together and is followed by the intimate mixing of the solids into a
molten polymer matrix. The conventional solid bed often found in single-screw extruders
is absent in this case. This melting mechanism has been labeled as dissipative mix-melting
by Tadmor and Gogos (1979).

The dominant source of energy for melting is mechanical energy introduced by the
rotors and converted into thermal energy by continuous deformation of the solid particles,
friction of the solid particles between themselves and with the metal walls, and viscous en-
ergy dissipation in the molten material. The flow patterns prevailing in the mixing section
readily mix the molten material with the solids, and the particulates are gradually converted
from an inhomogeneous, partly fluxed mass into a uniform suspension of gradually smaller
and softer solid particles in a molten polymer matrix. Figure 4b illustrates this
transformation. The solidified sample of Figure 4a was sliced at different axial locations to
generate the cross sections of Figure 4b. Not shown in the cross sections of Figure 4a is
the material in the window of interaction between the two rotors, which could not be re-
moved without being destroyed.

2.3. Mixing
Extensive mixing of highly viscous polymeric materials in continuous mixers is
achieved by a combination of laminar and distributive mechanisms. Laminar mixing
mechanisms can be further distinguished by the prevailing flow patterns into shearing and
elongational or squeezing flows. Shearing flows occur in the melt pools formed in front of
each rotor wing, in the regions between the rotor wing tips and the housing walls, and in
the window of interaction of the two rotors. Elongational flow components can be found
in the wedgelike region between the leading face of the rotor wings and the chamber walls.

Squeezing flows are obtained in the window of interaction as opposing wings from the
rotors cross each other. These squeezing flow patterns can be maximized by operating the
rotors at even speed and by maintaining a preferred rotor to rotor alignment. Figure 5a
shows schematically the effect of tip-to-tip rotor orientation: the material in the window of
interaction between the rotors is subjected to periodic contraction and expansion. If the
opposing wings of the two rotors are pumping in the same direction, axial flow accompa-
nies the squeezing action. If the two wings pump in opposite directions, a “scissors” effect
takes place. Distributive mixing can be promoted in the window of interaction by main-
Mixing in the Farrel Continuous Mixer 5

taining a tip-to-flat rotor orientation as shown schematically in Figure 5b. Under these
circumstances, the material between the rotors is periodically caught by the rolling pools
generated by the wings of the two opposing rotors. The situation is not unlike that of a
cavity in which opposite walls are subject to periodic motions in the same direction but out
of phase (Figure 6). The chaotic mixing characteristics of this flow pattern are well known
(see, e.g., Ottino, 1989).

Superimposed on those laminar flow patterns is the constant stream splitting and re-
combination of the material leaving one melt pool and joining the pool of the adjacent rotor
wing, and moving from one chamber half into the other. Figure 7 shows typical flow
patterns as material dragged by each rotor wing converges in the window of interaction of
the two rotors. It is also clear that the material conforms to the shape of the rotor wings as
illustrated in the lower portion of Figure 7. The top housing has been removed to permit
observation of the flow pattern. Consequently, circulating pools are difficult to see, since
they are visible only in the areas that experience relative motion between the rotor wings
and the inner housing walls.

Intensive mixing is achieved mainly by the passage of material through high shear re-
gions located at the gap between the wing tips and the housing. The flow in this region is
highly non-uniform, with very large deformations imposed on a small fraction of material,
while most of the material is being subjected to a relatively low shear. In these conditions,
uniform dispersion requires repeated passages of the material through the high shear re-
gions, with redistribution and randomization of the material elements between passes.
Modeling of these processes is considered in some detail in the next sections of this chap-
ter. Intensive mixing is also achieved by inter-particle friction within partially molten
material during the fluxing process, in an action reminiscent of grinding, in connection
with the dissipative mix-melting mechanism.

2.4. Devolatilization
For process applications requiring some degree of devolatilization the rotor geometry
of Figure 2c is preferred. Here, a second screw section has been incorporated in the mixing
section of each rotor. The downstream screw section may be vented for the removal of
volatiles. The screw geometry ensures the positive conveyance of material in the vent
section to avoid plugging of the vent port. The degassing characteristics of the continuous
mixer arise from high surface renewal rates, which are obtained by continuous spreading
6 Mixing in the Farrel Continuous Mixer

followed by mixing of the molten material by the rotor wings. Those flow patterns are in
contrast to the ones observed in other continuous equipment, where only a small portion of
the material is wiped by the flights into the housing wall (Biesenberger and Sebastian,
1983). Figure 4b shows that the mixer is only partially filled, allowing the gases to escape
into the vent opening. A preliminary attempt at modeling the devolatilization charac-
teristics of the FCM based on these concept has been presented by Valsamis and Canedo
(1989a-b).

2.5. Pumping
The continuous mixer is a low pressurizing device and the final pressurization process
is usually done by means of an extruder or gear pump. In the first case, the material dis-
charged from the mixer is gravity fed to a single screw extruder, which is operated inde-
pendently of the FCM. In the second case, the gear pump is close coupled to the mixer.
As a matter of fact, the gear pump is used to control the operation of the mixer. The gen-
eral arrangement of a continuous mixer-extruder combination and a continuous mixer-gear
pump combination are illustrated in Figures 8 and 9, respectively. In the latter configura-
tion, the mixer is placed sideways to facilitate close coupling with the gear pump, while
providing a horizontal discharge.

In some cases the continuous mixer and the extruder are mounted in single frame; this
assembly is known as Compact Processor (CP-Series) and identified by a two number
designation, indicating in inches the nominal bore diameter of mixer and extruder, respec-
tively. Farrel Corporation manufactures Compact Processors in standard sizes from CP-12
to CP-68. The CP series can be operated manually, with mixer and extruder individually
controlled. However, process control systems are available which provide fully automated
machine operation.

In the UMSD/GP line, machine operation is exclusively controlled by a microprocessor


control system, thus ensuring that the desired mode of operation is achieved (Scharer,
1985; Patel, 1987). The control process constantly monitors the operating characteristics
of the entire line and provides the needed adjustments to ensure maximum process effi-
ciency at all times. To control the fill level of the mixer, adjustments are made primarily to
the gear pump speed. Essentially, the gear pump replaces the function of orifice and
provides the required back pressure needed for mixer operation and the die pressure needed
for pelletizing. Such an operation is feasible because the mixer is operated in starved
Mixing in the Farrel Continuous Mixer 7

mode, and under these conditions the rotor speed is independent of the throughput rate.
For a given material, the proper combination of mixer rotor speed, flow rate, and gear
pump speed can yield optimum mixer performance based on target work input (specific
energy input), exit melt temperature, and exit pressure (gear pump suction pressure). Any
of these three parameters can be used for process control. In all cases. the mixer control
system employed is a closed-loop system, which provides corrective action to deviations
from set points.

To operate the mixer-gear pump combination under constant specific energy input the
flow rate, the mixer rotor speed, and the specific energy must be specified. The target
torque value is then calculated by a microprocessor and compared to the actual measured
value, determining the error difference. This error is used in a standard composite control
mode algorithm. The controller uses a series of calculated errors to determine what cor-
rective action is required by the gear pump. This simple logic allows a constant specific
energy to be applied to different polymer resins.

The constant temperature loop provides a control scheme for maintaining the stock
discharge temperature at a level set by the operator. The basic loop accomplishes this by
adjusting the specific energy reference value based on temperature feedback from the
mixer orifice. The control logic takes the ratio of the average of a number of actual tem-
peratures and the set or target temperature. This result is then multiplied by the target
specific energy value, yielding a new target value. A timing function incorporated into the
logic compensates for the long lag times involved with temperature changes. The timing
circuit is used to prevent oscillations, which could be caused by overadjusting the gear
pump speed before the effects of the preceding adjustment had been realized.
The constant energy and constant exit melt temperature control schemes can be applied
universally to any mixer, with or without the gear pump. A motorized orifice gate is used
to provide the required adjustments of the mixer fill factor. The constant gear pump inlet
pressure control scheme uses a pressure reading at the pump inlet to adjust the gear pump
speed. As in the other control schemes, a proportional integral derivative (PID) algorithm
is used to adjust the gear pump speed to the target values. However, unlike the two control
schemes described earlier, operation of the mixer at reduced rates requires the specification
of new target values if the mixer fill factor and thus specific energy input values are to be
maintained constant.
8 Mixing in the Farrel Continuous Mixer

Figure 10 illustrates the response of the gear pump speed to changes in flow rate under
constant energy input at a fixed rotor speed. For a 50% reduction in flow rate, the mixer
torque value is also reduced by 50%, while the gear pump suction pressure is reduced from
40 psi to 12 psi (a 70% reduction). Close examination of Figure 10 shows an increase in
the gear pump speed followed by a decrease to a new constant level. It increases from 40
rpm to 60 rpm and then drops to approximately 20 rpm. During this transient period, ex-
cess material is removed from the mixer, and the adjusted gear pump speed reflects the re-
duction in throughput rate.

3. MODELING

The flow patterns in high intensity continuous mixers are extremely complex; as a con-
sequence, there is the lack of comprehensive theoretical models to describe the processing
characteristics of such equipment. Recently, however, there have been attempts to apply
global models, successfully used on twin-screw extruders, to the study of continuous mix-
ers. Preliminary results for one of them, based on a modified flow network analysis (FAN)
approach, have been presented by Kim and White (1992). Yet theoretical models for
continuous mixers cannot match those readily available for single-screw extruders, and to a
lesser extent for some twin-screw extruders, in sophistication or predictive reliability. To
explain this apparent anomaly, since continuous mixers are often considered to be a special
kind of extruder, it may be instructive to review the main differences in geometry and
processing characteristics between continuous mixers and typical single- and twin-screw
extruders.
In a typical extruder the different processing steps (solid transport, melting, mixing,
pumping) take place successively along the axial direction, leading to a series of functional
zones in that direction, with minimum overlap between them. Consequently, models of
individual steps, frequently developed neglecting end effects, can be combined in series to
obtain a realistic picture of the whole machine. Continuous mixers, with an overall length
to diameter ratio L/D in the range of 5-10, are usually significantly shorter than equivalent
single or twin screw extruders, leading to a considerable overlap of processing steps.
While the functional combination of steps has definite processing advantages (one case in
point is the combination of melting and mixing in the so called dissipative mix-melting), it
makes more difficult the development of theoretical models of machine behavior.
Mixing in the Farrel Continuous Mixer 9

Melt flow in extruders is dominated by the channel flow, and the passage of material
over the screw flights (leak flow) is a relatively minor, often neglected, component of the
overall flow pattern. In contrast, the circumferential flow over the wing tips of continuous
mixers plays a much more important role and can arguably be considered to be the major
flow component. Although wing tips and screw flights appear to be equivalent machine
elements, their processing functions are different. The function of the screw flights is to
generate a drag flow component in the down-channel direction and to promote the recircu-
lating cross-channel flow by blocking the circumferential flow. The rotor wings have a
dual function. First, they provide high shear regions at and near the wing tip, increase the
circumferential flow rate over the wing tip by dynamic viscous pressurization in the
wedgelike region between the leading face of the wing and the chamber wall, and generate
rolling pools behind them. Second, they contribute to the axial circulation and to the local
degree of filling by virtue of their helical twist. The primary function of the rotor wings
has no equivalent in extrusion, except in some dispersive mixing sections of single-screw
extruders and in certain kneading block configurations in twin-screw extruders. Continu-
ous mixer rotor wings have more in common with batch internal mixer rotor wings, from
which they directly derive.

Another difference between extruders and continuous mixers is the presence in the
latter of a relatively wide open area between the rotors, the so called window of interaction
between the rotors. This area not only controls the exchange of material between the rotors
but plays an important role in the axial transport of material, allowing a considerable
degree of backmixing. Both functions can be affected by means of special mixing inserts
or dams in the window of interaction and by changing the relative orientation one rotor to
the other.

We will examine in some detail two complementary partial models that can be used to
investigate the basic mechanisms of melt transport and mixing in continuous mixers: a
classical fluid mechanical model of the circumferential flow over the rotor wing tips and a
global semiempirical transport model based on population balances. An early summary of
these topics is available (Valsamis and Canedo, 1989a).
10 Mixing in the Farrel Continuous Mixer

3.1. Circumferential Flow


A significant amount of work has been devoted to modeling the circumferential flow
over the rotor wing tips, mostly in connection with batch internal mixers. Early treatments
showed the essential qualitative features of the flow (Bolen and Colwell, 1958; Bergen,
1959). Semiquantitative models based on the lubrication approximation were developed
(Meissner and Rehrer, 1979; Meissner et al, 1980). The lubrication approximation as-
sumes that the flow in the wedge formed between the rotor leading face and the chamber
wall is locally one-dimensional, a valid assumption if the wedge angle is small. Under
certain conditions (linear wedge, Newtonian material, isothermal flow) a closed-form so-
lution of the equations of motion can be obtained. More elaborate models, still based on
the lubrication approximation but relaxing one or more of the other simplifying assump-
tions, have been presented (Wagenknecht et al, 1986, 1987, 1988). These models consider
wedges of arbitrary shape, purely viscous non-Newtonian fluids, and nonisothermal con-
ditions (viscous heat generation inside the material and heat transfer through the walls) and
require the use of simple numerical techniques to obtain meaningful results. A numerical
scheme along these lines has been described in some detail by Canedo and Valsamis
(1990), and a recent article by Meissner and Poltersdorf (1992) reviews much of the work
on the subject.

It has been suggested that the relatively minor elongational flow components in the
flow over the rotor tips can have a major effect on the dispersive mixing performance.
Although the order of magnitude of these elongational components can be estimated with
models based on the lubrication approximation, its accurate description requires more so-
phisticated techniques. Two-dimensional finite element methods have been used to study
the circumferential flow in actual mixer wing geometries (Cheng and Manas-Zloczower,
1989a; Yagii and Kawanishi, 1990; Nassehi and Freakley, 1991), and in a full mixing
chamber cross-section under realistic rheological and thermal conditions (Cheng and
Manas-Zloczower, 1989b, 1990; Wong and Manas-Zloczower, 1992).

Nevertheless, models based on the lubrication approximation capture the essential


features of the circumferential flow. Since analytic or relatively simple numerical solutions
can be easily obtained, these models are useful as a guide to experimentation and as a basis
for scale-up. We present some results of practical importance developed with such
approximate models.
Mixing in the Farrel Continuous Mixer 11

Basic Equations

Let us consider a mixer wing cross section (Figure 11), composed of a wing tip, leading
and trailing faces, and a rotor flat or waist zone. From a processing point of view, the
geometry of a rotor wing cross section is characterized by the leading face wedge angle
α, the wing tip width e, and by the minimum and maximum clearances between the rotor
and the chamber wall:

=h ½ ( D0 − DMAX )
(1)
H 0 ½ ( D0 − DMIN )
=

where D0 is the chamber bore diameter, and DMAX and DMIN are the maximum and mini-
mum rotor diameters, respectively. In the vicinity of the wing tip the flow field can be ap-
proximated by a combination of linear wedge and uniform clearance, in which a rectangu-
lar coordinate system x-y with origin at entrance of the tip has been defined as shown in
Figure 12. The gap clearance H = H(x) is given by:

h − x tan α , − E0 < x < 0


H = (2)
h , 0< x<e

The linear wedge is a very good approximation of the gap between the leading face of the
rotor and the chamber wall in the Farrel Continuous Mixer. Since we are interested in the
flow pattern in the vicinity of x = 0, the wedge extension in the circumferential direction E0
is not well defined by the model and may depend on the local fill factor φ, but in general

H0 − h
E0 ≤ (3)
tan α

In the present case we will assume that the equals sign always applies in Eq.(3). According
to Eq.(2), it is possible to substitute H for x as an independent variable in the wedge
section, x 0.

In principle, the lubrication approximation is valid only if tan << 1. Most continuous
mixer designs are based on leading wedge angles in the range of 10-20° (tan in the range
0.18-0.36), which hardly meet the criteria above. However, numerical calculations per-
formed without the lubrication approximation are in very good agreement (within 5-10%)
with the analytical results for the velocity, shear rate, and pressure profiles obtained using
12 Mixing in the Farrel Continuous Mixer

the approximation. Under the lubrication approximation, and assuming incompressible,


isothermal, creeping flow of a Newtonian fluid, the equations of continuity and motion are
reduced to:

∂u ∂v
+ =
0 (4)
∂x ∂y

and

dp ∂ 2u
=η 2 (5)
dx ∂ y

where u = u(x,y) and v = v(x,y) are the velocity components in the x and y directions re-
spectively, subject to the conditions

− E0 < x < e, y= 0 : u = U 0 = π ND0 , v = 0


− E0 < x < e, y = H ( x) : u =v =0 (6)

and p = p(x) is the pressure field, subject to the conditions

x=− E0 , 0 < y < H 0 : p =p0


=x e, 0< y<h : =p p0 (7)

Here η is the melt viscosity, U0 is the linear speed of the moving wall, N is angular rotor
speed (rpm), and p0 is an arbitrary constant pressure. Eq.(4)-(5), with boundary conditions
given by Eq.(6)-(7), can be easily integrated. The solution is conveniently expressed in
terms of the parameter S defined by:

∫ u ( 0,=
y ) dy U 0 h(1 + S )
h
=q 1
2 (8)
0

where q is the flow over the rotor wing tip per unit axial length. In the present case
(Newtonian isothermal flow) the pressurization coefficient S is the ratio of pressure to drag
flow in the tip clearance. The resulting velocity profiles are

  h  y  y 
u = 1 − 3 1 − (1 + S )   1 − U 0 (9)
  H  H  H 

in the circumferential direction, and


Mixing in the Farrel Continuous Mixer 13

  h  y  y 
2

 tan α  2 − 3(1 + S )  1 −   U 0 , x<0


v=  H   H  H 
0 , x>0
(10)

across the gap, where H = H(x) is given by Eq.(2). As expected, the velocity profile across
the gap is negligible if the lubrication approximation is strictly enforced (tan α → 0). In the
present case, Eq.(10) can be used to obtain estimates v and its derivatives, although with a
larger margin of error that similar quantities derived from Eq.(9). The pressure profile can
be computed introducing Eq.(9) into Eq.(5) and integrating from both ends:

 3ηU 0   h h   h h 
 cot α  2 − (1 + S )  +   − , − E0 < x ≤ 0
 h   H H 0   H H 0 
∆p = p − p0 =  (11)
 6ηU 0 S
 h2 ( e − x ) , 0≤ x<e

An expression for the pressurization coefficient S can be obtained taking the limits

lim p ( x) = lim+ p ( x)
x → 0− x →0 (12)

in the two branches of Eq.(11), leading to:

2
 h 
1 − 
S=  H0 
(13)
2
e  h 
1 + 2 tan α −  
h  H0 

Notice that in the present case (Newtonian isothermal flow), the pressurization coefficient
is a purely geometric parameter with values, for most continuous mixer geometries, in the
range 0.25 < S < 0.50. Thus, dynamic viscous pressurization in the wedge increases the
flow over the wing tip by 25-50%.

Flow Patterns

An analysis of the velocity profile u(x,y) given by Eq.(9) reveals the existence of a
stagnation point on the stationary wall at H = HS
14 Mixing in the Farrel Continuous Mixer

H=
S
3
2 (1 + S )h (14)

Provided H0 > HS , which is usually the case, there is recirculation in the wedge. For H <
HS, the material flows forward (i.e., in the direction of the moving wall), all across the tip
clearance (region I) and the wedge (region II). On the other hand, for H < HS there is a
forward moving layer near the moving wall (region III) and a rolling pool near the station-
ary wall (region IV). Figure 13 shows schematically the flow patterns in the model wedge.

The streamline that separates region III from region IV, y = yS (x), can be obtained by
taking into account that the flow rate in the forward moving layer, 0 < y < yS, must be
equal to the flow rate over the wing tip given by Eq.(8):

yS
∫0
=
u dy 1
2 (1 + S )U 0 h (15)

while there is no net flow rate in the rolling pool:

H
∫ yS
u dy = 0 (16)

Eq.(15) or (16), along with Eq.(9) lead to:

(1 + S )h
1
yS = 2
(17)
h
1 − (1 + S )
H

The line y = yS (x) is a stagnation streamline ending on the stationary wall at the stagnation
point y = HS . The velocity along the stagnation streamline can be readily obtained substi-
tuting Eq.(17) into Eq.(9):

2
 3 h
1 − 2 (1 + S ) 
uS = 
H (18)
U0
h
1 − (1 + S )
H

Inside the rolling pool (region IV) the velocity profile reaches a minimum at y = yMIN (x):
Mixing in the Farrel Continuous Mixer 15

 3 h 
1 − 4 (1 + S ) H
2
3
=  
H
yMIN (19)
h
1 − (1 + S )
H

with velocity:

2
 3 h
1 − 2 (1 + S ) 
1
3
= − 
H (20)
uMIN U0
h
1 − (1 + S )
H

and u = 0 at y = y0 (x):

1
H
y0 = 3
(21)
h
1 − (1 + S )
H

Figure 14a shows a typical velocity profile in regions III and IV. If S > 1/3, the velocity
has a maximum in regions I and II for H < ¾(1+S)h. In the tip clearance this maximum is
located at:

3S − 1
yMAX = h (22)
6S

with velocity:

( 3S + 1)
2

uMAX = U0 (23)
12 S

Figure 14b shows a typical velocity profile in regions I-II.

An analysis of the pressure profile p(x) given by Eq.(11) reveals that it reaches a
maximum inside region II at H = HM

H M= (1 + S )h (24)

where
16 Mixing in the Farrel Continuous Mixer

2
3ηU 0  h 
=∆pM cot α 1 − (1 + S )  (25)
(1 + S )h  H0 

Figure 15 shows, in dimensionless form, a typical pressure profile along the model wedge.
The pressure profile is one of the very few quantities discussed here that can be directly
measured in processing equipment. Preliminary results of experiments conducted with a
high density polyethylene (HDPE) melt in a 4FCM show qualitative agreement with pre-
dictions based on Eq.(11). The measured maximum pressure is of the same order of
magnitude of the value predicted by Eq.(25), an encouraging result in view of the highly
non-Newtonian and nonisothermal nature of the actual flow field.

Shear Rate and Shear Strain

The wedge is predominantly a shear device, in which most of the deformation imposed
on the material passing through it is due to shear between fluid elements. The rate at which
the shear is imposed is closely related to the mixing characteristics of flow field. Extensive
(laminar) mixing, as measured, by the increase of “interfacial” area or by the reduction of
the “striation thickness”, is proportional to the shear strain, which can be obtained from the
shear rate by integration. Intensive mixing (dispersion) is, in first approximation,
proportional to the shear stress, which in the present case is proportional to shear rate
through the constant viscosity. In the wedge, the shear rate profile γ = γ ( x, y ) is given by:

2
∂u  ∂v  ∂u
2

γ =   +  ≈ (26)
∂ y ∂ x ∂y

The radial contribution (proportional to tan2 α) has been neglected. Taking into account
Eq.(9):

 h  y  U0
γ = 1 − 3 1 − (1 + S )  1 + 2  (27)
 H  H H

The shear rate profile is a piecewise linear function of the radial coordinate y, and with the
help of Eq.(17)-(23) a complete picture of the shear field can be easily drawn.

In the tip clearance, 0 ≤ x ≤ e, the gap thickness is fixed, H = h, and the shear rate is
independent of the position along the gap. If S ≤ 1/3 the shear rate reaches its minimum at
the moving wall (y = 0) and its maximum at the stationary wall (y = h):
Mixing in the Farrel Continuous Mixer 17

1 − 3S ≤ γ ≤ 1 + 3S (28)

The average shear rate imposed on the material is given by:

1 h U0
q ∫0
γ AVG
= = γ u dy
(1 + S ) h (29)

If S > 1/3, the velocity has a maximum inside the gap, and the shear rate reaches its mini-
mum ( γ = 0 ) at that point, y = yMAX given by Eq.(22), and its maximum again at the sta-
tionary wall:

0 ≤ γ ≤ 1 + 3S (30)

γ 3S − 1 , while the average shear rate is:


The shear rate at the stationary wall is =

1 h  u 
2
 U  (1 + 3S ) 4  U 0
γ AVG = ∫ γ u dy =  2  MAX
 − 1 0
=  − 1 (31)
 (1 + S )h  72 S  (1 + S )h
2
q 0   U 0 

where uMAX is given by Eq.(23). Typical velocity and shear rate profiles in the tip clearance
are presented below in Figure 19. At the entrance of region II, the shear rate reaches its
minimum ( γ = 0 ) at the stagnation point, on the stationary wall, and its maximum on the
moving wall, in contrast with the behavior observed in the tip clearance:

4U 0
0 ≤ γ ≤ (32)
3(1 + S ) H

At the maximum pressure point, H = HM given by Eq.(24), the shear rate is uniform across
the gap. While the maximum shear rate developed in region II is lower than the maximum
obtained in the tip clearance, the average shear rate has the same value, given by Eq.(29),
all along the wedge if S ≤ 1/3. If S > 1/3, the average shear rate is still given by Eq.(29),
except for an small section of the wedge near the tip clearance where H < ¾(1+S)h, in
which it gradually adjusts its value to the one given by Eq.(31) at the entrance of the tip
clearance. Regions I and II (the tip clearance and its immediate vicinity) are considered to
be the high shear zone of the mixer. Average shear rates in the range 500-1000 sec-1 are
not uncommon in typical continuous mixers. In region III the span of the shear rate, as
well as its magnitude, is reduced:
18 Mixing in the Farrel Continuous Mixer

(33)
 h U  h U
4 1 − 32 (1 + S )  0 ≤ γ ≤ 4 1 − 34 (1 + S )  0
 HH  HH

The average shear rate in region III is given by:

(34)
  3 h 
4

 u   U2   2 1 − (1 + S )  
γ AVG 1 −  S  
= 0
=1 −  H   U0
  U 0   (1 + S )h  h   (1 + S )h
2

 1 − (1 + S )  
  H 

Figure 15 shows, in dimensionless form, the average shear rate profile along the model
wedge in a typical case (S = 1/3, e / h = 3, H0 / h = 10). It is interesting to have an estimate
of the average shear rate imposed on the material trapped in the rolling pool (region IV):

uS + 2uMIN
2 2
1 H
=γ AVG = ∫
2qRP y0
γ u dy
4qRP
(35)

where qRP = qRP(x) is the circulation flow (per unit axial length) in the rolling pool, com-
puted as follows:

y0 H
qRP = ∫ yS
u dy = − ∫ u dy
y0
(36)

Figure 16 shows the average shear rate in the rolling pool as a function of the gap geome-
try. Notice that the shear rate in the rolling pool is about one order of magnitude less than
in the high shear regions I and II.

The amount of shear deformation or shear strain is defined, in differential terms and
within the lubrication approximation, as follows:

∂u ∂ u dx
dγ =
− dt =
− (37)
∂y ∂y u

Eq.(37) should be integrated along streamlines to obtain the shear strain distribution. Ex-
cept in the simplest case, numerical methods are required. However, the average shear
strain in a region can be easily computed by inverting the order of integration. Let us
consider the forward moving flow in regions I-II-III of the model wedge. The total average
shear strain imposed in region III can be computed as follows:
Mixing in the Farrel Continuous Mixer 19

yS H
h H
⌠ 0 dH ⌠  ∂u  2 ⌠ 0  uS  dH
γ
= ( III )

 =
−  dy

 1 −  (38)
q tan α H ⌡0  ∂ y (1 + S ) tan α
AVG

⌡H S ⌡H S  U0  H

resulting in

1   3 h  1  h 
γ AVG
= ( III )
6 1 − 2 (1 + S )  − ln 3  1 − (1 + S )  (39)
tan α   H0  1+ S  H0  
 

A similar procedure in region II leads to:

H
h ⌠
HS
dH ⌠  ∂u  2 ⌠
HS
dH
=γ ( II )



 =
−  dy 
 (40)
q tan α  ∂ y (1 + S ) tan α
AVG
⌡h H  ⌡h H
⌡0

resulting in

2
=γ AVG
( II )
ln 32 (1 + S ) (41)
(1 + S ) tan α

while in the tip clearance:

1 e h ∂u  e
γ AVG = ∫ dx ∫  − u dy=
(I )
 (42)
q 0 0
 ∂ y (1 + S )h

Collecting Eq.(39)-(41) results in the following expression for the average shear strain:

   
   
1 e 6(1 + S ) 1 − 3 (1 + S ) h  + ln 4 (1 + S ) 
3 3 2
γ=  +
1  
(43)
  2  
1 + S  h tan α
AVG
 H0 
1 − (1 + S )
h 
   
  H0  

The average shear strain imposed during one passage over the wing tip is critically de-
pendent on the wedge angle. In a typical continuous mixer approximately 20-25 units of
strain are imposed on the material during each pass. Comparison of Eq.(39), (40), and (41)
reveals that the contribution of region III is far from negligible: 65-75 % of the shear strain
is imposed on the material in region III and 5-10 % in region I, although the average shear
rate in region III is only one fifth of the average shear rate in regions I and II.
20 Mixing in the Farrel Continuous Mixer

Extensional Flow Components

The capacity to generate flows with significant extensional components is an important


characteristic of batch and continuous mixers. It is well known that extensional flows are
more efficient than shear flows in laminar mixing (Erwin, 1978). Even relatively minor
elongational flow components might play a critical role in dispersive mixing, helping in the
separation of agglomerate fragments broken by larger shear stresses (Manas-Zloczower and
Feke, 1988). In the model wedge, a significant extensional contribution is expected in
region II. However, models based on the lubrication approximation, which can provide
excellent estimates of the shear flow components for moderate values of the wedge angle
α, cannot be used to compute elongational flow components with adequate accuracy in the
same circumstances. Such components can be obtained by a simple procedure suggested
by Cogswell (1978) and developed by Tadmor (1986). According to this model, the aver-
age stretch rate near the entrance of the tip clearance εAVG = εAVG ( x) is:

h U
εAVG ≈ (1 + S ) tan α 0 (44)
H H

while the average total elongation imposed in region II can be estimated as:

ε AVG ≈ ln 32 (1 + S ) (45)

Comparison of Eq.(44)-(45) with Eq.(31) and (43) reveals that a significant elongational
flow component is generated only in the immediate vicinity of the tip clearance and, while
stretch and shear rates in that region are of the same order of magnitude in typical situ-
ations, the total elongational strain imposed on a material as it passes over the wing tip is
orders of magnitude less than the corresponding shear strain.

Shear Rate Distribution Function

For the purpose of this section it is convenient to define a dimensionless shear rate γ *
using the pure drag shear rate in the wing tip clearance U0 / h as the characteristic value:

γ h
γ* = (46)
U0

The cumulative shear distribution function F1 (γ*) is defined as the fraction of material
upon which a shear rate equal or lower than γ * is imposed during its passage over the ro-
Mixing in the Farrel Continuous Mixer 21

tor wing tip. Alternatively, F1 can be taken as the probability that a material element ex-
periences equal or lower than γ * during its passage over the rotor wing tip. The corre-
sponding differential distribution or probability density function f1 (γ*) is related to F1 by:

dF1
f1 = (47)
d γ *

Although the high shear zone of the model wedge includes regions I and II, and material
elements near the moving wall may experience a higher shear in region II than in region I,
the present treatment is restricted to the tip clearance (region I), since tracking material
particles in region II with purely analytical methods is almost impossible. The error
committed is minimal, and restricted to the low shear end of the spectrum, leading to more
“conservative” estimates.

The cumulative shear distribution function is given by:

1 y (γ *)
q ∫0
F1 (γ*) = u ( y′) dy′ (48)

where the upper limit of the integral, y = y (γ*) , is obtained by inverting the shear rate
profile, Eq.(27) with H = h. If S ≤ 1/3 the inversion is trivial resulting in:

y γ * − (1 − 3S ) (49)
=
h 6S

defined in the interval 1-3S ≤ γ * ≤ 1+3S; cf. Eq.(28), Introducing Eq.(9) with H = h into
Eq.(48), and taking into consideration Eq.(49), the following expression is obtained:

3(1 + 3S ) 2 − γ *2  γ * −2(1 − 3S )(1 + 12 S + 9 S 2 )


F1 = (50)
108S 2 (1 + S )

for the cumulative shear distribution function, and from Eq.(47):

γ *2 − (1 + 3S ) 2
f1 = (51)
36 S 2 (1 + S )

for the corresponding probability density. Both F1 and f1 are monotonic functions of γ *,
F1 increasing from 0 to 1 and f1 decreasing from 1 / [3S(1+S)] to 0, without relative max-
22 Mixing in the Farrel Continuous Mixer

ima, minima, or inflection points. Once the shear rate distribution function is known, it
can be used to obtain averages such as:

1+ 3 S 1
=γ *AVG ∫=
f (γ*) γ * d γ*
1−3 S
1
1+ S
(52)

cf. Eq.(29). If S > 1/3, the integration of Eq.(47) should take into account that the inverted
shear rate profiles is double-valued in the range 0 < γ * ≤ 3S-1, leading to:

 3(3S + 1) 2 − γ *2  γ *
  , 0 ≤ γ* ≤ 3S + 1
 54 S 2
(1 + S )
F1 =  (53)
 3(3S + 1) − γ *  γ * + 2(3S − 1)(1 + 12 S + 9 S )
2 2 2

 , 3S − 1 ≤ γ* ≤ 3S + 1
 108S 2 (1 + S )

for the cumulative shear distribution function, and

 (3S + 1) 2 − γ *2
 18S 2 (1 + S ) , 0 ≤ γ* < 3S − 1

f1 =  (54)
 (3S + 1) − γ * ,
2 2
3S − 1 < γ* ≤ 3S + 1
 36 S 2 (1 + S )

for the corresponding probability density. The probability density function f1 (γ*) has a
finite discontinuity at γ *= 3S-1. It is convenient to define f1 at that point as follows:

f1 (3S − 1) ≡ 1
2 ( lim
γ *→(3 S +1) +
f1 (γ*) + lim
γ *→(3 S +1) − )
f1 (γ*) =
1
2 S (1 + S )
(55)

The single-pass shear distribution functions presented here, along with the number of
passes distribution (NPD) functions, are the basis for the multiple-pass functions developed
in the next section.

Non-Newtonian Effects

Polymer melts seldom behave as Newtonian fluids. Since the wedge flow is essentially
a shear flow, a generalized Newtonian model that accounts for shear thinning at high rates
of shear is an accurate enough representation of the non-Newtonian characteristics of the
melt. The well known power law is the simplest of such models. Wagenknecht et al.,
(1986) use a power law in their calculations. However, since numerical methods are re-
Mixing in the Farrel Continuous Mixer 23

quired in order to obtain meaningful results, a more sophisticated model can be used with-
out increasing time and effort.

The rheological behavior of polymer melts in shear flow is well represented by the
Carreau-Yasuda model (Bird et al., 1990):

η0 e − β (T −T
R)

η= 1− n (56)
1 + ( λ0γ )a  a
 

where 0 is the low shear rate viscosity at the arbitrary reference temperature TR, is the
temperature coefficient of the viscosity, n is the power law index, 0 is a characteristic
time, and a is a dimensionless parameter. The parameters and n measure the sensitivity
of the melt viscosity to changes in temperature and shear rate, respectively, while 0-1 is a
measure of the shear rate at which the transition between shear independent (Newtonian)
and shear thinning behavior, and the Yasuda parameter a is related to the breath of that
transition, often connected with the breath of the molecular weight distribution (MWD) of
the material. Selecting a = 2 results in the standard Carreau model, which represents well
the rheological behavior materials with narrow MWD, a = 1 - n lead to the Cross model,
successfully used with broad MWD materials such as many commercial resins. Filled
materials and polymer blends require special treatment, beyond the scope of this presenta-
tion.

For non-Newtonian isothermal systems Eq.(5) is substituted by:

dp ∂  ∂ u 
= η  (57)
dx ∂ y  ∂ y 

where η = η( γ *) according to Eq.(56) with β = 0, and γ * is given by Eq.(27). Integration


of Eq.(4) and (57) with boundary conditions given by Eq.(6)-(7) in carried out by a nu-
merical scheme described elsewhere (Canedo and Valsamis, 1990). A series of numerical
simulations were conducted using realistic mixer geometries and rheological parameters
corresponding to commercial resins. The resulting flow patterns were qualitatively similar
to the Newtonian results presented earlier, but the non-Newtonian behavior of the melt had
a significant quantitative effect on many flow parameters related to the mixing character-
istics of the device. Flow quantities that depend on the material behavior along the whole
wedge, such as the maximum pressure rise, were affected by all rheological parameters of
24 Mixing in the Farrel Continuous Mixer

the material. Other quantities of interest, depending mostly on the material behavior in the
high shear zone, were strongly affected only by the power law index n, showing a weak
dependence on the characteristic time and the Yasuda parameter.

The pressurization coefficient S, defined by Eq.(8), increases as n decreases, as shown


in Figure 17, while the stagnation point is pushed toward the entrance of the wedge, result-
ing in a significant increase in the extension of region II, as shown in Figure 18. Conse-
quently, the high shear region is larger, as is the zone where the elongational components
of the flow play an important role. Some flow parameters, such as the location of the ma-
ximum pressure and the average shear rate in regions I and II, can be accurately estimated
using the equations for Newtonian materials presented in preceding sections, along with
the non-Newtonian value of the pressurization coefficient taken from Figure 17. Figure 19
shows calculated velocity and shear rate profiles in the rotor wing tip clearance. The ma-
ximum shear rate is significantly higher for shear thinning materials than for Newtonian
fluids, resulting in a highly nonuniform shear field in the gap (Figure 20).

Thermal Effects

The passage of a viscous polymer melt over the rotor wing tip results in energy dissi-
pation in the material. Part of the heat is removed through the chamber and rotor walls but,
in continuous mixers, a significant part is converted into internal energy, raising the
temperature of the melt. Moreover, since the melt viscosity is very sensitive to temperature,
temperature gradients inside the melt result in significant alterations of the velocity and
pressure profiles. A complete account of the thermal effects in the model wedge requires
the simultaneous solution of the momentum and energy balances, usually by numerical
methods (Wagenknecht et al., 1986, Meissner and Poltersdorf, 1992). Canedo and
Valsamis (1990) presented a summary account of the differential equations and boundary
conditions involved, along with a relatively simple numerical technique to solve them. The
treatment includes viscous heat dissipation, temperature and shear rate dependent viscosity,
and heat transfer through the walls. The numerical technique is restricted to the analysis of
regions I and II, the high shear regions where most of heat is generated, and uses the
lubrication approximation, neglecting thermal conduction along the wedge (x direction)
and convection across it (y direction). Figure 21 shows velocity, shear rate, and tem-
perature profiles for a typical, if somewhat extreme, case of a fractional melt index (MI =
0.1) HDPE in a large size FCM with adiabatic rotors. It is clear from this example that
Mixing in the Farrel Continuous Mixer 25

significant thermal effects are restricted to a thin layer next to the stationary wall, in the
wing tip clearance and its immediate vicinity.

The numerical approach just described can generate detailed results in particular cases,
but cannot replace an analytical approximation. Available models used in the analysis of
leak flow over extruder flights (see Rauwendaal and Ingen Housz, 1988, for a recent re-
view) cannot be applied in the present case. Extruder flight clearances are long and narrow
(e/h ≥ 50) compared to short and wide mixer wing tip clearances (e/h 3). Consequently,
the relatively slow flow in extruder clearances can be considered to be fully developed both
hydrodynamically and thermally (Meyer et al., 1978). In contrast, the fast flow in mixer
clearances is a developing (entrance) flow, in which velocity and temperature changes are
limited to a narrow boundary layer next to the stationary wall.

Pearson (1977, 1978) developed an approximate analytic solution for developing flows
with high heat generation, later modified to account for shear thinning behavior
(Richardson, 1986, 1989). According to this model the average temperature rise after one
pass of the material over the rotor wing tip is given by:

4 ln Na
∆TAVG ≈ (58)
β π Gz

and the maximum temperature rise, on the stationary wall, is given by:

1 Na
∆TMAX ≈ ln (59)
β Gz1− n

in terms of the dimensionless numbers:

qh (1 + S )U 0 h 2
=
Gz = (60)
αe 2α e

and

1+ n
ηβ q 2 (1 + S ) 2 mβ U 0 h1− n
=
Na = (61)
kh 2 4k

where k and are the thermal conductivity and diffusivity of the melt, is the melt vis-
cosity evaluated at the rolling pool temperature T0 and the pure drag tip clearance shear
26 Mixing in the Farrel Continuous Mixer

rate U0 /h, is the temperature coefficient of the viscosity, and m is the power law viscos-
ity parameter given in terms of the Carreau model, Eq.(56), by:

η0 e − β (T −T
0 R)

m≈ (62)
λ01− n

Gz is the Graetz number, which can be taken as measure of the importance of the convec-
tive heat transfer along the tip clearance relative to the conduction across it, and Na is the
Nahme-Griffith number, which measures the temperature increase due to viscous heat dis-
sipation relative to the temperature increase necessary to substantially change the melt vis-
cosity. Computations based on Eq.(58) reveal that in typical cases the average temperature
rise is very modest, usually less than 1 or 2°C. On the other hand, Eq.(59) predicts a sub-
stantial maximum temperature rise, frequently in the order of 20°C or more. The high
temperature is confined to a thin layer next to the stationary wall, while most of the mate-
rial flows through the tip clearance without being affected. Nevertheless, very high local
temperature increases (hot spots) are difficult to detect and may promote thermal degrada-
tion of the material, affecting the final product quality. It must be pointed out that Eq.(59)
probably overestimates ∆TMAX, since the simplified model on which is based assumes per-
fectly adiabatic rotors – seldom the case. Control of ∆TMAX is an important part of the
overall rotor design.

3.2. Global Flow Models


Global flow patterns in batch and continuous mixers are dominated by the flow in the
window of interaction between the rotors, a complex, three-dimensional, time-dependent
flow. The essential three-dimensionality of the flow field is exemplified by its three main
components: material flow along the mixer (axial component), material carried by each
rotor (tangential component), and material transfer between the rotors (radial component).
The three components are, in principle, equally important and interdependent. Circumfer-
ential flow is expected to be most important when the wing tips of the two rotors meet at
the window of interaction, while axial flow dominates when the opposing rotors are flat-to-
flat, and transfer of material from one chamber half to the other takes place mostly at
intermediate positions. For even speed operation the flow is periodic, with, a frequency
equal to the rotor angular speed. Rotor-to-rotor orientation changes not only with time (at
a given axial location) but also along the mixing chamber (at a given time), resulting in
Mixing in the Farrel Continuous Mixer 27

significant elongational components (squeezing flow). The local degree of filling of the
mixing chamber, a key factor in the distributive mixing performance of the machine, is
determined by the flow pattern in the window of interaction, along with the circumferential
flow over rotor wings.

The flow analysis network (FAN) technique has been applied to batch and continuous
mixers by White and coworkers (Kim and White, 1991, 1992; Hu and White, 1992) and by
Tadmor and coworkers (David et. al., 1990, 1992), but there are still questions about the
best way to use an essentially two-dimensional technique such as FAN to analyze a true
(and complex!) three dimensional flow problem. Manas-Zloczower and coworkers (Yang
and Manas-Zloczower, 1992; Wong and Manas-Zloczower, 1993) used two- and three-di-
mensional finite element method (FEM) techniques to model the flow in internal mixers.
Preliminary results of their analysis of the effect of rotor orientation on distributive mixing
(Wong and Manas-Zloczower, 1993) are in excellent agreement with the available, mostly
qualitative empirical evidence. Bigio and coworkers (Bigio et. al., 1991; Conner and
Bigio, 1992) conducted experimental and numerical studies of the flow in the nip region of
non-intermeshing counter-rotating twin screw extruders, a system similar in many ways to
the window of interaction of the FCM, including the effect of partially filled channels and
screw flight stagger, equivalent to rotor orientation in continuous mixers.

At present, the only tools readily available for the global flow analysis of continuous
mixers are semiempirical models based on population balances coupled with local fluid
mechanical analysis, whose output is usually expressed in terms of distribution functions.
In the next sections three of these models, relevant to the mixing characteristics of the
FCM, are examined in some detail,

Residence Time Distribution

The distribution of residence time or “exit age” in continuous mixers is closely related
to the global mixing characteristics of the machine. In addition, it is one of very few
mixing-related parameters that can be directly measured in actual systems under (almost)
normal operating conditions. A relatively large amount of information on the residence
time distribution (RTD) in single- and twin-screw extruders is readily available, but until
recently very little was known about the RTD in high intensity mixers. The mean residence
time of a material element inside a continuous flow device in steady state, such as the FCM
or part of it, is
28 Mixing in the Farrel Continuous Mixer

φ VF φ ρVF
t=
AVG = (63)
QV QM

where VF is the free volume of the processing chamber, φ is the global fill factor (hence
φVF is the occupied volume), QV and QM are the volumetric and mass flow rates, respec-
tively, and ρ is the material density. While QV and QM are operating parameters, and VF is
determined by the geometry of the machine, φ requires a knowledge of the global flow
patterns inside the mixer. In fact, Eq.(63) can be used to determine φ from experimental
measurements of the mean residence time. For quick calculations, φ 0.75 is considered
to be a “reasonable” approximation for the FCM under normal operating conditions.

The distribution of residence times depends on the detailed flow pattern, more specifi-
cally on the axial mixing characteristics of the machine. Two extreme cases of ideal flow
behavior can be identified: complete axial backmixing and complete axial segregation. In
the first case (ideal mixer: IM), all material elements inside the mixer at a given time have
the same chance of exit; in the second (plug flow: PF), material flows like plug from inlet
to outlet. In dimensionless terms, the RTD functions corresponding to these two ideal
models are (Danckwerts, 1953; Naor and Shinnar, 1963):

IM: e* = e − t* (64)

and

PF: e* δ (t * −1)
= (65)

where t* = t/tAVG is the dimensionless time and e* has been normalized so that the area
under it (the cumulative RTD) goes to 1 as t* goes to ¥ ; δ(ξ) is Dirac's delta function (a
spike at ξ = 0).

The two ideal models can be combined to form “mixed models”, widely used to study
backmixing in complex flow systems. One of the simplest is composed of N ideal mixers
(IM) units in series, each with the same effective volume φVF / N, so the mean residence
time in any one of them is tAVG / N. The dimensionless RTD of this tanks-in-series model
(N´ IM) is given by (Naor and Shinnar, 1963):

N
N´ IM: e* = t* N−1 e − N t* (66)
( N − 1)!
Mixing in the Farrel Continuous Mixer 29

and graphically presented in Figure 22 for several values of the parameter N. Increasing
values of N correspond to increasing axial segregation of the flow and decreasing axial
backmixing. It can be shown that in the limit N ® ¥ the plug flow RTD is recovered. A
quantitative measure of the breadth of the RTD is the dimensionless standard deviation or
coefficient of dispersion SD:

1
N´ IM:
SD = (67)
N

for the N´ IM model. The coefficient of dispersion can be taken as a good measure of the
degree of backmixing. Since the RTD of real systems can be measured and its coefficient
of dispersion computed from the experimental data, the parameter N corresponding to ac-
tual mixers can be determined. Figure 23 shows the dimensional RTD measured in a
2 FCM (bore diameter D0 = 50 mm) fitted with style 15 rotors, processing 100 kg/hr of an
extrusion grade polystyrene resin at 600 rpm (Canedo and Valsamis, 1991). Residence
time was measured at different settings of the exit orifice using a standard tracer technique.
The mean residence time and the coefficient of dispersion were computed from the ex-
perimental data, and the global fill factor φ and the backmixing parameter N computed
from these values using Eq.(63) and Eq.(67) respectively. The results, collected in Table I,
reveal that the degree of backmixing in the FCM is closely related to the fill factor, which
is controlled by opening of the exit orifice. A more complete study of the RTD in
continuous mixers, particularly on the effect of rotor design and operating conditions on
the backmixing characteristics, is now under way.

Table I. Experimental RTD and Model Parameters


(2 FCM, 100 kg/hr PS @ 600 rpm)

Orifice Fill Mean Residence Dispersion N´ IM Model


Setting Factor Time Coefficient Parameter
% φ tAVG , sec SD N

90 0.74 11.3 0.361 7.66

70 0.81 12.3 0.458 4.78

50 0.95 14.3 0.540 3.43


30 Mixing in the Farrel Continuous Mixer

Number of Passes Distribution

The analysis of the circumferential flow over the wing tips presented in the preceding
sections shows that material passing over the tip is frequently exposed to highly nonuni-
form shear fields. The nonuniformity of the shear is caused by the dynamic viscous pres-
surization in the wedge-like region preceding the wing tip and is further enhanced by
thermal and shear thinning effects. Consequently, a high shear is imposed on only a portion
of the material passing through and, since high shear levels are required for adequate
laminar and dispersive mixing, only a fraction of the material is properly mixed during one
pass over the wing tip. The way to overcome this problem in continuous mixers is to pro-
vide for the repeated passage of the material over the wing tips, with randomization of the
material elements between passes, to give a chance to all material elements to experience
the high shear levels generated in parts of the tip clearance.

This arrangement has some advantages. The circumferential “dynamic” pressurization


is an efficient way to transfer mechanical energy from the rotors to the polymer in a ma-
chine with partially filled mixing chamber. In contrast, designs that depend on blisters in-
terposed in the flow path to generate high shear stresses require a “static” pressure buildup
to overcome the flow restriction, leading to full mixing chambers, overheating, and poor
distributive mixing.

The average number of times the material passes over the tip of a rotor wing is simply
the ratio of the flow rate through the wing tip clearance to the flow rate throughout the
mixer. For a continuous mixer:

nW qL 2 ρπ ND0 hL(1 + S )
=K = (68)
QV QM

where nW is the number of rotor wings in parallel (nW = 4 for two-wing twin rotors), and L
is the axial length of the mixing chamber. The average number of passes depends on mixer
geometry and operating conditions, but is independent of the flow patterns in the mixing
chamber. Under normal operating conditions typical values of K in the standard 3.5 L/D
FCM mixing chamber are between 20 and 30. Not all the material flowing through the
mixer passes the same number of times through the high shear regions, and since effective
mixing requires repeated passages, it is important to know, for a given machine, material,
and set of operating conditions, the fraction that passes fewer times that some empirically
determined minimum number of passages.
Mixing in the Farrel Continuous Mixer 31

Manas-Zloczower and Tadmor introduced the number of passes distribution (NPD)


function in their modeling of mixing in batch intensive mixers, and computed the NPD
over the screw flights of single screw extruders from a theoretical analysis of melt flow in
the extruder channels (Manas-Zloczower et al., 1982; Manas-Zloczower and Tadmor,
1983). The concept was later generalized by Tadmor (1986, 1988) to describe the disper-
sive mixing characteristics of batch and continuous mixers based on population balances
and ideal models of global flow behavior. No satisfactory experimental methods are pres-
ently available to measure the NPD, or even the average number of passes through the high
shear region in polymer processing machinery. However, preliminary results of
measurements of stress levels in twin screw compounders based on the breakage of stress-
sensitive glass spheres (Curry and Kiani, 1990, 1991) suggest that this may change in the
near future.

The two extreme cases of ideal flow behavior, complete axial backmixing (IM) and
complete axial segregation (PF), are represented schematically in Figure 24a-b. Notice that
in the plug flow model, complete segregation is assumed in the axial direction and ideal
mixing is assumed between passes through the high shear region, The corresponding NPD
functions are:

Kp
IM: gp = (69)
(1 + K )1+ p

and

K p e− K
PF: gp = (70)
p!

where gp is the fraction of material that passes p times over the rotor wing tip during its
stay in the mixing chamber. The parameter K is the average number of passes given by
Eq.(68):


K ≡ ∑ pg p (71)
p =0

The coefficient of dispersion is a measure of the breadth of the distribution relative to the
mean value. For the two ideal models discussed here:
32 Mixing in the Farrel Continuous Mixer

1
IM: S=
D 1+ (72)
K

and

1
PF: SD = (73)
K

Figures 25 shows the NPD function of the plug flow model for several values of the pa-
rameter K.

As a first approximation and under normal conditions, the plug flow model appears to
be an adequate representation of the FCM’s behavior. However, in some circumstances
mixer design or operating conditions induce a significant amount of axial backmixing. A
typical case occurs when mixing inserts are employed to increase the fill factor in the
mixing chamber. The tanks-in-series model (N´ IM) described in the preceding section and
schematically represented in Figure 24c can be used to model a moderate amount of axial
backmixing. The number of ideal mixers in series N can be estimated from coefficient of
dispersion of the RTD measured in similar conditions, as indicated before. The NPD for
the tank-in-series model can be obtained by:

p i m
N´ IM: g p = ∑ g ′p −i ∑ gi′− j  ∑ g m′ − n g n′ ( N times) (74)
=i 0=j 0 =n 0

where g′p is the NPD of an ideal mixer with K′ = K/ N given by Eq.(69). The resulting
NPD function is (Canedo and Valsamis, 1991):

( p + N − 1)! ( K / N) p
N´ IM: =gp ⋅ (75)
p ! ( N − 1)! (1 + K / N )1+ p

with a coefficient of dispersion given by:

1 1
N´ IM:
=
SD + (76)
K N

Figure 26 shows the NPD function for the tanks-in-series model with K = 20 and several
values of the backmixing parameter N . Other models of nonideal flow, such as a combi-
nation of plug flow and ideal mixer in series (PF+IM), can be used to account for axial
Mixing in the Farrel Continuous Mixer 33

backmixing. In most cases, results very similar to the ones presented here for the tanks-in-
series model are obtained (Canedo and Valsamis, 1991).

Maximum Shear Distribution

Solid agglomerates and liquid drops suspended in a polymer melt are maintained to-
gether by cohesive forces between aggregates and interfacial tension, respectively, and to
disperse them, a larger force must be imposed by the flow field. In continuous mixers, that
force (per unit area) is essentially the shear stress generated in and near the rotor wing tips.
In polymer melts the shear stress is proportional to a positive power of the shear rate (1 for
Newtonian material, the power law index n for a shear thinning material). While other
factors, such as the time and manner in which the shear is delivered, and the elongational
components of the deformation, often play an important role in the dispersion process, the
intensive mixing characteristics of the machine are mainly related to the maximum shear
rate that the flow field imposes on the material. In most practical situations the shear rate
profile in the high shear zone of a mixer is highly nonuniform. However, the same
material passes repeatedly over the wing tips, with randomization of the fluid elements
between passes. A global analysis of the shear distribution inside the mixer should take all
these effects into consideration, leading to useful estimates of the maximum shear rate and
its distribution.

Canedo and Valsamis (1989) introduced the maximum shear distribution (MSD) func-
tion as a tool for modeling this process. The cumulative MSD function F = F( γ *) is the
fraction of material leaving the mixer after being subjected, at least once, to a dimension-
less shear rate equal to or higher than γ *. Alternatively, F can be taken as the probability
that a dimensionless shear rate equal or greater than γ * has been imposed on a material
element during its passage through the mixer. The MSD is computed by composing the
single-pass shear rate distribution function F1, obtained by an analysis of circumferential
flow over the wing tips, with the number of passes distribution (NPD) function gp, ob-
tained by the techniques outlined in the preceding section. Assuming complete randomi-
zation of the material elements after each pass:


F = 1 − ∑ F1 g p
p
(77)
p =0

A differential MSD or probability density function f can be defined as the slope of F( γ *):
34 Mixing in the Farrel Continuous Mixer


f = − f1 ∑ p F1
p −1
gp (78)
p =0

where f1 is the single-pass differential shear distribution in the tip clearance given by
Eq.(47).

For Newtonian isothermal flow, the single-pass shear distribution functions F1 and f1
are given by Eq.(50)-(51) or Eq.(53)-(54) according to the value of the pressurization co-
efficient S. For non-Newtonian and/or nonisothermal flow, the single-pass shear distribu-
tion functions can only be obtained numerically. The NPD function can be estimated once
a global flow model has been assumed. Eq.(70) for axially segregated, plug flow behavior,
or Eq.(75) to account for axial backmixing, may be used. In the simplest case, Newtonian
isothermal flow and ideal plug flow, the MSD depends on only two dimensionless pa-
rameters: the pressurization coefficient of the circumferential flow S and the average num-
ber of passes through the rotor wing tip clearance K. Other parameters are introduced by a
more realistic approach, such as the power law index n for non-Newtonian melts, and the
backmixing parameter N for nonideal axial flow. If S £ 1/3 the cumulative MSD function
is given by:

p

1  3(1 + 3S ) − γ *  γ * −2(1 − 3S )(1 + 12 S + 9 S ) 
2 2 2

F = 1− e −K
∑ 
p!  108S 2 (1 + S )
K

(79)
p =0  

within the range 1 - 3S £ γ * £ 1 + 3S. The probability density can be easily computed by
differentiation:

p −1
(1 + 3S ) 2 − γ* 2 ∞
1  3(1 + 3S ) − γ *  γ * −2(1 − 3S )(1 + 12 S + 9 S ) 
2 2 2

f =
36 S (1 + S )
2
Ke − K ∑
p =1 ( p − 1)!

 108S (1 + S )
2
K

(80)
 

Figure 27 plots f = f( γ *) for S = 1/3 and several values of K. In most situations of practi-
cal interest (K > 20), virtually all material passing through the FCM is subjected, at least
once, to shear rates substantially higher than the average shear rate in the wing tip clear-
ance.

Once the MSD has been computed, it can be used to obtain an estimate of the typical
maximum shear rate encountered by the material in the mixing chamber. The average
maximum shear rate, given by
Mixing in the Farrel Continuous Mixer 35

γ*MAX
U
< γ MAX > = 0
h ∫
0
γ * f (γ*) dγ * (81)

is a meaningful estimate of the prevailing shear rate inside the mixer, probably better than
γ AVG or γ MAX as far as the intensive mixing characteristics of the machine are concerned.

3.3. Scale-up Considerations


Mixer dimensions, with the possible exception of rotor tip clearance and land width
(and blister clearance in two-stage mixers), are geometrically scaled based on the mixer
chamber bore diameter D0 . Thus, the length of the processing chamber L is proportional to
the bore diameter, the mixer L/D is independent of size, and the free volume of the mixer
VF is, in first approximation, proportional to the cube of the bore diameter. It is customary
to scale the volumetric throughput QM based on equivalent mean residence time in the
mixer tAVG , Eq.(63). If the global fill factor φ is assumed to be independent of mixer size,
this leads to:

3
D 
QM 2 = QM 1  02  (82)
 D01 

where subscripts 1 and 2 represent conditions in two mixers (e.g. a small process lab unit
and a full size production unit). The usual practice is to scale rotor speed N based on
equivalent mean shear rate in the rotor tip clearance γ AVG , Eq.(29) or Eq.(31). If the dy-
namic pressurization factor S is assumed to be independent of mixer size, this leads to:

h2 / D02
N 2 = N1 (83)
h1 / D01

Eq.(82)-(83) are the two standard scale-up rules for continuous mixers and can be used to
compute two important operating parameters of a production unit, throughput and rotor
speed, based on the corresponding values measured in the process lab. This procedure
guarantees that two key parameters related to mixing performance, residence time in the
mixer and shear rate in the tip clearance, will have the same value in the two machines.
The standard scale-up procedure leaves other operating variables, such as discharge orifice
opening or gear pump suction pressure, free to be adjusted in the field, to match fill factors
and correct minor variations, allowing a flexible operation of the machine. Moreover,
36 Mixing in the Farrel Continuous Mixer

Eq.(83) does not completely determine rotor speed and tip clearance, but rather establishes
their values relative to each other, introducing a degree of freedom in mixer design that can
be used to accommodate other processing characteristics.

Dispersive mixing in continuous mixers depends on the repeated passage of material


through the high shear regions at or near the rotor wing tips, measured in first approxima-
tion by the average number of passes over the wing tips K, Eq.(68). If the mixing chamber
length is proportional to the bore diameter and the pressurization coefficient is assumed to
be independent of mixer size, then scaling the mixer based on equivalent number of passes
leads to:

h2 / D02 h /D
N2 3
= N1 1 01 3 (84)
QM 2 / D02 QM 1 / D01

Eq.(84) may be used along with Eq.(82) as an alternative to Eq.(83), or along with Eq.(83)
as a complement of Eq.(82), to determine rotor speed and tip clearance. In the later case,
throughput is proportional to the cube of the chamber bore diameter and Eq.(84) is reduced
to:

h1 / D01
N 2 = N1 (85)
h2 / D02

Comparison of Eq.(83) and Eq.(85) shows that scaling based on equivalent shear rate in the
rotor tip clearance, equivalent number of passes through it, and equivalent residence time
in the mixer is possible only if the rotor angular speed is independent of mixer size and the
tip clearance is geometrically scaled, that is:

D02
=N 2 N=
1 , h2 h1 (86)
D01

Eq.(82) and Eq.(86) account for all major factors that affect extensive and intensive mixing
in continuous mixers under isothermal conditions. Unfortunately, there is no way to
incorporate the scaling of thermal effects in a single, consistent set of scale-up rules.

The specific rate of heat transfer through the mixing chamber walls is proportional to
the ratio of heat transfer surface to internal volume. In a properly scaled continuous mixer,
surface and volume are approximately proportional to the square and cube of the chamber
bore diameter respectively, and the surface-to-volume ratio is inversely proportional to the
Mixing in the Farrel Continuous Mixer 37

diameter. As a result, larger mixers dissipate heat less efficiently than smaller ones. In
addition, the specific rate of heat generation in melt pools is directly related to the linear
speed of material circulation, which is proportional to U0 . Constant linear speeds imply
angular speeds inversely proportional to the chamber bore diameter and, if shear levels in
the tip clearance are to be preserved as well, the rotor wing tip clearance must be
independent of mixer size, resulting in:

D01
=N 2 N=
1 , h2 h1 (87)
D02

as an alternative to Eq.(86). However, Eq.(87) is seldom used to scale-up continuous mix-


ers.

A significant part of the heat generation in continuous mixers takes place at, or in the
immediate vicinity of, the rotor wing tips, and convection in the circumferential direction
(not conduction in the radial direction) is the main mechanism of heat dissipation. Tem-
perature profiles near the tips are far from developed, and under these conditions higher
linear tip speeds do not necessarily lead to higher temperatures; cf. Eq.(58). As a result,
the scaling rule that preserves rotor angular speeds, Eq.(86), may be preferred. However,
the maximum temperature developed in tip clearance, Eq.(59), has a logarithmic depend-
ence on the linear tip speed, and scale-up according to Eq.(86) may lead to the formation of
hot spots detrimental to final product quality.

In some circumstances (e.g., high local fill factors), most of the heat may be generated
in rolling pools under more developed thermal conditions, for which linear speeds are the
key factor. Other factors, such as axial recirculation patterns, axial variation of the local
fill level, and pool homogenization in the window of interaction (material exchange be-
tween the low and high shear regions of the pool), should also be taken into account. Thus,
scaling procedures must consider design features and operating conditions that affect mixer
fill factors, axial recirculation, and melt pool renewal. In practice, a compromise between
the requirements of Eq.(86) and Eq.(87) is reached, and intermediate values of rotor speed
and tip clearance are selected:

D01 D02
N1 < N 2 < N1 , h1 < h2 < h1 (88)
D02 D01
38 Mixing in the Farrel Continuous Mixer

The particular choice of scaling factors is determined by the specific process application
and is, at present, guided more by experience than by calculation, part of the art of mixing.

In conclusion, it is apparent that several design and operating parameters can be used
alone or in combinations to obtain a specific machine performance. However, these pa-
rameters are not totally independent; they are interrelated in complex ways, and in many
cases they have opposing effects on machine behavior. Therefore, the scale-up procedure,
in addition to satisfying certain minimum criteria, must be capable of adjusting these pa-
rameters into a compromise design that will provide optimum performance in a particular
application.

4. ROTOR DESIGN

The standard rotor designs shown in Figure 2 are classified in two distinct categories
based on their processing characteristics. Rotor styles 7 and 15 are single-stage designs, in
which a single orifice at the exit of the mixer is used to control the fill level in the mixing
chamber at any given rotor speed and flow rate. Rotor styles 22 and 24X are two-stage
designs, in which the performance of each section is independently controlled by means of
separate adjustable devices located at the end of each stage. The first stage is controlled by
an adjustable gate over a rotary blister. The second stage is controlled by an adjustable
orifice at the exit of the mixer or, in side discharge models (UMSD), by a gear pump di-
rectly attached to the exit opening. A more detailed account of the structure and processing
characteristics of the standard rotor designs for the FCM can be found elsewhere (Canedo
and Valsamis, 1993).

4.1. Single-Stage Rotors


Rotor style 15 provides a higher mixing intensity than rotor style 7. It is the most
widely used design, especially in compounding applications involving difficult dispersion
of additives or fillers. As a result of their lower intensity mixing characteristics, rotors of
style 7 are particularly useful in processing shear- and temperature-sensitive materials.
Style 15 rotors have a relatively short forward pumping section near the feed end, followed
by a reverse pumping section and a neutral section near the exit end of the mixer. Style 7
rotors have a longer forward pumping section, and the neutral section over the exit orifice
is absent. In style 15 rotors the apices of the wings are shifted toward the feed end,
Mixing in the Farrel Continuous Mixer 39

compared to style 7 rotors. Consequently, fluxing in style 15 rotors occurs earlier and in a
narrower section of the mixer, while the melting and mixing operations are spread over a
longer section in style 7 rotors. Rotor style 15 provides longer melt residence time and
number of passes than rotor style 7 at equivalent operating conditions. It should be noted
that the cross section is identical in both rotor styles; moreover, both operate on the prin-
ciple that dispersive mixing is accomplished by the repeated passage of material through
high stress regions, and both rotor designs ensure intimate mixing of the dispersed material
within the polymer matrix.

A special 7/15 rotor combination provides unusual processing characteristics when the
rotors are operated at synchronous speed and at a specific orientation to each other. This
arrangement has been found to be the most appropriate in processing fractional melt index
materials, where temperature homogeneity of the melt discharge is critical to end product
properties. The processing characteristics of the 7/15 rotor combination are further exam-
ined in this section, where the effect of rotor wing tip clearance on mixer performance is
established in compounding HDPE in a size 4 LMSD (100 mm chamber bore diameter)
attached to a gear pump. Details of the experimental program can be found elsewhere
(Valsamis and Canedo, 1991).

Figure 29 presents axial temperatures profiles, measured at the positions indicated in


Figure 28 (T6 and T7 were located in the transfer channels, before and after the gear pump,
respectively), for a flow rate of 200 kg/hr at rotor speeds in the range of 200-500 rpm, with
rotor tip clearances of 2.4 mm (Figure 29a) and 3.2 mm (Figure 29b): a temperature
increase is seen along the mixer length and with increasing rotor speed at equivalent flow
rates. Temperature levels are higher for reduced rotor tip clearances and at the early stages
of the mixing process, while exit temperatures are approximately the same for the two rotor
tip clearances. It appears that axial melt pools are formed before the apex of style 7 rotor
(position T2-T3) and prior the mixing insert (position T4-T5), as revealed by the flattening
of the temperature profiles. The extension of the pools is larger at lower rotor speeds and
higher rotor tip clearances. The melt pools are promoted by the opposite pumping action
of the rotor wings at these locations, along the principles discussed earlier.

Figure 30, which shows the effect of rotor orientation on the temperature profiles, in-
dicates that mixer performance can be affected by the orientation of one rotor relative to
the other. The lowest temperature was obtained when, at the beginning of the mixing sec-
tion, the two rotors were tip-to-tip, with the fed wing of one rotor opposing the unfed wing
40 Mixing in the Farrel Continuous Mixer

of the other. The unfed wing is the one that is the continuation, in the mixing section, of
the feed screw flight.

Figure 31 presents the flow rate dependence of the melt temperature at three different
locations (T1, T3, T5) for a rotor speed of 300 rpm, with rotor tip clearances of 2.4 mm
(Figure 31a) and 3.2 mm (Figure 31b). Figure 31 shows, as expected, a decrease in melt
temperature with increasing flow rate and rotor tip clearance. The effect of flow rate is
more pronounced at the earlier stages of the mixing process and the lower flow rates.
Furthermore, Figure 31b shows the formation of a large axial pool between the rotor apices
at high flow rates. The flow rate dependence of the specific energy input and the melt
index shift are presented in Figures 32 and 33, respectively, for the two rotor tip clearances
investigated. Specific energy inputs steadily decrease with increasing flow rate as shown
in Figure 32. The rate of decrease is higher at lower flow rates. The effect of rotor tip
clearance on specific energy input is minimal the low rates, but becomes significant at high
rates. At this point, an increase in rotor tip clearance results in a decrease of specific power
input. Figure 33 shows a decrease in melt index shift with flow rate at constant rotor
speed. The rotor tip clearance appears to have little effect on melt index shift, except at
high flow rates, where larger rotor tip clearances result in lower shifts. The discontinuities
observed in Figures 32 and 33 at 250 kg/hr are due to changes in mixer discharge pressure.

4.2. Two-Stage Rotors


Rotor styles 22 and 24X offer additional process options not available in single-stage
designs. Mixing intensity in the first stage can be controlled independently of the second
stage. These rotors also offer the option of downstream addition or venting of the molten
material. Both functions can be accomplished through a single opening in the mixer body
by the use of a special hopper design, in which case venting is restricted to atmospheric
pressures. The overall length of the style 22 rotor is identical to that of single-stage rotors,
and thus it can be interchanged in standard L/D mixer bodies. The style 24X rotor is about
1.5 L/D longer and is used in applications calling for longer residence times, such as post-
reactor processing of polypropylene (Valsamis and Canedo, 1989b).

The processing characteristics of the style 24X rotors are examined next, by a series of
tests conducted in a size 4 LMSD (100 mm chamber bore diameter) equipped with a gear
pump. Details of the experimental program can be found elsewhere (Valsamis et al., 1991;
Donoian et al., 1993). Two polypropylene resins were with initial MFIs (230°C / 2.16 kg)
Mixing in the Farrel Continuous Mixer 41

of 1.2 and 2.6 were used in the tests, with flow rates varying from 175 kg/hr to 375 kg/hr,
and rotor speeds of 375 and 500 rpm. Mixer performance was controlled by means of the
gear pump suction pressure.

Figure 34 shows specific energy input as a function of gear pump suction pressure at
different throughput rates and mixer rotor speeds for the two resin grades. Increased spe-
cific energy inputs are observed with increased gear pump suction pressures for both mixer
rotor speeds. The lower limit of gear pump suction pressures is established as the point at
which a stable mixer operation is obtained. It is important to note that with increased flow
rates, the minimum gear pump suction pressure shifts to higher values. However, as ex-
pected, the associated specific power input is lower at the higher rates. Although the mixer
fill level increases with increasing flow rate and gear pump suction pressure, for a fixed
mixer rotor speed, the average number of material passes though the high shear region (i.e.,
the rotor wing tip clearance) is lower at the higher throughput rates, which essentially
yields lower specific power inputs. There are also indications that at different rotor speeds
throughput can be adjusted to obtain equivalent fill levels and thus similar mixer operation.
This behavior is exhibited by the overlapping low specific energy curves of Figure 34.
This figure also shows that, at any given rotor speed and throughput, the specific energy
input can be controlled through the use of the gear pump suction pressure. Also, within
limits, specific power input levels can be maintained at different throughputs by adjusting
the gear pump suction pressure.

Figure 35 shows exit melt flow index as a function of gear pump suction pressure at
different throughput rates and mixer rotor speeds for the two resin grades. It indicates two
important trends: first, at any given speed and suction pressure, the lower the flow rate the
greater the specific energy, and thus the higher the MFI shift; second, the effect of suction
pressure on MFI shift increases with decreasing flow rate. In several runs, a peroxide in-
itiator was added to the feed to modify the material by controlled degradation.

Figure 36 plots the exit MFI (initial MFI 1.2) as a function of peroxide level with the
mixer operated at constant rotor speed and flow rate. By steadily increasing the level of
peroxide from 0 to 1600 ppm, the melt index of the resin can be shifted from the initial
value of 1.2 to the final value of 120, a cracking ratio of 100:1. The suction pressure did
not appear to have a clear effect on the cracking efficiency of the peroxide, suggesting that
the controlled degradation process was completed in the first stage of the mixer.
42 Mixing in the Farrel Continuous Mixer

5. CONCLUSION

In this chapter we have attempted to cover the basic technology associated with con-
tinuous mixers, from a description of structure and principles of operation, to the mathe-
matical modeling of flow and thermal fields, followed by some experimental results gen-
erated by different rotor designs. The analysis is by no means complete, and a substantial
effort must be devoted toward the elucidation of the detailed fluid and thermal characteris-
tics of the continuous mixers. This is part of an ongoing effort. However, even a simpli-
fied analysis, as presented here, increases our knowledge of the interaction of machine de-
sign and operating conditions on mixing mechanisms, and the performance of continuous
mixers is now better understood. As a result, complex multistage rotor designs with vari-
able rotor tip clearance profiles are now used in various process applications. With the
high level of sophistication now common in materials development, it is expected that ro-
tor designs even more specialized than the ones presented here will evolve to meet the
plastics industry's needs.
Mixing in the Farrel Continuous Mixer 43

NOMENCLATURE

All symbols are defined at their first appearance; symbols used only once are not included
here.

a constant in Carreau-Yasuda viscosity model (“Yasuda constant”)


D0 mixer chamber bore diameter
e rotor wing tip width, measured normal to the axis
e* dimensionless Residence Time Distribution (RTD) function
f Maximum Shear Distribution (MSD) function
f1 single-pass shear rate distribution function in wing tip clearance
F cumulative Maximum Shear Distribution (MSD) function
F1 cumulative single-pass shear rate distribution function in wing tip clearance
gp Number of Passes Distribution (NPD) function
Gz Graetz number in the wing tip clearance
h rotor wing tip to mixer chamber wall clearance
H gap thickness, in the model wedge
H0 maximum gap thickness, in the model wedge
HS gap thickness at the stagnation point, in the model wedge
HM gap thickness at the maximum pressure point, in the model wedge
k melt thermal conductivity
K average number of passes over the rotor wing tip
L mixing chamber axial length
n power law index
N rotor angular speed (rpm)
N number of ideal mixers in series, in N×IM backmixing model
Na Nahme-Griffith number in the wing tip clearance
p pressure; number of passes over the wing tip
q volumetric flow rate (per unit axial length) over the wing tip
44 Mixing in the Farrel Continuous Mixer

QV volumetric throughput
QM mass throughput
S pressurization coefficient, in the model wedge
SD coefficient of dispersion (relative standard deviation)
t residence time in the mixer
t* dimensionless residence time in the mixer
T temperature
u circumferential velocity, in the model wedge
uS velocity along the stagnation streamline
uMIN minimum velocity in the rolling pool,
uMAX maximum velocity in the high shear zone
U0 linear wing tip speed
x circumferential coordinate, along the model wedge
y radial coordinate, across the model wedge
y0 location of the zero velocity in the rolling pool
yS location of the stagnation streamline
yMIN location of the minimum velocity in the rolling pool
yMAX location of the maximum velocity in the high shear zone

α leading face wedge angle; melt thermal diffusivity


β temperature coefficient of the melt viscosity
γ shear strain
γ shear stress
γ * dimensionless shear stress
η melt viscosity
η0 low shear rate melt viscosity
λ0 constant in Carreau-Yasuda viscosity model (characteristic time)
ρ melt density
φ global fill factor
Mixing in the Farrel Continuous Mixer 45

REFERENCES

Ahlefeld, E. H., Baldwin, J. J., Hold, P., Rapetski, W. A., Scharer, H. R., U. S. Patent (to
Farrel Corporation) 3,154,808 (1964).
Ahlefeld, E. H., Baldwin, J. J., Hold, P., Rapetski, W. A., Scharer, H. R., U. S. Patent (to
Farrel Corporation) 3,239,878 (1966).
Bergen, J. T., “Mixing and dispersing processes”, in Processing of Thermoplastic Materi-
als, E. C. Bernhardt (ed.), Reinhold, New York (1959).
Biesenberger, J. A., Sebastian, D., Principles of Polymerization Engineering. Wiley-Inter-
science, New York (1983).
Bigio, D., Baim, W., Wigginton, M., “Mixing in non-intermeshing twin screw extruders”.
Intern. Polymer Proc. 6, 172-176 (1991).
Bird, R. B., Armstrong, R. C., Hassager, O., Dynamics of Polymeric Liquids, Vol. 1: Fluid
Mechanics, 2nd. ed. Wiley, New York, 1990.
Bolen, W. R., Colwell, R. E., “Intensive mixing”. SPE J. 14 (3), 24-28 (1958).
Canedo, E. L., Valsamis, L. N., “Modeling mixing in continuous mixers”. SPE ANTEC
Tech. Papers 35, 116-119 (1989).
Canedo, E. L., Valsamis, L. N., “Non-Newtonian, non-isothermal flow between non-paral-
lel plates in relative motion. Applications to mixer design”. SPE ANTEC Tech. Papers
36, 164-167 (1990).
Canedo, E. L., Valsamis, L. N., “Modeling mixing in continuous mixers. Effect of back-
mixing”. SPE ANTEC Tech. Papers 37, 141-145 (1991).
Canedo, E. L., Valsamis, L. N., “Farrel Continuous Mixer Systems for Plastics Com-
pounding”, in Plastics Compounding Handbook, D. Todd (ed.), Carl Hanser, Munich
(1993) in press.
Cheng, J. J., Manas-Zloczower, I., “Hydrodynamic analysis of a Banbury mixer”. Polymer
Eng. Sci. 29, 701-708 (1989a).
Cheng, J. J., Manas-Zloczower, I., “Hydrodynamic analysis of a Banbury mixer. 2-D flow
simulations for the entire mixing chamber”. Polymer Eng. Sci. 29, 1059-1065 (1989b).
Cheng, J. J., Manas-Zloczower, I., “Flow field characterization in a Banbury mixer”.
Intern. Polymer Proc. 5, 178-183 (1990).
46 Mixing in the Farrel Continuous Mixer

Cogswell, F. N., “Converging and stretching flow: a compilation”. J. Non-Newtonian Fluid


Mech. 4, 23-38 (1978).
Conner, J. H., Bigio, D., “A numerical analysis of the mixing performance on the non-in-
termeshing twin screw extruder nip”. SPE ANTEC Tech. Papers 38, 31-34 (1992).
Curry, J., Kiani, A., “Measurement of stress level in continuous melt compounders”. SPE
ANTEC Tech. Papers 36, 1599-1602 (1990).
Curry, J., Kiani, A., “Experimental identification of the distribution of fluid stresses in
continuous melt compounders”. SPE ANTEC Tech. Papers 37, 114-118 (1991).
Danckwerts, P. V., “Continuous flow systems. Distribution of residence times”. Chem.
Eng. Sci. 2, 1-13 (1953).
David, B., Sapir, T., Nir, A., Tadmor, Z., “Modelling twin rotor mixers and extruders, Part
1”. Intern. Polymer Proc. 5, 155-163 (1990).
David, B., Sapir, T., Nir, A., Tadmor, Z., “Modelling twin rotor mixers and extruders, Part
2”. Intern. Polymer Proc. 7, 204-211 (1992).
Donoian, G. S., Canedo, E. L., Valsamis, L. N., “Post-reactor processing in the Farrel
Continuous Mixer using two-stage rotors”. 8th. Intern. Conf. Polyolefins, Houston
(1993).
Erwin, L., “New fundamental considerations on mixing in laminar flow”. SPE ANTEC
Tech. Papers 24, 488-493 (1978).
Farrel Corporation, Understanding the Farrel Continuous Mixer. A Self-Instruction
Training Program. Farrel, Ansonia (1965).
Farrel Corporation, Principles of Operation for the FCM. Farrel, Ansonia (1971).
Hold, P., “Continuous mixer with counter-rotating non-intermeshing rotors”. Adv. Polymer
Technol. 4, 281-289 (1984).
Hu, B., White, J. L., “Simulation of 3-dimensional flow in internal mixers”. Intern.
Polymer Proc. 8, 18-29 (1993).
Kearney, M. R., “Mixing in continuous internal mixers”, in Mixing in Polymer Processing,
C. Rawendaal (ed.), Marcel Dekker, New York (1991).
Kim, M. H., White, J. L., “Non-Newtonian and non-isothermal modeling of 3D-flow in an
internal mixer”. Intern. Polymer Proc. 6, 103-110 (1991).
Mixing in the Farrel Continuous Mixer 47

Kim, M. H., White, J. L., “Simulation of flow in a Farrel Continuous Mixer”. Intern.
Polymer Proc. 7, 15-19 (1992).
Manas-Zloczower, I., Nir, A., Tadmor, Z., “Dispersive mixing in internal mixers. A theo-
retical model based on agglomerate rupture”. Rubber Chem. Technol. 55, 1250-1285
(1982).
Manas-Zloczower, I., Tadmor, Z., “The distribution of number of passes over the flights in
single screw extruders”. Adv. Polymer Technol. 3, 213-221 (1984).
Manas-Zloczower, I., Feke, D. L., “Analysis of agglomerate separation in linear flow
fields”. Intern. Polymer Proc. 2, 185-190 (1988).
Meyer, H. E. H., Ingen Housz, J. F., Gorissen, W. C. M., “Temperature development in the
leakage flow of screw extruders”. Polymer Eng. Sci. 18, 288-292 (1978).
Meissner, K., Reher, E. O., “Zur Modellierung des Mischprozesses hochviskoser Medien
im Innenmischer”. Plaste und Kautschuk 26, 272-275 (1979).
Meissner, K., Bergmann, J., Reher, E. O., “Über den Mischprozess im Innenmischer unter
Beachtung viskoplastischer Werkstoffeingenschaften”. Plaste und Kautschuk 27, 147-
152 (1980).
Meissner, K., Poltersdorf, B., “Model development for an internal mixer”. Intern. Polymer
Proc. 7, 3-14 (1992).
Meyer, H. E. H., Ingen Housz, J. F., Gorissen, W. C. M., “Temperature development in the
leakage flow of screw extruders”. Polymer Eng. Sci. 18, 288-292 (1978).
Naor, P., Shinnar, R., “Representation and evaluation of residence times”. Ind. Eng.
Fundam. 2, 278-286 (1963).
Nassehi, V., Freakley, P. K., “Spreader blade analogy of flow past an internal mixer rotor”.
Intern. Polymer Proc. 6, 91-97 (1991).
Ottino, J., The Kinematics of Mixing: Stretching, Chaos, and Transport. Cambridge Uni-
versity Press, Cambridge (1989).
Patel, S. R., “Farrel side discharge continuous mixer/gear pump”. 5th. Intern. Conf.
Polyolefins, Houston (1987).
Pearson, J. R. A., “Variable-viscosity flows in channels with high heat generation“. J.
Fluid Mech. 83, 191-206 (1977).
Pearson, J. R. A., “Polymer flows dominated by high heat generation and low heat trans-
fer”. Polymer Eng. Sci. 18, 222-229 (1978).
48 Mixing in the Farrel Continuous Mixer

Rauwendaal, C. J., Ingen Housz, J. F., “Temperature and velocity profiles in drag flow of a
temperature dependent power law fluid”. Intern. Polymer Proc. 3, 123-133 (1988).
Richardson, S. M., “Injection moulding of thermoplastics: freezing of variable-viscosity
fluids, I. Developing flows with very high heat generation”. Rheol. Acta 25, 180-190
(1986).
Richardson, S. M., “Simplified geometry models”, in Computer Modeling for Polymer
Processing, C. L. Tucker (ed.), Carl Hanser, Munich (1989).
Scharer, H. R., “Gear pump discharge FCM”. Adv. Polymer Technol. 5, 65-71 (1985).
Tadmor, Z., Klein, I., Engineering Principles of Plasticating Extrusion. Reinhold, New
York (1970).
Tadmor, Z., Gogos, C., Principles of Polymer Processing. Wiley-Interscience, New York
(1979).
Tadmor, Z., “Mixing processes in polymer processing”, in Integration of Fundamental
Polymer Science and Technology, L. A. Kleintjens, P. J. Lemstra (eds.), Elsevier,
Amsterdam (1986).
Tadmor, Z., “Number of passage distribution functions with application to dispersive
mixing”. AIChE J. 34, 1943-1948 (1988).
Valsamis, L. N., Canedo, E. L., “Mixing, devolatilization, and reactive processing in the
Farrel Continuous Mixer”. Intern. Polymer Proc. 4, 247-254 (1989a).
Valsamis, L. N., Canedo, E. L., “Devolatilization and degassing in continuous mixers”.
SPE ANTEC Tech. Papers 35, 257-259 (1989b).
Valsamis, L. N., Canedo, E. L., “Effect of rotor geometry and operating conditions on mix-
ing performance in continuous mixers. An experimental study”. SPE ANTEC Tech.
Papers 37, 629-632 (1991).
Valsamis, L. N., Canedo, E. L.,, Donoian, G. S., “Operating characteristics of 24X rotors
for polyolefin conversion in the Farrel Continuous Mixer”. 7th. Intern. Conf. Polyole-
fins, Houston (1991).
Wagenknecht, U., Meissner, K., Reher, E. O., Polsterdorf, B., “Modellierung des nichtiso-
thermen Knetprozesses im Innenmischer”. Plaste und Kautschuk 33, 301-302 (1986).
Wagenknecht, U., Meissner, K., Bothmer, D., Reher, E. O., Polsterdorf, B., “Zur Model-
lierung der Strömungsvorgänge im Innenmischer unter Beachtung nicht-Newtonscher
Werkstoffeigenschaften”. Plaste und Kautschuk 34, 238-241 (1987).
Mixing in the Farrel Continuous Mixer 49

Wagenknecht, U., Meissner, K., Reher, E. O., Polsterdorf, B., “Zum Einfluss der Rotor-
form auf die Strömungsverhaltnisse im Knetspalt eines Innenmischers”. Plaste und
Kautschuk 35, 175-177 (1988).
Wong, T. W., Manas-Zloczower, I., “Numerical studies of the flow field in partially filled
mixing equipment”. SPE ANTEC Tech. Papers 38, 1788-1795 (1992).
Wong, T. W., Manas-Zloczower, I., “Two-dimensional dynamic study of distributive
mixing of a Banbury mixer”. SPE ANTEC Tech. Papers 39, 773-778 (1993).
Yagii, K., Kawanishi, K., “Flow analysis in an internal mixer, Part I: Application of finite
element analysis”. Intern. Polymer Proc. 5, 164-172 (1990).
Yang, H. H., Manas-Zloczower, I., “3D flowfield analysis of a Banbury mixer”. Intern.
Polymer Proc. 7, 195-203 (1992).
View publication stats

You might also like