Modeling of Transport

Phenomena and Melting

Kinetics of Starch in a
Co-Rotating Twin-Screw
Extruder
LIJUN WANG, DAVID D. JONES, CURTIS L. WELLER,
MILFORD A. HANNA
Industrial Agricultural Products Center, Department of Biological Systems Engineering,
University of Nebraska-Lincoln, Lincoln, NE 68583-0726
Received: February 18, 2005
Revised: October 17, 2005

ABSTRACT: A mathematical model was developed to simulate fluid flow,

heat transfer, and melting kinetics of starch in a co-rotating intermeshing
twin-screw extruder (TSE). The partial differential equations governing the
transport phenomena of the biomaterial in the extruder were solved by a finite
element scheme. For validating the model, the predicted product pressure, bulk
temperature at the entrance of the die, and minimum residence time of the
biomaterial in the extruder were compared with experimental data. Standard
errors of product pressure, bulk temperature at the die entrance, and minimum
residence time were about 8.8, 2.8, and 17.3%. Simulations were carried out to
investigate profiles of product pressure, bulk temperature, and melt fraction
within the extruder during extrusion.  C 2006 Wiley Periodicals, Inc. Adv Polym

Techn 25: 22–40, 2006; Published online in Wiley InterScience

(www.interscience.wiley.com). DOI 10.1002/adv.20055

A contribution of the University of Nebraska Agricultural

Research Division, Lincoln, NE, Journal Series No. 14515.
Correspondence to: Lijun Wang; e-mail: [email protected].
Contract grant sponsor: U.S. Department of Energy and the
Hatch Act.

Advances in Polymer Technology, Vol. 25, No. 1, 22–40 (2006)


C 2006 Wiley Periodicals, Inc.
 TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

KEY WORDS: Biomaterials, Computer modeling, Extrusion, Numerical

analysis, Starch, Transport

heat was assumed to be transferred from the hot

Introduction barrel wall to the material bed by convection in a
twin-screw extruder and a pseudo-convection heat

T win-screw extruders are used widely in the

manufacture of food and feed products, phar-
maceuticals, and plastics because of their advantages
transfer coefficient was used.1,3,7,8 Rauwendaal9 ob-
served that the dispersed solids melting (DSM) phe-
nomenon occurred in a twin-screw extruder. During
such as excellent feeding and mixing capacities, as DSM, the solid particles were dispersed uniformly in
compared with a single-screw extruder. Twin-screw a melt matrix and melted gradually. Zhu et al.10 de-
extruders consist of two screws, having a “figure- veloped a model to describe the DSM in co-rotating
eight” cross section, mounted in a barrel. There are twin-screw extrusion.
a large number of types of twin-screw extruders. However, no publications were found on the si-
The co-rotating intermeshing twin-screw extruder multaneous analysis of fluid flow, heat transfer, and
is one of the most frequently used designs.1 The melting kinetics in a co-rotating twin-screw extruder.
main difference between the screw channels of a In this paper, such a mathematical model is devel-
co-rotating intermeshing twin-screw extruder and oped. The model was validated by experiments. The
a single-screw extruder is the existence of an inter- model was further used to predict the profiles of
meshing region in the twin-screw channel. There- the product pressure, bulk temperature, and melt-
fore, the twin-screw channel has two regions: the ing fraction during extrusion.
C-shaped region and the intermeshing region.1−3
The existence of the intermeshing region in the
twin-screw channel results in a special fluid flow re-
gion in the channel. In the co-rotating intermeshing Mathematical Model
twin-screw extruder, the flow from one screw chan-
nel is divided into two streams: one part remains The model was divided into four parts: (1) geo-
in the same screw channel and goes through the metrical description to determine the parameters of
intermeshing region, and the other part goes into the channel for fluid flow and heat transfer, (2) fluid
the other screw channel. In the intermeshing region, flow to predict the flow patterns and pressure gra-
the flow from one screw channel interacts with the dients in the extruder, (3) melting kinetics and heat
flow from the other screw channel. The interaction transfer to determine the melting rate and the tem-
of two flow streams increases the mixing and heat perature profiles of product along the extruder, and
transfer capacities of the extruder.4 The conveying (4) mixing in the extruder.
section of a co-rotating twin screw is always par-
tially filled.1 Barres et al.5 reported degrees of fill-
GEOMETRICAL DESCRIPTION
ing, in the conveying section, in the range of 1–40%.
Modeling of transport phenomena and melting ki- A typical co-rotating intermeshing twin-screw ex-
netics in a twin-screw extruder has been carried out truder is shown in Fig. 1a. The central part of the ex-
in the past two decades. A literature review showed truder has two screws. The twin-screw channel has
that most of the models focused on just one of the two repeated regions: the C-shaped region and the
complex phenomena such as heat transfer, melting intermeshing region as shown in Fig. 1b. In order to
kinetics, or fluid flow, which occur in a twin-screw simplify the analysis, the real extruder was depicted
extruder. Sastrohartono et al.6 analyzed numerically schematically as a sequence of C-shaped chambers
the fluid flow and heat transfer in the melt pump- and intermeshing chambers as shown in Fig. 1c.
ing section of twin-screw extruders. They divided The following assumptions were made for de-
the complex geometry of the twin-screw extruder scribing the geometry of the channel:
into repeated “C” regions and intermeshing regions.
Tayeb et al.3 developed a model to predict the pres- (1) The channel in the C-shaped region was as-
sure generation in a C-shaped chamber and pres- sumed to have a rectangular cross-section, with
sure drop in an intermeshing chamber. Usually, the average height Hcn and width Bcn .2

ADVANCES IN POLYMER TECHNOLOGY 23

TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

FIGURE 1. Cross-sections and flow diagram of a typical intermeshing, co-rotating twin-screw extruder.

(2) The width of the channel in the intermeshing by an equivalent rectangle with average height
region reduced gradually to the minimum in Hin and width Bin .2
the middle point and then increased gradu-
ally to its original value as shown in Fig. 1b. The geometric dimensions of the selected ex-
The cross-section of the channel was described truder are given in Table I. The total flight number,

24 VOL. 25, NO. 1

 TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

TABLE I
Extruder Geometry and Dimensions Used in Experiments and Model
Initial End Initial Axial Width Axial Width Pitch of Axial
Barrel Barrel Screw End Screw of Flight of Flight at Screw Length of
Diameter Diameter Diameter Diameter at the Tip the Bottom Channel Screw
(mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm)

42.5 27.5 27.5 12.5 2.5 7.0 12.5 350.0

N, was determined by The helix length of the nth intermeshing chamber

was calculated by
L
N= (1)
l (αn /2) D̄n
Sin = (8)
cos φ̄n
The down channel was characterized by its height,
width, and length. For the selected screw, the height The cross-section of the channel had a trapezoidal
of the channel, which was the flight height, was con- shape. The average width of the nth C-shaped
stant through the helix screw channel. The height of chamber was calculated by
the “C” and intermeshing region was given by
 
σ + σ
Db1 − Ds1 Db2 − Ds2 Bcn = l − cos φ̄n (9)
H̄ = Hcn = Hin = = (2) 2
2 2

The barrel and screw diameters decreased gradually The minimum width of the nth intermeshing
along the axis of extruder to obtain a given compres- chamber was calculated by
sion ratio. The internal diameter of barrel and the
diameter of screw root at the nth flight were calcu- Bin,min = [l − (σ + σ  )] cos φ̄n (10)
lated linearly by
Thus, the average width of the nth intermeshing
n chamber was given by
Dbn = Db1 − (Db1 − Db2 ) (3)
N
n Bcn + Bin,min
Dsn = Ds1 − (Ds1 − Ds2 ) (4) Bin = (11)
N 2

The average helix angle of the screw channel at the

FLUID FLOW
nth flight, φ̄n , was calculated by
  There are two types of flow in the extruder: (1)
l plug flow of granular particles in the conveying sec-
φ̄n = tan−1 (5)
π D̄n tion and (2) non-Newtonian melt flow in the other
sections and die channel. The following assumptions
where D̄n = (Dbn + Dsn )/2. were made for modeling the fluid flow and mixing:
The overlap angle of the two screws in the inter-
meshing region at the nth flight, α, was calculated by (1) The flow through the intermeshing chambers
was assumed to be controlled by the minimum
 
Dbn − H̄ cross-sectional area, which was given by
αn = 2 cos−1 (6)
Dbn
Ain,min = Bin,min × Hin (12)
The helix length of the nth C-shaped chamber was
calculated by The flow from each C-shaped chamber was
divided into two streams. There were two flow
(π − αn /2) D̄n streams from two C-shaped chambers enter-
Scn = (7)
cos φ̄n ing an intermeshing chamber in the opposite

ADVANCES IN POLYMER TECHNOLOGY 25

TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

directions, the ratio of each flow entering the When the material was fed into the twin-screw ex-
intermeshing chamber was determined by truder, the material was distributed into two identi-
cal screw channels. Thus, the mass flow rate of the
0.5Ain,min downstream in the C-shaped region of each screw
xf = (13)
Acn channel was half of the mass feed rate

The theoretical flow division ratio is be- ṁcz = 0.5ṁ f (15)

tween 0 and 1. The flow division ratio could be
affected by the viscosity of materials and op- Thus, the bulk density of the solid particle bed in the
erating conditions besides extruder configura- conveying section was determined by
tion. Sastrohartono et al.6 found that the flow ṁcz
division ratio for a tangential twin-screw ex- ρs = (16)
V̇cz
truder was close to 0.5, which is close to the
area ratio in our case. As the selected screw had a tapered shape, the volu-
(2) The flow configuration was simplified by treat- metric flow rate V̇z , which was calculated by Eq. (14)
ing the screw as stationary and rotating the decreased along the downstream channel, resulting
barrel in the opposite direction to that of the in an increase in bulk density of the solid particles.
screw.6,11,12 The barrel thus moved at a constant The screw finally became completely full.
relative velocity, ub , and at the helix angle, φb , The velocity of the plug flow in the intermesh-
to the down channel direction. ing region was assumed to be the velocity at the en-
(3) It was assumed that there was no pressure de- trance of the intermeshing chambers since no friction
velopment in the conveying section since twin- between solid particles and screw surface was con-
screw extruders operate predominantly under sidered. The mass flow rate of each downstream in
starved feed conditions.11,12 the intermeshing chambers was expressed as
(4) In order to simplify the transport phenomena
ṁi z = x f ṁcz (17)
of the non-Newtonian melt flow in the screw
channel, the flow was assumed to be laminar,
During extrusion, the temperature of the solid par-
the gravitational forces and the inertia terms
ticles increases along the down channel due to the
in the Navier–Stokes equation were neglected
heat transfer from the hot barrel. When melting oc-
due to the high viscosity of the flow, the flow
curs, pressure is built up to push the melt along the
was fully developed, there was no slip at the
rest of the downstream channel, forcing the melt flow
walls, and the diffusion of momentum in the
through a die. In a sequence of C-shaped chambers
down channel z-direction was negligible.12
as shown in Fig. 1c, the melt flow was described as
a combination of drag-pressure flow. Thus, the gov-
In the conveying section of the extruder, the gran- erning equations of momentum were written as
ular particles of starch were trapped between flights
of the screw and conveyed in a manner similar to the  
∂P ∂ ∂ux
action of a screw conveyor. Friction force between = η in the x-direction (18)
∂x ∂y ∂y
the solid particle bed and barrel surface controlled
the solid plug flow in the screw channel. The veloc- ∂P
= 0 in the y-direction (19)
ity of the plug flow in the C-shaped chambers was ∂y
   
determined by ∂P ∂ ∂uz ∂ ∂uz
= η + η
∂z ∂x ∂x ∂y ∂y
ucz = π Dbn Ns cos φbn (13)
in the z-direction (20)
where the helix angle of the screw channel at the nth The boundary conditions of the momentum equa-
flight tip is φbn = tan−1 ( π Dl bn ). tions were given by
The maximum conveying capacity of the C-
shaped chamber in each screw channel was calcu- ux | y = 0 = 0 and ux | y = Hcn = π Dbn Ns sin φbn (21)
lated by
uz |x = 0 = 0 and uz |x = Bcn = 0 (22)
V̇cz = Bcn Hcn ucz (14) uz | y = 0 = 0 and uz | y = Hcn = π Dbn Ns cos φbn (23)

26 VOL. 25, NO. 1

 TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

Since there was no drag effect of the barrel on the flow was expressed as
in the intermeshing chambers, the melt was pushed
through the intermeshing chambers in a manner of 
N 
N

positive displacement. The positive displacement re- Ps = Pcn − Pin (31)
n=1 n=1
sulted in a decrease in pressure. The positive dis-
placement through the intermeshing chambers was
where Pcn was the pressure generation through the
also governed by Eqs. (18)–(20). However, as there
nth C-shaped chamber, Pin was the pressure drop
was no effect of barrel drag force in the intermeshing
through the nth intermeshing chamber.
chambers, the boundary conditions became
The extrusion die serves as a major tool for the
production of products. Flow in the die was assumed
ux | y = 0 = 0 and ux | y = Hin = 0 (24) to be isothermal non-Newtonian fluid. Because the
uz |x = 0 = 0 and uz |x = Bin = 0 (25) cylindrical hole of the die was symmetrical with the
axis, only half of the axial cross-section was chosen
uz | y = 0 = 0 and uz | y = Hin = 0 (26) for analysis. The governing equation of momentum
was written as
Equations (18)–(26) were used to predict the flow  
∂P 1 ∂ ∂uz
patterns and pressure gradients in the C-shaped = ηr (32)
chambers and the intermeshing chambers. Both the ∂z r ∂r ∂r
pressure gradients and velocities in Eqs. (18) and (26)
The boundary conditions of the above equation were
were unknown. In order to evaluate the pressure
given by
gradients and flow patterns, continuity equations
were used to solve the momentum and the conti- 
∂uz 
nuity equations iteratively. = 0 and uz |r = R = 0 (33)
∂r r =0
The integral form of the continuity equations in
the C-shaped chambers was given by In order to solve the momentum equation, the conti-
 nuity equation was used to iteratively determine the
H
pressure gradient in Eq. (32). The continuity equa-
ρm ux dy = ṁcx in the x -direction (27)
0 tion was written as
 Bcn  H  R
ρm uz dx dy = ṁcz in the z-direction (28) (ρm × 2r uz )dr = ṁ f (34)
0 0 0

where ṁcx was the leakage flow rate across flights The melt was assumed to follow the power law, and
which was assumed to be zero for small fight clear- the viscosity of the melt was calculated by13
ance and ṁcz was the downstream channel flow rate  
2451
in the C-shaped chambers given by Eq. (15). η = 106.25γ −0.63 exp exp(−4.63Xw ) (35)
The integral form of the continuity equations in TK
the intermeshing chambers were given by
where the shear rate was calculated by
 H
 2  2  2  12
ρm ux dy = ṁi x in the x-direction (29) ∂ux ∂uz ∂uz
0 γ= + +
 Bin  H ∂y ∂x ∂y
ρm uz dx dy = ṁi z in the z-direction (30)
0 0 (in screw channel) (36)
∂uz
where ṁi x was the leakage flow rate across flights γ= (in die channel) (37)
∂r
which was assumed to be zero for small flight clear-
ance and ṁi z was the downstream channel flow rate The working point of the extruder for given pro-
in the intermeshing chambers given by Eq. (17). cessing conditions was determined by the pressure
The net pressure generation in the screw was the balance over the whole extruder. The net pressure
sum of pressure generation in each “C” chamber and generation in the screw, Ps , should equal the pres-
pressure drop in each intermeshing chamber, which sure drop through the die, Pd .

ADVANCES IN POLYMER TECHNOLOGY 27

TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

HEAT TRANSFER AND MELTING KINETICS (1) The solid particles which were fed into the ex-
truder were assumed to be identical. Therefore,
During extrusion, starch particles enter the ex-
one particle was chosen for the analysis of melt-
truder, transform to a melt at some point within
ing kinetics.
the extruder due to heating, and move toward the
exit of the extruder. As a co-rotating intermeshing (2) Initially, the particle was heated from initial
twin-screw extruder has high mixing capacity, it is feed temperature to the melting temperature
normally assumed that product temperature is uni- at 115◦ C 8 by the hot barrel. The energy due to
form in the direction normal to the screw shafts and internal friction between starch particles was
heat transfers from the hot barrel to the material by neglected. The melting of the particle occurred
convection.7,8,14 when its surface temperature reached the melt-
It has been reported that co-rotating twin-screw ing point.
extrusion follows a particle dispersed melting (3) When the percent conversion of starch par-
mechanism.9,10,15 In the particle dispersed melting ticles to melted starch reached about 37%,16
model, the solid particles are assumed to be uniform, the particle plug flow became non-Newtonian
spherical, and dispersed in a melt matrix as shown melt flow. The heat transfer from the hot bar-
in Fig. 2a. Rauwendaal9 reported that a minimum rel and the heat of viscous dissipation in the
volume of polymer melt has to be available to fill melt matrix continued to melt the particle and
the space between the solid particles before the solid increased the melt temperature.
particle flow becomes a non-Newtonian melt flow. (4) The convective heat from the hot barrel and
Wang et al.16 found that the powdery raw starch be- the heat of viscous dissipation in the melt were
came interconnected and behaved like a fluid when uniformly distributed to each particle.
the percent conversion of starch reached about 37%.
Therefore, the following assumptions were made During melting, there existed a moving interface
for modeling the melting kinetics and heat transfer: within the particle, which moved from the surface to
the center of the particle until the whole particle was
melted. The moving interface divided the whole par-
ticle into two regions: the core solid region and the
outer melt layer as shown in Fig. 2b. The heat transfer
through the solid core region was governed by
 
dTs 1 ∂ 2 ∂ Ts
ρs c s = 2 r ks (38)
dt r ∂r ∂r
The heat transfer through the outer melt layer was
governed by
 
dTm 1 ∂ ∂ Tm
ρm c m = 2 r 2 km + q̇ m (39)
dt r ∂r ∂r
where the viscous heat generation rate in the melt,
q̇ m , was given by

q̇ m = ηγ 2 (40)

Boundary conditions were


∂ Ts 
=0 (41)
∂r r =0
Ts |r =Rs = Tm |r =Rs = Tm0 (42)

∂ T 
k = q sur (43)
∂r 
r =R p
FIGURE 2. Particle dispersed melting mechanism in a
twin-screw extruder. where q s is the heat flux on the surface of the particle.

28 VOL. 25, NO. 1

 TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

The location of the moving interface, rs , was de- thus given by

termined by the energy balance on the interface. On
the interface, the net heat conduction was equal to Qb
q sur = (49)
the latent heat of melting, which was expressed as Np Ap,sur
   
∂ Ts ∂ Tm drs The temperature of the bulk flow was the mass av-
ks − km = λρs (44)
∂r ∂r dt erage temperature of the particle, which was given
by
It should be noted that before melting occurred, heat
transfer through the particle was governed only by
E 2  
T(r )dm π Ti2ri23 − Ti1ri13
Eq. (38) with boundary conditions given by Eqs. (41) Vp i=1 3
and (43). The melt fraction was expressed as the ratio Taver = = (50)
dm
E 4  
of melt volume to the total volume of the particle, Vp π ri23 − ri1
3
i=1 3
which was given by

vm MIXING
θm = (45)
vp
As mentioned above, the flow from a C-shaped
chamber was divided into two streams at the en-
When the flow advanced for a z step in the down-
trance of an intermeshing chamber and two streams
stream direction, the total particles enclosed by the
were merged at the exit of the intermeshing chamber.
small volume of z × B × H were determined by
For each repeated turn, all particles had 100% possi-
bility flowing through a C-shaped chamber and x f ×
ṁz t
Np = (46) 100% possibility of flowing through an intermeshing
ρpvp chamber. Therefore, at the exit of each intermeshing
and C-shaped chamber, the velocity, temperature,
where t is the residence time for the z distance. and melt fraction were adjusted by
When particles entered the intermeshing cham-
bers, there was no convection heat transfer between u = ucn + x f uin (51)
the particles and the hot barrel wall. However, in
the C-shaped chambers, convection heat transfer oc- T = Tcn + x f Tin (52)
curred. When particles advanced for a zc step in the θ = θcn + x f θin (53)
downstream direction of the C-shaped chambers, the
total heat transfer between the particles and the hot
barrel was calculated by SIMULATION APPROACH
A computer program was written to solve the
Qb = h Bcn zc (Tb,sur − Tp,sur ) (47) model. The program was divided into four parts to
obtain solutions for geometrical parameters, fluid
For the particle bed, a pseudo-heat transfer coeffi- flow, heat transfer, and melting kinetics. The sim-
cient, h, was expressed as the dependence on the ulation began with the input of geometries of the
density of the particle bed extruder, processing conditions, and relevant con-
stants. The total flight turns were determined by
h = a ρb (48) Eq. (1). The height, helix length and width of each
C-shaped chamber and intermeshing chamber were
In order to determine constants a and b in the above determined by Eqs. (2)–(11). The leakage flow ratio
equation, two sets of h and ρ were required. It was of a stream entering the intermeshing chamber was
assumed that for the channel filled with air only, determined by Eq. (13).
h = 30 W/(m2 K) and ρ = 1 kg/m3 and for the chan- The helix length, height, and width of the “C”
nel fully loaded with particles, h = 1500 W/(m2 K) chamber and intermeshing chamber were then di-
and ρ = 416 kg/m3 . In this case, a = 30 and b = 0.65. vided into ten sections along the downstream chan-
The total heat, Qb , was uniformly distributed to nel, respectively. The radius of a particle was di-
each particle in the small analyzed volume, and the vided into ten segments. Simulations were advanced
heat flux on the particle surface, q sur in Eq. (43) was from the first cross-section at the entrance of the

ADVANCES IN POLYMER TECHNOLOGY 29

TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

C-shaped chamber to the last cross-section at the exit, section was divided into 100 segments each. The ve-
and the entrance of the intermeshing chamber to the locity and pressure of each node formed the veloc-
last cross-section at the exit. In the program, a finite ity and pressure fields in the die channel. The pres-
element scheme was used to solve partial differential sure gradient and velocity profile in the die channel
Eqs. (18), (20), (32), (38), and (39). were predicted by Eqs. (32)–(34). The simulation was
For the solid particle flow in the conveying sec- stopped when the last cross-section at the exit of the
tion, the velocity and volumetric flow rate of the die channel was reached.
downstream were calculated by Eqs. (13) and (14).
The bulk density of the solid particle bed was then
determined by Eq. (16). The temperature distribu-
tion in the particle was predicted by Eqs. (38), (41),
Materials and Methods
and (43). The temperature of the bulk flow was then
determined by Eq. (50). The surface temperature Normal corn starch (∼25% amylose and ∼75%
of the particle was then compared with the melt- amylopectin) purchased from National Starch and
ing temperature. If the surface temperature of the Chemical Co. (Bridgeport, NJ) was used in the ex-
particle was smaller than the melting temperature, periments. The moisture content of the starch was
there was no melting and the analysis of plug flow adjusted to the different levels by blending with dis-
continued on to the next cross-section. Otherwise, tilled water. The samples were sealed in plastic buck-
melting occurred and the melting rate was deter- ets and stored at 4◦ C for one day. Before extrusion,
mined by Eq. (44). If the volumetric melt fraction the samples were removed from cold storage and al-
was less than 37% of the total particle volume, the lowed to come to room temperature. The moisture
analysis of plug flow continued on the next cross- contents of samples were measured by a moisture an-
section. The temperature distribution in the particle alyzer at the chamber temperature of 105◦ C (HG 53
was predicted by Eqs. (38)–(43). Otherwise, the ma- moisture analyzer, Mettler-Toledo GmbH, Labora-
terial became a melt flow. The velocity and pressure tory & Weighing Technologies, Greifensee, Switzer-
profiles in the melt were obtained by Eqs. (18)–(30). land). The properties of granular and melted starch
The melting rate was determined by Eq. (44). The are given in Table II.
temperature distribution in a particle was predicted The starch was extruded in a twin-screw labo-
by Eqs. (38)–(43). The analysis proceeded until the ratory extruder (model CTSE-V, C. W. Brabender,
last cross-section, at the exit of the “C” chamber or Inc., South Hackensack, NJ). Two conical screws co-
the intermeshing chamber, was reached. At the exit rotated inside the barrel of the extruder. The conical
of each C-shaped and intermeshing chamber, the ve- screws had diameters decreasing from 42.5 to 27.5
locity, temperature, and melt fraction were adjusted mm along an axial length of 350 mm from the feed
by Eqs. (51)–(53). The analysis advanced to the next end to the exit end. The detailed dimensions of the
C-shaped chamber and intermeshing chamber until extruder are given in Table I. Three die nozzles with
the exit of the screw channel was reached. different diameters were used. The dimensions of
Finally, the non-Newtonian fluid entered into the three die nozzles and operating conditions are given
die channel. As indicated in Fig. 3, there were four in Table III.
sections in the die. The radius of each section was Starch was fed into the extruder using a
divided into 50 segments, and the length of each FlexWall® Plus Feeder (Brabender Technologie, Inc.,

FIGURE 3. Geometry of die and finite element arrangement in the die channel.

30 VOL. 25, NO. 1

 TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

TABLE II
Properties of Granular and Melted Starch Used in Model
Granular Starch Starch Melt

Specific heat (J/(kg K)) c = 1291 + 2321Xw [18] c = 1291 + 2321Xw [18]
ṁz
Density (kg/m3 ) ρs = (P = 0 Pa) ρm = 1200 [13]
V̇z
1 (1 − εw ) εw
Thermal conductivity ks = (1 − εv )k p + εg kg [19], where = + [20], where
km kds kw
(W/(m K))
ρs ρds Xw
εg = 1 − εw =
ρp ρw + ρds Xw
k p = (1 − εw )kds + εw kw
kds = 0.0976 + 0.167 × 10−2 T kds = 0.21 + 0.41 × 10−3 (T + 273.15)
kw = 0.571 + 1.76 × 10−3 T − 6.70 × 10−6 T 2 kw = 0.571 + 1.76 × 10−3 T − 6.70 × 10−6 T 2
kg = 0.00324 + 5.31 × 10−4 T
Melting latent heat (J/g) – λ = 13.8 [21]

TABLE III
Experimental and Predicted Process Variables and System Variables
Die Nozzle Process Variable Experimental Data Predicted Data

Dn Ln N ṁ f Mw Tb1 Tb2 Tb3 Tdw Pd Td tmin Pd a Pd b Td tmin

(mm) (mm) (rpm) (kg/h) (%) (◦ C) (◦ C) (◦ C) (◦ C) (MPa) (◦ C) (s) (MPa) (MPa) (◦ C) (s)

3.0 14.6 140 7.73 16 67 134 139 113 11.13 142 22 10.49 10.41 148.8 30.6
3.0 14.6 140 8.86 24 72 141 140 125 6.78 139 24 6.65 6.24 141.8 26.6
3.0 14.6 160 7.34 20 64 136 139 124 8.56 140 20 7.90 7.11 144.9 25.5
4.0 14.6 140 8.86 24 67 130 139 127 4.48 139 22 3.81 4.03 133.5 19.3
4.0 14.6 120 6.11 20 65 125 130 107 6.27 139 30 6.48 5.84 136.0 29.8
5.0 14.7 140 7.73 16 69 129 132 104 7.78 137 19 7.21 6.32 137.5 23.1
5.0 14.7 140 8.86 20 65 134 137 112 5.36 142 20 5.66 4.82 138.0 21.9
a
Die pressure based on pressure rise through the screw channel.
b
Die pressure based on pressure drop through the die channel.

Ont., Canada), maintaining as constant a feed rate tal helix length of the C-shaped chambers on a screw.
as possible. The extruder was controlled by a plasti- The maximum velocity of downstream flow was ob-
Corder (type FE 2000, C. W. Brabender, Inc., NJ). The tained by the velocity distribution profile. All extru-
barrel temperatures and the product temperature at sion runs were conducted in triplicate.
the die entrance were recorded using temperature
probes (Omega Engineering Inc., CT). The operat-
ing pressure at the die entrance was recorded using
a pressure transducer (model PT 411-10M-6, Dynisco Results and Discussion
Inc., MA). The exact rotational speeds, barrel temper-
atures, and die pressure were recorded automatically
MODEL VALIDATION
during extrusion.
A stimulus response technique was used to mea- In order to validate the model, predictions of the
sure the minimum residence time. The stimulus was model were compared with the measured values
a color dye tracer. During extrusion, the tracer was for seven arbitrarily selected experimental condi-
added to the feeding zone and the minimum time for tions. The operating conditions and predicted and
the tracer to go through the extruder was recorded. measured results are given in Table III. Standard
In simulation, the minimum residence time was the errors between the predicted and experimental prod-
shortest distance divided by the maximum velocity uct pressure and bulk temperature at the die en-
of downstream flow. The shortest distance is the to- trance, and minimum residence time were about 8.8,

ADVANCES IN POLYMER TECHNOLOGY 31

TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

2.8, and 17.3%. On one hand, the accuracy of the heat of viscous dissipation and higher thermal con-
measurements was a likely reason for the variations ductivity of the melt. However, at the end of extru-
between the predicted and experimental data. On sion, the heating rate of starch slowed down due to
the other hand, for improving the accuracy of the the decrease in temperature difference between the
model, more research is needed particularly on the barrel surface and the melt, and a decrease in the heat
densities of the solid bed and melt, the rheological of viscous dissipation at high bulk temperature.
properties of the melt, the latent heat of melting, and In the conveying section of the extruder where rel-
the moisture loss of the starch during extrusion. atively free-flowing granular particles exist, internal
shear of the material was negligible and the pres-
sure was zero as shown in Fig. 4. The loose particle
PROFILES OF BULK TEMPERATURE
starch was initially conveyed toward the direction of
AND PRESSURE
the die by the rotating screw. Along the extruder, the
Typical profiles of bulk temperature and pres- particles were heated by the barrel. When the surface
sure along the extruder during extrusion are given temperature of the particles reached their melting
in Fig. 4. The curve of the bulk temperature profile temperature, melting occurred. The particles were
shows there are three different increasing rates of compressed along the extruder due to the decrease of
bulk temperature along the extruder (Fig. 4). When cross-sectional area in the tapered screw channel and
entering into the feed zone at an initial temperature the cohesive effect of melted particle surface. When
of 25◦ C, the material was heated to only 50◦ C at the the melt to particle fraction approached 37% by vol-
end of the first heating zone due to the poor thermal ume, the starch material became a non-Newtonian
conductivity of the starch particle bed and the low melt fluid and pressure developed in the starch melt
barrel temperature. The material then entered into as shown in Fig. 4. It can be seen from Fig. 4 that the
the second heating zone with a higher barrel tem- pressure rise in the screw channel equaled the pres-
perature. The heating rate of starch increased in the sure drop in the die channel. The positive pressure
second heating zone due to a higher barrel temper- difference between the entrance and the exit of the
ature. When melting occurred and the particle flow die forced the melt through the die channel by over-
became a melt flow in the second heating zone, there coming the flow resistance in the die nozzle. It should
was a sharp increase in bulk temperature due to the be noted that only when the extruder operates at its

FIGURE 4. Predicted pressure and temperature profiles within the extruder.

32 VOL. 25, NO. 1

 TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

FIGURE 5. Pressure generation in a C-shaped chamber and pressure drop in an intermeshing chamber.

operating point, a good match between pressure rise the die of the extruder. The velocities near the screw
in the screw channel and the pressure drop in the die surface were negative in the direction of pressure
channel can be achieved. Each extruder has its oper- flow toward the feeder of the extruder. Both drag
ating characteristics.17 Before operating an extruder, and pressure flows maintained the prescribed down-
it is essential to know its characteristic curves for set- stream mass flow rate.
ting the operating conditions. For a given extruder, The drag flow decreased in the C-shaped cham-
the characteristic curves can be generated by simu- bers along the extruder due to the decrease in cross-
lation or experiments.17 sectional area and the decrease in apparent viscosity
Figure 5 shows typical pressures in a C-shaped of the dough with increase in melt temperature. With
chamber and the pressure drop in an intermeshing a decrease in drag flow, the pressure flow simulta-
chamber. The predicted pressure in one C-shaped neously decreased to maintain a given downstream
chamber was 1–2 MPa, and the pressure drop in an mass flow rate in the C-shaped chambers. Therefore,
intermeshing chamber was 0.25–0.35 MPa. The data the pressure generated in the C-shaped chambers
are close to those for similar conditions reported by along the extruder gradually decreased as shown in
Tayeb et al.3 In the intermeshing chamber, there was Fig. 5. In the intermeshing chambers, the decrease
no drag effect on the flow from the barrel wall. The of cross-sectional area increased the pressure drop
material was pushed through the gap between the for a given downstream mass flow rate through the
two screws by the pressure difference at the entrance chambers. However, the decrease in apparent viscos-
and exit of the chamber. As a result, the pressure de- ity of the dough decreased the pressure drop. Both
creased gradually from the entrance to the exit of adverse effects controlled the pressure drop in the
the intermeshing chamber. In the C-shaped cham- intermeshing chambers.
ber, drag flow, and pressure flow existed in the non-
Newtonian melt. Drag flow was the flow toward the
discharge end of the screw resulting from viscous
MELT KINETICS
drag. Pressure flow was dependent on the pressure The profiles of melt fraction for particles with dif-
gradient in the screw channel. As shown in Fig. 6, ferent sizes are given in Fig. 7. Profiles of starch
the velocities near the barrel wall were positive as melting along the extruder are important for the
drag flow along the barrel wall was directed toward quality of final products as inefficient melting will

ADVANCES IN POLYMER TECHNOLOGY 33

TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

FIGURE 6. A typical velocity distribution of a combined drag and pressure flow along the flight height in down channel
direction of a C-shaped chamber.

FIGURE 7. Predicted melt fraction along the axis of extruder.

34 VOL. 25, NO. 1

 TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

deteriorate the quality of the final products. It can be in the radial direction of the large particle during
seen from Fig. 7 that melting of a small particle (e.g., extrusion. For the particle with a radius of 2.5 mm,
particle radius 0.5 mm or less) was almost instanta- when the surface temperature increased to the melt-
neous. However, it took a long axial distance in the ing point of 115◦ C, the central temperature of the
extruder to melt a large particle (e.g., particle radius particle was only about 30◦ C. The large temperature
2.5 mm). gradient in the particle existed until the particle was
On the melt fraction curve of the large particle in entirely melted. After the entire particle was melted,
Fig. 7, the melting rate at the beginning was much the temperature gradient in the melt became very
smaller than that at the final stage. Melting is an en- small because the melt was heated mainly by the
dothermic process. At the beginning of melting, the inner heat of viscous dissipation in the melt rather
material was still in a particle form and heat transfer than heat transfer from the hot barrel. It also can be
from the hot barrel contributed to the latent heat of seen from Fig. 8 that there was a melting interface
melting. When the melt to particle fraction reached a in the particle when the surface temperature of the
critical value (37% for starch given by Wang et al.16 ), particle reached the melting point. The whole par-
the space between particles became filled by melt ticle was divided into two regions by the interface:
and the particle flow becomes a non-Newtonian melt the shell melt region and the core solid region. The
flow. The heat transfer from the hot barrel and the melting interface moved from the surface to the cen-
heat of viscous dissipation in the melt matrix contin- ter of the particle and finally disappeared when the
ued to melt the remaining part of particles, increas- entire particle was melted. However, there was no
ing the melting rate. obvious temperature gradient in a small particle as
It also can be seen from Fig. 7 under the same shown in Fig. 9. The difference between the surface
operating conditions, the melting of a large particle temperature and the central temperature in a small
started earlier than that of a small particle. To fur- particle with a radius of 0.5 mm was less than 2◦ C
ther explain this phenomenon, predicted tempera- during the whole extrusion.
ture distributions in a large particle and in a small The temperature distributions in Figs. 8 and 9
particle are given in Figs. 8 and 9. It can be seen from show that heat transfer during heating and melting
Fig. 8 that there was a large temperature gradient of a small solid particle is mainly controlled by the

FIGURE 8. Temperature distribution in a large particle (radius 2.5 mm) during extrusion.

ADVANCES IN POLYMER TECHNOLOGY 35

TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

FIGURE 9. Temperature distribution in a small particle (radius 0.5 mm) during extrusion.

convection between the barrel surface and the mate- came melt flow earlier than the large particle flow as
rial bed. However, heat transfer during heating and shown in Fig. 7. The melt flow is accompanied by the
melting of a large particle is controlled both by con- heat of viscous dissipation. Therefore, at the end of
vection and by conduction within the particle. This extrusion, the surface temperature of the small par-
explains why the melting of a small particle was in- ticle was higher than that of large particle due to the
stantaneous and the melting time of a large particle heat of viscous dissipation as shown in Fig. 10.
was very long as shown in Fig. 7.
Under the same operating conditions, the surface
EFFECTS OF PARTICLE SIZE ON THE
temperature of a larger solid particle increased faster
PROFILES OF BULK TEMPERATURE AND
than that of a small particle as shown in Fig. 10. In
PRESSURE
the conveying zone, the temperature of powders in-
creases due to the heat transfer between powders Particles with different sizes have different melt-
and the hot inner surface of the extruder barrel. The ing kinetics, causing different product bulk tempera-
heat first transfers from the hot barrel surface to the ture and pressure profiles in the extruder. As shown
powder surface by convection and then from the in Fig. 11, the bulk temperature of small particles was
powder surface to its inside by conduction. Com- higher than that of large particles at the same loca-
pared to a small powder, a large powder receives tion along the extruder under the same operating
more heat from the hot barrel surface due to its larger conditions. Before the particle flow became a melt
surface area but transfers heat slowly into the inside flow, the particle bed was heated by convection. The
due to its larger dimension. So, the surface tempera- convective heat transfer was a function of convec-
ture for the large powder is higher than that of small tion heat transfer coefficient, barrel surface area, and
powders. As a result, the surface temperature of the the temperature difference between the barrel wall
large particle reached the melt point earlier than that and the particle surface. Under the same operating
of the small particle as shown in Fig. 7. However, the conditions, the convection heat transfer coefficient,
melting rate of the large particle was smaller than barrel surface area, and barrel wall temperature were
that of the small particle. The small particle flow be- the same. However, as shown in Fig. 10, the surface

36 VOL. 25, NO. 1

 TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

FIGURE 10. Predicted surface temperature profiles of particles with different sizes during extrusion.

FIGURE 11. Predicted bulk temperature profiles in the extruder for particles with different sizes.

ADVANCES IN POLYMER TECHNOLOGY 37

TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

FIGURE 12. Predicted pressure profiles in the extruder for particles with different sizes.

temperature of small particles was lower than that of ticle radius 2.5 mm). Different melting characteris-
large particles during heating. Therefore, the small tics, product pressure, and bulk temperature profiles
particles received more heat from the barrel wall by in the extruder were observed for different particle
convection than the large particles due to larger tem- sizes using the model. Adjustment of operating con-
perature differences, resulting in higher bulk tem- ditions is needed for the extrusion of starch-based
perature along the extruder. Figure 12 shows that the materials with different particle sizes for uniform
product pressure of small particles was higher than product quality.
that of large particles along the extruder. Pressure
was developed in the melt through the remainder of
the extruder after the particle flow became the melt
flow. The melting of small particles was completed
much faster than the melting of the large particles.
Acknowledgments

This study was conducted at the Industrial Agri-

cultural Products Center, University of Nebraska,
Conclusions Lincoln, NE. Authors would like to recognize
Dr. Girish M. Ganjyal and Mr. Robert Weber for their
assistance in the experiments.
A mathematical model was developed to describe
the co-rotating twin-screw extrusion process of bio-
materials. The standard errors between the predicted
and experimental product pressure and bulk temper-
ature at the die entrance, and minimum residence Nomenclature
time were 8.8, 2.8, and 17.3%. The melting of small
particles (e.g., particle radius 0.5 mm or less) was A Area (m2 )
almost instantaneous, while it took a long axial dis- B Channel width (m)
tance in the extruder to melt large particles (e.g., par- c Specific heat capacity (J/(kg K))

38 VOL. 25, NO. 1

 TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

D Diameter (m) iz The downstream direction in the

D̄ Average diameter of internal diameter of intermeshing chamber
barrel and diameters of screw boot (m) m Melt
E Total element number min Minimum
H Channel height (m) n The nth flight
h Convection heat transfer coefficient p Particle
(W/(m2 K)) r r -Coordinate in the radial direction
i The ith element s Screw or solid
k Thermal conductivity (W/(m K)) sur Surface
L Axial length of the screw (m) w Water
l Pitch of screw channel (m) x x-coordinate in the cross-width direction
m Mass (g) y y-coordinate in the height direction
ṁ Mass flow rate (g/s) z z-coordinate in the down stream direction
N Total flight number
NP Total particle number
Ns Screw speed (rpm) GREEK LETTERS
P Pressure (Pa) φ̄ Flight angle (◦ )
P Pressure difference (Pa) α Overlap angle of two screws (◦ )
q̇ m Viscous heat generation rate (W/m3 ) σ Axial width of flights at the tip (m)
Q Heat transfer rate (W) σ Axial width of flights at the bottom (m)
q sur Heat flux on the surface (W/m2 ) ρ Density (kg/m3 )
R Radius (m) η Apparent viscosity (Pa s)
S Channel helix length (m) γ Shear rate (1/s)
T Temperature (◦ C) λ Melting latent heat (J/g)
TK Temperature on the Kelvin scale (K) θ Melting fraction
t Time (s) ε Volumetric fraction
t Time step (s)
u Velocity (m/s)
v Volume (m3 )
V̇ Volumetric flow rate (m3 /s) References
x The mass ratio of two flow streams
Xw Moisture (%, db)
1. Yacu, W. A. J Food Eng 1985, 8, 1–21.
z Spatial step (mm)
2. Tayeb, J.; Vergnes, B.; Della Valle, G. J Food Sci 1988a, 53,
616–625.
SUBSCRIPTS 3. Tayeb, J.; Vergnes, B.; Della Valle, G. J Food Sci 1988b, 53,
1047–1056.
0 Initial 4. Harper, J. M. In Biotechnology and Food Process Engineering;
1 Entrance or node 1 Schwartzberg, H. G.; Rao, M. A. (Eds.); Marcel Dekker: New
2 Exit or node 2 York, 1990; pp. 295–308.
a Constant in Eq. (48) 5. Barres, C.; Vergnes, B.; Tayeb, J.; Della Valle, G. Cereal Chem
aver Average 1990, 67, 427–433.
b Barrel or constant in Eq. (48) 6. Sastrohartona, T.; Jaluria, Y.; Karwe, M. V. Polym Eng Sci 1995,
35, 1213–1221.
cn The nth C-shaped chamber
7. Mohamed, I. O.; Ofoli, R. Y. J Food Eng 1990, 12, 145–164.
cx The width direction in the
8. Barres, C.; Vergnes, B.; Tayeb, J. J Food Eng 1991, 15, 167–185.
C-shaped chamber
9. Rauwendaal, C. Adv Polym Technol 1996, 15, 135–144.
cz The downstream direction in the
10. Zhu, L.; Narh, K. A.; Geng, X. J Polym Sci, Part B: Polym Phys
C-shaped chamber
2001, 39, 2461–2468.
ds Dried solid
11. de Graaf, R. A.; Woldringh, D. J.; Janssen, L. P. B. M. Adv
f Feed Polym Technol 1999, 18, 295–302.
g Gas 12. Wang, L. J.; Ganjyal, G. M.; Jones, D. D.; Weller, C. L.; Hanna,
in The nth intermeshing chamber M. A. J Food Sci 2004, 69, 212–223.
ix The width direction in the 13. Padmanabhan, M.; Bhattacharya, M. J Food Eng 1993, 18,
intermeshing chamber 335–349.

ADVANCES IN POLYMER TECHNOLOGY 39

TRANSPORT AND MELTING KINETICS OF STARCH IN TSE

14. van Zuilichem, D.; van der Laan, E.; Kuiper, E. J Food Eng 18. Rahman, S. Food Properties Handbook; CRC Press: Boca
1990, 11, 187–207. Raton, FL, 1995.
15. Gogos, C. C.; Tadmor, Z.; Kim, M. H. Adv Polym Technol 19. Maroulis, Z. B.; Shah, K. K.; Saravacos, G. D. J Food Sci 1991,
1998, 17, 285–305. 56, 773–776.
16. Wang, S. S.; Qu, D.; Chiew, Y. Starch 1994, 46, 337–341. 20. Maroulis, S. N.; Saravacos, G. D. J Food Sci 1990, 55,
17. Wang, L. J.; Ganjyal, G. M.; Jones, D. D.; Weller, C. L.; Hanna, 1367–1372.
M. A. J Food Sci 2004, 69, 212–223. 21. Lai, L. S.; Kokini, J. L. Biotechnol Progress 1991, 7, 251–266.

40 VOL. 25, NO. 1