Journal of Food Engineering 198 (2017) 7e16

Contents lists available at ScienceDirect

Journal of Food Engineering

journal homepage: www.elsevier.com/locate/jfoodeng

Mathematical modeling of ow behavior and cell structure formation

during extrusion of starchy melts
Pavan Harshit Manepalli a, Hulya Dogan a, John M. Mathew b, Sajid Alavi a, *
a
Department of Grain Science and Industry, Kansas State University, Manhattan, KS 66506, USA
b
Frito-Lay Research & Development, Plano, TX 75024, USA

a r t i c l e i n f o a b s t r a c t

Article history: The primary purpose of this research was to develop a mathematical model for ow behavior of starchy
Received 5 April 2016 melts inside an extruder barrel and bubble growth dynamics after exiting the extruder using mass, heat
Received in revised form and momentum transfer equations and obtain the physical characteristics of the product, such as
16 September 2016
expansion ratio and bubble radius, using input parameters such as feed rate, screw speed, water input in
Accepted 26 October 2016
Available online 3 November 2016
the extruder and pre-conditioner, etc. The model was written in Visual Basic and experimentally vali-
dated using pilot-scale twin screw extrusion for processing of cereal-based cellular products. Process and
product data were measured at different in-barrel moisture contents (19e28% dry basis) and experi-
Keywords:
Extrusion
mental screw speeds (250e330 rpm). Experimental process parameters such as die temperature (Tdie)
Modeling (120.7e170.6
C) and pressure (Pdie) (3160e7683 kPa) and product expansion ratio (ER) (3.3e16.9) and
Flow behavior cell size (R) (435e655 m) compared well with simulated results viz., Tdie(116.8e176.1
C), Pdie(3478
Cell structure e6404 kPa), ER(4.6e19.4) and R(426e728 m).
Bubble growth dynamics 2016 Elsevier Ltd. All rights reserved.

1. Introduction out mechanical interactions and thermal processing. The changes

Extrusion cooking is a widely used method in various food and
non-food applications (Lai and Kokini, 1991; Tang and Alavi, 2012;
Devi et al., 2013; Koppel et al., 2014; Lakshmi Devi et al., 2014).
Cellular structure in extruded foods and the way they are processed
affect their mechanical properties and texture (Agbisit et al., 2007;
Babin et al., 2007). Cell structure also affects oil uptake during post-
extrusion steps such as frying and hence is related to oil content in
snack products (Bouchon et al., 2003). Studying ow behavior of
the melt inside an extruder helps in understanding the cell struc-
ture formation during extrusion. This can be done by using theo-
retical tools such as mathematical modeling.
The concept of modeling the extrusion process started in the
beginning of 1940's with the characterization of ow of polymers.
The understanding of the behavior of plastics and other polymers is
comparatively easier, because of the homogeneity in structure and
well characterized physical and rheological properties (Tadmor and
Gogos, 1979; Karwe and Jaluria, 1990; Gopalakrishna et al., 1992).
Thus the polymers possess a predictable behavior upon carrying
* Corresponding author.
E-mail address: salavi@ksu.edu (S. Alavi).
upon interactions between bio-molecules and extruder are more
complex (Chiruvella et al., 1996). In food extrusion, the rst model
of a twin-screw corotating extruder was published by Yacu (1985).
This model was too general and pointed out the need for better
denition of geometry of twin-screw extruder and behavior of
product. Tayeb et al. (1989) published a model computing the
pressure proles and lled length in the extruder where the
product is a homogeneous melt. They validated the model using
maize starch and a simple screw prole. This model also found
good agreement between predicted and literature experimental
residence time distributions.
The modeling of bubble growth surrounded by a polymeric uid
under a given pressure started many years ago. A large number of
studies were made for polymer systems with different degrees of
complication (Barlow and Langlois, 1962; Yang and Yeh, 1966;
Street, 1968; Mikic et al., 1970; Street et al., 1971; Amon and
Denson, 1984; Arefmanesh et al., 1990, 1992; Venerus and Yala.
1997, Venerus et al., 1998; Venerus, 2001; Pai and Favelukis,
2002). A detailed review of mathematical modeling of bubble
growth in polymer systems is provided by Alavi et al. (2003a).
Several studies have also focused on modeling bubble growth or
shrinkage in biopolymer systems including processes such as
extrusion, baking and drying (Shimiya and Yano, 1987; Fan et al.,

http://dx.doi.org/10.1016/j.jfoodeng.2016.10.025
0260-8774/ 2016 Elsevier Ltd. All rights reserved.
8 P.H. Manepalli et al. / Journal of Food Engineering 198 (2017) 7e16

1994, 1999; Achanta et al., 1997; Shah et al., 1998; Mitchell et al., degradation of starch components and phase transitions such as
1998; Huang and Kokini, 1999). Detailed models were provided starch gelatinization, protein denaturation etc. were neglected due
by Schwartzberg et al. (1995) for bubble growth at the microscopic to difculty in modeling such reactions. The amount of energy
level in vapor induced pufng in popcorn, and Alavi et al. (2003a, generated (DE) from an element of thickness Dx of screw (Yacu,
2003b) for supercritical uid (CO2) extrusion of starchy melts. 1985) is
The formation of cell structure in expanded extrudates is
p 
dependent on the expansion and subsequent collapse of the cells in 4mN2 Dx pD  2Dh
the melt. This is controlled by the complex balance between the DE (1)
h
forces which drive deformation and forces which resist deforma-
tion and extensibility of the melt (Alavi et al., 2011). However, there where m is viscosity of the melt given by Eq. (2), N is screw speed, D
are no studies in the literature which integrate the ow inside the is the diameter of the screw and h is screw channel depth.
extruder and bubble dynamics models. This shortcoming was
addressed in the current study. The goal of this study was to m K gn1 (2)
develop a mathematical model for the ow behavior and cell
structure formation during extrusion and obtain the physical  
2650 Xw
characteristics of the extruded product (output parameters) such as K 4:224  exp  25 Pa sn (3)
T 1 Xw
expansion ratio, bubble radius, open cell fraction etc using the in-
dependent variables (input parameters) such as feed rate, screw
pDN
speed, water input in the extruder and pre-conditioner etc. This g (4)
model was developed with the objective of eventually using in a h
stochastic simulation of the extrusion process, which is described
where n is ow behavior index and K is consistency coefcient of
elsewhere (Manepalli, 2014).
the melt at temperature T and moisture content Xw and was
modeled based on a similar equation from Parker et al. (1989) and
2. Model development
Padmanabhan and Bhattacharya (1993).
The increase in temperature (DTE) inside the extruder due to the
2.1. Modeling the ow behavior inside the extruder
energy generated (DE) (Yacu, 1985) was given by

Tayeb et al. (1989) and Yacu (1985) developed a basic model for a DE
twin screw extruder, which were used to develop the basic model DTE
 (5)
m f Cp
inside the extruder. Pressure prole was developed based on Tayeb
et al. (1989) and temperature prole was developed based on Yacu
(1985). They did not consider the effect of thickness of the screw where mf is mass ow rate of the product inside the barrel and Cp is
ight. Hence, the equations were modied to incorporate the screw specic heat.
thickness effect as well based on Rossen and Miller (1973). There Specic mechanical energy (SME) during the extrusion was
were two sections inside the extruder: Solid conveying section and calculated by
melt pumping section. The actual length of the melt pumping P
DE
section was the parameter which was affected by screw prole, SME (6)
screw speed, throughput, viscosity of the material and the overall mf
ow resistance (Yacu, 1985). The temperature increases very
quickly in the melt pumping section. Viscosity of the material was
based on temperature and moisture content and hence it changes
as well. Hence, initial length of the melt was assumed. The tem- 2.1.2. Pressure prole
perature and pressure proles were developed along the barrel and No pressure was developed in the solid conveying section as the
nal temperature and pressure at the end of the barrel were screws were only partially full. Hence, the entire pressure was
obtained. generated in the melt pumping section. The pressure developed
The pressure drop across the die was calculated based on the was calculated using the ow rate equation. Leakage ow i.e. the
nal temperature obtained at the end of the barrel. If the pressure material leaking past the screw in the small gap between screw and
drop across the die was greater than the pressure at the end of the barrel was ignored. Hence, the ow rate consists of only two
barrel, the actual length of the melt was increased. If the pressure components: Drag ow along the direction of ow and pressure
drop across the die was lesser than the pressure at the end of the ow acting in opposite direction to the direction of ow due to the
barrel; the actual length of the melt was decreased and the same generation of pressure (Tayeb et al., 1989).
pressure and temperature proles were developed. This iterative "  2 !# 
evaluation was done till the pressure drop was almost same as the Qv 1 D2i D nf e
FD pND cos F 1 
2
ln 1 
nal pressure at the end of the barrel. lv 4 D2  D2i Di t
"   #
2.1.1. Temperature prole 1 DPE
2  2DDi D 2
 FP D  D2i 1  ln
The viscous heat dissipation in the conveying section was 32m Dq D  Di
2 2 Di
negligible as the screws were only partially full. The effect of  
nf e
thermal energy provided by the barrel on the temperature evolu-  1
t
tion inside the extruder was assumed as negligible. Hence, the
(7)
temperature does not increase in the conveying section. The tem-
perature prole along the barrel was developed based on the me-
where Qv is the volumetric ow rate, lv is channel width, Di is in-
chanical energy generated. The material owing inside the
ternal screw diameter and F is screw pitch angle. The term
extruder was considered to be homogeneous non-Newtonian uid  
ne
following the power law. The complex chemical reactions, 1  tf is incorporated to include the effect of screw thickness
 P.H. Manepalli et al. / Journal of Food Engineering 198 (2017) 7e16
based on Rossen and Miller (1973) where nf is number of threads, e
is screw ight thickness and t is the pitch of the screw element. The
Eq. (7) helps in calculating the pressure gradient DPE/Dq where
Dq Dx/pD. This helps in calculating the increase in pressure from
an element of thickness Dx of screw.
FD and FP are the correcting shape factors for drag ow and
pressure ow respectively and are dependent on depth/width ratio
of the channel (Rauwendaal, 1986). The values of these parameters
were obtained from Jiang (2008).

h
FD 1  0:5356  n0:4
lv

h
FP 1  0:6216  n0:4
lv
2.1.3. Pressure drop across the die
The pressure drop across the die is given by the uid ow
equation for laminar ows in non-Newtonian uids and was
calculated as described in Alavi et al. (2003a).
2.2. Microscopic model for extrudate expansion
The microscopic model comprises of bubble growth in the melt.
It takes place due to the pressure difference between vapor pres-
sure and opposing pressure components which include elastic
stress, yield stress, tensile stress and atmospheric pressure. The
growth of bubble also depends on the rheological properties of the
melt. As the bubble grows due to conversion of moisture to steam,
the steam migrates into the bubble from the microscopic layers of
the matrix. The process dynamics at the microscopic level is studied
by considering all bubbles to be uniform and spherical in shape.
Schwartzberg et al. (1995) developed a model for vapor induced
pufng (VIP) in popcorn kernels, which was used as the base to
develop microscopic model. The ow inside the extruder is coupled
Fig. 1. Pressure components acting on microscopic layers of the individual bubble during expansion and shrinkage.
9
to the microscopic model by linking the temperature and moisture
of the melt at the die with that in the microscopic layers. Brief
description was provided below, but detailed model development
approach and equations can be obtained from Manepalli (2014).
2.2.1. Bubble expansion and shrinkage
The domain is considered to be nite. It consists of a single
bubble lled with vapor and surrounded by material. The bubble is
considered to be spherical in shape and it is divided into Nmi layers
for nite element modeling. Radius of the bubble is R and the cell
(8) wall thickness of the domain is W and radius of the domain is L;
which is R W. Initial values of these parameters are indicated
with subscript o. Bubble expansion and shrinkage takes place
(9) depending on the pressure components acting on the wall of an
individual bubble. These pressure components include vapor
pressure (Pw), elastic stress (Pe), tensile stress (2s/R), yield stress
(Py) and atmospheric pressure (Pa) (Fig. 1).
If the resultant pressure of these pressure components acts in
the outward direction, it leads to expansion of bubble. Therefore,
the rate of expansion is dependent on Pw  Pa  Pe  Py  2s/R and
the rheological properties of the domain. As the water keeps
diffusing into the bubble and temperature falls, the vapor pressure
decreases with time. Hence, if the resultant pressure acts in the
inward direction, the bubble shrinks and the rate of shrinkage is
dependent on Pa Pe 2s/R  Pw  Py. Pw was calculated as
described in Schwartzberg et al. (1995). Elastic stress (Pe) and yield
stress (Py) pressure components acting on the cell wall is given by
Eq. (10) and Eq. (14) respectively (Schwartzberg et al., 1995; Alavi
et al., 2003b; Manepalli, 2014)
"   # 3 
5 2Ro 1 Ro 4 L  R3
Pe E   (10)
2 R 2 R L3 2R3
E Eb Xa  0:14 (11)
10 P.H. Manepalli et al. / Journal of Food Engineering 198 (2017) 7e16

h i
E Eb exp bf 0:14  Xa Xa < 0:14
Dt Dw;i1=2 Ai ri1 Xi1  ri Xi
(12) DXi
DM Ri1  Ri

Dw;i1=2 Ai1 ri Xi  ri1 Xi1

Eb Er exp0:1383  T (13)  (21)
Ri  Ri1
Er, bf are constants (Er 5 kPa, bf 73).where, E is the elastic
modulus modeled as a function of average moisture content (Xa) Dw;i Dw;i1  
and temperature (T) (Alavi et al., 2003b; Manepalli, 2014). Dw;i1=2 and Dw;i1=2 Dw;i Dw;i1 2
2
(22)
Py 3:464to 1=3 lnL=R  (14)
where Dw, i is the diffusion coefcient of ith layer, Ai is the area of
to tb Xa  0:14 (15) diffusion for ith layer, Ri is the radius of ith layer and ri is the density
of the dry matter in ith shell of the individual bubble. They were
h i calculated as described in Schwartzberg et al. (1995). DM in the
to tb exp bf 0:14  Xa Xa < 0:14 (16) above equations represents the mass of dry solid in each layer of the
bubble (DM M/Nmi). Mass of dry solid in domain M is calculated
by using M rdry/Nbubble where rdry is the density of dry unex-
tb tr exp0:1383  T (17) panded melt and Nbubble is the number of bubbles/m3 of unex-
panded material.

tr, bf are constants (tr 4.65 kPa, bf 73)

2.3. Macroscopic model for extrudate
where to is the ow yield stress modeled as a function of average
moisture content (Xa) and temperature (T) based on a similar The macroscopic model accounts for the heat and mass transfer
equation from Wang et al. (2005). that takes place at the extrudate level. The process dynamics at the
Neglecting the inertial effects, Eq. (18) is used to evaluate the macroscopic level was studied by considering a cylindrical shaped
rate of change of bubble radius (DR) in timestep Dt during expan- extrudate. Alavi et al. (2003a) was used as a base to develop
sion based on different pressure components (Schwartzberg et al., macroscopic model.
1995; Alavi et al., 2003a; Manepalli, 2014).
2.3.1. Heat transfer
2 31=n
For heat transfer at the macroscopic level, the cylindrical
Pw  Pa  Py  Pe  2Rs 7
DR RDt6
4 p n1 h 5 (18)
extrudate was discretized into nite number of concentric cylin-
42= 3 R3n i drical layers (Nb). The extrudate cools down after it exits the die due
n x K c  K s L to conduction, convection and evaporation. The fall of temperature
due to evaporation is discussed later in section 2.4.2 where the
NX
mi 1
 3n microscopic and macroscopic models are coupled. The heat transfer
2R
x Ki  Ki1  (19) in the outermost layer takes place due to conduction with the
i1
Ri Ri1 adjacent layer and convection with the atmosphere and heat
transfer in the inner layers takes place due to conduction. The
x in Eq. (19) accounts for the changes in consistency index (K)
change in temperature of each layer DTj (j 1 to Nb) in timestep Dt
across the domain due to differences in moisture content across the
was calculated as described in Alavi et al. (2003a).
different layers of the bubble and subscript i refers to the ith shell in
the discretization of the spherical bubble.
If the resultant pressure on the cell wall acts in the inward di- 2.3.2. Mass transfer
rection, the bubble shrinks and the rate of change of bubble radius Mass transfer takes place at both microscopic and macroscopic
is given by Eq. (20) (Alavi et al., 2003a; Manepalli, 2014). level. As the outermost layer is in contact with atmosphere, water
diffuses from the outermost layer into the atmosphere. The change
2 31=n in moisture content of each layer DCj (j 1 to Nb) in timestep Dt
Pa Pe 2s  Pw  Py 7 was calculated as described in Alavi et al. (2003a).The average
DR RDt6
4 p n1
R

5 (20)
42= 3  3n moisture (Cavg) is obtained by taking the mean of all moisture
n x Kc  Ks RL contents.

2.4. Micro-macro coupling

The linking of microscopic modeling with macroscopic

2.2.2. Diffusion of water
During the expansion of the bubble, water vapor diffuses into
the bubble from the domain. Part of the moisture present in the
spherical layers of the bubble is converted into steam which mi-
grates into the bubble. As more and more moisture is converted to
steam inside the bubble, diffusion takes place in the spherical layers
of the bubble in the inward direction. Xi represents the moisture
content of the ith spherical layer at any instant of time and the
change in moisture content DXi in timestep Dt is calculated using
the Eq. (21) (Schwartzberg et al., 1995).
modeling includes the linking of moisture contents of microscopic
layers of bubble to macroscopic layers of the extrudate; tempera-
ture fall of the extrudate due to evaporation of water and coupling
the growth of the bubble with the expansion of the extrudate.
2.4.1. Linking of moisture content
In order to link the macroscopic diffusion model to microscopic
diffusion, the microscopic moisture content of all the layers Xi is
reduced by same extent as reduction in Cavg (calculated in section
2.3.2) at each timestep.
 P.H. Manepalli et al. / Journal of Food Engineering 198 (2017) 7e16 11

2.4.2. Temperature reduction due to evaporation Table 1

DQ is the amount of water evaporated from a bubble of mass (M) Input parameters for model.

in a timestep Dt. It is calculated as described in Schwartzberg et al. Parameter Value

(1995). This is linked to macroscopic model by decreasing the Number of discretized microscopic spherical layers Nmi 10
temperatures across all the macroscopic layers by DQl/MCp where l Number of discretized macroscopic cylindrical tubes Nb 7
is the latent heat of vaporization. Bubble nucleation density Nbubble (bubbles/cm3 of unexpanded melt) 27500
Ambient temperature (K) 300
Initial bubble Size Ro (m) 3E-04
Surface tension at pore-shell interface s (kN/m) 1E-05
2.4.3. Expansion ratio of the extrudate Flow behavior index (n) 0.4
The expansion of the extrudate is coupled to the growth of the Density of dry unexpanded melt rdry (kg/m3) 1160
Convective heat transfer coefcient (W/m2K) 15
bubble by using a modied form of equation used by Schwartzberg
Mass transfer coefcient of water in starch matrix (m/s) 1E-05
et al. (1995). Ambient moisture content 0.04
Expansion ratio (ER) is calculated using the Eq. (23). Latent heat of vaporization (kJ/kg) 2250

 3 X k  3
Lk Lj 3. Materials and methods
ER 1  fo k Dfo j (23)
Lo j1
Lo
3.1. Extrusion run
where fo is open cell fraction and L is the domain radius which
were calculated as described in Schwartzberg et al. (1995). The Degermed corn meal purchased from Bunge (Atchison, KS) was
rst term in the equation accounts for expansion due to closed used for production of corn puffs. The corn puffs were extruded
cells whereas the second term accounts for expansion due to open using a pilot scale twin screw extruder (TX-52, Wenger
cells. Manufacturing, Sabetha, KS, USA) with a differential diameter pre-
conditioning cylinder. The extruder had a screw diameter of 52 mm
and length to diameter ratio of 19. The screw prole and the barrel
temperatures used were reported in Fig. 3. A 3  2 factorial design
2.5. Algorithm development
was used with 3 in-barrel moisture contents (19%, 23.5%, 28% (db))
and two screw speeds (250 rpm and 330 rpm). The notations used
The model equations were written in Visual Basic. The basic
for this treatments are shown in Table 2. Extrusion conditions were
algorithm used in developing the code is shown in Fig. 2. The input
allowed to stabilize for ~10 min. The product from each treatment
parameters, other than that of screw prole, die prole and
was collected for about 10 min.
extrusion conditions are shown in Table 1.

Fig. 2. Algorithm for the mathematical model.

12 P.H. Manepalli et al. / Journal of Food Engineering 198 (2017) 7e16

Fig. 3. Screw prole.

Table 2


Notations for treatments.   tto
kJ 100 N
Treatment notation Moisture content (d.b.) % Screw speed (RPM) SME   Pr (24)
kg m_ Nr
M.C. LO, RPM LO 19 250
M.C. LO, RPM HI 19 330
where t is the % motor load torque, to is the no load torque %, N is
M.C. MD, RPM LO 23.5 250
M.C. MD, RPM HI 23.5 330 the screw speed of extruder, Nr is the rated screw speed (336 rpm),
M.C. HI, RPM LO 28 250 Pr is the rated motor power (22.37 kW) and m _ is the mass ow rate
M.C. HI, RPM HI 28 330 (kg/s).

The raw material feed rate was maintained at 110 kg/h. There
was no water or steam added in the preconditioner. The water was
added only in the extruder. The die used was a circular die of
4.2 mm diameter. The product was cut immediately after exiting
the die with a face-mounted ex knife (6 blades) rotating at
539 rpm. Detailed experimental description can be obtained from
Manepalli (2014).
3.2. X-ray microtomography
For determining the microstructure parameters i.e. average pore
radius and cell wall thickness, representative samples were
collected from each treatment were selected for image analysis. A
desktop X-ray microtomography imaging system (Model 1072,
20e100 kV/0e250 mA, SkyScan, Aartselaar, Belgium) was used to
scan the samples. A set of two-dimensional virtual slices were
obtained after reconstruction for each sample. Image analysis
software in XMT uses marching cube algorithm to calculate
microstructural parameters such as average pore radius and cell
wall thickness.
3.4. Expansion ratio
The Sectional Expansion ratio (SER) is the ratio of the extrudate
cross-sectional area to the die orice cross sectional area, and was
calculated using Eq. (25).
D2e
SER (25)
d2ex
where De is the extrudate diameter measured using calipers and dex
is the exit diameter of the die. Longitudinal Expansion ratio is
assumed as 1 which makes the overall or volumetric expansion
ratio (ER) equal to SER. The video of the product extruded out of the
die was recorded using a digital video recorder (DVR 840XHD,
Vivitar, Santa Monica, CA, USA). The maximum expansion ratio of
the extrudate was obtained by calculating the maximum diameter
from the video using pixel ruler software.
3.5. Statistical analysis

3.3. Specic mechanical energy The experimental and simulated values of the parameters are
compared by calculating the mean relative percentage deviation
Specic mechanical energy (SME) was calculated experimen- (MRPD) given in Eq. (26). Lower the MRPD, the better is the
tally using Eq. (24). goodness of t of the mathematical model developed.
 P.H. Manepalli et al. / Journal of Food Engineering 198 (2017) 7e16
Fig. 4. Predicted and experimental specic mechanical energy comparison.
Fig. 5. Predicted and experimental die temperature and die pressure comparison.
13
4. Results and discussion
Predicted expansion ratio, die temperature, die pressure and
average pore radius and cell wall thickness for each of the nal
extrudates were compared with experimental data at different
moisture contents and screw speeds to validate the model.
4.1. Pressure, temperature and energy inside the extruder
The comparisons between predictions and experimental values
of temperature at die, specic mechanical energy and pressure at
die were shown in Figs. 4 and 5. The MRPD for temperature at die,
pressure at die and specic mechanical energy were 4.76%, 5.66%
and 3.22% respectively which shows that the simulated values were
very similar to experimental values.
Simulated results (Figs. 4 and 5) show that SME and tempera-
ture at die decrease with the increase in moisture. This was due to
the decrease in viscosity of melt with increase in moisture thus
decreasing the SME and die temperature. Previous studies (Cheng
et al., 2007; Alavi et al., 2011) also reported similar effect of mois-
ture content. Also the simulated results showed that SME and die
temperature increased with increase in screw speed. This was due
to the increase in shear rate when the screw speed was increased.
The increase in SME with increase in screw speed is in agreement
with previous studies (de Mesa et al., 2009; Zhu et al., 2010). Die
pressure decreased with increase in screw speed (Fig. 5) as it
increased the shear rate in the barrel which decreased the viscosity
of the melt. Fletcher et al. (1985), Jin et al. (1994) also reported the
similar effect of screw speed on die pressure.
4.2. Expansion ratio
The change in expansion ratio of the extrudate with time is
shown in Fig. 6. The product diameter increases at a relatively lower
rate as soon as it exits the die but increases rapidly at the end of
expansion. Expansion ratio decreased with the increase in moisture
content. As explained earlier, higher moisture content led to
decrease in specic mechanical energy and die temperature. Lower

No 
 die temperature leads to decrease in vapor pressure (Pw) which is a
100 X Zobs;i  Zpre;i  major driving force for expansion. Hence, lower die temperature
MRPD (26)
No i1 Zobs;i resulted in lower expansion. Similar trends were observed in pre-
vious studies (Chinnaswamy and Hanna, 1988; Cheng et al., 2007;
where Zpre, i is the ith predicted or simulated value of the param- Karkle et al., 2012).
eter, Zobs, i is the ith observed value and No is number of Expansion ratio increased with increase in screw speed which
observations. can be explained using the same phenomenon i.e. increase in screw

Fig. 6. Expansion ratio versus time for various treatments.

14 P.H. Manepalli et al. / Journal of Food Engineering 198 (2017) 7e16
Fig. 7. Predicted and experimental maximum expansion ratio and nal expansion ratio
comparison.
speed led to higher die temperature which led to higher expansion.
Previous studies (de Mesa et al., 2009; Zhu et al., 2010) also re-
ported the similar effect of screw speed.
The comparison between the predicted and experimental
expansion ratio (both maximum and nal expansion ratio) were
shown in Fig. 7. The MRPD for maximum expansion ratio and nal
expansion ratio were 21.06% and 25.95% respectively, with pre-
dicted values generally higher than experimental.
4.3. Temperature fall in extrudate
Fall in temperature of the extrudate is caused due to convection
with the atmosphere as well as evaporation of water. As the bubble
expands, water diffuses into the bubble and evaporates. Latent heat
of vaporization causes the drop in energy which leads to drop in
temperature. The temperature prole of the extrudate after it exits
the die is shown in Fig. 8.
4.4. Bubble growth
Fig. 8 also shows the growth and shrinkage of bubble (pore)
with time of exit from the die. The bubble grows relatively slowly
as soon as it exits the die due to high cell wall thickness but it
grows very rapidly at the end of expansion as the cell wall
Fig. 8. Temperature after exiting the die and bubble radius versus time for treatment M.C. LO, RPM LO.
thickness keeps decreasing. As the vapor pressure decreases due
to fall in temperature, the vapor pressure inside the bubble be-
comes less than the outer pressure and the bubble
starts shrinking, causing the shrinkage of the extrudate as well.
Also, a fraction of the bubbles rupture and become open cells
during expansion. The open cells lose their ability to expand and
shrink.
The experimental average bubble radius and cell wall thickness
were obtained from X-ray microtomography (XMT). The XMT im-
ages for some treatments are shown in Fig. 9. The comparisons
between predictions and experimental values of average pore
radius and cell wall thickness are shown in Table 3. The trend shows
that the average bubble radius decreases with increase in moisture
and increases with increase in screw speed. The reasoning behind
this was the same as explained earlier for expansion ratio i.e. the
major driving force for expansion (vapor pressure) is dependent on
die temperature. Increase in screw speed or decrease in moisture
content leads to higher die temperature which increases the bubble
radius. The MRPD for bubble radius was 15.67% which shows that
the simulated values were similar to experimental values. Also,
same trend was observed in experimental and simulated values of
bubble radius.
The same trend was observed in experimental and simulated
cell wall thickness i.e. decrease in moisture content and increase
in screw speed lead to lower cell wall thickness. But the simulated
cell wall thickness obtained for all the treatments were very low
compared to the experimental cell wall thickness. This can be due
to the high bubble number density used (27500 bubbles/cm3 of
unexpanded melt) which led to very low predicted cell wall
thickness. A proper bubble number density and initial cell radius
can improve the model prediction for cell wall thickness and
bubble radius. One other reason for higher value of the experi-
mental cell wall thickness may be its overestimation by X-ray
microtomography as it uses marching cube algorithm. The
marching cube algorithm is sensitive to pixelation effects (Porter
and Wildenschild, 2010) which can lead to overestimation of
values.
The model can be further improved by using different viscosity
equations for viscosity inside the extruder and after the product
exits the die. This is due to the differences in the shear rate inside
the extruder and during the bubble growth. The shear rate inside
the extruder is very high, whereas after the extrudate exits the
die, the shear rate rapidly approaches zero due to free-surface
ow.
 P.H. Manepalli et al. / Journal of Food Engineering 198 (2017) 7e16 15

Fig. 9. X-Ray Microtomography results for various moisture contents at RPM LO.

Table 3
Predicted and experimental bubble radius and cell wall thickness comparison.

Treatment Bubble radius Bubble radius Cell wall thickness Cell wall thickness
(experimental) Microns (predicted) Microns (experimental) Microns (predicted) Microns

M.C. LO, RPM LO 655 794.5 68.3 4.15

M.C. LO, RPM HI 728.4 855 75.8 3.56
M.C. MD, RPM LO 554.9 642 114.2 6.6
M.C. MD, RPM HI 561.6 626 131.2 6.94
M.C. HI, RPM LO 426.4 579.9 179.4 8.65
M.C. HI, RPM HI 434.9 544 191.5 9.83

4.5. Literature comparison although the predicted values of the cell wall thickness were very
less (3.5e9.8 m).
Apart from the experimental validation of the mathematical
model, the results from mechanistic model were compared with
5. Conclusion
experimental data in literature for further validation. The values for
expansion ratio of corn-based expansion products (Table 4) in
A deterministic mathematical model was developed for simu-
previous studies ranged from 3 to 19, depending on the formulation
lation of ow inside the extruder and expansion and shrinkage of
and processing conditions which is in the same range of the value
bubbles after the extrudate exits the die. The model was validated
for expansion ratio predicted by the mathematical model
using pilot scale twin screw extrusion for processing of cereal-
(4.5e19.4). Also, the range of predicted cell size (544e855 m) and
based cellular products. There were many assumptions for mate-
experimental values (426e728 m) obtained for the cell size
rial properties approximated from literature. Despite several as-
demonstrated a good match with the values of the cell sizes ob-
sumptions, the predicted results of pressure, temperature, energy
tained from literature (375e1900 m). The experimental values ob-
inside the extruder; nal and maximum expansion ratio demon-
tained for cell wall thickness (68.3e191.5 m) is in the same range of
strated a good match with the experimental measurements and the
the cell wall thickness obtained from literature (36e312 m)
values from literature. The model was useful in understanding the

Table 4
Literature comparison for expansion ratio, microstructure of corn based expanded products at different processing conditions.

Reference Material Moisture content Screw speed Cell size (radius) Cell wall thickness Expansion
(db) (RPM) (microns) (microns) ratio

Karkle et al. (2012) Corn our 21-33% 350 375e525 90e165 5.9e10.5
Chinnaswamy and Hanna Corn starch 15-40% 80e200 e e 3.8e16.1
(1988)
Cheng et al. (2007) Corn starch 29-37% 300 785e1470 36e80.5 e
Trater et al. (2005) Corn starch 5% WPC 35-51% e 575e850 65e75 e
Anton et al. (2009) Corn starch 28.2% 150 e e 6.45
Desrumaux et al. (1998) Corn grits 22.5% 130e200 e e 9.14e12.8
Agbisit et al. (2007) Corn starch 30-41% 200e400 1035e3160 65e125 e
de Mesa et al. (2009) Corn starch 28.2% 230e330 e e 17.5e19
Babin et al. (2007) High amylose maize 25% e 1650e1900 176.5e312 e
starch
Ahmed (1999) Corn grits 22% 200 e e 3
Barrett (1992) Corn meal 17.6e25% 300e400 726e1052 e e
Mezreb et al. (2003) Corn our e 200e500 e e 7.6e11.8
16 P.H. Manepalli et al. / Journal of Food Engineering 198 (2017) 7e16
various phenomena involved in the formation of cellular structure
and estimating the relative signicance of various material and
processing parameters. This model was developed with the
objective of eventually using in a stochastic simulation of the
extrusion process as described elsewhere (Manepalli, 2014).
References
Achanta, S., Okos, M.R., Cushman, J.H., Kessler, D.P., 1997. Moisture transport in
shrinking gels during saturated drying. AIChE J. 43 (8), 2112e2122.
Agbisit, R., Alavi, S., Cheng, E.Z., Herald, T., Trater, A., 2007. Relationships between
microstructure and mechanical properties of cellular cornstarch extrudates.
J. Texture Stud. 38 (2), 199e219.
Ahmed, Z.S., 1999. Physico-chemical, structural and sensory quality of corn-based
ax-snack. Nahrung 43, 253e258.
Alavi, S.H., Rizvi, S.S.H., Harriott, P., 2003a. Process dynamics of starch-based
microcellular foams produced by supercritical uid extrusion. I: model devel-
opment. Food Res. Int. 36 (4), 309e319.
Alavi, S.H., Rizvi, S.S.H., Harriott, P., 2003b. Process dynamics of starch-based
microcellular foams produced by supercritical uid extrusion. II: numerical
simulation and experimental evaluation. Food Res. Int. 36 (4), 321e330.
Alavi, S., Karkle, E., Adhikari, K., Keller, L., 2011. Extrusion research for addressing
the obesity challenge. Cereal Foods World 56 (2), 56e60.
Amon, M., Denson, C.D., 1984. A study of the dynamics of foam growth: analysis of
the growth of closely spaced spherical bubbles. Polym. Eng. Sci. 24, 1026e1034.
Anton, A.A., Fulcher, R.G., Arnteld, S.D., 2009. Physical and nutritional impact of
fortication of corn starch-based extruded snacks with common bean (Pha-
seolus vulgaris L.) our: effects of bean addition and extrusion cooking. Food
Chem. 113, 989e996.
Arefmanesh, A., Advani, S.G., Michaelides, E.E., 1990. A numerical study of bubble
growth during low pressure structural foam molding process. Polym. Eng. Sci.
30 (20), 1330e1337.
Arefmanesh, A., Advani, S.G., Michaelides, E.E., 1992. Accurate numerical solution
for mass diffusion-induced bubble growth in viscous liquids containing limited
dissolved gas. Int. J. Heat Mass Transf. 35 (7), 1711e1722.
Babin, P., Della Valle, G., Dendievel, R., Lourdin, D., Salvo, L., 2007. X-ray tomography
study of the cellular structure of extruded starches and its relation with
expansion phenomenon and foam mechanical properties. Carbohydr. Polym.
68, 329e340.
Barlow, E.J., Langlois, W.E., 1962. Diffusion of gas from a liquid into an expanding
bubble. IBM J. Res. Dev. 6, 329e337.
Barrett, A.M., 1992. Cell size distributions of puffed corn extrudates. J. Food Sci. 57,
146e148.
Bouchon, P., Aguilera, J.M., Pyle, D.L., 2003. Structure oil-absorption relationships
during deep-fat frying. J. Food Sci. 68 (9), 2711e2716.
Cheng, E.M., Alavi, S., Pearson, T., Agbisit, R., 2007. Mechanicaleacoustic and sen-
sory evaluations of cornstarchewhey protein isolate extrudates. J. Texture Stud.
38, 473e498.
Chinnaswamy, R., Hanna, M.A., 1988. Optimum extrusion-cooking conditions for
maximum expansion of corn starch. J. Food Sci. 53 (3), 834e836.
Chiruvella, R.V., Jaluria, Y., Karwe, M.V., 1996. Numerical simulation of the extrusion
process for food materials in a single screw extruder. J. Food Eng. 30 (3),
449e467.
de Mesa, N.J.E., Sajid, A., Singh, N., Shi, Y.C., Dogan, H., Sang, Y., 2009. Soy protein-
fortied expanded extrudates: baseline study using normal corn starch. J. Food
Eng. 90 (2), 262e270.
Desrumaux, A., Bouvier, J.M., Burri, J., 1998. Corn grits particle size and distribution
effects on the charactersitics of expanded extrudates. J. Food Sci. 63 (5), 1e7.
Devi, N.L., Shobha, S., Tang, X., Shaur, S.A., Dogan, H., Alavi, S., 2013. Development of
protein-rich sorghum-based expanded snacks using extrusion technology. Int. J.
Food Prop. 16, 263e276.
Fan, J.T., Mitchell, J.R., Blanshard, J.M.V., 1994. A computer simulation of the dy-
namics of bubble growth and shrinkage during extrudate expansion. J. Food
Eng. 23 (3), 337e356.
Fan, J.T., Mitchell, J.R., Blanshard, J.M.V., 1999. A model for the oven rise of dough
during baking. J. Food Eng. 41 (2), 69e77.
Fletcher, S.I., Richmond, P., Smith, A.C., 1985. An experimental study of twin-screw
extrusion-cooking of maize grits. J. Food Eng. 4, 291e312.
Gopalakrishna, S., Jaluria, Y., Karwe, M.V., 1992. Heat and mass transfer in a single
screw extruder for non-Newtonian materials. Int. J. Mass Heat Transf. 35,
221e237.
Huang, H., Kokini, J.L., 1999. Prediction of dough volume development which con-
siders the biaxial extensional growth of cells. In: Campbell, G.M., Webb, C.,
Pandiella, S.S., Niranjan, K. (Eds.), Bubbles in Food. Eagan Press, St Paul, MN,
USA., pp. 113e120
Jiang, Q., 2008. Modeling Flow, Melting, Solid Conveying and Global Behavior in
Intermeshing Counter-rotating Twin Screw Extruders (Ph.D. thesis). University
of Akron, Akron, Ohio.
Jin, Z., Hsieh, F., Huff, H.F., 1994. Extrusion cooking of corn meal with soy ber, salt
and sugar. Cereal Chem. 71 (3), 227e234.
Karkle, E.L., Keller, L., Dogan, H., Alavi, S., 2012. Matrix transformation in ber-added
extruded products: impact of different hydration regimens on texture, micro-
structure and digestibility. J. Food Eng. 108, 171e182.
Karwe, M.V., Jaluria, Y., 1990. Numerical simulation of uid ow and heat transfer in
a single screw extruder for non-Newtonian uids. Numer. Heat Transf. 17,
167e190.
Koppel, K., Gibson, M., Alavi, S., Aldrich, G., 2014. The effects of cooking process and
meat inclusion on pet food avor and texture characteristics. Animals 4 (2),
254e271.
Lai, L.S., Kokini, J.L., 1991. Physicochemical changes and rheological properties of
starch during extrusion (a review). Biotechnol. Prog. 7 (3), 251e266.
Lakshmi Devi, N., Shobha, S., Alavi, S., Kalpana, K.M., Soumya, M., 2014. Utilization of
extrusion technology for the development of millet based complementary
foods. J. Food Sci. Technol. 51 (10), 2845e2850.
Manepalli, P.H., 2014. Stochastic Modeling of Flow Behavior and Cell Structure
Formation during Extrusion of Biopolymer Melts (MS thesis). Kansas State
University, Manhattan, Kansas.
Mezreb, K., Gouilleux, A., Ralainirina, R., Queneudec, M., 2003. Application of image
analysis to measure screw speed inuence on physical properties of corn and
wheat extrudates. J. Food Eng. 57 (2), 145e152.
Mikic, B.B., Rohsenow, W.M., Grifth, P., 1970. On bubble growth rates. Int. J. Heat
Mass Transf. 13, 657e666.
Mitchell, J.R., Fan, J., Blanshard, J.M.V., 1998. Simulation of bubble growth in heat
processed cereal systems. In: Campbell, G.M., Webb, C., Pandiella, S.S.,
Niranjan, K. (Eds.), Bubbles in Food. Eagan Press, St Paul, MN, USA, pp. 107e112.
Padmanabhan, M., Bhattacharya, M., 1993. Effect of extrusion processing history on
the rheology of cornmeal. J. Food Eng. 18, 335e349.
Pai, V., Favelukis, M., 2002. Dynamics of spherical bubble growth. J. Cell. Plast. 38
(5), 403e419.
Parker, R., Ollett, A.L., Lai-Fook, R., Smith, A.C., 1989. The rheology of food melts and
its applications in extrusion processing. In: Carter, R.E. (Ed.), Rheology of Food,
Biological and Pharmaceutical Materials. Elsevier Science, London.
Porter, M.L., Wildenschild, D., 2010. Image analysis algorithms for estimating porous
media multiphase ow variables from computed microtomography data: a
validation study. Comput. Geosci. 14, 15e30.
Rauwendaal, C., 1986. Throughput-pressure relationships for power law uids in
single screw extruders. Polym. Eng. Sci. 26, 1240e1244.
Rossen, J.L., Miller, R.C., 1973. Food extrusion. Food Technol. 27, 46e53.
Schwartzberg, H.G., Wu, J.P.C., Nussinovitch, A., Mugerwa, J., 1995. Modelling
deformation and ow during vapor-induced pufng. J. Food Eng. 25 (3),
329e372.
Shah, P., Campbell, G.M., McKee, S.L., Rielly, C.D., 1998. Proving of bread dough:
modelling the growth of individual bubbles. Food Bio-products Process. 76 (2),
73e79.
Shimiya, Y., Yano, T., 1987. Diffusion-controlled shrinkage and growth of an air
bubble entrained in water and in wheat our particles. Agric. Biol. Chem. 51 (7),
1935e1940.
Street, J.R., 1968. The rheology of phase growth in elastic liquids. Trans. Soc. Rheol.
12 (1), 103e131.
Street, J.R., Fricke, A.L., Reiss, L.P., 1971. Dynamics of phase growth in viscous, non-
newtonian liquids. Ind. Eng. Chem. Fundam. 10 (1), 54e64.
Tadmor, Z., Gogos, C., 1979. Principles of Polymer Processing. Wiley, New York.
Tang, X., Alavi, S., 2012. Structure and physical properties of starch/poly vinyl
alcohol/laponite RD nanocomposite lms. J. Agric. Food Chem. 60, 1954e1962.
Tayeb, J., Vergnes, B., DellaValle, G., 1989. A basic model for a twin-screw extruder.
J. Food Sci. 54, 1047e1056.
Trater, A.M., Alavi, S., Rizvi, S.S.H., 2005. Use of non-invasive X-ray micro-
tomography for characterizing microstructure of extruded biopolymer foams.
Food Res. Int. 38, 709e719.
Venerus, D.C., 2001. Diffusion-induced bubble growth in viscous liquids of nite and
innite extent. Polym. Eng. Sci. 41 (8), 1390e1398.
Venerus, D.C., Yala, N., 1997. Transport analysis of diffusion-induced bubble growth
and collapse in viscous liquids. AIChE J. 43 (11), 2948e2959.
Venerus, D.C., Yala, N., Bernstein, B., 1998. Analysis of diffusion-induced bubble
growth in viscoelastic liquids. J. Newt. Fluid Mech. 75 (1), 55e75.
Wang, L., Ganjyal, G.M., Jones, D.D., Weller, C.L., Hanna, M.A., 2005. Modeling of
bubble growth dynamics and nonisothermal expansion in starch-based foams
during extrusion. Adv. Polym. Technol. 24 (1), 29e45.
Yacu, W.A., 1985. Modeling a twin screw co-rotating extruder. J. Food Process Eng. 8
(1), 1e21.
Yang, W.J., Yeh, H.C., 1966. Theoretical study of bubble dynamics in purely viscous
uids. AIChE J. 12 (5), 927e931.
Zhu, L.J., Shukri, R., de Mesa-Stonestreet, N.J., Alavi, S., Dogan, H., Shi, Y.C., 2010.
Mechanical and microstructural properties of soy protein e high amylose corn
starch extrudates in relation to physiochemical changes of starch during
extrusion. J. Food Eng. 100 (2), 232e238.