Deformação de Alimentos

You might also like

You are on page 1of 16

Chemical Engineering Science 66 (2011) 6482–6497

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Science


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

Transport in deformable food materials: A poromechanics approach


Ashish Dhall a,1, Ashim K. Datta b,n
a
Department of Biological and Environmental Engineering, Cornell University, 175 Riley-Robb Hall, Ithaca, NY 14853, United States
b
Department of Biological and Environmental Engineering, Cornell University, 208 Riley Robb Hall, Ithaca, NY 14853, United States

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

Article history: A comprehensive poromechanics-based modeling framework that can be used to model transport and
Received 17 February 2011 deformation in food materials under a variety of processing conditions and states (rubbery or glassy)
Received in revised form has been developed. Simplifications to the model equations have been developed, based on driving
26 August 2011
forces for deformation (moisture change and gas pressure development) and on the state of food
Accepted 1 September 2011
material for transport. The framework is applied to two completely different food processes (contact
Available online 16 September 2011
heating of hamburger patties and drying of potatoes). The modeling framework is implemented using
Keywords: total Lagrangian mesh for solid momentum balance and Eulerian mesh for transport equations, and
Mathematical modeling validated using experimental data. Transport in liquid phase dominates for both the processes, with
Food processing
hamburger patty shrinking with moisture loss for all moisture contents, while shrinkage in potato stops
Porous media
below a critical moisture content.
Solid mechanics
Pressure & 2011 Elsevier Ltd. All rights reserved.
Shrinkage

1. Introduction alternate approach is volume-averaging, i.e., begin with conserva-


tion equations at the microscale and then use averaging or
Factors affecting food safety (presence of pathogens and macroscopization to obtain relationships at the macroscale
toxins) and food quality (porosity, pore size distribution, texture, (Whitaker, 1977). In both approaches, the constitutive relation-
and color) are functions of the state (temperature, moisture, and ships can be written either empirically or by invoking second law
composition) of the food material and its processing history. of thermodynamics through entropy inequality (nonequilibrium
Fundamentals-based understanding of physics of food processing thermodynamics). Lewis and Shrefler (1998) provide a detailed
can help a long way in predicting the state and the history of a review of the similarities and dissimilarities, and the pros and
food material and, thus, its safety and quality. The underlying cons of these poromechanics theories. Although applied exten-
physics of many food processes, such as drying, rehydration sively to non-food materials, there are no examples of a compre-
(soaking), frying, baking, grilling, puffing and cooking, is essen- hensive poromechanics-based approach in food science literature.
tially energy and moisture transport in a deforming porous Majority of the existing transport models in food science
medium (Datta, 2007). Although transport in deformable porous literature are either curve fits of lumped empirical data (Ateba
media has been extensively studied for non-food applications and Mittal, 1994; Ikediala et al., 1996; Bengtsson et al., 1976;
such as geomaterials (soils, rocks, concrete, and ceramics), Chau and Snyder, 1988; Fowler and Bejan, 1991) or, in a slightly
biomaterials (plant and animal tissues), gels and polymers, still the improved version, assume purely conductive heat transfer for
combination of specific characteristics (softness, hygroscopicity and energy and purely diffusive transport for moisture (Dincer and
phase transitions) and processing conditions of food materials result Yildiz, 1996; Williams and Mittal, 1999; Shilton et al., 2002;
in unique complexities that have rarely been studied. Wang and Singh, 2004; Kondjoyan et al., 2006), solving a transient
The general mathematical framework of deformation in satu- conduction (or diffusion) equation using experimentally deter-
rated and unsaturated porous media (also known as poromecha- mined effective conductivity (or diffusivity). One notable excep-
nics) was developed by Biot (1965). The theory was later tion to lumped analysis is the application of Stefan’s moving
extended to include multiphase transport using theory of mix- boundary approach to track liquid–vapor interface during internal
tures by various studies (discussed by Schrefler, 2002). An vaporization (Farkas et al., 1996; Farid and Chen, 1998; Bouchon
and Pyle, 2005). In this type of modeling, the liquid–vapor inter-
face, where all the vaporization occurs, separates completely
n
Corresponding author. Tel.: þ1 607 255 2482; fax: þ1 607 255 4080.
saturated and completely dry regions of a food material. Some
E-mail addresses: ad333@cornell.edu (A. Dhall), akd1@cornell.edu (A.K. Datta). examples of detailed description of transport mechanisms based
1
Tel.: þ1 607 255 2871; fax: þ1 607 255 4080. on a porous media approach are: inclusion of vaporization

0009-2509/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2011.09.001
A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497 6483

generated pressure-driven flow during intensive heating pro- tensor, r, in the representative elementary volume (REV) around
cesses by Ni and Datta (1999), Halder et al. (2007) and the location (Lewis and Shrefler, 1998):
Yamsaengsung and Moreira (2002); nonequilibrium thermody- Z
1
namics based hybrid mixture theory approach towards Case-II r¼ r dV ð1Þ
V V
diffusion by Singh (2002) and Achanta (1995); and, more recently,
application of Flory–Rehner theory to predict swelling-pressure Now, total volume of an REV can be written as a sum of volumes
driven moisture transport in meat by van der Sman (2007). of the solid and the fluids present in the pores:
Fundamentals-based description of deformation in food mate- X
V ¼ Vs þ Vi ð2Þ
rials is even less frequent than the detailed description of i
transport itself. Two different approaches are generally followed:
either the experimental shrinkage data is empirically fitted as a Therefore, the total stress tensor can also be written as a sum of
function of moisture content, or the additivity of volumes of averages in the individual phase volumes:
different components is used to predict deformation from moist- Z XZ !
1
ure loss data (Mayor and Sereno, 2004; Katekawa and Silva, r¼ r dV þ r dV
V Vs i Vi
2006). Modeling of transport in deformable food materials as a  Z  X  Z 
Vs 1 Vi 1
solid mechanics problem and solving the linear momentum ¼ r dV þ r dV
V V s Vs V V i Vi
balance for the solid matrix is rare in food, although this approach X
i
is frequently used to study drying of some other materials such as ¼ es r s þ ðei r i Þ ð3Þ
wood and ceramics. Notable exceptions are study of hygrostress i
cracking (Izumi and Hayakawa, 1995), bread baking (Zhang et al.,
where ei and r i are, respectively, the volume fraction and the
2005) and microwave puffing of potatoes (Rakesh, 2010). For
volume-averaged stress of a phase, i. Given that shear stress is
detailed review of drying models that include shrinkage effects,
negligible in fluids, stress in a fluid, ri can be approximated as
including pioneering works by Perre and May (2001), Kowalski
(2000) and others, the reader is referred to the review by r i ¼ pi I ð4Þ
Katekawa and Silva (2006). Substituting fluid stresses from Eq. (4) in Eq. (3), we obtain
With this background, the current study is an attempt to X
develop a poromechanics-based modeling framework for the r ¼ es r s  ðei pi ÞI
i
coupled physics of transport and large deformation in food X
materials. The macroscale governing equations are based on ¼ ð1fÞr s f ðSi pi ÞI
i
extended Biot’s theory of poromechanics (Lewis and Shrefler, !
X X
1998). Classical constitutive laws are used in both mass transport ¼ ð1fÞ r s þ ðSi pi ÞI  ðSi pi ÞI ð5Þ
(Darcy and Fick’s laws) and conduction (Fourier’s law) and i i
deformation (hyperelastic solid) equations.
Defining the first term on the right-hand side of Eq. (5) as the
effective stress on the solid skeleton, r 0 , and the second term as
2. Mathematical model pore pressure, pf ð ¼ Sg pg þSw pw Þ, the well-known effective stress
principle of Terzaghi is recovered:
A mathematical model is developed that describes deforma- r ¼ r 0 pf I ð6Þ
tion and transport (energy and moisture) inside a food material
during thermal processing. First, the physics of deformation of the Now, by invoking the quasi-steady state assumption for deforma-
solid matrix is described, followed by a discussion on special tion (acceleration term equal to zero), the solid momentum
cases based on the driving mechanism behind deformation. Later, balance leads to divergence-free field of overall stress:
transport modeling in a deforming food material and special cases rr ¼0 ð7Þ
are described.
which implies divergence of effective stress is equal to gradient of
pore pressure:
2.1. Assumptions
r  r 0 ¼ r pf ð8Þ
(1) Food is treated as a multiphase porous material, in which In case of two-phase flow, when the pores are occupied by liquid
all the phases are in continuum. (2) Local thermal equilibrium is water and gas (comprising air and water vapor), the pore
assumed, i.e., temperature is shared by all the phases. Also, pressure, pf, can also be written as pg Sw pc . Inserting this
pressure in the liquid water phase is given as the gas pressure relationship in the solid momentum balance (Eq. (8)), we obtain
minus the capillary pressure (or the water potential). (3) The solid
skeleton is an incompressible hyperelastic material. Solid volume r  r 0 ¼ rpg rðSw pc Þ ð9Þ
remains constant during any process. Biological materials exhibit where the first term on the right-hand side is the gas pressure
non-linear stress–strain behavior, often following rubber and gradient, and the second term is a function of the temperature and
polymers (Ogden, 1972). For rubber, complex stress–strain beha- moisture content of the food material. Gas pressure gradients are
viors are accommodated using strain energy density functions. significant either for processes involving intensive internal vapor-
Neo-Hookean model is used in the present study, since this model ization such as microwave heating (Ni et al., 1999) or for processes
has been found to fit experimental data for rubber-like materials involving gas generation reactions such as carbon dioxide in bread
for large (30–70%) strains with sufficient accuracy. baking (Zhang et al., 2005). For most other processes, such as drying
and rehydration (soaking), gas is at atmospheric pressure and, thus,
2.2. Deformation of the solid matrix: model development for a the solid momentum balance reduces to
general case
r  r 0 ¼ rðSw pc Þ ð10Þ
Macroscopic total stress tensor, r , at any given location in a In Eq. (10), capillary pressure, pc, has a physical meaning only when
food material can be defined as volumetric average of total stress capillary suction is the only attractive force between the solid
6484 A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497

Volume
surface and the liquid water. In the presence of other attractive
forces like monolayer surface adsorption, multilayer absorption, etc.,
water potential, Cw , is a more appropriate term. Kelvin’s law is
usually applied to relate water potential, Cw (expressed in units of
pressure) to water activity, aw (Lu and Likos, 2004):
RT Gradual Ideal
Cw ¼ lnðaw Þ ð11Þ Transition Transition
vw in foods
After replacing pc by Cw , Eq. (10) can be used for liquid water in
the presence of multiple attractive forces. On the other hand, some Critical
high moisture food materials (with water activity, aw  1), which Volume, V*
Mainly Vapor Liquid + Vapor Mainly Liquid
undergo a change in their capacity to hold water with temperature Transport Transport Transport
rise, require a different approach for estimation of pressure in liquid
water. van der Sman (2007) applied Flory–Rehner theory to estimate
swelling pressure (equal to pore pressure in the absence of gas Moisture Content
phase) for such materials (more in Section 3.1.3).
Fig. 1. Volume change vs. moisture content curve for a typical food material.

2.3. Deformation of the solid matrix: special cases


V0, to final volume at which shrinkage stops, Vn, we obtain
Usual factors that lead to deformation in food materials are Z Vn
K
moisture change (examples include drying and rehydration) and dV ¼ pnc ð15Þ
V0 V
internal pressure generation (examples include puffing and bread
baking). Between the two, deformation due to moisture change is For a simple material with uniform pore size and a known bulk
a complex phenomena and is highly dependent on the state of the modulus-moisture content relationship (hardening of the mate-
food material. The physics of deformation due to gas transport is rial with moisture loss), an explicit value for critical volume, Vn,
relatively easy as the effect of gas pressure can be easily can be established from Eq. (15). However, due to the highly
expressed as a source term in the solid momentum balance (see heterogeneous and hygroscopic nature of food material, we can
Section 2.3.2). only say that K and pc are functions of moisture content, M, and
temperature, T. Thus, critical volume, Vn, will also be a function of
temperature and moisture at equilibrium:
2.3.1. Processes with moisture change as the driving mechanism
Most wet food materials are initially in a soft rubbery state. For V n ¼ V n ðM,TÞ ð16Þ
such materials, it is usually observed that total volume change at
equilibrium is equal to volume of moisture lost or gained Also, for a general food material with range of pore sizes, the
(Achanta, 1995). In other words, as long as the material is in a capillaries will empty at different values of shrinkage. Thus, as
rubbery state and the drying rate is not too high to cause surface shown in Fig. 1, in food materials, we may observe a gradual
cracks, the solid matrix remains saturated and the gas phase does decrease (rather than a sharp change which is expected for
not enter the pores. In such a case, the pore pressure is simply the uniform pore size material) in the slope of volume vs. moisture
pressure of liquid water, and Eq. (8) can be written as content plot to zero. Fortunately, volume vs. moisture content
data is available for many food materials from experiments and
r  r 0 ¼ r pw ð12Þ can be used to estimate free strain due to moisture loss, eM , and
other strain measures such as deformation gradient tensor due to
In a series of papers, moisture transport has been investigated in
detail by Scherer (1989), Smith et al. (1995) for soft and deform- moisture loss, FM. Volume change due to moisture loss can then
be treated as free strain analogous to thermal expansion
ing polymer gels, which behave in a similar fashion. Scherer
argued that for a uniform pore size medium with inert liquids in (discussed below for both small deformation and large deforma-
tion cases).
its pores, effective stress at equilibrium (or during a slow drying
process) is equal to pore pressure: Small deformation: For small deformation, volume changes due
to temperature and moisture change, i.e., the moisture and
r 0 ¼ pw ð13Þ thermal strains (eM and eT , respectively) are subtracted from the
total strain to get the mechanical strain, em :
As a soft material dries out, two important phenomena happen:
the pores shrink and the bulk modulus of the material increases, em ¼ eeM eT ð17Þ
turning a soft, rubbery food into a rigid, glassy state. For uniform
moisture distribution, the volume change is equal to the volume Now, with the effect of liquid (moisture) pressure accounted for
of water lost. The material will stop shrinking when the liquid– as a free strain, the mechanical strain, em , can be related to the
vapor meniscus moves inside the matrix and, with the increased stress due to mechanical load only, r 00 , i.e., the effective stress, r 0 ,
bulk modulus, the solid stresses can balance the compressive minus the pressure of water, pw:
capillary pressure, pnc . Until that point, the solid skeleton is too ðr 0 pw Þ ¼ r 00 ¼ D  em ð18Þ
soft to allow the meniscus to move inside and create compressive
pressure. Assuming the solid skeleton to be elastic, the normal The solid momentum balance, Eq. (12), can also be written in
effective stress (shear stress will be zero at equilibrium as there terms of r 00 :
are no pressure gradient or external shear load) can be related to
r  r 00 ¼ 0 ð19Þ
volume change (Scherer, 1989):
Depending on the time scales of the process and deformation, the
dr 0 ¼ K dV=V ð14Þ
food material can be treated as elastic or viscoelastic and the
Inserting the stress–strain relation from Eq. (14) into differential corresponding stress–strain relationship can be used along with
form of Eq. (13) and integrating from initial stress-free volume, the solid momentum equation.
A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497 6485

2.4. Heat and moisture transport: model development for a general


case
T, M, σ' Fel T, M, σ'= 0
Transport modeling for food processes using the multiphase
porous media approach has been reviewed elsewhere (Datta, 2007).
In this section, only equations relevant to deformable materials are
summarized and the reader should refer to Datta (2007) for details for
F FM rigid materials.

2.4.1. Governing equations


The governing equations for non-isothermal transport of two-
phases (liquid water and gas) in an unsaturated porous medium
are comprised of energy conservation and mass conservation of
gas phase, water vapor and liquid water phase, respectively:
FT @T X
To, Mo, σ'= 0 T, Mo, σ'= 0 ðreff cp,eff Þ þ n i,G  rðcp,i TÞÞ ¼ r  ðkeff rTÞlI_
ð~ ð24Þ
@t

Fig. 2. Steps indicating multiplicative split in the deformation tensor, separating @cg
moisture, temperature and mechanical effects.
n g,G Þ ¼ I_
þ r  ð~ ð25Þ
@t

@ðcg ov Þ
Large deformation: For large deformation analysis, a multi- n v,G Þ ¼ I_
þ r  ð~ ð26Þ
@t
plicative split (Vujosevic and Lubarda, 2002) in deformation
gradient tensor, F, can be used to separate volume changes due @cw
n w,G Þ ¼ I_
þ r  ð~ ð27Þ
to moisture and temperature changes from volume change due to @t
mechanical effects. As shown in Fig. 2, the material is first The energy equation is used to solve for temperature and the
assumed to go under stress-free deformations due to moisture mass conservation equations for their respective concentrations.
and temperature changes and, then, mechanical stresses act on The gas concentration, cg, is related to pressure by invoking the
this stress-free deformed material. The deformation tensor, F, can ideal gas law. Note that not all four equations are needed for all
be split as processes (Fig. 3). Just as the energy equation is needed only for
F ¼ FT FM Fel ð20Þ non-isothermal processes, the gas phase equation is solved only
in case of significant internal pressure generation when pressure
The dilatation (volume-changing) stress is related to elastic driven flow and/or deformation due to gas pressure gradients
Jacobian, Jel ¼ detðFel Þ, which is obtained as the ratio of total becomes important. Also, the vapor equation is rarely required as
volume change and volume change due to moisture and tem- vapor can be assumed to be at equilibrium with the liquid
perature effects (details in Section 3.1.2). Thermal Jacobian, moisture (more later).
JT ¼ detðFT Þ is often small for food materials and is usually In a deforming medium, since the solid has a finite velocity,
ignored. Moisture Jacobian, JM, can easily be obtained from ~
v s,G , the mass flux of a species, i, with respect to stationary
volume change vs. moisture content relationship (Fig. 1). observer, ~ n i,G , (used in Eqs. (24)–(27)) can be written as sum of
flux with respect to solid and flux due to movement of solid:
2.3.2. Processes with gas pressure as the driving mechanism ~
n i,G ¼ ~
n i,s þci~
v s,G ð28Þ
For some processes, such as microwave heating or bread-
baking, large internal pressure generation (due to water vapor in
microwave heating and carbon dioxide in baking) can cause
swelling/puffing of the material. In such cases, the gas pressure 2.4.2. Mass fluxes
gradient term of Eq. (9) (first term on the right-hand side) may Mass fluxes in an unsaturated porous medium can be attrib-
dominate. Swelling due to gas pressure in such cases can be much uted to two primary mechanisms—convection (for both gases and
larger than shrinkage due to moisture loss, and, therefore, stresses liquids) and binary diffusion (between vapor and air). Reynolds
and strains due to the latter can be ignored. In the absence of number is very low (usually less than one) for transport in food
thermal strains, the total strain is approximately equal to the materials and, therefore, Darcy’s law is applied to determine
mechanical strain: convective fluxes. For binary diffusion between vapor and air in
em  e ð21Þ the gas phase, Fick’s law is used:

Also, as the stress due to moisture transport is neglected, the solid kg


n g,s ¼ rg
~ ðrpg rg ~
gÞ ð29Þ
momentum balance (Fig. 9) becomes mg
r  r 0 ¼ r pg ð22Þ !
kg c2
n v,s ¼ rv
~ ðrpg rg ~
g Þ Mv Ma Dbin rxv ð30Þ
with effective stress, r 0 related to strain, e. mg rg
Of course, if deformation due to both phenomena (moisture
change and gas pressure) need to be accounted for, the governing kw
n w,s ¼ rw
~ ðrpw rw~

equation and the constitutive law will take the form mw
kw
r  r 00 ¼ rpg ¼ rw ðrðpg pc ðM,TÞÞrw ~

mw
   
kw @p @p
¼ rw r pg  c rM c rT rw~ g ð31Þ
r 00 ¼ D  em ð23Þ mw @M @T
6486 A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497

Fig. 3. A framework for modeling of transport and deformation in food materials based on the state of the material and its processing conditions.

Dry basis moisture content, M, is defined as than other relevant time scales for the process), vapor pressure
cw cw becomes a function of moisture and temperature (through Clau-
M¼ ¼ ð32Þ sius–Clapeyron equation and moisture sorption isotherms) and its
cs ð1fÞrs
conservation equation does not need to be solved. In such cases,
Taking density of solid, rs , as constant, moisture content, M, can vapor flux (ignoring gravity) can be written as
be expressed as M ¼ Mðcw , fÞ (where cw ¼ rw ew ), and Eq. (31) can !
be re-written as kg c2
    n v,s ¼ rv
~ r pg  Mv Ma Dbin rðpv ðMðcw , fÞ,TÞ=pg Þ
kw @p @M @p @M @p mg rg
n w,s ¼ rw
~ r pg  c rcw  c rf c rT rw~ g ! !
mw @M @cw @M @f @T
kg c2 pv
¼  rv þ Mv Ma Dbin 2 rpg Dv,cw rcw
kw mg rg pg
¼ rw ðrpg rw ~
g ÞDw,cw rcw Dw, f rfDw,T rT ð33Þ
mw
Dv, f rfDv,T rT ð35Þ
where diffusivity due to moisture gradient, Dw,cw , diffusivity due
to porosity gradient, Dw, f , and diffusivity due to temperature where vapor diffusivity due to moisture gradient, Dv,cw , vapor
gradient, Dw,T, are defined as diffusivity due to porosity gradient, Dv, f , and diffusivity due to
temperature gradient, Dv,T, are defined as
kw @pc @M !
Dw,cw ¼ rw
mw @M @cw c2 D @pv @M
Dv,cw ¼  Mv Ma bin
rg pg @M @cw
kw @pc @M
Dw, f ¼ rw !
mw @M @f
c2 Dbin @pv @M
Dv, f ¼  Mv Ma
kw @pc rg pg @M @f
Dw,T ¼ rw ð34Þ
mw @T !
Eqs. (24)–(27), along with fluxes from Eqs. (29), (30) and (33), c2 Dbin @pv
Dv,T ¼  Mv Ma ð36Þ
solid velocity, ~
v s,G , from solid momentum balance and an explicit rg pg @T
expression for evaporation rate, I, _ complete the model develop-
Now, adding liquid water and water vapor conservation equations
ment. Estimation of evaporation rate, however, is not always easy _ and inserting flux relationships,
to eliminate evaporation rate, I,
(Halder et al., 2010) and an accurate determination of I_ is possible
we obtain the equation for overall moisture balance
only in some special situations, e.g., when local equilibrium
between liquid water and vapor can be assumed. Details of @
ðcw Þ þ r  ðcw~
v s,G Þ ¼ r  ðK1 rpg þ Dcw rcw þ Df rf þ DT rTÞ ð37Þ
estimation of I_ are mentioned elsewhere (Halder et al., 2010). @t
where
2.4.3. Overall moisture equation !
kw kg c2 pv
If water vapor can be assumed to be in equilibrium with liquid K1 ¼ rw rv  Mv Ma Dbin ð38Þ
water (i.e., time-scale required to achieve equilibrium is smaller
mw mg rg p2g
A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497 6487

Dcw ¼ Dw,cw þ Dv,cw ð39Þ 2.5.1. Wet-rubbery state: liquid moisture transport as the
dominating mechanism
In the rubbery state, free shrinkage/swelling compensates for
Df ¼ Dw, f þ Dv, f ð40Þ
moisture loss/gain which means, at equilibrium, change in
volume of a food material is equal to the volume of water lost/
DT ¼ Dw,T þ Dv,T ð41Þ
gained (Section 2.3.1). During rehydration/dehydration of such
are the effective permeability and the effective diffusivities due to materials, the evaporation front stays at the surface of the
moisture concentration gradient, porosity gradient and tempera- material and there is no vapor generation or transport within
the food. So, the evaporation rate, I,_ is equal to zero, there is no
ture gradient, respectively. In Eq. (37), it is assumed that water
vapor can contribute to transport terms but not to accumulation gas pressure gradient term in Eq. (37), and the effective diffusiv-
term (this is because density of vapor is three orders of magnitude ities reduce to just those of liquid moisture. Therefore, the model
smaller than density of liquid water). reduces to Eq. (37) for moisture and Eq. (43) for temperature (gas
For a majority of food processes, moisture fluxes due to phase and vapor equations are not required), with solid velocity,
temperature, porosity and pressure gradients are considered ~
n v,G , from the solid momentum balance:
small as compared to that for moisture gradients (sometimes @T
without justification). The conditions under which these assump- ðreff cp,eff Þ þ ð~
n w,G  rðcp,w TÞÞ ¼ r  ðkeff rTÞ ð43Þ
@t
tions can be justified are:
For soft materials, shear modulus is very small as compared to
the bulk modulus, which means shear stresses (for an uncon-
 Gas pressure is atmospheric ðrpg ¼ 0Þ.
strained material) that restrict free swelling/shrinkage are also
 The material is either saturated (f  cw =rw and the porosity
small, and volume change at every point in the material can be
gradient term can be merged with moisture gradient term) or
approximated by the free volume change, even under large
the material is rigid ðrf ¼ 0Þ.
moisture gradients. Thus, if the only deformation information
 Water activity (in turn, capillary pressure, pc) is independent of
required is volume change at every point and estimation of
temperature gradient (DT ¼0).
stresses and shear strains is not important, solid momentum
balance can be skipped. Divergence in solid velocity can be
In such cases, the overall moisture balance reduces to the well-
estimated from the solid mass balance (assuming constant and
known equation:
uniform solid density):
@cw
þ r  ðcw~
v s,G Þ ¼ r  ðDcw rcw Þ ð42Þ @ðrs es Þ
@t þ r  ðrs es vs,G Þ ¼ 0 ð44Þ
@t
After ignoring the flux due to solid velocity (again, usually done
without justification), Eq. (42) is extensively used in the food Ds es
þ es r  vs,G ¼ 0 ð45Þ
literature to model drying-like processes. Its great advantage lies Dt
in the fact that rate of evaporation, I,_ is not required. Also,
1 Ds e s 1 Ds e w
effective diffusivity, Dcw , can be easily estimated by fitting r  vs,G ¼  ¼ ð46Þ
es Dt 1ew Dt
experimentally observed drying curves. However, the rate of
evaporation may be required to solve Eqs. (24) and (25) (if where Ds =Dt stands for material derivative in the reference frame
pressure gradients are significant). of the solid. Divergence of solid velocity, vs,G, from Eq. (46) can
now be inserted in the liquid water and energy equations.

2.5. Heat and moisture transport: special cases


2.5.2. Almost-dry, glassy state: vapor transport as the dominating
mechanism
As discussed in case of deformation, transport models can also
Food at very low moisture content exists in a rigid-glassy state.
be simplified. Energy and gas phase equations are only required
As discussed earlier in deformation analysis, there is no deforma-
when temperature and pressure gradients, respectively, are
tion below a certain moisture content. The material can be
significant. In the following sections, simplifications based on
assumed to be rigid and deformation analysis is not required.
the state of a food material, as illustrated in Fig. 3, are discussed.
Also, the food material can be highly unsaturated at low moisture
Two extreme states of a food material are: (1) wet, rubbery state
contents, which means the permeability of liquid water, kw, can
(above glass-transition temperature); and (2) almost-dry, glassy
become very low, while the binary diffusivity of vapor and air,
state (below glass-transition temperature). In the intermediate
Dbin, can be very high. In such conditions, the transport can be
region, near glass transition, moisture transport may exhibit non-
dominated by vapor transport terms, i.e., Dw,cw 5 Dv,cw ,Dw,T 5 Dv,T ,
Fickian behavior (Case-II diffusion). Traditional form of Darcy’s
and transport in liquid phase can be ignored. From Eq. (27),
law (which assumes that the flux is proportional to pressure
ignoring transport terms we get
gradients) breaks down for such regions and needs to be mod-
ified. Various approaches have been explored (especially in the @cw
I_ ¼  ð47Þ
polymer science literature) to account for non-Fickian or Case II @t
diffusion. The most fundamental of these approaches is developed Also, solid velocity terms in all transport equations go to zero and
by Cushman and coworkers (Singh, 2002; Achanta, 1995) to the diffusivities in Eq. (37) are those of water vapor. The model
derive modified constitutive equations such as Darcy’s law, Fick’s (for processes in which transport due to temperature and pres-
law, and solid stress–strain relationship based on nonequilibrium sure gradients is small) reduces to Eq. (42) (for moisture) and
thermodynamics. The approach Cushman and coworkers Eq. (48) (for temperature)
followed, known as Hybrid Mixture Theory, is described in detail
elsewhere (Cushman, 1997), and not discussed further in this @T @cw
ðreff cp,eff Þ ¼ r  ðkeff rTÞ þ l ð48Þ
manuscript. We now discuss simplifications in governing equa- @t @t
tions of transport based on the state (rubbery or glassy) of a food This assumption of neglecting liquid transport terms is, however,
material. justified only when the material is very dry and may happen only
6488 A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497

Free surfaces
No evaporation (Natural convection
or drip loss heat transfer)

Hamburger Patty Axial


Symmetry

1.8 cm
10 cm Drip loss only

Evaporation and
Heated Plate No axial displacement drip losses
(Heat transfer coefficient)
Schematic of the Simulated geometry
contact-heating process (showing deformation)

Free surfaces
Symmetry
Forced convection heat transfer
Plane
Surface evaporation moisture loss

2 cm
x cm

x = 0.4, 0.7, 1.0

No axial displacement
Insulated for energy & moisture
equations
Cross-section of potato slab Simulated geometry
(perpendicular to length) (showing deformation)

Fig. 4. Schematic of the two processes simulated: (a) single-sided contact heating of hamburger patties, and (b) drying of potato slabs, showing the modeled geometry and
boundary conditions. Input parameters are listed in Tables 1 and 2.

for a small range of moisture content such as during rehydration 3.00


of dry cereals due to high humidity levels.
Moisture Content, dry basis

2.50
3. Model implementation and validation

In the following section, the modeling framework developed is


applied to two food processes: single-sided contact heating of a 2.00
hamburger patty and hot-air drying of a potato slab (Fig. 4) to
predict deformation, mass and energy transport kinetics. Ham-
burger patty cooking is selected as an example of single phase 1.50
(liquid water only) transport as the patty remains largely rubbery
throughout the cooking the process. On the other hand, potato
drying involves development of air porosity and two-phase 1.00
(liquid water and water vapor) transport as the potato undergoes 20 40 60 80 100
transition from a soft and rubbery to rigid-glassy state during Temperature, °C
drying. In each case, the model predictions are validated using
Fig. 5. Water holding capacity (WHC) Dhall and Datta (submitted for publication)
experimental results.
in terms of moisture content (dry basis, kg water per kg of dry solids) as a function
of temperature showing a large drop in WHC near 60 1C. The error bars are for
3.1. Contact heating of a hamburger patty standard error for measurements done on three patties.

Meat can be processed and cooked in a variety of ways. For the temperature of 140 1C. As temperature rises, water at the surface
purpose of this study, single-sided contact heating of hamburger of the patty evaporates. Since ground meat is in a rubbery state,
patties (Fig. 4) bought from a local grocery store (USDA Nutrition the patty shrinks with loss of moisture, and, at equilibrium (in the
Database, 2010, entry no. 23557, 95% lean and 5% fat) is selected. absence of gradients of any temperature and moisture fields) the
A refrigerated hamburger patty of cylindrical shape (diameter shrinkage should be equal to the volume of water lost (Fig. 1).
10 cm and height 1.8 cm), initially stored at 5 1C, is heated on a With further rise in temperature, denaturation of muscle proteins
commercial griddle (HotZoneTM Griddle Model No. GR0215G, occurs, which leads to decrease in water holding capacity of the
Applica Consumer Products Inc., Miramar, Florida) at a fixed plate meat. Since the surface of meat in contact with the griddle gets
A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497 6489

heated up quickly, the water holding capacity near the surface Deviatoric part of elastic deformation gradient is related to elastic
1=3
may drop faster as compared to the drop in moisture concentra- deformation gradient, Fel, and its dilatation part, Jel , as
tion due to evaporation. This leads to dripping of water from the 1=3
Fel ¼ Jel Fel ð54Þ
patty. The variables of interest for predicting quality and safety
aspects of meat cooking are temperature, moisture content, Now, to estimate elastic Jacobian, Jel, we need to calculate
shrinkage, evaporation rate and their histories. Jacobian due to moisture change, JM (Eq. (20)). This is easy, as
under stress-free conditions, a patty shrinks/swells by the amount
3.1.1. Problem details of moisture lost/gained. Let V be the REV volume at moisture
The patty is simulated as a 2D axisymmetric geometry, as the volume fraction, ew . Then, change in volume of REV can be
exchange of heat and mass with the outside environment does equated to change in volume of moisture in REV:
not have angular dependence and only a cross-section of the VV0 ¼ ew Vew,0 V0 ð55Þ
cylindrical patty needs to be simulated. The effect of gravity on
mass transfer is ignored as the effect of pressure gradients is V 1ew,0
JM ¼ ¼ ð56Þ
much larger on moisture velocity. Since the patty is in a soft and V0 1ew
rubbery state, evaporation stays on the surface during the entire Similarly, porosity at any time t, fðtÞ, can be determined using
cooking process. Even if a rigid glassy region develops at the incompressibility of the solid skeleton, equating the initial
heated surface, it is assumed to be small and its effect can be volume of solid in an REV to solid volume at time, t:
neglected. Therefore, according to the modeling framework
outlined in Fig. 3, the rubbery state of food can be selected. Also, ð1fðtÞÞVðtÞ ¼ ð1f0 ÞV0 ð57Þ
as the temperature gradients are significant, the energy equation
1f0 1f0
needs to be solved along with the moisture transport and solid fðtÞ ¼ 1 ¼ 1 ð58Þ
VðtÞ=V0 JðtÞ
momentum balance equations. Since there is no internal gas
pressure generation, vapor and gas equations are not required. Note that while Jacobian due to moisture change, JM, is a state
function (depending on the moisture content), porosity, fðtÞ, is a
3.1.2. Solid momentum balance process variable, depending on the actual Jacobian, J(t).
A patty can shrink by 30% or more of its initial volume during
the contact heating process, which necessitates the use of large 3.1.3. Moisture and energy transport equations
deformation analysis for solid deformation. Since the evaporation Moisture flux in case of meat needs to be treated differently
front stays at the surface and there is no internal gas pressure from the discussion in Section 2. Water activity of meat at room
generation, gas pressure gradient term can be ignored for the temperature is  1, which gives capillary pressure, pc, or water
solid momentum balance. For large deformation, Lagrangian potential, Cw , close to zero (using Kelvin’s law (Lu and Likos,
measures of stress and strain are used, and the solid momentum 2004). Thus, Eq. (31) cannot be used to calculate moisture flux.
balance (Eq. (19)) is written in Lagrangian coordinates: Also, with increase in temperature, meat proteins denaturate
leading to a drop in water holding capacity (Tornberg, 2005). As
rX  ðS00  FTel Þ ¼ 0 ð49Þ time scales of temperature rise in the patty during intensive
00
where S is the second Piola–Kirchhoff (PK2) stress tensor, and Fel cooking such as contact-heating are smaller than time scales of
is the elastic deformation gradient tensor. PK2 stress, S00 , is related moisture transport, moisture concentration in much of the patty
to Cauchy stress, r00 , by the following relationship: is more than its water holding capacity at equilibrium.
Liquid water pressure (called swelling pressure) in meat has
S00 ¼ J  F1 00 T
el  r  Fel ð50Þ been estimated (van der Sman, 2007) by using the Flory–Rehner
PK2 stress is energy conjugate to the Green–Lagrange elastic theory. Taking the swelling pressure to be zero at equilibrium
strain tensor, Eel: moisture volume fraction, and linearizing the Flory–Rehner
expression near equilibrium, it can be shown that the swelling
Eel ¼ 12ðFTel Fel IÞ ð51Þ pressure is proportional to the difference between the actual and
00
and, thus, S and, Eel are related as follows: equilibrium moisture concentrations:

@Wel pw ¼ Cðcw cw,eq ðTÞÞ ð59Þ


S00 ¼ ð52Þ
@Eel where cw,eq is the equilibrium moisture concentration at a given
Now, we need a constitutive equation for the elastic strain energy temperature and the constant of proportionality, C, though
density, Wel. Rubbery state means the stress relaxation time constant here, can be temperature dependent. Inserting this
scales are expected to be small (as compared to the time scale expression of liquid pressure, pw, in Darcy’s law (line 1 in
of the cooking process which is in minutes, Deborah number  0) Eq. (31)), we get (ignoring gravity)
and the solid skeleton can be treated as a hyperelastic material. ~
n w,s ¼ ðDw,cw rcw þ Dw,T rTÞ ð60Þ
Also, the fibers in ground meat are randomly oriented. Therefore,
although meat fibers are anisotropic with different properties where the new definitions of diffusivities due to moisture
along and across the fibers, the averaged mechanical properties gradient and temperature gradient are:
are isotropic. A modified Neo-Hookean constitutive model is kw
Dw,cw ¼ rw C
chosen which accounts for the volume change due to moisture mw
loss also
kw @cw,eq
K m Dw,T ¼ rw C ð61Þ
Wel ¼ ðJ 1Þ2  ðI 1 3Þ ð53Þ mw @T
2 el 2
where K and m are the bulk modulus and the shear modulus, Thus, the moisture transport equation reduces to Eq. (62) with
respectively. Jel is the elastic Jacobian as defined earlier, and I is new definitions of diffusivity (Eq. (61)):
T
the first invariant of the right-Cauchy Green tensor, Cð ¼ Fel Fel Þ, @cw
þ r  ðcw~
v s,G Þ ¼ r  ðDw,cw rcw þ Dw,T rTÞ ð62Þ
for deviatoric part of elastic deformation gradient, i.e., Fel . @t
6490 A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497

Table 1
Input parameters Dhall (2011) used in the simulations of single-sided contact heating of hamburger patties. Number under source column refer to bibliographic order.

Parameter Value Units Source

2D axisymmetric patty dimensions


Height 1.8 cm Measured
Diameter 10 cm Measured
Patty composition Actual (used) Weight USDA Nutrition Database (2010)
Water 73.28 (72.74) %
Protein 21.41 (21.25) %
Fat 5.00 (4.96) %
Ash 1.05 (1.04) %
Initial conditions
Air volume fraction 0 –
Temperature 5 1C Measured
Processing conditions
Ambient temperature 60 1C Measured
Plate temperature 120 1C Measured
Heat transfer coefficient 400 W/m2 K Wang and Singh (2004)
Mass transfer coefficient 0.01 m/s Ni and Datta (1999)
Properties
Water holding capacity Fig. 5 – Measured
Density Choi and Okos (1986)
3
Water 997.2 kg/m
Fat 925.6 kg/m3
Protein 1330 kg/m3
Specific heat capacity Choi and Okos (1986)
Water 4178 J/kg K
Fat 1984 J/kg K
Protein 2008 J/kg K
Thermal conductivity Choi and Okos (1986)
Water 0.57 W/m K
Fat 0.18 W/m K
Protein 0.18 W/m K
Diffusivity 10  7 m2/s van der Sman (2007)
Bulk modulus Kw Pa Hashin (1985)
ew
Poisson’s ratio 0.49 – Rubbery state

The energy balance equation remains the same as discussed for In Eq. (66), the first term on the right hand side is the convective
rubbery materials in Section 2 (Eq. (43)). heat transfer coefficient multiplied by the temperature difference,
the second term is the latent heat taken up by surface evapora-
3.1.4. Boundary and initial conditions tion, and the third term is energy carried by convection terms
Solid momentum balance: Normal displacement of the axisym- normal to the boundary.
metric boundary and the bottom surface (lying on the griddle) is Initial conditions: Initially refrigerated at 5 1C, the composition
set to zero. The other two boundaries are unconstrained and free of the patty is taken from USDA Nutrition Database (2010) and is
to move (Fig. 4). listed in Table 1. Since the weight percentages of the proximates
Liquid water equation: The boundary condition for liquid water added up to 100.74, the weight percentages were normalized. The
equation consists of two flux terms: evaporation and drip. The volume fraction of air in the patties is considered small and, thus,
magnitude of the evaporation flux, nw,s,surfe , is simply given by ignored. From this data, the initial concentrations of water and
mass transfer coefficient multiplied by the vapor density differ- solid (protein, fat and ash) can be calculated.
ence between the surface and the boundary:
nw,s,surfe ¼ hm ðrv,surf rv,amb Þ ð63Þ
3.1.5. Input parameters and numerical solution
Water is lost from the matrix in liquid form (as drip) only when Input parameters used in the hamburger patty cooking simu-
surface moisture concentration, cw,surf , is more than the water lation are given in Table 1. Bulk modulus and Poisson’s ratio were
holding capacity, cw,eq . The drip loss, nw,s,surfd , under such condi- estimated considering the patty to be saturated and in a soft,
tions is equal to the total moisture flux reaching the surface rubbery state throughout the heating duration. In a soft material,
subtracted by that taken by surface evaporation, nw,s,surfe : the Poisson’s ratio is expected to be about 0.5. A value of 0.49 was
~ surf hm ðr used to help convergence. For saturated porous materials with
nw,s,surfd ¼ ~
n w,s  N v,surf rv,amb Þ ð64Þ
incompressible solid skeleton, the bulk modulus (for small elastic
Therefore, the total moisture flux at the surface with respect to a strains) is given by Hashin (1985)
stationary observer is equal to the sum of drip loss, evaporation  
1 4Gs
loss and flux due to movement of the surface itself: K¼ Kf þ ð1ef Þ ð67Þ
ef 3
nw,G,surf ¼ nw,s,surfe þnw,s,surfd þcw~ ~
v s,G  N ð65Þ
Since the bulk modulus of water, Kw (2.2  109 Pa) is much greater
Energy equation: For energy equation, forced convection heat than the shear modulus of the solid matrix, Gs ( o 106 Pa), it
transfer boundary condition is applied to get the heat flux at the justifies a Poisson ratio close to 0.5, and Eq. (67) reduces to
surface, qsurf :
X Kw
qsurf ¼ hðTamb Tsurf Þlnw,s,surf  ð~ ~ surf
n i,G cp,i TÞ  N ð66Þ K¼ ð68Þ
ew
A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497 6491

which is noted in Table 1. Heat transfer coefficient is estimated Moisture content


from the experimentally measured data of Wang and Singh (dry basis)
(2004). A commercially available finite element software,
COMSOL Multiphysics 3.5a (Comsol Inc, Burlington, MA), was
used to solve the equations. The solid momentum balance is
solved in the total Lagrangian reference frame (i.e., frame moving min
15
with the solid) for the axisymmetric geometry equation in the
structural mechanics module, while convection–conduction and
convection–diffusion equations (in the main COMSOL Multiphy-
sics module) were used for energy and moisture transport,
respectively. Deformed mesh equations (again, in the main
COMSOL Multiphysics module) were used to track the material
deformation in the Eulerian reference frame, and move the mesh
accordingly. The transport equations were solved in the Eulerian
reference frame (i.e., frame of the stationary observer) on the
deformed mesh. The computational domain was rectangular,
5 cm  1.8 cm, and had an unstructured quadrilateral mesh
consisting of 3864 elements. Linear shape functions were used. Unheated
The simulation of 900 s of heating took approximately 4 h of CPU Surface
time for an adaptive timestepping scheme (maximum time step
size of 0.05 s) on a 3.00 GHz dual-core Intel Xeon workstation
with 16 GB RAM. Mesh and timestep convergence were ensured in
3m
by checking that any dependent variable (temperature, moisture
content or displacement) did not change by more than 1% of the Axis of
total change (at any time at all four vertices of the geometry) by symmetry Heated
reducing the timestepsize or mesh-size by half. Surface

Fig. 7. Contours of moisture content (dry basis) after 3, 6, 9, 12 and 15 min of


single-sided contact heating of hamburger patties showing low moisture at the
3.1.6. Results and discussion heated surface and some accumulation in the center. Moisture gradients can be
Spatial and temporal distribution of moisture content: Fig. 6 seen primarily in the axial direction.
shows a comparison between predicted and experimentally
observed (Dhall, 2011) total moisture loss history of the patty
for 15 min of heating time. Total moisture loss is almost linear loss with time is slightly concave upwards (rate of loss always
with time, with the patty losing about 17% (26 g for a 155 g patty) increases throughout the heating duration). On the other hand,
of the initial moisture content in 15 min. The predicted moisture cumulative drip loss curve with time is S-shaped and stabilizes
loss history follows the observed history closely, and the differ- (rate of drip loss goes to zero) at around 5 min as moisture
ence between the two at any time is 5% or less. The cumulative concentration at the patty surface falls below equilibrium
evaporation and drip losses are also plotted in Fig. 6. Evaporation concentration. Evaporation loss and its rate exceed the drip loss
and the drip loss rate at any time during heating. Contours of
0.18 moisture content (dry basis) after every 3 min of heating (starting
at 3 min) are plotted in Fig. 7. It can be seen that the moisture
0.16 gradients dominate in the axial direction and end effects are
Moisture loss (fraction of initial patty mass)

Predicted restricted to a small region near the lateral surface of the patty.
0.14 Also, even at the end of heating, the minimum moisture content
(near the griddle plate) is still high (0.891), which means the
surface has not dried up. On the other hand, moisture content
0.12 Experiment
close to the exposed top surface (away from the griddle) rises to
2.731 (from an initial value of 2.6) during the process.
0.1
Spatial and temporal distribution of temperature: Fig. 8 shows a
Evaporation loss comparison between predicted and experimentally observed
0.08
temperature histories at two locations on the central axis of the
patty: (1) at the mid-point between the heated and exposed
0.06 surfaces, and (2) on the exposed top surface. With the initial lead
time of about 50 s, temperature at the midpoint follows the
0.04 concave downwards curve reaching a value of 56 1C after
15 min. The predicted curve follows the observed one closely,
0.02 Drip loss with the difference between the two at any time being 1 1C or
less. Temperature history at the surface is more interesting. While
0 the observed history is similar to that of the midpoint, having an
0 3 6 9 12 15 initial lead time followed by a concave downwards curve; the
Time (min) predicted history shows a quick initial heating period which is
absent in the observed history. The discrepancy between the
Fig. 6. Cumulative total (evaporation þ drip) moisture loss (predicted and experi- predicted and observed histories for the first 300 s of heating can
mentally observed, Dhall and Datta, submitted for publication), evaporation
moisture loss (predicted) and drip loss (predicted) for single-sided contact heating
be attributed to changing ambient conditions of temperature and
of hamburger patties. It can be seen that drip loss levels off after 5 min and relative humidity at the exposed surface during the cooking
evaporation loss dominates for the rest of the heating duration. process. At the top surface, a fixed ambient air temperature of
6492 A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497

Spatial and temporal distribution of deformation field: Fig. 10


Center
60 compares the histories of experimentally observed diameter with
(Predicted)
the predicted diameter (averaged for diameter at different
heights). The patty diameter reduces to about 91% of the original
50 Center value in 15 min, which is as predicted by the simulations. For
(Exp.) reference, the diameter, D(t), assuming uniform shrinkage
throughout the patty, is computed from the equation:
Temperature (ºC)

40
 
DðtÞ VðtÞ 1=3
¼ ð69Þ
D0 V0
30 Surface
(Predicted) and also plotted. Eq. (69) is obtained by assuming the same
Surface uniform linear shrinkage along the thickness and in the diameter
20 (Exp.) and relating this to volume shrinkage, using the equation
ðV0 VðtÞÞ=V0 ¼ ðpD20 L0 pD2 ðtÞLðtÞÞ=pD20 L0 Þ with DðtÞ=D0 ¼ LðtÞ=L0 .
Surface As shown in Fig. 10, the diameter assuming uniform shrinkage
10 is much larger than the predicted or observed diameters at any
Center
time, indicating the non-uniformity in patty shrinkage. Also, this
means that such a simplified relationship as Eq. (69) cannot be
0
used to predict diameter with solid deformation equations not
0 3 6 9 12 15
solved. Predicted thickness (normalized) and thickness assuming
Time (min)
uniform shrinkage are plotted in Fig. 11. The final value of
Fig. 8. Temperature histories (predicted and experimentally observed, Dhall and thickness is approximately 95% of the initial value, which means
Datta, submitted for publication) at the midpoint and the surface on the central the patty shrinks by less than 1 mm in thickness in 15 min.
axis for single-sided contact heating of hamburger patties. Predicted values of thickness were not compared to its
observed values because of high variability in patty thickness (it
varied by more than 2 mm at different locations on a single patty)
Temperature
and also due to variability in shear effects that cause rise of the
(°C)
bottom surface of the patty near the center. In this simulation, the
bottom surface was considered fixed in the z-direction that could
not be achieved in all the experiments at all times. Some patties
rose by 1–2 mm in the middle, while some others stuck to the
min
15 griddle plate. Therefore, uncertainty (more than 2 mm) in height
was more than the total expected change in height (  1 mm) and,
thus, it was meaningless to compare the observed and predicted
thickness values.
Fig. 12, which plots the contours of elastic Jacobian, Jel ð ¼ J=JM Þ,
at different times, helps us arrive at a very good (albeit, more
involved) method to predict shrinkage. Fig. 12 shows that the
ratio of actual Jacobian, J, to the Jacobian due to moisture change,

Unheated 1.01
Surface

0.99 Diameter (assuming uniform


in
3m shrinkage throughout the patty)
Diameter (normalized)

Axis of Heated 0.97


symmetry Surface

Fig. 9. Temperature contours (in 1C) after 3, 6, 9, 12 and 15 min of single-sided 0.95
Prediction
contact heating of hamburger patties showing constant heated surface tempera-
ture and gradients primarily in the axial direction.
0.93
Experiment
60 1C (Table 1) and zero moisture flux (moisture loss from the top
surface is negligible as compared to the moisture loss from the 0.91
bottom surface) were used as boundary conditions which may not
be valid at initial times when patty is just put on the plate.
Therefore, an error in surface temperature prediction may be seen 0.89
when the effect of boundary conditions dominates (in the model 0 3 6 9 12 15
the surface gets heated fast due to air temperature of 60 1C). Time (min)
Similar to moisture content, temperature contours (Fig. 9) show
Fig. 10. Diameter change histories (prediction and experimental observation) for
small end effects. The heated surface reaches around 90 1C early single-sided contact heating of hamburger patties. Also, diameter calculated
in the heating process and stabilizes. Temperature at the exposed assuming uniform shrinkage throughout the patty (Eq. (69)) is plotted showing
surface rises slowly and reaches about 50 1C after 15 min. assumption of uniform shrinkage will lead to erroneous results.
A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497 6493

1 only. Displacements due to this deformation field can now be


calculated and used in the deformed mesh equations to get new
geometry.

3.2. Convective drying of a potato slab


0.98 Prediction
Drying of potato slabs, as described by Wang and Brennan
Height (normalized)

(1992, 1995), is numerically implemented as a second example.


The potato slabs (Desiree variety) are 45 mm long and 20 mm
wide, with thickness varying from 4 to 10 mm. The drying
0.96
experiments were carried out by Wang and Brennan at air
Height (assuming uniform temperatures between 40 1C and 70 1C, at a constant absolute
shrinkage throughout the patty) humidity of 16 g (vapor)/kg (dry air). Initially, the potato slab is in
a rubbery state and shrinks with loss of moisture. However,
0.94 unlike meat, it becomes rigid towards the end of drying and stops
shrinking with moisture loss, allowing the evaporation front to
move in. As in meat cooking, the variables of interest are
temperature, moisture content, shrinkage, evaporation rate and
their histories.
0.92
0 3 6 9 12 15 3.2.1. Problem formulation and modeling details
Time (min) To reduce computational complexities, a 2D cross-section of
the potatoes (perpendicular to length) is modeled and the end-
Fig. 11. Height change history for single-sided contact heating of hamburger
patties. Note that height change was too small to be compared with experiments. effects are ignored (Fig. 4). Only half of the width is simulated as
Also, height calculated assuming uniform shrinkage throughout the patty is all the physics is symmetric about the center. Initially, the potato
plotted, showing assumption of uniform shrinkage will lead to erroneous results. is in a soft and rubbery state, and gradually transitions to a rigid
state. According to the modeling framework outlined in Fig. 3, the
Elastic Jacobian transition state of food can be selected as both rubbery and glassy
(Jel) states exist at different times and positions during potato drying.
Since this transition occurs at a very low moisture content and
there is no evidence of Case-II diffusion (as discussed in Section
2.5) in potatoes, the traditional constitutive relationship for
min moisture flux (Darcy’s law) holds. In this case, the energy balance
15
(Eq. (24)) is solved along with the moisture balance (Eq. (42)) and
solid momentum balance (Eq. (49)). Assuming equilibrium
between liquid water and water vapor, evaporation rate, I, _ is
estimated using Eq. (26). Also, reduction of volume with removal
of moisture stops at M ¼0.3, Jacobian due to moisture change, JM
is written as
1ew,0
JM ¼ M 4 0:3 ð70Þ
1ew

JM ¼ JM 9M ¼ 0:3 M r0:3 ð71Þ


Unheated
Surface For boundary condition of the solid momentum equation, the bottom
and the left edges are treated as a roller (zero normal displacement)
and a symmetry, respectively. The other two edges are free. The
in bottom and the left edges are insulated for energy and moisture
3m
transport equations, while surface evaporation and convective heat
Axis of and mass transport takes place at the other two edges. Thus, Eqs. (65)
symmetry Heated
and (66) (with no drip loss) are used as boundary conditions for
Surface
moisture and energy transport. Other input parameters used in the
Fig. 12. Elastic Jacobian, Jel, (ratio of actual volume to free volume) contours after simulation are listed in Table 2. The solution strategy remains the
3, 6, 9, 12 and 15 min of single-sided contact heating of hamburger patties. It can same, with the simulation of 1000 min of drying taking approxi-
be seen that the surface is stretched and the heated interior is compressed by a
mately 30 min of CPU time for a maximum timestep size of 60 s (784
maximum of 2% from free volume.
linear quadrilateral elements) on a 3.00 GHz dual-core Intel Xeon
workstation with 16 GB RAM.
JM, lies in the narrow range of 0.98–1.01%. The region near the
heated surface is under tension, while the other cooler regions are 3.2.2. Results and discussion
under compression. The narrow range of elastic Jacobian, Jel, is Figs. 13 and 14 compare model predictions with the experi-
due to the high bulk modulus to shear modulus ratio (Poisson mental observations: (a) temperature history at the top surface
ratio, n  0:5). For such cases, if estimation of stresses is not (spatially averaged) for drying a 7 mm thick slab at a drying
important, solid momentum balance can be avoided and Jacobian, temperature of 55 1C; (b) moisture content histories for slabs of
J, can be assumed to be equal to the moisture change Jacobian, JM. thickness 10 mm, 7 mm and 4 mm at drying temperature of
In the absence of significant shear strains, the dilatational strains 55 1C; and (c) normalized volume as a function of moisture
and, thus, deformation field, can be estimated from Jacobian, J, content for a 10 mm thick slab at drying temperatures of 70 1C
6494 A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497

Table 2
Input parameters used in the simulations of drying of potato slabs. Number under source column refer to bibliographic order.

Parameter Value Units Source

2D Slab dimensions
Height 4,7,10 mm Wang and Brennan (1992, 1995)
Half width 10 mm Wang and Brennan (1992, 1995)
Initial conditions
Moisture vol. frac.
0.865 (Tamb ¼ 55 1C) – Wang and Brennan (1992)
0.838 (Tamb ¼ 40,70 1C) – Wang and Brennan (1995)
Air vol. frac. 0 –
Temperature 32.5 1C Wang and Brennan (1992)
Drying conditions
Temperature 40, 55, 70 1C Wang and Brennan (1992),
Wang and Brennan (1995)
Absolute humidity 0.16 g/kg Wang and Brennan (1992)
Heat transfer coeff. 40 W/m2K Laminar flow
Mass transfer coeff. 0.01 m/s Lewis analogy
Properties
Water activity – – Ratti et al. (1989)
Density Choi and Okos (1986)
3
Water 998 kg/m
Air Ideal gas kg/m3
Solid 1592 kg/m3
Specific heat capacity Choi and Okos (1986)
Water 4178 J/kg K
Solid 1650 J/kg K
Thermal conductivity Choi and Okos (1986)
Water 0.57 W/m K
Air 0.026 W/m K
Solid 0.21 W/m K
 
Moisture diffusivity 2172 m2/s Wang and Brennan (1992)
4:49  105 exp
T
Binary diffusivity 2:6  106 eg m2/s Halder et al. (2007)
Bulk modulus
109 M 40:3 Pa Estimated for rubbery
106 M o 0:3 Pa Estimated for glassy
Poisson’s ratio
0:49M 40:3 – Estimated for rubbery
0:3M o 0:3 – Estimated for glassy

and 40 1C. The surface temperature rises from 32.5 1C to 50 1C in potato drying (10 mm thick slab at 70 1C). The large value of maxðJel Þ
200 min and stabilizes, reaching about 54 1C after 800 min of in the case of potato is because of the greater shear modulus for
drying. The predicted temperature history closely follows the potato which leads to deviations from free shrinkage. For hamburger
observed one. The predicted moisture content histories for three patties, Poisson’s ratio, n, stays close to 0.5 and, thus, much smaller
different values of slab thickness also follow the observed history deviations from free shrinkage are observed. As a potato slab is under
very well. The shrinkage of the potato slabs at the two drying a much larger expansive strains (near the surface as it dries up) as
temperature values (70 1C and 40 1C) is a little less than the compared to meat, its surface is more prone to cracking. Thus,
volume of moisture lost until moisture content of 0.3, with maxðJel Þ can be used as a criteria to predict and avoid drying
volume at 70 1C equal to or more than that at 40 1C. The situations most prone to cracking. Apart from cracking, other impor-
simulations capture the trends very well, apart from the small tant quality parameters, such as porosity development, case hard-
difference in the observed volumes at the two drying tempera- ening (surface drying leading to large increase in shear modulus and
tures, which the predictions could not capture. As the experi- reduced shrinkage), etc., can also be predicted from deformation
mental error values are not available, it is difficult to conclusively analysis.
say if the small difference in volumes at the two temperatures is
real. Nevertheless, the accurate predictions of moisture loss,
temperature and shrinkage histories for the drying process serve 4. Conclusions
to validate the modeling approach followed.
A poromechanics-based approach to mathematically model
3.3. Importance of solid mechanics analysis the coupled physics of transport and deformation during proces-
sing of food materials is developed. Following this comprehensive
Since the volume change is almost equal to moisture change for approach, food materials existing in a range of states (glassy to
the two food materials studied (above a critical moisture content for rubbery) and being processed under a variety of conditions, can
potato), the advantage of solving the solid momentum equation does be simulated to predict important food quality and safety para-
not lie in predicting volume change due to moisture content. The real meters (spatial and temporal histories of temperature, moisture
value of solid mechanics analysis lies in predicting small deviations and deformation). For deformation, primary driving forces are
from free shrinkage, which lead to stresses and can be important identified and their effect on the solid momentum balance is
indicators of food quality, related to cracking, for example. As an discussed in detail. The driving forces are: (1) gas pressure, which
example, Fig. 15 plots the maximum value of elastic Jacobian, Jel, as a causes the food material to swell (gas pressure gradient can be
function of normalized moisture content for hamburger cooking and directly treated as a source term for the solid momentum
A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497 6495

60 1.2
Surface temperature (ºC)

50 1.15

Max elastic Jacobian, Jel


Predicted Potato drying

40 1.1
Experiment
(Wang and Brennan, 1992)

30
0 200 400 600 800 1.05
Hamburger
Time (min) patty cooking

4
1
0 0.2 0.4 0.6 0.8 1
Moisture content (normalized)
Moisture content (d.b.)

3
10 mm
Fig. 15. Maximum value of elastic Jacobian, Jel, (ratio of actual volume to free
7 mm volume) vs. moisture content (normalized with respect to initial moisture
content) for the two processes simulated showing larger expansive strains for
2
4 mm Predicted potato drying as compared to hamburber patty cooking.

Experiment (Wang
1 heating processes such as microwave cooking and processes with
and Brennan, 1992)
internal generation such as bread baking. Also, solution of vapor
equation is not required unless local equilibrium between vapor
and liquid moisture breaks down. Assuming equilibrium vapor
0
concentration, liquid water and water vapor flux can be added to
0 200 400 600 800 1000
get the total moisture flux relationship, which with further
Time (min) simplifications takes the form of Fick’s law.
Fig. 13. (a) Spatially averaged surface temperature and (b) moisture content Two different food processes are simulated as implementa-
histories for drying of potato slabs. Moisture content histories are shown for three tions of the modeling framework developed: (1) single-sided
different slab thicknesses of 4, 7 and 10 mm, respectively. Drying temperature is cooking of hamburger patties for which shrinkage is equal to
55 1C and other input parameters are provided in Table 2. moisture loss throughout the process and (2) convective drying of
potato slabs for which shrinkage stops under a critical moisture
content. For both the cases, transport of moisture in liquid form
1
dominates. The difference lies in greater strains experienced by
the potato due to greater shear modulus at low moisture
0.8 contents. Accurate predictions of the experimental observations
for two completely different processes show the versatility of the
Shrinkage, V/V0

Experiment (70ºC) modeling framework. Being comprehensive and fundamentals-


0.6 (Wang and Brennan, 1995) based, the framework can be widely applicable in food product,
process and equipment design, accounting for both food quality
0.4 and safety as design parameters.
Predicted (40ºC &
70ºC, coincident)
0.2 Nomenclature
Experiment (40ºC)
(Wang and Brennan, 1995) aw water activity
0 ci concentration of species i, kg m  3
0 1 2 3 4
cp specific heat capacity, J kg  1 K  1
Moisture content (d.b.) c molar density, kmol m  3
Fig. 14. Volume change vs. moisture content (drying temperatures 40 and 70 1C,
C constant of proportionality in Eq. (59)
10 mm thickness). Dotted line is for shrinkage equal to moisture loss. D diameter, m
D stiffness tensor
balance) and (2) moisture change, which can be treated analogous Dbin effective gas diffusivity, m2 s  1
to thermal expansion/contraction to get the free volume change. Db effective diffusivity due to gradients of b, m2 s  1
For transport, temperature, moisture, vapor concentration and gas Da,b diffusivity of a due to gradients of b, m2 s  1
pressure are the primary variables of interest. As gas does not E Green–Lagrange strain tensor
enter the pores during processing of wet-rubbery materials, gas F deformation tensor
phase equation is not required for such materials. Even if gas is ~
g acceleration due to gravity, kg m  3
present, significant pressure generation occurs only for intensive h heat transfer coefficient, W m  2 K  1
6496 A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497

hm mass transfer coefficient of vapor, m s  1 Acknowledgements


I_ volumetric evaporation rate, kg m  3 s  1
I identity tensor This project was supported by National Research Initiative
J Jacobian Grant 2008-35503-18657 from the United States Department of
keff effective thermal conductivity, W m  2 K  1 Agriculture National Institute of Food and Agriculture.
ki permeability of phase i, m2
K bulk modulus, Pa
K1 defined by Eq. (41) References
L thickness of hamburger patty
M moisture content (dry basis) Achanta, S., 1995. Moisture Transport in Shrinking Gels During Drying. Ph.D.
Ma, Mv molecular weight of air and vapor Purdue University, Indiana, United States.
~ normal vector Ateba, P., Mittal, G.S., 1994. Modeling the deep-fat frying of beef meatballs. Int. J.
N Food Sci. Technol. 29 (4), 429–440.
~
n i,j mass flux of species i with respect to j, kg m  2 s  1 Bengtsson, N.E., Jakobsson, B., Dagerskog Sik, M., 1976. Cooking of beef by oven
pi pressure of phase or species i, Pa roasting—study of heat and mass-transfer. J. Food Sci. 41 (5), 1047–1053.
Biot, M.A., 1965. Mechanics of Incremental Deformations: Theory of Elasticity and
~
q heat flux, J m  2 s  1 Viscoelasticity of Initially Stressed Solids and Fluids, Including Thermody-
R universal gas constant, J kmol  1 K  1 namic Foundations and Applications to Finite Strain. Wiley, New York.
REV representative elementary volume Bouchon, P., Pyle, D.L., 2005. Modelling oil absorption during post-frying cool-
ing—I: model development. Food Bioproducts Process. 83 (C4), 253–260.
S00 Piola–Kirchoff stress tensor, Pa Chau, K.V., Snyder, G.V., 1988. Mathematical-model for temperature distribution
Si saturation of phase i of thermally processed shrimp. Trans. ASAE 31 (2), 608–612.
t time, s Choi, Y., Okos, M.R., 1986. Thermal properties of liquid foods—review. In: Okos,
M.R. (Ed.), Physical and Chemical Properties of Food, American Society of
T temperature Agricultural Engineers, Saint Joseph, Michigan, pp. 35–77.
vi,j velocity of species i with respect to j, m s  1 Cushman, J.H., 1997. The Physics of Fluids in Hierarchical Porous Media: Ang-
vw molar volume of water, m3 mol  1 stroms to Miles. Kluwer Academic Publishers, Norwell, MA.
Datta, A.K., 2007. Porous media approaches to studying simultaneous heat and
V volume, m3 mass transfer in food processes. I: problem formulations. J. Food Eng. 80 (1),
Vn critical volume at which shrinkage stops, m3 80–95.
W strain energy density, Pa Dhall, A., 2011. Multiphase Transport in Deformable Phase-Changing Porous
Materials. Ph.D. Cornell University, Ithaca, United States.
xi mole fraction of species i Dhall, A., Datta, A.K. Transport and deformation during single-sided contact-
heating of hamburger patties. J. Food Eng., submitted for publication.
Greek symbols Dincer, I., Yildiz, M., 1996. Modelling of thermal and moisture diffusions in
cylindrically shaped sausages during frying. J. Food Eng. 28 (1), 35–44.
e strain tensor, volume fraction Farid, M.M., Chen, X.D., 1998. The analysis of heat and mass transfer during frying
r density, kg m  3 of food using a moving boundary solution procedure. Heat Mass Transfer 34
(1), 69–77.
l latent heat of vaporization, J kg  1 Farkas, B.E., Singh, R.P., Rumsey, T.R., 1996. Modeling heat and mass transfer in
m shear modulus, Pa immersion frying. 1. Model development. J. Food Eng. 29 (2), 211–226.
mi dynamic viscosity of a phase, i, Pa s Fowler, A.J., Bejan, A., 1991. The effect of shrinkage on the cooking of meat. Int. J.
Heat Fluid Flow 12 (4), 375–383.
n Poisson’s ratio Halder, A., Dhall, A., Datta, A.K., 2007. An improved, easily implementable, porous
r stress tensor, Pa media based model for deep-fat frying—part I: model development and input
r0 effective stress tensor, Pa parameters. Food Bioproducts Process. 85 (C3), 209–219.
Halder, A., Dhall, A., Datta, A.K., 2010. Modeling transport in porous media with
r00 effective stress tensor due to mechanical load only, Pa
phase change: applications to food processing. J. Heat Transfer—Trans. ASME
f porosity 133 (3), 031010.
Cw water potential, Pa Hashin, Z., 1985. Large isotropic elastic deformation of composties and porous
media. Int. J. Solids Struct. 21 (7), 711–720.
ov , oa mass fraction of vapor and air in relation to total
Ikediala, J.N., Correia, L.R., Fenton, G.A., Ben-Abdallah, N., 1996. Finite element
gas modeling of heat transfer in meat patties during single-sided pan-frying. J.
Food Sci. 61 (4), 796–802.
Izumi, M., Hayakawa, K., 1995. Heat and moisture transfer and hygrostress crack
Subscripts formation and propagation in cylindrical, elastoplastic food. Int. J. Heat Mass
Transfer 38, 1033–1041.
amb ambient Katekawa, M.E., Silva, M.A., 2006. A review of drying models including shrinkage
effects. Drying Technol. 24 (1), 5–20.
a, g, s, v, w air, gas, solid, vapor, water Kondjoyan, A., Rouaud, O., McCann, M.S., Havet, M., Foster, A., Swain, M., Daudin,
c capillary J.D., 2006. Modelling coupled heat-water transfers during a decontamination
eff effective treatment of the surface of solid food products by a jet of hot air. I. Sensitivity
analysis of the model and first validations of product surface temperature
el elastic
under constant air temperature conditions. J. Food Eng. 76 (1), 53–62.
eq equilibrium Kowalski, S.J., 2000. Toward a thermodynamics and mechanics of drying pro-
f fluid cesses. Chem. Eng. Sci. 55 (7), 1289–1304.
Lewis, R.W., Shrefler, B.A., 1998. The Finite Element Method in the Static and
G ground (stationary observer)
Dynamic Deformation and Consolidation of Porous Media, second ed. John
i ith phase Wiley and Sons, Inc., New York.
m mechanical Lu, N., Likos, W.J., 2004. Unsaturated Soil Mechanics. John Wiley and Sons.
M moisture Mayor, L., Sereno, A.M., 2004. Modelling shrinkage during convective drying of
food materials: a review. J. Food Eng. 61 (3), 373–386.
0 at time t ¼0 Ni, H., Datta, A.K., 1999. Moisture, oil and energy transport during deep-fat frying
surf surface of food materials. Food Bioproducts Process. 77 (C3), 194–204.
surfd drip at the surface Ni, H., Datta, A.K., Torrance, K.E., 1999. Moisture transport in intensive microwave
heating of biomaterials: a multiphase porous media model. Int. J. Heat Mass
surfe evaporation at the surface Transfer 42 (8), 1501–1512.
T temperature Ogden, R.W., 1972. Large deformation isotropic elasticity—correlation of theory
and experiment for compressible rubberlike solids. Proc. R. Soc. London Ser.
A—Math. Phys. Sci., vol. 328; 1972, pp. 567.
Superscripts Perre, P., May, B.K., 2001. A numerical drying model that accounts for the coupling
between transfers and solid mechanics. Case of highly deformable products.
f Volumetric average of f over an REV Drying Technol. 19 (8), 1629–1643.
A. Dhall, A.K. Datta / Chemical Engineering Science 66 (2011) 6482–6497 6497

Rakesh, V., 2010. Transport in Rigid and Deformable Hygroscopic Porous Media van der Sman, R.G.M., 2007. Soft condensed matter perspective on moisture
During Electromagnetic and Combination Heating. Ph.D. Cornell University, transport in cooking meat. A.I.Ch.E. J. 53 (11), 2986–2995.
New York, United States. Vujosevic, L., Lubarda, V.A., 2002. Finite-strain thermoelasticity based on multi-
Ratti, C., Crapiste, G.H., Rotstein, E., 1989. A new water sorption equilibrium plicative decomposition of deformation gradient. Theor. Appl. Mech. 28–29,
expression for solid foods based on thermodynamic considerations. J. Food Sci. 379–399.
54 (3), 738–742. Wang, L., Singh, R.P., 2004. Finite element modeling and sensitivity analysis of
Scherer, G.W., 1989. Drying gels. 8. Revision and review. J. Non-Cryst. Solids 109 double-sided contact-heating of initially frozen hamburger patty. Trans. ASAE
(2–3), 171–182. 47 (1), 147–157.
Wang, N., Brennan, J.G., 1992. Effect of water binding on the drying behaviour of
Schrefler, B.A., 2002. Mechanics and thermodynamics of saturated/unsaturated
potato. In: Majumdar, A.S. (Ed.), Drying’92 Part B, Elsevier, Amsterdam, New
porous materials and quantitative solutions. Appl. Mech. Rev. 55 (4), 351–388.
York, pp. 1350–1359.
Shilton, N., Mallikarjunan, P., Sheridan, P., 2002. Modeling of heat transfer and
Wang, N., Brennan, J.G., 1995. Changes in structure, density and porosity of potato
evaporative mass losses during the cooking of beef patties using far-infrared
during dehydration. J. Food Eng. 24 (1), 61–76.
radiation. J. Food Eng. 55 (3), 217–222. Williams, R., Mittal, G.S., 1999. Low-fat fried foods with edible coatings: modeling
Singh, P.P., 2002. Effect of Viscoelastic Relaxation on Fluid and Species Transport in and simulation. J. Food Sci. 64 (2), 317–322.
Biopolymeric Materials. Ph.D. Purdue University, Indiana, United States. Whitaker, S., 1977. Simultaneous heat, mass and momentum transfer in porous
Smith, D.M., Scherer, G.W., Anderson, J.M., 1995. Shrinkage during drying of silica- media: a theory of drying. Adv. Heat Transfer 13, 114–203.
gel. J. Non-Cryst. Solids 188 (3), 191–206. Yamsaengsung, R., Moreira, R.G., 2002. Modeling the transport phenomena and
Tornberg, E., 2005. Effects of heat on meat proteins—implications on structure and structural changes during deep fat frying—part II: model solution and
quality of meat products. Meat Sci. 70 (3), 493–508 (Sp. Iss. SI). validation. J. Food Eng. 53 (1), 11–25.
USDA Nutrition Database /http://www.nal.usda.gov/fnic/foodcomp/search/S. Zhang, J., Datta, A.K., Mukherjee, S., 2005. Transport processes and large deforma-
Accessed July 31 2010. tion during baking of bread. A.I.Ch.E. J. 51 (9), 2569–2580.

You might also like