Sei sulla pagina 1di 26

Blackwell Science, LtdOxford, UKJFPEJournal of Food Process Engineering0145-8876Copyright 2005 by Food & Nutrition Press, Inc., Trumbull, Connecticut.

282107132Original Article FREEZE-DRYING SIMULATIONS. KHALLOUFI,

J.-L. ROBERT and C. RATTI

SOLID FOODS FREEZE-DRYING SIMULATION AND


EXPERIMENTAL DATA

S. KHALLOUFI1, J.-L. ROBERT2 and C. RATTI1,3


1
Soils Science and Agri-Food Engineering Department
2
Civil Engineering Department
Laval University
Sainte-Foy G1K 7P4
QC Canada

Accepted for Publication June 6, 2004

ABSTRACT

This article presents a mathematical model describing the unsteady heat


and mass transfer during the freeze drying of biological materials. The model
was built from the mass and energy balances in the dried and frozen regions
of the material undergoing freeze drying. A set of coupled nonlinear partial
differential equations permitted the description of the temperature and pres-
sure profiles, together with the position of the sublimation interface. These
equations were transformed to a finite element scheme and numerically solved
using the Newton-Raphson approach to represent the nonlinear problem and
the interface position. Most parameters involved in the model (i.e., thermal
conductivity, specific heat, density, heat and mass transfer coefficients etc.)
were obtained from experimental data cited in the literature. The dehydration
kinetics and the temperature profiles of potato and apple slabs were experi-
mentally determined during freeze drying. The simulation results agreed
closely with the water content experimental data. The prediction of tempera-
ture profiles within the solid was, however, less accurate.

INTRODUCTION

Freeze drying is a dehydration process where water vapor is removed by


sublimation from frozen materials, usually under conditions of low pressure
and temperature. The advantage of this process is mainly related to the high
final quality of products (Mellor 1978; Sagara and Ichiba 1994; Krokida and
Maroulis 1997). This attribute is due to two main characteristics of freeze

3
Corresponding author. TEL: 418-656-2131 # 4593; FAX: 418-656-3723; EMAIL:
Cristina.Ratti@sga.ulaval.ca

Journal of Food Process Engineering 28 (2005) 107–132. All Rights Reserved.


© Copyright 2005, Blackwell Publishing 107
108 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

drying: (1) Water, representing more than 80% of food products, is frozen and
therefore cannot serve as a solvent reactant throughout the sublimation
process; (2) Process temperatures are low (product temperatures are between
-40 to -20C during sublimation and final temperatures, in the range of
20–50C, depending on the heating plate temperature), so the deterioration
reactions (i.e., nonenzymatic reactions) are slow or are completely stopped
because of weak rate constant (Cheftel et al. 1977).
Despite the excellent quality of final products, the cost of the freeze-
drying process is fairly expensive (Lombraña et al. 1997; Sadikoglu et al.
1998) and its applicability remains mainly reserved to products of high market
value such as pharmaceutical products and some special foods (nutraceutical,
baby foods, spices and coffee). This major disadvantage of freeze drying is
directly related to the long processing times caused by the high material
resistance to heat transfer during the operation (Sagara and Ichiba 1994; Kuu
et al. 1995), the sample configuration, its own properties and the operating
conditions. In addition, the condenser temperature and the vacuum level
should be maintained at their lowest values during the freeze-drying process.
Energy constitutes a major cost factor, which is composed of the energy
necessary for sublimation and desorption as well as the energy to support
vacuum and refrigeration.
An important part of previous investigations aimed to reduce freeze-
drying times and consequently to lower the energy consumption, analyzing
the possibilities of controlling the heat intensity and the vacuum pressure and
therefore investigating possible ways of optimizing the freeze-drying pro-
cess. Several studies were carried out in laboratories and in pilot scale plants
(Sagara and Ichiba 1994; Kuu et al. 1995; Liapis et al. 1996). Simulation was
also used as a preliminary tool for the evaluation of the freeze-drying pro-
cess. Several theoretical models concerning the heat and mass transfer
phenomena during freeze drying can be found in the literature (Karel 1975;
Mellor 1978; Liapis and Bruttini 1995a; Lombraña and Izkara 1996;
Lombraña et al. 1997).
One of the earliest and most known models is that for solids with heat
and mass transfer through the dry layer (Karel 1975):

L2 rms ( Xo - X f )
t= (1)
8 Per( Pi - Ps )
which, because of simultaneous heat and mass transfer process, is equivalent
to:

L2 rms ( Xo - X f )
t= (2)
8kd (Ts - Ti )
FREEZE-DRYING SIMULATION 109

Although easy to use, Eqs. (1) and (2) have several key assumptions that are
not usually applicable. (1) The maximum allowable surface temperature, Ts is
reached instantaneously. (2) The heat output of the external supply is adjusted
to maintain Ts constant throughout the drying cycle. (3) Partial pressure in the
drying chamber is constant. (4) All the heat is used for the sublimation of
water vapor (Karel 1975).
More recently, numerical models with highly detailed equations have
been developed (Liapis and Bruttini 1995a; Lombraña and Izkara 1996; Lom-
braña et al. 1997; Brülls and Rasmuson 2002; George and Datta 2002). How-
ever, in most cases, adjustable parameters are needed to match the model
predictions to experimental data (Liapis and Marchello 1984; Millman et al.
1985; Sharma and Arora 1993; Sadikoglu and Liapis 1997; Sheehan and
Liapis 1998; George and Datta 2002). In other cases, no comparison with
experimental data is presented (Nastaj 1991; Liapis and Bruttini 1995b). In
addition, most of the models were developed for liquids and not for solid
products (Sadikoglu and Liapis 1997; Sheehan and Liapis 1998; Brülls and
Rasmuson 2002).
It is clear that one way to decrease the costs of freeze drying is to reduce
the operation time by increasing the temperature and therefore, a precise
understanding of the simultaneous heat and mass transfer occurring during
the process is required. Thus, the main goal of the present article is to
investigate the temperature, the vapor pressure and the water remaining within
a solid food during the process through a dynamic and unidirectional mathe-
matical model of the primary stage of freeze drying.
The structure of the present work is as follows: (1) a brief description of
the physical phenomena taking place during freeze drying, (2) a theoretical
analysis of partial differential equations necessary to describe the underlying
phenomena in an elementary volume, (3) the numerical solution with details
on the building of the computer program and (4) an experimental validation
of the model and a discussion of the effect of some important freeze-drying
parameters through simulation.

MODEL DEVELOPMENT

Device Description and Problem Formulation


The main components of the freeze-drying device (Fig. 1A) are (1) the
heating plates, to ensure the temperature control in the sample; (2) the vacuum
pump, to ensure the pressure control in the chamber and (3) the condenser,
which condenses the vapor and avoids its passage to the pump.
During the freeze-drying process, the product is placed between the
heating plates (Fig. 1B). Therefore, a radiant source, as well as heat by
110 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

Upper heating plate

Vapor

Sample Condenser
`
Ice
Lower heating plate

Vacuum
pump

Upper plate
q upper

Top

Sample Interfaces

Lower plate Bottom

q bottom

FIG. 1. (A) COMPONENTS OF A FREEZE-DRYER. (B) DETAILS ABOUT THE MECHANISMS


OCCURRING DURING FREEZE DRYING

conduction from the bottom, is imposed to the material slab. There are two
mechanisms of water elimination during freeze drying: sublimation (which
eliminates the frozen water) and desorption (which eliminates the bounded
unfrozen water). Sublimation occurs at the interface of the ice front as a
result of the transferred heat (Simatos et al. 1974). As dehydration proceeds,
the ice front retreats and the sublimated water vapor (at the ice front) is
removed by diffusion through the porous layer (Liapis and Bruttini 1995a).
The water vapor flows through the dried layer countercurrent to the heat flow.
FREEZE-DRYING SIMULATION 111

Then, it passes through the chamber to be finally collected on the condenser


plate.

Theoretical Analysis and Governing Equations


The physical model representing the foodstuff being freeze-dried has a
rectangular cross-section, with thickness 2E much smaller than the length
L and width l (Fig. 2A). The heat and mass transfer occurs in a direction
perpendicular to the largest side (Fig. 2A). To simplify the analysis, the fol-
lowing key assumptions were made:
(1) One-dimensional heat and mass flow in the normal direction from the
ice interface and the surface of the product.
(2) During freeze drying, sublimation occurs at the ice front interface, par-
allel to both bottom and top surfaces and situated at distance S from
them (Fig. 2B).
(3) The problem is symmetric and the uniform retreating ice front moves on
both sides to the centerline at the same speed (Fig. 2A). Some experi-
ments were done to validate this assumption. The results showed that
the temperature in the bottom and in the top of the sample were generally
comparable.
(4) In the dried porous region, the solid matrix and the enclosed vapor are
in thermal equilibrium.
(5) The thickness of the ice front interface is considered to be infinitesimal.
(6) At the ice front interface, the water vapor concentration is in equilibrium
with the ice.
(7) Each frozen and dried region is considered to be homogenous, having
uniform thermal conductivity, density and specific heat.
(8) Only the primary drying period (free water removed by sublimation) was
considered. The secondary drying (removal of the bound water) was
neglected, as suggested by Sharma and Arora (1993), Lombraña et al.
(1997) and Sheehan and Liapis (1998). The consideration of the second-
ary drying stage would have required, in addition, the knowledge of
sorption isotherms below and above 0C. However, sorption isotherms
below 0C are not available in the literature.
(9) Conduction was considered as the main energy transfer mechanism in
the frozen region, while both conduction and convection are considered
in the dried portion.
(10) Shrinkage is not considered in this model.
Due to the symmetry of the slab, the analysis will be applied only to half of
the sample. The equations describing the physical system, as applied to the
elementary volume that is shown in Fig. 2B, are as follows:
112 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

z a
A
x

E
y

b
B

0
qxd mxd

dxd Dried
zone
qxd+dxd mxd+dxd

S (t )
qxf

Frozen
Fr
dxf zone

qxf+dxf
E

FIG. 2. (A) GEOMETRICAL DIMENSIONS. (B) DISCRETIZATION USED IN THE


NUMERICAL SOLUTION

Dried Layer
(a) Temperature profile. The complete energy balance presented by
Sadikoglu and Liapis (1997) for the tray freeze-drying process was applied to
the freeze drying of a food slab under vacuum (the inert gas flow is negligible).
FREEZE-DRYING SIMULATION 113

The flow of vapor through the dry matrix was represented through a perme-
ability-type equation. The resulting equation is thus:

∂ Td ∂ Ê ∂ Td ˆ ∂ Ê ∂ Pv ˆ
= j + j (3)
∂t ∂xË 1 ∂x ¯ ∂xË 2 ∂x ¯

kd PerCpv
where j1 = and j 2 = T
Cpd r d Cpd r d d
The first term on the right hand side of Eq. (3) refers to energy transfer
by conduction, while the second refers to the energy that is transferred due to
vapor flow.
The permeability coefficient Per, necessary for the determination of
parameter f2 of Eq. (3), was calculated as reported by Simatos et al. (1974)
and Liapis and Litchfield (1978):

M C2 DAB Kn
Per = (4)
RT C2 DAB + Kn(PT - Pv )

The Knudsen (Kn) constant was calculated as follows (Liapis and Litchfield
1978; Mellor 1978):
12
È M ˘ 2 È 8 ˘1 2 (5)
Kn = C1 Í ˙ , where C = r
Î RTd ˚
1
3 ÎÍ p ˚˙
C2 is a dimensionless constant, which depends on the internal texture of the
product (Simatos et al. 1974). According to Gunn et al. (1969), this parameter
characterizes the porous medium but does not depend on the gas species
present in the matrix.
The molecular diffusivity (DAB) of the binary mixture was estimated by
the Fuller equation (Skelland 1974):

DAB = 1.16 10-4T1.75 (6)

(b) Pressure profile.

∂ Pv ∂ Ê ∂ Pv ˆ ∂T
= b + b2 d (7)
∂t ∂xË 1 ∂x ¯ ∂t

PerR P
where b1 = and b 2 = v
Me Td
The permeability value Per, required for the determination of the
parameter b1 of Eq. (7), was calculated using Eq. (4).
114 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

Frozen Layer
In the frozen region, there is only heat transfer by conduction. Thus, the
temperature profile is represented by the following differential expression:
∂ Tf ∂ Tf Ê ∂ Tf ˆ
= w (8)
∂t ∂x Ë ∂x ¯
kf
where w =
Cp f r f

Boundary Conditions
At the surface, ¥ = 0
(a) Temperature. The boundary condition at the material surface can be
described by:
∂ Td
kd = - hT (T• - Td [ 0, t ]) (9)
∂x
where hT is the heat transfer coefficient. Two heat transfer coefficients were
considered for the simulations:

(1) one was obtained from the following theoretical expression:

hT = sF(T•2 + Td2 [ 0, t ])(T• + Td [ 0, t ]) (10)

where s is the Stefan-Boltzmann constant and F is the radiation view


factor.
(2) another was an experimental heat transfer coefficient, which was reported
by Wang and Shi (1997, 1998).

(b) Pressure. A convection resistance coefficient (hP), which depends on the


total pressure (Lombraña and Izkara 1996), was considered for mass transfer
at the surface:
∂Pv
Per = hT ( P• - Pv [ 0, t ]) (11)
∂x
At the interface, ¥ = S(t)
S(t) represents the interface position of the sublimation front, which
depends on time. At this position, an additional equation was obtained from
the energy balance on a differential element of thickness dS:
FREEZE-DRYING SIMULATION 115

∂ Td ˘ ∂ Tf ˘
DHsub msub = - kd + kf (12)
˙
∂ x ˚s ∂ x ˙˚ s
The term on the left-hand side corresponds to the latent heat of sublimation.
The right-hand side terms are the conductive heat fluxes in and out of the
differential element. The mass flux of sublimated vapor can be expressed as
follows:
dS
msub = r f WC FF (13)
dt
There should be no discontinuity in temperature at the interface (S[t]), thus:
Td S (t ) = Tf S (t ) (14)
For vapor pressure calculation, the thermodynamic equilibrium existing
between the water vapor and the frozen product. was used at the interface
(Clapeyron equation):

Ê B ˆ
Pv S (t ) = expÁ A + ˜ (15)
Ë T S (t ) ¯

At the center, ¥ = E
At the matrix center, because of the symmetry, adiabatic or zero heat flux
condition can be supposed.
∂ Tf
=0 (16)
∂x
Initial Conditions
Uniform temperature and vapor pressure are chosen as initial conditions,
thus:
0<x<E T(x,0) = Tini (17)
0 < x < S(0) P(x,0) = Pini (18)
Calculation of Water Remaining
(a) Initial water. The initial water, (X0) existing in the matrix before freeze
drying, is:
X0 = WC r f VT (19)
(b) Sublimated water. The sublimated water represents the frozen water that
is eliminated from the matrix at each moment. This amount depends on the
position of the interface and can be expressed by:
116 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

msub (t ) = WC FF r f l L S(t ) (20)


(c) Remaining water. The amount of remaining water (Xr) is the sum of the
frozen water that has not been sublimated yet and the bound water, therefore:
Xr (t ) = WC r f L l( E - FF S(t )) (21)
The progress of remaining water can be followed by its comparison to the
initial water content (water remaining/initial water content, %):
Xr (t ) S( t )
(%) = 100 È1 - FF ˘˙ (22)
X0 Í
Î E ˚

MATERIALS AND METHODS

Experimental Data
Slabs (14-mm and 8-mm thickness) of potato and apple were chosen for
freeze-drying kinetic experiments. The samples were frozen in a medical
freezer at -40C (Sanyo MDF-235, Japan) and freeze-dried in a Unitop 400 L
(Virtis, NY) drying chamber connected to freezemobile 25 EL (Virtis). The
frozen samples were placed on the freeze-dryer shelf, initially at -40C. When
the condenser temperature was less than -85C, the vacuum pump was turned
on. The heating system (1C/min) was started after the total pressure was less
than 10 mtorr. The shelf temperature was set at 40C. Each experimental
determination of water content, after different process times, was done inde-
pendently from the others. At the initial time, the 8- or 14-mm-thick food slab
was weighed and introduced into the freeze drier. It was left inside the freeze
drier for a predetermined process time (1–8, 10, 12, 14 or 16 h) and then it
was removed and weighed. Then the experiment started all over again for
another predetermined process time. The experimental data for each process
time and for the slab thickness was obtained in triplicate. The maximum
process times were 8 and 16 h for an 8-mm or 14-mm-thick slab, respectively.
The duration of the freeze drying was calculated from the moment
when the heating system was started. To calculate the amount of water that
was eliminated, the samples were weighed using a Mettler Toledo balance
(PB 1502, ±0.01 g, Switzerland) before and after the freeze-drying opera-
tion. Once freeze drying ended, the dry matter of each sample was deter-
mined in a vacuum oven. The moisture contents of the samples were
estimated from the differences between the final freeze-dried weight and the
dry matter. A comparison of the experimental data with the simulation
results was done using the ratio percentage Xr(t)/X0 (water remaining/initial
water content, %).
FREEZE-DRYING SIMULATION 117

Upper plate

Top

Sample

Thermocouples
Bottom

Lower plate

1 2 3 4 5 6
Data
Logger
logger

Computer

FIG. 3. EXPERIMENTAL PROTOCOL FOR MONITORING TEMPERATURE EVOLUTION IN


THE SAMPLE DURING THE FREEZE-DRYING PROCESS

The temperature profiles during the freeze drying of a 28-mm-thick


sample were obtained by placing five standard 1-mm diameter K-type ther-
mocouples in the product at different locations (shown in Fig. 3): one at the
center, two inside the product very close to each bottom and top surface and
the last two at half of the thickness. The thermocouple insertion was per-
formed in a fresh food slab before freezing and then the product, with the
inserted thermocuple, was frozen. Before the thermocouple insertion, a tiny
118 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

needle was inserted in the product where the thermocouple was to be placed
(to serve as a guide in order to ease the thermocouple insertion) up to the
position where the thermocouple was supposed to be (i.e., 7 mm for the
measurement of the center temperature in a 14-mm-thick slab) and then a
determined length of the thermocouple was inserted. The placements of the
thermocouples were supposed to be the length up to where the needle guide
was inserted. All measures of position and distances were determined by using
a micrometer Digimatic (CD 6 B, Mitutoyo Corporation, Japan). During
freeze drying, temperature variations were recorded by a digital data logger
(21X Micrologger, Campbell Scientific Inc., UT), which was directly plugged
to the computer. The temperature acquisition was done at 1-min intervals.

Numerical Solution
To obtain a finite element (characterized by two node linear 1D element)
form of the governing Eqs. (3), (7), (8) and (12), each region (frozen and
dried) was divided into n elements of thickness dxd and dxf (Fig. 2B). Eqs. (3),
(7) and (8) can be organized in a matrix form (Eq. 23):
Ï ∂ Td ¸ È ∂ Ê ∂ˆ ∂ Ê ∂ ˆ˘
1 0 Ô dt Ô Í ∂ x Ëj 1 ∂ x ¯ ∂ x Ëj 2 ∂ x ¯ ˙ÏTd ¸
Ì ˝= ∂ ˙ÌP ˝
- b2 1 Ô ∂ Pv Ô Í ∂ Ê (23)
12
4 4 3 Í 0 b1 ˆ ˙Ó v ˛
m Ó ∂t ˛ 1 Î 4444 ∂ x Ë ∂ x3
4244444
¯˚
k

The application of the finite element method transforms the matrices m and
k into discrete forms, whose size depends on the number of elements chosen.
Then an implicit Euler time marching method is applied, in which the non-
linearity is treated by a Newton-Raphson method. Eventually, the condition
(14) is applied (Eq. 24):
Ï {Rd }n ¥1 ¸
È[ K d ]n ¥ n {K d }n ¥1 [ 0 ]n ¥ l ˘ Ï{DF d }n ¥1 ¸Ô
Í ˙Ô Ô Ô rd ( n + 1) rf 1 ÔÔ
( )
Í ÒK d ·1¥ n 0 ÒK f ·1¥ l ˙Ì DTi ˝ = -Ì - ˝ (24)
kd ( n + 1, n + 1) k f (1,1) Ô
Ô Ô Ô
Í [ 0]l ¥ n
Î {K f }l ¥1 [ K f ]l ¥l ˚Ó{DF f }l ¥1 ˛ Ô
˙
{R f }l ¥1 Ô˛
Ó
where DFd and DFf are the discrete variable increment vectors (Td and Pd for
the dried part and Tf for the frozen part), DTi is the increment of the common
temperature of the dried and frozen parts, Rd and Rf are the residuals, Kd and
Kf are matrices obtained from m and k. Originally, the sizes of the matrices
were (n + 1) ¥ (n + 1) and (l + 1) ¥ (l + 1). By the introduction of the
condition (14), the last temperature equation of Kd and the first of Kf are
FREEZE-DRYING SIMULATION 119

combined to give the central equation of the system (24), which ensures the
coupling of the system. Indeed, the matrix system allows calculating the
variation of the two parameters (DT and DPv) at a given node as a function of
the temperature and the vapor pressure values at the same node and at the
neighborhood nodes.
A computer program was developed in Matlab 5.2 based on the previ-
ously described numerical scheme. The program calculates for each value of
time, the temperature and vapor pressure spatial distributions, their average
values as well as the ice front position. It also calculates the instantaneous and
the cumulative values of the sublimating water. The specific heat, thermal
conductivity, density and porosity of the complete physical model were con-
sidered constant, but the developed numerical scheme could be easily adapted
to use complex expressions involving temperature, pressure- and/or humidity-
dependent terms.
The knowledge of the interface position S(t) has a fundamental impor-
tance in solving the freeze-drying problem. Since this position recedes con-
tinuously, it adds additional calculations and complexity to the problem.
Equation 13 was used to calculate the interface position at each time step (Dt).
For this, the program considers the S value as a variable and its calculation is
done by iteration using the Newton Raphson method. This method was chosen
upon the more traditional methods of evaluating S because it avoids interpo-
lation without increasing the running time.
To avoid pronounced gradients and consequently, a high number of
iterations, the total pressure variation in the chamber and the vapor pressure
in the condenser were approximated at the beginning of each simulation by
the following equation:

( Dt + G )
P(t ) = Pmax - ( Pmax - Pmin ) t (25)
Dt (t + G )

where Pmax is the pressure at the beginning, Pmin the pressure at the end of the
process, Dt the time required for P to be Pmin, t the time and G, a constant. It
can be seen from Eq. (25) that if the value of G is low, the end value of
pressure will be reached quicker. Several tests were made to assure that this
equation does not affect the final solution. The proposed technique allowed
achieving the final results in reasonable computation times.
In addition, preliminary tests showed that in the vicinity of the physical
model surface and in the neighborhood of the sublimation front, the temper-
ature (in both zones) and the vapor pressure (in the dried zone) gradients could
be very pronounced. This strongly affects the number of iterations. To find a
remedy for this situation, the network-refining technique was used. This
method involves more nodes where the gradient is steep and less nodes where
120 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

the gradient is flat, through an exponential function. Therefore, the network


was refined at each small displacement of the sublimation front.
Even if an implicit scheme was used to solve the coupled equations
(temperature and vapor pressure), the numerical solution stability was found
to be strongly dependent on initial conditions and particularly on the choice
of time step (Dt). To avoid the divergence of the solution, the time step (Dt)
was continuously adjusted while running the program. After each solution,
the Dt was reevaluated as a function of the number of iterations (from an initial
0.01 s step up to 100 s or more) without affecting the solution convergence.
To be able to start the numerical calculation, an arbitrary initial thickness
of the dried layer was chosen (0.5% of the slab thickness), thus:
S(0) = 0.01E (26)
This initial thickness of dry layer should be considered as formed during the
start-up of the process before energy is supplied to the sample. This period
represents the time required for passing from the atmospheric to the vacuum
pressure.

Physical Properties and Other Parameters


Table 1 shows all the parameters and expressions for physical properties
used for the simulations, obtained from published values of similar materials
in the literature. It should be signaled, however, that some parameters such as
C2 (see Eq. 4) were not easily available in the literature and, in addition,
physical properties generally depend on the product composition, the freeze-
drying device and on operating conditions. Therefore, the value C2 was esti-
mated through preliminary simulations. The final values of permeability that
were obtained using Eq. (4) for the two products (apple and potato) were
found to fall within the range of those reported by Kessler (1981) for some
food products, corroborating the predictions done on C2.

RESULTS AND DISCUSSION

Comparison with Experimental Results


Figure 4 shows the experimental freeze-drying curves for 14- and 8-mm-
thick apple and potato slabs at 40C. As expected, the effect of thickness on
the freeze-drying time was marked. Indeed, the sublimation of 50% of the ice
in the apple requires about 240 min if the thickness is 8 mm and about 400 min
for 14-mm thickness. As can be seen, freeze-drying times did not vary with
the square of the thickness. Other authors previously found a linear relation-
TABLE 1.
PARAMETERS AND PHYSICAL PROPERTIES USED FOR SIMULATION

Parameters Potato Apple

Values References Values References

rf (kg/m3) 1022 Kerr et al. (1996) 787.93 Khalloufi and Ratti (2002)
Cpf (J/kgK) 1.8 ¥ 103 Lewis (1987) 2.84 ¥ 103 Rao and Rizvi (1995)
Kf (W/m.K) 1.09 Mujumdar (1995) 1.289 Rao and Rizvi (1995)
rd (kg/m3) 180 Krokida and Maroulis (1997) 123.17 Khalloufi and Ratti (2002)
Cpd(J/kgK) 1.66 ¥ 103 a = 1.35 ¥ 10-7. Kostaropoulos and Saravacos (1997) a = 1.65 ¥ 10-7 Kostaropoulos and Saravacos (1997)
Kd(W/m.K) 0.0404 Kessler (1975) 0.0337 Kessler (1981)
Π(%) 88 Krokida and Maroulis (1997) 90 Khalloufi and Ratti (2002)
FF (g/g) 4 Rao and Rizvi (1995) 5 Rao and Rizvi (1995)
WC (g/g) 78.62 Determined for this study 84.36 Determined for this study
r (m) 48 ¥ 10-6 Estimated 43 ¥ 10-6 Mellor (1978)
C2 (dimensionless) 0.07 Determined in order to obtained the value of 0.05 Determined in order to obtained the value
permeability cited by Kessler (1981) of permeability cited by Kessler (1981)
F (dimensionless) 0.95 Given

Common expressions and parameter Expressions References


FREEZE-DRYING SIMULATION


Equilibrium equation (vapor-glass) (P in Pa and T in K) Psat = expÊ A + A = 28.80073 and B = -6142.89 Poling et al. (2001)
Ë Tsat ¯
2
Latent heat of sublimation (DHsub) (kJ/kg). T in K DHsub = -0.0037T + 1.7502T + 2630.4 Poling et al. (2001)
External mass transfer coefficient (hP) (kg/Pa m2.s) P in Pa hP = 15.912 ¥ 10-3 P-0.7731 Lombraña and Izkara (1996)
hT (W/m2.K) 6.7 Wang and Shi (1997, 1998)
Cpv (J/kgK) 1.6747 ¥ 103 Litchfield and Liapis (1982)
s (W/m2.K4) 5.676 ¥ 10-8 Liapis and Marchello (1984)
R (J/kmol.K) 8.314 ¥ 103 Sadikoglu and Liapis (1997)
M (kg/kmol) 18 Sadikoglu et al. (1998)
121
122 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

100

90
Potato
80
Water remaining (X /X0)x100

70

60

50
14 mm
40

30
8 mm
20

10

0
0 100 200 300 400 500 600 700 800 900 1000

100

90
Apple
80
Water remaining (X /X0)x100

70

60

50

40 14 mm
30

20
8 mm
10

0
0 200 400 600 800 1000 1200
Time (min)

Experimental data (14 mm) Experimental data (8 mm)


Model predictions (14 mm) Model predictions (8 mm)

FIG. 4. COMPARISON BETWEEN EXPERIMENTAL DATA AND MODEL PREDICTIONS OF


FREEZE DRYING OF APPLE AND POTATO (8 AND 14 mm THICKNESS) AT 40C HEATING
PLATE TEMPERATURE
FREEZE-DRYING SIMULATION 123

ship between freeze-drying time and thickness, as Sharma and Arora (1995),
for yoghurt under different heating transfer modes and Saravacos (1967), in
the case of apple and potato. In the literature, however, freeze-drying time is
often presented as proportional to the square of the piece size (see Eqs. 1 and
2), because the process is usually explained from the diffusion theory. King
(1968) explained the anomaly with respect to pure diffusion theory, based on
an externally controlled (boundary layer) freeze-drying process, which could
clearly explain the linear relationship between freeze-drying time and product
thickness.
The simulation of apple and potato freeze-drying kinetics was done by
solving the proposed model with the values listed in Table 1. Figure 4 shows
a comparison between predicted and experimental results on water remaining
in apple and potato as a function of freeze-drying time. As can be seen, the
agreement between both values is close, indicating that the present one-
dimensional model gives reasonably accurate results to predict freeze-drying
curves. The RSMEs (root square mean errors) between predicted and exper-
imental data were determined as 18.6 and 45.3 min for 8- and 14-mm-thick
apple slabs and 9.7 and 50.8 min for 8- and 14-mm-thick potato slabs, respec-
tively. The main differences between experimental and predicted values were
found for thicker slabs at the end of the process. An explanation for this could
be (1) the absence of desorption in the model and/or (2) the mass transfer,
happening by lateral surfaces, which could be more significant in the case of
thicker slabs.
Figure 5 shows the experimental and predicted temperature profiles for
28-mm-thick apple and potato slabs. It should be noted that because the
present model does not take into account the desorption phase (assumption
8), the comparison will be limited only to the first stage of freeze drying (from
the start of the heating until the complete disappearance of ice). From Fig. 5,
it can be seen that

(1) The sublimation temperature is seen as a plateau at the first stage of


freeze drying (x = E). The model gives a reasonably good prediction of
the sublimation temperature for both products.
(2) The temperature predictions, at different locations within the product,
substantially underestimates the obtained experimental data. This dis-
agreement can be interpreted from different viewpoints:

• most physical parameters used in the model were obtained from the litera-
ture or predicted from theoretical expressions (hT, k etc.)
• there was a clear difficulty encountered during the experiments to measure
an accurate frozen core temperature in specific locations within the sample
during freeze drying, particularly because the heat passing through the probe
124 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

35

25

15
x=0 Predicted
Temperature (C)

temperatures
5
x = E/2
–5

Experimental
–15
x=E temperatures

–25

p le
Ap
–35
Sublimation Desorption

–45
0 500 1000 1500 2000 2500 3000 3500 4000

35

25

Predicted
15
temperatures
Temperature (C)

x=0
Experimental
–5
x = E/2 temperatures

–15

–25
x=E
a to
P ot
–35
Sublimation Desorption
–45
0 500 1000 1500 2000 2500

FIG. 5. EXPERIMENTAL AND PREDICTED TEMPERATURE PROFILES DURING FREEZE


DRYING OF APPLE AND POTATO

generated some freeze drying around the thermocouple, which probably did
not help in reflecting the real temperature.
• the model did not take into account all the phenomena involved (i.e.,
desorption)
• the sample was too thick to assume unidirectional heat transfer.
FREEZE-DRYING SIMULATION 125

100
2E = 14 mm
90

80
Water remaining (X /X0) 100

70

60

50

40
Radiation transfer
30
Thermal resistance transfer
20

10
Experimental data
0
0 200 400 600 800 1000 1200
100
2E = 8 mm
90

80
Water remaining (X /X0) 100

70

60

50
Radiation transfer
40

30
Thermal resistance transfer
20
Experimental data
10

0
0 100 200 300 400 500 600
Time (min)

FIG. 6. COMPARISON BETWEEN PREDICTIONS CONSIDERING AND EXPERIMENTAL


HEAT TRANSFER COEFFICIENT AND PURE THEORETICAL RADIATION (8 AND 14 mm
THICKNESS APPLE SLABS)

Figure 6 shows the comparison between freeze-dying kinetics that was


obtained using theoretical (pure ideal radiation) and experimental heat
transfer coefficients. It can be seen that the coefficient, which was experi-
mentally found by Wang and Shi (1997, 1998) and which includes the total
heat transfer including radiation, gives much better predictions than the
use of the equation for ideal blackbody radiation. For the freeze-drying
process, Wolf and Gibert (1998) reported that the heat transfer coefficient
126 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

(hT) depends on pressure and its value varies from 4 W/m2.K at 1 Pa to


50 W/m2.K at 100 Pa.
Figure 7A shows that the heating plate temperature significantly affects
the freeze-drying kinetics. The sublimation process finishes at 12.6 h for a
heating plate temperature of 70C, while only 47% of sublimation is achieved
during the same period if the heating plate temperature is 10C. The impact of
the heating temperature on the surface and sublimation temperatures of the
sample is given in Fig. 7B. It can be seen that, while increasing the heating
plate temperature does not markedly increase the frozen core temperature, the
product surface temperature does. This finding can greatly help when opti-
mizing the product quality during freeze drying. Hammami and René (1998)
indicated that to avoid the collapse phenomena and the quality problems
during freeze drying, the temperature of the frozen solid should be below the
melting onset temperature and the temperature of the product in the dry zone,
below the glass transition of the dry solids (Tgs). Two temperature limits were
shown to be essential during the freeze drying of a particular product: the ice
melting onset temperature (Tm) and the glass transition temperature of the dry
solids (Tgs). Nevertheless, from Fig. 7B, it can be concluded that the second
temperature limit (glass transition of the dry solids) is a stronger optimization
parameter during freeze drying than the former.

CONCLUSIONS

Mathematical simulations were used to predict freeze-drying kinetics and


temperature profiles of food slabs and to study the impact of some parameters
on the process. The nonlinear governing equations were solved numerically
and the results were compared to experimental data. The comparison revealed
a very good agreement between predicted and experimental freeze-drying
kinetics and sublimation temperature, while the predicted temperature profiles
did not closely represent the experimental data. To improve the prediction of
the present model, the following considerations should be taken into account:
(1) to include a complex formulation of permeability, (2) to add the second
stage (desorption) of freeze drying (3) to consider two different thermal
resistances (one for the bottom, taking into account the thermal contact with
the heating plate and the other for the top of the sample) and (4) to consider
two dimensional transfers. In addition, extra experimental research should be
done to determine physical properties, such as thermal conductivity as a
function of pressure, C2 coefficient, pore radius and isotherm function at two
temperatures (below and up 0C) etc.
The simulations showed that the use of an experimental heat transfer
coefficient is a better option for simulation than to consider pure radiation
FREEZE-DRYING SIMULATION 127

100
A
90

80
Water remaining (X /X0) 100

70

60

50

40

30

20

10

0
0 5 10 15 20 25 30 35 40
Time (h)
40
B
30
Temperature of product (C)

20

10

-10

-20
0 10 20 30 40 50 60 70 80

Heating temperature (C)

Surface of sample Front of sublimation

FIG. 7. (A) EFFECT OF THE HEATING PLATE TEMPERATURE ON FREEZE-DRYING


KINETICS. (B) EFFECT OF HEATING PLATE TEMPERATURE ON SAMPLE
TEMPERATURE PROFILE
128 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

through a theoretical expression. In addition, the heating plate temperature


showed a major effect on freeze-drying kinetics and surface temperature. This
simulation program can therefore be used as a tool for the optimization of the
freeze-drying process.

NOMENCLATURE

r Mass density (kg/m3)


s Stephan-Boltzmann constant values (J/s.m2.K4)
DH Latent heat (kJ/kg)
e Porosity
C1 Knudsen constant value (m)
C2 Dimensionless constant
Cp Specific heat (J/kgK)
DAB Molecular diffusivity (m2/s)
E Half the thickness (m)
F Radiation view factor
FF Frozen fraction of water (kg frozen water/kg total water)
hP Mass transfer coefficient (kg/Pa m2 s)
hT Heat transfer coefficient (J/m2.s.K)
k Thermal conductivity (J/m.K.s)
L Length (m)
l Width (m)
M Molecular weight (kg/kmol)
m Vapor flux (kg/s.m2)
P Pressure (Pa)
Per Permeability (s)
Pini Initial pressure (Pa)
Pmax Pressure at the beginning (Pa)
Pmin Pressure at the end (Pa)
q Heat flux (J/s)
R Gas constant value (J/kmol.K)
r Pore radius of dried layer (m)
S Interface position (m)
T Temperature (K)
Tini Initial temperature (K)
t Time (s)
Tg Glass transition temperature (K)
Tm Melting temperature (K)
V Volume (m3)
WC Initial water content (kg water per kg fresh product)
FREEZE-DRYING SIMULATION 129

X Amount of water (kg)


x Spatial position (m)

Subscript
• chamber conditions
d dry
f frozen
r remaining
s solid
sat saturation
sub sublimation
T total
v vapor
0 initial

REFERENCES

BRÜLLS, M. and RASMUSON, A. 2002. Heat transfer in vial lyophilization.


Int. J. Pharm. 246, 1–16.
CHEFTEL, J.C., CHEFTEL, H. and BESANÇON, P. 1977. Introduction
a la Biochimie et a la Technologie Des Aliments. Technique et
Documentation-Lavoisier, Paris, France.
GEORGE, J.P. and DATTA, A.K. 2002. Development and validation of heat
and mass transfer models for freeze-drying of vegetable slices. J. Food
Eng. 52, 89–93.
GUNN, R.D., CLARK, J.P. and KING, C.J. 1969. Mass transport in freeze
drying basic studies and processing implications. In Recent Develop-
ments in Freeze Drying, Vol 1, p. 260, International Institute of Refrig-
eration, Paris, France.
HAMMAMI, C. and RENE, F. 1998. Determination of freeze-drying process
variables for strawberries. J. Food Eng. 32, 133–154.
KAREL, M. 1975. Heat and mass transfer in freeze drying. In Freeze Drying
and Advanced Food Technology, Chapter 14 (S.A. Goldbith, L. Rey and
W.W. Rothamayr, eds.) pp. 177–202, Academic Press, New York, NY.
KERR, W.L., KAUTEN, R.J., OZILGEN, M., MCCARTHY, M.J. and REID,
D.S. 1996. NMR imaging calorimertic, and mathematical modeling stud-
ies of food freezing. J. Food Process Eng. 19, 363–384.
KESSLER, H.G. 1975. Heat and mass transfer in freeze drying of mixed
granular particles. In Freeze Drying and Advanced Food Technology,
130 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

Chapter 16 (S.A. Goldbith, L. Rey and W.W. Rothmayr, eds.) pp. 223–
239, Academic Press, Ltd, London, England.
KESSLER, H.G. 1981. Food Engineering and Dairy Technology. Kessler, VA;
Freising, Germany.
KHALLOUFI, S. and RATTI, C. 2003. Quality deterioration of freeze-dried
foods as explained by glass transition temperature and internal structure.
J. Food Sci. 68(3), 892–903.
KING, J. 1968. Rates of moisture sorption and desorption in porous dried
foodstuffs. Food Technol. 22(4), 165–171.
KOSTAROPULOS, A.E. and SARAVACOS, G.D. 1997. Thermal diffusivity
of granular and porous foods at low moisture content. J. Food Eng. 33,
101–109.
KROKIDA, M.K. and MAROULIS, Z.B. 1997. Effect of drying method on
shrinkage and porosity. Drying Technol. 15(10), 2441–2458.
KUU, W.Y., MCSHANE, J. and WONG, J. 1995. Determination of mass
transfer coefficient during freeze drying using modeling and parameter
estimation techniques. Int. J. Pharm. 124, 241–252.
LEWIS, M.J. 1987. Physical Properties of Foods and Foods Processing Sys-
tems. Horwood, E. Ltd., Chichester, England.
LIAPIS, A.I. and BRUTTINI, R. 1995a. Freeze drying. In Handbook of
Industrial Drying, 2nd Ed., Chapter 10 (A.S. Mujumdar, ed.) pp. 309–
343, Marcel Dekker, Inc., New York.
LIAPIS, A.I. and BRUTTINI, R. 1995b. Freeze-drying of pharmaceutical
crystalline and amorphous solutes in vials: Dynamic multidimen-
sional models of the primary and secondary drying stages and quali-
tative features of the moving interface. Drying Technol. 13(1 and 2),
43–72.
LIAPIS, A.I. and LITCHFIELD, R.J. 1978. Optimal control of freeze dryer-
I. Chem. Eng. Sci. 34(7), 975–981.
LIAPIS, A.I. and MARCHELLO, J.M. 1984. Freeze drying a frozen liquid in
a phial. Drying Technol. 2(2), 203–217.
LIAPIS, A.I., PIKAL, M.J. and BRUTTINI, R. 1996. Research and develop-
ment needs and opportunities in freeze-drying. Drying Technol. 14(6),
1265–1300.
LITCHFIELD, R.J. and LIAPIS, A.I. 1982. Optimal control of freeze dryer-
II. Dynamic analysis. Chem. Eng. Sci. 37(1), 45–55.
LOMBRAÑA, J.I., DE ELVIRA, C. and VILLARÁN, M. 1997. Analysis of
operating strategies in the production special foods in vials by freeze
drying. Int. J. Food Sci. Technol. 32, 107–115.
LOMBRAÑA, J.I. and IZKARA, J. 1996. Experimental estimation of effec-
tive transport coefficients in freeze drying for simulations and optimiza-
tion purposes. Drying Technol. 14(3 and 4), 743–763.
FREEZE-DRYING SIMULATION 131

MELLOR, J.D. 1978. Fundamentals of Freeze Drying. Academic Press Inc,


London, England.
MILLMAN, M.J., LIAPIS, I.A. and MARCHELLO, J.M. 1985. An analysis
of lyophilization process using a sorption–sublimation model and various
operational policies. Aiche J. 31(10), 1594–1604.
MUJUMDAR, A.S. 1995. Handbook of Industrial Drying, 2nd Ed., Marcel
Dekker, Inc., New York, NY.
NASTAJ, J. 1991. A mathematical modelling of heat transfer in freeze drying.
In Drying 91 (A.S. Mujumdar and I. Filkova, eds.) pp. 405–413. Elsevier,
London.
POLING, B., PRAUSNITZ., J. and O’CONNELL, J. 2001. Vapor pressures
and enthalpies of vaporization of pure fluids. In The Properties of Gases
and Liquids, 5th Ed., Chapter 7, p. 768, McGraw-Hill, New York, NY.
RAO, M.A. and RIZVI, S.S.H. 1995. Engineering Properties of Foods, 2nd
Ed., Marcel Dekker, New York, NY.
SADIKOGLU, H. and LIAPIS, A.I. 1997. Mathematical modeling of the
primary and secondary drying stages of bulk solution freeze-drying in
trays: Parameter estimation and model discrimination by comparison of
theoretical results with experimental data. Drying Technol. 15(3 and 4),
791–810.
SADIKOGLU, H., LIAPIS, A.I. and CROSSER, O.K. 1998. Optimal control
of the primary and secondary drying stages of bulk solution freeze drying
in trays. Drying Technol. 16(3–5), 399–341.
SAGARA, Y. and ICHIBA, J. 1994. Measurement of transport properties for
the dried layer of coffee solution undergoing freeze drying. Drying
Technol. 12(5), 1081–1103.
SARAVACOS, G. 1967. Effect of the drying method on the water sorption of
dehydrated apple and potato. J. Food Sci. 32, 81–84.
SHARMA, N.K. and ARORA, C.P. 1993. Prediction transient temperature
distribution during freeze drying of yoghurt. Drying Technol. 11(7),
1863–1883.
SHARMA, N.K. and ARORA, C.P. 1995. Influence of product thickness,
chamber pressure and heating conditions on production rate of freeze-
dried yoghurt. Int. J. Refrig. (UK) 18(5), 297–307.
SHEEHAN, P. and LIAPIS, A.I. 1998. Modeling of primary and secondary
drying stages of freeze drying of pharmaceutical products in vials:
Numerical results obtained from solution of dynamic and spatially mul-
tidimensional lyophilization model for different operational policies.
Biotechnol. Bioeng. 60(6), 712–728.
SIMATOS, D., BLOND, G., DAUVOIS, P. and SAUVAGEOT, F. 1974. La
Lyophilisation: Principes et Applications. Collection de l’Association
Nationale de la Recherche Technique, Paris, France.
132 S. KHALLOUFI, J.-L. ROBERT and C. RATTI

SKELLAND, A.H.P. 1974. Molecular diffusivities. In Diffusional Mass


Transfer, 1st Ed., Chapter 3, p. 510, Wiley, New York, NY.
WANG, Z.H. and SHI, M.H. 1997. Effects of heating methods on vacuum
freeze drying. Drying Technol. 15(5), 1475–1498.
WANG, Z.H. and SHI, M.H. 1998. Numerical study on sublimation con-
densation phenomena during microwave freeze drying. Chem. Eng. Sci.
53(18), 3189–3197.
WOLFF, E. and GIBERT, H. 1998. Développements technologiques
nouveaux en lyophilisation. J. Food Eng. 8, 91–108.

Potrebbero piacerti anche