You are on page 1of 13

Journal of Food Engineering 161 (2015) 55–67

Contents lists available at ScienceDirect

Journal of Food Engineering


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

A mathematical model of multi-dimensional freeze-drying for food


products
Kyuya Nakagawa a,⇑, Takaaki Ochiai b
a
Division of Food Science and Biotechnology, Graduate School of Agriculture, Kyoto University, Sakyo-ku, Kitashirakawa Oiwakecho, Kyoto 606-8502, Japan
b
Amano Jitsugyo Co., Ltd, 4215 Satomi Satosho, Asaguchi-gun, Okayama 719-0393, Japan

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

Article history: A simple mathematical model was developed to predict the kinetics of freeze-drying where sublimation
Received 15 January 2015 progresses multi-dimensionally in a product, and the model was adopted to the freeze-drying of peeled
Received in revised form 28 March 2015 apple cubes. This model consists of classical heat and mass transfer equations, which are solved by
Accepted 29 March 2015
assuming a quasi-steady-state energy balance at the sublimation interface. The radiative heat coefficient
Available online 3 April 2015
and mass transfer coefficient between the chamber and the condenser are included in the model, and
were experimentally obtained by carrying out the ice sublimation tests with the freeze-dryer employed.
Keywords:
The mass transfer property is a specification of a given freeze-dryer. This work suggested that it strongly
Freeze-drying
Lyophilization
influenced the drying kinetics and it would be an important parameter to meet scaling-up issue. When
Mathematical model the sublimation progresses multi-dimensionally, the surface areas of the product exterior, the product
Multi-dimension bottom and the sublimation interface varied as a function of the extent of drying. These relationships
Food drying were obtained from experiments with freeze-drying of peeled apple cubes and employed for estimating
the mean thickness of the dried and frozen layers with a hollow spherical model proposed in the study.
The model equations were solved using commercial spreadsheet software, and the weight loss during
drying was predicted as a function of time. The calculation results were in good agreement with the
experimental data. This approach allows to set a heating program and/or localized variances of
microstructural parameters by manual input, and easily simulate drying kinetics and temperature
history. This makes big advantage for planning a desired drying operation.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction equation of the sublimated vapor in the dried cake and the heat
balance equation at the sublimation interface. Assuming that the
Freeze-drying, also denoted as lyophilization, has long been heat supplied to the sublimation interface is used only for sublima-
known as the best drying method for preserving the original prop- tion, the solutions to the differential equations give predictions of
erties of the resultant dried product. It favorably maintains the bio- the sublimation rates. A clear example can be seen in the report of
logical activity of pharmaceuticals, flavors, and aromas of food, Toei et al. (1975), which explains the instability of the sublimation
viability of cell biological products, etc. However, it is also known interface with this simple model, by introducing the concept of
to be an expensive drying method, due to its complexity in opera- zone sublimation. Theoretical approaches developed by the Liapis
tion and long drying time. After the water in a product is frozen, research group were based on classical mass and heat transfer
the ice is sublimated. The sublimated water is trapped by a con- equations, and simulated realistic freeze-drying processes for
densation device, and removed by thawing after completion of industry, especially for pharmaceuticals (Liapis and Bruttini,
the freeze-drying operation. Studies on freeze-drying have a long 1994, 1995; Millman et al., 1985; Sadikoglu and Liapis, 1997).
history, since great contributions on the basic principles were They modeled a freeze-drying system of a frozen solution in a glass
given by Flosdorf in early 1940s (Flosdorf, 1949). A variety of prod- vial, assuming one-dimensional movement of the sublimation
ucts has been investigated, and many theoretical approaches have interface. The model counted conductive heat flow from the shelf
also been taken. Mathematical models of freeze-drying consist of and radiation heat to the glass vial. This modeling approach was
simultaneous differential equations; for example, the mass balance later improved, including multi-dimensional freeze-drying pro-
cesses, so that the primary and secondary drying stages could be
⇑ Corresponding author. successfully simulated (Sheehan and Liapis, 1998).

http://dx.doi.org/10.1016/j.jfoodeng.2015.03.033
0260-8774/Ó 2015 Elsevier Ltd. All rights reserved.
56 K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67

Nomenclature

Aext external surface area of product contacting to ambience Q rad radiative heat to product (W)
(m2) Q sh
Rad radiative heat to shelf (basket) (W)
Abtm surface area of frozen layer contacting the sample tray Q sh heat flow from shelf (basket) to product (W)
(m2) Qs heat consumption at sublimation interface (W)
Ai effective surface area receiving radiative heat (m2) R ideal gas constant (J mol/K)
As surface area of sublimation interface (m2) t time (s)
dp mean pore diameter (m) T btm temperature of product bottom (K)
ed dried layer thickness at the product bottom (m) T ch temperature of chamber wall (K)
DA effective diffusion coefficient in dried layer (m2/s) Tc temperature of condenser (K)
Dk Knudsen diffusion coefficient (m2/s) T ext temperature of product external surface (K)
DH latent heat of sublimation (J/kg) Th temperature of electric heater (K)
hbtm overall heat transfer coefficient at the frozen bottom of Tp temperature of product (during ice sublimation test) (K)
the product (W/m2 K) T sh temperature of shelf (basket) (K)
hsh heat transfer coefficient at the bottom of the product Ts temperature at sublimation interface (K)
(W/m2 K) V ff volume of frozen fraction (m3)
hR1—4 radiative heat transfer coefficient (W/ K)
K cc mass transfer coefficient from product to condenser Greek letters
(m3/s) e porosity (–)
m mass of ice fraction (kg) ei emissivity (–)
MW molecular weight of water (kg/mol) r Stefan–Bolzmann constant (J/s m2 K4)
Pc pressure at condenser surface (Pa) kdry thermal conductivity of dried layer (W/m K)
Pmeas pressure measured by pressure gauge (Pa) kice thermal conductivity of frozen layer (W/m K)
Ps pressure at sublimation interface (Pa) g conversion applied to coefficients obtained from ice
Q dry heat flow through dried layer (W) sublimation test (–)
Q ice heat flow through frozen layer (W) s tortuosity (–)

Many mathematical models of freeze-drying have been applied sublimation of the sample is in practice controlled only by the
to pharmaceutical freeze-drying, where a product solution is set in heating devices. A detailed study of radiative heat for freeze-drying
a glass vial and sublimed from the solution top by receiving con- was made by Zamzow and Marshall (1952). The heating program is
ductive heat from the vial bottom. Sublimation of the frozen solu- therefore of critical importance for a practical drying operation.
tion in the vial progresses rather one-dimensionally from the top Drying engineers test their operational conditions with lab-scale
to bottom. A simple model thus assumes one-dimensional progress and/or pilot-scale dryers in order to fix a condition for actual pro-
of sublimation. When simulating the curvature of the sublimation duction. It is usually difficult to find an optimal condition with an
interface, well sophisticated computational method is required. actual scale, especially for an expensive product and in the case
Finite element analysis is a successful method, however it could where the production mass is enormous. A mathematical model
make a difficulty in applying moving boundary to the sublimation could be beneficial for exploring an optimal condition and reducing
interface (Ferguson et al., 1993; Mascarenhas et al., 1997). When the time required for validation. This is a common issue for food
products such as fruits or vegetables placed directly on a shelf engineers who are required to produce various products with a
are dried, sublimation progresses multi-dimensionally, depending same facility. An optimized heating program for drying a given
on geometries of the products. Therefore, a modeling approach to product is not always optimal for other products. Furthermore,
calculate the velocity of the sublimation interface based on classi- an optimal condition for a given dryer is not always optimal for
cal mass and heat transfer equations, as is described above, could the others. The motivation of this study is, thus, the development
be significantly complicated. An empirical and statistical approach of a mathematical model that includes specification parameters
using the response surface method would be a solution to obtain of the dryer and the conditions for operation.
parameters to realize optimized operation (Hammami and René, The mass transfer resistance between the chamber and the con-
1997). This is a method to predict optimal solution from the fitting denser often limits the sublimation rate of the drying products,
surfaces obtained from trial experiments. This approach does not particularly in situations when an excess amount of ice is subli-
give the status of products during drying, but gives a predicted mated. This resistance is a specification of a freeze-dryer, and its
property of the resultant dried product. Therefore, the database value would be expected to vary greatly between lab-scale and
used for the prediction must be made from the system that we industrial-scale freeze-dryers. Therefore, a model that includes this
wish to optimize. coefficient could be useful for tackling scale-up issues.
A temperature controllable shelf is a common component of In this study, a simple mathematical model was developed to
pharmaceutical freeze-dryers, which enables freezing in a drying predict the drying process in a batch freeze-dryer, which is widely
chamber and precise control of product temperature during drying, used for food products. The radiative heat coefficient and mass
by heat conduction between the shelf and products through a vial. transfer coefficient between the chamber and the condenser were
Freeze-dryers that do not have a cooling device in the drying included in the model, and they are determined as specified coeffi-
chamber are commonly used for freeze-drying of food. Products cients of the freeze-dryer by the water sublimation experiments.
are frozen in a refrigeration room, and immediately transferred The model was validated by freeze-drying experiments with apple
to a freeze-dryer. The dryer is usually equipped with heating cubes. The calculation carried out on a Microsoft Excel spreadsheet
devices to supply radiative heat to a drying product. The is also presented as an appendix.
K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67 57

2. Mathematical modeling By defining hRT and T Rad as follows,


hRT ¼ hR1 þ hR2 ð7Þ
2.1. Thermal flow equations
hR1 T h þ hR2 T ch
T Rad ¼ ð8Þ
Heat transfer equations were modeled as illustrated in Fig. 1A. hR1 þ hR2
The radiative heat delivered to a product, Q rad , originated from sev- Eq. (6) can be written as:
eral radiation sources, such that it can be described by the follow-
ing summation formula. Q rad ¼ hRT ðT Rad  T ext Þ ð9Þ
X   As illustrated in Fig. 1A, heat is transferred to the sublimation
Q rad ¼ Ai i r T 4i  T 4ext ð1Þ interface by conduction via the dried and frozen layers. These con-
i
ductive heat flow can be calculated with the thickness of these
T i is the temperature of the radiation source i. Ai is the effective sur- layers, however, it is not easy to estimate a mean thickness that
face area that receives radiative heat from the heat source i. ei is the is effective for the mass and heat transfer where the flows were
emissivity and r is the Stefan–Bolzmann constant. In this study, we multi directional. So the authors attempted to estimate the mean
consider two heat sources, namely, the heating plate (T h ) and the thicknesses of the layers by the sublimation, external and bottom
chamber wall (T ch ). Thus, surface areas of a product. The surface area of the sublimation
    interface, As , and the surface area of the frozen layer that con-
Q rad ¼ A1 1 r T 4h  T 4ext þ A2 2 r T 4ch  T 4ext ð2Þ tributes to the conductive heat flow from the sample tray, Abtm ,
vary as a function of the degree of drying. In the present study,
The right side of this equation can be expanded and simplified as we did not take the shrinkage of product into account, so the sur-
follows. face area of the product surface that contributes to the radiation
heat flow, Aext , is not variable. In order to model heat transfer in
Q rad ¼ hR1 ðT h  T ext Þ þ hR2 ðT ch  T ext Þ ð3Þ the present system, a hollow sphere model was introduced, as
where hR1 and hR2 are defined as the radiation heat transfer coeffi- illustrated in Fig. 2. This hollow sphere has two layers; that is,
cients, namely the outer layer corresponds to the dried cake and the inner to
  the frozen layer. The outer surface area of the sphere corresponds
hR1 ¼ A1 1 r T 2h þ T 2ext ðT h þ T ext Þ ð4Þ to Aext , and the inner core surface area to Abtm . The surface area that
  divides two layers corresponds to As . Considering heat flux in a
hR2 ¼ A2 2 r T 2ch þ T 2ext ðT ch þ T ext Þ ð5Þ spherical coordinate system (see Appendix A), the conductive heat
flow from the product surface to the sublimation interface via the
Although these coefficients are functions of T h ; T ext , and T ch , they are dried layer, Q dry ð¼ Q rad Þ, can be written as:
regarded as constant values in the present calculation, so that Q rad
can be expressed: 2kdry
Q dry ¼  pffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffiffiffiffi ðT ext  T s Þ ð10Þ
  1= pAs  1= pAext
hR1 T h þ hR2 T ch
Q rad ¼ ðhR1 þ hR2 Þ  T ext ð6Þ The conductive heat flow from the frozen bottom to the sublimation
hR1 þ hR2
interface can be written as:

Fig. 1. Schematic illustration of heat transfer (A) and mass transfer (B).
58 K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67

Fig. 2. Hollow sphere model for multi-dimensional drying.

2kice As addressed in the introduction, we are targeting a freeze-


Q ice ¼  pffiffiffiffiffiffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffi ðT s  T btm Þ ð11Þ
1= pAbtm  1= pAs dryer with a shelf of which temperature is not controllable. In this
case, the shelf temperature is determined by the heat balance
Products are placed in the sample tray on the basket, to consider the between the radiative heat flow and the conductive heat flow from
conductive heat from the basket to the bottom of the product, the shelf to the products. Here, we consider radiative heats from
Q sh ð¼ Q ice Þ. the heater and the chamber wall. The radiative heat transfer coeffi-
cients hR3 and hR4 are defined to describe the heat balance in the
Q sh ¼ Abtm hbtm ðT btm  T sh Þ ð12Þ shelf.
When drying progresses multi-dimensionally, the bottom of the  
product is dried, especially, in the later stage of drying. This dried hR3 ¼ A3 3 r T 2h þ T 2sh ðT h þ T sh Þ ð15Þ
layer formation at the bottom affects to the conductive heat flow  
hR4 ¼ A4 4 r T 2ch þ T 2sh ðT ch þ T sh Þ ð16Þ
rate. Therefore, hbtm is the overall heat transfer coefficient at the fro-
zen bottom of the product that is in contact with the basket via the In this calculation, we consider these coefficients to be constant.
sample tray and the dried layer of the product bottom, namely The radiative heat delivered to the shelf can be written as:
1 1 ed
¼ þ ð13Þ Q sh
Rad ¼ hR3 ðT h  T sh Þ þ hR4 ðT ch  T sh Þ
hbtm hsh kdry  
hR3 T h þ hR4 T ch
where hsh is the heat transfer coefficient at the bottom of the pro- ¼ ðhR3 þ hR4 Þ  T sh ð17Þ
hR3 þ hR4
duct that contacts to the basket via the sample tray. The dried layer
thickness at the product bottom, ed , is regarded as a function of the This equation can be simplified as follows.
degree of drying, in the present study. hRsh ¼ hR3 þ hR4 ð18Þ
By using Eqs. (9)–(13), the total heat delivered to the sublima-
hR3 T h þ hR4 T ch
tion interface, Q s , can thus be expressed as follows. T Rsh ¼ ð19Þ
hR3 þ hR4
Q s ¼ Q dry  Q ice Q sh
Rad ¼ hRsh ðT Rsh  T sh Þ ð20Þ
T  Ts T  T sh
¼  pRad
ffiffiffiffiffiffi  pffiffiffiffiffiffiffiffi   spffiffiffiffiffiffiffiffiffi   pffiffiffiffiffiffi ð14Þ When the system is at a stationary state, the radiative heat is
1 1= pAs  1= pAext 1 1= pAbtm  1= pAs
hRT
þ 2kdry Abtm hbtm
þ 2kice
equivalent to the conductive heat to the products, namely:

Q s , corresponds to the heat used for sublimation at the steady state. Q sh


Rad ¼ Q sh ¼ Q ice ð21Þ
K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67 59

Therefore, T sh can be related to T s with the following equation. and the time required to complete sublimation of the fraction can
be determined. The calculated values in each fraction are finally
T Rsh þ aT s
T sh ¼ combined to produce a drying curve, temperature profile, etc. The
1þa
various coefficients required for the calculation were obtained by
1=hRsh ð22Þ
a¼  pffiffiffiffiffiffiffiffiffi  pffiffiffiffiffiffi ice sublimation experiments, as explained in the following sections.
1 1= pAbtm  1= pAs
Abtm h
þ 2k
btm dry
2.5. Acquisition of essential coefficients by ice sublimation tests

   
2.2. Mass flow equations The coefficients hR1 ; hR2 ; hR3 ; hR4 ; hsh , and K cc are experimentally
obtained. This section explains the governing equations employed
The sublimated water transfers through the dried layer, the for the parameter acquisitions. Ice sublimation tests are conducted
boundary layer at the product surface, the chamber, and is then in a freeze-dryer, where the temperature of the sublimating ice, T p ,
finally condensed on the condenser. The mass flux in the dried cake the sublimation rate, dm , and the shelf temperature, T sh , are col-
dt
can be expressed with the diffusion coefficient, DA . The mass flux lected with the selected operating conditions, the heater tempera-
from the product surface to the condenser in the chamber is usu- ture, T h , the chamber wall temperature, T ch , and the condenser
ally complicated because it is highly dependent on the machine temperature, T c . The heat input, Q s , to the sublimating ice can be
assembly and geometry. In this study, therefore, we did not con- expressed as:
sider the detailed combination of the mass transfers (i.e., in the
       
boundary layer at the product surface, in the chamber, resistance Q s ¼ Q rad  Q sh ¼ hR1 T h  T p þ hR2 T ch  T p þ Abtm hsh T p  T sh
at the piping and in the pipe, in the condenser and the boundary ð28Þ
layer at the condenser surface). Instead, we introduced, as illus-
trated in Fig. 1B, a mass transfer coefficient between the product This heat flow is balanced to the enthalpy flow due to sublimation.
surface and the condenser, K cc , namely dm
Q s ¼ DH ð29Þ
dm C s  C con dt
¼  pffiffiffiffiffiffi  pffiffiffiffiffiffiffiffi ð23Þ  
dt 1= pAs  1= pAext Using these equations, hR1 ; hR2 , and hsh are determined with experi-
2DA
þ K1cc
mentally obtained values of T p ; dm
dt
; T sh ; T heat , and T cham . The radiation
 
Here, K cc includes the effective surface area that concerns the over- heat coefficients (hR1 and hR2 Þ contain effective surface areas that
all mass transfers from the product surface to the condenser. This represent the product surface receiving radiative heat. The effective
coefficient will be highly dependent on the dryer design. Eq. (23) surface area is closely related to, although not perfectly equivalent
can be written in terms of vapor pressures and temperatures, based to, the visible surface areas of the products, and depends on the lay-
on the ideal gas equation. out and geometry of the products on the shelf. In order to apply
these coefficients (obtained from ice sublimation experiments) to
dm M W P =T  Pc =T c
¼  pffiffiffiffiffisffi s pffiffiffiffiffiffiffiffi  ð24Þ model the drying process of a product, a conversion, gRi , is
dt R 1 pAs  1 pAext
2DA
þ K1cc introduced.
gRi hRi ¼ hRi ð30Þ
2.3. Diffusion coefficient in the dried cake If freeze-drying is carried out with the same geometry as the ice
used for the sublimation test, gR should be equal to 1. The value
The diffusion coefficient, DA , in the dried cake is calculated with of gR varies depending on the size, number, geometry, and layout
the Knudsen diffusion coefficient, Dk . This is dependent on the ice of the products. In the present study, the system was regarded as
microstructure, which can be described in terms of variables such a gray body. The temperature range of the present system was quite
as pore size, dp , tortuosity in the cake, s, and porosity in the cake, e. limited, so this assumption could be fair. But, in case this model is
e applied to a product with largely different surface color, smooth-
DA ¼ DK ð25Þ
s ness, and shape, further conversion can be made by adjusting the
where the Knudsen diffusivity is expressed by:
gR values.
 
sffiffiffiffiffiffiffiffiffiffiffiffi The values of hR3 and hR4 can be estimated with the following
1 8RTs procedure. When the system is at a stationary state,
DK ¼ d ð26Þ
3 pM W p Q sh
Rad ¼ Q sh ð31Þ
Thus,
2.4. Calculation  
hR3 T h þ hR4 T ch  
ðhR3 þ hR4 Þ  T sh ¼ Abtm hsh T sh  T p ð32Þ
Assuming that the system is at a quasi-steady-state, the heat hR3 þ hR4
flow is balanced with the mass flow.  
The values for hR3 and hR4 are determined using this equation with
dm values of T p ; dm ; T sh ; T h ; T ch , and hsh . As addressed above, these radia-
Q s ¼ DH ð27Þ dt
dt tive heat coefficients contain effective surface areas that correspond

This equation can relate Eqs. (14) and (24), and the solution gives to the surface of the shelf receiving radiation heat. The values for hR3

the value for T s . The calculation was carried out with a spreadsheet and hR4 would be constant as long as the same shelf is employed for
 
software (Microsoft Excel for Mac 2011, version 14.1.2), as pre- the actual drying process (i.e., hR3 ¼ hR3 ; hR4 ¼ hR4 Þ. However, if the
sented in detail in the appendix. In brief, a product was divided into number and type of shelf is rearranged, these values must be prop-
finite number of volume fractions, assuming that the temperature erly converted.
in each fraction was homogeneous. Calculation was carried out The mass transfer coefficient, K cc , can be determined with the
for each volume fraction to find a value of T s that satisfied Eq. following equation, by applying the experimentally obtained
(27). From the value of T s , the sublimation rate of each fraction, values.
60 K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67

dm K cc M W capacitance manometers (622B, MKS, USA) at the points shown


¼ ðPs =T s  Pc =T c Þ ð33Þ
dt R in Fig. 3. The gas flow in the system could be adjusted by the dia-
phragm valve. The evacuation rate, sublimation rate, leak rate, and
It should be noted that the design of the drying apparatus and the
condensation rate always balance at a steady state of drying. This
location of the products in the dryer usually affect the efficiencies
balance can be adjusted by the diaphragm valve. In this study,
of the heat and mass transfers. The inherent characteristics of the
the pressure, Pmeas , was adjusted to 20, 30, and 40 Pa. It should
dryer are incorporated into the estimated values of hR1 ; hR2 ; hR3 ;
be noted that these values indicate pressures at the measured
hR4 ; hsh , and K cc . It is expected that these parameters will enable an
point. However, for the purpose of this study, they are taken as
accurate prediction of drying kinetics with the present simplified
representative of the pressure in the dryer system.
equation sets.
All specimens were frozen with convective air in a refrigeration
device (SU-262, Espec Co., Ltd., Japan) at a controlled temperature
3. Experiments (held at 20 °C for 10 min, cooled down to 40 °C at 2.0 K/min, and
held at 40 °C for around 1 h). Six cubes of frozen apple were
3.1. Materials orderly placed on a thin plastic tray (50  60 mm side lengths,
20 cm height). The tray was set on the shelf in the freeze-dryer,
Distilled water was used for the ice sublimation test. Apple (Fuji which was immediately evacuated to carry out freeze-drying.
apple, purchased at a domestic market) was cut into a cubic shape The dryer chamber was covered with a sheet of black curtain in
(ca 1.1 cm of side length) and used for the drying test. order to avoid unfavorable external radiation. Drying experiments
were conducted several times (n = 2–3) for each condition, where
3.2. Freeze-drying experiments the coefficient of variation (CV) was less than 4.0%.

A freeze-dryer was set up with the geometry shown in Fig. 3. A 3.3. Protocol for ice sublimation tests
basket made from 0.5-mm aluminum plate was suspended from a
load-cell (A&D, Japan). The weight loss of specimens on the basket In the present study, essential heat and mass transfer coeffi-
during sublimation was monitored with the load-cell, which was cients were obtained by the ice sublimation test. 50 mL of distilled
calibrated under vacuum at around 10 Pa. The specimens were water was set in a plastic tray (50  60 mm side lengths, 20 cm
radiatively heated by electric heaters set at 15 mm above and height, ca 18 cm depth of water), and frozen in a refrigerator at
below the specimens. A ceramic electric heater was adhered to 40 °C. This ice was set in the freeze-dryer, and weight loss during
the 70  80 mm anodized aluminum plate, and its surface tem- sublimation was recorded under various operating conditions.
perature was controlled by the external controller. The tempera- Heater temperature settings of 40, 60, and 90 °C, chamber pressure
tures during sublimation were measured by directly set (Pmeas ) settings of 20, 30, and 40 Pa, and condenser temperature
thermocouples (5SC-TT-K-30-36, Omega, USA) in the specimens settings of 40 and 60 °C were applied. The temperature of the
(thin thermocouples were inserted prior to freezing). This experi- sublimating ice, T p , was separately measured with three
ment was conducted separately from the weight change measure- thermocouples set at the points 8, 12 and 16 mm from the bottom
ments. Pressure in the freeze-dryer was monitored with of the tray. The temperature of the basket, T sh , was also measured

Fig. 3. Experimental freeze-drying setting.


K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67 61

with the sublimating ices. The sublimation tests were conducted parameters for hR1 ; hR2 , and hsh that satisfy Eqs. (28) and (29) were
for 30 min, and the linearity of the weight loss time was confirmed obtained by using the generalized reduced gradient (GRG2) code
at each combination of operating conditions. run on the Microsoft Excel Solver tool. Considering that the heat
transfer coefficients take constant values under the different heat-
3.4. SEM observation ing temperature, condenser temperatures and evacuation rates,
optimal parameters that satisfy all the data sets in Table 1 were
The microstructures of the freeze-dried apples were observed obtained. Subsequently, values for hR3 and hR4 were obtained in
by scanning electron microscopy (SEM, S-2400; Hitachi, Japan). the same manner, based on Eq. (32), and are listed in Table 2.
The images obtained were used to estimate mean pore size, dp , Appendix will be a help to set up this calculation on a spread sheet.
in the dried cake. The parameter values used for the calculation are listed in Table 3.
The choice of data set for obtaining and optimized value of K cc is
3.5. Observation of sublimation interface critical. As noted above, K cc incorporates a range of distinct mass
transfer resistances, such as that of the chamber, the exhaust open-
Partially freeze-dried specimens were taken from the dryer, and ing, tubing, condenser room, and condenser surface. The evacua-
their cutting surfaces were observed and sketched. Observations tion rate, sublimation rate, leak rate and condensation rate are
were quickly carried out on the various cross sectional cuts with balanced at a steady state. The evacuation rate can also be consid-
numbers of specimens, and the results were summarized in a ered almost equal to the leak rate, and the sublimation rate is equal
sketch. Based on the sketches, 3D images of the frozen fractions to the condensation rate. The pressure, P meas , is determined by the
were redrawn with CAD software (AUTODESK, 123D Design evacuation rate, leak rate, and the partial pressure of the water
21.1.5). The volume of the frozen fraction, V ff , and the correspond- vapor at the measured position. This pressure value relates to the
ing surface areas of the frozen layer in contact with the sample fluid flow in the drying system, thus it could affect the value of
tray, Abtm , and the sublimation interface, As , were estimated using K cc . The condenser temperature, T c , provides a driving force for
this software. water vapor transfer from the product to the condenser surface.
This temperature is a major factor that controls the mass flux in
4. Results and discussion the boundary layer of the condenser surface. The electric heater
is a major radiation source, so the temperature, T h , is the main con-
4.1. Essential coefficients obtained via sublimation experiments trol of the temperature at the product surface, and consequently
controls the sublimation rate. Considering that the fluid flow (i.e.,
Ice sublimation experiments were conducted to determine the convection) in the chamber and condenser are influenced by T h
values for hR1 ; hR2 ; hR3 ; hR4 ; hsh , and K cc in the present freeze-drying and T c , these temperatures could affect the value of K cc .
system. The measurement results are listed in Table 1. The optimal However, we believe this influence is negligible. In fact, the

Table 3
Table 1 Values used for calculation.
Results of ice sublimation experiments.
gR ð¼ gR1 ; gR2 Þ 0.62
ID T c (°C) P meas (Pa) T h (°C) dm
(kg/s) T p (°C) T sh (°C) 
hR1 ð¼ gR hR1 Þ (J/K s) 2.1  102
dt

1 40 20 40 1.2  106 14.1 2.7 hR2 ð¼ gR hR2 Þ (J/K s) 1.1  102

60 1.4  106 12.8 7.5 hR3 ð¼ hR3 Þ (J/K s) 1.3  102

90 1.9  106 12.1 13.6 hR4 ð¼ hR4 Þ (J/K s) 5.1  103
hsh (W/m2 K) 12.9
2 40 30 40 1.1  106 9.7 4.4
K cc (m3/s) 7.4  104
60 1.4  106 8.5 9.3
DA (m2/s) 7.4  104
90 1.8  106 6.7 16.5
kice (W/K) 1.4
3 40 40 40 1.0  106 7.8 5.4 kdry (W/K) 0.09
60 1.3  106 6.4 10.7 Density (kg/m3) 790
90 1.72  106 4.7 19.0 Mw (kg/mol) 0.018
4 60 20 40 1.1  106 13.8 0.8 R (J/K/mol) 8.314
60 1.3  106 11.9 3.7 DHf (J/kg) 2,800,000
90 1.8  106 9.8 10.2 Side length, L (m) 0.011
Abtm (0) (m2) 7.26  104
5 60 30 40 1.0  106 11.8 0.5 Aext (0) (m2) 3.63  103
60 1.3  106 9.9 6.7 e (–) 0.85
90 1.7  106 8.1 15.8 Number of fraction (–) 100
6 60 40 40 9.6  107 7.6 5.2 Total volume of product (m3) 7.986  106
60 1.2  106 5.7 10.1 T ch (K) 298.15
90 1.6  106 3.4 18.9 T c (K) 213.15

Table 2
Estimated coefficient values.

ID 
hR1 (J/K s)

hR2 (J/K s)

hR3 (J/K s)

hR4 (J/K s) hsh (W/m2 K) K cc (m3/s)
2 2 2 3
1 3.4  10 1.7  10 1.3  10 5.1  10 12.9 9.2  104
2 5.4  104
3 4.6  104
4 7.4  104
5 6.0  104
6 4.0  104
62 K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67

Fig. 4. SEM images of freeze-dried apples.

optimal K cc values were calculated from the data sets that have 4.4. Estimation of variable surface areas during sublimation
common T h settings, and we confirmed that the variance of the
value was not significant (i.e., K cc ¼ 5:4  104 , 5.9  104, In the present mathematical model, we have introduced as
6.7  104 were obtained for T h = 40, 60, 90 °C, respectively). parameters the sublimation surface area, As , and surface area of
Therefore, the optimal values were obtained from the data set with the frozen layer contacting to the sample tray, Abtm . At the begin-
common Pmeas . The convective flow in the dryer would be different ning of drying, the value of As is almost identical to the external
at different P meas . As seen in Table 2, the value of K cc is likely to surface area contacting with ambient space, Aext . As drying pro-
depend on the P meas values; however, the details must be clarified gresses, As changes as a consequence of sublimation, and that
in a future study. In the present study, a value of 7.4  104 was change is dependent on the product geometry. Inhomogeneity of
adopted for K cc , regarding the selected condition for drying apple sublimation and/or shrinkage due to drying may also alter the
cubes. sublimation surface area. However, we have ignored the influence
of these effects in this work. We introduce the functions, f ðV ff Þ and
gðV ff Þ, that relate As and Abtm to the volume fraction of the frozen
4.2. Diffusion coefficient in the dried cake
zone, V ff .
As is evident in the SEM images shown in Fig. 4, the pore sizes in f ðV ff Þ ¼ As ðV ff Þ=As ð0Þ ð34Þ
the dried cake varied in the 100–300 lm range. An image analysis
g ðV ff Þ ¼ Abtm ðV ff Þ=Abtm ð0Þ ð35Þ
gave a mean pore size 188 lm, which was adopted as the value of
dp . The solid content of the apples used in this study was around In this study, f ðV ff Þ and gðV ff Þ were experimentally obtained. Some
0.15 [kg/kg]. Thus, the porosity, e, was estimated to be roughly sketches of the frozen interfaces observed on partially freeze-dried
0.85. This value is close to those reported in the literature apple cubes are shown in Fig. 5. These sketches were redrawn in a
(Karathanos et al., 1996; Krokida et al., 1998; Li et al., 2007). CAD software to estimate Abtm and As values. The plots of
Tortuosity, s, should be determined carefully as a representative Abtm ðV ff Þ=Abtm ð0Þ and As ðV ff Þ=As ð0Þ as a function of V ff are shown
value of porous microstructure. Pores in the freeze-dried cake are in Fig. 6. The following expressions for f ðV ff Þ and gðV ff Þ were
a skeleton of ice microstructure, thus they are significantly affected obtained by fitting functions to these plots.
by the manner of ice formation (Nakagawa et al., 2006, 2007,
2010). It could be meaningful to assess the relationship between f ðV ff =V ff ð0ÞÞ ¼ 1:3ðV ff =V ff ð0ÞÞ3  2:2ðV ff =V ff ð0ÞÞ2 þ 1:9ðV ff =V ff ð0ÞÞ
ice microstructure and tortuosity, however, in the present work, ð36Þ
the value of 2 was borrowed from a former report without making
a deep investigation (Li et al., 2007). With these values of e and s, g ðV ff =V ff ð0ÞÞ ¼ 1:5ðV ff =V ff ð0ÞÞ3 þ 2:0ðV ff =V ff ð0ÞÞ2 þ 0:5ðV ff =V ff ð0ÞÞ
the diffusion coefficients were calculated from Eqs. (25) and (26),
and a value for DA of 7.4  103 [m2/s] was estimated. ð37Þ

These equations were applied to estimate mean dried and frozen


4.3. Physical properties of apples layer thickness based on the hollow sphere model. This model is
valid only when Abtm ; As and Aext values are in the order of
The thermal conductivities of frozen and dried apples reported Abtm < As < Aext . Therefore, the applicability of this model largely
in former works were employed for the calculation (Lozano et al., depends on the product shapes and the trends of change in interface
1979; Ramaswamy and Tung, 1981). Values applied to the frozen geometries.
and dried layers were 1.4 and 0.09 W/K, respectively. The value The sketches in Fig. 5 enable the estimation of values for the
of density of frozen apple (790 kg/m3) was also taken from the dried layer thickness at the product bottom, ed . Drying at the bot-
literature (Ramaswamy and Tung, 1981; Sweat, 1974). tom progressed sluggishly, and the thickness of the dried layer was
K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67 63

Fig. 5. Observation of frozen layer geometries of semi-freeze-dried apples (A) and 3D CAD images redrawn from the observation results (B). The values of V ff =V ff ð0Þ in (A)
were experimentally obtained, whereas the values in (B) were given by the CAD illustrations.

around 0.5–1.0 mm when the volume fraction of frozen layer was


0.08. In the present calculation, we assumed that ed has a linear
relationship with V ff , and it increased from zero to 1.0 mm as V ff
decreased.

ed ¼ 0:001ð1  V ff =V ff ð0ÞÞ ð38Þ

4.5. Calculation results and comparison with experimental data

Calculations were carried out with the properties estimated as


above (see Table 3), and the resultant drying curves are shown in
Fig. 7. Experimentally obtained drying curves are also plotted in
this figure. The CV values from experimentally obtained and simu-
lated drying curves were 3.4%, 2.4% and 2.3% for the heater tem-
perature setting to 60, 90 and 125 °C, respectively. It can be
concluded that the present model is reasonably successful in pre-
dicting the experimental results. With the heater temperature set
to 60 °C, deviations between experiment and predicted results
were seen in the middle stages of drying, whereas no significant
deviations were seen for higher heater temperatures. At the higher
heater temperature settings, the contribution of radiative heat to
the total heat input is higher. The ratios of Q ice to Q dry are plotted
in Fig. 8 as a function of time and remaining moisture content. It
Fig. 6. Acquisition of fitting equations from plots for As and Abtm as a function of can be confirmed that, in the present freeze-drying system, the
volume fraction of frozen layer. radiative heat was the dominant heat source for drying, and that
64 K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67

of equipment. Thus, the position, material, color, etc. could all


affect the values of these coefficients. If the drying chamber is large
enough, the mean effect of radiative heat from the chamber wall
 
becomes negligible, so that hR2 and hR4 have much smaller values
than those used in this study. The heat transfer from the shelf to
the products is reflected in the value of hsh . The arrangement of
product placements (e.g. product layout on a shelf, use of plastic
tray, etc.) in a dryer could be optimized by tuning these values.
 
Higher hR1 ; hR3 , and hsh values indicate improved heat transfer,
but are not always effective for reducing drying time. The calcula-
tion of the present model could assist in the estimation of the
resultant efficacies against the tuning of heat transfer efficiencies.
The mass transfer coefficient from the product to condenser,
K cc , is one of the key values in this model. It reflects the major char-
acteristics of a freeze-drying system. For the system modeled here,
the value of K cc is sufficiently large that the mass transfer velocity
of sublimated moisture is limited by the resistance in the dried
cake. Thus, the sublimation process can be tuned to improve the
sublimation rate and/or mass transfer velocity, for example, by
increasing product temperature, freezing to produce large ice crys-
tals, etc. However, in systems where the value of K cc is smaller, the
Fig. 7. Comparison of experimentally obtained and simulated drying curves.
mass transfer velocity of sublimated moisture may be limited by
the resistance in the drying system. This could be particularly sig-
the contribution of conductive heat from the shelf dramatically nificant in cases where a lot of product is loaded into the dryer, and
decreased after ca 40% of the frozen fraction was sublimated. an excessive amount of moisture is produced by drying.
Interestingly, the profiles did not significantly differ among the dif- Determining the value of K cc is thus important for planning a dry-
ferent heating conditions. It means that the radiation heat and con- ing scheme, and is useful for designing scale-up. An optimal drying
ductive heat synchronously changes depending on the heating condition obtained for a laboratory-scale (or pilot-scale) freeze-
condition. dryer will not be optimal for full-scale dryer. Knowledge of the
K cc value of each dryer allows the conditions that realize the
4.6. Discussion desired drying kinetics to be explored. As noted above, there is
much still to be investigated to understand the contributions to
The present calculation requires various empirical coefficients this mass transfer coefficient in detail. Comparison of the freeze-
and properties. However, it makes the accurate estimation of dry- drying kinetics of dryers that have different K cc values would be
ing kinetics possible with the application of simple equations, and an interesting study.
without requiring any fitting parameters. The coefficients The parameters f ðV ff Þ and gðV ff Þ were of critical importance to
   
hR1 ; hR2 ; hR3 ; hR4 , and hsh relate to heat transfer from the dryer to the calculations in this study. Their introduction makes it possible
products. Their values are largely dependent on the design of the to model the kinetics of multi-dimensional freeze-drying, which

dryer and the layout of the products. The values for hR1 and hR3
 are applicable to most situations where foods and agricultural
relate to the efficiency of the heating devices with which the materials are freeze-dried. In this study, it was assumed that
freeze-dryer is equipped. Heating efficiency depends on the design f ðV ff Þ and gðV ff Þ are simply a function of V ff , regardless of the

Fig. 8. Ratio of conductive heat via frozen layer to radiation heat via dried layer.
K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67 65

Fig. 9. Temperature profiles during freeze-drying (simulated values).

kinetics of drying. This assumption must be validated in future to the spreadsheet and applied to the calculation. The localized
work. In practice, the deformation of the drying product (e.g., variances of ice microstructures that reflect the diffusion coeffi-
due to shrinkage) is likely to have an influence on the drying cients and/or tortuosities in the corresponding dried cakes can also
kinetics. Apple was selected for this work because the degree of be set by manual input. It is easy to generate temperature profiles
its deformation during drying is rather limited. For this reason, during drying, as shown in Fig. 9. Control of the temperature at the
we set the value for Aext to a constant. However, occasional shrink- sublimation interface is important to avoid the risk of collapse.
age of specimens was observed. Shrinkage lead to a decrease in Knowledge of the temperature at the external surface is important
Aext , thus causing an overestimation of the mean thickness of the to predict the loss of product quality during drying (e.g. thermal
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
dried cake (i.e., Aext =4p  As =4pÞ, and resulting in an over- degradation, non-enzymatic browning, etc.). In situations where
estimation of the drying time. The model could be improved by temperature related loss of product quality can be predicted in
allowing Aext to vary as a function of the degree of shrinkage. advance, this model allows to an operation engineer to plan a heat-
However, shrinkage depends largely on the type of food material, ing program that minimizes drying time to avoid quality loss.
so the acquisition of As ðV ff Þ; Abtm ðV ff Þ and Aext ðV ff Þ from various
materials may be a useful future study.
A major advantage of the model that we have presented is its 5. Conclusion
simplicity. The calculation can be performed on common spread-
sheet software, in which it is easy to set parameters manually. A simple mathematical model, based on classical heat and mass
For instance, a heating program (rather than constant values for transfer equations, was developed to predict the kinetics of freeze-
T h , as applied in the results shown in Fig. 7) can easily be input drying where sublimation progresses multi-dimensionally in a
66 K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67

Fig. A. Example of a calculation on spreadsheet software.

product, and the model was adopted to the freeze-drying of peeled coefficient between the chamber and the condenser were applied
apple cubes. to the simulation. And these coefficients were obtained by the
Important features of the present model was that the empiri- ice sublimation tests carried out under various heating tempera-
cally obtained radiative heat coefficient and mass transfer ture and chamber pressure settings. It could be fair to assume that
K. Nakagawa, T. Ochiai / Journal of Food Engineering 161 (2015) 55–67 67

the heat transfer coefficients took constant values under the differ- been divided into 100 fractions. As explained in the main manu-
ent heating temperature, condenser temperatures and evacuation script, As and Abtm can be simply calculated from the empirically
rates. Optimal thermal parameters that satisfied various experi- obtained function of V ff (Step 2). The diffusion coefficient, DA , are
mental data sets were successfully obtained. The result of trial estimated from Eq. (25) (Step 3). DA and T h are the values that
calculations of the mass transfer coefficient values suggested that we can set manually (Step 4). In the present paper, common DA
the values were not largely affected by the heater and condenser (Step 3) and T h (Step 4) values are applied to each fraction.
temperatures but by the evacuation rate. Therefore, fair optimal T rad ; T Rsh , and T sh are respectively calculated by using Eqs. (8),
values of the mass transfer coefficient could be obtained from a (19) and (22) with estimated heat transfer coefficients
   
data set with common chamber pressure settings. hR1 ; hR2 ; hR3 ; hR4 , and hsh (Step 5). T s for each fraction is a trial value
In order to model sublimation process that progresses multi-di- (Step 6). Ps and P c are saturated vapor pressures at T s and T c ,
mensionally, experimentally obtained relationships between the respectively (Step 7). Based on the trial value, Q s is calculated from
surface areas (i.e. the product exterior, the product bottom and the heat flow Eq. (14) (Step 8), and separately calculated DH dm dt
the sublimation interface) and the extent of drying were employed from the mass flow Eq. (24) (Step 9). The value of T s is subse-
for the simulation. The simulation results suggested that the quently optimized to equalize Q s and DH dm (Eq. (29)) by using
dt
relationship was applicable to simulate various drying kinetics. It
the Solver add-in (Step 10). This optimization is carried out for
was thus a fair assumption that the surface areas are functions of
each fraction to determine each T s value. The time required for dry-
the extent of drying. Furthermore, the mean thickness of the dried
ing this fraction can be simply given by dividing the weight of this
and frozen layers could successfully be estimated with the hollow
fraction by dm dt
(Step 11). The calculation of all the fractions in a pro-
spherical model.
duct provides the weight change during drying and the tempera-
The calculation method adopted to use commercial spreadsheet
ture profiles during drying.
software gave fair simulation results. It was not suitable to estimate
an accurate end point of drying. However it allows to set a heating References
program and/or localized variances of microstructural parameters
by manual input. It provides an easy-to-use simulation platform Ferguson, W.J., Lewis, R.W., Tömösy, L., 1993. A finite element analysis of freeze-
that gives drying kinetics and temperature histories of products. drying of a coffee sample. Comput. Methods Appl. Mech. Eng. 108 (34), 341–
352.
Flosdorf, E.W., 1949. Freeze-Drying: Drying by Sublimation. Reinhold, New York.
Appendix A. Conductive heat equation in a hollow sphere at Hammami, C., René, F., 1997. Determination of freeze-drying process variables for
strawberries. J. Food Eng. 32 (2), 133–154.
steady state
Karathanos, V.T., Kanellopoulos, N.K., Belessiotis, V.G., 1996. Development of porous
structure during air drying of agricultural plant products. J. Food Eng. 29 (2),
We consider a hollow sphere in spherical coordinate with inner 167–183.
radius R1 , temperature T 1 and outer radius R2 , temperature T 2 . At Krokida, M.K., Karathanos, V.T., Maroulis, Z.B., 1998. Effect of freeze-drying
conditions on shrinkage and porosity of dehydrated agricultural products. J.
steady state, the heat equation can be written as: Food Eng. 35 (4), 369–380.
  Liapis, A.I., Bruttini, R., 1994. A theory for the primary and secondary drying stages
d dT
4pr 2 k ¼0 ðA:1Þ of the freeze-drying of pharmaceutical crystalline and amorphous solutes:
dr dr comparison between experimental data and theory. Sep. Technol. 4 (3), 144–
155.
This equation gives, Liapis, A.I., Bruttini, R., 1995. Freeze-drying of pharmaceutical crystalline and
amorphous solutes in vials: dynamic multi-dimensional models of the primary
C1 and secondary drying stages and qualitative features of the moving interface.
TðrÞ ¼ þ C2 ðA:2Þ Drying Technol. 13 (1–2), 43–72.
kr
Li, S., Stawczyk, J., Zbicinski, I., 2007. CFD model of apple atmospheric freeze drying
where the constants C 1 and C 2 can be determined with the bound- at low temperature. Drying Technol. 25 (7–8), 1331–1339.
Lozano, J.E., Urbicain, M.J., Rotstein, E., 1979. Thermal conductivity of apples as a
ary conditions: T ¼ T 1 at r ¼ R1 and T ¼ T 2 at r ¼ R2 . The heat flux function of moisture content. J. Food Sci. 44 (1), 198–199.
qðrÞ can be written as: Mascarenhas, W.J., Akay, H.U., Pikal, M.J., 1997. A computational model for finite
    element analysis of the freeze-drying process. Comput. Methods Appl. Mech.
dT k T2  T1 Eng. 148 (1–2), 105–124.
qðr Þ ¼ k ¼ 2 R1 R2 ðA:3Þ Millman, M.J., Liapis, A.I., Marchello, J.M., 1985. An analysis of the lyophilization
dr r R1  R2
process using a sorption–sublimation model and various operational policies.
At a steady state, heat flow at r ¼ R1 is equal to that at r ¼ R2 . So the AIChE J. 31 (10), 1594–1604.
Nakagawa, K., Hottot, A., Vessot, S., Andrieu, J., 2006. Influence of controlled
heat flow Q can be written as: nucleation by ultrasounds on ice morphology of frozen formulations for
pharmaceutical proteins freeze-drying. Chem. Eng. Process. 45 (9), 783–791.
2k
Q ¼ Q ðR1 Þ ¼ Q ðR2 Þ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðT 2  T 1 Þ ðA:4Þ Nakagawa, K., Hottot, A., Vessot, S., Andrieu, J., 2007. Modeling of freezing step
1=pA1  1=pA2 during freeze-drying of drugs in vials. AIChE J. 53 (5), 1362–1372.
Nakagawa, K., Thongprachan, N., Charinpanitkul, T., Tanthapanichakoon, W., 2010.
where A1 and A2 are inner and outer surface areas of the hollow Ice crystal formation in the carbon nanotube suspension: a modelling approach.
Chem. Eng. Sci. 65 (4), 1438–1451.
sphere, namely:
Ramaswamy, H.S., Tung, M.A., 1981. Thermophysical properties of apples in relation
to freezing. J. Food Sci. 46 (3), 724–728.
Ai ¼ 4pR2i ðA:5Þ Sadikoglu, H., Liapis, A.I., 1997. Mathematical modelling 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
Appendix B. Calculation on spreadsheet (calculation of drying experimental data. Drying Technol. 15 (34), 791–810.
Sheehan, P., Liapis, A.I., 1998. Modeling of the primary and secondary drying stages
curve) of the freeze drying of pharmaceutical products in vials: numerical results
obtained from the solution of a dynamic and spatially multi-dimensional
The drying curve is calculated on commercial spreadsheet soft- lyophilization model for different operational policies. Biotechnol. Bioeng. 60
(6), 712–728.
ware (Microsoft Excel) with Eqs. (14), (22), (24) and (27). A product
Sweat, V.E., 1974. Experimental values of thermal conductivity of selected fruits and
is divided into finite number of fractions, and the drying time vegetables. J. Food Sci. 39 (6), 1080–1083.
required for each fraction is calculated. Summations of the time Toei, R., Okazaki, M., Asaeda, M., 1975. The stability of plane sublimation and a
give a drying curve when plotted as a function of the dried fraction. model of zone sublimation in freeze-drying of porous bodies. J. Chem. Eng. Jpn.
8 (4), 282–288.
The volumes of the remaining frozen fractions, V ff , are entered into Zamzow, W., Marshall, W., 1952. Freeze drying with radiant energy. Chem. Eng.
a column as shown in Fig. A (Step 1), where a total volume has Prog. 48 (1), 21–32.

You might also like