Moving Boundary Model For Simulating Moisture Movement

In Grapes
S. SIMAL, A. MULET, P.J. CATALÁ, J. CAÑELLAS, and C. ROSSELLÓ

ABSTRACT microscopic mass balance. For spheres the differential equation

A diffusional model for simulating grape dehydration with simultaneous is:

@]]rW 1 2r ]W]r #
shrinkage was developed using a finite difference method to solve dif-
ferential equations for moisture transfer within a sphere. Volume de- ]Wl 2
5 2Deff l l
(1)
crease was assumed to be equivalent to the volume of moisture removed. ]t 2
An effective diffusivity coefficient was developed from drying data at
air temperatures between 30 and 907C and constant air velocity. Effec- where Wl is the local moisture content.
tive diffusivity varied with drying air temperature following the Arrhe- This differential equation (Eq. 1) may be solved by assuming
nius equation. Experimental data at two different diffusional periods in that: (a) The initial moisture content was uniform throughout
drying were adequately represented by this moving boundary diffusivity
model. Drying kinetics at temperatures different from those used to de- the solid (Karathanos et al., 1990), (b) The solid surface and
velop the model and different grape sizes could be predicted. drying air were at equilibrium within the time frame under con-

FoodScience95 4994 Mp
sideration (Fusco et al., 1991); (c) Solid volume remained con-
Key Words: grapes, dehydration, moving boundary, moisture movement. stant (Tolaba et al., 1989). We used boundary conditions which
considered thermodynamic equilibrium and solid symmetry. The
initial condition was moisture content (Wc) of the solid at the
INTRODUCTION beginning of drying (Karathanos et al., 1990). We assumed a
QUALITY OF DEHYDRATED FRUITS can be improved by efficient diffusivity model with an important volume reduction which
drying (Fusco et al., 1991). Fruits and vegetables, highly sen- was linear with moisture loss.
sitive to microbial spoilage and deteriorative chemical reactions, The sphere was subdivided into n concentric shells of ho-
can be stabilized by dehydration. Heat and moisture transfer mogeneous material surrounding a spherical core. Each shell
within and on the surface of particles being dried control drying thickness was Dr except for the outermost shell which had a
thickness Dr/2 (Patil, 1988; Chau and Gaffney, 1990). The node

157
rate (Singh and Heldman, 1984). When external (interface to
gas) resistance to moisture transfer is negligible, drying is as- in each volume element, where the mass balance occurred, was
at the midpoint of the shell thickness. The mass balance for the

Friday Jan 26 07:54 PM

sumed to be controlled by internal diffusion in a pseudo-ho-
mogeneous fixed volume solid (Fusco et al., 1991). Internal outermost shell was made on its external surface. As drying
moisture gradient drives moisture flow and consequently dehy- proceeded the shell thickness and core size decreased due to
dration rate (Patil, 1988). Moisture transfer follows Fick’s sec- water losses, thus moving the boundary of each element. Mean-
ond law of diffusion with the moisture content gradient as the while, dry matter content remained constant during the process.
driving force. The diffusion equation was solved using Eq. (2), (3) and (4):
Nevertheless, shrinkage is common during drying and volume
changes may affect moisture removal rate. A study of the shrink- ]Wl W (r; t1Dt) 2 Wl(r,t)
age phenomena is essential to better understand drying kinetics 5 l (2)
and improve the reliability of diffusivity-based models. Re- ]t Dt
searchers have proposed that the decrease in volume during dry- ]Wl W (r1Dr,t) 2 Wl(r,t)
ing is equivalent to the volume of moisture removed (Misra and 5 l (3)
Young, 1980; Rosselló et al., 1992). ]r Dr
Numerical techniques using finite difference methods have ]2Wl W (r1Dr,t)1Wl(r2Dr,t) 22Wl(r,t)
been applied to solve such problems (Mulet et al., 1989; Chau 5 l (4)
and Gaffney, 1990; Fusco et al., 1991). Finite element methods ]r2
Dr2
have also been used (Vanegas and Marinos-Kouris, 1991; Clou- Local moisture content change was determined by mass bal-
tier et al., 1992). For spherical particles, the finite difference ance on a shell, in a Dt, time interval. Water losses (WL) during
method seemed to be suitable for modeling moisture diffusion. the time interval were calculated through a global mass balance
Our objective was to evaluate a diffusivity model with mov- with accumulation equated to the difference between input and
ing boundary conditions to simulate the drying kinetics of output (Eq. 5) around the control volume. The mass balance for
grapes. A finite difference method was used to solve differential one volume element at t 1 Dt time was obtained as a function
equations. of neighboring volume element properties at time t (Arpaci,
1966).
MATHEMATICAL MODEL
ASSUMING that internal resistance controls drying rate, mass
WL 5 2Deff Pss (NR21) ~W (R2Dr,t)2We
l

Dr 2

transfer can be calculated in terms of both Fick’s law and the 1

2 Wl(R2Dr,t)
R2Dr Dr
Dt ! (5)
Authors Simal, Cañellas, and Rosselló are with the Dept. of
Chem., Univ. of Illes Balears, Crta. Valldemossa km. 7.5, 07071 An effective diffusivity coefficient was identified from drying
Palma de Mallorca, Spain. Author Mulet is with the Food Tech- data at each drying temperature. An optimization technique,
nology Dept., Univ. Politècnica, 46071 València, Spain. Author Ca- based on the Gauss-Newton method (Kuester and Mize, 1973),
talá is with Gas and Electricity S.A., Joan Maragall, 16,07006 was used to estimate a diffusivity value which gave good agree-
Palma de Mallorca, Spain. Address inquiries to Dr. Rosselló. ment between experimental and calculated drying data. For par-

Volume 61, No. 1, 1996—JOURNAL OF FOOD SCIENCE—157

SIMULATING MOISTURE MOVEMENT IN GRAPES . . .

Fig. 1—Drier for sample dehydration. 1- Frame. 2- Fan. 3- Ane-

mometer. 4- Heating elements. 5- Pneumatic valve. 6- Tempera-
ture control measurement. 7- Sample holder. 8- Weighing
element. 9- Lifter. 10- Air compresor. C11- Computer. Fig. 2—Effect of different treatments before drying on drying ki-
netics. Air temperature 5 60&C. Air flow rate 5 11000 kg m22 h21.
Particle diameter: 1.8 5 0.1 cm. n (a) NaOH 6 g/L at 100&C for 20
ametric estimation a mixed criterion was used: the sum of sec, water at 25&C for 5 min, 40 g/L Na2S2O5 for 5 min, 1 mL/L
relative and absolute squared moisture differences in a ratio acetic acid for 5 min. ▫ (b) NaOH 6 g/L at 100&C for 20 sec, water
(1/0.2) (Richalet et al., 1978). at 25&C for 5 min, 40 g/L Na2S2O5 for 5 min, 1 g/L Chitosan in 1
A FORTRAN computer program was written to solve the set mL/L acetic acid for 5 min. l (c) 70 g/L K2CO3 and 4 mL/L olive
of equations. This allowed calculation of both the moisture dis- oil for 10 min. L (d) 40 g/L Na2S2O5 for 3 min, 25 g/L K2CO3 with
tribution inside the sphere and the average moisture content as 30 g/L etyl-oleate 4 min. ¶ (e) 6 g/L NaOH at 100&C for 20 sec. g
functions of time, and the effective diffusivity coefficient. Since (f) no treatment.
deterioration reactions are moisture dependent, moisture profiles
within the sample would be of interest. Effective diffusivity was
used to predict average and local moisture content with the sim-
ulation program.
The identified effective diffusivity values were fitted to an
exponential form (Eq. 6).

Deff 5 Doexp ~R (T1273)

2E
!
a
(6)

MATERIALS & METHODS

SEEDLESS GRAPES (Flame variety) from the island of Mallorca were used
in all experiments. Drying experiments were performed in a laboratory
hot air drier, at an average air flow rate of 11000 kg m22 h21. At this
air flow rate drying was controlled by internal resistance and rate was
not dependent on mass transfer from the solid surface to the gas phase
(Berna et al., 1991). Air humidity was 65—-70% and product was dried
on a monolayer.
The drier (Fig. 1) was equipped with 4-500W electric resistances (at Fig. 3—Influence of air drying temperature on drying kinetics.
380 V) serially connected and regulated by an automatic temperature (Treatment b). Air flow rate 5 11000 kg m22 h21. Particle diame-
controller Honeywell UDC3000 (50.17C) linked to a computer PCs ter: 1.8 5 0.1 cm.
Vectra QS/20 Hewlett-Packard. A 0.5 C.V. fan forced the air through
perpendicular to the bed. The air velocity was measured by a Data Log-
ger DTA4000 digital anemometer (accuracy 50.1 m/s) placed in the air 1.8 5 0.1 cm diam. Data were used to determine diffusivity and Arrhe-
duct. Samples were dried on a perforated plate of 13 mesh. A Mettler nius activation energy. The second series used air at 45 and 657C and
PM2000 balance linked to the computer measured weights. A pneumatic 1.8 5 0.1 cm diam grapes. The third series included diameters of 2.0
3-way valve diverted the air stream when weighing. Temperature con- 5 0.1 and 1.5 5 0.1 cm and drying air temperatures, of 50 and 707C.
trol, data acquisition and storage, as well as the general supervision of Dehydration was terminated when moisture content of ca 0.2 kg water/kg
the unit, start-up and shut down were all done by computer program. dry matter was achieved.
Drying kinetics of grapes were determined on samples (a, b, c, d, and Volume changes were calculated by immersion in distilled water of
e), dipped in different solutions before drying: dried samples with different moisture contents, and measurement of wa-
(a) NaOH 6 g/L solution at 1007C for 20 sec, distilled water at 257C ter displacement.
for 5 min, a 40 g/L Na2S2O5 solution for 5 min, and finally a 1 mL/L Ambient air was at 22 5 27C and 68 5 4% humidity. Moisture was
acetic acid solution for 5 min. determined by the AOAC method (1990).
(b) NaOH 6 g/L solution at 1007C for 20 sec, distilled water at 257C
for 5 min, a 40 g/L Na2S2O5 solution for 5 min, and finally a 1 g/L
Chitosan solution in 1 mL/L acetic acid solution for 5 min. RESULTS & DISCUSSION
(c) 70 g/L K2CO3 and 4 mL/L an olive oil emulsion for 10 min. MOISTURE CHANGES in grapes subjected to different pretreat-
(d) 40 g/L Na2S2O5 solution for 3 min followed by an inmersion in ments were compared (Fig. 2). The ordinate is the average mois-
25 g/L K2CO3 with a 30 g/L ethyl-oletate solution for 4 min.
(e) 6 g/L NaOH solution at 1007C for 20 sec. ture content (Eq. 7). Equilibrium moisture content (We) was
These samples, along with a control, were dehydrated at 607C. Product from the moisture isotherm proposed by Cañellas et al. (1993)
color was measured in a Minolta C300 colorimeter with C illuminant for grapes at the %RH of drying air.
using CIELab coordinates a* and b*. Results were reported as saturation
indexes (S.I.) 5 [(a*2 1 b*2)1/2]. After drying, the treatment which gave W2We
C5 (7)
fastest dehydration rate and highest S.I. value was selected for further Wc2We
tests.
Three additional experiments were carried out using pretreatment (b). Two different drying kinetics could be observed (Fig. 2), one
One series used air temperatures between 30 and 907C and grapes with representing pretreatments a, b and e, and the other representing

158—JOURNAL OF FOOD SCIENCE—Volume 61, No. 1, 1996

Fig. 4—Effect of air temperature on effective diffusivity coeffi-
cient for the two diffusional periods. (Treatment b). Particle di-
ameter: 1.8 5 0.1 cm.

Fig. 6—Experimental and model predicted dimensionless mois-

ture at different temperatures. (Treatment b). Particle diameter:
1.8 5 0.1 cm.

Fig. 5—Influence of drying time on average moisture content for

the two diffusional periods. (Treatment b). T 5 40&C. Particle di-
ameter: 1.8 5 0.1 cm.

c and d pretreatments. Pretreatments increased drying rates with

highest rates following the a, b and e treatments. Color values
for treatments a, b and e were 14.9 5 2.6, 14.3 5 3.6 and 9.6
5 2.4, respectively, expressed as saturation index. Although
color values for a and b were similar, treatment b was selected
for further tests due to our interest in evaluating chitosan as a
preservative.
The variation of particle volume with moisture content was
linear, described by Eq. (8).
V/Vo 5 0.2276 1 0.2151 W r250.998 (8)
Drying kinetics were studied for average moisture contents
from ca 3.4 to 0.2 kg H2O/kg dm. Induction periods or constant
drying rate period were not observed. The drying curves exhib-
ited only one diffusional period when drying air temperature Fig. 7—Experimental and computed dimensionless moisture. Ex-
was higher than 607C. Two diffusional periods were found when periments not used in parametric identification. (Treatment b).
drying air temperatures were between 30 and 607C, the first
from 3.4 to 1.0 kg H2O/kg dm, and the second from 1.0 to 0.2
kg H2O/kg dm. The activation energy for two diffusional periods was 7.9
The drying curves showed an important increase of drying kcal/mol for the first period and 8.8 kcal/mol for the second.
rates with increasing air temperature (Fig. 3). These values were similar to 10.3 kcal/mol reported by Berna
The effective diffusivity coefficients (Deff) fitted Eq. (9) for et al. (1991).
the first period and Eq. (10) for the second (Fig. 4). Experimental and average moisture content were calculated
using the proposed model and plotted vs drying time at 407C
7940.1 (Fig. 5). Two diffusional periods are shown. Experimental and
ln (Deff) 5 21.048 2 r250.988 (9)
1.987 (T1273) calculated average dimensionless moisture data at different tem-
peratures were compared (Fig. 6). The model accurately simu-
8761.9 lated the drying kinetics of grapes at different drying air
ln (Deff) 5 20.805 2 r250.996 (10)
1.987 (T1273) temperatures.

Volume 61, No. 1, 1996—JOURNAL OF FOOD SCIENCE—159

SIMULATING MOISTURE MOVEMENT IN GRAPES . . .

R gas constant (1.987) cal/mol K

Syx standard deviation (estimation) (kg water/
kg d.m.)2
Sy standard deviation (sample) (kg water/
kg d.m.)2
t time h
T drying air temperature 7C
V volume of the sphere m3
Vo initial volume of the sphere m3
W average moisture content kg water/
kg d.m.
Wc critical moisture content kg water/
kg d.m.
We equilibrium moisture content kg water/
kg d.m.
Wl local moisture content kg water/
kg d.m.
WL Water losses kg water
v percentage of variance
Fig. 8—Prediction of moisture distribution by proposed model in d local dimensionless distance
the grape sphere as related to local dimensionless distance and C average dimensionless moisture
drying time. (Treatment b.) T 5 40&C. Particle diameter: 1.8 5 0.1
cm.
REFERENCES
AOAC. 1990. Official Method of Analysis, No. 934.06. Association of Official
The percentages of explained variance (v) were calculated Analytical Chemists, Arlington, VA.
Arpaci, V.S. 1966. Conduction Heat Transfer. Adison-Wesley, Reading. MA.
using Eq. (11). Sy is the standard deviation of the experimental Berna, A., Rosselló, C., Cañellas, J., and Mulet A. 1991. Drying kinetics of
data and Syx the corresponding estimation. a Majorcan seedless grape variety. In Drying ’91, E. Mujumdar, (Ed.), p.
455-461. Elsevier, London.

@1 2 (SS )#
Cañellas, J., Roselló, C., Simal, S., Soler, A., and Mulet, A. 1993. Effects of
2 1/2 storage conditions on the quality of dehydrated grapes. J. Food Sci. 58(4):
v5 2
yx
100 (11) 805–807.
y Chau, K.V. and Gaffney, J.J. 1990. A finite-difference model for heat and
mass transfer in products with internal heat generation and transpiration.
Percentages of explained variance values showed good agree- J. Food Sci. 55(2): 484–487.
Cloutier, A., Fortin, Y., and Dhatt, G. 1992. A wood drying finite element
ment of experimental average moisture content and simulated model based on water potential concepts. Drying Technol. 10(5): 1151–
values. All v values were equal or higher than 99.3 except in 1181.
607C first period (98.2%). Fusco, A.J., Avanza, J.R., Aguerre, R.J., and Gabitto, J.F. 1991. A diffusional
model for drying with volume change. Drying Technol. 9(2): 397–417.
The proposed model was further tested by evaluating drying Karathanos, V.T., Villalobos, G., and Saravacos, G.D. 1990. Comparison of
kinetics at different temperatures and particle sizes, not used in two methods of estimation of the effective moisture diffusivity from drying
data. J. Food Sci., 55(1): 218–231.
the parameter estimation experiments (Fig. 7). The model ac- Kuester and Mize, 1973.
curately simulated drying kinetics of different sized spherical Misra, R.N., and Young, J.H. 1980. Numerical solution of simultaneous
moisture diffusion and shrinkage during soybean drying. TransASAE:
grapes. In all cases the percentages of variance explained by the 1277–1282.
model were 99%. These results point out that heat transfer at Mulet, A., Berna, A., and Rosselló, C. 1989. Drying of carrots. I. Drying
the sample surface had no significant influence on drying ki- models. Drying Technol. 7: 537–557.
Patil, N.D. 1988. Evaluation of diffusion equation for simulating moisture
netics, in this case, when modeling. movement within an individual grain kernel. Drying Technol. 6(1): 21–42.
Local moisture as a function of both drying time and dimen- Richalet, J., Rault, A., and Poulinquen, R. 1978. Identification des processus
par la methode du modele. Gordon & Breach, Paris.
sionless distance from the sphere-center to surface was plotted Rosselló, C., Cañellas, J., Simal, S., and Berna, A. 1992. Simple mathemat-
(Fig. 8). The plot shows moisture content variation in the dif- ical model to predict the drying rates of potato. J. Agric. Food Chem. 40:
2374–2378.
ferent shells from the center towards the surface. Singh, R.P., and Heldman, D.R. 1984. Introduction to Food Engineering.
Academic Press, Inc., New York.
Tolaba, M.P., Aguerre, R.J., and Suarez, C. 1989. Shape characterization for
NOMENCLATURE diffusional analysis of corn drying. Drying Technol. 7: 205–217.
Vanegas, G.K., and Marinos-Kouris, D. 1990. An analysis of mass transfer
Deff effective diffusivity m2/h in air-drying of foods. Drying Technol. 8: 323–342.
Valkó, P.; Vajda, S. 1989. Advanced Scientific Computing in BASIC with
Do Preexponential factor Arrhenius equation m2/h Applications in Chemistry, Biology and Pharmacology. Elsevier, New
Ea Activation Energy kcal/mol York.
Ms received 11/4/94; revised 6/13/95; accepted 6/26/95.
NR shell number
Pss dried matter content kg
r radius of the shell m We acknowledge the financial support of CICYT (ALI92-0422).
R radius of the sphere m

160—JOURNAL OF FOOD SCIENCE—Volume 61, No. 1, 1996