You are on page 1of 9

Journal of Food Engineering 79 (2007) 208216 www.elsevier.

com/locate/jfoodeng

Analytical solution for food-drying kinetics considering shrinkage and variable diusivity
I.I. Ruiz-Lopez *, M.A. Garca-Alvarado
Chemical and Biochemical Engineering Department, Instituto Tecnologico de Veracruz, Av. Miguel Angel de Quevedo 2779, 91897 Veracruz, Ver. Mexico, Mexico Received 10 April 2005; accepted 23 January 2006 Available online 15 March 2006

Abstract In this work, an analytical solution of a mass transfer equation for drying considering both shrinkage and variable diusivity was proposed. Experimental drying kinetics for mango slabs at dierent air temperatures (50, 60 and 70 C) and 2.5 m s1 of air velocity were obtained in order to t proposed model by means of non-linear regression with four dierent degrees of complexity. Results showed that proposed model is able to reproduce accurately experimental drying kinetics, and that shrinkage must be considered as averaged or moisture-dependent lengths for water diusion so as to obtain reliable values for eective diusivity. 2006 Elsevier Ltd. All rights reserved.
Keywords: Analytical solution; Drying; Mass transfer; Shrinkage; Variable diusivity

1. Introduction With the purpose of describe drying kinetics of foodstus, when water transport in the solid is the controlling mechanism, many authors have used diusive models (Balaban & Pigott, 1988; Banga & Singh, 1994; Bialobrzewski & Markowski, 2004; Efremov & Kudra, 2004; Hernandez, Pavon, & Garca, 2000; Lima, Queiroz, & Nebra, 2002; Maroulis, Kiranoudis, & Marinos-Kouris, 1995; McCarthy, Perez, & Ozilgen, 1991; Mulet, 1994; Ruiz-Lopez, Cordova, Rodrguez-Jimenes, & Garca-Alva, rado, 2004; Rosello Canellas, Simal, & Berna, 1992; Sapru & Labuza, 1996; Wang & Brennan, 1995). Non-steady mass transfer equation for moisture diusion within a homogeneous and isotropic material is expressed as (Ruiz-Lopez et al., 2004)

oqX r Deff rqX 1 ot Deff rqX i k C1 q1 Y i Y at airproduct interface Deff rqX 0 Y i f X i at product center 2 3 4

In addition to mass transfer, simultaneous heat transfer occurs during drying of foodstus, which may be expressed as non-steady heat conduction within the product as (Ruiz-Lopez et al., 2004): oqCpT r k eff rT ot k eff rT i DH W Deff rqX i h1 T i T 1 at airproduct interface k eff rT 0 at product center 5

6 7

Corresponding author. Tel.: +52 229 934 5701. E-mail address: miguelg@itver.edu.mx (I.I. Ruiz-Lopez).

Eqs. (1)(7) with variable properties do not have an analytical solution; so they must be solved by means of complicated numerical schemes (Balaban & Pigott, 1988; Banga &

0260-8774/$ - see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.jfoodeng.2006.01.051

I.I. Ruiz-Lopez, M.A. Garca-Alvarado / Journal of Food Engineering 79 (2007) 208216

209

Notation a specic contact area (cm1) A particle area (cm2) aW food water activity (dimensionless) Bim mass transfer Biot number C1 C5 empirical parameters for water activity Cp specic heat (dry basis) (J g dry matter1 K1) D1 D2 empirical parameters for diusion equation De eective water diusivity in food (cm2 s1) E activation energy (J mol1) h1 external heat transfer coecient (J cm2 s1 K1) kC, kC1 internal and external mass transfer coecients, respectively (cm s1) ke eective thermal conductivity (J cm1 s1 K1) Keq average moisture partition constant between the air and the product (dimensionless) LD path length for diusion (cm) P, pW total pressure and partial pressure of water vapor, respectively (Pa) R ideal gases constant (8.314 J mol1 K1) t drying time (s) T ; T i ; T ; T 1 food temperature: punctual at time t, at foodair interface, averaged over diusion path length, and air temperature, respectively (K) V particle volume (cm3) X ; X e ; X i ; X food moisture content: punctual at time t, at equilibrium, at foodair interface, and averaged over diusion path length, respectively (g water g dry matter1) Y, Yi z air moisture content: bulk air and foodair interface, respectively (g water g dry air1) coordinate parallel to water diusion ux

Greeks a, b empirical parameters for dimensionless diusion equation d eective lm thickness for mass transfer (cm) DHW water sorption heat (J g water1) DLDf fraction of the initial diusion path length at the end of the drying period e shape factor for eective lm thickness (dimensionless) c shape factor for diusion equation (dimensionless) j, g empirical parameters for dimensionless shrinkage equation q food density (dry basis) (g dry matter cm3) q1 air density (dry basis) (g dry air cm3) W, Wi dimensionless moisture content: averaged over diusion path length; at foodair interface, respectively f shape factor for specic area (dimensionless) Subscripts 0 at the beginning of the drying period j implies a spatial node in numerical solution

Singh, 1994; Lima et al., 2002; Maroulis et al., 1995; McCarthy et al., 1991; Mulet, 1994; Ruiz-Lopez et al., 2004; Sapru & Labuza, 1996; Wang & Brennan, 1995). System (1)(4) may have an analytical solution if it is assumed that there is no shrinkage, water diusivity into the product is constant, and process is diusion controlled (i.e. Xi = Xe at t = 0). These analytical solutions for rectangular, cylindrical, and spherical geometries are, respectively (Crank, 1975): W X Xe X0 Xe 1 8 X

2n 1 p2 Deff t 2 exp 2 p n1 2n 1 4L2 D ! 1 2 X 1 X Xe k Deff t 4 exp n 2 W 2 X0 Xe LD n1 kn ! 1 X Xe 6 X 1 n2 p2 Deff t W 2 exp X 0 X e p n1 n2 L2 D 1

"

# 8 9 10

In Eq. (9), kn is the nth positive root of J0(kn) = 0 where J0(x) is the Bessel function of the rst kind of order zero.

Nonetheless, shrinkage is not negligible during drying. Most foods shrink during drying, some of them shrinking more than 50% of their original dimensions, depending upon the degree of drying (Balaban & Pigott, 1988). Because of this, most of the authors have considered product shrinkage in their models using linear functions of moisture content (Balaban & Pigott, 1988; Banga & Singh, 1994; Hernandez et al., 2000; Lewicki, Witrowa-Rajchert, & Nowak, 1998; Lima et al., 2002; Mulet, 1994; Ruiz-Lopez et al., 2004; Wang & Brennan, 1995) or some other functional relations (Bialobrzewski & Markowski, 2004). It have been recognized that the eect of shrinkage cannot be ignored in establishing reliable values of De (Bialobrzewski & Markowski, 2004; Hernandez et al., 2000; Mulet, 1994). But even if shrinkage is neglected, simplied diusional models still describe satisfactorily the experimental data, although eective diusivity is then reduced to an empirical parameter that ts experimental data (Hernandez et al., 2000; Mulet, 1994; Ruiz-Lopez et al., 2004). Eective diusivity may be estimated directly from Eqs. (1)(7) (Bialobrzewski & Markowski, 2004; Lima et al., 2002; Maroulis et al., 1995; Mulet, 1994), but the numerical eort may not compensate the advantages of

210

I.I. Ruiz-Lopez, M.A. Garca-Alvarado / Journal of Food Engineering 79 (2007) 208216

simplied models for most of the common applications (Ruiz-Lopez et al., 2004). Therefore, simplied models still remains popular in obtaining values for De (Hernandez et al., 2000; Lewicki et al., 1998; Mulet, 1994; Rosello et al., 1992; Ruiz-Lopez et al., 2004; Sapru & Labuza, 1996; Zogzas & Maroulis, 1996). Food shrinkage produces a variation in the distance required for the movement of water molecules, therefore making that eective diusivity be overestimated when obtained from analytical solutions (8)(10). For that reason, it is desirable to obtain a simple model taking into account dimensional changes. With the aim of incorporate dimensional changes of product in an analytical solution of the simplied boundary problem dened by Eqs. (1)(4), Hernandez et al. (2000) showed that shrinkage may be considered in Eq. (8) if: Deff t=LD > 1 11

where a = A/V. Eq. (12) may be integrated under the following assumptions: 1. Drying is diusion-controlled, i.e. Xi = Xe and mass transfer into the drying media has no eect. This assumption is valid when (q1/q)KeqBim ) 1 (Cordova-Quiroz, Ruiz-Cabrera, & Garca-Alvarado, 1996), where Bim is the mass Biot number dened as Bim k C1 Deff =LD 13

and Keq represents an average moisture partition constant between the air and product: R Xe K eq
X0

Y i =X i dX i X0 Xe

14

This only is valid at the end of the drying, when practically shrinkage had concluded. By the other hand, analytical solutions (8)(10) produce an ideal linear behavior for moisture when it is represented on a semi log plot (at W < 0.7). However, deviations of predicted values with respect to experimental data can be observed for foodstus, mainly at lower moisture levels (W < 0.1) when practically shrinkage had nished (Hernandez et al., 2000; Mulet, 1994; Ruiz-Lopez et al., 2004; Zogzas & Maroulis, 1996). Thus, deviations at lower moisture levels cannot be explained by shrinkage. Mulet (1994), Maroulis et al. (1995), and Hernandez et al. (2000) reported that diusivity is a moisture function at W < 0.1, but it may be moistureindependent at higher moisture levels. In order to include moisture dependence of diusivity, many authors have used exponential functions of moisture content when modeling drying kinetics (Banga & Singh, 1994; Bialobrzewski & Markowski, 2004; Hernandez et al., 2000; Lewicki et al., pez et al. 1998; Maroulis et al., 1995; Mulet, 1994). Ruiz-Lo (2004) showed that, even when diusivity and diusion path length change along the drying time, analytical solutions with averaged diusivity and characteristic length for path diusion reproduce moisture evolution acceptably. The objective of this work was to obtain and validate an analytical solution from a mathematical model developed from macroscopic mass transfer with simultaneous shrinkage and variable diusivity. 2. Model development During drying of a single particle or during xed-bed drying, the average moisture evolution may be described as (Herman-Lara, Salgado-Cervantes, & Garca-Alvarado, 2005) dX k C aX X i dt 12

2. Food temperature is constant. This assumption is not generally true during food drying. However, it has been extensively used in order to obtain an analytical solution of mass transfer equation for diusive drying of food stus (Hernandez et al., 2000; Lewicki et al., 1998; Mulet, 1994; Rosello et al., 1992; Ruiz-Lopez et al., 2004; Sapru & Labuza, 1996; Zogzas & Maroulis, 1996). Therefore, is more rigorous the assumption that food temperature is equal to its average value during process. 3. Volumetric concentration of dry product is constant (average value) with respect to moisture content. 4. Shrinkage is present and occurs primarily in the minor dimension of the product, but it keeps its geometry (there is no deformation). For handiness, Eq. (12) can be expressed in terms of the dimensionless moisture content dened in Eq. (8) as dW k C aW Wi dt 15

Since it was assumed a diusion-controlled process, Xi = Xe and Wi = 0. Therefore, Eq. (15) simplies to: dW k C aW dt 16

If the eective coecient for mass transfer and the specic surface are moisture functions, then we can express a general solution of Eq. (16) given by Z
W

W0

1 dW k C aW

Z
t0

dt

17

Mass transfer coecient can be calculated by considering water diusion inside a product of some particular geometry, and it is generally accepted as having the following form (Viollaz, Suarez, & Alzamora, 1980): k C Deff =d 18

I.I. Ruiz-Lopez, M.A. Garca-Alvarado / Journal of Food Engineering 79 (2007) 208216

211

In Eq. (18), the problem is to know the eective lm thickness (d) where the moisture change takes place. From mass transfer equation solution under unsteady state (Crank, 1975), the resulting relation for the eective lm thickness is d eLD 19

where e = 4/p2 for an innite slab, e 2=k2 for an innite 1 cylinder (k1 = 2.4048 is the rst positive root of the Bessel function of the rst kind of order zero), and e = 3/p2 for a sphere. Accordingly, eective coecient for mass transfer is k C Deff =eLD Specic area (a) has the following form: a f=LD 21 20

Eqs. (22), (24), and (26) may be combined to obtain: Z Z W j expbW 2jg W dW expbW dW W a W0 a W0 Z g2 W expbWW dW a W0 Z f t dt 27 e t0
2

Then, analytical solution for the dimensionless moisture of the food when t0 = 0 and W0 = 1 is given by " # 1 X bn j2 2jg n expbW expb lnW W 1 ab a n n! n1 ! g2 expbW expb expb W expbW b b ab f t e 28

where f = 1 for an innite slab, f = 2 for an innite cylinder, and f = 3 for a sphere. Eqs. (17), (20) and (21) may be combined to yield e f Z
W W0

L2 D dW Deff W

Z
t0

dt

22

Water diusivity in Eq. (22) may be assumed to be modeled by an exponential moisture function (Bialobrzewski & Markowski, 2004; Mulet, 1994; Ruiz-Lopez et al., 2004), with an Arrhenius type temperature inuence (Balaban & Pigott, 1988; Banga & Singh, 1994; Bialobrzewski & Mar kowski, 2004; Hernandez et al., 2000; Lewicki et al., 1998; Lima et al., 2002; Wang & Brennan, 1995): Deff D1 expD2 X expE=RT 23

Eq. (28) is an algebraic expression that represents food drying kinetics with both shrinkage and variable diusivity when the process is diusion-controlled. This equation is implicit in W, and therefore requires some numerical algorithm to solve it. Yet, evaluation of an algebraic equation is mathematically simpler than numerical solution of a partial dierential equation. If there is not moisture eect over diffusion, i.e. D2 = 0, then a = D1exp(E/RT) and b = 0. Analytical solution for this particular case is j2 2jg g2 f W 1 W2 1 t ln W a e a 2a 29

Series solutions (8)(10) converge at the rst term when W < 0.7. Under such circumstances, Eqs. (28) and (29) are mathematically equivalent to these assuming that there is not shrinkage (i.e. DLDf = 1). 3. Methodology 3.1. Experimental Proposed model was tted to experimental drying kinetics of mango slices. Mangoes were purchased from the local market. They were peeled and cut into 1 3 5 cm3 slices. For following the moisture evolution the samples were isolated at the lateral faces and placed in a plant-pilot cabinet dryer in a manner that air owed only at top and bottom faces (LD0 = 0.5 cm) in the way previously described by Hernandez et al. (2000). Drying was carried out at air dierent temperatures (50, 60 and 70 C) and with an air velocity around 2.5 m s1. Air velocity was measured with an anemometer at the same site of the sample. In order to follow moisture evolution, weight losses of the sample were measured with a balance every 15 min during the rst 2 h, every 30 min the next hour, and every 60 min the rest of the experiment (about 8 h). Initial moisture content was determined by the vacuum oven method.

Eq. (23) expressed in terms of the dimensionless moisture is, Deff a expbW 24

where a D1 expD2 X e expE=RT , and b = D2(X0 Xe). If product shrinks without deformation, then the diusion path length (LD) must change along the drying time. In this work, we used a linear dependence of characteristic length for diusion with moisture concentration (Hernandez et al., 2000): LD =LD0 DLDf 1 DLDf X =X 0 25

Rewriting Eq. (25) in terms of the dimensionless moisture yields: LD j gW 26

where j = DLDfLD0 + (1 DLDf)LD0Xe/X0 and g = (1 DLDf)(X0 Xe)LD0/X0.

212

I.I. Ruiz-Lopez, M.A. Garca-Alvarado / Journal of Food Engineering 79 (2007) 208216

Initial and nal slab dimensions were measured with a vernier. DLDf was estimated by tting initial and nal slabs thickness to Eq. (25). Additionally, a dierent set of samples was prepared with a K-type thermocouple in the center with the purpose of follow temperature evolution during drying. Temperature measurements of the product were taken every 5 min during the rst 2 h, and every 15 min the rest of the experiment (about 8 h). Wet and dry bulb temperatures of the room air were measured with a glass thermometer in order to estimate the bulk air moisture content (Y). All experiments were carried out by duplicate. 3.2. Model tting Experimental data were used to estimate the tree parameters D1, D2, and E by means of non-linear regression over Eq. (28) with moisture content as dependent variable, and time and temperature as independent variables. These parameters were obtained with shrinkage, without shrinkage, and using an averaged value for diusion path length. Additionally, D1 and E were obtained assuming no moisture eect over diusion and without shrinkage (D2 = 0, DLDf = 1) by tting Eq. (29). Model assumptions are presented coded in Table 1. Following data tting, 95% non-simultaneous condence intervals for the nonlinear least squares parameter estimates were calculated. Nonlinear regression and condence intervals were obtained using the Matlab Statistics Toolbox 5.0 (MathWorks Inc.). Temperature values used with Eq. (23) were the averaged ones obtained from the temperature evolution data of the product, and were calculated using Eq. (30). Z t T T dt=t 30
0

Yi

aW pW =P 18 1 aW pW =P 29   2 aW 1 exp C 1 T C2 X C3 C4 T C5 T i

31 32

Equilibrium moisture content (Xe) is dened as the moisture value that satises Eqs. (31) and (32) with the bulk air moisture (Y). Eq. (32) is known as Henderson modied equation (Cordova-Quiroz et al., 1996). Deviations between predictions and experimental data of the analytic solution were compared with those produced using a detailed model (model E). This model was solved using the method of lines with a longitudinal scheme and central nite dierences for space derivatives. This method generates an ordinary dierential equation (ODE) system with t as independent variable. The scheme obtained was ! ! # !" dX j 1 Deff DX DX mj1=2 mj1=2 Dzj1=2 m j Dz j Dz j1 dt ! ! Deff DX j DX j1 c 33 z j Dzj Dzj1 " ! ! # dT j 1 DT DT k eff j1=2 k eff j1=2 qCpj Dzj1=2 Dz j Dz j1 dt ! ! ! k eff DT j DT j1 #T dX j c 34 qCp j dt qCpz j Dzj Dzj1 where m = De[q + [dq/dX]X] and # = q[dCp/dX] + Cp[dq/ dX], for 1 6 j 6 n. Dyj is the rst forward dierence operator over yj and means subtraction between discrete values for variable or property y evaluated in j + 1 and j nodes. j + 1/2 means the arithmetic average of variable or property evaluated in j and j + 1 nodes, j 1/2 means the arithmetic average of variable or property evaluated in j and j 1 nodes. In Eqs. (33) and (34), c = 0 for an innite slab, c = 1 for an innite cylinder, and c = 2 for a sphere, and z denotes coordinate (axial or radial) parallel to water diusion ux. With the intention of evaluate the points outside of the boundary (variable or property evaluated in j + 1/2 node where j = n), Eqs. (2) and (6) were solved for Xj+1 and Tj+1 in the central nite dierences scheme giving ! 2Dz X j1 X j1 k C1 q1 Y i Y 35 m j ! !! 2Dz DX j DX j1 T j1 T j1 hT j T 1 DH W mj k eff j 2Dzj 36

3.3. Model solving Proposed model is implicit in W, and thats why it must be solved by means of a numerical method. In this work, bisection algorithm was used using 12 terms of the series solution. In order to estimate the equilibrium moisture content (Xe), water vapor and air were assumed to be ideal gases, therefore, function expressed in Eq. (4) is equal to

Table 1 Codes for model assumptions Assumptions De and LD as moisture functions, Eq. (27) De as a moisture function and LD as a constant initial value, Eq. (27) De as a moisture function and LD as a constant averaged value, Eq. (27) De as a moisture independent function and LD as a constant averaged value, Eq. (28) Numerical solution of Eqs. (1)(7) with De and LD as moisture functions Numerical solution of Eqs. (1)(4) with De and LD as moisture functions Model A B C D E F

I.I. Ruiz-Lopez, M.A. Garca-Alvarado / Journal of Food Engineering 79 (2007) 208216 Table 2 Parameters, thermophysical properties, and detailed model assumptions used in model solutions Sorption isotherms: Hernandez et al. (2000) for mangoes C1 = exp(0.914), C2 = 0, C3 = 0.5639, C4 = 0, C5 = 0 Initial and equilibrium values for moisture content (g water g dry matter1) X0 = 3.675, Xe = 9.73 103 at 50 C, Xe = 2.71 103 at 60 C Xe = 4.92 104 at 70 C Initial and averaged values for temperature (C) T 0 28; T 41:7 at T1 = 50, T 47:2 at T1 = 60, T 55:8 at T1 = 70 Fraction of the initial diusion path length at the end of the drying period DLDf = 0.3 Averaged diusion path length (cm) LD = 0.325 Dry solids density (g dry matter cm3): experimentally determined for mangoes q = 350 30X Specic heat (dry basis) (J g dry matter1 K1) Cp = CpD + CpWX CpW = 4.185, Geankoplis (1978) CpD = 1.657, Mujumdar (1993) Thermal conductivity (J cm1 s1 K1): Singh and Heldman (1993) for foods 3 k eff 1:418 103 4:9310 X 1X Sorption heat (J g1): pure water (Perry & Chilton, 1973) 0:31993:276104 T i 6:1596107 T 2 Ti i DH W 2889:4 1 647:13 Model E: Eqs. (1)(7), all properties as X or T functions, LD as moisture function, De is calculated with the parameter set obtained for model A Model F: Eqs. (1)(4), Xi = Xe and T T for t P 0, volumetric concentration of dry matter is constant, LD as moisture function, De is calculated with the parameter set obtained for model A

213

Average dimensionless moisture content was calculated as W Z


0 LD

R LD ! X dz=z X e X Xe dz=z 0 X0 Xe X0 Xe 37

X Xe X0 Xe

Simulations for model E were performed using parameters found for model A. External heat and mass transfer coecients were calculated with the equations reported by Perry and Chilton (1973) for the case of forced convection around a submerged at plate parallel to ow. Additionally, detailed model was solved considering the same assumptions as proposed model, that is, using a constant average food temperature (dTj/dt = 0 for 1 6 j 6 n) and interfacial moisture content instantaneously equilibrated with air (Xj+1 = Xe in j = n) (model F). All parameters used in model tting and thermophysical properties jointly with the codications for the dierent detailed model assumptions are summarized in Table 2. 4. Results and discussion Parameters obtain by non-linear regression (D1, D2, and E) jointly with the 95% condence intervals (CI) for parameters and the root mean square errors (RMSE) between experiments and predictions are listed in Table 3. Averaged temperatures used in Eq. (23) were calculated with Eq. (30) from the experimental temperature evolutions plotted in Fig. 1. It can be seen that model A (which takes into account the shrinkage and moisture eect over diusivity) exhibited the better t (lower RMSE), yet models B and C represent good t alternatives. However, observed dierences in estimated parameters have important implications over the estimation of water mobility in food medium. The De values obtained with Eq. (23) and the parameters listed in Table 3 are plotted in the Fig. 2 (at 50 C). The greater values for De were obtained for model B. This fact is predictable since the diusion path length is kept constant at its initial value, and because the product shrinks, a greater value in De is required in order to predict the same drying rate. On the

In the same way, to evaluate the variable or property in j 1/2 node where j = 1, Xj1 = Xj+1 and Tj1 = Tj+1 were stated. The resulting set of ODEs was integrated forward in t using the Euler method.

Table 3 Parameters obtained by non-linear regression of proposed modela Model A 95% B 95% C 95% D 95% 0 Db 95%
a b

D1 102 (cm2 s1) 2.6709 0.8071/4.5348 51.4573 13.2447/89.6699 21.7407 5.5945/37.8869 2.2411 0.3905/4.8727 4.2604 104 4.0867 104/4.4341 104

D2 (g dry matter g water1) 0.8520 0.8250/0.8790 0.2137 0.1872/0.2401 0.2137 0.1872/0.2401

E (J mol1) 27680.5 25750.5/29610.5 29963.9 27918.3/32009.4 29963.9 27918.3/32009.4 22951.2 19792.5/26110.0

RMSE 0.0235 0.0242 0.0242 0.0512 0.0967

CI CI CI CI CI

Condence intervals for parameters in non-linear regression are non-simultaneous. Fitted after the temperature dependence over diusivity had been eliminated.

214

I.I. Ruiz-Lopez, M.A. Garca-Alvarado / Journal of Food Engineering 79 (2007) 208216

Simulation results for all models and experimental data are presented in Figs. 35. It can be observed that model D shows the typical linear behavior of Eqs. (5)(7). This model is not as good as models A, B, or C to reproduce the experimental behavior, thus exhibiting the greatest deviations from experimental data at all moisture contents. These results demonstrate that assumptions made in model D are inadequate for a trustworthy estimation of De when experimental data separate from ideal linear behavior and

Fig. 1. Experimental temperature evolution of mango slices.

Fig. 3. Drying kinetics of mango slices at 50 C.

Fig. 2. Eective water diusivity predictions in mango (at 50 C).

other hand, the De values obtained for models C and D represent approximately an average of those ones obtained for model A. Therefore, it is fundamental to allow for shrinkage (either with an averaged LD or with a variable LD) in order to obtain a representative water diusivity into the product. If water diusion is estimated using only the initial product thickness, then the obtained value is overestimated. It is important to note that signicance lost of parameter D1 for model D was due to the covariance between the independent variables time and temperature, which is reected as the failure of the non-linear regression procedure to estimate all model parameters with statistical signicance. This fact was demonstrated by tting experimental data to model D 0 , obtained after eliminating temperature eect over diusivity (parameter E) in model D. The new value for D1 jointly with its 95% condence intervals and the RMSE between experiments and predictions are also listed in Table 3. As it can be seen, this time D1 was signicant. Experimental data for temperature evolution are presented in Fig. 1. It can be seen that after a continuous increase of temperature during the rst 2 h of the experiment, the product temperature reaches a nearly stationary period in which it can be considered constant. During this period, product temperature increased very slowly until reach the drying air temperature, so it can be assumed that all supplied energy is expended in water evaporation.

Fig. 4. Drying kinetics of mango slices at 60 C.

Fig. 5. Drying kinetics of mango slices at 70 C.

I.I. Ruiz-Lopez, M.A. Garca-Alvarado / Journal of Food Engineering 79 (2007) 208216

215

agree with previous works in which is pointed out that at low moisture contents, any model, with or without shrinkage, that assumes a moisture-independent diusivity will be insucient to make a mathematical description of drying curves (Bialobrzewski & Markowski, 2004; Hernandez et al., 2000; Mulet, 1994). Nevertheless, if experimental data do not deviate too much from linear behavior, then assumptions made in model D are still adequate in estimating De. Models A, B, and C reproduce accurately the experimental shape at all moisture contents, tracking deviations from linear behavior in a precise way, but as we seen before, model B overestimates diusion. Thus, assumptions made with models A or C represent the best alternatives in estimating dependable values of De. Concluding, if moisture dependence is going to be considered in estimating De, then shrinkage must be considered as moisture-dependent or averaged values for LD. Figs. 35 shows simulations results for detailed model E [Eqs. (1)(7)]. Some deviations between predicted values and experimental data are observed, but these deviations take place at the lower moisture levels. Nevertheless, it must be remembered that experimental data were not tted to this model, but model E is being used for prediction. Furthermore, model E uses a set of thermophysical properties not obtained from experimental data tting, but reported by other authors. Hence, these results exemplify that proposed model is able to generate reliable values of De for prediction purposes. Simulations results for model F [Eqs. (1)(4)] are shown in Figs. 35. This time, even though model F uses the same assumptions as proposed model, great deviations between experimental data and predicted values are observed at all moisture contents. Model F is a microscopic model, therefore during its numerical solution a diffusivity value is calculated at each node. As this model assumes that interface reaches instantaneously the equilibrium moisture content and there is not a preheating period, drying occurs with the higher moisture gradient, thus making moisture content diminished rapidly in the nodes located near surface. As a consequence, calculated diusivities at these nodes have a value too small, which make external nodes behavior like a barrier, avoiding water migration from interior nodes. Yet, as seen with model E, if temperature evolution is considered, then a microscopic model using a tted expression for diusivity obtained from proposed model gives a good approximation to experimental data. Although some dierences are observed between predictions produced by proposed model (a macroscopic model) and detailed model (a microscopic model), it must be remembered that properties estimation from detailed models uses experimental data that are in fact averaged values for moisture or temperature and that process design uses preferentially macroscopic models. Consequently, as diusivity may be overestimated from detailed models if temperature evolution is not considered, a macroscopic model will be more adequate in estimating De.

Fig. 6. Eect of shrinkage over drying kinetic predictions produced by proposed model and detailed model for an innite cylinder geometry.

Fig. 7. Eect of diusivity over drying kinetic predictions produced by proposed model and detailed model for a sphere geometry.

Fig. 6 shows simulated drying kinetics at 60 C for model A and model E when product has a cylindrical shape. This predictions were calculated for extreme values of DLDf, that is, when product does not shrink (DLDf = 1) and for the ctitious case when product disappear (DLDf = 0). These simulations were intended only for testing both models under extreme circumstances. It can be seen that both models produce very similar responses along all the moisture range. Simulations results for models A and E using conditions for product being dried at 60 C when it has a spherical shape are presented in Fig. 7. These predictions were calculated varying the moisture dependence on diusivity, that is, reducing and increasing the value for D2. Again, it can be observed that both models produce similar drying kinetics. All these results conrm the fact that proposed model represents a straightforward alternative to detailed models to estimate De in an easy and consistent way. 5. Conclusions Proposed model provides a simple mathematical description for food drying kinetics considering both shrinkage and a moisture dependent diusivity and solu-

216

I.I. Ruiz-Lopez, M.A. Garca-Alvarado / Journal of Food Engineering 79 (2007) 208216 Lewicki, P. P., Witrowa-Rajchert, D., & Nowak, D. (1998). Eect of drying mode on drying kinetics of onion. Drying Technology, 16(1&2), 5981. Lima, A. G. B., Queiroz, M. R., & Nebra, S. A. (2002). Simultaneous moisture transport and shrinkage during drying of solid with ellipsoidal conguration. Chemical Engineering Journal, 86, 8593. Maroulis, Z. B., Kiranoudis, C. T., & Marinos-Kouris, D. (1995). Heat and mass transfer modeling in air drying of foods. Journal of Food Engineering, 26, 113130. McCarthy, M. J., Perez, E., & Ozilgen, M. (1991). Model for transient moisture proles of a drying apple slab using the data obtained with magnetic resonance imaging. Biotechnology Progress, 7, 540543. Mujumdar, A. S. (1993). Handbook of industrial drying. New York: Dekker. Mulet, A. (1994). Drying modelling and water diusivity in carrots and potatoes. Journal of Food Engineering, 22, 329348. Perry, R. H., & Chilton, C. H. (1973). Chemical engineering handbook. Kogakusha: McGraw-Hill. Rosello, C., Canellas, J., Simal, S., & Berna, A. (1992). Simple mathematical model to predict the drying rates of potatoes. Journal of Agriculture and Food Chemistry, 40, 23742378. Ruiz-Lopez, I. I., Cordova, A. V., Rodrguez-Jimenes, G. C., & GarcaAlvarado, M. A. (2004). Moisture and temperature evolution during food drying: eect of variable properties. Journal of Food Engineering, 63, 117124. Sapru, V., & Labuza, T. (1996). Moisture transfer simulation in packaged cerealfruit systems. Journal of Food Engineering, 27, 4561. Singh, R. P., & Heldman, D. R. (1993). Introduction to food engineering (2nd ed.). San Diego, CA: Academic Press. Viollaz, P. E., Suarez, C., & Alzamora, S. (1980). Temperature prediction in air drying of food materials: a simple model. Journal of Food Technology, 15, 361367. Wang, N., & Brennan, J. G. (1995). A mathematical model of simultaneous heat and moisture transfer during drying of potato. Journal of Food Engineering, 24, 4760. Zogzas, N. P., & Maroulis, Z. B. (1996). Eective moisture diusivity estimation from drying data. A comparison between various methods of analysis. Drying Technology, 14(7&8), 15431573.

tion is available for various geometries. Furthermore, this model represents adequately the experimental behavior at all moisture contents and represents a simple way to evaluate a moisture dependent function for eective diusivity in a reliable way. References
Balaban, M., & Pigott, G. M. (1988). Mathematical model of simultaneous heat and mass transfer in food with dimensional changes and variable transport parameters. Journal of Food Science, 53(3), 935939. Banga, J. R., & Singh, R. P. (1994). Optimization of air drying of foods. Journal of Food Engineering, 23, 189211. Bialobrzewski, I., & Markowski, M. (2004). Mass transfer in the celery slice: eects of temperature, moisture content, and density on water diusivity. Drying Technology, 22(17), 17771789. Cordova-Quiroz, V. A., Ruiz-Cabrera, M. A., & Garca-Alvarado, M. A. (1996). Analytical solution of mass transfer equation with interfacial resistance in food drying. Drying Technology, 14(7&8), 18151826. Crank, J. (1975). The mathematics of diusion (2nd ed.). Oxford: Oxford University Press. Efremov, G., & Kudra, T. (2004). Calculation of the eective diusion coecients by applying a quasi-stationary equation for drying kinetics. Drying Technology, 22(10), 22732279. Geankoplis, C. J. (1978). Transport processes and unit operations. Newton, MA: Allyn and Bacon. Herman-Lara, E., Salgado-Cervantes, M. A., & Garca-Alvarado, M. A. (2005). Mathematical simulation of convection food batch drying with assumptions of plug ow and complete mixing of air. Journal of Food Engineering, 68, 321327. Hernandez, J. A., Pavon, G., & Garca, M. A. (2000). Analytical solution of mass transfer equation considering shrinkage for modeling fooddrying kinetics. Journal of Food Engineering, 45, 110.

You might also like