13 Osmotic Dehydration An Approach to the Modeling of Solid Food-Liquid Operations

Pedro Fito and Amparo Chiralt

Introduction Solid food-liquid systems are common in the field of food technology (Table 13-1) which usually bases related heat and mass transfer operation modeling on the consideration that a food solid is a continuous phase. Fickian equations have been the typical way to evaluate the kinetics of mass transfer phenomena, without taking into account the role of a food's structure on the mechanism taking place inside the solid. Despite the relatively large number of published works about the fundamentals of osmotic phenomena related to the transport mechanisms involved and their effects on food structure, an effort to systematize the mathematical treatment of this unit operation seems to be necessary to make such data more usefulespecially when temperature or

Table 13-1. Some Solid-Liquid Systems in Food Engineering Osmotic dehydration Vacuum impregnation Solid extraction Food hydration Cooking (conventional, under pressure, "sous vide") Canning Pickles Salting (cheese, meat, fish) Treatments of liquid food with solid particles (heating, electric pulses, high pressures, microwaves)

pressure changes are relevant. Because of the complexity of microscopic alterations at the cellular level (Le Maguer and Yao, 1995), a semiempirical boarding of osmotic dehydration is more frequently carried out for the kinetics model, but this approach has meant the use of kinetic constants containing complex information about the phenomena determining the mass transfer rate. The aim of this chapter is to establish an approach to the design and evaluation of an osmotic dehydration operation model that allows prediction of changes in the total food mass and composition of the food-liquid fraction over time. This will enable the forecasting of food-water activity values and therefore product stability. The total weight loss of the product is basic for any economic evaluation of the industrial operation. The proposed model takes into account: 1. Thermodynamic aspects in terms of the chemical potential of components in the present phases at equilibrium. 2. Kinetic analysis of mass transfer for each component, taking into account pseudo-Fickian (osmotic and diffusion), vaporizationcondensation, hydrodynamic mechanisms (HDM), as well as the role of structure in their coupling. 3. Prediction of the liquid retention capacity of the solid matrix.

Development of the Mathematical Model Applied to Osmotic Dehydration Operation Scheme and System Definition Figure 13-1 shows the scheme for an osmotic dehydration operation (variables used in the model are defined in the notation), and Figure 13-2 illustrates the main changes produced in a structured food when a hydration or osmotic dehydration operation occurs [the solid food matrix (FM) sample is represented as a solid container full of a food-liquid phase (FLP)]. As osmotic dehydration progresses, some structure changes occur that involve sample shrinkage, which leads to the diminishing of its capacity to hold liquid. The composition values calculated for apple dehydration reveal that from a thermodynamic point of view the solid food could be defined as a multiphase and multicomponent system (because any cell represents a phase itself); but a food engineering approach assumes the food is composed of water, solutes, and inert. The latter would represent the macromolecular components that constitute the solid matrix of the food. Mean concentrations of each component in a food system are represented as x (mass fraction). The solid food could also be considered as a two-phase system in which the solid stage would be the food solid matrix (SM) with some adsorbed water, and the food-liquid phase (FLP) a

RAW PRODUCTS Initial values:

Osmotic Dehydration Operation Operation Variables: P(mbar), T(0C), f(s), recirculation ratio, agitation Osmotic solution: y w ^X0

FINAL PRODUCT Values at time t:

Figure 13-1. Scheme for an osmotic dehydration operation.

solution that includes the major part of water and solutes present in the food. Concentrations of each component in this phase [assumed to be composed of water (W) and total solutes (ST)] are represented by z (mass fraction), and compositions in the external osmotic solution by v (mass fraction). Relationships between the different variables are shown in Equations 1-4, where it is clear that the FLP composition (z values) may be easily calculated from the overall compositions (x values). Nevertheless, to calculate x values from the FLP composition does require that one know the value of the FLP/SM rate.

(D

COMPONENTS Solid Matrix Solutes Water COMPONENTS

Osmosed Material

Compositions

Hydrated Material

FOOD ENGINEERING APPROACH

System Components
Solid Matrix I = 4%

TREATMENT EFFECTS ONTHESTRUCTURE Shrinking Pore reduction Less FLP/SM rate Loss of turgor press Cells collapse DEHYDRATION Water W = 84% PHASES EXT. LIQUID PHASE INT. PHASES Solid Matrix SM= Food Liquid Phase FLP= Compositions Osmosed material Raw material

Solutes ST = 12%

System Phases
Solid Matrix SM = 4%

Water W= Solutes ST= Solid Matrix I= Food Liquid Phase FLP = 96%

Figure 13-2. The solid food-liquid system.

(2) (3) (4) Mass Balances The signs convention was set considering the mass fluxes into the product as positive; Equations 5-10 define the relationships between the following variables: (5) (6) (7) (8) (9) (10) Equation 10 may be used to verify the agreement among the different experimental determinations (M and x values). Plotting the experimental values of (H) requires that they lie on the diagonal, because scattering in the distribution of experimental points around the diagonal denotes the occurrence of experimental errors. Equilibrium Relationships According to the thermodynamic description of the system the equilibrium between the food and its liquid surroundings will be reached when the chemical

potential of any component would be the same in all phases present in the system. Thus, for water it would be (12) where the chemical potential can be expressed as (13) For foods with high-moisture and low-macromolecular content (i.e., fruits), the contribution of the matric potential Q) can be neglected; however, during an osmotic dehydration (OD) operation there are normally no pressure gradients except for the contribution from the cell turgor, which is supposed to disappear in the first stages of the process. Therefore, the equilibrium relation remains
^OsmoUcsoluUon^FLP. (14)

When the main solute is the same in the osmotic solution and FLP, the relation can be further simplified to z7 = / Kinetic Considerations The OD model assumes that mass transfer is controlled by internal resistances, so experimental data used for verification were obtained in conditions designed for making sure that this assumption was reached. Mechanisms and Rate Equations Recently a group of researchers (Fito et al., 1996) classified the mechanisms acting on OD into two groups: (1) those depending on aw gradients (namely, concentration gradients described in this chapter as "pseudodiffusionals," which include Fickian and osmotic mechanisms that involve individual cell interaction, cells in the external surface of the solid external solution, and cells in the internal surface of the pores); and (2) those depending on pressure gradients and strongly affected by food structure (including capillary and external pressure effects), which are both addressed by the HDM (Fito and Pastor, 1994; Fito, 1994; Fito et al., 1996). (15)

Furthermore, when the process pressure is close to the osmotic solution's vapor pressure [as may occur in vacuum osmotic dehydration (VOD)], a condensation-vaporization mechanism could appear since the concentration gradients in the food imply vapor pressure gradients in the FLP (Fig. 13-3). In cases involving low temperatures (T < 400C), active transport may also play an important role in solute transport (Yamaki and Ino, 1992). In this model, however, it is assumed that the HDM acts very quickly, and the effect on the food is supposed to take place at t = O. Rate Equations for the Pseudo-Diffusional Mechanism The use of driving forces with overall concentrations (x) is quite common in the literature, but this procedure usually leads to the use of different water and solute diffusivity values. It should be noted that there is also some degree of indetermination in the values of water and solute concentration in the limits of the system since they are dependent not only on thermodynamic considerations but also structural changes as they affect the FLP/SM rate evolution. In fact, several papers have proposed related empirical equations, so it seems more convenient to assume that the kinetics of this group of mechanisms is controlled by the gradients of component activities in the FLP that can be simplified by

(16)

where i is the mass fraction of the j component in the FLP. When only two components are transferred (water and total solutes), the kinetic parameter De would be a pseudo-diffusivity (or apparent diffusivity), and the following relation for a binary system could be established: (17) This apparent diffusivity would be related to the food structure by (18) where Dws is the diffusivity of solutes in water in a solution and T is the "tortuosity" of the pores in the solid food. The reduced driving force in FLP is defined as

Structured Food = Solid Matrix & Food Liquid Phase (SM) (FLP)

External Liquid (OS) Pmin

Working Pressure P Ideal Pore Volume where ECM occurs

P'min distance

Pwa w

Figure 13-3. Scheme of evaporation-condensation mechanism (ECM). When vacuum is working the minimal working pressure is usually considered as P' min instead of P min. The ECM occurs when the working pressure P lies between: P' min <P<P min.

(19) where Ze 1S the water concentration of FLP in equilibrium with the osmotic solution. Utilizing Equation 14 results in (20) (Note: the equivalence between y^and Y?T justifies Eq. 17). With this hypothesis, the evolution of Y during the process will be the result of the HDM's action over a very short period (assumed to be an instantaneous effect), followed by that of those mechanisms summarized in Equation 17. T will be calculated by (21) HDM's Effect on the Global Mass Transfer Kinetics The term YY HDM,/=O *n Equation 21 will be calculated from the composition ' of the FLP as described by Fito et al. (1996): X-Y where: (23) (24) (25) (26) The initial value of xifi changes according to the effect of HDM on
-*0
1

= (ee + y)(l-l/r)-Yl

(22)

HDM'

(27)

where y = O for OD, Y = Y1 for VOD, and Y = Y f r pulsed vacuum osmotic dehydration (PVOD) (see Eq. 22). Mass Transfer Kinetics Calculations The term Ytw\ in Equation 21 can be evaluated by an integrated

solution of Equation 16 (Crank, 1975). For semi-infinite plates the expression is (28)

The following simplified expression can next be used (29)

Then: (30)

Utilizing the definitions of the reduced driving forces produces

(31)

In this equation, it could be expected that if the Fickian mechanism is predominant and the solute concentration distribution in the product is uniform at time zero a constant K = 8/rc2 would be obtained, and in that case, De would represent a true diffusivity. Retention of the Liquid Phase by the Solid Matrix Kinetic modeling as described above allows one to analyze and predict the evolution of FLP concentration throughout the process. This information is also

useful for estimating the aw (or even jiw) in the product after any time of treatment; however, the evolution of the global concentrations and the net mass fluxes can only be calculated if the mass ratio of FLP/SM during the process is known, and this can be achieved by one of the following procedures: 1. Plotting the experimental values of the water, solutes, and inert mass fractions in a triangular diagram, as is usually done with solid-liquid operations in food engineering when two phases and three components are present in a system. (A relationship between the water and total solute mass fraction can be obtained if the inert is considered invariant as in Fig. 13-4.) 2. Empirically analyzing the mass changes as a kinetic process. The first procedure allows one to graphically analyze the operation's evolution and relate the FLP concentrations with those on a global scale, but the prediction of global mass fluxes and weight changes can raise important errors when the inert concentration in the food is very low, so the second procedure is recommended if this is the case. Procedure 1 Typical water and solute mass fraction evolutions are shown in Figure 13-4. Data obtained with fruit usually fit a straight line well; for example, point O would represent the raw material, P the final product, and slope b = 1 means there is no change in the FLP/FM rate throughout the process. The slope of the line is also related to the structure changes suffered by the product due to osmotic dehydration. The equation of the straight line can be written as: (32) where b is the slope. This value, which should be calculated experimentally for each process, is dependent upon product and osmotic solution characteristics as well as process conditions. Procedure 2 Weight loss has already been defined in Equation 5, but the following represents the empirical expression:

Figure 13-4. Typical plot of water and solute concentrations during an OD operation.

(33)

where It1 is the HDM's contribution to the total mass variation of the sample, and the latter term is the global effect of the pseudodiffusional mechanisms on the sample's weight change. Then:

(34)

The above quotient may be calculated theoretically from the HDM equations as:

(35)

Mathematical Model Assessment Figure 13-5 shows the simplified general layout of the mathematical model as has been explained above; previous work (Barat et al., 1996) has provided results for OD and PVOD (Fito, 1994) on apple (var. Granny Smith) rings that were presented as a way to assess model behavior. Experiments were run at 400C by using an osmotic solution of sucrose-water (65 kg sucrose/100 kg solution), and PVOD experiments were applied at 50 mbars. Conditions were maintained to assure that internal resistance controlled the mass transfer kinetic, with the time of treatment ranging from 15 to 540 minutes. Figure 13-6 shows the data coherence in all the experiments between the values of the weight losses and the sum of solutes and water transferred as predicted by Equation 11. Analysis of changes in fructose and glucose concentration in the sample during the experiments showed that sucrose represented the major solute transferred (Fig. 13-7), probably because glucose and fructose are preferentially located inside the cells and were not transferred to the external liquid (Yamaki and Ino, 1992). (In this situation Eq. 15 may be used to illustrate the equilibrium relationship.) The values of the parameters obtained by applying the mathematical model to the experimental results of apple osmotic dehydration can be observed in Table 13-2 (Barat et al., 1996) which shows that De values are similar to those published in other works (Pastor, 1994). It is interesting to also note the values of b since it is evident that the more important reduction in liquid retention capacity is produced when OD is applied. In the case of PVOD, however, the effect of HDM increased the values of the FLP/SM rate in the samples throughout the operation. Table 13-2 parameters were used to predict the evolution of concentrations and mass transfer rates with the time of treatment, and Figure 13-7 shows the coherence achieved between experimental and predicted values in all the cases. The mathematical model was also applied to some previously published experimental results from OD and PVOD of pineapple (Pensaben, 1994) and banana (Sousa, 1996). Table 13-3 contains the values of the obtained parameters, and Figures 13-8 and 13-9 plot the predicted values for concentrations and mass transfer rates against the experimental ones (with close agreement). As noted in the case of apples, the obtained values of De were similar to those previously published.

General Conclusions The approach made to model the solid-food liquid operation permits one to preliminarily propose a group of mathematical equations to predict the evolu-

PREDICTED MODELTARGET

Overall Compositions FLP/SM rate

FLP Composition
Predictive equations

Product Properties

CONCERNED WITH : Food quality and stability

Shelf life

Eqs. 27 and 31 Mass transfer rates Model parameters

Some physical and sensorial properties

Economic benefits
MODEL APPLICATION Option A using I and III

FLP hold-up

Good prediction of composition Bad prediction of mass fluxes

Option B using I and II

(FLP/SM) (t)

One parameter more than A Good prediction of composition and fluxes

Eq. 32

Figure 13-5. Model layout.

Experimental (OD) Predicted (OD) Experimental (PVOD) Predicted (PVOD)

Figure 13-6. Checking of mass balances in OD and PVOD experiments on apple slices.

tion of sample concentrations, mass transfer rates, and liquid retention capacities with times of treatment. In the case of OD and PVOD treatments with watersucrose as the osmotic solution, the model was tested with the experimental results from 272 experiments arranged in 92 triplicate series with different fruits and work conditions. The experimental results were always close to those calculated with the model; nevertheless, the model must be improved by looking for relationships among the parameters and operation variables (T, P9 osmotic solution composition), as well as testing more results obtained in wider range of experimental conditions.

Predicted (%)

Experimental (%)

Predicted (%)

Experimental (%)

Figure 13-7. Predicted versus experimental values of concentration and fluxes in apple slices subjected to (A) OD and (B) POD with sucrose (65% w/w) at 40 Cfrom0 to 540 minutes

Table 13-2. Values of Model Parameters for Apples (Barat et al, 1996) Parameters Dm {mis1) k Ic1 Ic2 ^ b OD 4.34 E-IQ 1.0710 0.0220 -0.0199 0.2000 -1.0490 PVOD 4.85 E-IO 0.9740 0.0940 -0.0155 0.2000 -0.9880

Notation Y *F a pa \ij pr aJ De ee eT M P P1 P2 Pc / rso T t V V X JC1 Y y i Strain-relaxation module (mVm3) Matric potential (N/m2) Surface tension (N/m) Apparent density (kg/m 3 ) Chemical potential of j component in phase i (J/mol) Real density (kg/m 3 ) Water activity in phase i Effective diffusivity (m/s2) Effective porosity Total porosity Sample mass (kg) Pressure (N/m2) Work pressure (N/m2) Atmospheric pressure (N/m2) Capillary pressure (N/m2) Half thickness of dehydrated product (m) Density of osmotic solution (kg/m 3 ) Temperature ( 0 C) Time (s) Volume (m3) Partial molar volume of water (mVmol) Pore volume fraction occupied by impregnating solution Mass fraction of j in food ( kgj/total kg) Reduced driving force Mass fraction of j in osmotic solution Mass fraction of j in food liquid phase

Table 13-3. Values of Parameters from the Mathematical Model Applied to Experimental Data of Pineapple (Pensaben, 1994) and Banana (Sousa, 1996) Product Pineapple Pineapple Pineapple Pineapple Pineapple Pineapple Banana Banana Banana Banana Banana Banana Treatment OD PVOD (50) OD PVOD (50) OD PVOD (50) OD PVOD (100) OD PVOD (100) OD PVOD (100) T (C) 40 40 50 50 60 60 25 25 35 35 45 45 k 0.16 0.16 0.16 0.16 0.16 0.16 0.10 0.10 0.10 0.10 0.10 0.10 0.963 0.863 0.913 0.810 0.763 0.747 1.003 0.994 0.954 0.968 0.957 0.951

D.E10
(mVs) 3.43 4.55 5.13 5.11 4.68 5.95 4.19 5.02 3.82 4.06 3.89 4.51

K
0.000 0.050 0.000 0.066 0.022 0.094 0.036 0.036 0.000 0.007 0.024 0.078

-Jfc2E3
1.16 0.67 1.54 0.83 1.75 1.68 2.51 2.52 2.30 2.66 2.56 1.90

1.14 0.98 0.84 1.00 0.90 0.97 1.38 1.15 1.18 1.15 1.26 1.33

I I I I I I I I I I I I

Predicted (%)

Experimental (%)

Predicted (%)

Xst AIVW AMs AM

Experimental (%)

Figure 13-8. Predicted versus experimental values of concentration and fluxes in apple slices subjected to (A) OD and (B) POD with sucrose (65% w/w) at 40-60 C from 0 to 180 minutes

Predicted (%)

Experimental (%)

Predicted (%)

Experimental (%)

Figure 13-9. Predicted versus experimental values of concentration and fluxes in apple slices subjected to (A) OD and (B) POD with sucrose (65% w/w) at 25-35-45 C from O to 130 minutes

Superindexes I j=\=w j o Sj ST Subindexes FLP HDM o PD t Liquid fraction of food Hydrodynamic mechanism Initial values (t = 0) Pseudodiffusion Values at time t Inert (solid matrix of food) Water Food components (j = 7, 2, ...n) Total mass Solute; (j = 2...n) Total solutes

References Barat J, Alvarruiz A, Chiralt A, Fito P (1996). Aproximacion a un modelo matematico de Ia operaci6n de deshidrataci6n osmotica de alimentos. Proceedings del Primer Congreso Iberoamericano de Ingenieria de Alimentos. Crank J (1975). The Mathematics of Diffusion. London: Oxford University Press, pp. 44-68. Fito P (1994). Modelling of osmotic dehydration of foods. In Fito P, Mulet A, Mckenna B (Eds.): Water in Foods: Fundamental Aspects and their Significance in Relation to Processing Foods. London: Elsevier Applied Science, pp. 313-328. Fito P, Pastor R (1994). On some non-diffusional mechanism occurring during vacuum osmotic dehydration. J Food Eng 21:513-519 Fito P, Andres A, Chiralt A, Pardo P (1996). Coupling of hydrodynamic mechanism and deformation-relaxation phenomena during vacuum treatments in solid porous food-liquid systems. J Food Eng 21:229-240.

Le Maguer M, Yao ZM (1995). Mass transfer during osmotic dehydration at the cellular level. In Barbosa-Canovas GV, Welti-Chanes J (Eds.): Food Preservation by Moisture Control: Fundamentals and Applications. Lancaster, PA: Technomic Publishing Co., Inc., pp. 325-350. Pastor R (1994). Modelizacion de los Fenomenos de Transporte en Ia Deshidratacion Osmotica a Vacio de Manzana. PhD thesis, Universidad Politecnica de Valencia, Spain. Pensaben M (1994). Deshidratacion Osmotica a Vacio de Pina y Mango. Ph.D. thesis. Universidad Politecnica de Valencia, Spain. Sousa R (1996). Aplicacion de Ia Impregnacion y de Ia Deshidratacion Osmotica al Desarrollo de Producto sde Banana (Musa Cavendish). Ph.D. thesis, Universidad Politecnica de Valencia, Spain. Yamaki S, Ino M (1992). Alteration of cellular compartmentation and membrane permeability to sugars in immature and mature apple fruit. J Amer SocHortSci 117:951-954.