Proceedings of the 17th World Congress

The International Federation of Automatic Control

Seoul, Korea, July 6-11, 2008

Fuzzy logic application to drying kinetics modelling

Henry A. Váquiro *. José Bon **. José L. Diez ***

*Faculty of Agronomic Engineering, University of Tolima, Ibagué, Colombia (e-mail: hvaquiro@ tal.upv.es).
** Department of Food Technology, Polytechnic University of Valencia, Spain (e-mail: jbon@tal.upv.es)
*** Department of Systems Engineering and Control, Polytechnic University of Valencia,
Camino de Vera 14, 46022 Valencia, , Spain, (Tel: +34-963877007 ext. 75794; e-mail: jldiez@isa.upv.es)

Abstract:
A Fick’s model that includes a Takagi-Sugeno fuzzy model to estimate the effective diffusivity was
analyzed. The modelling of drying kinetics on mango trough this diffusional-fuzzy model was compared
with the theoretical Fick’s model and the empirical Peleg and Weibull models. The identification and
validation was performed from experimental drying curves of ripe mango slices (Mangifera indica L. cv.
Tommy Atkins) at constant air velocity (4 m/s) and different drying temperatures (40, 50, 60 and 70 ºC).
The fuzzy sets for the antecedent of Takagi-Sugeno system were identified by the Gustafson-Kessel
clustering algorithm and approximated by membership functions of piecewise exponential form. On
identification and validation, the diffusional-fuzzy model showed best results than the Fick’s model,
whereas it showed little difference with the Weibull and Peleg models. The diffusional-fuzzy model keeps
the interpretability of Fick’s model, improves the process simulation and avoids phenomenon and property
considerations which require additional experimental and modelling work.

levels of production and importation, mango is ranked as the

1. INTRODUCTION
third tropical fruit and occupies the fifth place on all fruits on
Over the last years, dehydration food processes have the world (FAO, 2005). Its dehydration is an alternative in
advanced through merging of new technologies that make order to utilize production surplus and increase the
possible a more deep and complete study of processes. On diversification and offer of added-value products (De La
this way, the modelling techniques have been applied for Cruz Medina & García, 2003).
process design, control and optimization, as well as the study
The heterogeneity on solid-soluble and fiber content within a
and research of involved phenomena (Mujumdar, 2006a).
mango fruit could influence on the mechanisms that govern
These efforts has been realized on hot-air drying, one of the the drying process and difficult the representation by the
most ancient, common and diverse processes of conservation. diffusion model when the material properties are considered
Also, it is one of the industrial activities that demand more constant.
energy resources (Mujumdar, 2006b), in consequence, it
In this study the modelling of drying kinetics on mango
causes considerable environmental effects, at the same time
trough a Fick’s model that includes a Takagi-Sugeno fuzzy
that affects significant parameters of commercial acceptance
model to estimate the effective diffusivity was analyzed.
as organoleptic and nutritional quality of products
Thus, an alternative model to improve the process simulation
(Jayaraman & Das Gupta, 2006, Raghavan & Orsat, 2006).
was obtained. The resulting estimations were compared with
In order to advance on drying technologies, R&D approach the theoretical Fick’s model and the empirical Peleg and
leads to techniques that offer quality products and reduce the Weibull models.
environmental impact by means of industrial energy
efficiency. This approach includes, in relevant way, the 2. PRELIMIMARIES
establishment of mathematical models to study, develop or
In the case of foods and other biological materials, the
improve new or existing process and equipments (Mujumdar,
convective drying normally describes the water removal
2006a).
process from a solid material by exposition to a hot-air flux.
The fuzzy modelling uses artificial intelligence techniques to It is an operation realized to minimize the chemical and
model non-linear systems by sub-models integration. It is an enzymatic reactions or prevent the microbial growing and
alternative methodology on some complex processes that do degradation by reducing to a given level the moisture content.
not enable an effective application of traditional modelling This technology occupies an important place on the foodstuff
techniques and require approximated and fast solutions (Diez, transformation and post-harvest treatments. Also, it entails
2003, Labadini & Baker, 2006, Ross, 2004). the diminishing of volume and weight of products that, as
fruits and vegetables, minimize packaging, storage and
The mango (Mangifera indica L.) performs an important
economic role for several nations world-wide. According to

978-3-902661-00-5/08/$20.00 © 2008 IFAC 2206 10.3182/20080706-5-KR-1001.2212

17th IFAC World Congress (IFAC'08)
Seoul, Korea, July 6-11, 2008

transport costs (Jayaraman & Das Gupta, 2006, Raghavan & Other important empirical model is the probabilistic Weibull
Orsat, 2006). model, which describe the degradation kinetic of foodstuff
submitted to stress conditions by a given time. This has been
The food responses during the drying depend on product and employed to describe the drying on diverse products in the
hot-air properties of mass and heat transfer. The knowledge form of Page’s model (García-Pascual et al., 2006):
of the temperature and moisture distribution and its evolution
through time is essential to process and equipment design, ⎛ ⎛ t ⎞
k w1 ⎞
quality control and selection of suitable storage conditions M = M e + ( M 0 − M e ) exp⎜ − ⎜⎜ ⎟
⎟
⎟. (6)
⎜ k ⎟
and handling practices (Özilgen & Özdemir, 2001). ⎝ ⎝ w2 ⎠ ⎠
On the drying process, the variables of interest can be Here, kw1 and kw2 were the shape and scale parameters
estimated by mathematical models that represent the mass respectively.
and heat transport phenomena within the product. In this
sense, the diffusion model based on the Fick’s second law is 3. METODOLOGY
one of the most used and cited for agro-food applications.
This theoretical model relates the experimental results to From experimental information, three kinds of models were
physic laws (García-Pascual et al., 2006). used to describe the drying kinetic: the theoretical Fick’s
model, the empirical models of Peleg and Weibull, and the
The variation of moisture content across an infinite slab of Fick’s model that integer the fuzzy model to calculate the
material as function of time using the Fick’s second law, effective diffusivity.
considering that moisture transport from interior to surface is
mostly performed by liquid diffusion and the material is Matlab® was used as computational platform to program the
homogeneous and isotropous, can be expressed as (Crank, algorithms of fuzzy-model identification, theoretical and
1975, García-Pascual et al., 2006): empirical model identification, and numerical solving of
partial differential equations of Fickian models.
∂M(r,t) ∂ ⎛ ∂M ⎞
= ⎜ De ⎟. (1)
∂t ∂r ⎝ ∂r ⎠ 3.1 Experimental information
M(r,t) is the local moisture content at time t (kg water / kg
dried solid), t is the time (s), De is the effective diffusivity Physical and chemical product properties are significant on
(m2/s) and r is the distance at slab centre in thickness movement moisture mechanisms that govern the mango
direction (m). drying process. The mango fruit commonly has fibrous pulp,
with high content of soluble solids and water. Its
The formulation of the mass transport process on the slab heterogeneous ripening process is more pronounced at the
proceeds defining its initial and boundary conditions. The exterior and the sunlight incidence zones. This characteristics
initial moisture content in the slab is uniform (2). The influence on the existence of zones with different solid-
condition of system symmetry for the geometry is assumed soluble and fiber content, fiber orientation and consistency
(3). The moisture content at surface corresponds to within a mango fruit. Likewise, on drying process they cause
equilibrium moisture content with hot air (4) when the effect a product response of difficult representation by conventional
to external resistance mass transfer is negligible (Simal et al., theoretical models. Particularly, in the models based on the
2003). diffusion theory, the influence of those characteristics on the
shrinking and internal resistance to transport phenomena
M(r,0) = M 0 . (2)
should be assumed, raising in a substantial manner the
∂M ( 0,t ) complexity without to improve the solving accuracy.
=0↔t =0. (3)
∂t
M ( L ,t ) = M e . (4) The identification and validation of the considered models
was performed from experimental data for drying curves of
M(L,t) is the moisture content in the surface at time t ripe mango slices (Mangifera indica L. cv. Tommy Atkins,
(kgw/kgds), Me is the equilibrium moisture content (kgw/kgds), ºBrix = 13.01 ± 0.93, acidity = 0.516 ± 0.078 % malic acid,
L is the slab half-thickness (m). thickness = 5 mm), dried by hot air at constant conditions
during the process. The experiments was realized four times
In the empirical models the proposed by Peleg, to describe at 40, 50, 60 and 70 ºC and 4 m/s hot air conditions until a
the sorption curves, also has been used on the dehydration average moisture content around of 0.25 kgw/kgds (Fig. 1).
and rehydration modelling on diverse foodstuffs (Ruíz-Díaz Three of the temperature series were used to fuzzy sets and
et al., 2003, Simal et al., 2003, García-Pascual et al., 2006). model parameters identification, while the last one was used
The Peleg model applied to dehydration as shown in (5), to model validations.
where M is the average moisture content (kgw/kgds), M0 is
the initial moisture content and, kP1 and kP2 are the model Through the discrete differentiation of average moisture
parameters. content as function of time the drying rate was calculated.
The drying rate and the average moisture content (Fig. 2)
t were selected as input variables to identify the fuzzy sets in
M = M0 − . (5)
k P1 + k P 2 t the Takagi-Sugeno model, since the diffusion usually
depends on moisture content and symbolizes the transport

2207
17th IFAC World Congress (IFAC'08)
Seoul, Korea, July 6-11, 2008

coefficient in the diffusional mass transfer phenomena that zj −z

govern the drying mechanisms (Saravacos, 2005). z*j = . (7)
σ
6 Here, zj* is the normalized datum of zj, z and σ are the
40ºC mean and the standard deviation of data respectively. The
50ºC asterisk denotes the normalized data.
M (kgw / kgds )

4 60ºC
70ºC 4

2
⏐
2

*
(dM/dt)
__
0
0 0.5 1 1.5 2 0
t (s) 4
x 10

Fig. 1. Drying curves: average moisture content vs. time. -2

Experimental data for one temperature series. -2 -1 0 _* 1 2 3
M
-3
x 10
3 Fig. 3. Fuzzy clustering of normalized drying rate and
40ºC average moisture content by Gustafson-Kessel algorithm.
dM/dt (kgw / kgds⋅s)

50ºC Data at 40ºC.

2 60ºC
70ºC The rule-base for the Takagi-Sugeno model were established
as (8), where A is the fuzzy partition projected on M , Ai is
1 the antecedent fuzzy set for each class i (i = 1, 2, 3), Di is the
⏐
effective diffusivity coefficient (m2/s) in the consequent
associated to each antecedent fuzzy set.
0
0 1 2__ 3 4 5 6
R1 : If M is A1 then D1 .
M (kgw / kgds )
R2 : If M is A2 then D2 . (8)
Fig. 2. Drying curves: drying rate vs. average moisture R3 : If M is A3 then D3 .
content. Experimental data for one temperature series.
The fuzzy partitions were approximated by membership
functions of piecewise exponential form (Fig. 4) to represent
3.2 Diffusional-fuzzy model formulation
mathematically each antecedent fuzzy set.
In this work the diffusional-fuzzy model concerns to Fick’s 1
model that integer a Takagi-Sugeno model for estimating the
effective diffusivity. 0.8
Membership degree

0.6 * *
The Takagi-Sugeno model is a base-rules fuzzy system that A3 A2 *
A1
combines three fuzzy sets in the antecedent and associates 0.4
each of them with effective diffusivity coefficients in the
consequent. It is used to estimate the effective diffusivity as 0.2
function of its membership to high, medium and low
0
moisture ranges. -2 -1 0 _* 1 2 3
M
The identification of fuzzy sets was carried out by clustering
through application of the Gustafson-Kessel algorithm on the Fig. 4. Projection of fuzzy partition A* over normalized
normalized input variables (average moisture content and moisture content M * and approximation of membership
drying rate) (Fig. 3). Thus, antecedent fuzzy sets were functions. Data at 40ºC.
established from the fuzzy partition projected on average
The denormalized membership functions are shown in (9),
moisture content.
(10) and (11).
The Zscore normalization (7) was applied on input variables
⎧1 − A2 − A3 if M < α 1 .
before clustering because distance norm can be sensitive to A1 (M , α 1 , α 1 ,β) = ⎨ (9)
variations in the numerical ranges of the different data (Chen ⎩1 − A2 if M > α 1 .
et al., 2001). This normalization translates and scales the
experimental data so that all the values have zero mean and
unit variance.

2208
17th IFAC World Congress (IFAC'08)
Seoul, Korea, July 6-11, 2008

⎧ ⎛ M − α1 ⎞ 2
⎪ −⎜⎜⎝ β ⎟⎟⎠
⎪e if M < α 1 .
⎪ ⎛ M −α 2 ⎞2 3.4 Statistical analysis of parameters
A2 (M , α 1 , α 2 ,β) = ⎨ −⎜⎜ ⎟ (10)
β ⎟⎠
⎪e ⎝ if M > α 2 . As regards Fick, Peleg, Weibull and diffusional-fuzzy
⎪
⎪1 models, the explained variance (16, 17 and 18) was used to
⎩ if α 1 < M < α 2 . assess and compare how close the fit of models were. Note
⎧1 − A2 if M < α 2 . that the explained variance consider the number of model
A3 (M , α 1 , α 2 ,β) = ⎨ (11) parameters, that vary in each case, to achieve an adequately
⎩1 − A1 − A2 if M > α 2 . comparison among them.

( )
Posterior to definition of the denormalized membership
2
functions, the global structure of the consequent for the VAR = 1 − S yx S y2 . (16)
effective diffusivity estimation (12) was established as a 1
∑ [M j − M mean ] 2
N
function of the effective diffusivity coefficients (D1, D2 and S y2 = (17)
N −1 j = 1
D3) for each range of moisture content:
D f (M , α 1 , α 2 ,β ) = A1 D1 + A2 D2 + A3 D3 . (12)
2
S yx =
1
N − p −1 j = 1
[
∑ M j − M mean
N s
] 2
(18)

The diffusional-fuzzy model (13) is established by including N is the number of data, s superscript denotes the estimated
the effective diffusivity Takagi-Sugeno model (12) in the data by the model, VAR represents the relative variance
classic Fick’s model: explained by the model Syx2 with respect to total variance Sy2,
mean subscript denote the data average, and p is the number
∂M ( r ,t ) ∂ ⎛ ∂M ⎞
= ⎜Df ⎟. (13) of parameters.
∂t ∂r ⎝ ∂r ⎠
3. RESULTS AND CONCLUSIONS
The average moisture content was calculated by integration
of the moisture content profile (local values) in thickness The parameters of membership functions for the antecedent
direction using the equation (14). of Takagi-Sugeno model were defined from averages values
(Table 1) since the parameters exhibit similar values to
1L
L 0∫
M ( 0 ,t ) = M ( r ,t )dr . (14) different drying temperatures (fig. 5).
Table 1. Parameters of denormalized membership functions.
3.3 Solving and parameter identification Sample Drying temp. (ºC) α1 α2 β
1 40 2.090 3.174 0.296
Fick (1) and diffusional-fuzzy (13) models were solved
2 40 2.054 3.102 0.252
numerically considering the initial and boundary conditions
(2), (3) and (4). Although the analytical solution of the classic 3 40 1.963 2.979 0.304
Fick’s model is possible, the numerical solution was 4 50 2.167 3.091 0.398
considered to use a same algorithm for the identification and 5 50 2.131 2.881 0.384
simulation of the models based on Fick’s theory (classic and 6 50 2.254 2.743 0.422
fuzzy). Then, the numeric solution of the partial differential 7 60 1.980 2.961 0.282
equation was accomplished by the pdepe Matlab function,
8 60 2.253 3.106 0.432
tool that solves systems of elliptic and parabolic partial
differential equations in one spatial dimension and time. 9 60 2.229 2.989 0.256
10 70 1.971 2.988 0.318
Optimization problems to find the parameters for Fick, Peleg, 11 70 2.298 3.199 0.388
Weibull and diffusional-fuzzy models were formulated to 12 70 2.191 3.182 0.378
identify the optimal parameters that minimize the root mean
Mean 2.153 3.038 0.342
square deviation between experimental and estimated data
(15). As parameters of Fick and diffusional-fuzzy models a
constant effective diffusivity (De) and three effective
The identification and validation of the models for each
diffusivity coefficients (D1, D2 and D3) were considered,
drying temperature was done. The parameters identified are
respectively.
shown in Table 2, the models’ validation in Fig. 6 and the
root mean square deviations and the explained variances in
∑ j =1 (M j − M sj ) .
1 N 2
RMSD = (15) Table 3. The effective diffusivity calculated by Takagi-
N
Sugeno fuzzy model is represented in Fig 7.
Thus each mathematical model was solved in an optimization
algorithm that uses the fminsearch Matlab function The effective diffusivity coefficients identified for the
(multivariate unconstrained non-linear optimization based on Takagi-Sugeno model were inversely proportional to
the Simplex method). moisture content of the sample: 8.99e-11, 3.84e-10 and
5.19e-10 m2/s at 40ºC; 9.31e-11, 4.57e-10 and 7.15e-10 m2/s

2209
17th IFAC World Congress (IFAC'08)
Seoul, Korea, July 6-11, 2008

at 50ºC; 1.60e-10, 6.73e-10 and 1.03e-9 m2/s at 60ºC; 1.17e- 6

(a) Fick's model
6
(b) Peleg's model
10, 1.08e-9 and 1.65e-9 m2/s at 70ºC; for moisture ranges of Experimental
high (3.3 to 6 kgw/kgds), medium (1.8 to 3.3 kgw/kgds) and low Estimated

M (kgw / kgds )
(0 to 1.8 kgw/kgds) respectively (Table 2). 4 40ºC 4
50ºC
1 60ºC
(a) T = 40ºC 2 70ºC 2

__
0
0 1 2 3 4 5 0 0
1 0 1 2 0 1 2
(b) T = 50ºC t (s) 4 t (s) 4
x 10 x 10
Memberschip degree

0 (c) Weibull's model (d) Diffusional-fuzzy model

0 1 2 3 4 5 6 6
1
(c) T = 60ºC

M (kgw / kgds )
4 4
0
0 1 2 3 4 5
1
2 2
(d) T = 70ºC

__
0
0 1 2 3 4 5 0 0
1 0 1 2 0 1 2
t (s) 4 t (s) 4
(e) Mean x 10 x 10

0 Fig. 6. Model validation: Experimental and estimated data

0 1 2 __ 3 4 5
M using Fick (a), Peleg (b), Weibull (c) and diffusional-fuzzy
(d) models for each drying temperature.
Fig. 5. Denormalized membership functions.
Table 3. Root mean square deviations and explained
Table 2. Model parameters: Averages of three identified variances of the considered models. (*) On the identification,
values for each temperature. RMSD and VAR correspond to averages of three values for
Model Drying temperature (ºC) each drying temperature.
parameters 40 50 60 70
Identification* Validation
Fick Model T(ºC)
RMSD VAR RMSD VAR
De 3.56E-10 4.28E-10 6.35E-10 9.68E-10
40 0.0440 0.9712 0.0457 0.9689
Peleg
50 0.0534 0.9594 0.0441 0.9609
kp1 827.0 806.9 455.3 372.7 Fick
60 0.0502 0.9640 0.0550 0.9568
kp2 0.119 0.120 0.104 0.118 70 0.0563 0.9558 0.0465 0.9677
Weibull 40 0.0538 0.9988 0.0437 0.9979
kw1 1.011 1.063 1.046 1.080 50 0.0511 0.9988 0.0492 0.9981
kw2 5758 4787 3244 2112 Peleg
60 0.0543 0.9990 0.0435 0.9978
Diffusional-fuzzy 70 0.0392 0.9993 0.0463 0.9982
D1 8.99E-11 9.31E-11 1.60E-10 1.17E-10 40 0.0588 0.9986 0.0609 0.9971
D2 3.84E-10 4.57E-10 6.73E-10 1.08E-10 50 0.0793 0.9971 0.0903 0.9962
Weibull
D3 5.19E-10 7.15E-10 1.03E-9 1.65E-9 60 0.0907 0.9972 0.0774 0.9969
70 0.0795 0.9970 0.0723 0.9964
40 0.0046 0.9997 0.0071 0.9983
On identification and validation, the diffusional-fuzzy model Diffusional-fuzzy
50 0.0055 0.9996 0.0062 0.9985
showed best results than the traditional Fick’s model, 60 0.0051 0.9996 0.0067 0.9982
whereas it showed little difference with the empirical Weibull 70 0.0095 0.9987 0.0057 0.9986
and Peleg models, such as is indicated by root mean square
deviations and explained variances in Table 3.
The resulting Takagi-Sugeno fuzzy model enables to interpret
In contradistinction to empirical models that lack of meaning the effective diffusivity as function of moisture content
as regards heat and mass transfer phenomena, the diffusional- ranges. In addition, diffusional-fuzzy model preserves the
fuzzy model keeps the interpretability of Fick’s model and mathematic structure and the assumptions of principal govern
presents an alternative to effective diffusivity estimation that equation, thus phenomenon and property considerations that
improves the process simulation. Likewise, it enables to require additional experimental and modelling work are
estimate the moisture profiles inside the solid and estimates avoided. However, the existence of membership functions
its evolution through time (Fig. 8), interesting advantages on and effective diffusivity coefficients increase the number of
study, design and optimization of some drying process. parameters.

2210
17th IFAC World Congress (IFAC'08)
Seoul, Korea, July 6-11, 2008

x 10
-9
De La Cruz Medina, J. and H.S. García (2003). Mango: post-
2
harvest operations, In: Compendium on post-harvest
40ºC operations, chapter XX, AGSI/FAO, Online resource,
50ºC
1.5 last checked: 7.Jan.2008. http://www.fao.org/inpho
60ºC
García-Pascual, P., N. Sanjuán, R. Melis and A. Mulet
Df (m 2 /s)

70ºC
1 (2006). Morchella esculenta (morel) rehydration process
modelling. J. Food Eng., Vol.72, 346-353.
Jayaraman, K.S. and D.K. Das Gupta (2006). Drying of fruits
0.5
and vegetables. In: Handbook of Industrial Drying, 3rd
Edn (Mujumdar, A.S.(Ed)). CRC Press, Boca Raton,
0
0 1 2 3 4 5
Florida.
__
M (kgw / kgds )
Labadini, H.M.S. and C.G.J. Baker (2006). Fuzzy modelling
an alternative approach to the simulation of dryers.
Fig. 7. Effective diffusivity estimated by Takagi-Sugeno Proceedings of the 15th International Drying Symposium
fuzzy model. IDS 2006, Vol.A, 258-264.
Martín, B. and A. Sanz (2001). Redes neuronales y sistemas
x 10
4
(a) Fick's model x 10
4
(b) Diffusional-fuzzy model
borrosos, 2nd Edn. Ra-Ma, Madrid, Spain.
1.5 1.5
6 Mujumdar, A.S. (2006a). Innovation and globalization in
5
drying R&D. Proceedings of the 15th International
Drying Symposium IDS 2006, Vol.A, 3-17.
M (kgw / kgds )

1 1
4
Mujumdar, A.S. (2006b). Principles, classification and
t (s)

3
selection of dryers. In: Handbook of Industrial Drying,
0.5 0.5 2
3rd Edn (Mujumdar, A.S.(Ed)). CRC Press, Boca Raton,
1 Florida.
0 0 Nelles, O., A. Fink, R. Babuska and M. Setnes (2000).
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
r (m) x 10
-3
r (m) x 10
-3 Comparison of two construction algorithms for Takagi-
Sugeno fuzzy models. Int. J. Math. Comput. Sci., Vol.10,
Fig. 8. Estimated moisture profiles by Fick and diffusional- 835-855.
fuzzy models. Data at 40ºC. Özilgen, M., M. Özdemir (2001) A Review on Grain and Nut
Deterioration and Design of the Dryers for Safe Storage
From obtained results, it is deduced that fuzzy logic with Special Reference to Turkish Hazelnuts. Crit. Rev
integration to effective diffusivity within the Fick’s model Food Sci., Vol. 41, 95-132.
improves the simulation of drying process on Mangifera Raghavan, G.S.V. and V. Orsat (2006). Recents advances in
indica L. drying of bio-materials for superior quality bio-products.
Proceedings of the 15th International Drying Symposium
4. ACKNOWLEDGMENTS IDS 2006, Vol.A, 44-53.
Reznik, L. (1997). Fuzzy controllers. Newnes, Boston.
The authors would like to acknowledge the financial support Ross, T.J. (2004). Fuzzy logic with engineering applications,
of MCyT project (ref: AGL2006-14146-C02-01). 2nd Edn. John Wiley and Sons Ltd., London.
H.A. Váquiro received a grant from the University of Tolima Ruíz-Díaz, G., J. Martínez-Monzó, P. Fito and A. Chiralt
and COLCIENCIAS (Colombian governmental institute for (2003). Modelling of dehydration-rehydration of orange
science and technology development). slices in combined microwaveyair drying. Innovative
Food Sci. Emerging Tech., Vol.4, 203-209.
5. REFERENCES Saravacos, G.D. (2005). Mass transfer properties of foods. In:
Engineering properties of foods, 3rd Edn (Rao, M.A.,
Abonyi, J., R. Babuska, L. Nagy and F. Szeifert (2000). Rizvi, S.S.H, Datta, A.K. (Eds)). CRC Press, Boca
Local and global identification for fuzzy model based Raton, Florida.
control. Proceedings of the Intelligent Systems in Control Simal, S., A. Femenia, P. García-Pascual and C. Roselló
and Measurement Symposium INTCOM 2000, 111- 116. (2003). Simulation of the drying curves of a meat-based
Chen, C.L., H.X. Li and H.P. Huang (2001). Data product: effect of the external resistance to mass transfer.
preprocessing. In. Fuzzy neural intelligent systems: J. Food Eng., Vol.58, 193-199.
mathematical foundation and the applications in
engineering. CRC Press, Boca Raton, Florida.
Crank, J. (1975). The mathematics of diffusion, 2nd Edn.
Oxford University Press, Oxford.
Diez, J.L. (2003). Técnicas de agrupamiento para
identificación y control por modelos locales. PhD
Thesis. Polytechnic University of Valencia, Spain.
FAO (2005). Trade: crops & livestock primary & processed.
In: FAOSTAT. FAO Statistics Division. Online resource,
last checked: 7.Jan.2008. http://faostat.fao.org

2211