You are on page 1of 10

Journal of Food Engineering 144 (2015) 138–147

Contents lists available at ScienceDirect

Journal of Food Engineering


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

Drying of shrinkable food products: Appraisal of deformation behavior


and moisture diffusivity estimation under isotropic shrinkage
B. Ortiz-García-Carrasco a, E. Yañez-Mota a, F.M. Pacheco-Aguirre a,b, H. Ruiz-Espinosa a,
M.A. García-Alvarado b, O. Cortés-Zavaleta b, I.I. Ruiz-López a,⇑
a
Colegio de Ingeniería en Alimentos, Facultad de Ingeniería Química, Benemérita Universidad Autónoma de Puebla, Av. San Claudio y 18 Sur, Ciudad Universitaria, Puebla,
Puebla, Mexico
b
Unidad de Investigación y Desarrollo en Alimentos, Departamento de Ingeniería Química y Bioquímica, Instituto Tecnológico de Veracruz, Av. M.A. de Quevedo 2779, Col. Formando
Hogar, C.P. 91860 Veracruz, Veracruz, Mexico

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

Article history: A new methodology based on image analysis was proposed to estimate the simultaneous size reduction
Received 20 May 2014 (shrinkage) and shape change (deformation) during food drying. Potato strips (9.525 mm 
Received in revised form 21 July 2014 9.525 mm  80 mm) were used as model system and subjected to convective drying at 50, 60, 70 and
Accepted 28 July 2014
80 °C with an air velocity of 2 m/s. Developed protocol was used to analyze the shrinkage-deformation
Available online 7 August 2014
behavior occurring in minor product dimensions, considered the dominant directions in mass transfer.
To this purpose, 2D perpendicular slices were obtained from original 3D product and their digital images
Keywords:
were processed to evaluate the changes in contour shape, perimeter, and cross-sectional and specific
Image analysis
Mass transfer
areas of samples. Product contours were averaged to extract relevant deformation characteristics of dried
Shape change samples. Drying and shrinkage data were further used to estimate variable water diffusivities in product
Water diffusivity with a previously reported analytical solution for shrinking solids, which was extended to allow for 2D or
3D mass transfer. Studied responses were successfully described as a function of free moisture fraction
(R2 > 0.84). It was demonstrated that shrinkage-deformation behavior was not affected by drying temper-
ature under the tested conditions (p < 0.05). The analysis of averaged contour shapes showed that,
although shrinkage occurs from the beginning of drying, deformation appears at the final stages, when
the free moisture fraction is below 0.2. Mean water diffusivities were estimated in the range of
3.04–5.36  1010 m2/s for studied drying temperatures.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction reliable estimation of diffusion coefficients and its description


remains an actual subject (Souraki and Mowla, 2008; García-
Food products are complex systems which ordinarily require Pérez et al., 2012; Ruiz-López et al., 2012).
the reduction of their water content in order to extend their In order to gather the relevant information to complement a
shelf-life. In this regard, convective drying with hot air is the most drying model, the appraisal of product shrinkage can be achieved
widely used technique for the production of dehydrated foodstuffs. through universal methodologies including direct measurement
Air drying involves several heat and mass transfer mechanisms; of product dimensions or volume displacement techniques
yet, water diffusion within the product described by Fick’s second (Panyawong and Devahastin, 2007; Yan et al., 2008; García-Pérez
law is usually considered the controlling factor for modeling and et al., 2012; Ponkham et al., 2012; Curcio and Aversa, 2014). In
simulation purposes (da Silva et al., 2014; Pacheco-Aguirre et al., any case, dimensional changes are lumped in a single variable,
2014). This operation is accompanied by several changes in food, such as thickness or volume. Nevertheless, foods not only shrink
but shrinkage and deformation, besides color, are the most evident. during drying but also may suffer a considerable deformation
Thus, the inclusion of food shrinkage is paramount for making a (i.e., shape change), which can affect both the visual appeal and
packing properties of product. In addition, geometry deformation
may reflect the collapse of microscopic porous structure of food,
⇑ Corresponding author. Tel.: +52 222 2295500x7250. leading to a drastic change in its physical properties (Ruiz-López
E-mail addresses: irvingisrael@gmail.com, irving.ruiz@correo.buap.mx et al., 2012). Both assessment and prediction of food deformation
(I.I. Ruiz-López).

http://dx.doi.org/10.1016/j.jfoodeng.2014.07.022
0260-8774/Ó 2014 Elsevier Ltd. All rights reserved.
B. Ortiz-García-Carrasco et al. / Journal of Food Engineering 144 (2015) 138–147 139

Nomenclature

a specific area (m1) Y characteristic length for water diffusion along y-axis
A affine transformation matrix (m)
A cross-sectional area of product (m2)
Bim Biot number for mass transfer (dimensionless) Greek letters
C number of contours to be averaged d distance criterion between two contours
D effective diffusivity of water in food (m2/s) Dx horizontal translation (m)
d0 parameter for moisture diffusivity equation (m2/s) Dy vertical translation (m)
d1 parameter for moisture diffusivity equation (m2/s°C) c denotes the c th cyclic order rotation of coordinates in
e basis vector of the orthogonal coordinate system product contour
H absolute humidity of drying air (kg water/kg dry air) / rotation angle (rad)
hm external mass transfer coefficient (m/s) j height-to-width ratio (dimensionless)
K water partition ratio between gas and solid phases q volumetric concentration of dry solids (kg dry solids/m3
k1, k2 parameters for moisture diffusivity equation (dimen- product)
sionless) h modified Fourier number for mass transfer in shrinkable
L product length (m) products with time-dependent diffusivity (dimension-
mp product mass (kg) less)
n power constant (dimensionless) n axial coordinate along x-axis (dimensionless)
n normal unit vector w, W free moisture fraction (dimensionless): local and aver-
N number of points on a product contour aged, respectively
n1, n2 parameters for moisture diffusivity equation (dimen- f axial coordinate along y-axis (dimensionless)
sionless)
P perimeter of product contour (m)
b P Subscripts
P, P, edge coordinates of product contour: original, mirrored 0 at the beginning of the drying process
and averaged, respectively a for specific area
S product surface (m2) A for cross-sectional area
t drying time (s)
e at equilibrium
T cyclic order rotation of a given contour i at the air-product interface
u moisture content (kg water/kg dry solids) P for perimeter
v humid volume of drying air (m3 humid air/kg dry air)
r any reference product contour
V product volume (m3)
X characteristic length for water diffusion along x-axis (m)

represents a challenging task both numerically and experimen- generated data are further used to evaluate water diffusivity in
tally. Thus, several studies have focused on the characterization food system corrected for product shrinkage.
of macroscopic (external) shrinkage-deformation (SD) characteris-
tics of food products, mainly through using novel image analysis 2. Methodology
techniques.
Common properties obtained from digital images related with 2.1. Drying experiments
product SD behavior include projected area and its correspond-
ing perimeter, as well as selected sample dimensions which Two sets of air-drying experiments were conducted in order to
are further used to evaluate secondary indices such as elonga- obtain the SD behavior of potato strips as a function of their moisture
tion, fractal dimensions, roundness, etc. (Campos-Mendiola content. Fresh, well-graded potatoes were locally purchased
et al., 2007; Yan et al., 2008; Yadollahinia and Jahangiri, 2009; (Puebla, Pue., México) and dried the same day. Potatoes were
Yadollahinia et al., 2009; Khazaei et al., 2013). Recently, volume washed, dried with a cloth and sliced with a vegetable chipper
changes in dried products have been calculated using dimen- (9.525 mm-square openings) to produce strips that were further
sions obtained from lateral and top digital images including cut to their desired length (80 mm). On average, 12 regular slices
dual-camera setups (Sampson et al., 2014). However, while the were obtained from each tuber while remaining portions were
aforementioned characteristics are related to product quality reserved for the initial water/dry solids analysis. Drying experi-
indices, reported methodologies do not pursue a further applica- ments were conducted with 20 potato strips placed flat on a stainless
bility of SD data in the modeling and simulation of drying steel welded mesh open tray (dimensions: 0.25 m  0.20 m, open-
processes. For example, projected area is measured in non- ings: 4.5 mm  5.0 mm, wire diameter: 0.7 mm) in a tunnel dryer
dominant mass transfer directions (top area in flat slices) or is (Armfield UOP8, Ringwood, UK) with airflow parallel to the longest
affected by product bending in lateral views of samples. product dimension. Samples were dried at 50, 60, 70 and 80 °C for
Product deformation during drying imposes an additional diffi- about 330–470 min with an air velocity of 2 m/s. Drying curves were
culty: no food sample shrinks and deform in the same way, even obtained in the first experiment set, where moisture evolution was
under the most controlled conditions. Thus, the use of SD data calculated by continuously recording the weight of the product
for advanced process simulations would require the development throughout the process with trays that are carried on a support
of new protocols to extract relevant descriptors of product behav- frame that was in turn attached to a digital balance mounted above
ior. The main objective of this study is to develop and validate a the tunnel. A schematic view of the experimental setup was recently
new methodology based on image analysis to estimate the simul- presented in Pacheco-Aguirre et al. (2014). Moisture content in
taneous SD during food drying and obtain the representative pat- product was expressed as the dimensionless free moisture fraction
terns of this behavior using potato as food model. Besides, W (the removable water portion left in product) according to
140 B. Ortiz-García-Carrasco et al. / Journal of Food Engineering 144 (2015) 138–147

u  ue mp  mpe k-means clustering algorithm (Press et al., 2007) and reduced to


W¼ ¼ ð1Þ
u0  ue mp0  mpe 3 dominant color descriptors, which were enough to retain the
image details in product border. This operation facilitated both
These data were used to estimate the required time to approx-
background extraction of image with a foreground mask and esti-
imately achieve a specific moisture content in product at each dry-
mation of the pixel fraction corresponding to product. Quantized
ing temperature (from W = 0.1 to W = 0.9 in 0.1 increments) in
image without background was transformed to grayscale and
order to obtain SD data regularly spaced over this variable in the
finally coordinates of product boundary were obtained from this
second experiment set.
image. A total of 400 points were used to describe each product
With the purpose of evaluating the SD product behavior, groups
contour. Fig. 3 shows the image analysis steps used to determine
with 5 samples each were formed and dried for the predefined
product deformation. On a separated procedure the relationship
times. Then, a single transversal slice (perpendicular to the largest
between pixel number and real dimensions of reference object
dimension) of about 1 mm-thick was cut with a sharp blade from
was obtained, allowing the estimation of the slice area A. Boundary
the central part of the strip. A schematic view of samples along
coordinates were further used to estimate contour perimeter P as
with its relevant characteristics is shown in Fig. 1. Digital images
the cumulative sum of the Euclidian distance between consecutive
of resulting slices were immediately taken. Remaining product
points. All image analysis operations were performed with the
portions were analyzed for their moisture content. The aforesaid
procedure was also applied to fresh (W = 1) and equilibrium-dried Matlab Image Processing Toolbox 7.0 (Matlab R2010a, MathWorks
Inc., Natick, MA, USA).
(W = 0) samples. A total of five slices were obtained for each mois-
ture content-temperature combination.
Required moisture contents were determined by oven-drying 2.4. Modeling of shrinkage characteristics
(Binder ED 53, Germany) the samples at 105 °C until constant mass
weight (when mass change was less than 0.001 g over an 8 h per- By assuming a constant cross-sectional area (A) along major
iod). Initial moisture content of product was 84.5 ± 3.1 g water/ dimension (L), the volume (V) can be estimated as the quantity
100 g product (mean ± s.d.). AL (Fig. 1). On the other hand, total surface of sample (S) can be
expressed as the sum of lateral (PL) and minor (2A) faces. If
2.2. Image acquisition PL  2A, then specific area can be estimated with the simple
expression
Potato slices jointly with a reference object of known dimen- S PL þ 2A PL P
sions (a black-anodized metal washer of 0.59 cm-diameter) were a¼ ¼  ¼ ð2Þ
V AL AL A
placed on a blue paper sheet to provide plenty contrast for back-
ground extraction, and their digital images were acquired (Coolpix In addition, changes in slice perimeter and cross-sectional area
L810, Nikon Corp., Japan). Digital camera was positioned with its are proportional to changes in mass transfer surface and product
sight line normal to the supporting base. Illumination was pro- volume if shrinkage is considered negligible along major dimen-
vided through ordinary 18 W fluorescent lamps without special sion. Thus,
specifications as color standardization was not needed between S PL P
images. Images were taken with the maximum available resolution   ð3Þ
S0 P 0 L 0 P 0
(4608  3456 pixels) in macro mode with a focal distance of about
V AL A
0.10 m, using automatic settings. Digital images were stored in   ð4Þ
V 0 A0 L0 A0
JPEG format. The schematic view of the experimental image acqui-
sition setup is shown in Fig. 2. The following relationships are proposed in this study to relate
dimensional changes of product with moisture content:
2.3. Image analysis
A
¼ DA þ ð1  DA ÞWnA ð5Þ
A0
Two rectangular portions containing either the potato slice or
P
the reference object were manually selected from the original ¼ DP þ ð1  DP ÞWnP ð6Þ
image for their subsequent handling. Color information in these P0
a na
images was transformed to the CIELAB color space for their analy- ¼ 1 þ Da ekW ð7Þ
sis. Thus, every pixel was represented as a vector of color compo-
a0
nents L*, a* and b*. Afterward, color data were quantized using Eqs. (5) and (6) have been used to express the shrinkage behav-
ior of food products during drying (Zielinska and Markowski, 2010;
García-Pérez et al., 2012; Ruiz-López et al., 2012). Here, parameters
nA and nP both control the deviations from the straight-line behav-
ior, while DA and DP represent the final fractions of cross-sectional
area and perimeter at the end of drying, respectively. On the other
hand, Eq. (7) is an exponential decay model, similar to Page’s
model (Ruiz-López et al., 2012), where Da controls the relative
increase of specific area and k and na define the decaying rate
and shape of the curve, respectively.

2.5. Mean SD behavior

Since all dried samples shrink and deform in a different way it is


desirable to develop a strategy for appraisal of relevant and common
characteristics of this behavior. In this study, the contours were
Fig. 1. Schematic view of potato strips with their initial geometry (right square combined to obtain a mean deformation profile. However, contour
prism) and relevant characteristics. coordinates cannot be simply averaged after their estimation.
B. Ortiz-García-Carrasco et al. / Journal of Food Engineering 144 (2015) 138–147 141

Fig. 2. Schematic view of experimental setup for image acquisition and sample preparation: (1) digital camera, (2) fluorescent lamps, (3) reference object, (4) product slice,
(5) contrast background, (6) adjustable support, (7) dried product strip and (8) sharp blade.

Fig. 3. Image analysis steps used to determine product deformation: (a) original image, (b) simplified image with three color clusters, (c) gray-scale image after removing
non-product color clusters and (d) product contour. Image corresponds to an equilibrium-dried potato strip at 50 °C (440 min). Initial contour shape is a square. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Instead, they need to be translated and aligned with respect to P0i ¼ APi ð9Þ
some reference point as both position and orientation of objects
in digital images changes with every shot. Otherwise, SD contour where A is both a clockwise rotation matrix and a translation
characteristics would cancel or produce unexpected results when matrix
averaged. Thus, the following algorithm was developed in order 2 3
cos / sin / Dx
to align contours. Let us consider that Pi contains the ordered 6 7
homogeneous coordinates, in clockwise or anticlockwise direction, A ¼ 4  sin / cos / Dy 5 ð10Þ
of every point on contour i. Thus, 0 0 1

2 3 2 3 The parameters Dx, Dy and / should minimize some distance


xi xi;1 ... xi;N criterion between Pi and Pr. However, the straight comparison of
6 7 6 7
Pi ¼ 4 yi 5 ¼ 4 yi;1 . . . yi;N 5 ð8Þ these contours relates point pairs (xi,1, yi,1) with (xr,1, yr,1), (xi,2, yi,2)
1 1 ... 1 with (xr,2, yr,2) and in general form (xi,k, yi,k) with (xr,k, yr,k), which
not necessarily produce the best results. Consequently, both the
The alignment of contour Pi with respect to contour Pr involves the cyclic order rotation and mirrored projection of Pi were considered
translation of Pi coordinates in the amounts Dx and Dy, as well as to render the best overlap with reference contour. The cyclic order
their rotation in the angle / by means of the affine transformation rotation Ti,c of contour Pi was defined as
142 B. Ortiz-García-Carrasco et al. / Journal of Food Engineering 144 (2015) 138–147

2 3 2 3
xi;c xi;c ... xi;N xi;1 ... xi;c1 diffusion model and its boundary condition are written as
6 7 6 7 (Ruiz-López et al., 2012)
Ti;c ¼ 4 yi;c 5 ¼ 4 yi;c . . . yi;N yi;1 . . . yi;c1 5 ð11Þ
   
1 1 ... 1 1 ... 1 @u @ @u @ @u
¼ D þ D ð15Þ
An Euclidian distance criterion (/c) between Pr and Ti,c is given
@t @x @x @y @y
by hm @ui @ui
ðH  HÞ ¼ n  Dq ex  n  Dq ey ð16Þ
v i @x @y
2 T T
/c ¼ ðxr  xi;c Þðxr  xi;c Þ þ ðyr  yi;c Þðyr  yi;c Þ ð12Þ Let us consider that both diffusion coefficient D and character-
which correspond to the squared sum of x and y distances istic lengths for diffusion X and Y are explicit drying time functions,
between both contours. Parameters Dx, Dy, / and c achieving the i.e., D = D(t), X = X(t) and Y = Y(t). Then, Eqs. (15) and (16) can be
best overlap between Pi and Pr (those minimizing /) can be esti- rewritten as
mated by means of nonlinear regression analysis. This procedure @w @ 2 w 1 @ 2 w
is repeated with the mirrored contour ¼ þ ð17Þ
@h @n2 j2 @f2
2 3 2 3 @w 1 @wi
^i
x maxðxi Þ  xi;1 . . . maxðxi Þ  xi;N Bim ðtÞwi ¼ n  i ex  n  e ð18Þ
bi ¼ 6 ^i 7 6 7 @n j @f y
P 4y 5¼4 yi;1 ... yi;N 5 ð13Þ u  ue
1 1 ... 1 w¼ ð19Þ
u0  ue
Only the best overlap of contour Pi (achieved using the original DðtÞ
@h ¼ @t ð20Þ
or mirrored image) with respect to Pr, designated as Pi?r, is consid- ½XðtÞ2
ered in the rest of the analysis. Finally, contour coordinates can be 1 1
averaged to obtain a single representative image of product defor- @n ¼ @x; @f ¼ @y ð21Þ
XðtÞ YðtÞ
mation as YðtÞ ð1=v Þ hm dHi
P j¼ ; Bim ðtÞ ¼ K ; K¼ ð22Þ
Pr þ Pi!r
XðtÞ q ½DðtÞ=XðtÞ dX i
P¼ ð14Þ
C The term K in Eq. (22) is the local slope of equilibrium curve (an
Preliminary results indicated that any contour can be chosen as instantaneous partition coefficient). The variable transformations
the reference since the same final image is obtained in all cases. in Eqs. (20) and (21) allowed expressing the original problem for
The adequacy of averaged contour to represent the global SD moisture diffusion in a shrinkable body with variable diffusivity
behavior of product can be estimated by calculating the evolution as the simpler case of mass transfer for constant D in a non-
of its enclosed area which should be as close as possible to the shrinkable product (Ruiz-López et al., 2012). However, the
mean value of original contours at each moisture content-temper- unknown variable nature of Bim(t) in boundary condition (18) rep-
ature combination. Please notice that it is expected that both resents a difficulty for providing an analytical solution to the afore-
perimeter and specific area of averaged contour slightly deviate mentioned problem. Ruiz-López et al. (2012) demonstrated that if
from the mean values of original contours, especially at low the internal resistance to mass transfer by diffusion is accepted as
moisture contents, since the averaging procedure smooths out the only mechanism controlling drying rate throughout the drying
roughness and irregularities developed in product border. Fig. 4 process then existing solutions could be used to solve Eqs. (17) and
shows the steps used to determine the representative SD behavior (18) considering product shrinkage and variable water diffusivity.
of potato strips. Under this assumption boundary condition (18) is simplified to
wi ¼ 0 ð23Þ
2.6. Modeling of drying data The procedure to both evaluate water diffusivities and describe
drying kinetics of shrinkable food products proposed by
By assuming (i) negligible heat transfer, (ii) constant properties Ruiz-López et al. (2012) is now extended to products with mass
of drying air, (iii) constant volumetric concentration of dry solids, transfer in more than one direction. It should be noticed that the
(iv) uniform initial distribution of water within the solid and (v) following procedure is only applicable for products with isotropic
two-dimensional mass transfer in product, the unsteady-state shrinkage (solid shrinks in the same proportions in involved mass

Fig. 4. Contour manipulation steps used to determine a representative image of product deformation: (a) original, (b) aligned and (c) averaged contours (black contour was
chosen as the reference one). Contours correspond to an equilibrium-dried potato strip at 70 °C.
B. Ortiz-García-Carrasco et al. / Journal of Food Engineering 144 (2015) 138–147 143

Fig. 5. Effect of moisture content on relative cross-sectional area of dried potato Fig. 7. Effect of moisture content on relative specific area of dried potato strips.
strips.

A XY jX 2 pffiffiffiffiffiffiffiffiffiffiffi
¼ ¼ or X ¼ X 0 A=A0 ð26Þ
A0 X 0 Y 0 jX 20

In this study, Eq. (24) was numerically solved for experimental


W values with bisection algorithm, while derivatives dh/dt in Eq.
(25) were estimated using finite differences with second-order
accuracy. Estimated diffusivities were further used to evaluate its
dependence on moisture content and drying temperature with
the model

DðWÞ ¼ ðd0 þ d1 TÞ½1  expðk1 Wn1 Þ expðk2 Wn2 Þ ð27Þ


Mean water diffusivities for every drying temperature were
estimated from the numerical integration of instantaneous values
using trapezoidal rule according to

Z W2 Z W2
D¼ DðWÞdW dW ð28Þ
Fig. 6. Effect of moisture content on relative perimeter of dried potato strips. W1 W1

transfer directions, that is, j is constant). The analytical solution 2.6.2. Drying simulation
for Eq. (17) and (23) with mass transfer in x and y directions can If the dependence of both D and X on time or moisture content
be obtained from the well-known flat-slab solution and the super- are known, for example with Eqs. (26) and (27), then drying curves
position principle to obtain can be simulated from Eq. (20) by solving the initial value problem,
" #
X
1  2 2

p
W¼ 8 1
exp  ð2n1Þ h  ...
p2 ð2n1Þ2 4 dt ½XðtÞ2 ½XðWÞ2
"
n¼1
# ð24Þ ¼ ¼ with tðh ¼ 0Þ ¼ 0 ð29Þ
X
1   dh DðtÞ DðWÞ
2 2
p
p2
8 1
ð2n1Þ2
exp  ð2n1Þ
4
h
j2
n¼1
where W is calculated from h in each iteration with Eq. (24), and W
Eq. (24) allows the calculation of W for a given drying time is further used to evaluate both D(W) and X(W) .
expressed as variable h (or the inverse problem). Generalization of the just described procedures to 3D mass
transfer or other coordinate systems should be evident.
2.6.1. Moisture diffusivity estimation
Drying data, in the form of W vs. t and X vs. t (or W), can be used
for moisture diffusivity estimation using the following procedure: 2.7. Data analysis
(i) calculate h for every W value in the drying curve by solving Eq.
(24), (ii) calculate the derivative dh/dt and (iii) estimate moisture The fitness quality of identified models was quantified by the
diffusivities from Eq. (20) at every experimental t as determination coefficient (R2) and statistical significance of param-
eter estimates was evaluated through their 95% confidence inter-
dh
DðtÞ ¼ ½XðtÞ2 ð25Þ vals (95% CI). Numerical procedures, nonlinear regression (based
dt on ordinary least squares) and statistical analyses were performed
In our case, as cross-sectional area can be calculated with with the Matlab software and its Statistics Toolbox 7.3 (Matlab
Eq. (5), then R2010a, MathWorks Inc., Natick, MA, USA).
144 B. Ortiz-García-Carrasco et al. / Journal of Food Engineering 144 (2015) 138–147

Table 1 As demonstrated in Eq. (4), changes in cross-sectional area of


Regression parameters for shrinkage models. the studied geometry are related with changes in product volume.
Response Parameters Value 95% CI R2 Drying of potato strips resulted in samples with about 17% of their
Cross-sectional area a
DA 0.1733 0.1469/0.1997 0.9320 original size (Fig. 5). This shrinkage value is comparable to those
nA 0.7222 0.6732/0.7712 found in other studies, regardless of the sample geometry and
b
Perimeter DP 0.6390 0.6233/0.6547 0.8460 the method used to estimate it. For example, Wang and Brennan
nP 0.9581 0.8625/1.0537 (1995) and Hassini et al. (2007) reported potato slabs
c
Specific area Da 2.2641 2.1730/2.3553 0.9226
k 5.8974 4.8650/6.9297
(10 mm  45 mm  20 mm) shrinking up to 17–20% of their origi-
na 1.1257 1.0002/1.2512 nal volume when dried at 40–85 °C with air velocities ranging from
0.5 to 4 m/s. Lozano et al. (1983) also reported a final shrinkage
a
A=A0 ¼ DA þ ð1  DA ÞWnA .
ratio of approximately 19% for potato cylinders (1 cm in diameter,
b
P=P 0 ¼ DP þ ð1  DP ÞWnP .
c na
a=a0 ¼ 1 þ Da ekW . 4 cm long) dried at 40 °C with an air velocity of 1 m/s. Hassini et al.
(2007) determined volume of potato slabs using direct local mea-
surements of length, thickness and width, while Lozano et al.
3. Results and discussion (1983) and Wang and Brennan (1995) used volume displacement
techniques with toluene and water, respectively. According to sta-
3.1. Shrinkage characteristics of potato strips tistical analysis, the dependence of cross-sectional area on mois-
ture content shows a noticeable deviation from straight line
Figs. 5–7 show the dependence of cross-sectional area, perime- behavior (nA – 1, p < 0.05), indicating that dimensional changes of
ter and specific area of slices on moisture content. Initial values of product not only depend on volume of evaporated water, but also
these variables were determined by image analysis as on the collapse resistance of cell structure. As shown in Fig. 5, the
86.5779 mm2, 37.4430 mm and 0.4325 mm1, respectively. These size of potato strips was initially reduced in about 40% when 60% of
values were slightly different than those expected from size of available water was eliminated, but a comparable size reduction
chipper openings (9.525 mm), corresponding to 90.7256 mm2, was observed thereafter, when a smaller amount of water was
38.1 mm and 0.4199 mm1. Real product metrics produce initial evaporated (the remaining 40%). We hypothesize that as water is
values for volume (V0), and surfaces of both lateral (P0L0) and eliminated from product cell turgor is gradually loss, but rigidity
minor faces (2A0) of 6926.232 mm3, 2995.44 mm2 and of cell walls offers a resistance against shrinkage. However, as dry-
173.1558 mm2, respectively. Therefore, square faces of potato ing proceeds, the continuous water flux across membranes might
strips amount to less than 6% of the total available surface for mass rupture them once a critic moisture content is reached, debilitating
transfer, validating the assumption made in the development of Eq. the inner structure of product and causing a pronounced shrinkage
(2) that P0L0  2A0 (by little over 17 times). As expected, perimeter (0.1 6 W 6 0.4). Finally, at very low moisture contents water could
and cross-sectional area of samples decreased with moisture con- be removed with minimum product shrinking if collapse of cell
tent, while the specific area exhibited the opposite behavior. No structure is not complete, causing the development of an air-filled
significant effect of drying temperature was observed on shrinkage porous network, which reflects as a subtle tail of data in plot at
characteristics of potato (p < 0.05), which is in agreement with pre- W < 0.1. Shrinkage deviations from the straight-line behavior have
vious studies (Hassini et al., 2007). Thus, Eqs. (5)–(7) were fitted been well-documented in several studies with current product
over the entire temperature domain. Regression coefficients of (Lozano et al., 1983; Ratti, 1994; Wang and Brennan, 1995;
the models used for the mathematical description of these data Hassini et al., 2007).
are presented in Table 1. A satisfactory reproduction of these According to Eq. (3), changes in contour perimeters are related
responses was obtained in all cases (R2 > 0.84). to changes in available surface for mass transfer. Surface area of

Fig. 8. Final edge deformation of slices cut from dried potato strips.
B. Ortiz-García-Carrasco et al. / Journal of Food Engineering 144 (2015) 138–147 145

Fig. 9. Mean edge deformation of slices cut from dried potato strips (80 °C). Inner numbers represent the free moisture fraction reached in product/elapsed drying time (min).

Fig. 11. Comparison of perimeter estimated from original and averaged contours.
Fig. 10. Comparison of cross-sectional area estimated from original and averaged
contours.

Product volume lowered faster than available surface for mass


potato strips decreased with moisture content up to about 64% of
transfer, as expected from its proportionality with linear dimen-
its initial value (Fig. 6), with no significant departure from the
sions raised to the third, during drying causing a 3.26-fold increase
straight-line behavior (nP = 0.9581, 95% CI = 0.8625/1.0537). Image
of the specific area (Fig. 7), with a near first-order exponential
analysis has been previously used to estimate the surface area evo-
behavior (na – 1, p < 0.05). Comparable results were obtained by
lution of potato slices. Selected studies include those from
Ratti (1994) during drying (40–60 °C, 1–5 m/s) of potato disks
Campos-Mendiola et al. (2007) where the projected lateral area
(4 cm-diameter, 0.9 cm-thickness) and cylinders (1 cm-diameter,
of air-dried (55 °C, 1.7 m/s) circular potato slices (2.5 mm-
5 cm-length), reporting 2.18 and 2.95-fold increments in this
thickness, 40 mm-diameter) was measured, observing a significant
variable, respectively, where sample volume was determined by
increase of this variable, caused by sample bending during process.
liquid displacement and surface area was roughly estimated from
In later studies, Yadollahinia and Jahangiri (2009) and Yadollahinia
product dimensions.
et al. (2009) followed the evolution of the upper area of
circular slices (10 mm-thickness, 35 mm-diameter) during drying
(60–80 °C, 0.5–1 m/s), reporting final reductions between 50% 3.2. Deformation characteristics of potato strips
and 65% of the original values. These authors reported a marked
deviation from the straight-line behavior for this response for Deformation and shrinkage characteristics of the cross section
W < 0.1, caused by upward bending of product developing an of potato strips were successfully estimated with the proposed
irregular shape and affecting the measured area. methodology. It was found that all samples suffered a similar size
146 B. Ortiz-García-Carrasco et al. / Journal of Food Engineering 144 (2015) 138–147

for an excellent reproduction of cross-sectional area from original


data; consequently, size reduction estimated from averaged con-
tours is representative of observed behavior in individual samples.
On the other hand, perimeter is slightly underestimated (Fig. 11) at
lower moisture contents (W < 0.3, < 12% of relative difference)
when calculated from averaged contours since this is smoothed
during averaging procedure (Fig. 4). Consequently, the specific area
is also underestimated (<10% of relative difference, Fig. 12). Never-
theless, for simulation purposes observed discrepancies would not
translate in notably different results, as numerical convergence
would be reached before mesh fineness had to be increased to
unpracticals level to replicate surface roughness. The use of mean
SD behavior data in the modeling and simulation of drying process
will be demonstrated in the second part of this study.

3.3. Moisture diffusivity and process simulation

Fig. 12. Comparison of specific area estimated from original and averaged contours.
Drying curves from the first experiment set are presented, in
both linear and semi logarithmic representations, in Fig. 13. These
data were used to evaluate moisture diffusivities plotted in Fig. 14.
reduction and shape change along process regardless of drying As expected, the use of higher drying temperatures caused a signif-
temperature. A comparison of the contour shapes at the end of dry- icant increase in water diffusivity values (p < 0.05). Water diffusiv-
ing is presented as example in Fig. 8 (for W = 0), where it can be ity behavior can be divided in three sections. In the first region, the
verified that product surface developed irregularities or roughness gradual increase of water mobility from the beginning of the
with drying, which is in agreement with other studies (Campos- drying process up to W  0.6 indicates that product temperature
Mendiola et al., 2007). It should be emphasized that shrinkage is in a transitory state (preheating period) from initial product
and deformation behavior was unique for each sample, but they temperature to air temperature. This unsteady-thermal state is fol-
clearly exhibit a repetitive pattern. Thus, the use of a mean defor- lowed by an approximately constant water diffusivity period for
mation profile is desirable for a representative description of prod- 0.2 6 W 6 0.6. From then on, deceleration in water mobility might
uct shape changes during drying, both mathematically and be caused by cell structure collapse as suggested by the severe
qualitative wise. For all studied temperatures, product reduced product deformation observed at these moisture levels. Similar
its dimensions without a significant shape change up to a free trends have been previously reported during convective drying of
moisture content of 0.2, with an important deformation occurring several foodstuffs including chayote, fish muscle, grapes and fresh
thereafter, mainly manifested in a preferential contraction of the and pre-osmosed carrot cubes (Azzouz et al., 2002; Pinto and
middle section of contour edges toward the sample center, as Tobinaga, 2006; Singh and Gupta, 2007; Ruiz-López et al., 2012).
evidenced in Fig. 9 for product dried at 80 °C. The estimation of rel- Consequently, any constant diffusivity model would fail to repro-
evant SD characteristics of product is also desirable for simulation duce the whole experimental behavior when product preheating
of drying process, as the effect of product deformation on water or severe shrinkage/deformation cannot be neglected. In this case,
diffusivity could be evaluated without requiring particular data the suggested procedure allows the objective identification of
of a specific sample, especially if drying is conducted with several these three regions. Regression parameters for moisture diffusivity
slices. Thus, it is important to determine if an averaged contour equation (26) are shown in Table 2. A good reproduction of exper-
retains some characteristics of those from which it was obtained. imental behavior was achieved with proposed model (R2 = 0.9404).
Figs. 10–12 show a comparison of shrinkage properties (cross- Statistical analysis revealed that all constants were significant
sectional area, perimeter and specific area) from original and (p < 0.05), thus the model is structurally identifiable with current
averaged contours. As shown in Fig. 10, averaged contours allowed data (i.e., its parameters can be uniquely estimated).

Fig. 13. Experimental (dots) and predicted (lines) potato drying curves in linear (left) and semi logarithmic representations (right).
B. Ortiz-García-Carrasco et al. / Journal of Food Engineering 144 (2015) 138–147 147

Acknowledgments

The authors wish to thank the Consejo Nacional de Ciencia y


Tecnología (CONACYT) for providing financial support through
project 130011. Betzabeth Ortiz-García-Carrasco and Estefanía
Yañez-Mota acknowledge their undergraduate scholarship from
CONACYT. Francisco Manuel Pacheco-Aguirre acknowledges his
doctoral scholarship from CONACYT.

References

Azzouz, S., Guizani, A., Jomaa, W., Belghith, G., 2002. Moisture diffusivity and drying
kinetic equation of convective drying of grapes. J. Food Eng. 55 (4), 323–330.
Campos-Mendiola, R., Hernández-Sánchez, H., Chanona-Pérez, J.J., Alamilla-Beltrán,
L., Jiménez-Aparicio, A., Fito, P., Gutiérrez-López, G.F., 2007. Non-isotropic
shrinkage and interfaces during convective drying of potato slabs within the
frame of the systematic approach to food engineering systems (SAFES)
methodology. J. Food Eng. 83 (2), 285–292.
Curcio, S., Aversa, M., 2014. Influence of shrinkage on convective drying of fresh
vegetables: a theoretical model. J. Food Eng. 123, 36–49.
Fig. 14. Water diffusivities as a function of moisture content during drying of da Silva, W.P., Hamawand, I., e Silva, C.M.D.P.S., 2014. A liquid diffusion model to
potato strips. describe drying of whole bananas using boundary fitted coordinates. J. Food
Eng. 137, 32–38.
Dissa, A.O., Desmorieux, H., Bathiebo, J., Koulidiati, J., 2008. Convective drying
characteristics of Amelie mango (Mangifera Indica L. cv. ‘Amelie’) with
Table 2 correction for shrinkage. J. Food Eng. 88 (4), 429–437.
Regression parameters for water diffusivity modela. García-Pérez, J.V., Ozuna, C., Ortuño, C., Cárcel, J.A., Mulet, A., 2012. Modeling
ultrasonically assisted convective drying of eggplant. Drying Technol. 29 (13),
Parameter Value 95% CI 1499–1509.
10 2
d0  10 (m /s) 1.2333 2.0408/0.0426 Hassini, L., Azzouz, S., Peczalski, R., Belghith, A., 2007. Estimation of potato moisture
d1  1011 (m2/s°C) 1.0246 0.8903/1.1590 diffusivity from convective drying kinetics with correction for shrinkage. J. Food
Eng. 79 (1), 47–56.
k1(dimensionless) 7.8981 4.0639/11.7323
Khazaei, N.B., Tavakoli, T., Ghassemian, H., Khoshtaghaza, H., Banakar, A., 2013.
n1(dimensionless) 0.6046 0.4932/0.7161
Applied machine vision and artificial neural network for modeling and
k2(dimensionless) 2.9294 2.4228/3.4361
controlling of the grape drying process. Comput. Electron. Agric. 98, 205–213.
n2(dimensionless) 7.1208 5.9278/8.3138 Lozano, J.E., Rotstein, E., Urbicain, M.J., 1983. Shrinkage, porosity and bulk density of
foodstuffs at changing moisture contents. J. Food Sci. 48 (5), 1497–1502.
a
D ¼ ðd0 þ d1 TÞ½1  expðk1 Wn1 Þ expðk2 Wn2 Þ.
Pacheco-Aguirre, F.M., Ladrón-González, A., Ruiz-Espinosa, H., García-Alvarado,
M.A., Ruiz-López, I.I., 2014. A method to estimate anisotropic diffusion
coefficients for cylindrical solids: application to the drying of carrot. J. Food
Mean water diffusivities were estimated as 3.04  1010, Eng. 125, 24–33.
Panyawong, S., Devahastin, S., 2007. Determination of deformation of a food
3.61  1010, 4.57  1010 and 5.36  1010 m2/s for drying product undergoing different drying methods and conditions via evolution of a
processes at 50, 60, 70 and 80 °C, respectively. Moisture diffusivi- shape factor. J. Food Eng. 78 (1), 151–161.
ties are comparable to those reported by other authors where Pinto, L.A.A., Tobinaga, S., 2006. Diffusive model with shrinkage in the thin-layer
drying of fish muscles. Drying Technol. 24 (4), 509–516.
values have been also corrected for product shrinkage. For potato, Ponkham, K., Meeso, N., Soponronnarit, S., Siriamornpun, S., 2012. Modeling of
Hassini et al. (2007) reported diffusivity values ranging from combined far-infrared radiation and air drying of a ring shaped-pineapple with/
3.55  1010 to 1.92  109 m2/s (40–85 °C). For other vegetable without shrinkage. Food Bioprod. Process. 90 (2), 155–164.
Press, W.H., Teukolsky, S.A., Vetterling, W.T., Flannery, B.P., 2007. Numerical
products, such as carrot, chayote and mango, values were between Recipes. The Art of Scientific Computing. Cambridge University Press, New York,
2.58  1010–1.72  109 m2/s (60–90 °C), 4.44  1010–8.60  USA, Third edition.
1010 m2/s (40–70 °C) and 2.61  1010–1.30  109 m2/s (40– Ratti, C., 1994. Shrinkage during drying of foodstuffs. J. Food Eng. 23 (1), 91–105.
Ruiz-López, I.I., Ruiz-Espinosa, H., Arellanes-Lozada, P., Bárcenas-Pozos, M.E.,
70 °C), respectively (Dissa et al., 2008; Zielinska and Markowski,
García-Alvarado, M.A., 2012. Analytical model for variable moisture diffusivity
2010; Ruiz-López et al., 2012). As shown in Fig. 13, identified diffu- estimation and drying simulation of shrinkable food products. J. Food Eng. 108
sivities jointly with shrinkage data allowed for an excellent repro- (3), 427–435.
duction of drying curves at all moisture contents (R2 > 0.99). Sampson, D.J., Chang, Y.K., Rupasinghe, H.P.V., Zaman, Q.U.Z., 2014. A dual-view
computer-vision system for volume and image texture analysis in multiple
apple slices drying. J. Food Eng. 127, 49–57.
4. Conclusions Singh, B., Gupta, A.K., 2007. Mass transfer kinetics and determination of effective
diffusivity during convective drying of pre-osmosed carrot cubes. J. Food Eng.
79 (2), 459–470.
The proposed image analysis methodology allowed estimating Souraki, B.A., Mowla, A., 2008. Axial and radial moisture diffusivity in cylindrical
representative shrinkage-deformation characteristics of dried fresh green beans in a fluidized bed dryer with energy carrier: modeling with
and without shrinkage. J. Food Eng. 88 (1), 9–19.
potato strips along dominant mass transfer directions. Shrinkage-
Wang, N., Brennan, J.G., 1995. Changes in structure, density and porosity of potato
deformation characteristics determined by current protocols were during dehydration. J. Food Eng. 24 (1), 61–76.
successfully applied in the estimation of water diffusivities and Yadollahinia, A., Jahangiri, M., 2009. Shrinkage of potato slice during drying. J. Food
Eng. 94 (1), 52–58.
drying simulation and results were comparable with other studies
Yadollahinia, A., Latifi, A., Mahdavi, R., 2009. New method for determination of
where different methodologies have been used. A previously potato slice shrinkage during drying. Comput. Electron. Agric. 65 (2), 268–274.
reported one-dimensional drying model for shrinkable solids was Yan, Z., Sousa-Gallagher, M.J., Oliveira, F.A.R., 2008. Shrinkage and porosity of
extended to consider two- or three-dimensional mass transfer, banana, pineapple and mango slices during air-drying. J. Food Eng. 84 (3), 430–
440.
with a remarkable reproduction of drying curves. It was demon- Zielinska, M., Markowski, M., 2010. Air drying characteristics and moisture
strated that individual shrinkage-deformation characteristics of diffusivity of carrots. Chem. Eng. Process. 49 (2), 212–218.
dried products could be combined in a single representative profile
which could be used in detailed simulations, as will be presented
in a following document currently in preparation.

You might also like