Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

0960–3085/04/$30.00+0.

00
# 2004 Institution of Chemical Engineers
www.ingentaselect.com=titles=09603085.htm Trans IChemE, Part C, March 2004
Food and Bioproducts Processing, 82(C1): 60–72

MODELLING DIFFUSION OF MOISTURE DURING


STOVING OF STARCH-MOLDED CONFECTIONS
M. B. SUDHARSAN1, G. R. ZIEGLER2,* and J. L. DUDA3
1
Nestlé Product Technology Centre, Orbe, Switzerland
2
Department of Food Science, Penn State University, University Park, USA
3
Department of Chemical Engineering, Penn State University, University Park, USA

S
tarch molding is a widely used shape forming technique in the confectionery industry.
Candy is deposited into cavities imprinted in dry granular starch contained in a shallow
tray. For products that cannot be deposited at their final moisture content, a drying or
stoving stage follows deposition, and largely determines the productivity of the process. In the
present work a finite difference model was developed that quantitatively predicted the moisture
diffusion from a single candy piece into the starch bed or air at a given temperature (50 C).
The model suggested that a ‘skin’ or region of of low moisture content (3–4% wet basis) and
diffusivity (1013 m2 s1) was formed near the surface of the candy at normal operating
conditions of the drying room that then slowed the drying process, and that the spacing
between the candies does not affect the drying rate at typical print densities. The model
adequately predicted experimental results from model systems and in production. Controlling
the relative humidity of the drying air, the initial moisture content in the starch bed, and the
temperature of the drying chamber may avoid skin formation, improve drying rate and
minimize the drying energy required.

Keywords: confectionery; diffusion; drying; mathematical modelling; starch-molding; stoving.

INTRODUCTION Once the candies are sieved out, the starch is dried and
recycled back to the mogul for reuse. The process of
Confectionery items such as jellybean centers, candy corn, recycling the molding starch is called starch conditioning,
marshmallows, starch gums, caramels, nougat, gummies and which is also energy-intensive. The typical amount of
cream centers may be manufactured using starch molding energy needed to reduce the moisture content of one
techniques. Hot candy liquor is deposited into cavities kilogram of starch by 1% is 64 kJ (27.5 BTU lb1; Collins,
imprinted in dry granular starch contained in a shallow 1989). The required ratio of molding starch per unit mass of
tray, and the finished candy piece is sieved from the molding candy deposited is a key factor in minimizing the starch
starch after setting. A variety of shapes can be produced on conditioning energy. Tray size is standardized to facilitate
the same equipment by simply changing the print board. mechanical handling by the mogul, but it is uncertain
Because of its versatility, starch molding is one of the most whether tray depth has ever been optimized in relation to
widely used shape-forming techniques for non-chocolate drying efficiency.
confectionery and confectionery centers. The relative rates of drying through the surface exposed
For products that cannot be deposited at their final moisture to air vis-à-vis that in contact with the molding starch has
content, a drying or stoving stage follows deposition. The trays not been clearly explained in the literature. Brock (1950)
are held in a drying atmosphere at 50–60 C, and 10–40% determined that the final moisture content in trays of
relative humidity until the candies reach the desired moisture molding starch was established by the relative humidity
content. The drying time varies from 24 to 72 h depending on (RH) of the drying air, but that the drying rate was
the type of candy, and is the rate-limiting step in the molding influenced by temperature alone. Grover (1947) studied
process. Drying is an energy-intensive process, and manufac- moisture permeation rates through confectionery jelly and
turers are interested in optimization to achieve maximum concluded that the rate of drying was independent of RH
production capacity with minimal energy consumption. below 60%. While drying temperatures are common in the
literature, much less attention is given to other environ-
mental conditions such as RH or air velocity, suggesting
*Correspondence to: Professor G.R. Ziegler, 116A Borland Lab, Department
atmospheric conditions have little effect on the process
of Food Science, Penn State University, University Park, PA 16802, USA. (Kintz, 1996). Motion of the air has no effect except as it
E-mail: grz1@psu.edu influences the RH by carrying moisture away from the

60
DIFFUSION IN STARCH MOLDING 61

The average mass transfer coefficient, Km, was calculated


using the correlation between Sherwood number (Kml=Dg),
Reynolds number (lrV=m) and Schmidt number (m=rDg) for
an external laminar flow along a flat plate given by equation
(2) (Cussler, 1996):

 1=2 !1=3
Km l rlV m
¼ 0:323 (2)
Dg m rDg

The velocity of air in the drying room is approximately


0.05 m s1 (Kintz, 1996), and this value has been used for
further calculations. The length (l) of candy pieces was
Figure 1. Transport mechanism of moisture during stoving of starch jellies. taken to be 5 cm and thickness (L) to be 0.5 cm. The
diffusion coefficient of moisture in candies is not well
reported. Yamamoto et al. (1997) has reported values of
the diffusivity of moisture in sucrose at 30 C, a major
starch and out of the conditioned area (Brock, 1950). These
ingredient in the candy, of the order of 1011–
observations lead to the conclusion that for most starch-
1014 m2 s1. Vagenas and Karathanos (1992) have reported
molded products drying is diffusion controlled.
diffusivities in gelatinized starch, another ingredient in
It is common to find a dry ‘skin’ on the surface of
candy, in the order of 1010–1011 m2 s1. Vagenas and
products manufactured using starch molding, formed from
Karathanos (1991) have estimated the effective diffusivity in
a region of low moisture content near the surface. This skin
granular Hylon 7 to be between 1–2  109 m2 s1 in the
is beneficial to products with a soft center since it prevents
temperature range 40–60 C at a porosity of 0.5. A conser-
pieces from sticking together and resists mechanical stresses
vative value of 5  109 m2 s1 was taken as the diffusion
during handling. Seager (1991) intuitively explained the
coefficient of moisture for the Biot number analysis, giving
formation of this moisture gradient by stating that the
a Bi value of the order 103, suggesting that internal diffusion
moisture transfer inside the candy cannot replace loss of
in the candy and starch phases would probably be the rate
moisture on the surface. The presence of a skin could
limiting mechanism for moisture transfer.
severely limit the drying rate since moisture diffusivity is
The variety of shapes of real candy pieces is complex,
concentration dependent. Initially, moisture must move
making modelling difficult. To circumvent this problem, the
rapidly from the outer surface of the product to encourage
shape of the candy pieces have been taken to be a hemi-
‘skinning’, which aids in shape definition (Best, 1990).
sphere (Figure 2), not an unreasonable shape. The shape of
However, if too rapid, moisture removal at the surface can
the starch bed has been taken to be a hollow hemisphere,
result in ‘case hardening’ that can actually slow moisture
with the candy residing in the center (Figure 2). All the
transfer overall. Moisture trapped in this way may cause
assumptions involved in developing the model have been
‘sweating’ during storage. Ziegler et al. (2003) observed
listed below:
skin formation in a model starch molding system using
magnetic resonance imaging, and determined that drying (i) no change in densities of the candy or the starch bed;
was effectively diffusion-controlled after the first 30 (ii) no shrinkage or expansion occurs in the candy or
minutes. Ziegler et al. also concluded that drying via the starch bed during the drying process;
candy–molding starch interface was more effective than that (iii) uniform moisture content in the starch bed and
through the candy–air interface.
candy at the beginning of the process;
The possible transport mechanism for moisture migration
(iv) heat transfer in the system is much faster when
during the actual drying process has been outlined in
Figure 1 (Ziegler and Sudharsan, 2000). The objective of compared to mass transfer and the system is consid-
the present work was to develop a model that would ered isothermal;
quantitatively predict the moisture diffusion from a single (v) diffusion coefficient in the candy was a function of
candy piece into the starch bed or air at a given temperature moisture content in the candy (Yamamoto et al.,
(50 C). 1997; Vagenas and Karathanos, 1992);
(vi) diffusion coefficient in the starch bed was a constant
MODEL FOR MOISTURE MIGRATION within the typical operating moisture range (between
DURING STOVING 6 and 12%);
(vii) the diffusion process in the system is assumed to be
The rate-limiting mechanism for moisture transfer Fickian;
(convection or internal diffusion limited mass transfer)
(viii) the air–candy and air–starch surfaces immediately
was determined by evaluation of the Biot number (Bi) at
the candy–air and starch–air interface. In the case of a flat reach equilibrium determined by the external relative
plate, Bi is given by equation (1). humidity;
(ix) the candy–starch interface immediately came
Km L to its equilibrium value as determined by the
Bi ¼ (1)
D isotherms.

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72
62 SUDHARSAN et al.

Figure 2. Finite difference grid used in the starch–candy system.

Moisture Diffusion in the Candy Phase about the center axis. Thus the ‘j’ term in equation (3) is
neglected giving equation (4):
The unsteady-state diffusion equation in spherical coor-
dinates, neglecting the convective terms is presented as
equation (3) (Bird et al., 1994):  
@X 1 @ 2 @X 1 @
  ¼ 2 r D(X ) þ
@X 1 @ 2 @X 1 @ @t r @r @r r sin (y) @y
¼ 2 r D(X ) þ  
@t r @r @r r sin (y) @y D(X ) sin (y) @X
   (4)
D(X ) sin (y) @X 1 @ r @y
 þ
r @y r sin (y) @f t¼0 X ¼ Xo 0  r  Rcandy y ¼ all (5)
 
D(X ) @X t ¼ all X ¼ Xequ 0  r  Rcandy y¼0 (6)
 0  r  Rcandy , t  0 (3)
r sin (y) @f
t ¼ all X ¼ Xequ r ¼ 0 y ¼ all (7)
Since the moisture content in the candy and starch bed is @X p
t ¼ all ¼ 0 0  r  Rcandy y ¼ (8)
uniform at the beginning of the process, there is symmetry @y 2

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72
DIFFUSION IN STARCH MOLDING 63

The fourth boundary condition is the flux condition at the the Taylor series was used to account for the steep gradients
starch-candy interface. This is dealt with separately. Figure 2 that might arise at the interface.
summarizes the boundary conditions. Differencing equations (4) and (9) resulted in tri-diagonal
matrixes with fringes (Sudharsan, 2001). In the case of the
candy phase the set of simultaneous equations obtained
Moisture Diffusion in the Starch Phase were non-linear and in the starch phase equations were
The process of moisture migration in the starch bed was linear. The Gauss–Seidel iterative technique was used to
by porous diffusion with adsorption of moisture on the solve the matrix (Press et al., 1992). The generalized
surface of the starch particles. Previous work on the diffu- algorithm to solve the starch–candy system completely is
sion of moisture in granular starch suggested that the explained in Figure 3. The concentration values at the
effective diffusivity at 50 C might be constant (Vagenas starch–candy interface were always solved before going
and Karathanos, 1991) in the operating moisture range of 6– onto the next time step and were checked for change,
12% (wet basis). The unsteady-state diffusion equation in because the flux condition solved initially was solved
spherical coordinates neglecting the convective term and the using the values at the previous time step and was not
‘j’ term is presented as equation (9) (Bird et al., 1994): truly the implicit solution at the interface.
  A 50  50 grid (50 nodes radial direction and 50 nodes in
@Y 1 @ 2 @Y 1 @ the ‘y’ direction) was overlaid on the candy and starch
¼ 2 r Ds þ
@t r @r @r r sin (y) @y
 
Ds sin (y) @Y
 (9)
r @y
t¼0 Y ¼ Yo Rcandy  r  Rstarch y ¼ all (10)
t ¼ all Y ¼ Yequ Rcandy  r  Rstarch y¼0 (11)
@Y p
t ¼ all ¼ 0 Rcandy  r  Rstarch y¼ (12)
@y 2
@Y
t ¼ all ¼ 0 r ¼ Rstarch y ¼ all (13)
@r
The fourth boundary condition is the flux condition at the
starch–candy interface and is explained below.
At the starch–candy interface, any moisture that leaves
the candy enters the starch bed, and the water activity of the
starch is equal to the water activity of the candy at this
interface.
astarch
w (Y jt¼t
r¼Rcandy ,y¼y ) ¼ aw
candy
(X jt¼t
r¼Rcandy ,y¼y ) (14)
t¼t t¼t
@Y  @X 
rstarch Ds  ¼ rcandy D(X ) 
@r r¼Rcandy ,y¼y @r r¼Rcandy ,y¼y
(15)

Solving the System of Diffusion Equations


To predict the moisture profiles in the starch-candy
system, it was necessary to solve unsteady-state diffusion
equations in the candy phase [equation (4)] and the starch
phase [equation (9)] simultaneously with the appropriate
boundary conditions. The Crank–Nicholson finite difference
scheme was used to solve the diffusion equations implicitly
with the grid shown in Figure 2.
To solve for the concentration profile at the starch-candy
interface, the flux condition [equation (15)] was solved
simultaneously with the water activity relation [equation
(14)] at the interface (Sudharsan, 2001). From equation (14)
a relation between the equilibrium moisture content in the
candy and the equilibrium moisture content in the starch
was derived. This relation was then back substituted into
equation (15) (Sudharsan, 2001). This converted equation
(15) into an equation with one variable, which was solved
using the bisection method (Press et al., 1992). While finite Figure 3. Generalized algorithm to solve the starch–candy system
differencing equation (15) a second-order approximation of implicitly.

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72
64 SUDHARSAN et al.

phases respectively, and the solution was obtained for 0.1 s values of relative humidity of various saturated salt solutions
time steps. The choice of time step as 0.1 s seemed appro- for a broad range of temperature. The starch samples took 2–4
priate as the implicit solution obtained using Dt ¼ 0.01 s was days to equilibrate at a given relative humidity. The moisture
not significantly different, while that obtained using Dt ¼ 1 s content in the samples was then determined using nuclear
was different. A 100  100 grid in the candy and starch magnetic resonance (NMR) (Minispec-MQ20, Bruker,
phases respectively took approximately 2 weeks to solve for Canada) as described below.
30 h of drying time on a 700 MHz desktop computer. The
implicit solution obtained using a 100  100 grid was not
significantly different from that obtained using a 50  50 Determination of Moisture Content
grid. in Starch using NMR
The mass average concentration of the candy was then The free induction decay signal intensity per unit mass of
obtained by integrating the concentration profile radially and starch was calibrated against moisture content determined
later in the ‘y’ direction according to equation (16). by potentiometric Karl Fischer titration (Koehler Instrument
H Company Inc., Bohemia, NY, USA) using methanol (Karl
V X dv
Xavg ¼ (16) Fischer grade, anhydrous, Fischer Scientific, Pittsburgh, PA,
(2=3)pR3candy USA) as the solvent to extract moisture from the starch
To integrate, Simpson’s 3 rule was used (Press et al., 1992). (Troutman, 1999). The GAB equation [equation (17)] was
Press et al. have commented that using a one-dimensional used to fit the data:
integral technique to evaluate a multidimensional integral Wm CKaw
would yield a solution of low accuracy. As a first step to M¼ (17)
evaluate the integral the above technique was used. (1  Kaw )(1  Kaw þ CKaw )
The integral solution obtained was checked by solving for where C and K are fitted parameters.
a simple case, the volume of a hemisphere (23 pR3candy ). The The GAB equation (Van den Berg, 1984) gave a good fit
predicted volume of the hemisphere was 8.9% less than the (r2 value ¼ 0.9892, average error ¼ 1.4%) to the data
true value while using 50  50 grid (50 nodes in radial (Figure 4). The values of the GAB parameters obtained
direction and 50 nodes in ‘y’ direction) and was 4.9% less using ‘Solver’ (Microsoft Excel, Microsoft Corp., Redman,
than the true value while using a 100  100 grid and was WA, USA) were Wm ¼ 0.087, K ¼ 0.574 , C ¼ 31.8.
1.59% less than the actual while using a 300  300 grid.
Reducing the grid size further improved the integral solution
but it was impractical to solve the system of diffusion Desorption Isotherm of Candy at 50 C
equation for such a refined grid. As a first approximation,
Commercial gummy bears were frozen and cut into small
the 50  50 grid was chosen and a 9% numerical error in
chips. Typical formulae used in manufacturing gummy bears
the integral solution was expected. This then was the source
is 40% sucrose, 41.36% corn syrup 42 D.E., 6.32% gelatin,
of error observed in Figure 11 (i.e., an initial moisture
1.04% citric acid and 11.28% water (Warnecke, 1991). Care
content lower than the given initial conditions). However,
was taken not to contaminate the product; the work area and
this approach still allows us to compare one case to another.
accessories used were cleaned with ethyl alcohol and
A ‘C’ program was written to solve the above unsteady-state
potassium sorbate, an anti-molding agent. The shredded
diffusion equation for a given set of initial and boundary
product (0.25–0.5 g) was placed in aluminum weighing
conditions in the candy phase and evaluate the mass average
dishes (44 mm i.d.  12 mm deep) over pure water in a
concentration. The code was developed on a Linux work-
desiccator. Once the product picked up moisture more
station with 128 MB RAM and an Intel 700 MHz PIII proces-
than twice its weight, it was taken out and equilibrated
sor. The GNU ‘C’ compiler from Free Software Foundation
over saturated salt solutions in Mason jars as described
(Boston, MA, USA) was used to compile the source code.
above. The dishes were weighed periodically after 2 days to
The experimental parameters needed to model the starch–
check if the sample had reached a constant weight. Samples
candy system were the adsorption isotherm of starch,
kept at a water activity of 0.45 and above equilibrated within
desorption isotherm of candy, diffusion coefficient of
14 days. For samples kept at water activities below 0.45, a
water in the starch bed and the diffusion coefficient of
clear sign of equilibration was not observed; a loss of
water in the candy as a function of moisture content.
approximately 103 g (0.2%) was observed for every 2
days. Although the samples did not equilibrate completely,
MATERIALS AND METHODS the percentage loss of moisture suggested that the samples
were near equilibrium. Methanol was used as a solvent for
Adsorption Isotherm of Starch at 50 C
moisture extraction. The solvent was titrated against Karl
Molding starch (unmodified cornstarch) was obtained from Fischer reagent and the amount of moisture in the product
National Starch and Chemical Company (Bridgewater, NJ, was back calculated using the dilution factor and amount of
USA). Cornstarch (2–4 g) was vacuum dried and placed in moisture present in the solvent (Troutman, 1999). The GAB
aluminum weighing dishes (44 mm i.d.  12 mm deep) on equation gave a good fit to these data (r2 value ¼ 0.9942,
wire mesh supports inside sealed Mason canning jars (473 ml). average error ¼ 5.35%). Values of the GAB equation para-
The jars were filled to a depth of about 1 cm with saturated salt meters obtained were Wm ¼ 0.041, K ¼ 1.13, C ¼ 31.3. The
solution, and kept in an environmental chamber (Environ cab, isotherm obtained (Figure 5) is comparable to the J-type
Lab-Line Inc., IL, USA) at 50  1 C. Different salt solutions isotherm obtained by Troutman (1999) for starch jelly
were chosen so that the samples were equilibrated over a broad centers at 20 C, and to the isotherm reported by Van den
range of relative humidity. Greenspan (1977) has tabulated Berg and Bruin (1981) for crystalline sucrose at 40 C.

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72
DIFFUSION IN STARCH MOLDING 65

Figure 4. Adsorption isotherm for cornstarch at 50 C.

Estimation of Diffusion Coefficient of


Moisture in the Starch Bed at 50 C
A tray filled with saturated salt solution (sodium chloride,
RH 74.4%), a small fan and a balance (Mettler PE-360) Figure 6. Experimental setup to determine sorption curves for starch and candy.
were placed inside a Plexiglas box (0.4  0.2  0.2 m) as
shown in Figure 6. The whole box was kept in an environ-
mental chamber at 50 C. A circular Plexiglas dish with an The experimental drying curve was fitted to the one-
inner depth of 1 cm and an inner diameter of 7.68 cm dimensional, unsteady-state diffusion equation [equation
(weight ¼ 67.09 g) was custom made, filled with vacuum (18)] with constant diffusivity and a flux condition at the
oven-dried starch, and placed on the balance. The Mettler starch–air interface:
PE-360 type balance was chosen since it was sensitive @Y @2 Y
enough (0.001 g) to record the change in weight of the ¼ Ds 2 (18)
starch yet not sensitive to other vibrations in the environ- @t @z
mental chamber. The weight of the drying dish was recorded Y ¼ Yo z ¼ all t¼0 (19)
periodically until equilibrium was attained, after approxi-
mately 39 h. It was observed that the relative humidity in the @Y
rstarch Ds ¼ Kg [astarch
w (Y jt¼t starch
z¼1 )  aw (Yequ )]
box did not reach its equilibrium value immediately. This @z
suggested that the fan speed was not fast enough and that z ¼ 1 cm t ¼ all (20)
there was a significant film resistance for mass transfer. @Y
From Figure 7 a small bend in the sorption curve can be ¼ 0 z ¼ 0 t ¼ all (21)
seen during the start of the experiment that can be attributed @z
to the film resistance in the air phase. The bulk density of An implicit finite difference scheme was used to solve the
the starch bed was 700.2 kg m3. Cracks and a 1 mm (10%) diffusion equation (Press et al., 1992). A ‘C’ program was
increase in the bed depth were observed in the starch bed written to do the fitting (Sudharsan, 2001). The solution
during adsorption. This increase in the depth was neglected obtained was checked against the solution obtained by
during estimation of the diffusivity. Crank (1975) for a linear isotherm. One-hundred nodes

Figure 5. Desorption isotherm for candy at 50 C. Figure 7. Experimental and fitted sorption curves for starch at 50 C.

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72
66 SUDHARSAN et al.

were used to discretize the length of the starch bed with the drying dish. The drying dish was placed on the balance
Dt ¼ 0.1 s. To solve for the flux condition at the starch–air and the weight of the dish was noted periodically until a
interface, the relationship between the equilibrium RH and constant weight was reached.
the moisture content of starch was needed. This was The candy drying experiment was stopped after 4 days
obtained from the GAB isotherm of molding starch. The even though complete equilibration of the sample was not
bisection method was used to solve the flux condition and obtained. The initial thickness of the sample in the drying
the technique was similar to that used in solving the flux dish was 1 mm. At the completion of the experiment, the
condition at the candy–starch interface. Once the concentra- thickness of sample at the center of the dish was observed to
tion profile in the slab was estimated as a function of time, be 0.5 mm and at the sides it was 0.95 mm, a uniform gradient
the profile was integrated using Bode’s five-point rule (Press was not observed. To account for this shrinkage, the average
et al., 1992) to estimate the mass of the sample at a given thickness of the sample was taken to be 0.75 mm. Also, the
time t. The dimensionless mass, M*, was then calculated bend seen in the drying curve of starch was less apparent in
using equation (22). the drying curve of candy (Figure 8) because the RH of the
Mt  Mo ambient air was between 30 and 40% at the time of closing the
M ¼ (22) Plexiglas box and the salt solution did not take long to bring
Mequ  Mo the relative humidity to 30.5%.
There was good fit between the experimental and The unsteady-state diffusion equation with variable diffu-
predicted adsorption curves (Figure 7; r2 ¼ 0.998, average sivity [equation (25)] was solved using implicit finite
error ¼ 0.66%). The estimated effective diffusivity in the differencing technique (Press et al., 1992). The exponential
starch bed was 4.1  109 m2 s1 and the mass transfer function in equation (26) was used to describe the concen-
coefficient in the system, Km, was 4.5  104 kg m2 s1. tration dependence of the diffusion coefficient of moisture in
Vagenas and Karathanos (1991) have estimated the effective the candy. A ‘C’ code was developed to solve equation (25).
diffusivity in granular Hylon 7 to be between 1  109 and One-hundred nodes were used to discretize the radius of the
2  109 m2 s1 in the temperature range 40–60 C at a candy with Dt ¼ 0.1 s. Once the concentration profile in the
porosity of 0.5. slab was estimated as a function of time, the profile was
The apparent diffusivity of moisture in the porous starch integrated using Bode’s five-point rule (Press et al., 1992) to
bed obtained using the experiment above was checked using estimate the mass of the candy at a given time t. The values
the procedure outlined below. Let the adsorption isotherm of of Do(1.7  1010 m2 s1) and Ao (0.23) were optimized by
starch be linear and given by equation (23). trial and error to give the best fit for the experimental drying
curve (Figure 8) (r2 ¼ 0.9973, average error ¼ 1.35%).
Cstarch ¼ ACair (23)  
@X @ @X
The apparent diffusivity of the system could be obtained ¼ D(X ) (25)
from the equation (24) (Weisz, 1995). Weisz (1995) has also @t @z @z
 
commented that in the case of non-linear isotherms the value A
of diffusivity derived might not change by more than a D(X ) ¼ Do exp  o (26)
X
factor of 1.6:
X ¼ Xo z ¼ all t ¼ 0 (27)
Dapparent ¼ Dg =A (24)
X ¼ Xequ z ¼ 1 mm t ¼ all (28)
In the present case the diffusivity of water vapor at 50 C @X
in air (Dg ¼ 2.47  105 m2 s1) was estimated (Wilke and ¼ 0 z ¼ 0 t ¼ all (29)
@z
Lee, 1955) and it was divided by the slope (A ¼ 2615) of the
starch isotherm at Y ¼ 6% (typical moisture content at the The experiment was repeated using a salt solution of
start of the drying process). The value of Dapparent thus sodium bromide of aw ¼ 0.51 (Greenspan, 1977). This was
estimated was 9.5  109 m2 s1. The estimated value does done to check the ability of the diffusion model proposed
not account for the tortuosity, usually between 1.5 and 2
(Weisz, 1995), or the porosity of the system (e 0.48), both
of which would decrease the estimated diffusivity value
even further, but not by orders of magnitude.

Diffusion Coefficient as Function of Moisture


Content in the Candy at 50 C
The experimental setup to obtain the drying curve of the
candy is similar to that of the starch (Figure 6), except that
magnesium chloride (aw ¼ 0.305) was used as the salt
solution and the depth of the drying dish used was 1 mm.
The inner radius of the custom made drying dish was
7.67 cm (weight ¼ 97.15 g). Frozen starch jelly slurry was
obtained from Harmony Foods Inc. (Santa Cruz, CA, USA).
The composition of the starch jelly was comparable to the
example formulation given by Warnecke (1991). The slurry
was reheated so that it was fluid enough to be deposited in Figure 8. Experimental and fitted sorption curves for candy at 50 C.

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72
DIFFUSION IN STARCH MOLDING 67

Figure 10. Diffusion coefficient for moisture as a function of water activity


Figure 9. Diffusion coefficient for moisture as a function of moisture for candy at 50 C.
content of candy at 50 C.

Case 1: Typical Operating Conditions


[equation (26)] to interpolate. The experimental drying This case is representative of the typical commercial
curve was compared with the predicted drying curve operating conditions at which drying is done. The initial
(Do ¼ 1.7  1010 m2 s1 and Ao ¼ 0.23). There was good moisture content in the candy was 25% but the predicted
agreement between the predicted and the experimental curve mass average concentration was 22.1% (Figure 11), this error
(r2 ¼ 0.9829, average error ¼ 5.85%). was due to the numerical error while evaluating the multi-
The diffusivity of moisture in starch jellies, as a function dimensional integral as explained earlier. At time t ¼ 0.1 s
of moisture content is not well reported in literature. (Figure 12) a skin of very low moisture concentration
Yamamoto et al. (1997) have estimated diffusivity in (approximately 3.8% wet basis) was formed around the
sucrose, a bulk ingredient in the candy, as a function of candy. The effective diffusivity at this concentration was
moisture content at 30 C using a power law model (Figure 9). 5  1013 m2 s1. The effective diffusivity at the center of the
They have reported effective diffusivity values between candy was in the order of 1011 m2 s1. The diffusivity in the
1015 and 1011 m2 s1 in a moisture content range of starch bed was in the order of 109 m2 s1, about 100 times
0.02–0.4 (wet basis). These values of the effective diffusivity faster than that in center of the candy. From this it was evident
are significantly lower than the present work (Figure 9). that the moisture in the bulk of the candy was not able to
In addition to the compositional differences, this could be replace the quick loss of moisture at the surface of the candy,
attributed to the lower temperature (30 C) at which they conforming with the intuitive explanation given by Seager
carried out their experimental work. Using the isotherm of (1991). The skin formed not only limited the moisture
the candy a relationship between the water activity and the escaping into the starch bed but also into the air above.
diffusion coefficient of the candy was obtained (Figure 10). There was no significant accumulation of moisture in the
starch bed even after 30 h of drying (Figure 13). This
suggested that the starch played an important role in
absorbing moisture from the candy. This is contrary to the
RESULTS AND ANALYSES
conclusion made by Kintz (1996); she had concluded that
The experimentally obtained adsorption isotherm of the starch played no role in the drying of the candies. At t ¼ 30 h
starch, desorption isotherm of the candy, the effective diffu- it could be seen that the moisture content in the center of the
sion coefficient in the starch bed and the effective diffusion candy was still between 20 and 22% (wet basis) and
coefficient in the candy as function of moisture content, were surrounded by skin of low moisture content. This might
used in the model to solve for the concentration profile in the explain, why the centers of Swedish fish (a low-moisture
starch–candy system. It was observed that the solution was product similar to wine gums) are soft yet the outer surface
most sensitive to the order of magnitude of the diffusion is hard enough to withstand any mechanical stresses. Also,
coefficients, i.e., if the prediction of the diffusion coefficients the candies do not stick together when the skin is very dry.
in the above experiments were off by an order of magnitude, Stickiness is an important quality criterion for manufactures
then the concentration profile obtained would have been very who wish to avoid lumping of candies within a package.
different. The isotherms mostly influenced the equilibrium At the end of 30 h of drying, it could be seen that the skin
values at the interfaces. formed at the starch–candy interface was thicker than that
The effect of the initial moisture content of the starch bed, formed at the starch–air interface. The surface area of the
the relative humidity of the drying air, and the spacing candy exposed to the starch bed was 2pR2candy and that exposed
between the pieces, on the drying process was investigated; to the air is only pR2candy (half that exposed to starch). Therefore
the starch–candy system was solved implicitly for five the amount of moisture removed into the starch bed was much
different cases (Table 1). The mass average concentration more than that removed into the air. This was more evident
of moisture in the candy as a function of drying time for the from a plot of the concentration profile in the candy, as shown
five different cases is shown in Figure 11. in Figure 14 for y ¼ p=2 and time t ¼ 30 h.

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72
68 SUDHARSAN et al.
Table 1. Operating conditions of different drying cases studied.

Initial moisture in Initial moisture in Air relative Diameter of Thickness of the


the starch (wet basis) the candy (wet basis) humidity (%) candy piece (cm) starch bed (cm)
Case 1 6% 25% 10 2 1
Case 2 10% 25% 10 2 1
Case 3 6% 25% 30 2 1
Case 4 6% 25% 10 2 1.5
Case 5 6% 25% 10 2 0.5

Case 2: Higher Initial Moisture in the Starch Bed 30%, with the other operating conditions as in case 1.
The initial moisture content in the starch bed was 6% and
The effect of increasing the initial moisture content of the
the equilibrium moisture content at 30% RH was 8.4%,
starch bed to 10% while keeping other operating conditions
therefore it could be clearly seen that starch absorbs moisture
the same as in case 1 was modeled. It can be observed that
from the external air. A similar situation can be seen in the
the center moisture in the candy had dropped by the same
manufacturing plant when the starch is exposed to uncondi-
3–4% (Figure 15) as observed in case 1 after 30 h of drying,
tioned air (  40% RH). In Figure 16 it can be seen that the
but that the moisture content in the skin was around 10–13%
moisture content at the starch–candy interface rises from
(w.b.) or about four times higher than in Case 1. At a
3.6% in case 1 to 9%. The mass average concentration is
moisture content of 8–13% (w.b.) the candy is very sticky
approximately the same as observed in case 1.
(by observation). The moisture content at the candy–air
interface was still around 3.6% due to the low relative
humidity of the air (10%). From Figure 11 it can be seen Case 4: Greater Spacing Between the Candies
that the mass average concentration of the candy was much
higher for case 2 than case 1, which could be attributed to The effect of increasing the space between candy pieces on
the lower concentration gradient, i.e. driving force, at the the drying process was studied by increasing the depth of the
starch–candy interface, again pointing to the importance of starch bed to 1.5 cm. The mass average concentration profile
moisture transport to the molding starch. Here a quasi- obtained was very similar to profiles in case 1 (Figure 11). It
steady state is reached in the starch bed. If the starch used was predicted that increasing the depth of the starch bed by
is not conditioned properly, it is possible that the initial 0.5 cm would not therefore improve the drying rate.
moisture content in the starch bed is around 10%. In such
cases it is common to see sticky products. Case 5: Lesser Spacing Between the Candies
The effect of decreasing the space between candy pieces
on the drying process was modeled by decreasing the depth
Case 3: Higher Relative Humidity of Air
of the starch bed to 0.5 cm. The mass average concentration
This case was chose to study the effect of the RH of profile obtained was similar to the profiles in cases 1 and 4,
the drying air on the process. The RH of air was taken to be suggesting that decreasing or increasing the spacing, at least

Figure 11. Mass average concentration of moisture in candy as a function of drying time.

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72
DIFFUSION IN STARCH MOLDING 69

in the range 1–3 cm edge-to-edge, might not affect the interface. This skin limited the moisture migration from
drying during typical drying conditions, perhaps because the candy into the starch bed. We hypothesized that moisture
of the larger diffusivity in starch (at least 100) when diffusivity could be improved by initiating the drying
compared with the diffusivity in candy. This does not process with higher moisture content in the starch bed, but
mean that spacing between the candies could be reduced perhaps not as high as case 2, limiting skin formation at the
completely, because making a good impression in the starch expense of a lower moisture gradient at the starch–candy
requires certain structural integrity and continuity with the interface. For example, let equation (30) determine the flux
air phase must be maintained. at the starch–candy surface.
@X
Moisture flux ¼ rcandy D(X ) (30)
DISCUSSION AND CONCLUSIONS @r
Stoving of starch jellies is necessary to remove excess Now by increasing the initial moisture content in the starch,
moisture from the candy. During stoving at typical operating the concentration gradient (@X =@r) decreases but the diffu-
conditions the moisture migration in the starch–candy–air sivity (D) improves. Therefore for some moisture content
system was via three different routes. the flux is maximized. This is an excellent optimization
problem, and has been left for future work. It might also be
(i) direct migration of moisture from the candy into the possible that the typical operating conditions used by
starch bed; manufactures are already optimal.
(ii) direct migration of moisture from the candy into
the air;
(iii) migration of the moisture from the starch bed into Thickness of the Starch Bed
the air, depending on the relative humidity of air. The thickness of the starch bed did not affect the drying at
At typical operating conditions, starch plays an important typical operating conditions. This does not mean that the
role in taking up moisture from the candy and giving it up to thickness could be reduced indefinitely since getting a good
the air. The effect of intial moisture content in the candy and shape definition on the starch might become a key factor.
the thickness of the candy piece in this diffusion-limited Typical tray heights vary from 46 to 56 mm with starch
drying process is more obvious, while the effect of initial depths typically from 23 to 33 mm (Pyrz, 1994). Since the
moisture content in the starch bed, relative humidity and thickness of the starch bed did not affect the drying rate at
thickness of the starch bed are less intuitive and have been typical operating conditions it would be possible to reduce
explained below. The effect of temperature on the drying the depth of the trays. Any reduction in the depth of the tray
process has not been modeled in this study. At higher would reflect as saving in the energy required to condition
temperature the diffusion coefficients in the candy would the starch, i.e. the amount of starch used per kg of candy
increase and may improve drying. In general, the conclu- deposition could be reduced. Reduction in the depth would
sions drawn from these modelling exercises confirm also mean more trays could be stacked in the drying room.
observations made in model systems (Ziegler et al., 2003) Any capacity increase should also be augmented by increas-
and during tests we undertook in a commercial manufactur- ing the amount of drying air used to avoid a higher RH of air
ing plant. in the stoving room.

Relative Humidity of Air


Initial Moisture Content in the Starch Bed
Air is the final sink for all the excess moisture in the
At typical operating conditions, a skin of low moisture candy. The relative humidity of the air indirectly controlled
content and diffusivity was formed at the candy–starch the concentration gradient at the starch–candy interface,
since the starch bed picked up or lost moisture as dictated
by the relative humidity of the air. Starting the drying
process at a higher relative humidity and decreasing this
parameter during the process might help in enhancing
diffusion during the initial stage of the process and
decreasing the humidity during the later stages might
give the mechanical stress needed for handling the candies.
The way in which the relative humidity should be
decreased as a function of time is also an excellent control
problem.

NOMENCLATURE
aw water activity
acandy
w water activity in the candy for given moisture content
astarch
w water activity in the starch bed for given moisture content
A sorption constant
Ao exponential constant
Bi Biot number
Figure 14. Moisture concentration profile in starch–candy system at C adsorption constants relating the energies of interaction
y ¼ p=2 and time t ¼ 30 h (case 1). between the first and further adsorbed molecules at indivi-

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72
70 SUDHARSAN et al.

Figure 12. Moisture concentration profile in starch–candy system at time t ¼ 0.1 s (case 1; candy section displaced upwards to aid in visualization).

Figure 13. Moisture concentration profile in starch–candy system at time t ¼ 30 h (case 1; candy section displaced upwards to aid in visualization).

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72
DIFFUSION IN STARCH MOLDING 71

Figure 15. Moisture concentration profile in starch–candy system at time t ¼ 30 h (case 2; candy section displaced upwards to aid in visualization).

Figure 16. Moisture concentration profile in starch–candy system at time t ¼ 30 h (case 3; candy section displaced upwards to aid in visualization).

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72
72 SUDHARSAN et al.

dual sorption sites Collins, J., 1989, Starch conditioning, Manufact Confect, 69(10): 53–56.
Cstarch concentration of water in starch, kg m3 Crank, J., 1975, The Mathematics of Diffusion (Clarendon Press, Oxford).
Cair concentration of water vapor in air, kg m3 Cussler, E.L., 1996, Diffusion: Mass Transfer in Fluid Systems (Cambridge
D diffusion coefficient, m2 s1 University Press, New York, USA).
Dapparent apparent diffusivity of the system, m2 s1 Greenspan, L., 1977, Humidity fixed points of binary saturated aqueous
Dg diffusion coefficient of moisture in air, m2 s1 solutions, J Res Nat Bur Stand, 81A: 89–96.
Do pre-exponential constant, m2 s1 Grover, D.W., 1947, The stoving of confectionery, B.F.M.I.R.A. Res. Rep.
Ds diffusion coefficient in the starch bed, m2 s1 no. 2 (as quoted in Kintz, 1996).
D(X) diffusion coefficient at a given moisture content X, m2 s1 Kintz, S.R., 1996, Moisture exchange in starch deposited confections. M.S.
K adsorption constants relating the energies of interaction Thesis, University of Georgia.
between the first and further adsorbed molecules at indivi- Press, W.H., Teukolsky, S.A., Vellerling, W.T. and Flannery, B.P., 1992,
dual sorption sites Numerical Recipes in C (Cambridge University Press, New York, USA).
Km mass transfer coefficient, m s1 Pyrz, E.J., 1994, The starch molding process, Manufact Confect, 74(10):
Kg mass transfer coefficient, kg m2 s1 75–80.
l length of plate, m Seager, R.H., 1991, The behavior of moulding starch, Proceedings of 45th
L thickness of the plate, m P.M.C.A. Production Conference, Hershey, PA, April 22–24, pp 154–158.
M moisture content at a given water activity—dry basis, Sudharsan, M.B., 2001, Modelling moisture diffusion during stoving of
kg kg1 starch molded confections, M.S. Thesis, The Pennsylvania State
M* dimensionless mass University.
Mequ mass of sample at equilibrium, kg Troutman, M.Y., 1999, Moisture migration and textural changes during
Mo intial mass of sample, kg manufacture of soft panned confections, M.S. Thesis, The Pennsylvania
Mt mass of sample at time t, kg State University.
r radius at a given position in the starch-candy system, m Vagenas, G.K. and Karathanos, V.T., 1991, Prediction of moisture diffusivity
Rcandy radius of the candy, m in granular materials, with special applications to foods. Biotechnol Prog,
Rstarch outer radius of the starch bed, m 7(5): 419–426.
t time, s Vagenas, G.K. and Karathanos, V.T., 1992, Prediction of the effective
v volume of the hemispherical candy, m3 moisture diffusivity in gelatinized food systems, J Food Eng, 18:
V velocity of air, m s1 159–179.
Wm monolayer value, kg of moisture. kg of dry sample1 Van den Berg, C., 1984, Description of water activity of foods for
X moisture content in the candy—wet basis, kg kg1 engineering purposes by means of the G.A.B. model of sorption, in
Xavg mass average concentration of candy—wet basis, kg kg1 Engineering and Food, McKenna, B.M. (ed) (Elsevier Applied Science,
Xequ equilibrium moisture determined by the candy isotherm— New York, USA), Vol. 1, pp 311–321.
wet basis, kg kg1 Van den Berg, C. and Bruin, S., 1981, Water activity and its estimation in
Xo initial moisture in the candy—wet basis, kg kg1 food systems: theoretical aspects, in Water Activity: Influences on Food
Y moisture content in the starch bed—wet basis, kg kg1 Systems, Rockland, L.B. and Stewart, G.F. (eds) (Academic Press, New
Yo initial moisture content in the starch bed—wet basis, York, USA), pp 1–61.
kg kg1 Warnecke, M., 1991, Gums and jelly products and formulations, in Proceed-
Yequ equilibrium moisture determined by the isotherm—wet ings of 45th P.M.C.A. Production Conference, Hershey, PA, April 22–24,
basis, kg kg1 pp 140–145.
z distance from the bottom of the drying dish, m Weisz, P.B., 1995, Molecular diffusion in microporous materials: formalisms
and mechanisms, Ind Eng Chem, 34: 2692–2699.
Greek symbols Wilke, C.R. and Lee. C.Y., 1955, Estimation of diffusion coefficients for
D difference gases and vapors, Ind Eng Chem, 47: 1253–1257.
y azimuthal angle, radian Yamamoto, Coumans, W.J. and Vlugt, T., 1997, Determining concentration-
j polar angle, radian dependent diffusivity in food materials, in Engineering and Food,
e porosity Jowitt, R. (ed.) (Sheffield Academic Press, Sheffield, USA), Vol. 1,
r density of air, kg m3 pp A164–167.
rstarch bulk density of the starch bed, kg m3 Ziegler, G.R. and Sudharsan, M.B., 2000, Moisture and starch molding, in
rcandy density of the candy, kg m3 Proceedings of 54th P.M.C.A. Production Conference, Hershey, PA,
m viscosity of air, kg m1 s1 May 1–3, pp 52–60.
Ziegler, G.R., MacMillan, B. and Balcom, B.J., 2003, Moisture migration in
starch molding operations as observed by magnetic resonance imaging,
REFERENCES Food Res Int, 36: 331–340.

Best, E.T., 1990, Gums and jellies, in Sugar Confectionery Manufacture,


Jackson, E.B. (ed.) (Van Nostrand Reinhold, New York, USA), Chap 10, ACKNOWLEDGMENTS
pp 190–217.
Bird, R.B., Stewart, W.E. and Lightfoot, E.N., 1994, Transport Phenomena The authors express their gratitude to General Mills Inc. for financial
(Wiley, Singapore). support, and to Harmony Foods Inc. for supplying samples and allowing us
Brock, F.H., 1950, Moisture equilibrium in confectioners’ molding starch, in to verify models in a production setting.
Proceedings of 4th Production Conference, Pennsylvania Manufacturing
Confectioners’ Association in Cooperation with The Lehigh Institute for The manuscript was received 30 June 2003 and accepted for publication
Research, Bethlehem, PA, April 27–28. after revision 30 January 2004.

Trans IChemE, Part C, Food and Bioproducts Processing, 2004, 82(C1): 60–72

You might also like