You are on page 1of 30

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/275596310

Computational model to design plastic multi-layer films for food packaging to


assure a shelf life at the best cost

Article  in  Journal of Plastic Film and Sheeting · December 2013


DOI: 10.1177/8756087913484920

CITATIONS READS

2 686

3 authors:

Maria Del Pilar Noriega Escobar Omar Augusto Estrada


Daabon Group Institute for Training and Research in Plastics and Rubber
65 PUBLICATIONS   124 CITATIONS    16 PUBLICATIONS   29 CITATIONS   

SEE PROFILE SEE PROFILE

Ivan Lopez
Institute for Training and Research in Plastics and Rubber
12 PUBLICATIONS   32 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Simulación y diseño CAE para alimentos View project

All content following this page was uploaded by Maria Del Pilar Noriega Escobar on 01 July 2016.

The user has requested enhancement of the downloaded file.


Original Article
Journal of Plastic Film & Sheeting
2014, Vol 30(1) 48–76
Computational model to ß The Author(s) 2013
Reprints and permissions:
design plastic multi-layer sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/8756087913484920
films for food packaging jpf.sagepub.com

to assure a shelf life at the


best cost

Marı́a del Pilar Noriega, Omar Estrada


and Iván López

Abstract
A computational model to design plastic food packaging is proposed. The model
minimizes the cost of the multi-layer structure satisfying the specific product
requirements, using a heuristic optimization algorithm. The product requirements
are defined by the expected shelf life, the storage conditions, the water sorption
isotherms of foods and the maximum allowable gain or loss of gases (O2, CO2,
N2, etc.) and moisture for the packaged food. In order to assure the food shelf life,
these product requirements should be fulfilled to estimate the maximum perme-
ance values of the plastic package. The computational algorithm automatically
generates different multi-layer film structures that satisfy the product require-
ments. This algorithm combines different polymeric materials taking into account
the barrier properties and cost of each layer, the compatibility between layers, the
maximum number of layers and the minimum and maximum film thickness for
each layer. Temperature and relative humidity corrections for the permeance
calculations are considered.
Permeance calculations of several barrier films are compared with oxygen
transmission rate (OTR) and water vapor transmission rate (WVTR) measure-
ments. The optimization model algorithm is evaluated by means of standard
numerical routines and numerical benchmarking.

ICIPC – Instituto de Capacitación e Investigación del Plástico y del Caucho, Medellı́n, Colombia
Corresponding author:
Omar Estrada, ICIPC – Instituto de Capacitación e Investigación del Plástico y del Caucho, Cra. 49, No.
5 sur 190, Medellı́n, Colombia.
Email: oestrada@icipc.org
del Pilar Noriega et al. 49

Keywords
Permeability, plastic films, oxygen barrier, water vapor barrier, food and beverage
packaging

Introduction
For the food industry, packaging is a key element in defining the food
shelf life. It helps to keep foods safe and provides protection during
manipulation, transport and display. The package is capable of promoting
or limiting gas and vapor exchange between food and the environment.
This gas or vapor is named a permeant. For a specific food, limiting the
presence of some permeants reduces chemical and biological degradation,
microbiological growth and preserves organoleptic food properties (prevents
color, taste, texture and odor changes), with the goal of maximizing shelf
life.1,2
In past decades, plastic packaging use has grown dramatically. The prin-
cipal advantages of plastics in packaging are their flexibility, low density,
processability and in many cases, they are more environmentally friendly
compared with other packaging materials after a complete life cycle analysis.3
Global plastics consumption presents exponential growth, increasing the gap
with steel and paper consumption over the past few decades.2,4
Food shelf life is the maximum time available between its production and
end use or consumption. This time is reduced by any chemical, biological or
physical change in foods. These changes can alter the consumer’s sensory
appreciation of foods, their nutritional value or their safety, potentially
making the food dangerous to the consumer. These changes are largely
promoted by the following factors:1–3,5–9

. Gain or loss of moisture: foods gain or lose humidity from environ-


mental moisture, the initial food moisture before packaging and the
specific sorption isotherm at storage temperature.
. Oxidation reactions promoted by the gain of oxygen: the oxidation
reactions depend on food composition. These reactions generate reac-
tion products that change the color, taste, odor or texture in foods.
The changes depend on the concentration of oxygen inside the package
and the reaction kinetics at storage temperature. In other cases, the gas
exchange is promoted in order to lengthen the shelf life, as in the
design of breathable packages that control the rate of respiration of
the product, as occurs in the packing of vegetables.
. Microorganism growth: the growth of the microorganism population
depends on the initial population, the conditions of moisture and
50 Journal of Plastic Film & Sheeting 30(1)

oxygen in the environment, moisture and food composition, as well as


the storage temperature, which determines kinetics growth.
. Loss or gain of aromas: some foods are known for their aroma, which
must be preserved in the interior of the package. The loss of aroma is
associated with aging and deterioration of the product.
. Loss or gain of other gases: especially important when the products are
packed in modified atmosphere packaging (MAP), these are the prod-
ucts that breathe or have dissolved gases to improve the texture and
palatability properties, as in carbonated beverages.
. Migration of low molecular weight substances from the packaging
material: most of the polymers used to produce packages have addi-
tives, low molecular weight fractions or substances produced by the
polymer degradation. These chemicals can migrate into the food. In
most cases, the migration levels of these substances are low and do not
pose a risk for human health, but they can change the organoleptic
food characteristics. In other cases, they can achieve concentrations
considered harmful during the shelf life.

Proper packaging design helps to ensure food shelf life, and a good design
begins with the complete food characterization. This characterization includes
determination of absorption or desorption moisture isotherms (see Figure 1),
the critical moisture or water activity value that can generate undesirable phy-
sicochemical and microbiological changes, the gain of oxygen that promotes
oxidation kinetics capable of deteriorating foods or accelerating aerobic micro-
bial growth (see Table 1) and the minimum concentration levels of other gases
or aroma molecules that keep the product viable, among others.1,2,10

Figure 1. Typical moisture adsorption isotherm, where mi is the initial moisture, mc is


the critical moisture and me is the equilibrium moisture (2).
del Pilar Noriega et al. 51

Table 1. Allowable oxygen gain for some food and beverages


products.1
Oxygen gain, G
Food and beverage products (mg O2/g of product)

Beer 0.001–0.004
Wine 0.003
Fruit juice 0.02
Soda beverages 0.04
Coffee 0.11
Cheese 0.42
Milk and other products like sauces, 0.015
meat, vegetables and dry nuts

The role of packaging in terms of product preservation is the control of the


entry or exit of permeant substances, established as determinants for shelf life
during the food characterization. The design target is to prevent the per-
meants from reaching the critical levels defined. Depending on the form of
control, there are two types of packaging: barrier packaging and breathable
packaging. Barrier packaging can provide active or passive barriers. Active
barriers capture or release the molecules of interest through the use of sub-
stances or additives, such as oxygen scavenger additives, desiccants, carbon
dioxide generators, deodorizers, ethylene scavengers and generators, among
others. Passive barriers are the result of polymer permeance properties and
the thickness of the layers in the package structure.
The design of passive barriers to ensure the shelf life is a complex process in
which the interactions between product, packaging, packaging conditions and
environmental conditions must be taken into account. For this reason, many
of the works found in the literature focus on the study of specific foods. For
example, Kulchan et al.7 show the effect of the water vapor permeance of the
packaging on the deterioration of cassava-flour-based baked products. They
use the activation energy and the Arrhenius model to predict the behavior of
LDPE water vapor permeance with temperature and the interaction with the
food through its moisture absorption isotherm. Van Bree et al.11 couple per-
meability models of multi-layer films with the respiration rate of fresh vege-
tables, modeled by the Michaelis–Menten equation. This model calculates the
way in which the oxygen concentration in the environment seeps into the
package or the free space. The model uses the Arrhenius equation to calculate
the variation of the permeability coefficients with temperature. Cárdenas
et al.8 performed an experimental study which, among other things, analyzes
the effect of using barrier packaging on microbial population growth in red
meat. This work does not model the interactions between the packaging and
52 Journal of Plastic Film & Sheeting 30(1)

the food. Rodriguez-Aguilera et al.12 integrate the gas permeability rate of the
packaging with the respiration kinetics of the cheese, predicted by the
Michaelis–Menten equation. This model takes into account consumption of
oxygen and generation of carbon dioxide. Genkawa et al.13 present a model
that couples the water vapor permeance of the plastic packaging with the
absorption of humidity into toasted rice, depending on its absorption
isotherm.
Prediction of gas and vapor permeance values is based on the solution to
Fick’s equation and the polymer’s capacity to dissolve the permeant accord-
ing to Henry’s law.
The mathematical model for permeability phenomena is based on the fol-
lowing assumptions:14–17

. Permeants are transferred from the environment to package and vice


versa by convection. Mass transfer convective coefficients are infinite.
In this case, there is no resistance to mass transfer in the internal and
external packaging environment, and the environment is considered a
fluid.
. The volume of fluid inside and outside the package is infinite.
According to this, the permeant concentration in the media and the
velocity of mass transfer at the same temperature are constant.
. There is intimate contact between the fluid media and the package
wall.
. The diffusion coefficients are constants. They are independent of crys-
tallization grade and the macromolecular orientations and only
depend on temperature and relative humidity (RH).
. The phenomenon is quasi-steady state.

The most studied permeants are oxygen and water vapor for their impact
on food shelf life, and both are considered in the computational model
presented.
In the case of carbonated beverages, MAP and breathable packaging,
other gases such as carbon dioxide and nitrogen may be important. For
other products, the permeability of odors can be decisive. In these cases,
there is less information about polymer barrier properties and fewer tech-
niques developed for their measurement. For this reason, these cases are
not covered in the computational model proposed in this article.
The perception of packaging as a complementary element for food prod-
ucts has changed. Today, the packaging is considered as an ingredient in food
formulation and its cost is a very important factor. There are a wide range of
types of polymeric materials with different barrier properties and cost. This
cost is widely variable depending on the availability of the materials and the
del Pilar Noriega et al. 53

international price of oil. The determination of the best structures, capable of


providing the barrier required by food at the lowest cost possible, involves the
assessment of a great number of alternatives and the thickness optimization
for each alternative. In order to satisfy this cost requirement, a suitable
mathematical model that describes the physical phenomena and an efficient
optimization algorithm are needed.16,17
Non-linear multi-variable constrained optimization is deeply explored in
the literature. Several numerical and deterministic methods have been pro-
posed for obtaining accurate and exact solutions. Dennis and Lewis,18 for
example, evaluate different non-linear programming approaches, such as the
black box approach (generalized reduced gradient, GRG), the all-at-once
approach and the domain decomposition method. However, those methods
have shown to have high dependence on the initial guess values, falling easily
in local minimums,19 and, according to the nature of the equations involved,
they can have several convergence problems. That situation has motivated the
development of other alternatives, such as heuristic methods. They do not
necessarily deliver the best solution, but they deliver a good enough solution
to be accepted.
When the objective, or target function, and the equations of the constraints
of the optimization problem present linear behavior, the problem can be
successfully solved by linear programming methods such as the simplex algo-
rithm.20,21 However, the optimization problem proposed in this article, cost
minimization of multi-layer structures, involves non-linear conditions, espe-
cially when the dependence of the permeance on the humidity is considered.
Initially, the optimization problem was solved using the GRG method,22
which delivered accurate solutions as long as the initial guess values were
close to or within the feasible region (the values that satisfy all the con-
straints). Finding an initial guess that satisfies all the constraints is not
always a trivial problem, and it can sometimes be quite difficult, especially
when the optimization algorithm must be run several times.23 This is the case
for the multi-layer optimization problem, which seeks to evaluate several
different plastic multi-layer structures, all of which require different initial
guess values.
This article presents the advances in a model that considers the interactions
between product, packaging, packaging conditions and environmental condi-
tions. This computational model integrates accepted phenomenological
models with a heuristic optimization algorithm in order to obtain packaging
structures with the lowest possible cost. At the same time, these structures
must take into account the specific requirements of oxygen and water vapor
permeability to ensure the food shelf life.
So far, the proposed model is the first that considers cost optimization and
integrates in a general way the requirements of the food and the variation of
54 Journal of Plastic Film & Sheeting 30(1)

the polymer barrier properties with the RH and temperature, in order to


design food packaging.

Computational model
The permeability model
A permeant is a gas or vapor of interest which diffuses through the package
wall. The diffusion process occurs when there are different permeant concen-
trations at the interior and exterior of the package. At steady state, the mag-
nitude of flow is defined by the magnitude of the permeant concentration
gradient and the resistance of the material to the diffusion. The flow direction
is defined by the direction in which the concentration decreases. This behavior
is described by Fick’s equation. This equation can be transformed in the
permeance equation when the following four assumptions are considered:
the diffusion coefficient is constant at any point across the thickness of the
wall of the package, the diffusion process is linear, permeants are ideal gases
and these gases can be dissolved on the polymer wall of the package according
to Henry’s law.1,2,15–17 The equation can be rewritten as
P VA
¼ ð1Þ
x ðpA1  pA2 ÞAs t

where P is the permeability coefficient, which is the product of the diffusion


coefficient and the solubility coefficient for the permeant A in medium B
(PAB ¼ DAB SAB ¼ P).
In the production of barrier packaging is very common to find multi-layer
structures. Each layer provides a specific barrier as a function of layer thick-
ness, the chemical structure of the material, temperature and RH.
Independent of the number of layers, the flow rate of the permeant A must
be equal in each layer. This fact makes possible the calculation of the total
permeance of the multi-layer structure using the following equation:1,2,15–17
 
P 1
¼  ð2Þ
x tot x 1 x 2 xn
P1 þ P2 þ    þ Pn

Permeability coefficients correction with temperature


Since the permeability coefficient changes with temperature, its value should
be obtained at the storage or shelf conditions of the multi-layer package/
product. In general, the gas permeability coefficients are reported in the lit-
erature at 23 C and 0% RH, while the water vapor permeability coefficients
del Pilar Noriega et al. 55

are measured at 37.5 C and 90% RH.24–27 In general, the permeability coef-
ficients should be given at STP conditions (1 atm and 0 C) for comparison
purposes. The relationship for the temperature correction follows the
Arrhenius equation:1,2,7
E 1 1
P ¼ Po eRðTTo Þ ð3Þ

Permeability coefficients correction with RH


Polar polymers are subjected to changes in permeance with RH. In the case of
these polymers in external or internal layers (in contact with food), a poly-
nomial regression is taken based on experimental data. The permeability
coefficients are corrected at the storage or shelf conditions of the multi-
layer package/product. If the polar polymer is placed in intermediate
layers, the influence of RH in permeance of layer j should be calculated
based on the water vapor permeance of the adjacent layers as shown in
equation (4).28
" ! !#
X
j1
xi xj RHout  RHin
RHj ¼ RHOUT  þ Pn x i ð4Þ
P 2Pj
i¼1 i i¼1 Pi

Determination of the product maximal permeance


Food products suffer physical, chemical and microbiological changes. When
these changes generate perceptible and unacceptable spoilage, the food shelf
life ends. The shelf life depends on the maximum gain or loss of gases for a
food. Table 1 shows some typical allowed oxygen gain values (G) for some
foods. For example, the presence of oxygen in the product promotes oxida-
tion reactions and significant changes in aroma and taste. With this value, the
calculation of the maximum allowable permeance to guarantee a determined
shelf life is possible.1
 
P GA Wprod
¼ ð5Þ
x max A ðpA1  pA2 ÞAs 

In the particular case of water vapor, the maximum allowed permeance is


calculated differently. Any kind of product can gain or lose moisture. The
moisture content is a function of the RH, or the water activity in equilibrium,
and the relationship between the moisture content and water activity is
56 Journal of Plastic Film & Sheeting 30(1)

obtained from the food sorption isotherm. The product absorption isotherms
show moisture gain and the product desorption isotherms show moisture loss
(see Figure 1). The moisture changes in the food product promote hydrolysis
reactions, growth of microorganisms or bacteria and loss of organoleptic
properties, among others. The moisture gain or loss into the package is rep-
resented by

dw P
¼ As ðp1  p2 Þ ð6Þ
dt x

In this case, the moisture gain or loss is expressed in mass unities. If the
total quantity of water in the package can be described by w ¼ W þ WHS ,
where W is water in food and WHS is water in the head space in package, the
additional moisture gain or loss is

dw ¼ dW þ dWHS ð7Þ

Air in the head space has a limited quantity and capacity to dissolve water
vapor before the saturation point in comparison with food (dW  dWHS ).
An acceptable supposition is that the moisture gain or loss into the package is
the food moisture gain or loss. Additionally, in equilibrium, the food water
activity is equal to the RH and is equal to aw ¼ p=pv . If the total water food
content is W ¼ mWprod , equation (6) can be re-written as

dm P As
¼ pv ðawe  aw Þ ð8Þ
dt x Wprod

When a linear model to describe sorption isotherms is used, the differential


equation has an analytical solution. In other cases, a numerical solution is
required. The calculation of the maximum allowable permeance for water
vapor to guarantee a determined shelf life, when the isotherm behavior
between the initial and critical food moisture can be considered linear, is as
follows:1,2,7,13
 
P
awe awi ð9Þ
x maxA¼lnðawe a
 wc Þ
As pv
Wprod ð b Þ


The maximum gas and water vapor permeabilities calculated are the max-
imum target values for any multi-layer structure with the goal of ensuring the
desired food shelf life.
del Pilar Noriega et al. 57

Formation of multi-layer structures


The computational model generates multi-layer structures considering the
number of desired layers (n) and the polymeric material selection (m) intended
for the package design. The calculation uses a statistical combination of m
materials and n layers satisfying following criteria:

. If compatibility is not guaranteed with respect to adequate adhesion


between layers, a tie layer is suggested. The algorithm considers that tie
layer only promotes adhesion and it does not provide a barrier. This
consideration is a good assumption because tie layers are generally
thinner.
. External and/or internal layers of the package are selected based on
the thermal sealability behavior of the polymer. This selection depends
of the faces that must be sealed to build the package (internal face/
internal face, external face/external face or internal face/external face).
. The algorithm suggests that hygroscopic polymers are placed in inter-
mediate layers of the film structure.
. The given maximum number of layers in the package structure as well
as the minimum thickness of each layer should be considered.
. The given thickness range (minimum and maximum) for the design of
the multi-layer film structure should be checked.

Optimization model
In order to establish a constrained optimization problem, it is necessary to
define the variables, the objective function and the equations that describe the
constraints. For the particular problem of cost minimization of multi-layer
structures, the thickness of each layer is considered a variable. The cost is
defined as the objective function that must be minimized (equation (10)). The
dimensional constraints are given by the maximum permeance of each per-
meant permitted through multi-layer structure (equation (11)) and the max-
imum and minimum thickness of each layer (equations (12) and (13)). If a
structure does not satisfy any one or more constraints, it is discarded.

Objective function: X
n  
Cost ¼ j Cj xj ð10Þ
j¼1

Constraints:
   
P P
 ð11Þ
x tot i x max i
58 Journal of Plastic Film & Sheeting 30(1)

xj  xmin j ð12Þ

xj  xmax j ð13Þ

Possible problem solutions


For any problem of constrained optimization which objective function pre-
sents a linear behavior, such as the cost function (10), the minimum is located
in one of the limits of the model constraints. For this particular case, the
minimum must be obtained in one of the following cases:

. The optimum is found when thicknesses are equal to the allowed min-
imum values (equation (12)), satisfying each permeability constraint
(equation (11)). Since the thicknesses are directly proportional to the
cost, and they have the lowest possible value, this is the trivial solution,
and the execution of the optimization algorithm is not required.
. When the minimum cost is obtained, at least for one permeant A, the
total structure permeance is equal to the maximum permeance food
requirement (equation (11)). In this case, the solution must satisfy the
constraints of the other permeants (equation (11)) and the layer (equa-
tions (12) and (13)) thickness constraints. The optimum solution is
not always obtained simply when a permeance value is reached. For
this reason, the optimization algorithm must be executed k times:
one for each permeant to which food is exposed. In this case, for the
permeant A:
   
P P
¼ ð14Þ
x tot A x max A

Algorithm
Deterministic methods, such as the Newton–Raphson method, are commonly
used to solve optimization problems. In that case, when a problem has mul-
tiple local minimums (see Figure 2), the result depends on an adequate selec-
tion of initial values, because the solution can be arrived to a local minimum
instead of the global minimum.
Initial values are not always easy to determine, especially when a high
number of variables and constraints are used. Several methods have been
developed in order to solve this problem. For example, Wah29,30 proposes
the constrained simulated annealing (SA) algorithm as a variation of the
traditional SA method. Johnston23,31 proposes the min-conflict heuristic algo-
rithm. This algorithm minimizes the number of constraint violations after
each iteration.
del Pilar Noriega et al. 59

Figure 2. Local minimum and global minimum.

This article proposes a similar method to the min-conflict method, but it


minimizes all violations at the same time and the flow diagram is presented in
Figure 3. The first step is to verify the trivial solution. If trivial solution is
invalid, the algorithm finds the minimum cost for each permeant, making
equal the total structure permeance and the maximum food permeance
requirement, using equation (14) and the SA method as the heuristic
method shown in Figure 3. The minimized cost for each permeant is com-
pared, selecting the lowest one and its respective thickness for each layer.
These thicknesses are used as initial guess values if these values are in the
domain of feasible solutions. A feasible solution must satisfy all constraints.
SA is sequentially executed for each constraint, regarding the equalization of
the permeant in question, looking for random near solutions until the restric-
tion or requirement is satisfied. SA must be executed as often as permeants
must be evaluated.
In Figure 3, R1 and R2 represent random numbers between zero and one,
[x] is the vector of thicknesses of each layer j, po and pmin are initial and
minimum probabilities to accept solutions with higher costs, " is the dumping
factor to reduce the probability p and x is the thickness gradient used to
modify the layers thickness in order to find the optimum. By accepting prob-
abilistically solutions with higher cost, local minimums are avoided. This
probability is reduced as the algorithm runs, assuming that the solution is
closer to the global minimum.
60 Journal of Plastic Film & Sheeting 30(1)

Figure 3. Optimization algorithm.

Experimental
Comparison between model and experimental results
The plastic film structures presented in Table 2 were evaluated in total oxy-
gen and water vapor permeance. The WVTR and OTR measurements
were carried out following international DIN and ASTM standards as
shown in Table 3.
del Pilar Noriega et al. 61

Table 2. Different film structures used to compare the model and the experimental
results.
Low and medium barrier High barrier coextruded
laminated film structures film structures

PET (12 mm)/PP (38 mm) PA (46 mm)/EVOH-F (8 mm)/PP


(28 mm)/PE-m (25 mm)
PET (12 mm)/PP (33 mm) PP (18 mm)/EVOH-F (4 mm)/PP (18 mm)
BOPP (25 mm)/BOPP(25 mm) PE (21 mm)/EVOH-L (4 mm)/PE (16 mm)
BOPP (20 mm)/BOPP PEARL (30 mm)
BOPP (20 mm)/BOPP (20 mm)
PET (12 mm)/PP WHITE (51 mm)
PET (12 mm)/BOPP PEARL (30 mm)
BOPP (17.5 mm)/BOPP METALLIZED
(17.5 mm)

EVOH-F has an ethylene content of 32%; EVOH-L has an ethylene content of 27%

Table 3. OTR and WVTR standard test and equipment.


Permeant Test Standard Conditions Equipment

Water vapor WVTR DIN 53380 II 40 C, 90% RH Bruger GDP-C


Volume: 375 mm3
Gas flow: 100 cm3/min
Transmission area: 78.4 cm2
Oxygen OTR ASTM D3985-95 23 C, 0% RH MOCON, OXTRAN 100
O2 flow: 10 (cm3/min)
N2 flow: 15 (cm3/min)
Transmission area: 20.3 cm2
Partial pressure of O2: 0.1568 bar

The experimental data were corrected for differences in the reference tem-
perature and RH using equation (12) and equation (13). Table 4 shows the
experimental and obtained results for water vapor permeance for various
films in Table 2. In the same way, Tables 5 to 7, show the experimental
and model calculated results for oxygen permeance using low, medium and
high barrier film structures.
62 Journal of Plastic Film & Sheeting 30(1)

Table 4. Water vapor permeance for several film structures: comparison between
measured and model calculated data.
Water vapor permeance

Measured Calculated
permeance permeance
Structure (g/m2 day atm) (g/m2 day atm) Variation (%)

PET (12 mm)/PP (38 mm) 18.11 17.32 4


PET (12 mm)/PP WHITE (38 mm) 19.38 20.90 8
PET (12 mm)/PP (33 mm) 22.09 19.86 10
BOPP (25 mm)/BOPP(25 mm) 7.76 7.45 4
BOPP (20 mm)/BOPP (20 mm) 10.41 10.77 4
BOPP (17.5 mm)/BOPP MET (17.5 mm) 1.04 1.04 0.04

Table 5. Oxygen permeance for several low barrier films: comparison between
measured and model calculated data.
Oxygen permeance: low barrier films

Measured Calculated
permeance permeance
(mL (STP)/m2 (mL (STP)/m2
Structure day atm) day atm) Variation (%)

BOPP (20 mm)/BOPP PEARL (30 mm) 715.89 959.08 34


BOPP (20 mm)/BOPP (20 mm) 774.96 956.91 23
BOPP (25 mm)/BOPP(25 mm) 783.07 579.96 26

Table 6. Oxygen permeance for several medium barrier films: comparison between
measured and model calculated data.
Oxygen permeance: medium barrier films

Measured Calculated
permeance permeance
(mL (STP)/m2 (mL (STP)/m2
Structure day atm) day atm) Variation (%)

PET (12 mm)/BOPP PEARL (30 mm) 96.95 90.52 7


PET (12 mm)/PP (38 mm) 113.52 90.12 21
PET (12 mm)/PP WHITE (38 mm) 113.19 90.12 20
PET (12 mm)/PP WHITE (51 mm) 96.34 88.56 8
PET (12 mm)/PP (33 mm) 109.92 90.73 17
del Pilar Noriega et al. 63

Table 7. Oxygen permeance for several high barrier films: comparison between
measured and model calculated data.
Oxygen permeance: high barrier films

Measured Calculated
permeance permeance
(mL (STP)/m2 (mL (STP)/m2
Structure day atm) day atm) Variation (%)

PA (46 mm)/EVOH-F (8 mm)/PP (28 mm)/PE-m (25 mm) 0.84 0.81 3.83
PP (18 mm)/EVOH F (4 mm)/PP (18 mm) 1.67 1.68 0.93
PE (21 mm)/EVOH L (4 mm)/PE (16 mm) 0.87 0.85 1.80

Numerical experiments
In order to check the validity of the proposed heuristic solution, a benchmark
test for linear optimization was designed. The solution cost of the algorithm is
compared with the known global minimum cost. In an optimization problem,
the minimum is known when the objective function and the constraints have a
linear behavior.2 To force that, it was supposed that the permeance of hygro-
scopic materials is not affected by humidity. In these conditions, the exact
minimum solution can be obtained by solving the problem using linear pro-
gramming methods such as the simplex algorithm.21 The evaluation was made
for several structures, comparing the heuristic solution with other numerical
method (GRG) and the exact known solution (calculated using a simplex
algorithm). The results are presented in Table 8.
When the variation of the barrier properties with regard to humidity is
considered, the constraint equation of allowable maximum permeance (11)
presents highly non-linear behavior and a non-linear optimization benchmark
test must be realized. In these conditions, the global minimum is not possible
to know. However, it is possible to compare the costs calculated by different
methods, checking that the constraints are fulfilled. The methods evaluated
were the GRG and SA. The results are presented in Table 9.
64

Table 8. Results of benchmark test for linear optimization.


Multi-layer structure 1

Thickness of each layer (mm)


Total
Method LLDPE Tie PET EVOH PET Tie LLDPE Perm. W.Va Perm. Ob2 Perm. COb2 thickness (m) Cost [¢/m2] Error %

Simplex 10.00 2.00 5.00 70.59 5.00 2.00 45.98 1.6000 0.505 6.0000 140.57 53.634 –
GRG 15.38 2.00 5.00 70.59 5.00 2.00 40.60 1.6000 0.505 6.0000 140.57 53.634 2.92 E06
Heuristic 27.04 2.00 5.01 70.60 5.00 2.00 28.94 1.5998 0.505 5.9996 140.60 53.640 1.01 E02
26.85 2.00 5.00 70.60 5.00 2.00 29.13 1.5998 0.505 5.9993 140.59 53.641 1.22 E02
26.52 2.00 5.00 70.60 5.00 2.00 29.46 1.5998 0.505 5.9995 140.59 53.638 7.93 E03
35.71 2.00 5.00 70.61 5.00 2.00 20.27 1.5998 0.505 5.9990 140.59 53.642 1.53 E02
25.58 2.00 5.00 70.60 5.00 2.00 30.40 1.5998 0.505 5.9993 140.59 53.641 1.22 E02

Multi-layer structure 2

Thickness of each layer (mm)


Total
Method ULDPE PP Tie EVOH PET Tie LLDPE Perm. W.Va Perm. Ob2 thickness (m) Cost [¢/m2] Error %

Simplex 10.00 10.00 2.00 8.85 5.00 2.00 60.26 1.6000 4.000 98.11 15.309 –
GRG 10.00 10.00 2.00 8.85 5.00 2.00 60.26 1.6000 4.000 98.11 15.309 4.21 E05
Heuristic 10.00 10.01 2.00 8.84 5.79 2.00 59.99 1.5998 3.996 98.63 15.447 6.05 E01
10.00 10.00 2.00 8.86 5.00 2.00 60.27 1.5998 3.996 98.13 15.315 3.89 E02
10.00 10.00 2.00 8.86 5.00 2.00 60.27 1.5998 3.996 98.13 15.316 3.93 E02
10.00 10.00 2.00 8.86 5.00 2.00 60.27 1.5998 3.996 98.13 15.316 3.99 E02
10.00 10.01 2.00 8.86 5.00 2.00 60.26 1.5998 3.996 98.13 15.316 4.48 E02

(continued)
Journal of Plastic Film & Sheeting 30(1)
Table 8. Results of benchmark test for linear optimization.
Multi-layer structure 3

Thickness of each layer (mm)


Total
Method LLDPE Tie PET EVOH Tie LLDPE Perm. W.Va Perm. Ob2 thickness (m) Cost [¢/m2] Error %

Simplex 10.00 2.00 5.00 8.8501 2.00 59.55 1.6000 4.000 87.40 12.692 –
del Pilar Noriega et al.

GRG 48.41 2.00 5.00 8.8501 2.00 21.14 1.6000 4.000 87.40 12.692 3.29E10
38.96 2.00 5.03 8.8622 2.00 30.59 1.5998 3.994 87.44 12.707 9.82E02
43.09 2.00 5.00 8.8605 2.00 26.46 1.5998 3.995 87.41 12.700 5.50E02
34.36 2.00 5.02 8.8582 2.00 35.18 1.5998 3.996 87.43 12.702 6.73E02
42.09 2.00 5.02 8.8726 2.00 27.45 1.5998 3.989 87.44 12.712 1.33E01
Heuristic 34.03 2.00 5.00 8.8610 2.00 35.52 1.5998 3.995 87.42 12.700 5.93E02

Multi-layer structure 4

Thickness of each layer (mm)


Total
Method LLDPE Tie PA-Amor. EVOH Tie LLDPE Perm. W.Va Perm. Ob2 thickness (m) Cost [¢/m2] Error%

Simplex 10.00 2.00 5.00 8.74 2.00 60.25 1.6000 4.000 87.99 13.253 –
GRG 60.26 2.00 5.00 8.74 2.00 10.00 1.6000 4.000 87.99 13.253 3.68E07
35.64 2.00 6.49 8.70 2.00 34.61 1.5933 3.990 89.45 13.710 3.58E02
44.71 2.00 5.00 8.75 2.00 25.78 1.5948 3.995 88.24 13.281 2.07E03
44.54 2.00 5.00 8.75 2.00 25.95 1.5948 3.995 88.24 13.281 2.09E03
39.76 2.00 6.91 8.67 2.00 30.42 1.5929 3.996 89.77 13.281 4.46E02
Heuristic 39.36 2.00 5.01 8.75 2.00 31.13 1.5948 3.993 88.25 13.285 2.39E03
a 2
Permeance of eater vapor. Units: (g/(m day atm)).
65

b
Permeance of gases. Units: (mL (STP)/(m2 day atm)).
66

Table 9. Results of non-linear optimization.


Multi-layer structure 1

Thickness of each layer (mm)


Total Diff. GRG vs.
Method LLDPE Tie PET EVOH PET Tie LLDPE Perm. W.Va Perm. Ob2 Perm. COb2 thickness (m) Cost [¢/m2] heuristic %

GRG 47.92 2.00 5.00 60.43 5.00 2.00 10.00 0.4994 1.6000 6.000 132.36 47.040 0.00Eþ00
39.07 2.00 5.16 62.69 5.11 2.00 18.33 0.4965 1.6000 5.985 134.37 48.554 3.09Eþ00
47.73 2.00 5.00 60.98 5.00 2.00 10.10 0.4951 1.6000 5.999 132.81 47.397 7.34E01
47.81 2.00 5.00 60.97 5.00 2.00 10.01 0.4950 1.6000 5.999 132.80 47.394 7.28E01
48.05 2.00 5.03 60.95 5.07 2.00 10.02 0.4952 1.5940 6.001 133.12 47.417 7.77E01
Heuristic 47.80 2.00 5.00 60.98 5.00 2.00 10.03 0.4950 1.6000 5.999 132.81 47.396 7.31E01

Multi-layer structure 2

Thickness of each layer (mm)


Total Diff. GRG vs.
Method ULDPE PP Tie EVOH PET Tie LLDPE Perm. W.Va Perm. Ob2 thickness (m) Cost [¢/m2] heuristic %

GRG 10.00 39.43 2.00 8.88 5.00 2.00 44.02 4.0000 1.600 111.33 16.634 0.00Eþ00
10.01 33.04 2.00 9.37 5.02 2.00 47.45 4.0000 1.600 108.89 16.675 9.01E02
10.00 35.03 2.00 9.22 5.00 2.00 46.39 3.9999 1.600 109.65 16.664 5.15E02
10.00 36.79 2.00 9.11 5.00 2.00 45.44 3.9991 1.600 110.35 16.670 6.47E02
10.01 25.07 2.00 10.24 5.01 2.00 51.68 3.9999 1.600 106.02 16.886 6.34E01
Heuristic 10.00 34.75 2.00 9.24 5.00 2.00 46.54 4.0000 1.600 109.54 16.662 7.49E02

(continued)
Journal of Plastic Film & Sheeting 30(1)
Table 9. Continued.
Multi-layer structure 3

Thickness of each layer (mm)


Total Diff. GRG vs.
Method LLDPE Tie PET EVOH Tie LLDPE Perm. W.Va Perm. Ob2 thickness (m) Cost [¢/m2] heuristic %
del Pilar Noriega et al.

GRG 59.84 2.00 5.00 7.32 2.00 10.00 4.0000 1.600 86.16 11.698 0.00Eþ00
52.60 2.00 5.01 7.49 2.00 17.21 3.9998 1.600 86.32 11.815 8.66E01
59.74 2.00 5.00 7.34 2.00 10.10 3.9998 1.600 86.19 11.713 1.12E01
57.62 2.00 5.00 7.38 2.00 12.21 3.9994 1.600 86.22 11.741 3.21E01
59.80 2.00 5.00 7.34 2.00 10.05 4.0000 1.600 86.18 11.712 1.06E01
Heuristic 57.23 2.00 5.00 7.39 2.00 12.61 3.9999 1.600 86.23 11.746 3.57E01

Multi-layer structure 4

Thickness of each layer (mm)


Total Diff. GRG vs.
Method LLDPE Tie PA-Amor. EVOH Tie LLDPE Perm. W.Va Perm. Ob2 thickness (m) Cost [¢/m2] heuristic %

GRG 60.76 2.00 5.00 7.27 2.00 10.00 4.0000 1.600 87.04 12.323 0.00Eþ00
60.98 2.00 5.00 7.25 2.00 10.63 4.0029 1.599 87.86 12.375 4.60E01
60.98 2.00 5.00 7.24 2.00 10.63 4.0038 1.599 87.86 12.375 4.57E01
61.55 2.00 5.00 7.25 2.00 10.07 3.9970 1.600 87.86 12.375 4.60E01
60.98 2.00 5.00 7.25 2.00 10.63 4.0025 1.600 87.86 12.375 4.59E01
Heuristic 61.56 2.00 5.00 7.25 2.00 10.05 3.9968 1.600 87.86 12.375 4.60E01
a 2
Permeance of water vapor. Units: (g/(m day atm)).
b
Permeance of gases. Units: (mL (STP)/(m2 day atm)).
67
68 Journal of Plastic Film & Sheeting 30(1)

Polymeric film structures, storage conditions and permeance restrictions


used in the benchmark tests are shown in Table 10 and the properties of
polymers required for the model are presented in Table 11.
For an additional check for the model proposed, the water vapor and oxygen
barriers of three seven layers structures were measured according to ASTM
F1249 and ASTM D3985 standards. The results are presented in Table 12.

Table 10. Constraints of each structure.


Maximum permeance
Area of package 0.0225 m2 for water vapor 1.6 (g/(m2 day atm))

Temperature 25 C
External humidity 80% Maximum permeance for O2 4 (mL (STP)/(m2 day atm))
Internal humidity 10%
Maximum thickness 250 mm Maximum permeance for CO2 6 (mL (STP)/(m2*day*atm))
of the structure
Multi-layer structure 1: it requires barrier against water vapor, O2 and CO2
Layers LLDPE Tie PET EVOH PET Tie LLDPE
Min thickness (mm) 10 2 5 5 5 2 10
Multi-layer structure 2: it requires barrier against water vapor and O2
Layers ULDPE PP Tie EVOH PET Tie LLDPE
Min thickness (mm) 10 10 2 5 5 2 10
Multi-layer structure 3: it requires barrier against water vapor and O2
Layers LLDPE Tie PET EVOH Tie LLDPE
Min thickness (mm) 10 2 5 5 2 10
Multi-layer structure 4: it requires barrier against water vapor and O2
Layers LLDPE Tie PA amor-phous EVOH Tie LLDPE
Min thickness (mm) 10 2 5 5 2 10

Table 11. Material properties.


Water
vapor CO2
permeability O2 permeability permeability Density Cost
Material Reference coefficienta,b coefficientb,c coefficientb,c (g/cm3)b [US$/kg]b

ULDPE Attane 4001, Dow Chemical 0.0209 46.8397 –d 0.905 2.75


PP Marlex 0.0217 30.0728 –d 0.910 0.99
PET Mylar, DuPont 0.0625 0.8205 2.3026 1.330 1.50
LLDPE Sclair 11F9, DuPont Canada 0.0088 32.4815 200.0000 0.921 0.90
EVOH Eval E, Eval Company 0.6100 0.0057e 0.0633e 1.140 5.83
PA AMORFO Selar PA 3426, Dupont 0.5500 0.1704e –d 1.100 2.93
Tie –d –d –d 1.00 –––
a
Permeability coefficient estimated at 25 C. Units: (g mm/(m2 day atm))
b
Values obtained from PDL-Handbook and www.matweb.com
c
Permeability coefficient estimated at 25 C. Units: (mL (STP mm/(m2 day atm))
d
Omitted values of permeability coefficient, considering low barrier of those materials.
e
Estimated value, omitting the variation of permeability coefficient with humidity for hygroscopic
materials (EVOH and PA).
Table 12. Results of measured and calculated permeances.
References

LDPE: polifen 640 Tie: bynel 4157 EVOH: evoh h101 LDPEþMLDPE*: LDPE 607a 60%þelite550g 40%
PA6 66: ultramid PP: ppsr257 LDPEþMLDPE**: polifen640þelite5401
del Pilar Noriega et al.

c35fn

Multi-layer structure 1

Thickness of each layer (mm)

LDPE LDPE Tie EVOH Tie LDPE LDPEþ Measured Calcula-ted Calcula-ted Measu-red Calcula-ted Cost per
MLDPE* Perm. W.V(1) perm. perm. perm. perm. unit area
at 38 C W.Va at 38 C W.Va at 23 C, Ob2 at 23 C, Ob2 at 23 C, [¢m2]
90%RH (3) 90% RH 0% RH 0% RHd 0% RH

Experimental 14.7 4.5 3.4 5.6 3.4 3.4 18.1 3.525 3.291 0.8983 1.735 1.767 7.10
Optimized 12.33 8.21 2.00 5.55 2.00 9.20 5.00 – – 0.8983 – 1.767 6.58
Multi-layer structure 2

Thickness of each layer (mm)

PP Tie PA6 66 PA 6 66 PA6 66 PA 6 66 LDPEþ Measured Calcula-ted Calcula-ted Measu-red Calcula-ted Cost per
MLDPE** Perm. W.V(1) perm. perm. perm. perm. unit area
at 38 C W.Va at 38 C W.Va at 23 C, Ob2 at 23 C, Ob2 at 23 C, [¢m2]
90%RH (3) 90% RH 0% RH 0% RHd 0% RH

Experimental 22.2 8.9 17.7 12.8 17.7 8.7 38.2 1.960 1.851 0.5223 0.763 0.751 25.07
Optimized 10.00 2.00 10.00 12.83 10.00 2.00 54.62 – – 0.5223 – 0.751 20.40
(continued)
69
70

Table 12. Continued.


References

LDPE: polifen 640 Tie: bynel 4157 EVOH: evoh h101 LDPEþMLDPE*: LDPE 607a 60%þelite550g 40%
PA6 66: ultramid PP: ppsr257 LDPEþMLDPE**: polifen640þelite5401
c35fn

Multi-layer structure 3

Thickness of each layer (mm)

PA6 66 PA6 66 PA 6 66 LDPE Tie LDPE LDPE LDPEþMLDPE** Measured Calcula-ted Calcula-ted Measu-red Calcula-ted Cost per
Perm. W.V(1) perm. perm. perm. perm. unit area
at 38 C W.Va at 38 C W.Va at 23 C, Ob2 at 23 C, Ob2 at 23 C, [¢m2]
90%RH (3) 90% RH 0% RH 0% RHd 0% RH
Experimental 20 16.7 6.86 23.74 25.22 14.08 52.44 1.330 0.960 0.2602 23.580 24.795 22.41
Optimized 18.79 17.90 2.00 38.51 26.50 45.32 10.00 – – 0.2602 – 24.795 21.80
a 2
Permeance of water vapor. Units: (g/(m day atm)).
b
Permeance of gases. Units: (mL (STP)/(m2 day atm)).
c
According to the norm ASTM F1249.
d
According to the norm ASTM D3985.
Journal of Plastic Film & Sheeting 30(1)
del Pilar Noriega et al. 71

Results and discussion


Experimental permeance data were compared with model calculated data for
several film structures. In Table 4 it can be observed that there is very good
agreement between experiments and calculated data for water vapor perme-
ance. The low barrier films show the highest deviation between measured and
calculated data such as is observed in Table 5. Medium barrier films show less
deviation than low barrier films and finally, high oxygen barrier films show a
very good agreement with the model calculated data as can be seen in Table 6
and Table 7, respectively. The high deviation for low oxygen barrier films can
be explained considering the high precision cell of the OTR equipment ade-
quate for high barrier measurements.
The summary of results, comparing the heuristic solution with GRG
method, and the exact known solution, calculated using complex algorithm,
are presented in Table 8. This table shows that the heuristic method offers a
good solution. The error is less than 1%. The other method, GRG, for all
cases delivered a better solution than the heuristic method. However, for that
method, convergence is highly conditional on initial guesses.
Table 9 presents the summary of results considering non-linear behavior of
permeance for hygroscopic materials (in this case focused on EVOH and
amorphous PA). For both materials, the permeability coefficient for O2 and
CO2 are not constant, therefore, they must be calculated by interpolation. The
results show for some cases lower costs for GRC and in other cases, lower
costs for heuristical method. Moreover, the heuristical method has the advan-
tage of lower dependence of initial guesses than GRG. Computing time using
SA has increased compared with the linear benchmark test. That is due to the
non-linear behavior and the iterative method, in order to calculate permeabil-
ity coefficient values.
Table 12 presents a comparison between permeance calculated and mea-
sured for three seven layer film structures using hygroscopic materials. It can be
observed that there is very good agreement between experiments and calculated
data. Additionally, the algorithm of optimization was executed to reduce the
cost of these structures guaranteeing the same permeability properties for water
vapor and oxygen at 0% RH and 23 C. The summary of results shows a sig-
nificant reduction of cost per unit of packaging area, changing the thickness of
each layer without affecting the desirable permeance of the structures and sat-
isfying the technical requirements for the end product.

Conclusion
According to the present study, it has been demonstrated that the developed
computational model is a valuable tool for predicting the permeance values of
72 Journal of Plastic Film & Sheeting 30(1)

plastic multi-layer structures, satisfying the requirements of a specific food or


beverage. The calculated values have a good accuracy for water vapor per-
meance and for oxygen permeance in a wide variety of films. The use of the
computational model facilitates multi-layer packaging design and it is an
excellent tool for film processing, film converting and food and beverage
companies.
The heuristical method algorithm proposed in this article for the optimiza-
tion offers a good alternative to the solution of non-linear constrained
approaches. The numerical experiments for the heuristical method delivered
good solutions, avoiding typical problems of other methods, such as conver-
gence problems, existence of local minimums and the required definition of
feasible initial guesses.

Funding
This research received no specific grant from any funding agency in the public, com-
mercial, or not-for-profit sectors.

Acknowledgments
The authors gratefully acknowledge invaluable support from ICIPC – Instituto de
Capacitación e Investigación del Plástico y del Caucho, Universidad EAFIT and
COLCIENCIAS.

References
1. Robertson GL. Food packaging: Principles and practice. New York, NY: Marcel
Dekker, 1993.
2. Hernandez RJ, Selke SEM, and Culter JD. Plastics packaging: Properties, process-
ing, applications, and regulations. Hanser Gardner Publishers, 2000.
3. Stoffers NH. Certified reference materials for food packaging specific migration
tests: Development, validation and modelling. Wageningen, the Netherlands:
Wageningen University, 2005.
4. Solvay Plastics sector. Major markets in 2001. Looking at Solvay Differently, June
2002.
5. Harte BR and Gray JI. The influence of packaging on product quality. In: Gray JI,
Mils J and Harte BR (eds). Food product-package compatibility proceedings. :
Technomic Publishing Co, 1987.
6. Robertson G. Food packaging: Principles and practice. Taylor & Francis/CRC
Press, 2006.
7. Kulchan R, Boonsupthip W and Suppakul P. Shelf life prediction of packaged
cassava-flour-based baked product by using empirical models and activation
energy for water vapor permeance of polyolefin films. J Food Eng 2010; 100:
461–467.
del Pilar Noriega et al. 73

8. Cárdenas FC, Giannuzzi L and Zaritzky NE. Mathematical modelling of micro-


bial growth in ground beef from Argentina. Effect of lactic acid addition, tem-
perature and packaging film. Meat Sci 2008; 79: 509–520.
9. Peychès-Bach A, Moutounet M, Peyron S, et al. Factors determining the trans-
port coefficients of aroma compounds through polyethylene films. J Food Eng
2009; 95: 45–53.
10. Piringer OG and Baner AL. Plastic packaging materials for food. Wiley, 2000.
11. Van Bree I, De Meulenaer B, Samapundo S, et al. Predicting the headspace
oxygen level due to oxygen permeability across multi-layer polymer packaging
materials: A practical software simulation tool. Innov Food Sci Emerg Technol
2010; 11: 511–519.
12. Rodriguez-Aguilera R, Oliveira JC, Montanez JC, et al. Mathematical modelling
of the effect of gas composition and storage temperature on the gas exchange rate
of soft cheese. J Food Eng 2009; 95: 82–89.
13. Genkawa T, Uchinob T, Miyamoto S, et al. Hamanaka. Development of math-
ematical model for simulating moisture content during the re-wetting of brown
rice stored in film packaging. Biosyst Eng 2008; 101: 445–451.
14. Scheichl R, Klopffer M-H, Benjelloun-Dabaghi Z, et al. Permeability of gases in
polymers: Parameter identification and nonlinear regression analysis. J Membr Sci
2005; 254: 275–293.
15. Mastromatteo M and Del Nobile MA. A simple model to predict the oxygen
transport properties of multi-layer films. J Food Eng 2011; 102: 170–176.
16. López ID, Estrada O, Noriega M del P, et al. Optimization model based on a
heuristic method for barrier films design. In: ANTEC, 2005. Nashville: Society of
Plastic Engineers.
17. Noriega M, Estrada O and Vargas C. Design of plastic multi-layer structure that
fit the requirements of a specific food or beverage. In: ANTEC, 2003, Nashville:
Society of Plastic Engineers.
18. Dennis JE and Lewis RM. A comparison of nonlinear programming approaches
to an elliptic inverse problem and a new domain decomposition approach. CRPC-
TR94468. Center for Research on Parallel Computation. Rice University, August
1994.
19. Alotto P and Nervi MA. An efficient hybrid algorithm for the optimization of
problems with several local minima. Int J Numer Methods Eng 2001; 50(4):
847–868.
20. Wrigth MH. Interior methods for constrained optimization. Acta Numerica-1992.
New York, NY: Cambridge University Press, pp.341–407.
21. Nelder JA and Mead R. A simplex method for function minimization. Comput J
1965: 7, New York, NY: Wiley.
22. Gabriele GA and Ragsdell KM. Large scale nonlinear programming
using the generalized reduced gradient method. ASME J Mech Des 1980;
102(3): 566–573.
23. Minton S, Johnston MD, Philips AB, et al. Minimizing conflicts: A heuristic
repair method for constraint satisfaction and scheduling problems. Artif Intell
1992; 58: 161–205.
74 Journal of Plastic Film & Sheeting 30(1)

24. DIN. Determination of water vapor transmission (density of moisture flow rate)
of plastics films, elastomer films, paper, board and other sheet materials. DIN
53122, Part 2.
25. DIN. Determining the gas transmission rate of plastic film, sheeting and mould-
ings by the carrier gas method. DIN 53380-3.
26. ASTM Standards. Standard test methods for water vapor transmission of mater-
ials. ASTM E-96.
27. ASTM Standards. Standard test method for oxygen gas transmission rate trough
plastic film and sheeting using a coulometric sensor, ASTM D-3985-95.
28. EVALCA. Gas barrier properties of resins. Technical Bulletin No. 110, 1996.
29. Wah BW and Chen YX. Optimal anytime constrained simulated annealing for
constrained global optimization. In: Sixth international conference on principles
and practice of constraint programming, September 2000, Springer-Verlag.
30. Wah BW and Wang T. Simulated annealing with asymptotic convergence for
nonlinear constrained global optimization. In: Principles and practice of constraint
programming. 1999, pp.461–475.
31. Johnston MJ and Minton S. Analyzing a heuristic strategy for constraint-satisfac-
tion and scheduling. In: Intelligent scheduling. Morgan Kaufmann, 1994.
32. Plastics Europe Market Research Group (PEMRG). Business data and charts.
Bruselas: Association of Plastics Manufacturers, 2007.
33. Vernaud JM and Rosca LD. Assessing food safety of polymer packaging.
Rapra, 2006.
34. PDL. Plastics and rubbers data collection. PDL Handbook Series, RoverTM
Inside, 1997.
35. Kirkpatrick S, Gelatt CD and Vecchi MP. Optimization by simulated annealing.
Science 1983; 220(4598): 671–680.
36. Ingber AL. Simulated annealing: Practice versus theory. J Math Comput Model
1993; 18(11): 29–57.
37. Ingber AL. Adaptive simulated annealing (ASA): Lessons learned. Control Cybern
1996; 25(1): 33–54.
38. Alexandrov NM. Multilevel and multiobjective optimization in multidisciplinary
design. AIAA paper 96- 4122, 1996.
39. Alexandrov NM and Dennis JE. Multilevel algorithms for nonlinear optimization.
In: Borggaard J, Burkardt J, Gunzburger M, Peterson J (eds) Optimal design and
control. Birkhauser, 1994, pp.1–22.
40. Cardoso MF, Salcedo RL and de Azevedo SF. Nonequilibrium simulated anneal-
ing: A faster approach to combinatorial minimization. Ind Eng Chem Res 1994;
33: 1908–1918.
41. Floudas CA and Pardalos PM. A collection of test problems for constrained global
optimization algorithms, Lecture Notes in Computer Science. Vol. 55, Springer-
Verlag, 1990.
42. Alexandrov NM and Dennis JE. A class of general trust-region multilevel algo-
rithms for nonlinear optimization: Global convergence analysis. SIAM J Optim
2000; (in revision).
del Pilar Noriega et al. 75

Appendix
Notation
aw ¼ water activity
awc ¼ critical water activity
awe ¼ equilibrium water activity
awi ¼ initial water activity
AS ¼ surface area h 3 i
3
ATR ¼ permeant A transmission rate LL2 t , i:e : mcm
2 day

b ¼ slope of sorption isotherm


cA ¼ permeant A concentration
Cj ¼ material cost of layer j per mass unity
Cost ¼ multi-layer structure cost per surface area unity
DAB ¼ diffusion coefficient of permeant A in B
E ¼ activation energy (J/mol)
GA ¼ permeant A gain or hloss i
3 3
JA ¼ flux of permeant A LL2 t , i:e : mm2 s
m ¼ food humidity
mc ¼ critical food humidity
me ¼ equilibrium food humidity
mi ¼ initial food humidity
pA ¼ permeant A partial pressure
pv ¼ water vaporh pressure i
3
PAB cm3
x ¼ permeance LL2 tP , i:e : m2 day atm
P
x maxA ¼ maximum permeance allowed of permeant A
 P
x totA ¼ total permeance of permeant A in multi-layer structure
QA ¼ permeant A volumetric flow
R ¼ value of the ideal gas constant (8314 J/(mol K))
RH ¼ relative humidity
SAB ¼ solubility coefficient of permeant A in B
t ¼ time
T ¼ storage temperature or any desired temperature (K)
To ¼ reference temperature (K)
VA ¼ volume of permeant A
w ¼ total quantity of water into the package
W ¼ water in food
WHS ¼ water in the head space in package
Wprod ¼ product weight
x ¼ thickness

 ¼ food shelf life


A ¼ permeant A density
76 Journal of Plastic Film & Sheeting 30(1)

Biographies
Marı́a del Pilar Noriega, PhD, is a Chemical Engineer from the Universidad
Pontificia Bolivariana, Medellı́n, Colombia with graduate studies at the
Technical University of Dresden, Germany and the Institute of Plastics
Technology (IKT, Stuttgart, Germany). She received her PhD from the
University of Wisconsin-Madison. Since 1990, she has worked for the
Research Institute for Plastics and Rubber (ICIPC, Medellı́n, Colombia);
she served as the technical director from 1997 to 2012 and currently acts as
the general director. She is the coauthor of many books and papers in the field
of polymers. She has several patents granted in Colombia, the US and other
countries. She is a member of Society of Plastics Engineers (SPE).

Omar Estrada, MsSc, is a Chemical Engineer from the Universidad Nacional


de Colombia, in Medellı́n, Colombia, with postgraduate studies in Plastic and
Rubber Transformation Processes at the Universidad EAFIT, Medellı́n,
Colombia, and a Master’s degree in Engineering in Polymer Processing at
the Universidad EAFIT. He worked in the Technical Division of ICIPC in the
area of extrusion from 1997 to 2006. Then he worked as the chief operating
officer at a plastics company until February 2009 and currently serves as the
Chief in the Extrusion Area at the ICIPC. He is the author of numerous
articles for national and international journals.

Iván Darı́o López, PhD, is a Production Engineer from the Universidad


EAFIT, Medellı́n, Colombia, with doctoral studies in Mechanical
Engineering with an emphasis on polymer processing at the University of
Wisconsin, Madison, United States. He worked as a teaching assistant at
the University of Wisconsin from 2006 to 2009. Author of several inter-
national publications in the SPE Annual Technical Conference among
others and a software developer at the ICIPC, Iván currently serves the
ICIPC as Technical Director.

View publication stats

You might also like