.3

Sm.

..

as

”5%

I
)’

Wm.

 

 

 

10w .LIpRARY
’é*¢:&5€5é’ M1ch1gan State
3 University

This is to certify that the
thesis entitled !

Development of Electronic Nose method for evaluation of
HDPE data, correlated with organoleptic testing

presented by

Rajarshi Das

has been accepted towards fulﬁllment
of the requirements for the

Master degree in Packaging

5”;an a” M

Major Professor’s Signature

 

12/10/03

 

Date

MSU is an Affirmative Action/Equal Opportunity Institution

L‘

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE | DATE DUE I DATE DUE

 

 

AN 2 61290“ 15 £33028? 2005

o

082 5 1'
immunity

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/ClRC/DateDue.p65-p.15

 

DEVELOPMENT OF ELECTRONIC NOSE METHOD FOR EVALUATION
OF HDPE DATA, CORRELATED WITH ORGANOLEPTIC TESTING

By

Rajarshi Das

A THESIS

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

MASTER OF SCIENCE

School of Packaging

2003

ABSTRACT

DEVELOPMENT OF ELECTRONIC NOSE METHOD FOR EVALUATION OF
HDPE DATA, CORRELATED WITH ORGANOLEPTIC TESTING

BY

Rajarshi Das

A standard method for analyzing food grade high density polyethylene
(HDPE) resins was developed using the electronic nose (E-nose) system. Eight
different HDPE resin grades were analyzed using the E-nose. The E-nose
system with Principal Component Analysis, was found to be capable of
discriminating between the resin grades. The resin samples were soaked in ultra-
pure water at 40 i 2° C for 1 week. Water samples stored in contact with the
resins were organoleptically evaluated at different concentrations by untrained
consumer sensory panels and the resins were ranked based on the degree of
off-ﬂavor perceived. The E-nose was used to correlate the sensory data and the
sensor responses for the various resin grades, using multivariate statistical
techniques.

A good correlation was obtained between the E-nose sensor responses
and the human sensory analysis data, when the resins were analyzed using
standardized experimental run conditions. The method is capable of predicting
the quality of an unknown food grade HDPE resin sample, in terms of its
possibility of producing an off-ﬂavor in drinking water, which could be a result of

migration of low molecular weight compounds from the polymer.

ACKNOWLEDGEMENTS

I would like to express my sincere gratitude to my advisor, Dr. Susan
Selke for her continuous support, inspiration and guidance. I also thank and
appreciate her for providing me valuable technical support during the project and
also for her critical review of the manuscript.

I would also like to extend my sincere thanks to Dr. Janice Harte for her
immense support and valuable advice in the area of sensory evaluation. A major
part of my research work was based on sensory analysis, which would not have
been possible without her. I thank all the sensory panelists, who were an integral
part of my project.

My sincere appreciation and gratitude is also extended to Dr. Bruce R.
Harte, for his valuable participation in my committee.

Acknowledgment and appreciation is extended to Alpha MOS for their
help in developing the experimental procedure and providing immense technical
support with the Electronic Nose. Sincere appreciation is also given to CFPPR
for ﬁnancial support.

Finally, I express my sincere gratitude and regards to my family and
friends for their motivation and support, which has a great contribution for the

success of this project.

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................ vi
LIST OF FIGURES .......................................................................................... vii
INTRODUCTION ............................................................................................. 1
CHAPTER I LITERATURE REVIEW ............................................................. 5
Polymers as packaging material ........................................................... 5
Interaction between plastics and food materials .................................... 7
Theory of migration ............................................................................... 7
Off-ﬂavor from packaging material ....................................................... 1O
Off-ﬂavor from polyethylene ................................................................. 12
Off-ﬂavor problem in drinking water ..................................................... 15
Analysis of off-ﬂavor from packaging materials .................................... 18
Sensory Evaluation .............................................................................. 20
Sensory evaluation of drinking water ................................................... 24
Instrumental analysis of off-ﬂavor ......................................................... 26
Electronic nose technology .................................................................. 28
CHAPTER II MATERIALS AND METHODS ................................................ 33
Electronic Nose System ....................................................................... 35
E-nose Operation Principle .................................................................. 41
Experimental Procedure ....................................................................... 43
E-Nose procedure ................................................................................ 43
Sensory Evaluation Procedure ............................................................. 45
Sample preparation ................................................................... 45
Experimental plan and design ................................................... 45
Correlation of E-nose and sensory results ........................................... 49
Multivariate Statistical Data Analysis ......................................... 49

CHAPTER III RESULTS AND DISCUSSION ............................................... 52

Evaluation of food grade HDPE resins using the E-nose system

(Standardizing E-nose run parameters) ............................................... 52

E-nose sensor response comparison ................................................... 65

Analysis of off—ﬂavor in water (stored in contact with resin samples) 69

Sensory Evaluation .............................................................................. 69
Correlation of E-nose and sensory responses ..................................... 72
E-nose analysis of off-ﬂavored water samples ..................................... 78

CHAPTER IV CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE

WORK .................................................................................... 80
APPENDICES ................................................................................................ 82
APPENDIX A Tray with water samples provided to consumer panel
(Triangle Test) — Photo ................................................ 83
APPENDIX B Kitchen area (sample trays are presented through the
booth windows) — Photo ............................................... 84
APPENDIX C Sensory evaluation in progress

APPENDIX D

APPENDIX E
APPENDIX F

BIBLIOGRAPHY .....

(consumer panel) — Photo ............................................ 85
Vials containing water sample,

used for sensory evaluation - Photo ............................ 86
Questionnaire for sensory evaluation (Triangle test).... 87

PCA - No discrimination among off-ﬂavored water
samples and reference water ....................................... 88

....................................................................................... 89

Table 1
Table 2
Table 3
Table 4

Table 5

LIST OF TABLES

Metal oxide sensors used in Fox E-nose systems ..................... 39
System conditions used for different experimental runs ............ 44
Sequence of sample presentation to the consumer panel ......... 47

Consumer panel response for off-ﬂavor in different water
concentrations at signiﬁcance level of 5% ................................. 70
Sensory scores

(resins categorized into 3 groups, based on the sensory panel
response) .................................................................................. 71

vi

Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9 (a)
Figure 9 (b)
Figure 10
Figure 11
Figure 12
Figure 13
Figure 14

Figure 15

Figure 16

LIST OF FIGURES

Experimental Summary (E-nose evaluation of HDPE resins

correlated with sensory results) ................................................. 34
Alpha MOS Fox 3000 E-nose system components ................... 35
Metal oxide sensor .................................................................... 36

Change in resistance of sensors

due to action of reducing chemical vapors ................................ 37
Architecture of gas array sensor system ................................... 40
Principle of Electronic Nose System .......................................... 42
Sensory evaluation experimental layout .................................... 46
Concept of PCA ......................................................................... 50
PCA (Run l conditions)

— Discrimination among 8 resin samples by E-nose .................. 53

PCA — magniﬁed portion (encircled in Fig. 9(a)) - Good

discrimination between DX, AP, EX, GX, BP ............................. 54
PCA (Run II conditions)
— Lower degree of discrimination than Run l conditions ............ 55
PCA (Run III conditions)
- Lower degree of discrimination than Run l conditions ............ 56

PCA (Run | conditions) - High degree of discrimination

between resins AP, BP, CP ....................................................... 58
PCA (Run II conditions) — resins AP, BP, CP
- lower degree of discrimination than Run I conditions ............. 59
PCA (Run MI conditions) — resins AP, BP, CP
- lower degree of discrimination than Run l conditions ............. 6O

PCA (Run I conditions) — high degree of discrimination

between resins DX, EX, FX, GX, HX ........................................ 62
PCA (Run II conditions) — resins DX, EX, FX, GX, HX
- Lower degree of discrimination than Run | conditions ............ 63

vii

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22(a)

Figure 22(b)

Figure 23

Figure 24

Figure 25

Figure 26

Figure 27
Figure 28

Figure 29
Figure 30

PCA (Run Ill conditions) - resins DX, EX, FX, GX, HX
— Lower discrimination than Run l conditions ........................... 64

Comparison of sensor responses (Resin CP)

— between Run I & ll .................................................................. 65
Comparison of sensor responses (Resin CP)
- between Run l & ll .................................................................. 66
Comparison of sensor responses (Resin CP)
— between Run I & III ................................................................. 67
Comparison of sensor responses (Resin CP)
— between Run l & llI ................................................................. 68

Correlation between expected and predicted values

for Run | conditions (assigned scores — 1,2,3) .......................... 73
Correlation between expected and predicted values

for Run I conditions (assigned scores - 1,3,5) .......................... 74
Correlation between expected and predicted values

for Run II conditions...................................... ............................. 76
Correlation between expected and predicted values

for Run lll conditions .................................................................. 77
Tray with water samples provided to consumer panel

(Triangle Test) ........................................................................... 83
Kitchen area

(sample trays are presented through the booth windows) ......... 84
Sensory evaluation in progress (consumer panel) .................... 85
Vials containing water sample, used for

sensory evaluation .................................................................... 86
Questionnaire for sensory evaluation (Triangle test) ................. 87
PCA — No discrimination among off-ﬂavored water samples and

reference water (negative index) ............................................... 88

viii

INTRODUCTION

The use of polymeric materials in packaging has been increasing at a very
fast pace. Polymers, in comparison to other materials, have proved to be highly
useable and competitive as packaging materials because of many advantages
such as durability, light weight, ﬂexibility, and low cost, engineered mechanical
properties, transparency, suitability for direct food contact and convenience
(Downes et al., 1996).

Food is one of the most important applications for plastic packaging.
Polymeric materials are widely used for food packaging. Developments and
ﬁndings in the area of food packaging can be applied to many other products and
systems, eg. pharmaceutical, chemicals, and other consumer products.
Polymeric packaging materials are not totally inert; hence there may be physical
and chemical interactions with the food product, which play a decisive role in the
selection of the packaging material. The nature of these interactions includes
permeation of gases and vapors across the package, migration of package
components and additives into the food, and sorption of food components. These
interactions can give rise to odors and degradation reactions in both the food and
the polymer.

One of the sources of off-ﬂavor and off-odor in food is the transfer and
migration of certain components from the packaging material. For example,
drinking water stored in plastic containers is often found to have a faint waxy

ﬂavor. Water, unlike other food items, cannot mask the off-ﬂavor that transfers

from some polyethylene packaging materials. Such migration could be because
of formation of carbonyl groups from overheated polyethylene, residual catalysts
in the polymer resin or antioxidant additives in the resins (Brody, 1989). These
off-ﬂavors are a concern to both the food industry and consumers as they can
affect the safety, organoleptic qualities, and nutritive values of the food products
(Thompson et al., 1994). Therefore, the need for effective quality control
procedures becomes more obvious and relevant to provide safe and quality food
product.

The compounds responsible for off-ﬂavors in the food have, in general,
very low concentrations, which sometimes become difﬁcult to detect even by
sensitive analytical instruments. Sensory analysis by consumer and expert
panels is extensively used to detect such off-ﬂavors, in addition to being used for
product development and batch quality control in the food and beverage industry
(Schmitt and Tan, 2001). Such organoleptic study has its own limitations as it is
highly subjective and lacks standardization. This is due to the fact that all
individuals are not equally sensitive to taste and odor perception. Human
subjects require repetitive training for each type of product and cannot work for
long stretches of time (Hansen and Wiedemann, 1999).

Gas chromatography-mass spectroscopy (GC-MS), which involves
separation of complex vapor mixtures into individual compounds, is one of the
most common techniques used to detect such off-ﬂavors. The GC-MS response
can be ambiguous, since strong off-ﬂavors might be generated by chemical

compounds even at very low concentrations (below the detection limits of the

instrument). It is usually difﬁcult to correlate such data with human sensory
analysis (Hansen and Wiedemann, 1999).

The electronic olfactory system, commonly called the electronic nose
system (E-nose), is an instrument that is analogous to the human olfactory
system. It generates olfactory response proﬁles for various aromatic components
of a food product. The E-nose is being extensively used in major food, cosmetic,
and packaging companies for applications such as quality control, spoilage
detection, flavor quantiﬁcation, and origin identification (Schmitt and Tan, 2001).
The E-nose results are objective in nature and can predict human sensory
response from physical measurements. This may make it a more reliable tool
than the human sensory panel in some applications and it minimizes
dependence on subjective sensory panel tests for product evaluation (Hansen
and Wiedemann, 1999). Electronic nose systems areused in the packaging
industry to detect residual solvents in flexible packages that are released from
printing inks, coatings, and adhesives; residual catalysts and antioxidants in
plastic resins; organic volatiles formed in a polymer during package fabrication;
and organoleptic changes in the product/package system.

The area of focus in this study is food-grade high density polyethylene
(HDPE) resin, which is one of the most versatile polymers for packaging
applications. These polymeric materials are used for drinking water pipeline
manufacturing in addition to manufacture of blow molded food containers. In both
cases, low residual organic compounds must be present in the HDPE pellets,

which can otherwise be a potential cause of off—ﬂavor in drinking water. Quality

control of these materials can be quite difﬁcult as only a few HDPE pellet grades
have residual organic compounds that are likely to cause off-ﬂavor in drinking
water (Schmitt and Tan, 2001).

The speciﬁc objectives of this study are:

1. To develop a standard method for evaluating food grade HDPE resins
using the electronic nose.

2. To correlate electronic nose results with sensory panel data for water off-
ﬂavor that is currently being used for the purpose.

3. To understand the experimental and statistical techniques and explore the
potential of the electronic nose as a quality control tool for a variety of
packaging applications.

This experimental method may eventually replace current sensory
evaluation techniques, with a simpler qualitative and quantitative objective
technique, offering better reliability and reproducibility of results and faster

operation.

CHAPTER I

LITERATURE REVIEW

Polymers as packaging materials

There has been tremendous growth in the development and design of
various packaging materials in the recent past, for a variety of applications. There
have been adaptations of many traditional packaging materials, eg. paper,
metal, and glass. A major contributor to this growth has been the development
and use of plastic packaging materials. The ﬂexible plastic packages may be
composed of a single polymeric material or a blend of more than one polymer or
a multi-layered laminated structure, composed of multiple components. Common
ﬂexible packaging materials include polyethylene, polypropylene, polyester,
polyamide, cellophane, ethylene vinyl acetate (EVA), polyethylene terephthalate
(PET), polystyrene, polyvinyl chloride, polyvinylidene chloride, and acrylonitrile
(Risch, 1988).

Food packaging is one of the major applications for such polymeric
packaging materials. Food packages can be in the form of single layered
extruded ﬁlms, blow-molded bottles, and multi-layered laminated structures,
depending on the degree of protection required from environmental, microbial,
and physical damage of the product. As there have been remarkable
improvements in the area of food package development, there has also been a
concern about interaction of plastic packaging materials and foods. One of the

major areas of concerns related to polymeric packages has been the presence of

compounds in toxicologically insigniﬁcant amounts, but at levels affecting the
ﬂavor (aroma/taste) of the packaged food. The concern is both from a consumer
standpoint and an industry perspective. The ﬂavor balance in the food may be
disrupted in three ways — absorption of components contributing to the desired
food ﬂavor by the packaging material (subtraction), chemical reaction of food with
the package components (reaction) or release of compounds from the package
into the food product (addition). Addition is the most common phenomenon for
causing off-ﬂavor in the packaged food (Kim-Kang, 1990). It is usually difﬁcult to
trace the source of off-odors and ﬂavors in food, primarily due to the fact that
most of the compounds that are responsible for such off-ﬂavors have very low
detection thresholds, even lower than the detection limits of the analytical
instruments available. Although such odor-active compounds may not pose any
health risk, they may cause concern in the consumer’s mind about possible
product contamination (Risch, 1988).

Drinking water, which is considered to be a food model (Cabral et al.,
1977), has a very subtle taste and can be affected by off-ﬂavors from packaging
materials very easily. Such off-ﬂavors and odors can be due to migration of
speciﬁc compounds from drinking water bottles made of high-density
polyethylene (HDPE), PET or PVC. A similar problem can be encountered when
plastic pipes (HDPE or PVC) are used for domestic water supply (Skjevrak et al.,
2003).

Interaction between plastics and food materials

As discussed earlier, plastic packaging materials are widely used for food
products due to many advantages. However they are not totally inert; hence
there may be physical and chemical interactions with the food product. There are
several different types of reactions that can occur, including scalping (sorption),
permeation, and migration. Scalping is the loss of a ﬂavor into the packaging
material. Such a phenomenon can also lead to damage to the package, e.g.
increase in movement of gases and water vapor across the package and
deterioration of physical characteristics and performance of the package.
Migration is the movement of components of the packaging material into the food
product, resulting in contamination of the product and potentially an undesirable
off-ﬂavor. Permeation of ﬂavors and gases through the packaging materials can
result in change of ﬂavor proﬁle and loss of ﬂavor intensity of the food product,
over time (Risch, 2000). Food-package interaction may also include a chemical
change in the food product, the package, or both, thereby leading to degradation

of the food or the package.

Theory of migration

The migration of low molecular weight compounds from the packaging
material, residual monomers and oligomers, residual solvents from printing inks,
adhesives, coatings, breakdown products of polymers and additives, are

primarily responsible for causing off-ﬂavor and off-odor in food products.

As explained by Kim-Kang (1990), the mechanism by which migration of
such low molecular compounds take place is based on the classic theory of the
diffusion of gases, described by Fick’s law. According to the law, the rate of
transfer, R, of a gas passing perpendicularly through the unit area of a section is
proportional to the concentration gradient through the section.

R = - D(C) [dC/dx] (1)

Where D(C) is the diffusion coefﬁcient in cm2 /sec, D can be a function of
the diffusant concentration, C is the concentration of the diffusant in mol/cm3 and
x is the thickness of the material in cm.

Kim-Kang (1990) also explained that the amount of package components
that may migrate from the packaging material to the solid or liquid food depends
on the physical and chemical properties of both the food and the polymer. The
controlling factors for the degree of migration are the original concentration of
migrants, the solubility at the contacting phase, the partition coefﬁcient between
the polymer and the contacting phase, temperature, time and morphology of the
polymer matrix (Gilbert et al., 1980). The properties of the polymer that can affect
migration include molecular weight, percent crystallinity, chain branching,
density, and afﬁnity for migrants (Shepherd, 1982). Kim-Kang (1990) further
explained that the diffusion mechanism depends on the volatility of the migrants.
The more volatile the components are, the higher is their concentration in the
package headspace. Migration does not always require direct contact between

the food and packaging material.

Ho et al. (1994) pointed out that there are three possible factors that
influence the degree of off-odor due to release of speciﬁc compounds from the
packaging material into the food, namely, odor threshold, molecular weight and
polarity. The lower the sensory threshold, the more readily the odor can be
detected, as in the case of aldehydes and ketones. The molecular weight of the
migrants controls their volatility. The lower the molecular weight, the more volatile
is the odor compound. This affects how rapidly the low molecular weight odor
compounds can desorb from the polymer matrix to the gas phase. Polar odor-
compounds have a greater tendency to desorb from the non-polar polymer matrix
(Ho et al., 1994).

Katan (1980) developed a model to deﬁne factors affecting migration in
the three phases: food, packaging material, and environment. The migration
variables included in the model were temperature, time, speciﬁc gravity of plastic
and solvent, original mass of migrant in plastic, mass of migrant transferred
during a specified time, mass of plastic and solvent, area of contact, and
geometry of the system.

Petrova et al. (1976) studied the possibility of using acrylonitrile—
butadiene-styrene (ABS) plastics for food contact applications. Three different
ABS plastics were tested and all three of them were found to impart some off-
ﬂavor to drinking water after 10 days contact due to migration of monomers, but
no off-ﬂavor was noticed after only 3 days. Water was found to have 0008-0019
mg/l of styrene and 0022-0057 mg/l of acrylonitrile, which was slightly higher

than the permissible limits.

Desobry (2000) studied the interaction between fatty foods and packaging
materials. It was observed that residual contaminants or microorganisms in
processed packaging materials could migrate into food and alter its quality. A
fatty food has high afﬁnity to the chemical contaminants in the package due to its
hydrophobic nature, leading to a potential safety problem of migration. The fat
molecules were found to enter the packaging absorbent sites and activate
contaminant migration. Such interactions were also found to modify the gas and
water permeability of the packaging material.

Sharma et al. (1990) reported the leaching of film anti-oxidants, e.g.
butylated hydroxyanisole and butylated hydroxytoluene, into refined peanut and
sunﬂower oils stored in contact with anti-oxidant containing polyethylene films at
37 deg. C. The storage stability of the oil was affected by film oxygen barrier

properties and the presence of antioxidants in the ﬁlm.

Off-ﬂavor from packaging material

Off-ﬂavors in packaged foods, which are often related to the packaging
materials, are causes of many customer complaints. An off-ﬂavor or off-odor
unfavorably changes the organoleptic properties of the food (Briston and Katan,
1974). Various sources of off-ﬂavor or off-odor compounds have been identiﬁed,
which primarily include residual monomers and oligomers, polymer volatiles and
additives, residual solvent from printing inks, types of adhesives and coatings,
and other factors like processing temperature, sealing temperature, sterilization

and extrusion processes (Thompson et al., 1994). The problem of off-ﬂavor in

10

food has major economic repercussions, which include recalls from the market
and compensation claims, lost consumer conﬁdence and shattered brand image
and trade relations (Huber et al., 2002).

Passy (1983) reported an off-ﬂavor in a fruit beverage due to migration of
certain components from an adhesive used for lamination of the packaging
material structure, polyester/aluminum/polyethylene, and also due to solvent
used in the printing inks. However very little contamination of dry powder
containing dextrose was found when it was packaged in the same packaging
material. Passy (1983) also reported an off-ﬂavor problem in chocolate and
lemon cream cookies packaged in polystyrene trays, wrapped with printed
cellophane. The prime reason for such off-ﬂavor was found to be due to styrene
monomer migration into the cookies.

Potts et al. (1990) explained the impact on taste characteristics of heated
apple juice and drinking water samples due to its interaction with low-density
polyethylene and ionomer. It was observed that the taste performance of the
polymers was associated with their processing conditions. An increase in melt
temperature or extruder output was found to increase the degree of oxidation,
thereby reducing taste properties. An increase in web speeds, causing less
surface oxidation, led to better taste properties.

Heydanek (1977) reported a “piney-spruce” off-ﬂavor in breakfast cereal,
which migrated from the package inner liner and speciﬁcally from the resin used

with microcrystalline wax to laminate two layers of kraft paper. This resin was

11

vacuum distilled and analyzed by gas chromatography, which eventually showed
nine peaks of “piney-spruce” off-odor.

Leong et al. (1992) investigated the development of off-ﬂavor in milk
packaged in polyethylene-coated paperboard cartons. Whole milk, low fat milk,
skim milk, and water were used as samples for study, and were ﬁlled into
polyethylene-coated cartons. The products were evaluated using a 10 member
sensory panel. Packaging off-ﬂavor was found to develop in the milk and water
samples after 1 day storage, which did not increase signiﬁcantly even after 3
days of storage. It was concluded that the off-ﬂavor compounds, which migrated
from the internal polyethylene (PE) coating, were formed due to oxidative
changes on the PE surface, which were largely soluble in water and quite
volatile. Interestingly, milk packaged in half-pint cartons was found to have more
off-ﬂavor than milk in quart or half-gallon cartons. HenCe it was conﬁrmed that
off-ﬂavor problems that occur due to migration from package components would
increase with decrease in the container size, due to the higher surface area to

volume ratio.

Off-ﬂavor from polyethylene

As described in the earlier sections, packaging materials may contain
components that may transfer into the packaged food product and can adversely
affect product ﬂavor characteristics.

The source of off-odor in polyethylene materials might be i) formation of

carbonyl groups due to overheated polyethylene, which typically occurs during

12

extrusion and package manufacturing processes; ii) residual catalysts in the
polymer resins, which depends on the resin manufacturing process and polymer
characteristics and iii) antioxidant additives in polyethylene resins, which prevent
the development of oxidized odors but can also create their own off-odors
(Brody, 1989). Catalysts and antioxidants can sometimes interact to generate off-
odor compounds.

Polyoleﬁns such as polyethylene are often processed at high
temperatures in the presence of oxygen and high shear stress, which may
eventually lead to oxidative and thermal degradation of the polymer. Usually, low-
density polyethylene (LDPE) tends to degrade to shorter chain hydrocarbons at
temperature greater than 300°C and then further undergoes oxidation to form a
variety of low molecular weight oxidation products such as aldehydes, ketones
and acids (Hoff and Jacobsson, 1981; Kim-Kang, 1990). Such products of
thermo-oxidative reactions in polyolefins can develop adverse ﬂavors in the food
product (Kim-Kang, 1990).

Bravo et al. (1992) collected volatiles from thermal oxidation of
polyethylene and analyzed them by a gas chromatography/olfactometry
technique. Fourteen odor-active compounds were detected, mainly saturated and
unsaturated aldehydes and ketones, which were primarily responsible for off-
odor associated with themno-oxidation of PE. About 46% of the odor-active
compounds detected from oxidized polyethylene were hexanal, 1-hepten-3-one,
1-octen-3-one, octanal, 1-nonen-3-one, nonanal, trans-2-nonenal, and diacetyl.

The overall odor was termed as “waxlike”.

13

In a similar study, Hoff and Jacobsson (1981) identiﬁed 44 compounds
resulting from thermo-oxidative degradation of polyethylene between the
temperatures of 264 — 289° C, using GC-MS analysis. These compounds were
mainly hydrocarbons, alcohols, aldehydes, ketones, acids, cyclic ethers, cyclic
esters, and hydroxy-carboxylic acids. Out of all the compounds identiﬁed, fatty
acids and aldehydes were predominant.

Ho and Yam (1999) reported the effect of vitamin E formulation in HDPE
resin pellets on off-odor release. An antioxidant formulation of vitamin E, glycerol,
polyethylene glycol-400 (PEG-400), and glyceryl monocaprylate/caprylate (GMC)
was used for the study. Off-odor release from the resin pellets was evaluated
using a sensory panel and GC/MS technique. The results conﬁrmed that GMC
was the main off-odor contributor.

Yam et al. (1996) studied the effects of three types of HDPE resins (A, B,
and C) and three antioxidants (vitamin E, lrganox 1010, and BHT) on the release
of off-ﬂavor from blow-molded HDPE bottles, using sensory evaluation and
GC/MS analysis. The sensory results conﬁrmed that off-ﬂavor intensity was
inﬂuenced by both resin type and antioxidant. The GC/MS study identiﬁed more
than 60 volatile compounds released from the bottles, which belonged to the
groups of n-alkane, 1-alkene, aldehyde, ketone, phenolic, oleﬁn, and parafﬁn.
The aldehydes and ketones were found to have very low odor thresholds. Bottles
made with Resin A, which had Vitamin E as the antioxidant, yielded the least off-

ﬂavor compared to the other two bottles made with Resins B and C. This was

14

primarily due to the fact that the Resin A bottle yielded less aldehydes and
ketones than the other two.

Fauconnier et al. (2001) developed a mathematical model for migration of
odor components from HDPE into hexane, ethanol, lemon terpenes, and their
emulsions, which are concentrated solutions used as food ﬂavorings. Analysis
indicated the presence of 2,4-bis(1,1—dimethylethyl)phenol in all the liquid
solutions at all temperatures, which was probably a degradation compound of

the antioxidant in the polymer.

Off-ﬂavor problems in drinking water

Off-ﬂavor in drinking water is a very common problem, both in cases of
bottled water and domestic water transported in plastic pipes. The off-ﬂavor in
water is usually attributed to the plastic aftertaste, which is reportedly caused by
migration of low molecular weight compounds and degradation products from the
plastic.

Taste and odor complaints from consumers are a major problem for
suppliers of drinking water. Small traces of chemicals, which could be naturally
present in water or produced during water treatment or even leach out from the
package, can alter the organoleptic properties of drinking water (Young et al.,
1995). Brody (1989) reported that water, which has practically no aroma, cannot
mask the faint waxy or burnt odor that transfers from polyethylene containers.

Fayad et al. (1997) studied the migration of vinyl chloride monomer and

plasticizer migration from polyvinyl chloride (PVC) packaging material into bottled

15

drinking water. It was reported that concentration of vinyl chloride monomer
(VCM) in various brands of bottled water was less than 0.6 ppb (parts per billion),
which is far below the 2 ppb maximum concentration limit set by the US
regulatory body. However, the GC-MS analysis of the bottled water sample
revealed the presence of several volatile and semi-volatile organic compounds,
after the bottled water was exposed to sunlight. Dichloroacetic acid and 2,3-
dichloro-1-propanol were the volatile compounds identiﬁed. In addition, the
presence of benzene was conﬁrmed in some bottled-water samples. Di-n-octyl
adipate and bis(2-ethylhexyl) phthalate, which are widely used plasticizers for
PVC, were the major semi-volatile organic compounds identiﬁed. Factors
affecting migration were found to be the storage time, temperature, and exposure
to sunlight.

Solin et al. (1988) studied the effects on the taste of drinking water stored
in PVC, HDPE, polycarbonate, and polyethylene terephthalate (PET) bottles at
room and elevated temperature (120° F), for 4 weeks. Water stored in PVC
bottles was reported to have the least off-ﬂavor, followed by PET, polycarbonate,
and HDPE.

Crompton (1979) reported an increase in pH and total solids in bottled
water, which could be due to migration of polymeric residues, non-polymeric
additives and adhesive compounds from the plastic package.

Calvosa et al. (1994) analyzed the taste quality of drinking water samples
stored in low-density polyethylene (LDPE) containers, which were exposed to

direct sunlight for 2 weeks. GC-MS analysis of the water samples revealed the

16

presence of ketonic compounds, from the photo-decomposition of LDPE, in
addition to 3,5-dimethoxybenzaldehyde, n-butyl phthalate and i-butyl phthalate,
from the ink.

Off-ﬂavor in drinking water is a problem with plastic pipes. The
compounds causing off-ﬂavors are mainly carbonyl compounds. HDPE has
gained more acceptance as a material for drinking water pipe manufacture
because of its favorable mechanical properties, ease of handling during
manufacture, and low permeability to external contaminants (Villiberg at al.,
1998)

Anselme et al. (1985) performed a study on the cause of an intense plastic
taste and odor problem in drinking water transported using polyethylene pipes in
southern Paris, France, in summer 1984. Analysis of the water samples from the
pipe by GC-MS revealed the presence of polymer additives, e.g. lubricants,
antioxidants, stabilizing agents, and polar compounds such as aldehydes. The
primary causes of such organoleptic changes in the water were attributed to
dissolution of polymer additives and oxidation of the interior surface of the pipe
during extrusion, with subsequent release of polar compounds. The burnt plastic
odor of the water was due to migration of butylated hydroxytoluene (BHT), a
common antioxidant used during plastic pipe and bottle manufacture.

Brocca et al. (2002) reported various organic chemicals, which migrated
from four different types of polyethylene (used commonly for pipeline
manufacture) into drinking water. Most of the compounds detected had a basic

common structure characterized by a phenolic ring typically substituted with alkyl

17

groups in the 2 and 6 positions of the aromatic ring. The presence of some of
these compounds was attributed to impurities or by-products in phenolic
additives, which are typically used as antioxidants during pipeline manufacture.
The presence of a few other compounds was attributed to degradation products
from the polymer additives during the extrusion process at ZOO-250° C in the
pipeline manufacturing process.

In a similar study, Skjevrak et al. (2002) analyzed the migration of volatile
organic compounds from HDPE, cross-bonded polyethylene (PEX), and PVC
pipes into drinking water, which eventually caused off-odor in water. A wide
range of esters, aldehydes, ketones, aromatic hydrocarbons, and terpenoids
were identiﬁed as migration products from the HDPE pipes. However the major
source of off—odor in water flowing in HDPE pipe was attributed to leaching of
2,4-di-tert-butyl-phenol, which is a degradation product from the antioxidants
used in pipeline manufacture. The leaching of phenolic antioxidants such as BHT
and alkylbenzenes has also been reported as a signiﬁcant cause of off-odor in
drinking water. The predominating volatile organic compound in the water from
PEX pipes was methyl-tert-butyl ether (MTBE), whose concentration was found
to be higher than the permissible limits. MTBE was considered to be a probable
cause of off-odors in the water sample. The organic compounds in the test water

from PVC pipe included hexanal, octanal, nonanal, and decanal.

18

Analysis of off-ﬂavor from packaging materials

As discussed in the earlier sections, the potential package-related cause
of off-ﬂavor in the food is the migration of residual monomers and oligomers,
breakdown products of polymers and specific additives, and residual solvents
from printing inks, adhesives, and coatings. This demands strict quality control by
the package and polymer manufacturer, so that the taste and odor problems
associated with leaching of volatile compounds from the package can be kept at
a minimum.

It has been really difﬁcult to evaluate taste and odor problems both
subjectively and objectively. The situation becomes even more difﬁcult when very
low levels of odor-active compounds are present. Typically, sensory evaluation
by human subjects is used to detect such off-flavor, as it is easier to detect some
chemicals by subjective organoleptic evaluations, as they are present in levels
far below the detection threshold of analytical instruments such as GC-MS, solid
phase micro extraction, and E-nose. Determination of odor threshold and the
concentration of the odor-compound is a very important step in aroma research.
The concept of odor threshold is very useful in deﬁning odor purity, describing
intensity and odor quality and evaluating which of the components contributes to
a characteristic aroma (Teranishi et al., 1991).

Hodgson et al. (2000) reviewed literature on volatile organic compounds
that originate from polyethylene during its manufacture, processing, storage, and
service life. It was reported that analysis of such volatile organic compounds and

off-ﬂavors associated with them was performed using methods such as sensory

19

evaluation, chromatograph techniques and their associated sampling techniques,
including the "hot-jar" method and dynamic headspace sampling, gas
chromatograph-olfactory sensing, and artiﬁcial olfaction (electronic nose)
technology.

Vom Bruck and Hammerschmidt (1977) used headspace gas
chromatography to determine off-ﬂavor compounds associated with packaging
materials such as LDPE and polystyrene, which would alter the organoleptic
properties of the food.

Kiritsakis et al. (2002) explained the various methods to perform ﬂavor
analysis of olive oil, which include GC-MS technique, aroma extract dilution
analysis (AEDA), electronic olfaction technology (E-nose) and advanced
methods such as nuclear magnetic resonance spectroscopy, fourier-transfomn
infrared technique, differential scanning calorimetry, high performance liquid
chromatography (HPLC) and solid phase micro extraction (SPME). These
methods can also be used to study oil-package interaction and off-ﬂavor

analysis.

Sensory Evaluation

Sensory analysis is a formalized, structured and codiﬁed methodology
used for evaluation of physical appearance and organoleptic qualities of a food
product. The various sensory attributes of a food item are appearance,

odor/aroma/fragrance, consistency and texture, and ﬂavor (aromatics, texture,

20

chemical feelings). However most of the attributes overlap with each other
(Meilgaard et al., 1991).

In some cases, sensory analysis works better than instrumental analysis
for determining off-flavors in food products, since some off-ﬂavors can be
detected by taste but cannot be effectively and accurately measured using
analytical instruments. In general, sensory panels can detect off-ﬂavors and off-
odors present at much lower concentrations than can analytical techniques
(Peled and Mannheim, 1977).

Linssen et al. (1991) evaluated a taint in water packed in test pouches
made of LDPE-lined aluminum, using 48 sensory panelists. Eight descriptive
attributes were used by the panel: metallic, synthetic, dry, rough, astringent,
musty, sickly and penetrating. Fourteen panelists were used to judge intensities
for these attributes on a visual analog scale. Factor analysis reduced the original
data matrix to a 6 dimensional one. Synthetic and penetrating loaded high on one
factor where as rough and astringent on another one.

Peled and Mannheim (1977) studied off-ﬂavor in milk using sensory and
GC techniques. It was reported that the organoleptic tests were more reliable
than gas-chromatograph tests, in determining packaging off-ﬂavors.

Olsen and Ashoor (1987) studied the impact on ﬂavor, odor, and appearance of
retail milk samples stored in 3.8 liter HDPE plastic containers and 1.9, 0.95, and
0.48 liter ﬁberboard containers and displayed under normal ﬂuorescent light. 25
untrained sensory panelists were employed to detect any light induced off-ﬂavor

in milk. The riboﬂavin content was determined by the HPLC method. The results

21

indicated that under typical production and storage conditions in the area, type
and size of container, fat content, and season of production had no major
damaging effects on ﬂavor, odor, appearance, or riboﬂavin content of retail milk.

Adebiyi et al. (2002) used sensory analysis to evaluate the quality
attributes of dry roasted peanuts which were processed in five different ways,
packed in four different packaging materials, and stored under three different
relative humidity conditions for 3 months at ambient temperature. Ten trained
sensory panelists were employed to evaluate the color, taste, ﬂavor, texture,
crunchiness, and overall acceptability using a nine point hedonic scale. The
sensory results indicated that there was no signiﬁcant difference in color scores
among all peanut samples. Salting was found to improve the taste, ﬂavor,
palatability, and overall acceptability of dry-roasted peanuts. However salting had
no effect on the level of crunchiness and texture of the dry roasted peanuts.

Al-Bachir and Mehio (2001) conducted sensory evaluation to detect
differences between irradiated and non-irradiated luncheon meat during an effort
to study the effect of gamma irradiation on shelf life of luncheon meat. A
consumer panel comprised of 20 subjects was employed to evaluate the taste
and ﬂavor of the meat and rank the samples on a 5-point scale. The sensory
evaluation indicated that no signiﬁcant differences were found between irradiated
and non-irradiated samples in terms of taste and ﬂavor.

Church and Parsons (2000) employed sensory evaluation techniques to
study the quality of chicken breast and of sliced potatoes in cream, both

immediately after cooking under vacuum (sous vide) and following subsequent

22

chilling, chilled storage and reheating. Trained panels comprising of 25 subjects
were employed and the quality of the food was evaluated using attribute scaling
methods. Results revealed that sous vide signiﬁcantly increased the ﬂavor
intensity of both products (p<0.05) and the juiciness and moistness of the
chicken and potato respectively (p<0.05), compared to non-vacuum packed
freshly cooked products.

Wheeler et al. (1999) employed a trained descriptive attribute sensory
panel to evaluate the palatability of gamma irradiated vacuum-packaged frozen
ground beef patties and the taste of hamburgers made with those patties. The
trained panel evaluated grilled patties for ground beef aroma intensity, off-aroma,
and off-ﬂavor on 4-point scales (4 = intense; 1 = none) and ground beef ﬂavor
intensity, tenderness, and juiciness on 8-point scales (8 = extremely intense,
tender, or juicy; 1 = extremely bland, tough, or dry). Control patties had more
intense (p<0.05) ground beef aroma, less off-aroma, and more intense ground
beef ﬂavor than irradiated patties. Hamburgers made with patties treated with 4.5
kGy radiation were rated lower (p<0.05) in taste than hamburgers made with
either control patties or those treated with 3.0 kGy radiation.

Okayasu and Naito (2001) employed sensory panels to evaluate the
sensory characteristics of unclariﬁed apple juice and to compare unclariﬁed and
clariﬁed types. 140 consumers and 10 trained panelists evaluated 16 apple juice
samples (4 clarified and 12 unclarified). It was difﬁcult to predict consumer
preference by regression models using trained panel results and analytical

attributes.

23

Durst and Laperle (1990) employed a 12 member trained sensory panel to
determine the presence of styrene off-ﬂavor in apple juice stored in polystyrene
containers. A standard triangle test method was used, where each panelist was
presented with three glass samples of apple juice, comprised of two glass
controls and one stored in plastic, or one control and two samples stored in
plastic. The panelists were required to choose the odd sample and detect styrene
related off-ﬂavor. The sensory panel did not perceive a difference in ﬂavor
between juice packed in polystyrene containers and glass jars after 1 and 2 week
storage times at 24° C. More than 50% panel members detected a styrene-like
off-ﬂavor after 4 weeks storage. A signiﬁcant difference at 99% confidence level

was found after a span of 8 weeks.

Sensory evaluation of drinking water

Drinking water is very sensitive to any off-ﬂavor or odor, which can be due
to any contamination from an external source or due to migration of certain
compounds from the package into the water. Recently, there have been rapid
developments in the area of powerful analytical techniques. However the human
nose can still easily detect trace amounts of chemicals at levels many times
lower than the analytical detection limits. Although many taste and odor
assessments of drinking water have been conducted using sensory panels, such
sensory data played up till now only a minor role in the management of the water
treatment plant and understanding packaging requirements of water (Koster et

al., 1981).

24

Young et al. (1995) used sensory evaluation to determine taste and odor
threshold concentrations of 59 potential drinking water contaminants. These
contaminants can be naturally present in raw water or can be introduced from
industrial sources or during water treatment. The panelists included trained
female subjects who ranged between 22 and 25 years of age and were above
average in their basic sense of taste and odors. The subjects were not allowed to
eat or drink or wear any kind of cosmetics or perfumes during the organoleptic
evaluation, as those might interfere with the evaluation. The results indicated that
there was no correlation between organoleptic effects of drinking water
contaminants and their toxicity. It was found that chemicals with lowest taste and
odor threshold values included chemicals such as geosmin, 2-methyl-isoborneol,
and chlorinated phenols and anisoles.

Righi et al. (1999) performed a study on off-odor problems in non-
carbonated mineral water packaged in 750 ml. disposable glass bottles and
sealed with aluminum screw caps. The off-odor developed during storage of
water in the glass bottle. A ﬁve member sensory panel evaluated a set of 120
samples. The water odor problems were attributed to screw caps contaminated
with a Penicillium fungus.

Ho et al. (1994) evaluated the organoleptic properties of HDPE containers
produced using Vitamin E, lrganox 1010, and BHT as antioxidants, using sensory
evaluation of water. 32 untrained panelists were employed. A speciﬁed amount of
water was ﬁlled in the HDPE containers and stored in controlled test conditions

for 4 days. The taste of the water was evaluated by a duo-trio test. The odor of

25

the water sample was judged by snifﬁng the headspace of the bottle. A 15 cm
line scale was used to assign scores to the water samples based on the degree
of off-ﬂavor or off-odor present. “0” was assigned if there was no detectable odor

and “15” was assigned if there was a strong off-ﬂavor.

Instrumental analysis of off-ﬂavor

Instrumental analysis of trace organic compounds in food requires very
sensitive analytical instruments, which can detect volatile vapors at very low
concentrations and have high discrimination capability. The most common
instrumental methods used for detection and identiﬁcation of unknown volatile
compounds are gas chromatography - mass spectrometry (GC-MS), electron
impact (El); chemical ionization (Cl); and gas chromatography — infrared
spectroscopy (GC-IR) (Maneesin, 2001). Among these, gas chromatography -
mass spectrometry (GC-MS) is the most widely used method to detect taste and
odor compounds in food.

Steele at al. (1994) determined styrene levels in 12 different food
commodities using analytical measurement techniques such as the dynamic
heated headspace purge-and—trap extraction technique, which was followed by
quantiﬁcation using selected ion monitoring capillary gas chromatography/mass
spectrometry.

Tombesi (2002) developed a method to detect the presence of butylated

hydroxytoluene (BHT) in mineralized bottled drinking water, using solid-phase

26

microextraction (SPME) and GC-MS. BHT is a common antioxidant used in bottle
manufacture, which has a high chance of migration into the water sample.

Wyatt (1986) developed a dynamic headspace technique, to analyze the
taste and odor compounds imparted from packaging materials. In this method,
volatile aroma and ﬂavor compounds were generated by heating the packaging
material into a vapor stream, subsequently swept away by nitrogen gas and
purged. The volatiles were then trapped by adsorbents and analyzed by GC-MS.
The procedure was found to be rapid and reproducible, with a low detection limit
and high precision.

Doust et al. (2003) studied the migration of organic compounds from
polyethylene to packaged water. Liquid — liquid extraction and GC-MS were used
to study the possible migration of certain organic compounds, such as oligomers,
from polyethylene to packaged water. Primarily, tridecane, tetradecane,
hexadecane, octadecane, and eicosane were detected in the polymer, using the
analytical techniques, which was believed to alter the organoleptic properties of
drinking water.

Giese (2000) pointed out that the classical analytical techniques such as
GC-MS could separate, quantify, and identify individual volatile chemicals.
However such techniques cannot tell if the component has an odor and it is often
very difﬁcult to correlate this data with the sensory evaluation data. Human
sensory evaluation, as discussed earlier, is a powerful method, but it has
limitations, such as it is time consuming and expensive. The sensory study

conducted by trained human subjects is subjective, and illness and other factors

27

can adversely affect their performance. Giese (2000) also explained that
instruments like the electronic nose could be highly advantageous compared to
other analytical instruments and sensory evaluation methods. The prime
advantages of the electronic nose include rapid, real-time detection of volatiles,

less preparation time, greater safety, and lower costs.

Electronic nose technology

The electronic nose is deﬁned as an instrument comprised of electronic
chemical sensors with partial speciﬁcity and an appropriate pattern recognition
system, capable of recognizing simple or complex odors (Gardner and Barlett,
1993)

The basic mechanism of an electronic nose is to generate headspace over
the sample being tested, present the headspace gas to the sensors, record the
sensors’ response, and analyze the data. Different types of sensors are
commercially used in electronic noses, which include metal oxide sensors,
conducting polymers, and quartz crystals. Metal oxide sensors are made from
zinc or tin oxide. Such sensors are operated between two electrodes at high
temperatures - 300° C. The aroma compound gets oxidized on the surface of the
sensor and changes the resistance of the sensor. Conducting polymer sensors
are obtained by electro-polymerization of a thin ﬁlm of polymer across the gap
between gold-plated electrodes. The electrical conductance of the ﬁlm changes
according to the odor compounds adsorbed on its surface. In the quartz crystal

category, two different types of sensors are used. One is based on sensing the

28

mass of the aroma compound absorbed into the stationary phase coated on the
crystal surface. The adsorption changes the frequency of vibration of the crystal,
due to change in mass. These sensors are called quartz microbalances. The
second type of sensor is a surface acoustic wave device. It operates similarly to
the quartz microbalance, apart from the fact that a surface wave is used to
measure the absorbed quantity of aroma compound (Culter, 1999).

As reviewed by Schaller et al. (1998), there is another type of metal oxide
semiconductor sensor used in electronic noses commercially, known as a metal
oxide semiconductor field-effect transistor (MOSFET) sensor. A MOSFET sensor
is comprised of three layers: a silicon semiconductor, a silicon oxide insulator,
and a catalytic metal such as palladium, platinum, iridium or rhodium. The
catalytic metal is also called the gate. In the MOSFET transistor, the gate and the
drain contacts are shortcut. The applied voltage on the gate and drain contact
creates an electrical ﬁeld, which alters the conductivity of the transistor. Hence
when polar odor compounds interact with the metal gate, the electric ﬁeld is
modiﬁed, which eventually modiﬁes the current ﬂowing through the sensor.

The electronic nose is an analytical instrument that can recognize ﬂavors,
odors, and volatile compounds. It has many advantages over the subjective
sensory panel evaluation of odors and flavors as it eliminates the factor of
fatigue, inconsistency, and high cost involved in human sensory analysis. An
electronic nose is composed of a chemical sensing system such as a sensor
array and a pattern recognition system. Each sensor is sensitive to a certain

volatile compound and generates a signature or pattern characteristic of the

29

vapor. Different chemicals can be presented to the sensor array, which can be
used to build a database of signatures. Such a database can be used to train the
pattern recognition system of the electronic nose (Giese, 2000).

Culter (1999) pointed out that analytical measurement technique such as
GC-MS can detect individual components in a volatile vapor, but such
components do not necessarily represent the combined sensory effect of the
vapor. Moreover, the trained human subjects are not always available to perform
a sensory analysis.

Schaller et al. (1998) pointed out that the electronic nose has been
successfully used as a quality control tool to evaluate quality of various food
products such as meat, grains, coffee, beer, mushroom, cheese, sugar, ﬁsh,
blueberry, orange juice, cola, and alcoholic beverages. It is also being widely
used to analyze off-ﬂavors in food due to packaging.

Heinio and Ahvenainen (2002) used the E-nose to analyze the taints
caused by pigments of printed solid boards. The objective of the experiment was
to determine the effect of printing inks on the sensory properties of the packaging
material, using the E-nose, which was correlated with human sensory evaluation
and other headspace methods such as GC-MS. Twenty samples were studied,
which included unprinted solid board, lacquered solid board, offset printed solid
board with 14 different colors, and offset printed, lacquered solid boards with 4
colors. The E-nose was found to be very successful in discriminating the different
board samples based on their coloring agents or lacquering. The results also

indicated correlation with the off-ﬂavor perceived during sensory evaluation.

30

Winquist et al. (1993) used the E-nose to study the quality of ground beef
and pork and also estimate storage time in a refrigerator, based on the
organoleptic property of the meat after storage. The electronic nose used
consisted of a gas sensor array combined with a pattern recognition routine.
Samples of ground beef and pork, stored in a refrigerator, were studied. The E-
nose was successful in identifying the type and quality of meat.

Benedetti et al. (2002) explored the use of the electronic nose to study the
shelf life of ripened Taleggio cheese packaged in paper. The electronic nose
used for the study had an array of 10 MOSFET sensors and 12 MOS sensors.
The E-nose was found to effectively classify and discriminate among cheese
samples based on differences in their storage time and temperature. The
different storage times and temperatures inﬂuenced the aroma characteristics of
the cheese, which was sensed by 6 of the 22 sensors, which had good
discrimination power.

Van Deventer and Mallikarjunan (2002) analyzed and compared the
performance of three electronic nose systems as a quality control tool, used for
detecting retained printing solvents in packaging. Three electronic nose systems
were used, which had different sensor technologies — metal oxide
semiconducting sensors, conducting polymer sensors, and quartz microbalance
sensors. Each system was used to test 3 different ﬁlm classes, with varying
retained solvents. It was concluded that the E-nose with conducting polymer

sensor technology had the highest discriminatory power. However, all the

31

electronic noses were found to be capable of discriminating among the film
samples at different levels of retained solvents.

Willing et al. (1998) used an electronic nose for measuring odors from
paperboard, intended for packaging applications. Nine different paperboards
from a wide range of board grades were analyzed in the electronic nose. The
electronic nose was equipped with 10 MOSFET sensors, 4 Tagushi sensors, and
1 carbon dioxide sensor. The partial least squares regression (PLS) method was
used to correlate electronic sensor responses with sensory panel descriptors.
Some electronic sensor responses correlated well with a selected number of
panel descriptors, while others did not ﬁt with any panel descriptors.

Electronic nose technology is still in its development phase, both in
respect to hardware and software development. It still has a few disadvantages.
It cannot provide sufﬁcient quantitative information for certain aroma differences
(Harper, 2001). In addition, the electronic hose system is prone to sensor drift,
which occurs due to aging and degradation of individual sensors. Drift results in
gradual change in output over time without any signiﬁcant change in input. Thus
it hinders the reproducibility of the system. Calibration of sensors and sensor
replacement after a ﬁxed time interval can help in minimizing this problem
(Maneesin, 2001). Moreover, the sensors are sensitive to moisture. Conducting
polymer sensors are more sensitive to moisture than the metal oxide sensors,
which can be minimized by using a ﬁlter and an air conditioning unit (Culter,

1999).

32

CHAPTER II

MATERIALS AND METHODS

This research was directed towards developing a standard experimental
procedure for evaluating food grade HDPE resins using an electronic nose,
correlated with the traditional sensory evaluation technique, for detecting off-
ﬂavor in drinking water. As a part of the procedure, the electronic nose system
was used to analyze the degree of discrimination between eight different grades
of HDPE resin. The experimental run conditions in the electronic nose that gave
the highest degree of differentiation between the resin samples were
standardized. Speciﬁc quantities of resin samples were soaked in ultra-pure
water at 40 : 2° C for 1 week. Water samples stored in contact with the resins
were evaluated for off-ﬂavor by a human consumer sensory panel. The water
samples were organoleptically evaluated at different concentrations and the
resins were ranked based on the degree of off-ﬂavor generation. Finally, the
electronic nose was used to correlate the sensory data and the e-nose sensor
responses for the various resin grades, using multivariate statistical techniques.

The experimental design is summarized in Figure 1.

33

 

I (8 different food grade resins)

I

Electronic Nose
Analysis

I

 

 

 

 

2 g of HDPE resins

(analyzed at three

different incubation
temperatures and times)

 

 

 

I

 

Olfactory response proﬁles
of the resins
(Metal oxide sensor
responses)

 

 

 

I

Establishing E-nose
method (parameters) to
evaluate HDPE resins

 

 

 

 

 

Correlation of e-nose
and sensory results

statistical techniques

 

 

 

HDPE resins I

I
I

Organoleptic Study
(Sensory Evaluation)

I

Resin pellets soaked in

ultra-pure water at 40 3:

2° C for 7 days (ratio —

8:250, 8 g resin in 250
ml water)

I

Water ﬁltered after 7
days (yielding off-
ﬂavored water)

 

 

 

 

 

 

 

 

Off-ﬂavor analysis 1

 

Sensory Evaluation of
water samples at different
concentrations
(Triangle test)
(consumer study)

I

Resin samples
ranked based on the

 

 

 

 

using multivariate ‘

degree of off-ﬂavor
generated (based on
sensory results)

 

 

 

Figure 1 : Experimental Summary (E-nose evaluation of HDPE resins correlated

with sensory results)

34

 

Electronic Nose System

The Fox 3000 electronic nose (E-nose) system (Alpha M.O.S. SA,
Toulouse, France) was used in this study, which has three main components -
Static Headspace Autosampler (H8100), Sensor Array System (Fox 3000), and

controlling computer software (Fox 3000 software). Figure 2 shows the various

components of the E-nose.

 

Syringe (Headspace injectionLI

 

Oven (Headspace generation)

 

 

 

Autosampler Tray

I Sensor Array System I

 

 

 

Figure 2 : Alpha IVIOS Fox 3000 E-nose system components

The static headspace autosampler generates the headspace (vapors)
over the sample being tested. It includes a sample holding tray (capable of
holding a maximum of 64 vials), an oven to heat the sample vial (10 or 20 ml) to
the temperature required for headspace generation, and an agitator to facilitate

homogeneous headspace generation. The temperature for heating the sample

35

(incubation temperature) and the time of heating (incubation time) must be
standardized and controlled for a specific sample, so that the same amount of
sample vapors is generated each time. The quantity of sample loaded in each
vial must also be standardized to ensure injection repeatability. The autosampler
also has a gas syringe (1 to 5 ml) to collect headspace vapor and inject it into the
sensor array system. The gas syringe temperature is maintained slightly above
the sample temperature to avoid any condensation.
The E-nose is based on an array of 12 metal oxide sensors (Figure 3).
The sensors have partial responses to specific volatile compounds due to
reactions with various kinds of molecules and produce olfactory fingerprints of
odors. The sensor names are as follows.
Chamber 1 : SY/LG, SY/G, SY/AA, SY/GH, SY/gCTl, SY/gCT
Chamber 2 : T30/1, P10/1, P10/2, P40/1, T70/2, PA2
Each of the sensors listed above has its own characteristics in terms of
sensitivity and selectivity to different volatile compounds. The sensors have thick
metal oxide (tin or zinc oxide) films deposited onto ceramic tubes, which are

heated to temperatures between 300-500° C.

 

  

I‘ll:

 

 

 

Figure 3 : Metal oxide sensor (Source : Alpha M.O.S. SA, Toulouse, France)

36

Figure 5 depicts the architecture of the gas array sensor system. Metal
oxides are doped with various catalytic metals to shift the sensitivity and
responsiveness of the sensors to particular chemical volatiles. The detection
principle is based on conductivity measurements. The sensor adsorbs oxygen
from the air on its metal oxide layer by trapping free electrons from n-type
semiconductors, producing a highly resistive layer. This adsorbed oxygen is
removed from the surface temporarily, in the presence of volatile reducing
compounds, resulting in an increase of free electrons. This increases conductivity
of the metal oxide ﬁlm. The time-dependent electrical response due to change in

conductivity of the sensor is interpreted as an olfactory fingerprint of an odor

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(Mielle, 1996).
Carrier Gas ' O ’ H' hl _ t'
xygen 19 y resrs Ive
’ I Adsorption > I layer
Adsorption
Volatile reducing , e- e- e- e-
vapors Oxygen e- e- e- e- ’ Highly
De-sorption conductive layer
De-sorption

Figure 4 : Change in resistance of sensors due to action of reducing chemical
vapors

37

These reactions with the chemical volatiles are inﬂuenced by parameters
such as the carrier gas purity, operating temperature, carrier gas ﬂow rate and
humidity, which need to be carefully controlled for rapid response and recovery
time. The sensors operate at high temperature to avoid interference from water.

Table 1 shows the responsiveness of various metal oxide sensors to a
variety of chemical compounds, associated with different applications.

The Fox 3000 software package is used for data acquisition and
processing in addition to driving the E-nose operational steps. The electrical
response proﬁles obtained from the gas sensors are summarized in a data bank
or library. Multivariate statistical data analysis methods such as Principal
Component Analysis (PCA), Discriminant Function Analysis (DFA), Soft
Independent Modeling of Class Analogy (SIMCA), Partial Least Squares (PLS),
and Statistical Process Control (SPC) are used for sample group identiﬁcation,

discrimination, and sensory score correlation.

38

Table 1 : Metal oxide sensors used in Fox E-nose systems (Maneesin, 2001)

 

Metal Oxide Sensors

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Description Application P Type T Type SX Type SY Type
SX13
Cooking, SX13p
Non polar 3322:2313 roasting of P10/1 T10/1 sx14 SY/WM
volatile ro a ne coffee, petro— P10/2 T10/4 SX14d SY/CT
p p chemical SX15
SX42
Milk industry,
Hydrogenate :ydg’ge” ;°°°h P10/9 T10/9 sx
d Molecules °“ '"9 '9? "335' . PA3 TA3 21
Aldehydes animal ranCld
odors
Polar Liquors,
Organic compounds beers P30“ T30“
Alcoholic 8X22 SYNVC
Solvents $133221: perfumes, PA2 TA1
fermentation
Alcohol & Paint & P70 ,0 1.778,]?-
aromatic polymer P70/1 T70 /2_
A t' compounds industry 70/3
roma IC
Compounds Smoke 8X23 SY/GC
detection
Hydrogen PA3
bonding
Amlnes and Meat and
amlne . ﬁsh
°°”t°'”'”% freshness ””2,” T50/3 3x24 SY/GA
and sulphur derivatives Envrronment
Environment,
inﬁrm THT in 3:3? T50l1 3x25
p butane y
Sensor for
. ﬂuorinated 8 Environment SX30
ELISE: and chlorinated packaging, I343; 128;; SX31 SY/LG
compounds, TCA 8X32
aldehydes
. Alcohol Petrochemic T70/2
Cooking compounds al T70/3 8X82
control (food Natural
aroma & Food aroma aroma P70/0 T70/0 8X83
volatile and volatile ’ P70/1 T70/1 SX83T
coffee
Tobacco Environment
. . smoke and and SXOO
gang?” quality of air applications PA3 $2; SX80
carburant in smoke 8X84
vapor detection

 

 

39

 

 

 

 

 

 

F-II Baseline (synthetic air) I

Irv -
0 I4way valve I

 

 

Odor or organo-volatile
compounds I

 

 

 

 

 

 

 

chambers

 

Metal oxide sensorJ

1o 11 12

 

 

 

 

IrFlow-meter I__,> I Pump I

 

I Exhaust I

 

Figure 5 : Architecture of gas array sensor system (Source : Alpha M.O.S. SA,
Toulouse, France)

40

E-nose Operation Principle

The basic operating principle of the electronic nose system involves
generation of headspace over the sample being tested using heat and agitation,
presentation of the headspace vapor to the sensor array system, and recording
and analysis of the sensor response data using artiﬁcial intelligence or statistical
data processing (Culter, 1999). Using pattern recognition software and previous
human training of the E-nose, the system predicts the most likely response to the
new odor ﬁngerprint pattern, in the form of qualitative or quantitative information.
The E-nose measures the aroma or odor in a way analogous to humans.
However, it does not analyze or measure the components of an odor. Each
chemical sensor is similar to an olfactory receptor of a human nose. The pattern
recognition software used by the e-nose is analogous to the human olfactory
system, which can discriminate and memorize odor responses in the cerebral
cortex of the brain (Bartlett et al. 1997).

Figure6 represents the principle of the e-nose system compared to the

human nose.

41

 

 

 

 

 

 

 

 

 

 

 

 

 

ARES...— .om:o_:o._. .<m .m.0.5_ «:26. H 8.509 E2m>m omoz oEobooﬁ ho o_&o:_..n_ " o 2:9".
l . m
Enchant
coEcmouwm 9d 65% wow coamom
I... Egan. comcmw
sesame“... ,. Masseuse
coeo sen.

 

 

:oSoZ

 

Emcm

 

    

 

EOUO

 

 

owtoO
:m___Nm._m

   

 

 

{31.
.I.

 

42

 

 

Experimental Procedure
E-Nose procedure

In this study, the Fox 3000 E-nose system was used to analyze the
olfactory profiles of eight different high-density polyethylene (HDPE) resin grades
and to correlate them to the data obtained from human sensory evaluation of the
drinking water samples, stored in contact with these resins. The eight different
HDPE resin grades were AP, BP, CP, DX, EX, FX, GX, and HX. The first three
grades were procured from a particular resin manufacturer and the remaining ﬁve
were from a different manufacturer. A ﬁxed quantity of each resin (29) was
weighed into a 10 ml glass vial and sealed. Replicates were made for each resin
grade sample. The samples were loaded in the autosampler tray of the E-nose
and the E-nose operation cycle was activated using the software. During the
cycle, each vial was automatically transferred to the oven and agitator, to
generate the headspace volatiles. The headspace volatiles were collected from
the heated vial using a syringe and injected into the sensor array chamber, to
generate the olfactory response profiles. Three experimental runs were
conducted with variations in system parameters (Table 2).

Twelve sensor responses were obtained for the injected volatile. A pre-
processing response plot was generated by the software between AR/Ro and
time (3). AR = (R0 — R)/Ro, where RC) is the resistance at t=0 and R is the
resistance at the selected time. Between each sample volatile injection, the gas
syringe was thoroughly ﬂushed by carrier gas to avoid cross-contamination. The
data obtained for the various replicates of the resin samples were analyzed by

multivariate statistical treatments such as PCA and PLS to understand the

43

degree of sample discrimination and correlation with the sensory scores of the
samples. The experimental run conditions described in Table 2, which gave the
highest discrimination index and a good correlation with the sensory results, were
standardized for the HDPE resin grades. The response ﬁngerprint pattern was
also studied to understand the efﬁciency of the e-nose in discriminating various

resin grades, which can be used as a valuable quality control tool.

Table 2 : System conditions used for different experimental runs

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

RUN

System Parameters I II III
Sample replicates 6 6
Sample quantity (9) x* x
Incubation time (sec) 600 1200 2400
Incubation temperature (deg) 80 90 x
Agitation speed (rpm) 500 x x
Syringe type (ml) 5 x x
Syringe ﬁll speed (pl/sec) 500 x x
Syringe temperature (deg) 85 95 x
Flushing time (sec) 180 x x
Vial type (ml) 10 x x
Injection volume (ul) 2000 x x
Injection speed (pl/sec) 2000 x x
Acquisition time (sec) 120 x x
Acquisition period (sec) 0.5 x x
Delay (sec) 900 1080 900
Flow (ml/min) 150 x x

 

 

 

 

x* : same value as previous run

44

Sensory Evaluation Procedure

The procedure for the sensory study was formulated based on guidelines
obtained from Alpha MOS. An untrained consumer panel (Appendix C) was
employed to detect off-ﬂavor in drinking water samples, stored in contact with the

resins.

Sample preparation

All the 8 different grades of HDPE resin pellets were soaked in ultra-pure
water (in the ratio of 8:250, 8 grams of resin in 250 ml. water) for a period of one
week at 40 i 2° C, in an environmental chamber. After 7 days, the stored water
was vigorously agitated and ﬁltered, to yield the 100% concentrated off-ﬂavored

water sample pertaining to each resin grade (Appendix D).

Experimental plan and design

The water samples, prepared by the above method, were presented to the
untrained consumer panel for detection of off-ﬂavor (Appendix A & B). The
triangle difference test method was used, which involved determination of the
odd sample in a set of three given samples. Any two given samples out of three
in a triangle test are identical. This method is highly useful to determine whether
an overall difference exists, which could be a result of change in ingredients,
processing, packaging, or storage (Meilgaard et al. 1991). The detailed

experimental layout is summarized in Figure 7.

45

 

 

Sensory evaluation
100 % conc. off-ﬂavored sample
(off-ﬂavor detection in water)

 

Identify samples Sensory
QIVIHQ off-ﬂavor evaluation
> 66 % conc. off-

 

 

 

 

8 HDPE
resin samples
(each soaked in ultra -
pure water)

 

 

 

ﬂavored sample

 

 

 

Identify samples
giving off-ﬂavor

 

Sensory
evaluation
90 % conc. off—
ﬂavored sample

 

 

 

Identify samples
giving off-ﬂavor

 

Sensory
evaluation
33 % conc. off-
ﬂavored sample

 

 

 

 

Rank samples in
order of off-ﬂavor
generation

 

 

 

Figure 7 : Sensory evaluation experimental layout

46

The sample presentation was based on a randomized plan. Each subject was
presented with a set of four sample sets (each being a triangle). The number of
sample sets was restricted to four, in order to minimize sensory fatigue. A
randomized pattern was followed for sample presentation in the triangle test, to
eliminate any factor of bias during the sensory evaluation. Each sample was
presented the same number of times to various human subjects (students,
faculty, and staff) during each sensory experiment, which made the design quite
balanced. Table 3 illustrates a section of the typical sample presentation design
plan that was used.

Table 3 : Sequence of sample presentation to the consumer panel

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

51:?“ I II III IV

1 W8 7 W88 8 SW8 2 WSW 3
2 W88 8 WSW 5 SSW 6 W88 3
3 SW 7 SW 1 WSW 3 SW8 6
4 W3 7 SW 4 SSW 1 SW8 3
5 W88 2 WSW 4 SW8 8 W8 6
6 SW8 7 SW 2 WSW 3 SSW 5
7 wss 1 wsw ' 5 wss 4 wss 8
8 W8 4 W88 8 WSW 5 SSW 3
9 SSW 3 W8 6 W8 5 SW 2
10 W8 5 W88 8 SSW 7 SW8 3
W — Puriﬁed reference water

8 - Sample (off-ﬂavored) water (corresponding to the resin sample

number in the subsequent column)

47

 

Based on the above design plan, it can be observed that the position in
which a particular resin was offered to subjects was varied. For instance, Sample
8 was offered to subject 1 in the second position, subject 2 in the ﬁrst position,
subject 5 in the third position and subject 7 in the fourth position. Another
important feature of this design was the presentation of a resin in different ways
within the triangular pattern. For instance, Sample 8 was presented to subject 1
as WSS (pure water sample followed by two off-ﬂavored samples) whereas it
was presented to subject 5 as SWS (pure water sample in between two off-ﬂavor
samples). Three digit codes were used to codify the various samples so that no
information could be derived by the subjects about the nature of samples or
pattern of sample presentation (Appendix E).

The 100% off-flavored samples for all the eight different resin grades were
presented to a consumer panel of 96 subjects, for taste evaluation. Each sample
was presented 48 times according to the design plan. The samples that were
found to have the most off-ﬂavor were identiﬁed, based on the number of correct
responses by the consumer panel, at a signiﬁcance level of 5%. The triangle
difference test method is statistically more efﬁcient than the other sensory
difference test methods eg. paired comparison and duo-trio. The off-ﬂavored
water samples, which were differentiated by the consumer panel at 100%
concentration, were further diluted to 66% concentration and presented to
another untrained consumer panel, comprising of 90 human subjects. Each
sample appeared 72 times in the experimental design. None of the samples were

detected for off-ﬂavor at 66%. Hence a higher concentration of 90% was

48

prepared for the same set of off-ﬂavored samples (detected initially at 100%) and
presented to another consumer panel of 90 subjects, where each sample was
presented 72 times, similar to the earlier presentation pattern. The samples
having the most off-ﬂavor were identiﬁed in a similar way.

Based on all consumer panel responses, the resins were categorized into
three broad groups, depending on the severity of off-ﬂavor generated. Sensory
scores were assigned to the groups of resins, which were eventually used for
correlation with the e-nose profiles. A high score was assigned to the resin

producing the maximum off-ﬂavor.

Correlation of E-nose and sensory results

Multivariate statistical techniques were used for analysis of the E-nose
olfactory response data. The degree of discrimination between different resin
grades was studied using principle component analysis (PCA). The correlation
between the E-nose and the sensory results was determined using the partial

least square technique (PLS).

Multivariate Statistical Data Analysis

Principal component analysis (PCA) was used to study the similarity or
dissimilarity between the resin samples as well as to understand the relationship
between the variables (E-nose sensor responses). In other words, PCA involves
recognizing patterns of association in multivariate data sets. When PCA is
applied to a data set, the original variables (which are the E-nose sensor

responses in this case) are mathematically converted to a new set of variables,

49

called components (Figure 8). Each component is expressed as a linear

combination of the original variables.

 

Variable 2
‘0
9

    

 

 

 

 

 

Figure 8 : Concept of PCA

The ﬁrst principle component (PC1) explains the maximum amount of
variation possible in one direction, for a given data set. Thus, PC1 contains the
maximum amount of information. The second principle component (PC2) is
orthogonal to PC1 and explains the maximum amount of remaining variation. The
degree of discrimination indicates how well the sensor responses are able to
distinguish between the different resin grades, based on their olfactory
fingerprints. A high degree of discrimination would imply that the E-nose is
capable and efﬁcient in discriminating the various HDPE resin samples and such
a procedure can be validated and standardized for similar applications.

The partial least squares (PLS) method was used to correlate the E-nose
sensor responses of the different resin grades with the sensory analysis results.

PLS is based on the linear regression technique, which is used to extract

50

quantitative information. In this case, it was used to build a model that can predict
the sensory panel score for any unknown HDPE resin grade, based on its E-nose
sensor responses. Y is the matrix containing quantitative measurements,
whereas Y’ is the matrix containing the predictive values and X is the matrix built
with detector (metal oxide sensor) measurements. The PLS model generates a B
matrix that minimizes the distance between Y and Y’ with Y’ = XB. The B matrix
is used to predict quantitative information (the sensory panel score) for an
unknown sample. The measurement matrix is multiplied by B to obtain the

prediction (Alpha MOS Fox 3000 Manual).

51

Chapter III

RESULTS AND DISCUSSION

Evaluation of food grade HDPE resins using the E-nose system
(Standardizing E-nose run parameters)

Four replicates of eight different grades of HDPE resins were analyzed by
the E-nose with Run l condition and six replicates with Run II and Ill conditions
(Table 2). The primary idea was to establish standard run conditions, suitable to
analyze such food grade HDPE resins. Principle Component Analysis (PCA) was
performed on the olfactory responses of the E-nose sensors, to understand its
capability to differentiate between the resin samples, based on the degree of
discrimination. Twelve metal oxide sensors were used in the E-nose system. A
set of twelve sensor responses was generated for each sample. PCA reduces
the factor of variability between various sensor responses by a linear
combination of the responses. The location of a resin in a two-dimensional PCA
plot would give an idea about how different or similar it is, in comparison to other
resins.

The E-nose was found to be very efficient in discriminating the resin
samples, which is evident from the high discrimination indices (Run l — 93%, Run
II — 88% and Run lll - 88%). The high discrimination percentages in PCA proﬁles
of the sensor responses indicated that the E-nose was successful in
distinguishing between the constituent volatile components present in the various
resin grades. Figure 9 (a&b) show the PCA results for Run 1 conditions. Figure 10

and 11 show the PCA results for Run II and Ill conditions, respectively.

52

omocm >9 moEEmm Ewe. w mcoEm cosmEEtomE 13:36:00 _ cam: <00 ” Ev m 0.59“.

 

e\e®m.om U v0

 

 

o3 om .o om .o o E 8 .o o to. 8.0. on .o- 26. B .o. 8.9 and. 8 .o. om .o. 8. F. o F. P. om. F. mm. P.
L _ . a _. _ a _ _ _ . a _ L. -_.l .. _ oomd.
W Vi . . . . . . W

E A. _
s _ _
_.

. . _. _ _ . . . .
_. . . .r m , M. H _ _ H. .. _, . . o
_. . _ _ .. . . . . A , . . . __ , . _ . 1-183.
. .. . . n m .. . . .. _ . .
n _ H W .. u , .H .n a . H E W _.

. p _. . L . . ., ..
~ I m E E , .. v , V ‘ n _ .

.E
_ u . . . a.

L. - *-4._..< -

 

a _ J . , . , _ . , . . . . .
H .XI . . _ .. . . . , _ . . low—b.
,. I. . . ., , N. s ,. _. U H . .. . . .
.. w ,. H, V u , ,. . _ H . ﬂ . H
. w 2 W s m _. W .. ._ H H w E w _ _
. H _ w . , ,_ _ H .
. . .. . . H _ . iogo.
. . . . . ., N A .
. . . z
z . N A . A. _ . z
.. L , W h. . M _ . . n . w
. H . . .. m L , q. A . logo.
. , z , . A . t .
_. w M u m _ m .4 .
_ .. M M . .. u .
. ,. . , , 183.
., . W . .
,. w u _ v w
s z . s

.
W
. ,ﬂ . a , m .
. . , a m m . . w a s m H .H .
_. W m c c .m M . M m c H - r83.
M , w W H . H M .__, a .
_
_
_
A
_
_

 

961881 33

 

 

 

. .s ,, . m H
s. _ m w m _ _ m _ . . .
W U. .. H .H _. U _

. _ E .
u. M W m . _ . - rowan.
w. . W m . .
. .. m m m M . -939.
.. ., c. E ., H z. . M
_. .w . . n u z . . .
a w _ .2 M , W . r82.
.H E _ W W L
w A W M . M H m . areas
an .m m . w W A W m w , ”A .
m m m ., m m. m . H M
H n a . w M .83
a .. . L s . s . .
. m w E s M W W w
.. . H E M L — .2 H m H
. m . _ . H s _. .
m H H A W m . W .M ,Wt-roopo
" ,.:;,..*..1:". : ﬂ. . m . “ ..M.- .w.. u m .m UltTDNFO
m H ._. .. w _ .u _ E H
E .. . .e ﬂ 2 L _ W W _ _.l #26

 

 

 

 

mm r x005 coszEtomE

 

 

 

 

 

53

 

 

.6 .XM .n_< .xo e858 8855:er eooo - 83m 9:5 E 8283 8E8 85:85 l «on. n 3. a 2:9".

 

 

somemneo

ewe Ra Ra R: 3d 3d :d ﬂu 3d 36 Ba 3d Ba Ba mﬁwsq
, , -meod.

m , N. M _ A W W , w lose
, _ -mOod.
-Doad
. -mooe
. m . W A. W . W, a . W W -oeod

rmrmd

%I£'€ 1 33

named
x0

 

rmmod
tamed

.w. -mmad

 

x O f) 3,2 .. . .
. x, . W .e -Deod
. , . a<

m . , _ . e . , . . . A , -meod
w M . _ ex. . . . A -omod
z , , -Nmod
_ mm - chE coszEtowE #

 

 

 

 

54

 

mcozﬁcoo _ :31 can“ coumEEtowE Co 00600 .033 l AmcoEucoo = :3: (on. u 2. 0.59“.

$wm.mm H PO
mi. 96 . 5 mod. o3. m3. omd. 95. £3.

 

3
Z
.3
9
7w
%

 

mm - x005 coszEtowE

 

55

0:050:00 _ cam :05 coszEtomE “6 005.00 .0264 - 6:036:00 :. any: <00 ” 5 0.59".

 

$on©mnr0

 

mmd Q+o cud omd DVO cm¢ awc omd cad owe. owe- amd. med. omd. mmd.

 

_ _ _ L . u w u -33.
M W . . _ . -83
. my -88.

lowed.

 

lowed.

laced

tamed

9696 V ZS

as age

lemma

 

 

lowed

H H . W . m W w W W W laced

”New:

 

 

 

 

 

 

 

ww-x0c:_:o:0::::o£o

 

56

 

Based on the PCA proﬁles, it can be seen that Run l conditions gave the
highest discrimination index, Le. 93%. The high discrimination percentage
validated the experimental conditions used for Run 1, which can be standardized
as the most ideal set of E-nose conditions, to analyze such food grade HDPE
resins. These conditions provided higher E-nose sensor responses, possibly due
to higher generation of volatile compound vapors. However, the degree of
discrimination was signiﬁcantly high even in Runs II and III, which reiterates the
fact that the E-nose is very effective in distinguishing between the olfactory
ﬁngerprints of such food grade HDPE resins.

PCA was also performed on the sensor responses of the groups of HDPE
resins obtained from each source. As mentioned earlier, grades AP, BP, and CP
were procured from one manufacturer whereas DX, EX, FX, GX, and HX were
procured from another manufacturer. Figure 12, 13 & 14 show the PCA plots for
resins AP, BP, and CP, based on Run I, II, and Ill conditions, respectively. It can
be seen even in this case that Run 1 conditions gave the highest discrimination

percentage (96%) for the resin set. Run II and Ill gave only 87% and 75%.

57

00 .0m .0< 0500: c0050: :o=0:_E:ow_u Co 00600 :9... I 3:036:00 _ :35 <00 “ «r 0.520

 

 

$5w.mm ” F0
Ed.

 

95.

 

mm - x005 :o:0:_E_._ow_o

am .a.

Still-0123

 

 

58

0:050:00 _ :sm :9: :osmEEtowﬁ 96 00500 .026. I 00 .0m .0< 0500: l 0:050:00 = :35 <00 H 2 0.590

 

$2.3 H .0
mm: 949 mm: :2. m d a .a mad 8.: mad. mg. mg. END- mmd. and.

 

 

 

 

 

00 c

 

 

 

 

 

A

In

8. .

0
96902130

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

F 5 - x005 :o_ﬁ:_E:o0_o _

 

 

59

0:036:00 _ :30 :93 :030:_E:006 6 00506 .032 - 00 .0m .0< 0500: I 0:036:00 =_ :33 <00 " 3 0.590

 

 

$2.2 H 30
m S a 8 mad Q .a mod. n S. m S. 03.
_ r83.

 

.l.... ...- .l I. -.. II. .- :.l.-l. .I-l..l: ..l..-. .. .. - .. .- i. I... .i-.-.I-.l.-IIIII.- I -.....l...-ll -I-..-Il..- . . ...I 11.. l.- .l I... . IDmOD.

. . -.l. .. .l .l I I l I II II I I I l l I I - ...- .. -...- .- l-...l. ...l.-... IDmD.Dr
.runnnumm
- -- . .... .-. ..... It 1 ll I I l I l IDNDD.

-23.

 

%88'I.I 3 ZS

-. ...r . .. .. ..- .- I .. ...r. ..lI-... .. .. l. ..l. r .. . . -. . ..l .. . . ..... .. .. . l .. .. 1 IOWDD
I... -l.. I. .-......-....- ..... -.-.l.-...l . . . .. . .-. . -.. .- -. . .. ..r... .I... . .- I..:. .. I.. . .. . . .-.-..-..:-. .-.. IDVOD

- .- . : - . 183

.l. . .. . l. .-l. . .. ..... ..ﬁIﬁllIVW-Mmﬂd/J/W : - -. . .. .. lowed
IJICIIIIIO

. - . ..-DBD

 

 

 

 

 

 

 

 

Iowa .3

 

 

_ mn - x065 :030:_E:00_o

 

60

 

A similar analysis was performed for the resin grades DX, EX, FX, GX,
and HX, which were procured from a different manufacturer. It can be seen in
this case that Run 1 gave the highest discrimination index (96%) whereas Run II
and III gave 91% and 93% respectively (Figure 15, 16 & 17). Hence, E-nose Run
l experimental conditions can be standardized for food grade HDPE resins, since
the highest degree of discrimination was achieved in all the different cases. The
Incubation temperature and incubation time in Run 1 conditions were ideal to

generate the right concentration of headspace vapors for the resins.

61

x: .5 .x“. .00 .xo 050.2 5052 8:055:86 3o 090% :02 I 08328 _ as”: <00 ” me 050:.

 

 

mod

e\emo.vw H r0

...—No. mm .3.

am.

 

8.0 8.3- s S.

 

..e
s.\ x=.\_
H x0
7r .
. $.07.
c as”..-
>\ 50
W .... M

 

Dr

18 to.

180.3

I m8 .3

1030.0

Immod

loo—.3

. .lmmrd

 

 

 

 

_ om - x065 :030:_E:00_o

ram 3

rmncd.

lemma.

Immad.

9698'” 1 30

 

 

62

0:036:00 _ :30 :05 :030:_E3006 6 00506 0026-. I XI .x0 .00 .xm JR. 0500: l 0:036:00 __ :3”: <00 ” or 0590

 

$5M? r0

 

loo—d.

 

 

Dad 0mg omd 0N0 ems men 030 mad Dec 030. are. mpd. ems. 0mg. 0mg. mma. Dec. 030. emu.

.. M W . W H m . . H

a w w M M W W M W M . _ -meod.

x: . m n H m . .

U . m .. . see

0 W M W .2 . -Deod

M W W W W -emoond

M m A n m W W . W W W m . -oeqo,s

W s H .. m m w H 0 . E w H. . W
mm

w M . 0 M . m ., m . .0 m H W W . W -00od

roma .o

 

w W W W , W -ee_d

 

3w;

 

m w M W . m . -0430

 

 

 

 

 

w m m
m . . m . m -0030
w m . m . . . . a W m . w . m -0036
w a m .. _ . m 0 ﬂ ..

c . n m . . . -eeme

 

Fm - x065 :030:_E300_o #

 

 

 

63

085:8 _ es: 85 83056506 .033 I x: .6 .x“. .x0 .x0 0509 108328 z. 5:0 <00 ” t 2:9...

 

e\emm.®m H v0

3.3. 3.3. cm

1
.. .

.0. mm .3. cm.

.3.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1

 

 

 

 

-
t

0...... --.”. -._..A. -

 

 

.
E
m
_.

m

. 2..

 

 

 

 

3
0
0
r
0
..
0 m
. _.
._ M
..., ﬂ It
._ H _
z. m .0
3 .
j z .
n _ _
M a .
w .. .
_ h H
n m .
0 E .
. M _
0 M H
0 m .
E . .
z 0 .n
.w m ._
0 . w
0 0 z
. W
H— . v
.. m .u
_ . n
0 . _ ,
s.
r.. . .Inq ................
0 H
. .0
E n
E n
W .
_
. f
.. w .
u . m
E s
m . N
a . H.
.. a m
M. w m
I |.. .\II.I.-.l.... . Ir. r. W .....t
m u
m
,

 

 

m.w
0::

 

 

 

cm...o med 3.3 mmo omd mmd omd mg a; 83 8.3 8.3. 9.3. m3. DNQ mmdr Qua. £3. a

.
w
E
.2
M

a
N
m
2
.0

 

 

 

_. h
m ..
_ n
. .
. _
w ._
. A
m ..
. ., ._
0 0
., .
. M
_ v
. .
. ..
.u z
_ E
u L.
.

_ .
_ .
u

_

_

n

_

.

 

 

192.9

rch.

tamed.

. Immod.

., 1900.0

138.0

romeo

. -00;

.. 132......

-mNS

 

 

 

 

mm .. x005 coszEtOwﬁ—

 

1902...

96093 3 33

 

 

 

 

64

E-nose sensor response comparison

A comparative study was done with the olfactory sensor proﬁles of Resin
CP, which was understood to have a distinctively differentiating odor response,
as evident from the PCA proﬁles. The location of Resin CP in the PCA map was
found be signiﬁcantly isolated from the other resin grades. This indicated the fact
that the volatile composition of this grade was signiﬁcantly different than the other
grades. Figure 18 shows the comparison between sensor responses of Resin

CP, generated by Run l and II conditions.

 

 

 

0.3:

0.5-

0.4-

 

0.2-

0.0- ._ ..

 

 

 

 

Figure 18 : Comparison of sensor responses (Resin CP) — between Run 1 & ll
Run l — Incubation time : 600 sec; Incubation temperature : 80° C
Run II — Incubation time : 1200 sec; Incubation temperature : 90°C

65

 

The proﬁles shown in Figure 18 are the pre-processing plots between
AR/Ro and time (s). AR = (R0 - R)/R°, where Ro is the resistance at t=0 and R is
the resistance at the selected time. It can be seen that the sensor responses in
Run l conditions are higher than those generated by Run II conditions. The

responses can be visualized in a different format (Figure 19).

 

 

[15- RUN I . RUN II

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

-5'.D D'tl 5'0 100 15.0 211.0 25'.D

 

 

 

Figure 19 : Comparison of sensor responses (Resin CP) — between Run 1 8. ll
Run 1 - Incubation time : 600 sec; Incubation temperature : 80° C
Run II - Incubation time : 1200 sec; Incubation temperature : 90°C

66

A similar comparison can be made between the sensor proﬁles of Resin CP,

generated by Run l and Run Ill conditions (Figure 20 & 21).

 

 

0.8:

RUN I , RUN III
l'fP —.

 

 

 

 

 

 

 

Figure 20 : Comparison of sensor responses (Resin CP) - between Run 1 & Ill
Run l - Incubation time : 600 sec; Incubation temperature : 80° C
Run Ill — Incubation time : 2400 sec; Incubation temperature : 90° C

67

 

 

0.8-
0.8-
0.4—
1125

l

RUN I RUN III

13.0—
D2—
{1.4—

'U-B—I I I I I I l
-5.U 0.0 5.0 IUD 15.0 20.0 25.0

 

 

 

 

Figure 21 : Comparison of sensor responses (Resin CP) - between Run | & III
Run l— Incubation time : 600 sec; Incubation temperature : 80°C
Run Ill — Incubation time : 2400 sec; Incubation temperature : 90°C

It can be observed even in this case that the sensor response generated
by Run | conditions was higher than the one generated by Run Ill conditions.
Thus it can be seen that the sensor response generated using Run l conditions
was the highest of the three. This reiterates the validation of the Run |
experimental conditions, which are ideal for analysis of food grade HDPE resin
grades. The higher sensor response was attributed to a higher concentration of
volatiles generated in the headspace, which eventually led to a higher degree of
discrimination between the resin grades, as evident from the PCA profiles. This
also indicates the reliability and effectiveness of the E-nose system in

differentiating between the resin samples, based on their volatile compositions.

68

Analysis of off-ﬂavor in water (stored in contact with resin samples)
Sensory Evaluation

Odor-threshold, molecular weight, and polarity are three main factors that
contribute to the release of off-ﬂavor compounds from HDPE (Ho et al. 1994).
The lower the odor threshold, the easier it is to detect off-ﬂavor for the consumer.
Compounds having very low odor thresholds may cause off-ﬂavor problems in
water, even in trace amounts. Typically, the compounds that migrate into the
food product, which is drinking water in this case, have low molecular weight.
The lower the molecular weight, the higher is the volatility of the odor
compounds. Hence such compounds can desorb from the polymer matrix to the
gas phase at a faster rate than the other high volatile compounds, which may
further migrate into the water sample. Polar compounds have a higher tendency
to escape out of a non-polar matrix. Hence low molecular weight polar volatiles
can easily migrate into water and cause off-ﬂavor problems (Maneesin, 2001).
The volatiles generated in the I-IDPE resin sample headspace, by the E-nose,
can be attributed to release of such lower molecular weight polar compounds
from the polymer matrix.

An organoleptic study was conducted on drinking water samples that were
stored in contact with the eight different grades of HDPE resins at 40 1: 2° C for 1
week. The idea was to rank the resin grades based on the degree of off-ﬂavor
generated. The triangle difference test method was followed, using an untrained
consumer panel.

As a part of the sensory evaluation procedure, water samples were initially

presented to the consumer panel at a 100% concentration level. Five out of eight

69

samples (which were stored in contact with the respective resin grades) were
found to having undesirable off-ﬂavor. The off-ﬂavored water samples were
further diluted to 66% concentration and presented to another untrained
consumer panel. None of the samples were detected for off-ﬂavor at 66%. Hence
a higher concentration of 90% was prepared for these ﬁve off-flavored samples
(detected initially at 100%) and presented to another consumer panel. Resin
grade CP was the only resin detected for off-ﬂavor at the 90°/o concentration
level. Table 4 shows the consumer panel response in detecting off-ﬂavor at

different concentrations of water.

Table 4: Consumer panel response for off-ﬂavor in different
concentrations at signiﬁcance level of 5% (probability, p s 0.05)

water

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Resin 100% conc. 66% conc. 90% conc.
_grades Correct Incorrect Correct Incorrect Correct Incorrect

AP M22 _ " 26 26 46 28 44

BP 16 32 - - - -

CP 22 26 24 48 36 36

DX 25 . 23 22 50 27 45

EX 18 30 - - - -

FX 30 18 24 48 28 44

GX 27 21 19 53 27 45

HX 10 38 - - -
Number of subjects = 96 Number of subjects = 90 Number of subjects = 90
Each sample presented Each sample presented 72 Each sample presented

48 times in this plan times in this plan 72 times in this plan

 

As seen in Table 4, resins AP, CP, DX, FX, and GX were detected for off-

ﬂavor at 100% concentration (shaded), and were further presented at 66%

70

 

concentration to a later panel. No off-ﬂavor was detected for any of the samples.
Off-ﬂavor was detected for the water stored in contact with resin grade CP at
90°/o concentration. The minimum required number of correct responses for
number of respondents, n = 48, is 22 at the 5% signiﬁcance level. Similarly, the
minimum required number of correct responses for number of respondents, n =
72, is 32 at the 5% signiﬁcance level (Meilgaard et al. 1991, Table T7).

Based on the consumer panel responses, the resins were categorized into
3 broad groups, depending on the severity of off-flavor generated. Sensory
scores were assigned to the groups of resins, which were eventually used for
correlation with the e-nose proﬁles. A high score was assigned to the resin
producing the maximum off-ﬂavor. Table 5 shows the sensory scores assigned to

different groups of resins.

Table 5: Sensory scores (resins categorized into 3 groups, based on the
sensory panel response) — A high score was assigned to the resin
that generated the most off-ﬂavor in water

 

 

 

 

 

 

Scores
1 2 3
BP AP CP
EX GX
I-IX DX
FX

 

 

 

 

 

71

Correlation of E-nose and sensory responses

The olfactory data generated by the E-nose for all the resin samples were
correlated with the sensory panel results using the Partial Least Squares (PLS)
linear regression model. Figure 22 (a&b) shows the correlation between the
expected (based on actual consumer panel response) and predicted values
(based on the e-nose sensor responses), for the resin olfactory data obtained
using Run l conditions. Any scoring scale can be used for the sensory data,
provided the difference between the assigned scores remains the same. This is
evident from Figure 22 (a) where scores 1, 2 & 3 were assigned and Fig 22 (b)
where 1, 3 8- 5 were assigned. The correlation percentage did not change in both

cases.

72

8N; I 00500 00:9003 0:036:00 _ :00 :03 00:_0> 60306000 6:0 00300000 :002300 530.0300 u 00 «N 0590

 

99m omm cod

003000xw
8m 8 .m SN 5mm Ed R .F cm: 3.0
- coma

 

 

   

S.— ome Om: 2.0
W

    
 

 

8m:

 

 

I 8:

 

 

cam:

183.— .

m
Paloma-Id

 

Bow

 

 

8m .N

00

 

 

comm

 

 

 

 

 

 

 

 

ﬂ . .. . . . . . -800
-00“
_ ”40005.0“:0306300

 

 

 

 

 

 

 

 

 

 

73

6.0; I 00000 00:90.08 0:036:00 _ :00 :03 00:_0> 0030605 0:0 00300000 :002300 :030_0:00 ” 3. Na 0590

 

80800
8.0 000 8.0 a: 8.4 9; 8.4 8.4 8.0 8.0 9.0 .00 8.2.. 80 80 $0 8.0 80 em: 80 9: 8.3 of 8.9 owe
lommd

- oomd
.. .- 1 .1-2.: -. . - .. . ... ..... . - . L ..- Iommd

..--III. .-I.. -.--l- 111-21111: --.1--I1--:-.-1111 ---I-l III\M /( com;
...-1.. I. ...-1.1.... i l... --l. -.l-II-I..-..-I-;- l1... -.-- -. .1. . .1 .-.-:- -\ llama
&/ x:

Tl. 1.1. .- 1-...-. .-.-l 1 I --.---.--.l :11, - ..--1 .---i111- . / Pox-a

I- I --.-I .-.11-..l1.-Ill...--.-- .1. III. II. --11- . ...-1.- -1---.. ....... - ;:-:I\1\ . . I. --.-21-:- -..-m-mIIDDDN

 

 

 

 

 

Ir III-.11 r- l IIITII

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.

I I--- l- I - - I .....-.---.1- .l..--...--.-11-1Il.;:-.Iwum1l-if - -IIDmN N
.11.11- I - -1 .- - ...... --.-1 -. ---1- ,- --1-- -.--l ----... .11. NM com. N m
'0
I II 1 I \«I I .u Leat|. . .2 I. il 0 \Q. \ I ‘ Ila-arr I 34... II I..ic..l\.-... .rfcfll‘v. ..1» 33“....7 ..II-I :0!!th ...-ii! . 113-... E... ‘11VI...III,I\I. {it'll-IL 5.1.4) 'DmN N m

I1 I... - _ I _ .1 1_.:-..-.1 “I . _ 1..- ...I - _I..- I--1-t..~111 1.1-- IDDD m

- -1- 1.1820

l- ..... . -;-. -.----.1.I-\.0-111111-..-11.---1-1 -.---.----1-1---..-.11..‘-..---- .1...-.... 1800

I III...II.I1IIII llLr.1 .11II.11..1 ..I. 1 \k .IIAr.-IIII-l. LTIII-IL.IIIIII :1111.11-..11I... a ......... . l... IIIIIIL-IIIII..11IIIIerIIl III .l1r..lv..I'DQDv
IIIIIFIIIII . . Iii-IIIIIIIII. . \. . -..I.-......-..I.I.IIII .I.FIIII. .III..II .. 1 lrI.1II.lL IIII. 1 1 .II II; .r. l. .IIIII...I III! leNov

If :..---...\ -..--. 1I1l --.-11:13.- Is... III! I. III-II -.I: II- -..lloom.v

1...--11. . Ill-II- III-1111.111 Poke
Iooo.m

vmmmnwd H 0030.00.00

 

 

 

 

 

 

I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

74

As is evident from Figure 22 (a&b), a good correlation percentage (87.5%)
was obtained between the expected and predicted values. The predicted values
are generated by the E-nose, based on the sensor responses of the different
resin grades. If an unknown HDPE resin is analyzed by E-nose, a predicted
value can be obtained for it based on its sensor response. Based on the linear
correlation model, an expected sensory score can be estimated, which would be
a true indication of an actual sensory score based on an organoleptic response
from a consumer panel. The need for continual sensory evaluation by a real time
consumer panel can be eliminated. The correlation models obtained for Run II

and Ill conditions are shown in Figure 23 and 24, respectively.

75

 

0:036:00 __ :3. :2 wo:_0> 092090 0:0 0.20098 502000 5:06:00 “ mm 059".

 

 

0063xm

ovN

 

(\i
paoIpaId

 

vawmmd “ cozﬂotoo

 

 

76

0:020:00 =_ :01 :00 000_0> 09200.0 0:0 000.0098 :00500 :0=0_0:00 ” 0N 059“.

 

om-m Sm

 

em .m

8 .m

omN

000000xm
8N ova om-N oo-m 89 am: a: oh: 8.: 8d 8.0

 

 

wovomnd H cozmﬁtoo

9913!Pe~1d

 

 

77

It can be seen that high correlation percentages were not obtained for Run
conditions II and Ill (63% and 76%). This can again be attributed to the fact that
higher sensor responses were not obtained for Run II and Ill conditions. In
addition, a high correlation percentage for Run | conditions validates that the Run
l conditions are the most ideal to analyze food grade HDPE resins. Hence the
standardization and usage of the most ideal run parameters becomes critical for
generating reliable and reproducible data from the E-nose.

A correlation between the actual human sensory responses and the E-
nose sensor responses becomes very useful in predicting the quality of an
unknown HDPE resin. A successful prediction can also be made about how
similar or different the unknown resin is in comparison to known resin qualities,
as the E-nose would assign an expected sensory score to the unknown, based
on the sensor responses of the unknown resin. This cOuld eliminate the need for

a formal human sensory evaluation of the unknown sample.

E-nose analysis of off-flavored water samples

As discussed in the earlier sections, a correlation was established
between the E-nose sensor responses of the HDPE resins and the sensory
characteristics of the water samples, stored in contact with the resins. Such a
correlation would be useful in predicting the quality of the resins, in terms of its
likelihood of generating off-ﬂavor in the water sample. However, a more intimate
correlation would be between the taste characteristics of the water samples,

evaluated by sensory panel and E-nose sensor responses of the same off-

78

 

flavored water samples. An effort was made to analyze the off-ﬂavored water
samples (stored in contact with the resins), using the E-nose. The idea was to
understand whether the E-nose was capable of discriminating between the water
samples and effectively correlating the sensor responses with the human
sensory scores. Two different experimental run conditions were tried in the E-
nose, involving variations in incubation time and temperature. It was found that
E-nose sensors were not capable of discriminating between the different water
samples in both cases. This is evident from negative discrimination percentage
obtained in the PCA proﬁle (Appendix F). It was also difﬁcult to establish

standard experimental run conditions, suitable for analyzing water samples.

79

CHAPTER IV
CONCLUSIONS
AND

RECOMMENDATIONS FOR FUTURE WORK

In this study a standard method for analyzing the quality of food grade
HDPE resins was developed, using the E-nose. The method is capable of
predicting the quality of an unknown food grade HDPE resin sample, in terms of
its possibility of producing an off-ﬂavor in drinking water, as a result of migration
of low molecular weight compounds from the polymer. This is evident from the
high correlation percentage in the PLS model, which indicates a good correlation
between the E-nose sensor responses and human sensory scores for the resin
samples. Such an experimental method may completely substitute for the
sensory evaluation techniques, which are currently being used for this
application, with a simpler qualitative and quantitative data analysis, offering
better reliability and reproducibility of results.

A much better correlation between the sensor responses and the human
sensory scores could be obtained if the water samples were discriminated more
distinctly by the human sensory panel. This would have facilitated a better
ranking of the resins based on the degree of off-ﬂavor generated in the water
samples, unlike categorizing into three broad groups, as done in this case. One
of the main reasons for not being able to achieve a clear sensory ranking, with a

distinct sensory score attributed to each resin, is the use of untrained consumer

80

 

panel. Determining off-taste in water sample is often a very difﬁcult task and
needs good understanding and sense of taste perception, which can be
developed only among trained panelists. Thus use of a trained sensory panel
would certainly be a positive step ahead in this area of research.

The work also validated the sensitivity and efﬁciency of the E-nose
sensors for analyzing HDPE resin samples. This is evident from the high
discrimination percentage in the PCA proﬁles. The ideal experimental conditions
required to analyze food grade HDPE resins have also been validated and
standardized. However the E-nose sensors were not found to be capable of
discriminating between off-ﬂavored water samples. A more sensitive set of E-
nose sensors is recommended for this application, which might even require
training of the E-nose to some extent. A good discrimination between the water
samples can prove highly useful in achieving more meaningful correlation
between the E-nose olfactory responses and human sensory data, which is a
potential area of future research.

The E-nose is capable of performing a variety of multivariate statistical
techniques, which can be used for numerous packaging and food applications.
The E-nose can certainly be considered as a powerful and effective quality

evaluation and control tool for diverse packaging applications.

81

APPENDICES

82

APPENDIX A

c00-r 0_m:0:b _0:00 :0E:0:00 9 00030.0 00.0E00 .0003 53> >0; " mm 0.59".

 

83

APPENDIX B

0300:? £000 05 300:5 09:0005 0:0 0:0: 03:03 00:0 c055.

u on 0590

 

 

84

 

 

APPENDIX C

 

85

Sensory evaluation in progress (consumer panel)

Figure 27

APPENDIX D

 

 

 

Figure 28 : Vials containing water samples, used for sensory evaluation

86

APPENDIX E

 

Triangle Test

Name : Gender: Age : Date

 

Sample : Water (stored in contact with plastic) Panelist # 5

INSTRUCTIONS

- Four independent sets (A, B, C, D) of samples are presented sequentially.

- Each set has 3 samples; 2 are identical. Determine which one is the odd sample.

If no cﬂference is apparent. you must guess.
- Taste samples from left to right.

. Take at least half amount of the water into your mouth from each vial, to get

a feel of its taste.
You may spit the sample out after judging (foam cup provided)
Rinse your mouth by water between samples.

Which one is the odd sample in each Set ? (Check the circle)

 

 

 

 

 

 

 

 

 

Set A O 950 O 150 O 581
Set B O 615 O 407 176
Set C O 977 O 813 764
Set D O 352 O 525 718

 

 

 

 

 

Figure 29 : Questionnaire for sensory evaluation (Triangle test)

 

 

 

APPENDIX F

080E 0200005 :0003 00:0:000: 0:0 003600 :0003 00:90:00

m:0E0 :o_00:_E_:00_0 02 I <00 ” on 0590

 

 

$5.0m H ..0
mod 9.3 Nod.

7w ..- H x005 :0_:0:_::_:00_Q

3d.

mod.

mod.

 

o to.
1:3 d.

- 180.9

- Iowa .0.

- 195.c-

ID8.D-

:-

8.

qI
%LV'€II 5 33

195.:-

.1899

105.:

famed

 

18nd

 

 

88

BIBLIOGRAPHY

Adebiyi, A.P.; Adeyemi, LA. and Olorunda, AC. 2002. Effects of processing
conditions and packaging materials on the quality attributes of dry-roasted
peanuts. Journal of the Science of Food and Agriculture. 82(13): 1465-
1471.

AI-Bachir, M and Mehio, A. 2001. Irradiated luncheon meat : microbiological,
chemical and sensory characteristics during storage. Food Chemistry. 75:
169—175.

Alpha MOS Fox - 3000 operating manual. Release Jan. 2002.

Anselme, C.; N'Guyen, K.; Bruchet, A. and Mallevialle, J. 1985. Can polyethylene
pipes impart odors in drinking water? Environmental Technology Letters.
6(11): 477-488.

Bartlett, P.N; Elliott, J.M. and Gardner, J.W. 1997. Electronic noses and their
application in food industry. Food technology. 51 (12): 44-48.

Benedetti, S.; Toppino, P. M.; Riva, M. 2002. Shelflife of packaged Taleggio
cheese: Use of electronic nose. Scienza e Tecnica Lattiero-Casearia.
53(4): 259-282.

Bravo, A.; Hotchkiss, J.H. and Acree, TE. 1992. Identiﬁcation of odor-active
compounds resulting from thermal oxidation of polyethylene. Journal of
Agriculture and Food Chemistry. 40: 1881-1885.

Briston, J.H. and Katan, LL. 1974. Plastics in contact with food. London: Food
Trade Press.

Brocca, D.; Arvin, E. and Mosbaek, H. 2002. Identification of organic compounds
migrating from polyethylene pipelines into drinking water. Water Research.
36: 3675—3680.

Brody, A. 1989. Flavor interactions with packaging. Prepared Foods. 158(10):
128-132.

89

 

Cabral, Antonio C. D.; Madi, Luis F. C.; Soler and Roger M. 1977. Detection of
off-ﬂavors in foods packed in ﬂexible packages. Boletim do Instituto de
Tecnologia de Alimentos (Campinas, Brazil) 54: 13-40.

Calvosa, L.; Chiodini, G.; Coretti, W.; Donaggio, P.; Orlandi, M.; Paratici, V. and
Rindone, B. 1994. Taste and odor development in water in polyethylene
containers exposed to direct sunlight. Water Research. 28(7): 1595-1600.

Church, l.J. and Parsons, AL. 2000. The sensory quality of chicken and potato
products prepared using cook-chill and sous vide methods. International
Journal of Food Science and Technology. 35(2): 155-162.

Crompton, TR. 1979. Additive migration from plastics into food. Oxford:
Pergamon Press.

Culter, JD. 1992. Minimizing plastic package/product interactions — An unﬁlled
need. Journal of Plastic Film & Sheeting, 8(3): 208-226.

Culter, JD. 1999. The control of product and package quality with the electronic
nose. Tappi Journal. June(1999): 194-200.

Culter, J.D.; Hernandez, R.J. and Selke, S.E.M. 2000. Plastics Packaging:
Properties, Processing, Applications, and Regulations. Hanser Gardner
Publications.

Dean, A and Voss, D. 1999. Design and analysis of experiments. Springer —
Verlag Inc., USA: 339 — 347.

Desobry, S. 2000. Packaging/fatty food interactions. Oleagineux, Corps Gras,
Lipides. 7(5): 427-430.

Doust, A.M.M.; Mousavi, S.M.; Oromiehie, A. 2003. Liquid-liquid extraction and
GC-MS methods to determine the migration of organic compounds from
polyethylene to packaged water. Environmental Chemistry: 1-7.

Downes, T. W.; Giacin, J. R. and Hernandez, R. J. 1996. The spectrum of
product/package interaction and the advanced technologies being
developed to stop them. Polymers, Laminations & Coatings Conference,
Boston. 2: 523-526.

90

 

Durst, G.L. and Laperle, EA. 1990. Styrene monomer migration as monitored by
purge and trap gas chromatography and sensory analysis for polystyrene
containers. Journal of Food Science. 55(2): 522-524.

Fauconnier, M.-L.; Panhaleux, V.; Vanzeveren, E.; Marlier, M. and Wathelet, J.
P. 2001. Modelling migration from high-density polyethylene containers
into concentrated solutions used as food ﬁavourings. Food Additives and
Contaminants. 18(11); 1040-1045.

Fayad, N.M.; Sheikheldin, S.Y.; AI-Malack, M.l-l.; El-Mubarak, AH. and Khaja, N.
1997. Migration of vinyl chloride monomer (VCM) and additives into PVC
bottled drinking water. Journal of Environmental Science and Health, Part
A: Environmental Science and Engineering & Toxic and Hazardous
Substance Control. A32(4): 1065-1083.

Gardner, J.W. and Bartlett, RN. 1993. A brief history of electronic noses.
Sensors and Actuators. 18. 211-220.

Giese, J. 2000. Electronic Noses. Food Technology. 54(3): 96-100.

Gilbert, S.G. 1978. Origin of food ﬂavors through packaging materials. American
Chemical Society Abstracts. 176 : AGFD 5.

Gilbert, S.G.; Miltz, J. and Giacin, JR. 1980. Transport considerations of
potential migrants from food packaging materials. Journal of Food
Processing and Preservation. 4: 27

Goldenberg, N. and Matheson, HR, 1975. Off-ﬂavors in foods, a summary of
experience 1948 — 1974. Chemical Industry. 5: 551.

Hanlon, JP. 1992. Handbook of Package Engineering. Lancaster, PA:
Technomic.

Hansen, W.G. and Wiedemann, 8.0.0. 1999. Evaluation and optimization of an
electronic nose. In. “Electronic noses and sensor array based systems.”

Technomic publishing co. USA: 131-143.

Harper, W. J. 2001. The strengths and weaknesses of the electronic nose.
Advances in Experimental Medicine and Biology: 59-71.

91

 

Heinio, R.L. and Ahvenainen, R. 2002. Monitoring of taints related to printed solid
boards with an electronic nose. Food Additives and Contaminants. 19:
209-220.

Heydanek, Menard G., Jr. 1977. Tracing the origin of off-flavors in a breakfast
cereal. Analytical Chemistry. 49(11): 901A-902A, 904A.

Ho, Y. C. and Yam, K. L. 1999. Effects of individual components in a vitamin E
formulation on off-odor release, melt ﬂow index, and yellowness index of
an HDPE polymer. Polymer-Plastics Technology and Engineering. 38(1):
1941.

Ho, Y.C.; Yam, K.L.; Young, 86. and Zambetti, PF. 1994. Comparison of
Vitamin E, lrganox 1010 and BHT as antioxidants on release of off-ﬂavor
from HDPE bottles. Journal of Plastic Film & Sheeting. 10: 194-212.

Hodgson, S.C.; Casey, R. J.; Bigger, SW. and Scheirs, J. 2000. Review of
volatile organic compounds derived from polyethylene. Polymer-Plastics
Technology and Engineering. 39(5): 845-874.

Hoff, A. and Jacobsson, S. 1981. Thermo-Oxidative degradation of low-density
polyethylene close to industrial processing conditions. Journal of Applied
Polymer Science. 26: 3409-3423.

Huber, M; Ruiz, J and Chastellain, F. 2002. Off-ﬂavor release from packaging
materials and its prevention - a foods company’s approach. Food
Additives and Contaminants. 19: 221-228.

Hurst, W.J. 1999. Electronic noses and sensor array based systems. Technomic
publishing co. USA.

Katan, LL. 1980. A generalized dimensional analysis of migration, and physical
application of this to plastics for contact with food, cosmetics and medical

products. 9-28

Kim-Kang, H. 1990. Volatiles in packaging materials. Critical Reviews in Food
Science and Nutrition. 29(4): 255-271. Boca Raton, FL: CRC Press.

92

Kiritsakis, A.; Kanavouras, A. and Kiritsakis, K. 2002. Chemical analysis, quality
control and packaging issues of olive oil. European Journal of Lipid
Science and Technology. 104: 628-638.

Koster, E.P.; Zoeteman, B.C.; Piet, G.J.; de Greef, E; van Oers, H; van der
Heijden, B. G. and van der Veer, A. J. 1981. Sensory evaluation of

drinking water by consumer panels. Science of the Total Environment. 18:
155-166.

Leong, C.M.O.; Harte, B.R.; Partridge, J.A.; Ott, DB. and Downes, T.W. 1992.
Off-Flavor development in milk packaged in polyethylene-coated
paperboard cartons. Journal of Dairy Science. 75: 2105-2111.

Linssen, J.H.; Janssens, J. M.; Roozen, JP. and Posthumus, MA. 1991.
Sensory descriptors for a taint in water packed in test pouches made of
polyethylene-lined aluminum. Journal of Plastic Film & Sheeting. 7(4):
294-305. '

Maneesin, P. 2001. GC-MS and Electronic Nose analysis on off-ﬂavor
components in HDPE containers and correlation with sensory evaluation.
Doctor of Philosophy dissertation. Michigan State University, East
Lansing, MI.

Meilgaard, M; Civille, G.V and Carr, ET. 1991. Sensory evaluation techniques.
CRC press Inc., USA.

Mielle, P. 1996. ‘Electronic noses’: Towards the objective instrumental
characterization of food aroma. Trends in Food Technology & Science. 7:
432-438.

Okayasu, H.; Naito, S. 2001. Sensory characteristics of apple juice evaluated by
consumer and trained panels. Journal of Food Science. 66(7): 1025-1029.

Olsen, J. R.; Ashoor, S. H. 1987. An assessment of light-induced off-ﬂavors in
retail milk. Journal of Dairy Science. 70(7): 1362-70.

Passy, N. 1983. Off-ﬂavors from packaging materials in food products; some

case studies. Instrumental Analysis of Foods: Recent Progress, 3rd
International Flavor Conference, 1: 413-21.

93

 

Peled, R. and Mannheim, CH. 1977. Off-ﬂavors from packaging materials.
Modern Packaging. 50: 4548.

Petrova, L. I. and Boikova, Z. K. 1976. Possibility of using ABS plastics in contact
with food products. Plasticheskie Massy . (12), 31-2.

Pierce, L.M. 2002. Sensory testing ensures ‘tasteless’ packaging. Food & Drug
Packaging: 38-41.

Potts, M. W.; Baker, S.L.; Hansen, M. H. and Hughes, M. M. 1990. Relative taste
performance of plastics in food packaging. Journal of Plastic Film &
Sheeting 6(1): 31-43.

Righi, F.; Zarabini, L. A.; Legnani, P. and Leoni, E. 1999. Organoleptic alterations

of bottled natural mineral water. Industrie delle Bevande. 28(164): 589-
591.

Risch, SJ. 1988. Migration of toxicants, ﬂavors, and odor-active substances from
ﬂexible packaging materials to food. Food Technology. 42: 95-102.

Risch, SJ. 2000. Flavor and package interactions. Flavor Chemistry. 756: 94-
100.

Schaller, E.; Bosset, JD. and Escher, F. 1998. ‘Electronic Noses’ and their
application to food. Academic Press. 31: 305-316.

Schmitt, V and Tan, T.T. 2001. Simplify quality control with Electronic Nose.
G.I.T. Laboratory Journal. 1 : 2-4.

Sharma, G. K.; Madhura, C. V. and Arya, S. S. 1990. Interaction of plastic ﬁlms

with foods. II. Effect of polyethylene and polypropylene films on the
stability of vegetable oils. Journal of Food Science and Technology. 27(6),
328-31.

Shepherd, M.J. 1982. Trace contamination of foods by migration from plastic
packaging - A review. Food Chemistry. 8: 129-136.

Skjevrak, I; Due, A; Gjerstad, KC. and Herikstad, H. 2003. Volatile organic

components migrating from plastic pipes (HDPE, PEX and PVC) into
drinking water. Water Research. 37: 1912-1920.

94

 

Solin, C.F.; Fazey, AC. and McAlIister, P. R. 1988. Which plastic package has
the lowest contribution to taste and odor? Journal of Vinyl Technology.
10(1): 302.

Steele, D.H.; Thornburg, M.J.; Stanley, J.S.; Miller, R.R.; Brooke, R.; Cushman,
JR. and Cruzan, G. 1994. Determination of styrene in selected foods.
Journal of Agricultural Food Chemistry. 42: 1661-1665.

Stuetz, R.M; White, M. and Fenner, RA. 1998. Use of an electronic nose to
detect tainting compounds in raw and treated potable water. Journal of
Water Supply Research and Technology — Aqua. 47(5): 223-228.

Teranishi, R.; Buttery, R.G.; Stern, DJ. and Takeoka, G. 1991. Use of odor
thresholds in aroma research. USDA. Agricultural Research Service.
Albany, CA.

Thompson, L.J.; Deniston, DJ. and Hoyer, CW. 1994. Method for evaluating
package-related ﬂavors. Food Technology. 48: 90-94.

Tombesi, N.B.; Freije, H. and Quimica, A. 2002. Application of solid-phase
microextraction combined with gas chromatography-mass spectrometry to
the determination of butylated hydroxytoluene in bottled drinking water.
Journal of Chromatography. 963(1-2): 179-183.

Van Deventer, D. and Mallikarjunan, P. 2002. Comparative performance analysis
of three electronic nose systems using different sensor technologies in
odor analysis of retained solvents on printed packaging. Journal of Food
Science. 67(8): 3170-3183.

Villiberg, K.; Veijanen, A. and Gustafsson, I. 1998. Identiﬁcation of off-ﬂavor
compounds in high-density polyethylene (HDPE) with different amounts of
Abscents. Polymer Engineering and Science. 38(6): 922-925.

Vom Bruck, C. G.; Hammerschmidt, W. 1977. Determination of the foreign ﬂavor

threshold of foods and the importance in the selection of packaging
materials. Verpackungs-Rundschau. 28(1): 1-4.

95

 

Wheeler, T.L.; Shackelford, S.D.; Koohmaraie, M. and Roman L.H. 1999. Trained
sensory panel and consumer evaluation of the effects of gamma
irradiation on palatability of vacuum-packaged frozen ground beef patties.
Journal of Animal Science (Savoy, Illinois). 77(12): 3219-3224.

Willing, B-I.L.; Brudin, A. and Lundstrom, I. 1998. Odor analysis of paperboard,
the correlation between human senses and electronic sensors using
multivariate analysis. Packaging Technology and Science. 11: 59-67.

Winquist, F.; Hoernsten, E. G.; Sundgren, H. and Lundstroem, l. 1993.
Performance of an electronic nose for quality estimation of ground meat.
Measurement Science and Technology (1993), 4(12), 1493-1500.

Wyatt, OM. 1986. Analytical analysis of tastes and odors imparted to foods by
packaging materials. Journal of Plastic Film & Sheeting. 2(2): 144-152.

Yam, K. L.; Ho, Y. C.; Young, S. S. and Zambetti, P. F. 1996. Effect of resin
types and antioxidants on release of off-ﬂavor from HDPE bottles.
Polymer-Plastics Technology and Engineering. 35(5); 727-755.

Young, W.F.; Horth, H.; Crane, R.; Ogden, T. and Arnott, M. 1995. Taste and

odor threshold concentrations of potential potable water contaminants.
Water Research. 30(2): 331-340.

96

 

 

   

IIIIIIIIIIIII