(Woodhead Publishing India in Textiles) Agrawal, Sweety A. - Shaikh, Tasnim N - Engineering Cotton Yarns With Artificial Neural Networking (ANN) - Woodhead Publishing India PVT LTD (2017)

Engineering Cotton Yarns with Artificial Neural
Networking (ANN)
Engineering Cotton Yarns
with Artificial Neural
Networking (ANN)
Dr. (Mrs.) Tasnim N. Shaikh

and
Mrs. Sweety A. Agrawal
WOODHEAD PUBLISHING INDIA PVT LTD

New Delhi, India
Published by Woodhead Publishing India Pvt. Ltd.
Woodhead Publishing India Pvt. Ltd., 303, Vardaan House, 7/28, Ansari Road,
Daryaganj, New Delhi - 110002, India
www.woodheadpublishingindia.com
First published 2017, Woodhead Publishing India Pvt. Ltd.

© Woodhead Publishing India Pvt. Ltd., 2017
This book contains information obtained from authentic and highly regarded
sources. Reprinted material is quoted with permission. Reasonable efforts have
been made to publish reliable data and information, but the authors and the
publishers cannot assume responsibility for the validity of all materials. Neither
the authors nor the publishers, nor anyone else associated with this publication,
shall be liable for any loss, damage or liability directly or indirectly caused or
alleged to be caused by this book.
Neither this book nor any part may be reproduced or transmitted in any
form or by any means, electronic or mechanical, including photocopying,
microfilming and recording, or by any information storage or retrieval system,
without permission in writing from Woodhead Publishing India Pvt. Ltd. The
consent of Woodhead Publishing India Pvt. Ltd. does not extend to copying
for general distribution, for promotion, for creating new works, or for resale.
Specific permission must be obtained in writing from Woodhead Publishing
India Pvt. Ltd. for such copying.
Trademark notice: Product or corporate names may be trademarks or registered

trademarks, and are used only for identification and explanation, without intent
to infringe.
Woodhead Publishing India Pvt. Ltd. ISBN: 978-9-38505-920-9

Woodhead Publishing India Pvt. Ltd. WebPDF ISBN: 9781351186261
Typeset by Third EyeQ Technologies Pvt Ltd, New Delhi

Contents
Preface xiii
1 Classification of textile yarns 1
1.1 Introduction 2
1.2 Types of textile fibers 3
1.3 Types of textile yarn 5
1.4 Significance of cotton yarn in textile industry 9
1.5 Production pattern in brief for most significant
cotton fiber in textile industry 10
1.6 Impact of natural fiber variations on yarn
production process 13
References 14
2 Attributes of cotton mixing 17
2.1 Need for mix formulation 17
2.2 Interrelationship between fiber characteristics
and yarn quality 18
2.3 Contribution of fiber parameters on ring spun
yarn quality & cost 21
2.3.1 Length & length variations 21
2.3.2 Fineness 23
2.3.3 Maturity 25
2.3.4 Strength 26
2.3.5 Trash 27
2.3.6 Moisture 28
2.3.7 Colour 29
2.4 Importance of mix homogeneity 30
2.5 Impact of technological changes on homogeneity of mix 30
Reference 30
vi Contents
3 Testing techniques used in yarn engineering 33

3.1 Introduction 33
3.1.2 Attributes of cotton ring spun yarn engineering 34
3.2 Role of testing in cotton selection 35
3.3 Various fiber testing techniques 35
3.3.1 Fiber length 36
3.3.2 Fiber fineness 44
3.3.3 Fiber maturity 48
3.3.4 Fiber strength 51
3.3.5 Trash 55
3.3.6 Moisture 58
3.3.7 Colour 59
3.4 Stages of developments in testing techniques and its impact
on mix formulation process 61
3.4.1 Classical testing techniques 62
3.4.2 Semi-automatic mode of testing 63
3.4.3 Automatic mode of testing 65
References 69
4 Statistical techniques used in yarn engineering 71
4.2 Analysis of test data 72
4.2.1 Measures of central tendency 73
4.2.2 Measurement of dispersion 74
4.3 Compatibility test methods 78
4.4 Statistical techniques for defining technological value of
cotton 80
4.4.1 Drafting quality index (Q) 81
4.4.2 Fiber quality index (FQI) 82
4.4.3 Modified fiber quality index (MFQI) 84
4.4.4 Spinning consistency index (SCI) 85
4.4.5 Premium discount index (PDI) 86
4.4.6 Multi criteria decision-making (MCDM) 87
Contents vii
4.4.7 Geometric properties index (IG) 88

4.5 Statistical techniques for defining proportion of cotton
constituents in mix 89
4.5.1 Judicious mixing 89
4.5.2 Linear programming 89
4.5.3 Fuzzy linear programming approach 92
References 93
5 Artificial neural networking (ANN) 95
5.1 Introduction 95
5.2 Historical background for the development of ANN 97
5.3 Basic concept of ANN (Artificial Neural Network) 98
5.4 Types of neural network 98
5.4.1 Single layer feed forward network 100
5.4.2 Multilayer feed forward network 100
5.4.3 Recurrent network 100
5.4.4 Learning of a network 101
5.5 Architecture of ANN 101
5.6 Designing the network 102
5.7 Operational mode of ANN 103
5.7.1 Training the network 103
5.7.2 Verification testing 104
5.8 Applications areas of ANN 104
5.9 ANN applications in the field of textile engineering 105
5.9.1 Fibers 106
5.9.2 Yarn 108
5.9.3 Fabric 109
5.9.4 Garment 114
5.9.5 Non-woven 115
5.10 Connotation of ANN offered solutions
over the other methods 116
References 116
viii Contents
6 Changes in mix formulation approach with the

technological developments 127
6.1 Introduction 127
6.2 Basic objectives of mix formulation 128
6.3 Constrains for accurate mixing 129
6.4 Different approaches of mix formulation 129
6.4.1 Classical visual judgment approach 130
6.4.2 Mix formulation with non–automatic instrumental
technology 135
6.4.3 Mix formulation with automatic instrumental
technology 143
References 149
7 Cotton fiber engineering 151
7.2 Importance of cotton fiber engineering 152
7.3 Attributes of cotton fiber engineering 153
7.3.1 Cotton purchasing strategy 154
7.3.2 Cotton testing 154
7.3.3 Bale management 155
7.3.4 Cotton fiber selection 156
7.4 Bale inventory analysis system (BIAS) 159
7.5 Engineered fiber selection (EFS) 160
7.5.1 Determination of cotton specifications 160
7.5.2 Opening line configuration and availability 161
7.5.3 In-house inventory management 163
7.5.4 Mix profiles 165
7.5.5 Bale selection 165
7.5.6 Mix evaluation and performance verification 166
7.5.7 Benefits offered by EFS® 166
7.6 Linear programming 167
7.6.1 Assumptions of linear programming 168
Contents ix
7.6.2 Types of linear programming 168

7.6.3 Effect of inventory constraints 169
References 170
8 Yarn engineering by back propagation algorithm concept of
ANN 173
8.2 Reverse yarn engineering 175
8.2.1 Importance 175
8.2.2 Basic steps of networking 175
8.3 Procedure for cotton yarn engineering 176
8.3.1 Defining aim of yarn engineering 177
8.3.2 Database creation 178
8.3.3 Construction of desired artificial
neural network 179
8.3.4 Modelling 183
8.3.5 Testing of neural network 190
8.3.6 Computing prediction error 191
References 192
9 Optimisation of yarn quality, cost and process parameters 195
9.2 Components for optimisation 196
9.3 Technological value of cotton mix 197
9.3.1 Cotton fiber quality 197
9.3.2 Cotton cost 198
9.4 Optimisation of process parameters 200
9.5 Optimisation of yarn technological value 202
9.5.1 End product added value 204
9.5.2 Optimum fiber quality utilisation for target
yarn by the use of EFS® system 206
9.5.3 Optimum fiber quantity utilisation efficiency for
target yarn by the use of EFS® system 211
9.6 Summary 214
x Contents
References 215
10 Case study 217
10.2 Case study I 218
10.2.1 Basic attributes used for ANN reverse yarn
engineering model 218
10.2.2 Sample preparation 219
10.2.3 Optimisation of ANN parameters 220
10.2.4 Validation of ANN prediction 221
10.2.5 Linear programming 222
10.2.6 Impact analysis 222
10.3 Case study II 225
10.3.1 Basic attributes of artificial neural network 225
10.3.2 Sample preparation 226
10.3.3 Experimentation 227
10.4 Case study III 229
10.4.1 Artificial neural networking 229
10.4.2 ANN parameters 230
10.4.3 Experimental 231
10.4.4 Prediction performance of ANN 231
10.4.5 Analysis of input parameters influence 232
10.4.6 Trend analysis by ANN model 233
References 234
Appendix 237
Abbrevations 249
Index 251
Preface
Cotton fiber is holding an unbeaten position in the apparel sector especially in

the tropical zone due to its superior comfort associated properties. Although
several new fibers have been launched in the commercial market to compete
in terms of quality and cost with this natural fiber but unable to create big
threat to this costly and highly preferred staple fiber. However, the fiber to
yarn conversion process of this fiber is complex due to multivariate non-linear
relationship present between input fibers and output yarns irrespective of type
of spinning system followed. This is mainly attributed to inherent variations
in the natural fiber characteristics. Use of cotton mix instead of single cotton
variety is the best way for minimising effects of input material variations on
quality and cost of yarn produced.
Precise way of cotton mix formulation demands efficient working on three
perspectives. They are ease of availing more varieties of cotton along with
due knowledge about fiber parameters and cost, statistical techniques defining
optimum mix conditions for getting desired yarn quality at an economic rate
and decision-making system for verifying correlation amongst fiber-process-
yarn parameters before starting with process. Thus, it is mainly influenced by
various developments took place in testing, statistic and blending techniques
used apart from supplementary techniques like agriculture, harvesting,
transport, communication, etc. Today ANN is the most preferred and
accurate technique in the replacement of human brain for addressing complex
multivariate non-linear problem of decision-making. Continuous effort have
been made by the researchers and supported by technological developments
to resolve these stringent problems of mix formulation by ANN to maintain a
pace in a competitive market. Critical analysis of various vital changes took
place in all technological aspects of yarn engineering is the main focus of this
book.
The authors are in teaching professions for many years and felt that there
is a need to have a text book, with a compressed source of knowledge on
widespread topics of scientific mix formulation and thereby engineering
cotton spun yarn using artificial neural networking (ANN). There are so many
xii Preface
Cotton fiber is holding an unbeaten position in the apparel sector especially in

the tropical zone due to its superior comfort associated properties. Although
several new fibers have been launched in the commercial market to compete
in terms of quality and cost with this natural fiber but unable to create big
threat to this costly and highly preferred staple fiber. However, the fiber to
yarn conversion process of this fiber is complex due to multivariate non-linear
relationship present between input fibers and output yarns irrespective of type
of spinning system followed. This is mainly attributed to inherent variations
in the natural fiber characteristics. Use of cotton mix instead of single cotton
variety is the best way for minimising effects of input material variations on
quality and cost of yarn produced.
Precise way of formulation of cotton mix demands efficient working on
three perspectives. They are ease of availing more varieties of cotton along with
due knowledge about fiber parameters and cost, statistical techniques defining
optimum mix conditions for getting desired yarn quality at an economic rate
and decision-making system for verifying correlation amongst fiber-process-
yarn parameters before starting with process. Thus, it is mainly influenced by
various developments took place in testing, statistic and blending techniques
used apart from supplementary techniques like agriculture, harvesting,
transport, communication, etc. Today ANN is the most preferred and accurate
technique in replacement of human brain for addressing complex multivariate
non-linear problem of decision-making. Continuous effort have been made
by the researchers and supported by technological developments to resolve
these stringent problems of mix formulation by ANN to maintain a pace in
a competitive market. Critical analysis of various vital changes took place in
all technological aspects of yarn engineering is the main focus of this book.
The authors are in teaching professions for many years and felt that there
is a need to have a text book, with a compressed source of knowledge on
widespread topics of scientific mix formulation and thereby engineering
cotton spun yarn using artificial neural networking (ANN). There are so many
excellent research publications available for ANN application in this field, but
information is scattered. Students need to struggle for getting well compiled
information from one source. Therefore, the authors have thought to prepare
a book by keeping in mind the undergraduate and post graduate students who
took textile engineering and applied mathematics. The book explains the basic
aspects of mix formulation and various technological advances took place
from judiciary decision time to till date (step by step)|. It is expected that
the content of this book will be useful to all researchers, academic institutes
engaged in teaching textiles and professionals working in textile industries.
The book is divided into ten chapters. The initial chapter briefs about
various types of yarn in use and out of them the significant role of cotton
ring spun yarn. It is followed by the set of chapter giving details on various
Preface xiii
attributes of cotton mixing: testing techniques, correlation amongst fiber

properties – process parameters – yarn properties, blending techniques.
Fundamental concept of ANN system with necessary diagrams is given in
Chapter 5. Changes in the mix formulation approach with technological
changes are summarised in Chapter 6. Cotton yarn engineering and use of
back propagation algorithm for the purpose are described in the next set of
chapters. Usefulness of process and yarn technological value optimisation is
explained in Chapter 9. Complexity of ANN based yarn engineering can be
well understood by case study, therefore, three different approach research
works is incorporated in Chapter 10. The objects behind this is to make ANN
based mix formulation and its functions easily understandable by the reader.
They are well incorporated in a proper sequence and the sincere efforts are
given to discuss thoroughly and technically.
This book is the fruit of the joint efforts of the authors. We would like
to acknowledge them who helped us directly or indirectly while preparing
this book, special thanks to Mr. Hardik Pujara for his inputs on ANN. We
would also like to convey our thanks to authors whose research paper we have
included in case study. We are also thankful to our family members, without
their encouragement and moral support this book could not be materialised.
Vadodara, Gujarat (India) Dr. T.N. Shaikh

Mrs. S.A. Agrawal
February 2016
1
Classification of textile yarns

Mrs. Sweety A. Agrawal*
Department of Textile Engineering, Faculty of Technology & Engineering, The Maharaja

Sayajirao University of Baroda, Kalabhuvan, Vadodara, Gujarat 390 001, India.
Corresponding author
*
Abstract
Fibers are the fundamental units of the yarn structure. Such yarns are then
served as founder units for a fabric structure produced by either weaving or
knitting techniques. The basic classification of yarn is thereby done as per
its use in fabric manufacturing. The second largest method of classification
adopted deals with the origin of fiber. Until the turn of 20th century,
all the fibers were obtained from natural resources but later on due to
increased demands and scarcity of natural resources, manmade fibers were
developed. These yarns based on the technology of their manufacturing
were classified as synthetic or regenerated multifilament yarn. Blended or
doubled (ply) yarns were introduced for compensating the limitations of
natural fiber yarns in one or other way round to meet basic needs of end
use application. Simultaneously to compete on economical front various
unconventional spinning techniques were introduced in the market and
given one more tag for the recognition of the yarn; rotor spun, air jet spun,
etc. Thus wide varieties of yarn are available till date in terms of material,
structure and production technology.
The day when spinning and weaving systems were mechanized, i.e.
operated by power instead of human being, cotton took over the charge as
the most widely used fabric material and retained that position till today. The
territory of cotton yarn remains untouched due to its preferential performance
irrespective of production technique or yarn form. Apart from the cotton yarns
other most popular yarns in usage were made up of silk, linen and wool, but
having restricted application areas. No doubt in last century, so many new
2 Engineering cotton yarns with Artificial Neural Networking (ANN)
textile fibers have stricken the market every day, but did not able to shake
cotton position by satisfying umpteen expectations of the customers.
Key Words: textile fiber, spun yarn, plying, blending, spinning technology,
economical.
1.1 Introduction
Yarns are generally made of fibers or filaments, which can come from various
sources. The origin of fiber can be natural or man-made; accordingly they
are classified as natural fibers or man-made fibers. The major resources for
obtaining natural fiber are plant (e.g. cotton, jute), animal hair (wool) or saliva
(silk) and minerals (basalt). Man-made fibers on the other hand differs in
terms of their production system; regenerated from natural resources (viscose
Rayon) or synthesized from chemicals (nylon, polyester). The beauty of man-
made fibers over the natural is their consistency in size or linear density and
limitations in their properties.
Conversely, high performance fibers are driven by special technical
functions that require specific physical properties unique to these fibers.
Some of the most prominent of these properties are: tensile strength, operating
temperature, limiting oxygen index and chemical resistance. Each fiber has a
unique combination of the above properties which allows it to fill a niche in
the high performance fiber spectrum. For comparative purposes carbon, glass
and high density polyethylene are also referenced. Although these fibers do
not necessarily meet all of the requirements of the stated definitions, they
commonly compete in the high performance market and should therefore be
referenced.
The key distinction between filament yarns and staple fiber (commonly
known as spun) yarns are in the type of material making up the yarn:
filaments consist of very long; thin cylinders of extruded material,
either in single strands (monofilament) or in grouped multiple strands
(multifilament). The filament is <1 mm in maximum cross section (i.e.,
diameter, in most cases). A filament can be compared to a length of wire
or fishing line. Multifilament yarns may be, but need not necessarily be,
twisted. Staple fiber yarns generally consist of much shorter lengths of
material 25–180 mm in length. Generally, to form yarn from such short
fibers, the fibers must be first aligned in a parallel fashion (carded, or
carded and combed), then wound together (“spun”) either clockwise or
counter-clockwise so that the fibers adhere to each other. A yarn made in
this way is also called a spun yarn. Spinning is a twisting process wherein
Classification of textile yarns 3
the fibers, which are first aligned in a parallel fashion, are twisted together
until they cling to each other well enough to remain in the form of a yarn.
This type of twisting is to be distinguished from the process of taking two
or more yarns and twisting them together to form a plied yarn. Silk is a
naturally occurring filament; the silkworm produces a single filament of
silk that can be thousands of feet in length. Cotton and wool are examples
of naturally occurring staple fibers. The fibers removed from the cotton
plant are an inch or two in length, while the fibers cut from a sheep are of
several inches in length.
1.2 Types of textile fibers

Yarns are generally made of fibers or filaments, which can come from various
sources. According to basic definition, textile fiber is a long thin object
with a high ratio of length to thickness. Fiber usually having shorter length,
ranging between 10 to 500 mm is called staple fiber, whereas a fiber with
continuous length such as silk which is minimum around 2 km is refereed
as filament. Manufactured fibers are produced either as continuous infinite
length filaments or cut staple of desired length value. Staple fibers irrespective
of natural sourced or manufactured, have to be twisted to make them adhere
into a continuous length of yarn.
The main bifurcation of textile fibers is based on the origin of fiber, either
obtained from natural resources or manufactured by some chemical processing;
accordingly they are classified as natural fibers or man-made fibers (Fig. 1.1).
The natural fibers are further classified on the basis of natural source from
which it is obtained, viz; plant, animal, minerals, etc. The major resource
amongst all for obtaining natural fiber is plants. Fibers can be obtained usually
from plant parts like seed hair (cotton), leaf (sisal, abaca), fruit (coir) and
inner bark also popularly known as bast fibers (flax, jute, hemp, kenaf and
ramie) and are cellulosic in nature. These are categorized under the banner
of cellulosic natural fiber. The other important source is animals. The fibers
can be obtained mainly from animal’s hair, like sheep hair (wool), goat hair
(cashmere), fur (mink) and are proteinase in nature. Another animal fiber is
produced by the silkworm in the form of single filament that is silk. It is
known as filament rather than fiber as it can be thousands of feet in length.
Man-made fibers are basically differs in terms of their production system;
(i) regenerated from natural resources (viscose Rayon) or (ii) synthesized
from chemicals (nylon, polyester). Since chemical engineering is the pioneer
of this manufacturing system, the further diversity is based on the chemical
groups like organic (ester-polyester, amide-nylon) or inorganic (glass).
Different types of textile fibers.

Figure 1.1
Conversely, high performance fibers are driven by special technical

functions that require specific physical properties unique to these fibers.
Some of the most prominent of these properties are: tensile strength, operating
temperature, limiting oxygen index and chemical resistance. Each fiber has a
unique combination of the above properties which allows it to fill a niche in
the high performance fiber spectrum. For comparative purposes carbon, glass
and high density polyethylene are also referenced. Although these fibers do
not necessarily meet all of the requirements of the stated definitions, they
commonly compete in the high performance market and should therefore be
referenced.
1.3 Types of textile yarn

Yarn is “a generic term for a continuous strand of textile fibers, filaments, or
material in a form suitable for knitting, weaving, or otherwise intertwining to
form a textile fabric. Yarns play an important role in the fabric manufacturing
process since a majority of the textile materials are constructed with yarns.
Yarns are also used for products such as sewing and embroidery thread, string,
and rope. Yarns are produced in various sizes and textures, and also vary in
other characteristics. Performance, end use, and fabric care are affected by the
yarn characteristics.
Fiber length is mainly used to broadly divide the yarns into – spun
yarns (made from short or long staple fibers) or filament yarns (made from
continuous filament fibers). Yarn production methods for spun yarns are very
different from those of filament yarns. Spun yarns are composed of either
short staple fibers (cotton), or long staple fibers (jute, manmade). Filaments
have been cut into short or long staple fibers to meet spinning requirements.
Spun yarns may contain fibers of the same type or a blend of different fibers
and accordingly designated. Just for an example spun yarn from 100% cotton
fibers is known as cotton yarn but if it is made up of polyester: cotton blend
than known as PC yarn.
The key distinction between filament yarns and staple fiber (commonly
known as spun) yarns is in the type of material used for making yarn.
Filaments consist of very long; thin cylinders of extruded material,
either in single strands (monofilament) or in grouped multiple strands
(multifilament). A filament is usually <1 mm in cross section i.e., diameter,
in most cases. A filament can be simulated to a length of wire or fishing line.
Multifilament yarns may or may not necessarily be twisted; the decision is
purely influenced by its end use application. Staple fiber yarns generally
consist of much shorter constituent material lengths, viz; 25–180 mm in
length. Generally, to form yarn out of such short length fibers, the fibers
must be first need to be aligned in a parallel fashion and then twisted (or
spun) together either by clockwise or counter-clockwise twisting, so that the
constituent fibers get adhere to each other. A yarn produced in this way is
called a spun yarn in general. But, if such yarn is composed off short length
fibers than referred as short staple spun yarn and similarly yarn with longer
length constituents as long staple spun yarn.
Alignments of staple fibers require spinning preparation that includes
removal of unwanted matters from the fibrous material before spinning. If
this task done by using blow room and carding machine than the yarn spun
out of such preparation is known as carded yarn. But if spinning preparation
is still going one step ahead for yarn quality enhancement, that is removing
short fibers out of process at comber, before spinning (and after blow room
and card), the yarn so produced is known as combed yarn. The difference
in such spinning method is used to manufacture yarn, affects the properties
such as uniformity and strength greatly. Just for example; finer, smoother and
better quality cotton yarns are earned by combing the fibers prior to spinning
(Fig. 1.2).
Figure 1.2 Basic steps of spun yarn production.

Spun yarns are made by mechanical assembly and twisting together
(spinning) of staple fibers. Different system of spinning are used to produce
these yarns like ring spinning, rotor spinning, wrap spinning, air-jet spinning,
etc. Accordingly the yarns produced are known, like ring yarn, rotor yarn, etc.
and differs in their characteristics, like ring spun yarns are soft and bulkier
than respective size rotor yarns (Fig. 1.3).
Twisting is the inherent feature of spinning. If yarn is produced by twisting
process wherein the fibers, which are first aligned in a parallel fashion and
then twisted together until they cling to each other well enough to remain
in the form of yarn, than it is known as single yarn. This type of twisting
is to be distinguished from the process of taking two or more yarns and
twisting them together, the yarn so formed is known as a plied yarn. Further
twisting together of plied yarn can results in the production of thick and strong
cable yarn.
Figure 1.3 Types of yarn spinning.
Single yarns are used in the majority of fabrics for normal textile and clothing
applications, but in order to obtain the special yarn features, particularly high
strength and modulus for technical and industrial applications, ply yarns are
often needed. A folded or ply yarn is produced by twisting two or more single
yarns together in one operation, and a cabled yarn is formed by twisting
together two or more folded yarns or a combination of folded and single
yarns. The direction of twisting is designated as S or Z, just as in single yarns.
Normally the folding twist is in the opposite direction to that of the single
yarns (Fig. 1.4).
Figure 1.4 Different types of ring spun.
Majority of the filament yarns are extruded yarns that are made of natural
extruded fibers or manufactured (chemically synthesized) fibers extruded
through the spinneret. These yarns can be broadly divided into two categories;
(i) monofilament yarn: it is made up of a single and relatively thicker
continuous strands called filament and (ii) multifilament yarn: it is made up of
twisted or twist less multiple filaments, where each component filament runs
the whole length of the yarn. As continuous filament longer lengths requires
little or no twisting to hold the multifilament yarn together (Fig. 1.3).
Yarns can be divided into two categories according to their fabric
manufacturing system use, viz; Woven yarns and Knitting yarns. Woven
yarns are further subcategorized into warp and weft. Yarns to be used in
warp, i.e. the lengthwize direction of a cloth, are generally stronger and
carrying a higher twists than the respective size of weft yarn. They are also
smoother and more even as compared to yarns used for filling, the cross
wise direction of a cloth. Novelty yarns may be used in the warp, but they
are generally found in the filling. Highly twisted crepe yarns are usually
found uses as filling yarns.
Knitting yarns are divided into yarns for hand knitting and yarns for machine
knitting. They are usually more slackly twisted than yarns for weaving. Hand
knitting yarns are generally plied, where as those for machine knitting can
be either single or ply. The popularly used yarns for handknitting are listed
below;
1. Knitted worsted: The four-ply all-around yarn is used for accessories,

for the house, and for apparel. This is the most common weight of
hand-knitting yarn, comprising 90% of the handmade yarn business.
2. Fingering (baby or sock) yarn: The fine yarn that was originally
wool, but is found most commonly in acrylic for comfort and ease of
care.
3. Sport yarn: The three-ply yarn is used for socks, sweaters, and hats.
4. Shetland yarn: The two-ply yarn used for sweaters.
5. Fashion or novelty yarn: Any novelty structure.
1.4 Significance of cotton yarn in textile industry

Cotton plant is essentially grown for its fiber, which is universally used
as a textile raw material. Cotton is an important commodity in the world
economy. Cotton is one of the important cash crops and raw material for
the textile industry. Cotton and cotton textile industries are engines of
economic growth in both developed and developing countries. A revolution
on cotton production is occurring in many developing countries. Many
Governments have recognized that this can increase their export earnings
and can contribute to socio-economic development of rural areas. In view
of the importance of cotton in the Indian economy, Government of India
has launched a technology mission on cotton, which aims to address the
issues of productivity, improving the quality and reducing the cost of
production. Research and development on cotton is being strengthened
for genetic improvement of cottonseeds, resource management for quality
cotton production and post-harvest technology. Ginning and pressing
factories, which use the technology of colonial times, are being provided
assistance for modernization and acquisition of latest machinery and
equipment.
The textiles industry in India accounts for 14% of industrial production,
which is 4% of GDP; employs 45 million people and accounts for nearly
11% share of the country’s total exports basket. Exports of textiles and
clothing products from India have increased steadily over the last few years,
particularly after 2004 when textiles exports quota stood discontinued.
India’s textile exports for the year 2011 was reported at 15,016 million
US $ by the WTO in International Trade and Market. It is found that from
year 2008 to 2012, Cotton Exports in India have increase from 4803 to
11321 (in mn US $).
Cotton is the most widely used textile fiber. The structure consists
of more 90% of cellulose, remaining is waxes, monomer and polymeric
sugars, residual protoplasm and minerals. The main advantage of natural
fibers like cotton is that they are abundantly available in nature. Cotton
possesses some unique fibers properties like higher moisture absorption,
providing great bodily comfort which makes it superior over man-made
fibers. Being a non-allergenic natural fiber, it does not irritate sensitive
skin. Its softness makes it a preferred fabric for other garments worn close
to the skin (e.g. underwear). Cotton fiber has large amorphous portion and
cotton breathes easily as a result of its unique fiber structure. This attribute
makes cotton more comfortable to wear than artificial fibers, which unable
to provide ventilation. So, the fabric made up of cotton fiber is quite
comfortable to user. It has low thermal-conductivity, therefore it is an
ideal material for both summer and winter clothes; in summer it prevents
your skin from heat, and in winter it preserves the warmth of your body. It
can be blended easily with most of the other fibers including synthetics as
well as natural fibers. This permits it to acquire additional strengths from
sharing material, like strength and good creasing properties from polyester
on blending. Cotton fiber has high absorbency rate and holds up to 27
times its own weight in water. This property allows it to undergo hassle
free dyeing. The absorbency also makes it ideal for medical and personal
hygiene products. These properties signifies the role of cotton in textile
industry.
In today’s world, researches are being done to produce textiles which
can perform more technical functions. Nanotechnology has created a range
of new textiles for medical and surgical purposes. Cotton is being made
such that it catches no dirt and stain without altering softness and pliability
of cotton because of this surface finish. Thus this technology has further
widened up the horizon of this popular fiber. Under such circumstances
it becomes mandatory to optimize the cotton yarn engineering process in
terms of quality and cost.
1.5 Production pattern in brief for most significant

cotton fiber in textile industry
Cotton is the highly favourable material for apparel sector due to its preferable
comfort associated properties. Various stages involved in its production
process and their outcomes are briefly summarized in Fig. 1.5.
Figure 1.5 Cotton textile production sequence for apparel.
The manufacturing process commences with the arrival of raw materials

in form of ginned cotton bales in spinning mill. This cotton is trash and dust
laden so, entire yarn production sequence are divided into two compartments:
spinning preparatory and spinning. The cotton undergoes through the process
of opening, cleaning, mixing and blending during spinning preparatory in
order to get rid of unwanted foreign matters embedded with lint at blow room
and card, and also acquire uniformity of fiber profile. Carding is the basic
process which straightens and individualizes fibers to form a round rope
like mass, having structure well simulated with yarn and called card sliver
Basic tasks of cotton spinning.

Figure 1.6
(Figs. 1.2–1.6). As per the quality requirements, sliver is subjected to the

process of combing for the removal of short fibers. This optional process
provides an added advantage of fibers parallelization and straightening. The
combing process is identified with better quality yarn production because it
constitute of all long staple fibers and thereby gives stronger, smoother and
more serviceable fabrics.
Spinning starts with drawing process mainly consists of doubling and

drafting which helps in straightening and parallelization of fibers in the
continuous soft strand of fibers known as drawn sliver. After drawing, sliver
is slightly twisted and wound on bobbins with the aid of flyer to produce
a roving bobbin. This roving is finally taken onto the ring spinning frame
to produce the desired yarn type. Drafted roving is given enough twist to
hold the fibers together and impart strength enough that it does not create any
problem in subsequent operations. Carded and combed yarns differ not only
in terms of processing path but also combed yarn is much more precious in
terms of quality than respective carded yarn.
Apart from the ring spinning there is another methods of yarn production
as mentioned earlier, viz; rotor spinning, dref spinning, air jet spinning, etc.
but they are all having restricted end use application area. Hence this book is
targeted towards engineering cotton yarns by Artificial Neural Networking,
the yarn which is the most versatile in application, viz; ring spun yarn is
considered in the forthcoming discussion.
The yarn so produced is then translated into fabric structure by either
weaving or knitting. The grey fabric is then subjected to finishing process
like dyeing, mercerizing, bleaching, etc. Many a time yarn stage finishing is
conducted to create peculiar patterns in fabric. So, sequence followed is not
necessarily same, liable to change as per need. Finally the two dimensional
fabric is converted into three dimensional garment to fit the customer’s
demand. This process sequence is true for apparel sector mainly. There are
vivid sectors for textile application, but here description of manufacturing
process is restricted to major consumption area only.
1.6 Impact of natural fiber variations on yarn

production process
The raw material cost is more than 50% of the total manufacturing costs of a
spinning mill. Also, raw material accounts for about 80–90% of yarn quality
and 50–70% of yarn cost. Therefore, the spinning mill has a considerable
interest to know the quality characteristics of fibers to reduce the costs and to
optimize the spinning process. Thus, the main technological challenge lies in
any textile production set up is to convert the high variability of input fiber
characteristics to a uniform end product.
Cotton is a major consumed natural fiber, which displays great variation in
its properties. Being an agricultural product, high level of genetic variability
is found in quality parameters, like, diameter, maturity level, fiber length and
uniformity, etc. amongst others. The selection of suitable cotton fiber to meet
the customer’s end use requirement has remained a major challenge to the
spinner because the consistency of yarn quality depends on the consistency of

the raw material and on the spinning processes.
Production of natural fibers is not totally in control with human as far as the
quality and quantity of the fiber is concerned. Variation in length, fineness, maturity
level, uniformity, etc. of the natural fibers causes lot of problems during processing.
These variations lead to the concept of blending. Blending reduces the variations
occurs due to geographical, growing and production factor. The economic impact
of blending is seen from its ability to reduce the costs by providing an opportunity
to replace the expensive cotton with cheaper one in a blend without affecting the
quality. Several problems that occur during the processing the staple yarns can
be explained mainly by the spinning preparation, as a consequence of improper
blending. The knowledge of fiber properties and careful selection of fiber bales
are considered critical stages of spinning preparation. Thus, high quality testing of
bales to know its specific fiber properties will enhance the chances of using cotton
more efficiently. Because of the technical base and know-how for the processing
of this fiber, the application of this fiber can be usually being made less expensive
than using other fiber types. Further, it is suitable for blending with many of
the other fiber types in order to achieve a unique combination of properties not
available in cotton alone.
Cotton quality is defined on the basis of interaction of several physical
properties of the fibers and the inherent variation existing in all these
properties among the fibers. Thus it becomes utmost important to measure all
fiber properties precisely before subjecting to the production process. It is also
equally true that testing should not be based on a small sample drawn from
the whole population but drawn in such a way that it represents the population
accurately. This gives way to develop automated sampling and testing
techniques like AFIS and HVI, for 100% quality check with due accuracy at
faster rate. This is a compulsory feature for assured quality of mix formulation
from large number of cotton varieties, where the raw materials are showing
high degree of variability between the lot and within the lot.
In ring spinning, raw cotton is the prime factor that influences cost and
quality of yarn. Mixing thereby plays an important role in controlling both;
quality and cost of yarn as well as any form of end product availed from
these yarns. If any mistakes happen in selecting the raw material during
the preparatory process before and within the blow room then it becomes
apparently impossible to rectify the mistakes even after having modern
machines in the later processes.
References
1. Banerjee P.K., Gupta Kaushal K., et al., 1991, Publication of Two

Week Course on Testing and Evaluation of Geotextiles, IIT Delhi.
2. Gowayed Yasser, 1995, Theory and practices of cotton fiber selection.

Part I: Fiber selection techniques and bale picking algorithms, Textile
Research Journal, 65(1), 32–40.
3. EI Mogazhy & Yehia E, 1992, Optimizing cotton blend costs with
respect to quality using HVI fiber properties and linear programming.
Part I: Fundamentals and advanced techniques of linear programming.
Textile Research Journal, 62(1), 1–8.
4. Gupta V.B. and Kothari V.K., Manufactured Fiber Technology,
Chapman and Hall, London, 2nd Edition.
5. Murthy H.V.S., 1999, Introduction to textile fibers, the textile
association India.
6. Tandon N., et al, 2013, A study on emerging trends in textile industry
in India, International Journal of Advancements in Research &
Technology, 2(7), ISSN 2278-7763.
7. http://cottonaustralia.com.au/cotton-library/fact-sheets/cotton-fact-
file-cotton-properties-and-products visited on 15/10/15.
8. http://nptel.ac.in/courses/116102026/visited on 12/10/15.
9. http://textilelearner.blogspot.com/2011/05/physical-properties-of-
cotton-fiber-end visited on 5/10/15.
10. http://www.cottonguide.org/cotton-guide/the-world-cotton-market/
the-importance-of-cotton-in-world-trade/ visited on 14/10/15.
11. http://www.slideshare.net/GolamKabir/final-mdgolam-kabir visited
on 22/9/15.
12. https://wcrc.confex.com/wcrc/2007/techprogram/P1780.HTM visited
on 12/09/15.
13. https://www.icac.org/meetings/cgtn_conf/documents/05_mohapatra.
pdf visited on 10/11/15.
14. www.slideshare.net/vaibhavmathankar/classification-of-yarn-yarn-
classification?related=5 visited on 22/9/15.
2
Attributes of cotton mixing


*
Abstract
Cotton is the most versatile natural fiber for its different end use applications.
But its high variability accounting for precise mix formulation to maintain
consistency in the quality and cost of yarn produced. The strong interrelationship
is prevailing between fiber characteristics and yarn quality; this demands not
only the knowledge about fiber properties like length, fineness, strength,
maturity, trash and colour etc., but also their contribution toward desired
yarn quality. Fiber parameters like length, fineness, strength etc., also play
a detrimental role in defining cotton cost. Apart from that behaviour of fiber
during production process, fiber parameters and their influence differs quite
considerably as per production technology. So, this chapter is designed to
elucidate the aspect of yarn engineering. The entire discussion is concentrated
round to versatile ring spinning technology only to avoid undue complications.
Key Words: formulation of mix, yarn quality, ring spun yarn, mix homogeneity,
fiber parameters.
2.1 Need for mix formulation

Raw material represents major share in the manufacturing cost of a short
staple spun yarn. This fact alone is sufficient to indicate the significance of
the raw material for the yarn producer. The condition becomes more critical
while dealing with natural fiber like cotton with inherent variations in its
characteristics. These variations are affected by region and time of cultivation,
crop cultivation conditions, monsoon schedule and ginning practices. These
lot to lot variations influence becomes still more apparent when the ease
in processing fiber material of first lot is compared with the difficulties,

annoyance, additional effort, and the decline in productivity and quality
associated with subsequent lots of similar material. Through blending of large
volume of cotton become necessary for any spinner to use a problem-free raw
material and retain consistency in producing yarn quality for a long period
of time. But it would normally be too expensive and demands huge initial
investments with added burden of inventories and maintenance.
Yarn quality gets ensured based on two aspects, they are:
1. Number of fiber per cross section should be uniform throughout the

length of yarn.
2. The proportion of fibers of each component of mix should be identical
in cross section throughout the length.
The first aspect can never be satisfied with natural fiber like cotton, even
if only one variety of cotton is used. This condition is mainly attributed to
the natural variations. The second aspect requires the cotton mix formulated
should be uniform for a long time interval, i.e. consistency of mix should
be maintained. Limitations of crop availability, price fluctuation and
varying government policies for this natural fiber complicates spinner’s job
in maintaining consistency of mix and thereby yarn quality and cost. The
situation can be over ruled statistically by formulating the mix from more
than one variety of cotton (higher the number lesser the per cent of CV). This
facilitates in optimising yarn quality by averaging out the cotton variations and
also minimises every possibilities of economic impact. As more replacement
options for the given cotton are available along with its reduced share in raw
material cost from 100%. Ultimately mix formulation prevents use of cotton
component, which is either costlier on the account of larger investments done
for huge inventory or due to hiked cotton prices and also nullifies or at least
minimises undesirable cotton characteristics variations arises from a lot to lot.
However, optimal mix formulation conditions can be obtained only through
systematic raw material selection. These demand an extensive knowledge
about the starting material and its behaviour in processing and subsequent
stages to address the expected practicing difficulties in getting desired output.
2.2 Interrelationship between fiber characteristics

and yarn quality
As mentioned in the previous section it becomes utmost important for the
spinner to well understand inter relationships amongst several parameters
of fiber quality, process and yarn quality. This can prevent deterioration of
cotton process ability in the spinning mill. The contribution of different fiber
Attributes of cotton mixing 19
characteristics for a particular yarn property varies as per type of spinning

system used. While working on similar ground Sasser (1991) had given
contribution coming from different fiber properties for one of the most
important characteristics, viz; yarn strength by considering similar fineness
(26s Ne/23 tex ), produced out of ring spinning and rotor spinning. According
to him, fiber length was playing a detrimental role for ring spun yarn but
for rotor spun yarn fiber strength was found highest contributing factor in
the group. Similar shuffle was observed for rest of the fiber characteristics
contribution (Table 2.1).
Table 2.1 Contribution of fiber properties towards yarn strength
Fiber properties % Contribution

Ring spun yarn Rotor spun yarn
(26 Ne/23 tex )
s
(26s Ne/23 tex )
Length 22 12
Length uniformity 20 17
Strength 20 24
Micronaire 15 14
Elongation 5 8
Color/reflectants 3 6
Trash content 3 6
Others 12 13
Thus, correlation amongst fiber parameters, yarn parameters and process
parameters never remains constant but varies with type of spinning system
(conventional or unconventional), technological condition of machine (old or
modern), end use application etc.
Uster Technologies had also worked out correlation between fiber properties
and yarn tenacity for ring spun yarn (Table 2.2).
Highly significant correlation was found between all yarn properties and
fiber length and length variation. Comparatively lower influence on yarn
properties arises from fiber fineness. Little or no correlation was found
between rest of the fiber properties and yarn properties except few, e.g.
nep content and cleanliness of cotton has highly significant influence on
yarn irregularities and faults but have no or a little correlation with rest of
properties. Thereby it becomes interesting at this stage to have a brief review
about the influence of various fiber properties on yarn quality and cost for a
given process parameters.
20
Table 2.2 Correlation between fiber properties and yarn tenacity for ring spun yarn
Fiber parameters Yarn parameters

Breaking Breaking Classimat Imperfections: thick Evenness Hairiness
strength elongation (%) faults/100 km places, thin places, (U%/CV%)
(Fmax/tex) Neps/1000 m
1. Length & length uniformity Highly Highly Highly Highly significant Highly Highly
significant significant significant correlation significant significant
correlation correlation correlation correlation correlation
2. Micronaire value Significant Significant Significant Significant Significant Significant
correlation correlation correlation correlation correlation correlation
3. Nep content, trash content & Highly significant
trash type correlation
4. Colour — — — — — —
5. Breaking strength (3 mm) Highly
significant
correlation
6. Breaking elongation (3 mm) Highly
significant
correlation
Engineering cotton yarns with Artificial Neural Networking (ANN)
2.3 Contribution of fiber parameters on ring spun yarn

quality & cost
The characteristics of starting materials have a great influence on process
performance as well as product yarn quality. No doubt extent of correlation
between fiber parameters and process parameters gets varied as per type
of spinning system. So, the entire discussion is focused round to versatile
ring spinning and also for highly preferred cotton yarn only. The major
fiber properties influencing ring spun yarn characteristics as per order of
contribution are listed below:
1. Length & length variations

2. Fineness
3. Maturity
4. Strength
5. Trash
6. Colour etc.
2.3.1 Length & length variations
Cotton fiber length plays a decisive role in determining its commercial as well
as qualitative value as the selection of cotton and its price are generally based
on this character. This statement gets substantiated by the fact that cotton with
other parameters being equal, longer length cotton on spinning gives better
performance than shorter ones and sold at higher price also.
Fiber length has a great influence on process parameters. The upper quartile
length (effective length or 2.5% span length) defines various setting values
on spinning machines, viz; draft zone setting, feed roller to beater setting,
carding zone as well as stripping zone settings etc., and also adopted for
trading purposes. Similarly mean length (or 50% span length) is generally
regarded as providing a better measure of spinning performance and yarn
quality. Longer cottons, which are often finer, are generally more prone to
form neps during carding and are therefore often carded at lower speeds and
also combed to remove neps for the better alignment of the fibers. Longer
fibers as compared to shorter one also enable lower roving and yarn twists for
acquiring same strength, as the twist inserted into longer fibers transverses
and entwines over a longer length of yarn. Spinning with low twist level can
allow working with higher delivery speed for a given spindle speed, thus,
permits higher production rate and reduces cost/kg. Spinning of finer, stronger,
even and less hairy yarns, as well as stronger fabrics with better appearance
becomes possible with long staple fibers due higher allowable optimum twist
level. Thus selection of cotton fiber is usually based on cotton fiber length for
different count ranges to be spun. It has also a great impact on each and every
process of spinning.
Count pattern spun in spinning mill mainly rely on fiber length. SITRA has
specified norms for fiber properties required while processing on modern high
production spinning set up (running with more than 50 kg/hr. card production)
for different count ranges. They are given in table 2.3 as ready reference over
here. Hence in modern high production spinning set up cotton fiber length
cannot be selected as per traditional testing system, so testing values given are
based on HVI calibration mode.
Table 2.3 SITRA norms for cotton fiber properties required
for ring spun yarn
Count range (Ne) Fiber length Micronaire Strength Maturity index

(mm) (g/tex)
Carded yarn
Below 20s Ne 26–27 4.0–4.6 24–27 More than 0.85
Combed yarn
20s Ne to 32s Ne 28–30 3.6–4.2 27–30 More than 0.84
33 Ne to 45 Ne
s s
30–32 3.5–3.8 29–31 More than 0.84
46 Ne to 60 Ne
s s
32–34 3.3–3.6 30–32 More than 0.84
However, length of cotton is a variable quantity as in natural fibers vary

enormously even in small random bunch, although belonging to same location
of cultivation and species of cotton. This is mainly attributed to the variations
in its agronomic and environmental conditions with respect to time during
fiber development stage and also due to stresses fibers get subjected by the
mechanical processes at and after harvest. For example damage caused to
fiber on ginning mainly depends on cotton variety, seed, moisture, temperature
applied in gin and fiber conditions at the feed end to the gin (e.g. weathered
fibers). But the length distribution pattern differs markedly for hand-harvested
and hand-ginned samples as compared to those mechanically harvested and
ginned. This happens mainly due to difference in extent of mechanical stresses
undergone by the fibers, which are resulted in varying rate of fiber rupture
during these processes.
These length variations invite spinning limit, due to adverse impact on
major yarn qualities, viz; strength, evenness, hairiness and also on spinning
productivity. Excessive length variations (e.g. CV% of fiber length, uniformity
ratio or uniformity index) reduce fiber control at spinning triangle and results
in increased yarn hairiness, fly generation and end down at ring frame. Higher
end breakage rate increases manufacturing waste in the mill thereby reduced
yarn realization, lowers down machine efficiency with more tendency of

traveller loading, excessive EYC cuts and cop rejections at autoconer and
pilling or barre problems in fabric due to introduction of piecing faults. Even
wide length variations bring about high amplitude drafting wave generation and
increased fiber rupture at blow room. Cotton fibers with low length uniformity
should not process together as they create great difficulty for the spinner in
setting up process parameters. Fiber length must stay consistent as variations
in length can cause severe problems at all processing stage and lead to an
increase in waste, deterioration in processing performance and yarn quality.
Thus achieving only average fiber length value as per the standards is not
enough to guarantee better spinning performance; hence it should go together
with minimum lot to lot length variations. Cotton lots with higher variations
should preferably run in controlled quantity only otherwise adversely impacts
yarn quality and cost as mentioned above.
2.3.2 Fineness
Fiber fineness is one amongst three important fiber quality characteristic

influencing ring spinning performance. Thereby knowledge about fiber
fineness becomes mandatory in deciding the spinning value of cottons. This
can be well explained with simple example of geometry; assuming yarn as
cylinder filled with small diameter cylinders which represents constituent
fibers. If yarn size (outer cylinder diameter) is kept constant and fiber size
gets varied, i.e. same size yarns are produced from two varieties of cotton;
coarse (bigger diameter) and fine (smaller diameter). Then the yarn spun from
the variety having finer fibers will definitely carry more number of constituent
fibers/cylinders. Owing to higher number of fibers per cross section available
in yarn structure, genetic variations of this natural fibers gets average out and
such yarn execute higher strength, fullness as well as better uniformity as
compared to that spun from the coarser fibers. Fiber fineness thereby plays an
important role in defining fiber spinning limits, spin plan, process performance
in terms of quality of products, viz; yarn and fabric (obtained out of fibers),
and economy.
2.3.2.1 Spinning limit
Minimum number of fibers per yarn cross section is required for desired
spinning stability. However, this number varies from fiber to fiber and system
to system which is based on fiber length and fineness. For example ring
spinning of cotton fibers requires minimum fibers per cross section; 33 for
combed yarn and 75 for carded yarn, whereas rotor spinning needs 100 fibers
per cross section. Finer fibers can reach this spinning limit with smaller value
of resultant yarn diameter and thus spinning of finer yarn counts becomes
possible. Against that rise in resultant yarn diameter with coarser constituent
fibers is quite high, restricts yarn size spun out of them. Thus spinnability
of fibers depends on fiber fineness in addition to its length properties. The
spinning limit can then be calculated approximately by following equation:
Yarn Fineness(Tex y )
Number of Fibers(N f ) =
Fiber fineness (Tex f )
≥ 33;for combed yarn; 2.1
≥ 75;for carded yarn
It is practically found for coarse and medium ring spun yarns that yarn
produced with fiber per cross section below 90 leads major working problems
due to low yarn strength and above 300 leads higher unevenness.
2.3.2.2 Performance of product
Finer fibers offer more fiber cohesion due to higher available specific surface
area and thereby inter fiber friction, results in stronger yarn production.
Yarn formed of finer fibers thereby need less amount of twist for the desired
strength as compared to one spun out of coarser fibers. Lower the twist more
voluminous and softer the yarn produced.
The irregularity in the yarn depends upon the average number (n) of the
fiber in the cross section. With a higher number of fibers per yarn cross section
provides better distribution of fiber in yarn matrix and its basic irregularity
gets reduced. The average number of fibers in the cross section for given count
of yarn will depend upon the fiber fineness. The uniform yarn also brings
about other desirable properties such as better tenacity, extensibility and
lustre, low count CV%, strength CV% and twist CV%, thus, facilitates better
performance at forthcoming processes due to fewer end breakages and yarn
inbuilt variation.
Finer fibers are softer with less resistance to bend; thereby fabric made
from yarn composed of finer fiber is less stiff in feel and also drapes better.
Similarly, torsional rigidity means ability to twist or resistance against twisting
also depends on the flexibility of fiber. As fiber fineness increases, torsional
rigidity of the fiber along with yarn reduces proportionally. Such fibers can
be twisted easily during spinning and also execute less snarling and kink
formation tendency in the yarn.
Lustre represents uniformity of light reflection rate. Hence, yarn formed
out of finer fibers possesses more number of fibers per unit area, average out
individual’s variations drastically and produce a soft sheen. On the contrary,
hard glitter produced by the yarn produced out of coarser fibers. Going in
agreement, the apparent depth of the shade will be lighter in case of fabrics
made with finer fibers than in case of coarser fibers. Hence amount of dye
absorbed depends upon the amount of surface area accessible for dye out of
a given volume of fibers. Finer fibers lead to quicker exhaustion of dyes than
coarser fibers.
Yarn characteristics mainly born by the distribution fibers especially in the
production of blends. It must be borne in mind that, at least in conventional
ring spinning processes, fine fibers accumulate to a greater extent in the yarn
core and coarser fibers at the periphery.
2.3.2.3 Performance of process
Low micronaire value causes higher nep generation in blow room and carding,
which leads to higher neps and imperfections in final yarn and also, white
spots or dots in dyed fabric. Finer fibers always have tendency to roll round
to form fiber entanglement or neps. Hence it is recommended to reduce blow
room beater speed and carding production rate while processing fine fibers.
This reduces production rate of the system.
Spinning productivity is influenced via the end breakage rate, the number
of turns per inch required in the yarn (giving improvement of the handle),
which generally leads to better spinning conditions. Finer yarn spun out of
finer fibers with low strength thereby restricts higher production rates due to
low strength.
Higher fineness variations can lead towards barre problems or shade
variations in the fabric. So, spinning of such fibers together must be avoided.
2.3.3 Maturity
Fiber maturity is an important character of cotton and is an index of

development of the fibers. The maturity of the cotton fibers varies not only
between fibers of different samples but also between fibers on the same seed.
A cotton fiber consists of a cuticle, a primary layer and secondary layers of
cellulose surrounding the lumen or central canal. In the case of mature fibers,
the secondary thickening is very high, and in some cases, the lumen is not
visible. In the case of immature fibers, due to some physiological causes, the
secondary deposition of cellulose has not taken sufficiently and in extreme
cases the secondary thickening is practically absent, leaving a wide lumen
throughout the fiber (Fig. 2.1). Thus, the maturity of the fiber is concerned
with development of cell wall. The cell wall thickening is highly sensitive
to growing conditions. Adverse weather, poor soil plant, plant diseases and
pests etc., will increase the proportion of immature fiber and lead to trouble
in processing.
Figure 2.1 Fiber maturity.

Maturity index (MI) is the measure of fiber maturity and it depends upon
the thickness of the cell wall. The fibers are considered mature (ripe) if they
have MI between 50–80% immature if they have MI between 30 to 45% and
dead when they have it < 25%.
Mature fiber achieves and exhibits better fiber properties in all aspects as
per their origin of variety characteristics. On the contrary, immature fibers have
neither adequate strength nor adequate longitudinal thickness. They break into
multiple pieces in blow room and carding section during metallic actions of
beaters and carding wires, results in generation of short fibers and micro-dust,
further drops fiber strength as well as length. Fiber rupture so caused leads to
increased yarn breaks, yarn defects, spinning waste and yarn imperfection level.
Subsequently brings about low working efficiency of spinning machineries and
thereby reduced production rate with due hike in yarn cost.
The immature fibers are more prone to nepping effect, things becomes
acute when fine cottons are being processed, since even the mature fibers are
likely to cause neps by faulty processing. In addition, the neps so formed are
usually more prominent because of their size relative to the diameter of the
yarn, which ultimately reflect as white dots on dyeing the fabric formed out of
such yarn and causing fabric rejection.
2.3.4 Strength
Strength is a dominating characteristic. This can be seen from the fact that
nature produces countless fiber most of which are not usable for textiles only
because of inadequate strength. This can be understood from the fact that the
binding of the fiber into the yarn is achieved mainly by twisting thus explicit
at most 30–70% of the strength of the material.
Fiber strength is generally considered to be next to fiber length and
fineness in the order of importance amongst fiber properties. Fiber strength
denotes the maximum tension the fiber is able to sustain before breaking and
influenced mainly by: molecular structure, number and intensity of weak
places, coarseness or fineness of fiber, relative humidity and fiber elasticity or
extension (percentage of the starting length). Elastic elongation is of utmost
importance as textile products without elasticity are hardly usable. They must
be able to deform in order to withstand high loading but they must return to
shape. The fiber elongation should be at least 1–2% and preferably slightly
more.
There is a high correlation between fiber strength and yarn strength. Cotton
with high fiber strength is more likely to withstand breakage during the
manufacturing process. Fiber strength is affected by plant nutrient deficiencies
and weather. However, it is less influenced by adverse growing conditions than
the length and micronaire.
Fiber bundle strength plays a decisive role for spinning machine speeds.
If fiber is not having adequate strength leads to higher end down adversely
affecting quality, increases waste level and lowers efficiency. The situation can
be avoided by only compromising with production rate or forcing spinning
machines to operate at lower speeds than desirable. Ultimately results in
higher cost for yarn with low strength.
2.3.5 Trash
Trash refers the amount of non-lint materials present in cotton, such as leaf
and bark from the cotton plant. At the point of picking fiber balls from the
plant and collecting at one place in the fields unwanted elements, viz; leaves,
broken branches, dirt, dust, string, metallic pieces etc., come along with cotton
fibers. Such collected cotton sends for ginning, where the seeds get broken
and some of the seed coat fragments with small fibers stuck to it remain in the
cotton. These foreign materials affect the appearance of the yarn and fabric
badly. Moreover, if it remains in the fibers in large amount drafting would be
affected.
A high percentage area of trash results in greater textile mill processing
waste and lower yarn quality and quantity produced. Even trash type and
content in given cotton has a great influence on cleaning efficiency of spinning
preparatory machines, viz; blow room and card. More efforts made to separate
trashy matter which is badly embedded with fibrous matter prone to increase
fiber damage and not allow its true realisation into yarn quality. At the same
time, improper cleaning of cotton leads to higher yarn imperfections and
classimate faults, gives increased end down during ring spinning as well as
winding. It is also appearing as black dots in greige fabric, usually referred
as kitty particles which adversely affect process efficiency as well as product
quality.
Metal parts can cause fire as well as damage to metallic card clothing,
Cloth fragments and packing materials can lead to the presence of foreign
fibers in the yarn which may result in deterioration in dyed yarn quality.
Vegetable matters can lead to drafting disturbances, yarn breaks, loading of
card clothing and contamination in yarn. Mineral matter can cause deposits
and high wear rates in machines.
Trash content information is useful for estimating the net amount of
manufactured textile product obtainable from raw cotton, predicting the quality
of cotton textile products, particularly their aesthetic properties, assembling
and blending values in a mix on a trash content basis, adjusting ginning and
textile processing machinery for maximum efficiency in removing trash from
cotton, and relating trash content of cotton to processing efficiency and end
product quality.
This is a typical problem associated with plant fiber like cotton and therefore,
it becomes utmost important to know the trash type and its content not only
at the point of feeding to the spinning preparatory but also at different stages
of processing. This provides information about the cleaning efficiency of the
particular machine and quantum of trash left behind for the next cleaning
process in sequence. Gradation of the cotton is also influenced by its trash
content. Cotton with higher trash content is given lower grade.
2.3.6 Moisture
Cotton fiber is hygroscopic in nature, thereby absorbs moisture easily from

a high humid atmosphere and evaporates when stored in dry atmosphere.
Basic fiber properties like strength, fineness are being affected on acquiring
moisture. Location and seasonal factors due to the same reason leave traces
on fiber properties to more or less extent. Even moisture gets evaporated
during spinning gets accounted under invisible loss and reduced yarn
realisation value. Thereby information about amount of moisture contained
by the fiber becomes important at each stage in the mill processing. Possible
difference in weight even of 0.1% leads to considerable difficulties in those
market transactions where close margins are maintained. Thus, a moisture
standardisation program must be followed to resolve practical problems in
sampling cotton bales as well as smoothening trade resistance. Excessive

drying by heat and exposure to light slightly impair moisture regain capacity
of cotton, so record of previous treatment becomes mandatory at both the
ends.
Cotton should have moisture in the range of 6–8% for smooth spinning
operation. Moisture content below this limit liable to result in high fly liberation
and end down frequency. On the contrary cotton with higher moisture content
than this is difficult to open and clean during blow room and carding. It also
adversely affects sliding behaviour of fibers during drafting at draw frame,
hence gives rise in imperfections and yarn slubs. They again cause higher
end breaks at ring frame stage and lowering the production efficiency with
increase in yarn imperfections and defects.
2.3.7 Colour
Colour of cotton fiber has a considerable influence on raw material quality and
cost. Cotton with bright whitish tint is sold at higher price range as compared
to one with pale yellowish tint. Thus colour grade of cotton is decided in
conjunction with brightness and yellowness of cotton measured in terms of Rd
value and +b value, respectively by HVI.
Colour is related to other fiber properties having influence on processing
performance and fabric quality. As for example pale yellow cotton is
considered as degraded and always possesses low strength value. The colour
shade of cotton gets affected by exposure to adverse weathering like frost
or drought and attack of micro-organisms, viz; insects, fungi, etc., on this
cellulosic fiber. Stains caused due to soil, oil or grease used during mechanical
harvesting, or from green leaves and trash crushed by machinery is capable
of changing colour characteristics of cotton. Condition of cotton at the point
of storage before and after ginning in terms of moisture and temperature,
their excessive values are capable of changing cotton colour. As the colour
of cotton deteriorates because of environmental conditions, the probability
of reduced processing efficiency is increased and at the same time, a lower
market value. Colour deterioration also affects the ability of fibers and
thereby yarn and fabric to absorb and hold dyes and finishes. Monitoring of
colour grade is thus a very important job to control the shade variation or
barre problem in resultant yarn and fabric. Colour is a basic criterion which
decides on the quality classification of cotton raw materials. According to the
standardised procedures developed by the USDA (United States Department
of Agriculture), colour grade describes the colour of cotton lint. There are
standards for 25 colour grades of upland cotton and five categories of “below
grade” colour.
2.4 Importance of mix homogeneity

If different grades of the same fibers are kept together it is called mixing.
Homogeneity of mixing is leading to the homogeneity of physical–mechanical
characteristics of blended yarns. Thus, uniformity of the mix is an important
aspect and depends on radial homogeneity, axial homogeneity and mass
evenness. They can be defined as;
(1) Radial homogeneity: Uniform arrangement of fibers in the cross

section,
(2) Axial homogeneity: Uniformity of mixing degree or composition
between cross section,
(3) Mass evenness: Uniformity of number of fibers between cross sections.
Mass evenness is never been expected for inherently variable natural fiber
like cotton but radial and axial homogeneity can be well attained by precise
mix formulation.
2.5 Impact of technological changes on homogeneity

of mix
Precision of mixing arises from full proof mixing techniques and reduced
variation introduced by each component in mix. The significance of first
situation can be realised with scientific growth of technology but the latter
one largely depends on natural variations of cotton mix gradients. Conversely
there is no technique available till date to terminate natural variations but can
be resolved statistically. Accordingly higher the number of gradients used in
the mix, lesser will be their share and thereby contributed variations. Thus
modern mix formulation should be based on scientific selection of more
number of components in the mix to reduce aggregate variations. This is
discussed well in forthcoming chapter.
Reference
1. Arindam B., 2001, Textile Testing: Fiber, Yarn & Fabric, The SITRA
Publication.
2. Chapter 2 - Cotton value addition - Impact of varieties and production
practices. http://www.cottonguide.org visited on 12/9/16.
3. Duckett K., Zapletalova T., Cheng L., Ghorashi H., Watson M.D., 1999,
Color grading of cotton. Part I: Spectral and color image analysis.
Textile Research Journal, 69 (11), 878–886.
4. ISI Handbook of Textile Testing, Determination of bundle strength

(tenacity) of cotton fibers, 1982, 95, IS 3675–1966 (Bureau of Indian
Standards), New Delhi.
5. ISI Handbook of Textile Testing, IS 236–1968, Determination of
cotton fiber maturity by Sodium hydroxide swelling method, (Bureau
of Indian Standards, New Delhi), 1982, 86.
6. Matusiak M., Walawska A., 2010, Important aspects of cotton color
measurement. FIBRES & TEXTILES in Eastern Europe, 18 3 (80)
17–23.
7. Mohamed Ben, Hassen, Faouzi Sakli, 2008, Adjustment of cotton fiber
length by the statistical normal distribution: Application to binary
blends. Journal of Engineered Fibers and Fabrics, 3(3).
8. Nickerson D., 1953, Color measurement of cotton: Second report
on application of Nickerson-Hunter cotton colorimeter, Including a
Discussion of Recent Work on Standard for Grade, USDA.
9. Nickerson D., Tomaszewski J.J., 1958, Color change in raw cotton
related to conditions of storage. Textile Research Journal, 6, 485–497.
10. P.E. Sasser, et al., 1991, Interpretations of single fiber, bundle, and
yarn tenacity data. Textile Research Journal, 61(11), 681–690.
11. Schenek A., et al., 8/1982, MelliandTextilber, 551–554.
12. Schenek A., et al., 6/1984, Textil-Praxis, 559–563.
13. Schloesing and Hartshorne, 1912, Moisture in raw cotton and yarn. A
report of the standardization committee. Journal Textile Institute.
14. Shaikh T.N. and Agrawal S.A., 2011, Formulation of cotton mix:
development from indecisive to decision support systems. International
Journal of Engineering Research and Applications (IJERA), 1(3),
660–665.
15. Sohel Rana, Concept of High Volume Instrument (HVI) | HVI Used in
Bale Management System, http://textilelearner.blogspot.in/2012/10.
16. Standard Test Method for Length and Length Uniformity of Cotton
Fibers by Fibrograph Measurement1, D 1447–00, November 10,
2000. Published February 2001, Annual Book of ASTM Standards,
Vol 07.01.
17. Standard Test method for measurement of cotton fibers by high volume
instruments, Annual Book of ASTM Standards, ASTM D 5867–95,
Vol. 07.02, ( American Society for Testing Materials, Philadelphia,
USA) 1998, 883.
18. Sunil Kumar Sharma, Cotton yarn: Quality depends on mixing

strategy, 2014. The Indian Textile Journal, 124(6).
19. Uster AFIS PRO 2 Application Report, July 2008, A new Single Fiber
Testing System for the Process Control in Spinning Mills.
20. Xu B., Fang C., Huang R., Watson M.D, 1997, Chromatic Image
Analysis for Cotton Trash and Color Measurements, Textile Research
Journal, 67(12), 881–890.
21. http://www.cottonguide.org/cotton-guide/cotton-value-addition/
length-and-length-uniformity/#sthash.oDIGyTer.dpuf
22. http://www.jeffjournal.org.
23. http://textilelearner.blogspot.in/2012/10/concept-of-high-volume-
instrument-hvi_14.html
3
Testing techniques used in yarn engineering


*
Abstract
Quality of yarn is dictated by the quality of fibers used in the production
course that is the extent of contribution arises from fiber quality attribute in
yarn engineering. This statement is true irrespective of production technology;
conventional slow speed or modern high speed as well as spinning techniques,
viz; classical ring spinning or unconventional rotor, dref or air jet spinning
employed for the purpose. Thereby success involved in meeting consumer’s
demand is mainly attributed to right start or else way right selection of raw
material. But this requires detailed knowledge about fiber parameters and
they should be measured with due precision. Different test methods are in
use, starting from traditional to modern times for this purpose. However,
each of them has associated advantages and disadvantages. An attempt is
made to provide a brief mention about these techniques along with various
specifications used in the course of measure. This can help spinner in sketching
out the several relationship as per kind of technology available at his end.
Key Words: length, fineness, maturity, color, trash, strength, moisture
3.1 Introduction
Fiber to yarn conversion process is a complex process affected by many
parameters irrespective of kind of yarn desired. They include fiber properties,
level of technology employed, type of machinery and their condition, skill of
machine operators and the ambient conditions. Since term “textile fibers”,
covers up a large horizon with a wide variations in their types and properties,
the discussion is restricted up to cotton fiber only.
The classical and non-classical spinning techniques are available for cotton
fiber to yarn conversion. Among them most versatile is the classical ring
spinning system. It has undergone many technological changes during its long
run of development. The older version of ring spinning was labour intensive,
but with due modernisation labour compliment involved in the manufacturing
system got reduced quite considerably and diverted more towards the use of
less but skilled and trained labour. Behaviour of hygroscopic material like
cotton, changes significantly with atmospheric conditions and thereby due
care is taken to maintain desired ambient working conditions in modern ring
spinning plant. Thus, knowledge of raw material characteristic is the most
important tool in controlling yarn quality for the known technological level
of machinery set up available at production end. According to earlier survey
done in this direction the major share (80–90%) of yarn quality is born by raw
material quality. Thus selection of right quality cotton fibers or mix is utmost
important.
3.1.1 Attributes of cotton ring spun yarn engineering
Ring spinning is the oldest and versatile yarn manufacturing technology and
widely used to extract out best qualities of highly preferred cotton fibers on
yarn conversion. The practices of getting the desired quality cotton ring spun
are not new, as carried out from the judiciary time, only its shape has changed
with the developments took place in its various sectors. Two basic attributes
were used in the course of yarn engineering from the beginning; (1) Selection
of suitable cotton components in mix and (2) Decision-making.
(1) Selection of suitable cotton components in mix:
This step has further bifurcations as it depends on indepth knowledge about:
a. Cotton fibers properties and cost
b. Process parameters
c. Influence of various process parameters on the fiber during production
course, i.e. correlation between fiber parameters and process parameters
d. Influence on yarn parameters for the cotton with known parameters
when processed on fixed process set up that is correlation between
fiber parameters and yarn parameters.
(2) Decision-making:
This deals with defining
a. Compability of different cottons when processed together for getting
desired yarn quality. They should result in maximum realization of
fiber parameters in yarn quality and cost or other way round should not
impact adversely on process efficiency or product quality.
Testing techniques used in yarn engineering 35
b. Proportions of different mix components for optimised mix quality

and cost values.
c. Optimised process set up for getting desired yarn quality for fed cotton
mix quality.
Both the aspects of yarn engineering were taken care of by human expertise
in judiciary time but later on different techniques were established to address
them properly, viz;
• Testing technique
• Statistical techniques and
• Artificial neural networking.
Detailing of each in terms of their as well as change in yarn engineering
approach with the availability of various techniques has been given in the
forthcoming sections of this book. Hence ring spinning in itself is a broader
area not possible to accommodate within the scope of this book, only required
details of its process variables are included.
3.2 Role of testing in cotton selection

Despite of wide variations involved, cotton fiber application area is not
restricted due to favorable characteristics, especially in the hot tropical
regions. The machinery response is different for different values of fiber
properties as mentioned in previous chapter. Conversely the long term yarn
quality consistency demands the use of cotton mix to subsidise impact of
cotton variations. The selection and processing of number of cottons together
(mix) thereby becomes a complex function. At the same time it is also equally
important that two cottons mixed together should not differ critically otherwise
create problems during spinning. Altogether demand precise testing for fiber
parameters before proceeding for a mix formulation or rather than that cotton
procurement.
3.3 Various fiber testing techniques

Different test methods are used in measurement of fiber parameters, starting
from traditional to modern times. A brief mention is given for these techniques
along with various specifications used in the course of measure. This can help
spinner in sketching out the several relationship as per kind of technology
available at end.
3.3.1 Fiber length
The fiber length can be described by its distribution by weight function fw that
expresses the weight of a fiber within the length group, or it can be described
by its distribution by number function fn that expresses the probability of
occurrence of fibers in each length group. Thereby various measures employed
for fiber length measurement can be categorised into two groups;
(1) Length by weight measurement,

(2) Length by number measurement.
A weight-biased diagram can be obtained from the distribution by weight
by summing fw from the longest to the shortest length group and similarly,
the diagram by number can be obtained by summing fn from the longest to
the shortest length group. The particular fiber length can be described by its
distribution by number that expresses the probability of occurrence of a fiber
within the length group, or it can be described by its distribution by weight
that expresses the weight of fibers in each length group.
3.3.1.1 Specifications of length and length uniformity
A family of parameters has been developed over the years. The oldest measure
is staple length (Classer’s Length) whereas, mean length (ML), short fiber
content (SFC%), upper quartile length (UQL), upper half mean length
(UHML), upper quartile mean length (UQML) and span lengths (SL) are
the most used length measurement parameters today. Coefficient of length
variations (CV%), percent dispersion, uniformity index (UI%) and uniformity
ratio (UR%) are the measures employed for defining length variations in a
given fiber group.
The technical definitions of different length and its variations measures
obtained by various techniques and widely used in different countries all over
the world are appended below:
(1) Staple length:
It is the length, visually judged by classer from parallel fiber beard. It is the
length, where there are reasonably well defined fiber edges.
(2) Effective length:
There is no clear scientific definition available for this quantity. This measure
is obtained from Baer Sorter diagram. Accordingly it may be defined as the
upper quartile of such numerical length distribution. Thus, effective length is
a characteristic of the bulk of the longer fibers.
(3) Mean length:

As the name suggests it is the estimated quantity which theoretically signifies
the arithmetic mean of the length of all the fibers (100%) present in a small but
representative sample of the cotton. This quantity can be an average according
to either number or weight as per the method of measurement, viz; MLw and
MLn are the mean length by weight and the mean length by number.
Â i =1Li ;
n
MLn = 3.1
Âi
Where, L = fiber length, i = number of fiber
Â Wi Li ;
n
MLW = i =1 3.2
ÂWi
Where, L = Fiber length and W = Weight of fiber length group.
(4) Upper half mean length (UHML):
The upper half mean length by number (UHMLn) as defined by ASTM
standards is the average length by number of the longest one-half of the fibers
when they are divided on a weight basis. Thus, it is the mean length by number
of the longer 50% of fibers when clamped randomly and measured by optical
scanning on HVI.
This parameter can also be reported on weight basis (UHMLw) and it will
be the average length by weight of the longest one-half of the fibers when they
are divided on a weight basis. Thus, it is the mean length by number of the
longer 50% of fibers when clamped randomly.
(5) Upper quartile length:
When all the fibers present in a small but representative sample of the cotton
arranged in descending order of magnitude, the value of length for which
75% of all the observed values are lower, and 25% higher is referred as upper
quartile length (UQL). Thus, upper quartile length by weight (UQLw) is
defined as the length that is exceeded by 25% of fibers by weight. Similarly
upper quartile length by number (UQLn) is defined as the length that is
exceeded by 25% of fibers by number. Both the measures are obtained from
the fiber length frequency distribution curve plotted with Uster AFIS.
Upper quartile length = length exceeded by 25% of the fibers 3.3

(6) Upper quarter mean length (UQML):

The upper quarter mean length by number (UQMLn) as defined by ASTM
standards is the average length by number of the longest one-quarter of the
fibers when they are divided on a weight basis. So it is the mean length by
number of the fibers longer than UQLw.
This parameter can be also reported on weight basis (UQMLw) and it will
be the average length by weight of the longest one-quarter of the fibers when
they are divided on a weight basis.
(7) Span length (SL):
The percentage span length indicates the percentage (it can be by number or
by weight) of fibers that extends a specified distance or longer. The 2.5% and
50% are the most commonly used by industry. It can be calculated from the
fibrogram as:
a. 2.5% span length:

It is defined as the distance spanned by 2.5% of fibers in the specimen being
tested when the fibers are clamped in parallelised and randomly distributed
condition and where the initial starting point of the scanning in the test is
considered 100%. The guideline given for classifying Indian cotton is based
on this measure of the length (Table 3.1).
Table 3.1 2.5% span length based classification of Indian cotton
Class 2.5% span length (mm)

Extra long staple 33.0 and above
Long staple 29.5–32.5
Medium staple 25.0–29.5
Short staple A 20.5–24.5
Short staple B 20.0 and above
Table 3.2 HVI system grading for UI
Grading Uniformity index

Very low <76
Low 77–79
Average 80–82
High 83–85
Very high >86
b. 50% Span length:

It is defined as the distance spanned by 50% of fibers in the specimen being
tested when the fibers are clamped in parallelised and randomly distributed
condition and where the initial starting point of the scanning in the test is
considered 100%.
(8) Coefficient of fiber length variation (CV%):
The coefficient of variation of fiber length CV% is the ratio of σ divided by
the mean length (ML).
σ
CV % = ¥ 100 3.4
ML
Where σ is the standard deviation of fiber length and can be measured by
weight (σW) or by number (σN). It is the root-square of the respective variance
(Vw or Vn) and it expresses the dispersion of fiber length.
(9) Dispersion (%):
This measure is obtained from the sorter diagram that is fiber array of uniform
density formed in the descending order of length. Dispersion represents inter
quartile variations as a percentage of UQL (effective length here).
Inter Quartile Length Range

Dispersion ( %) = ¥ 100 3.5
Effective Length
(10) Uniformity ratio (UR%):

UR% is the ratio between the two span lengths expressed as a percentage of
the longer length (eq. 3.6). It is a smaller value than the UI% by a factor close
to 1.8.
50% Span Length

Uniformity ratio = ¥ 100 3.6
2.5% Span Length
According to SITRA guidelines 50% or a little over 50% uniformity ratio

is observed as the best for this genetically inherent high length variant fiber.
If it becomes <43% there is an increasing short fibers indicating higher length
variations and leads towards associated working problems during spinning.
(11) Uniformity index (UI%):
UI% is the ratio of the mean length divided by the upper half-mean length
(it can be by number or by weight). It is a measure of the uniformity of fiber
lengths in the sample expressed as a percent. Thus, UI is representing length
uniformity of the fibers, while testing on HVI.
MLN ¥ 100
UI N = 3.7
UHML N
Or
MLW ¥ 100
UI W = 3.8
UHML W
Classification of length uniformity index measured from HVI system is

given below:
Short Fiber Content (SFC%)
It can be measured by weight (SFCW%) as well as by number (SFCN%). It
is the percentage by weight or percentage by number of fibers less than one
half inch (12.7 mm).
Short fiber is not measured directly by any instrument employed in HVI
lines. Instead short fiber content (SFC), short fiber index (SFI) is estimated
indirectly using the fibrogram measurements of UHML and ML or 2.5% SL
and 50% SL as the main variables in prediction equations. Accordingly,
Short Fibre Index (SFI ) = 122.56 - (12.87 ¥ UHML ) - (1.22 ¥ UI ) 3.9
Where UHML = upper half mean length (inches), UI = uniformity index
Short Fibre Index (SFI ) = 90.34 - (37.47 × S.L.2.5% ) - (0.90 × UR) 3.10
Where SL2.5% = 2.5% span length (inches), UR = uniformity ratio

Conventional fiber array method defines short fiber content as the fibers
having length less than half the effective length (UQL) for the selected small
but representative sample. This method enables the sample (fibers) to be
fractionated into length groups for determining the cumulative fiber length
distribution by weight in parameters. It gives the measures in both systems; by
weight (SFCW %) as well as by number (SFCN %) as follows:
Area of fiber array covered by short fibers

SFCW % = ¥ 100 3.11
Total Area of fiber array
and
Base Length of fiber array covered by short fibers

SFC N % = ¥ 100 3.12
Total Base Length of fiber array
3.3.1.2 Instrumental measurement of fiber length and its

variations
Fiber length characteristics are determined by genetic (cotton variety) factors

as well as by ginning and textile processing conditions, the latter in terms of
fiber breakage. The staple length, upper half mean length (UHML) and 2.5%
span length all provide similar, but not identical, measures of the length of the
bulk of the long fibers in a sample, and approximate the length of the fibers
when carefully detached from the seed by hand. They are measured by various
classical to modern test methods like Hand stapling, Baer Sorter Diagram,
Digital Fibrograph, HVI and AFIS.
(1) Hand Stapling:
This is the oldest technique where fiber length measurement is carried out
by trained classer (highly experience spinning master) on subjective ground.
A random bale cotton sample is selected and the fibers are prepared by hand
stapling (doubling and drawing) to give a fairly well straightened tuft of about ½
inch wide. This is laid on flat black background and the staple length is measured
by the classer, which is the length, where there are reasonably well defined fiber
edges. Since it is a subjective measure, always remain operator bias
(2) Fiber array method:
Fiber length can be determined using fiber arrays or fiber staple length diagrams
produced using a comb-sorter apparatus. It consists of a bed of upright and parallel
combs which control the fibers and arranged it in the form of an array of uniform
density in the descending order of length. In this way, it enables the sample (fibers)
to be fractionated into length groups for determining cumulative fiber length
distribution by weight in parameters upper quartile length (UQL), mean length
(ML) and % short fibers (SFC) and dispersion percentage. Baer Sorter Diagram
technique is the best example of this class, use a series of hinged combs separated
at 1/8 inch intervals, to align, separate and allow the withdrawal and description of
weight-length or number-length groups from a sample. The disadvantages of this
device are time consuming (2 h per sample) and calls for considerable operator
skill in sampling and preparing the diagram. Thus comb sorter technique is found
unacceptably expensive in terms of operator cost and give results that are too
imprecise for routine testing and for commercial trading purposes.
(3) Digital fibrograph:
It deals with the optical method of measuring the density along the length of
a tuft of parallel fibers. Test specimens for this instrument are fiber beards
prepared manually (old version) or automatically (new version). Auto samples
are prepared by “fibro sampler”. The test fiber beard sample held in a comb
is inserted into the instrument and scanned by a light source. The variation
in fiber density (related to light intensity) of the different lengths of fiber are
recorded and reproduced in the form of a length-frequency curve called the
fibrogram. Two different kinds of fiber length measurement can be generated

from a fibrogram; mean lengths and span lengths. It can overcome the
limitations of comb sorter technique but its semiautomatic mode of operation
still limits its accuracy and testing speed levels.
(4) High volume instrument (HVI):
High volume instrument (HVI) system provides measurement of fiber length,
length uniformity and short fiber (fibers shorter than half inch) by measuring
the light intensity that goes through a sample of fibers. The less amount of
light can go through the sample, the longer the fibers are (use of light in the
distribution of staple length measurements), and therefore such measurement
leads to obtaining a chart called length-frequency curve or fibrogram beard.
The chart (Fibrogram) presenting the relationship between the amount of light
going through the sample and the fiber length was corresponding to this value.
Upper half mean length (UHML), mean length (ML) and SFI are the measures
obtained by this system. Short fiber is not measured directly by any instrument
employed in HVI lines. Instead, short fiber content (SFC), short fiber index
(SFI) is estimated indirectly using the fibrogram measurements of UHML and
ML or 2.5% SL and 50% SL as the main variables in prediction equations.
(5) Advance fiber information system (AFIS):
The USTER® AFIS (Advanced Fiber Information System) is used for the
measurement of individual fibers. The fibers are opened and individualised
with a pair of spiked rollers surrounded by carding segments. The fiber opening
unit works by aero-mechanical separation is to separate trash particles and
large seed-coat fragments from the fibers. These trash particles are extracted
through the trash channel, while individual fibers and neps pass through the
fiber channel. Both channels are equipped with optic-electrical sensors. The
individual fiber passing through the channel is scanned optically and length–
frequency curve is plotted to define various measures of fiber length, UQL,
ML, SFC% and dispersion percentage, which is expressed as (CV%) after
measuring the individual fiber length for a selected weight and number of
fibers which can be varied between 1000 and 10,000.
Since this technique deals with the length measurement of actual number
of fibers belonging to same length group and do not get influenced by the
weight of the fibers. Thereby for the given cotton sample tested length by
number results are always found shorter than using the length by weight
method. Hence it provides basic information: complete individual fiber length
distribution, followed by an accurate length measurement and also fibre
damage in terms of short fiber content by number it serves as an effective tool
for quality control and production efficiency in mills (process control).
The basic difference lies in these measuring techniques is in their working
concept:
(1) Full length measure concept: Comb Sorter technique (Baer Sorter
Diagram), AFIS
Figure 3.1 Different concepts of fiber length measurement. (a) Full length
measure concept.
(2) Partial length measurement concept (Random clamped tuft scanning):

Digital Fibrograph, HVI (Figs. 3.1 (a–b)).
Conventional fiber lengths, measures like mean length effective length
and UQL of modern AFIS technique are derived from the knowledge of
the full length of every fiber in the sample. As against this, span lengths
(conventional 2.5% span length, 50% span length) and UHML are based on
the partial fiber length measured from a randomly taken fiber bundle which
is fixed in a measuring grip. This measurement has good simulation with the
process as compared to the full length measures due to randomised clamping
arrangement of fiber tufts, e.g. drafting. So, it becomes utmost important to
derive inter relationship between two different natures of fiber measurement.
The South India Textile Research Association (SITRA) has worked in this
direction and given the following empirical relationships to estimate the full
length measures: effective length and mean length from the partial length
measures; Span Lengths.
Effective Length = 1.013 ¥ 2.5% Span Length + 4.39 3.13
Mean Length = 1.242 ¥ 50% Span Length + 9.78 3.14

(b) Random clamped tuft scanning.
3.3.2 Fiber fineness
Fiber fineness is one among the three important fiber quality characteristic
normally measured in terms of micronaire. Universal numbering system
defines fineness of this textile specimen in terms of mtex.
3.3.2.1 Specification of fineness
Similar to other solid objects like steel wire, fiber fineness cannot be specified
by reference to diameter with the exception of wool and hair fibers. This
is attributed to fiber cross section, which is seldom circular in addition to
compressible nature of fiber and thus it is not easy to measure its diameter.
Thereby as in the case of yarns, fiber fineness is usually specified by the
relation of mass (weight) to length or linear density; mtex (militex) or Micron/
Micronaire.
(i) Militex: It is the linear density of the specimen representing weight in

milligrams of one kilometers of length.
weight in mg
mtex = 3.15
Length in Km
(ii) Micron/Micronaire: It is the linear density of the specimen representing

weight in micrograms of 1 inch of length.
weight in mg
Micron = 3.16
Length in inches
Whereas for man-made fibers decitex is almost used exclusively, the

micronaire value is used worldwide for cotton. Instruments based on air-
flow method give fiber fineness measure in terms of micronaire. It should
be remembered that the micronaire value does not always represent the true
fineness of the spinnable fibers due to the use of the air-flow method. A low
average value is obtained when a high proportion of immature fibers present
in fiber plug than actual one. Specification by linear density (mtex/decitex)
is more accurate in such a case, but far harder to obtain. It becomes further
difficult for cotton, a natural fiber. It grows in various soils, in various climates,
and with annually changing cultivation conditions. The fibers therefore cannot
be homogeneous in their characteristics, including their fineness.
Experimentally, it was derived that the micronaire value varied, in an
extreme example, between 2.4 and 3.9 from bale to bale in a lot of 500 bales.
Long-staple cotton varieties are commonly finer than medium-staple. The
fiber fineness rating scale used is given in Table 3.3.
Table 3.3 Fiber fineness rating
Micronaire value Fineness
Up to 3.1 Very fine
3.1–3.9 Fine
4.0–4.9 Medium (premium range)
5.0–5.9 Slightly coarse
>6 Coarse
3.3.2.2 Instrumental measurement of fiber fineness
Different principles are used in the measurement of cotton fiber fineness (Fig.
3.2). Mostly adopted methods are:
1. Gravimetric method (Dimensional measure)
2. Air flow method
Figure 3.2 Basic concepts of fiber fineness measurement.
3. Vibrating String Method

4 Optical method
Test methods are either applicable to single fibers (vibrating string, optical)
or deals with a mass of fibers (air flow, gravimetric). Hence mass of fiber
measure is usually preferred as there is a considerable variation found in the
linear density from fiber to fiber even though belonging to the same seed and
also having a good correlation with yarn characteristics. Apart from these
single fiber methods are time-consuming and laborious as a large number of
fibers have to be tested to get a fairly reliable average value. Air flow technique
is widely used for measuring yarn fineness based on bulk of fibers during
routine checks.
(1) Gravimetric method (Dimensional Measure):
This method is based on manual mode of linear density measurement of mass
of fibers. It is performed next to fiber length measurement done by the comb
sorter technique. Parallel fiber tufts are collected from comb sorter diagram
and kept side by side with spacing of 1 cm. These tufts sections are sliced out
with the help of sharp razor and then 100 fibers are counted and weighed on a
sensitive balance either in µg/mg. Fiber fineness (linear density) is measured
as follows:
weight in mg ¥ 106
Fiber fineness ( mtex ) = 3.17
100 ¥ fiber length ( mm )
or
weight in microgram
Fiber fineness ( Micronaire ) = 3.18
100 fiber length ( inch )
(2) Air-flow method:

This test method is based on the measurement of resistance offered to the
flow of air through a plug of fibers compressed into a cylindrical container of
fixed dimensions. Compressed air is forced through the sample, at a definite
pressure for the purpose and the volume (rate of flow) of air is measured by
a roto-meter type flow meter. The sample for Micronaire test should be well
opened cleaned and thoroughly mixed (by hand fluffing and opening method).
The fiber resistance and thereby volume rate of air flow varies in inverse
proportion to the specific surface area (s) offered by the fibers plug.
Where;
p dl 1
Specific surface area ( s ) = 2
• 3.19
p ¥d ¥l d
4
Thus finer fibers own higher specific surface area as compared to coarser
one. However, the specific surface area which determines the flow of air
through a cotton plug is dependent not only upon the linear density of the fibers
in the sample but also upon their maturity. Hence the micronaire readings have
to be treated with caution particularly when testing samples varying widely in
maturity. Air-flow method based instruments, like Sheffield micronaire (old)
and now a days HVI are robust in construction, easy to operate and presents
little difficulty as regards its maintenance.
(3) Optical method:
This method is mainly applicable to the fibers with circular cross section
starting from the conventional projection microscope based measure to the
latest optical fiber diameter analysis (OFDA) test.
Classical method (Projection Microscope)
A random and representative cotton fiber test sample is conditioned for 24
h in standard atmospheric conditions. Then fibers are cut into suitable small
length and slide is prepared by carefully mixing the fibers in to mountant. The
mounting agent should be non-swelling and have a suitable refractive index,
not equals to the fiber refractive index, normally liquid paraffin is used. The
mixture of fibers and mount is spread thinly on the slide and covered with a
cover glass. The slide is traversed manually in zigzag fashion when viewed
under projection microscope so as to cover all the fibers randomly.
Optical fiber diameter analysis (OFDA)
It is designed to automate the conventional projection microscope system.
Here microscope stage moves with the help of two computer controlled
stepper motor (X-Y scan of slide). Image is controlled with a video camera
and digitised by frame grabber board with a 256×256 pixel matrix. The fiber
diameter is then measured with the help of pattern recognition software, which
identifies and measures fibers to resolution of up to about 1 µm.
Fiber diameter analyzer (FDA)
It works on light scattering method. As the name suggests it is a non-
microscopical method of measuring fiber diameter and operates by light
scattering. The fiber cut into snippets 1.8 mm long and suspended in
isopropanol (to give a slurry) are caused to intersect a circular beam of light
in a plane at right angles to the direction of the beam (not greater than 200 µm
diameter). The intensity of scattered light is proportional to the projected area
of fiber, i.e. diameter. Only fibers that completely cross the beam are recorded,
so that the scattered light pulse is then proportional to the fiber diameter. The
flow rate and concentration of the slurry are such that fiber intersects the
beam once at a time. The snippets which do not fully intersect the beam are
rejected. It is capable of measuring 50 fibers per second. The beam diameter
is maximum 200 µm to reduce the effect of any curvature due to fiber crimp.
(4) Vibrating string method
Vibroscope works on this method and gives indirect measure of mass per
unit length for an individual fiber. Resonance frequency “f ” of fiber clamped
under nominal 0.5 cN/tex tension (T) for applied load (W) is measured, from
which fiber linear density M (mass/unit length) is calculated as follows;
{ ( ) }
2
Ê ˆ
M = ÁW ¥ g 2 2 ˜ ¥ 9 ¥ 105 1 + R Ep 3.20
Ë l f ¯ l T
Where; R = radius of fiber, E = Young’s Modulus, l = length, λ = wavelength

and g = gravitational force constant.
At this point of discussion it should be pointed out that most of the fineness
determinations are likely to be affected by fiber maturity, which is another
important characteristic of cotton fibers.
3.3.3 Fiber maturity
Fiber maturity represents an index of developments of the fibers (Fig. 2.1).
3.3.3.1 Specifications of maturity
(1) Normal fibers (N): Mature fibers with a well-developed cell wall cotton
fiber become rod-like after swelling with no convolution and no continuous
lumen. These fibers are classed as “normal”.
(2) Thin-walled fibers: These category fibers lying between the other two
classes.
(3) Dead fibers (D): Convoluted fibers with wall thickness one-fifth or less of
the maximum ribbon width are classed as “Dead”.
(4) Maturity ratio: It is a combined index of normal and dead fibers to
specify maturity of tested cotton. Higher is the ratio, more the mature fibers in
a group. Its mathematical form is
Maturity ratio = [(Normal − Dead)/200] + 0.70 3.21
Where, N=Percentage of normal fibers and D=Percentage of dead fiber

(5) Mature fibers (M): A cotton fiber consists of a cuticle, a primary layer
and secondary layers of cellulose surrounding the lumen or central canal. In
the case of mature fibers, the secondary thickening is very high, and in some
cases, even the lumen is not visible.
(6) Half mature fibers (H): The fibers in which the secondary deposition of
cellulose has not taken sufficiently due to some physiological causes known
as half mature fibers.
(7) Immature fibers (I): Fibers with extreme cases, where the secondary
thickening is practically absent, leaving a wide lumen throughout the fiber.
(8) Maturity coefficient: It is a combined index of mature, half mature
and immature fibers to specify maturity of tested cotton. Its mathematical
expression is:
M + 0.6 H + 0.4 I 3.22

Maturity Coefficient =
100
3.3.3.2 Instrumental measurement of fiber maturity
An immature fiber will show a lower weight per unit length than a mature fiber
of the same cotton, as the former will have less deposition of cellulose inside
the fiber. This analogy can be extended in some cases to fibers belonging to
different samples of cotton also. Hence it is essential to measure the maturity
of a cotton sample in addition to define its fineness, to check whether the
observed fineness is an inherent characteristic or it is a result of the maturity.
Maturity deals with the measurement of the degree of cell wall thickening
and expressed as the ratio of the actual cross-sectional area of the wall to the
area of the circle with same perimeter (Fig. 2.1). The direct method for fiber
maturity is not practicable routine test. So indirect method can be used for the
same.
(1) Maturity ratio method:

The fibers after being swollen with 18% caustic soda are examined under the
microscope with suitable magnification. The fibers are classified into different
maturity groups depending upon the relative dimensions of wall-thickness and
lumen. However, the procedures followed in different countries for sampling
and classification differs in certain respects. The swollen fibers are classed into
three groups as normal fibers, thin walled fibers and dead fibers. A combined
index known as maturity ratio (equation 3.20) is used to express the results.
(2) Maturity coefficient method:
After sorter diagram test tufts of cotton are left on the velvet pad. Each tuft is
laid on a microscope glass slide (maturity slide). The overlapping fibers are
separated with the help of teasing needle such that the fibers are parallel but
separate. The free ends of the fibers are then held in the clamp on the second
strip of the maturity slide which is adjustable to keep the fibers stretched to the
desired extent. The fibers are then irrigated with a small amount of 18% caustic
soda solution which has the effect of swelling them and covered with suitable
slip. After the fibers are observed under projection microscope and examined
to class them into three categories: mature, half mature and immature, based
on lumen width (L) to cell wall thickness (W) ratio (Table 3.4).
Table 3.4 Classification of cotton fiber maturity
Lumen width/wall thickness Rating

<1 Mature
1< L/W <2 Half mature
>2 Immature
About four to eight slides are prepared from each sample and examined.
The results are presented as percentage of mature (M), half-mature (H) and
immature (I) fibers in a sample. The results are also expressed in terms of
“Maturity Coefficient” (eq. 3.21) and cotton is rated based on this measured
value (Table 3.5).
Table 3.5 Maturity coefficient based cotton rating
Maturity coefficient Rating

<0.7, Immature
Between 0.7 and 0.9 Medium mature
>0.9 Mature
(3) Air flow method for measuring maturity:
There are other techniques for measuring maturity using Micronaire
instrument, viz; Arealometer, FMT (Fiber Maturity Tester). As the fineness
value determined by the Micronaire is dependent both on the intrinsic fineness

(perimeter of the fiber) and the maturity, it may be assumed that if the intrinsic
fineness is constant at both pressure levels applied to fiber plug during testing,
then the Micronaire value is a measure of the maturity.
(4) Dyeing methods:
Mature and immature fibers differ in their behaviour towards various dyes.
Certain dyes are preferentially taken up by the mature fibers while some dyes
are preferentially absorbed by the immature fibers. Based on this observation,
a differential dyeing technique was developed in USA for estimating the
maturity of cotton. In this technique, the sample is dyed in a bath containing
a mixture of two dyes, namely Dipheny-l Fast Red 5 BL and Chlorantine Fast
Green BLL. The mature fibers take up the red dye preferentially, while the
thin walled immature fibers take up the green dye. An estimate of the average
of the sample can be visually assessed by the amount of red and green fibers.
3.3.4 Fiber strength
Fiber strength is a second important fiber parameter.
3.3.4.1 Specifications of fiber strength
Different measures used for strength and elongation measures of fiber are
defined below:
(1) Load: It is a force applied to a specimen in its axial direction cause
a tension to be developed in the specimen, usually expressed in grams or
pounds. If it results in the rupture (break) of the specimen, then it is referred
as breaking load (breaking strength).
(2) Stress: It is the ratio between the force and the area of cross-sectional
of the specimen. But for textile fibers, cross-section being irregular, linear
density of fiber is used instead of the cross-sectional area. Such measure of
stress is known as specific stress and also allows strength of different linear
density fibers to be compared. The preferred units are N/tex or mN/tex; other
units which are found space in the industry are gf/denier and cN/dtex.
(3) Tenacity: Coarse cottons generally give higher values for fiber strength
than finer ones. In order, to compare strength of two cottons differing in
fineness, it is necessary to eliminate the effect of the difference in cross-
sectional area by dividing the observed fiber strength by the fiber weight per
unit length. The value so obtained is known as ``Intrinsic strength or tenacity”.
Tenacity is found to be better related to spinning than the breaking strength.
Thus it is the mass stress at break or the specific stress corresponding
with the maximum force on a force/extension curve. The nominal denier or
tex of the fiber is the figure used in the calculation without considering any
allowance made for any thinning of the specimen as it elongates. Expressed in
terms of grams/denier or grams/tex.
Some significant tenacity (breaking strengths) of fibers are:
a. Polyester fiber 3560 cN/tex
b. Cotton 1540 cN/tex
c. Wool 1218 cN/tex
(4) Breaking length: It is theoretical but old measure of tenacity only
and thereby holds same magnitude as well as unit of tenacity grams/tex. It
represents kilometer of specimen length required to break it due to its own
weight.
(5) Strain: It refers a certain amount of elongation (stretching) taking
place in the specimen on the application of load. It is the expression of the
elongation as a fraction of the original length.
(6) Extension percentage: It is the percentage expression of strain. If
elongation is measured at the breaking point of specimen then it is said to a
breaking extension.
Three concepts for fiber extension measure must be clearly distinguished:
a. Permanent elongation: The part of the extension through which the
fiber does not return on relaxation;
b. Elastic elongation: The part of the extension through which the fiber
does return on relaxation;
c. Breaking elongation: The maximum possible extension of the fiber
until it breaks, i.e. The permanent elongation and the elastic elongation
together.
(7) Gauge length: It represents the original length of that portion of the speci-
men over which the strain or change of length is determined.
3.3.4.2 Instrumental measurement of strength
Fiber strength measurement system also goes in accordance to fiber fineness,

i.e. strength is measured by either testing individual fiber or group of fibers.
Again single fiber strength testing is suitable for more uniform nature man
made fibers, otherwise it becomes tedious and time consuming for inherently
highly variable natural fiber. Bundle fiber strength measure also execute good
correlation with yarn strength, thereby it is usually adopted practice for cotton
fiber. There are three different principles used in tensile strength measurement
of fiber. They are:
(1) CRE (Constant rate of extension): Rate of increase of specimen

length is uniform with time (the load measuring mechanism moves
a negligible distance). Uster HVI operates on this principle for
measuring bundle fiber strength.
(2) CRL (Constant rate of loading): Rate of increase of the load is
uniform with time and rate of extension is dependent on the load-
elongation characteristics of the specimen. Bundle fiber strength
measured by Stelometer on this principle.
(3) CRT (Constant rate of traverse): Pulling one clamp at a uniform
rate and the load is applied through the other clamp which moves
appreciably to actuate a load measuring mechanism so that the rate
of increase of either load or elongation is usually not constant. This
method is normally found for yarn strength measurement than fiber
strength.
a. Bundle fiber strength test: All routine checks for fiber strength
are carried out by this method only irrespective of classical
Pressley strength tester and Stelometer or modern Uster HVI
tester. The basic difference between classical and modern system
lies in their operation mode of testing only
i. Non-automatic mode: Strength measurements includes
preparation of fiber beard by hand stapling, followed by fine
and then coarse combing as used for measuring fiber length
(staple diagram). The beard is clamped in two sets mechanical
jaws spaced either at 0 mm or 3 mm (1/8”) gauge length, and
the amount of force required to break the fibers are measured.
Mass of the broken fiber beard is then measured on weighing
balance manually. Bundle fiber tensile strength is calculated from
these measured values in terms of grams per tex (g/tex) or cN/tex.
Pressley test bundle fiber strength and mass values are converted
into of Pressley index (eq. 3.22). It is an absolute measure, used
only for grading the respective cotton strength as weak or strong.
So, it can be converted either to Tenacity (Breaking length) or
tensile strength eqs (3.23–3.24).
Breaking Load (lbs)
Pressley Index ( PI ) = 3.23
Fiber Beard weight ( mg )
Ê g ˆ
Breaking Length OR Tenacity Á ˜ = 5.36 ¥ PI 3.24
Ë tex ¯
Ê lbs ˆ
Tensile Strength Á .inch˜ = [(10.8116 ¥ PI ) - 0.12] ¥ 103 3.25
Ë sq ¯
Similarly, for comparison purpose physical measures of Stelometer are

converted to tensile strength (equation 3.25).
Ê g ˆ Breaking Load ( Kg ) ¥ 1.5 ¥ 10

Tensile strength Á ˜ = 3.26
Ë tex ¯ Fiber Beard Weight ( mg )
The rating used for bundle fiber strength at 3 mm gauge length test for
Stelometer and Pressley is given in Table 3.6.
Table 3.6 Fiber tensile strength rating
Rating Pressley (1000 lb sq. inch) Stelometer (g/tex)

High 75 30
Average 60 24
Low 45 18
However, conversion of one physical measure to other system is normally
avoided, as measuring procedures are not exactly same. Now a days, high
speed, fully automated and more reliable HVI method is used.
ii. Automatic mode: Strength measurement includes preparation of fiber

beard by fibro sampler as used for measuring fiber length. The cotton
fibers are collected by fibro comb from fibro sampler. Such fiber laden
comb is then placed in comb track for brushing the beard for cleaning
and aligning fibers after removing loosely adhering fibers. Combed
beard is slide by a mechanical finger into comb holder that faces fixed
lens and jaws. Just after scanning by light, two pneumatic jaws hold
the fiber band of a fixed width at fixed distance [3 mm (1/8””)], and
the amount of force required to break the fibers as well as elongation
at break are measured. Mass of the broken fiber beard is then measured
by photoelectric sensing (difference in light absorption) and from these
values tenacity of cotton is calculated by using regression equation
in terms of grams per tex (g/tex) or cN/tex, as per USDA – or HVI
– calibration cottons. A tex unit is equal to the weight in grams of
1000 m of fiber. Therefore, the strength reported is the force in grams
required to break a bundle of fibers of 1 tex (unit) in size. Table 3.7
provides HVI-CC, 3 mm gauge length strength calibrations used as a
guide in interpreting cotton fiber strength measurements.
Table 3.7 HVI strength ratings for cotton (3 mm gauge length)
Strength rating HVI strength (g/tex)

Very Strong More than 32
Strong 30–32
Average 26–29
Intermediate 24–25
Weak 23–21
Very weak 20 and below
Fiber elongation (extension) is also measured simultaneously for 3 mm
gauge length test performed by Stelometer as well as HVI. The following
scale represents the cotton fiber HVI elongation rating:
Table 3.8 HVI elongation ratings for cotton (3 mm gauge length)
Elongation rating HVI extension (%)

Very low <5.0
Low 5.0–5.8
Average 5.9–6.7
High 6.8–7.6
Very high >7.6
3.3.5 Trash
Trash includes vegetable matters, mineral matters and foreign matters

embedded with fibrous matter and need to be separated before spinning.
3.3.5.1 Specifications of trash
(1) Percent trash content: It represents unwanted foreign matters carried

by raw cotton, need to be removed before spinning and normally expressed as
the percentage of total cotton stock weight.
(2) Percent lint content: Pure cotton fibers free of all trashy matter in
cotton stock is known as lint. It is also expressed as the percentage of total
cotton stock weight.
(3) Vegetable matter: This represents trashy matters arises from the plant,
like husk portions, Seed fragments, leaf fragments, wood fragments etc.
(4) Mineral matters: This represents trashy matters arises from mineral
resources, like; dust, dirt, sand, soil and coal picked up in transportation of
cotton.
(5) Foreign matters: This category of trash includes metal fragments,
cloth fragments, and other fibrous matters like polypropylene coming from
packing bags.
3.3.5.2 Instrumental measure of trash
Trash content is an element in determining the quality or use value of raw

cotton and trash is the primary factor of the leaf grade of cotton. Since cotton
is usually graded by visual judgment, an instrument that measures trash in
terms that are highly correlated with visual judgment is a decided asset.
Basically two methods are adopted for trash measurement, they are:
(1) Gravimetric method

(2) Opto electrical methods
(1) Gravimetric method: There are several trash analyser instruments
operating on this concept. The well known examples are classical Shirley
trash analyser and latest Microdust Trash Analyser (MDTA).
Shirley Trash Analyser: Separation of lint from the trash is carried out
by making use of the difference in their air buoyancies. Simulating with
production process 100 gm of feed cotton is opened to individual fiber state
before subjecting to air stream for separation of constituents (Aeromechanical
separation). The air stream is so adjusted that it carries only the cotton fibers
and dust leaving the trash to fall in to the collection tray placed beneath the
beater. Dust is filtered off by cages and collected separately in dust collection
box. Finally clean lint, trashy matter and dust are collected and weighed on
digital balance. The values are expressed as percentage of feed cotton to
calculate, Lint (% L), Trash (% T) and Dust (% D) in tested cotton. Since
gravimetric weight base measurement is involved in the system classified
under this heading.
Trash weight ¥ 100

Trash Content (T%) = 3.27
Weight of Feed cotton
Lint weight ¥ 100

Lint Content (L%) = 3.28
Dust Weight ¥ 100

Dust Content (D%) = 3.29
Microdust trash analyser (MDTA): Small quantity of feed cotton (4–6 g)

is subjected to gentle opening action to separate small dust particles and fiber
fragments. They are extracted in dust channel by a low vacuum suction. The
large trash particles are separated from the fibers by a trash knife (grid bar) and
an air baffle is extracted by means of centrifugal force. Separately collected
trash, microdust and fiber fragments are weighed to get their percentage share
in fed cotton.
(2) Optoelectric measurement: Uster HVI as well as AFIS operates on

this concept.
a. Uster HVI: It is provided with separate module to measure trash

properties of fiber. The module uses two lamps to illuminate the
sample during testing. The sample is placed on glass window through
which light comes to it. A pneumatic powered automatic pressure
plate presses the sample with a consistent amount of pressure and light
reflected through the sample. The reflected light passes through filters
and photodiodes. The automated video image processor measures
amount of visible leaf or trash. Thus, the HVI systems measure trash
or non-lint content by use of video camera to determine the amount
of surface area of the sample that is covered with dark spots. As the
camera scans the surface of the sample, the video output drops when
a dark spot (presumed to be trash) is encountered. The video signal is
processed by a micro computer to determine the number of dark spots
encountered (COUNT) and the percentage of the surface area covered
by the dark spots (AREA). The digitised image thereby produces
results in terms of three measurements: trash area, trash count and
trash code.
i. Trash area: It is the percent of sample viewing area occupied by
trash during HVI testing.
ii. Trash count: It is the measure of number of trash particles 0.01
inch in diameter or larger present in a given sample tested by HVI
technique.
iii. Trash code: It represents the range were the tested sample falls
relative to the levels determined during HVI calibration.
The area and count data are used in an equation to predict the amount
of visible non-lint content as measured on the Shirley analyser. The
HVI trash data output is a two-digit number which gives the predicted
non-lint content for that bale. For example, a trash reading of 28 would
mean that the predicted Shirley analyser visible non-lint content of that
bale would be 2.8%.
The ratio between percentage area of trash and trash particle count
is a good indicator of the average particle size in a cotton sample.
For instance, a low percentage trash area combined with a high trash
particle count indicates a smaller average particle size than a high
percentage trash area combined with a low trash particle count. Small
trash particles or “pepper trash” are very undesirable because they are
more difficult for the textile mill to remove from the cotton lint than
larger trash particles.
b. Uster AFIS: The more accurate modern AFIS method uses image
analysis technique for the evaluation of trashy matter content. The
fibers are separated and individualised by aeromechanical opening
unit (opening rollers and air drag). The dust and trash are extracted as
the individualisation of fiber takes place. One electro optical sensor
differentiates between fibers and neps and the second sensor assesses
trash and dust. The signals are processed by a microcomputer to
determine:
i. Sample size: Cotton sample size or weight in grams.
ii. Total: Trash and dust particles per grams in count/g.
iii. Mean size: Mean size of trash and dust particles in µm.
iv. Dust: Dust particles per gm. in count/g.
v. Trash: Trash particles per gm. in count/g.
vi. T.F.M.: Calculated gravimetric trash content.
Here again with well bifurcated information about trash type, size and
content facilitates in setting and defining cleaning efficiency of machines in
production line.
3.3.6 Moisture
Presence of amount of moisture in the hygroscopic fiber like cotton is utmost

important and not only accuracy is involved in quality measurement but also
cost point of view.
3.3.6.1 Specifications of moisture measurement
(1) Moisture regain (R): It is the amount of moisture (W) present in the
material expressed as its oven dry weight (D).
W 3.30
R%= ¥ 100
D
(2) Moisture content (C): It is the amount of moisture (W) present in the
material expressed as its original weight (D + W).
W
C %= ¥ 100 3.31
D +W
3.3.6.2 Instrumental measures of moisture
Different methods used for moisture measurement are broadly classified into
two groups, viz;
(1) Direct methods: Automatic moisture oven

(2) Indirect methods:
(1) Direct method: In this method, the amount of moisture in a definite
weight of the material is driven out by heating (by means of hot air at
105±2 º C) and the loss in weight is determined. Classical Conditioning
oven test is the good example of this category of test. Infrared drying is
another example in this category.
(2) Indirect method: There are two popularly used methods for doing
moisture measurement under this category. They are:
a. Electrical conductivity method: The quantum of moisture
present in cotton alters its electrical conductivity or resistivity
properties. As water is a good conductor of electricity in contrast
to cotton which is very poor conductor of electricity. Thereby
electrical resistance diminishes with an increase in regain or
other way round electrical conductivity increases with moisture
content. The electrode or probe is inserted into bale cotton and
the moisture content (moisture regain) value is displayed directly
on well calibrated scale. This is much faster method as compared
to direct method but affected by condition of sample; presence of
dye, antistatic agent or packing density etc. HVI measures fiber
moisture based on this concept (dielectric constant).
b. Capacitance change based method: The capacitance value
gets influenced by the moisture present in cotton and changes
capacitance value accordingly. Here no physical contact needs to
be made with the sample being tested by sensor. The method is
quick, accurate, passive and harmless. Forte electronic moisture
content analyser test fiber moisture based on this principle.
3.3.7 Colour
Colour of cotton fiber is the combined output basic tint of fiber (whiteness
or yellowness) and reflectance and having a considerable influence on raw
material quality and cost.
3.3.7.1 Specifications of colour
(1) Rd value: It represents degree of light reflectance. The degree of

reflectance (Rd) indicates how bright or dull a sample is. Low Rd levels
indicate dullness or greyness while high Rd levels indicate brightness
or lack of grey. Usually in Indian cotton Rd values lie between 72 and
82. For better control it is advisable to maintain the Rd value range <5
in daily mixing plan.
(2) +b value: It represents yellowness. Thus +b indicates the degree of

colour pigmentation. +b value significantly differs region-wise and
variety-wise. High +b levels indicate a higher degree of yellowness
while low +b levels indicate a low level of yellowness. In Indian
cotton +b lies in range of 6.0–11.0. It is better to maintain the +b range
<2.5 in one variety cotton mixing.
(3) Colour grade: It is the numerical value calculated on the basis of Rd
and +b values of cotton. The colour grade is determined by locating the
quadrant of the colour chart in which the Rd and +b values intersect.
Colour grade is represented in three digits as xx-y.
Example: A sample with Rd value of 72 and +b value of 9.0 would have a
colour code of 41-3. First digit represents brightness of cotton.
Lower the number, higher will be the brightness of cotton. Second digit
represents yellowness of cotton; its yellowish tint gets increased with increase
in second digit number. There are 25 colour grades and five categories of
colour grades. Indian cotton usually lies in white and light spotted categories
with middling to good middling colour grade. It is always advisable that only
four adjacent cotton grades be selected for mixing plan to avoid the shade
variation or barre problem. For example cotton with grades 11, 12, 21, 22 may
run together but running 11 with 31 or 11 with 13 should be avoided.
3.3.7.2 Measurement of colour
There are basically two methods adopted for this subjective matter: colour.
They are visual assessment and instrumental evaluation.
(1) Visual assessment: Cotton colour assessment is performed by cotton

classers in an organoleptic way. A specially trained expert classifies the
cotton sample by visual comparison with a set of physical standards
under standard illumination. The visual assessment is performed in a
room with grey walls, in which samples are placed on a black desk
illuminated by light of 1200 lx illuminance.
(2) Instrumental evaluation: Since 1999, USDA has utilised the
instrument instead of the human classer for determining colour grade
for official Upland cotton classification.
a. Nickerson and Hunter method: Cotton colour grade is
determined instrumentally using a two-filter colorimeter. The
Hunter scales used in a Nickerson Hunter cotton colorimeter
indicate the percentage reflectance (Rd) in a vertical direction,
which is a measure of the lightness of a sample, and in a horizontal
direction the colour code is determined by locating the point at
which the (Rd) and (+b) values intersect on the Nickerson Hunter
cotton colorimeter diagram for Upland cotton.
b. HVI method: It is based on colorimeter technology. The HVI
utilises a dual Xenon light source to illuminate a sample window
measuring 2.8” by 3.6” to give a colour sample area of 10.1
square inches. The two light sources are located at 45º with
respect to the viewing angle in order to comply with ASTM D
1729. This is to avoid the specular reflection component that is
highly related to the smoothness of the surface. In the colour
measurement process, cotton is placed on the sample window
and compressed to a predetermined pressure. This pressure is
necessary to avoid the influence of way of sample preparation
on the measurement results. Light is reflected from the surface
of the cotton sample through two proprietary interference filters,
which are wide band interference filters – one located in the blue
region of the spectrum and the other in the green region – chosen
to approximate CIE standard observer functions. The light is
measured by two separate detectors. The signals from these
detectors are used to calculate the sample colour (Rd) and (+b)
with a precision to the nearest tenth of the measurement unit.
The degree of reflectance (Rd) determined by the HVI shows the brightness
of the sample, which corresponds to the reflectance (Rd) represented in the
Nickerson Hunter colour chart. The yellowness (+b) according to the HVI
is determined using a yellow filter, which depicts the degree of cotton
pigmentation. The yellowness (+b) from the HVI corresponds to the (+b) value
represented in the Nickerson Hunter colour chart. The yellowness (+b) is used
in conjunction with the reflectance (Rd) value to determine the instrument-
measured colour grade of cotton.
The colour of cotton can also be measured by other instruments, such as the
FQT/ FibroLab (Lintronics) and IsoTester (Schaffner Technologies, Inc). The
FibroLab and IsoTester are implemented in industrial practice, but their usage
is not widespread. At present the HVI is considered as a universal method for
cotton colour measurement worldwide.
3.4 Stages of developments in testing techniques and

its impact on mix formulation process
The instrumental measure of fiber techniques described in the previous section
has made it clear that entire testing process invariably includes four steps.
They are:
(1) Sample preparation

(2) Testing including decision of end point
(3) Data record
(4) Statistical analysis of test results
Accordingly, testing techniques can be broadly classified into two groups,
viz; (i) manual mode and (ii) instrumental mode. As the name suggests, if
all the four steps performed manually, the technique is referred as manual
testing technique, e.g. hand stapling technique used for length measurement.
However, instrumental technique not necessarily involves use of instrument
fully in completing all the above mentioned tasks. This category can be further
divided into two more categories as per the extent of human involvement apart
from instrument in the testing course. It is said to be semiautomatic method, if
instrument performs only testing and rest jobs are done by testing personnel,
e.g. Stelometer used for bundle fiber strength. If entire testing exercise
starting from sampling is performed by instrument only then it is put under
the category of an automatic mode, e.g. HVI and AFIS.
3.4.1 Classical testing techniques
Classical testing technique was manual mode based. Hence judgment was
fully lying in the human hands and performed on visual screening based,
many times put under the heading of ``judicious method” also.
The system included random selection of fiber bunch from one of the
cotton bale arrived at mill end, followed by hand stapling (manual drawing
and doubling) to prepare a beard of parallel fibers (Fig. 3.1). The fiber quality
was then visually judged based on 3-parameters:
(1) Grade: Based on colour and trash content,

(2) Fiber length and
(3) Fiber fineness.
The entire process was carried out by the most experience person in the
mill. Testing cost involved was nil and could be performed very quickly but
did not have any scientific ground. The decision was highly influenced by
individual’s expertise and thereby showing variations from person to person.
Quantum of sample taken for testing was very small (few milligrams) in
comparison to lot (1000 kg ) arrived at production end. Thus system was not
able to scientifically address cotton values and also highly human bias.
3.4.2 Semi-automatic mode of testing
Limitations of classical testing mode have initialised instrumental evaluation.

This has introduced measurement techniques for important fiber quality
parameters, viz; length, fineness, maturity, strength, and trash in the market.
Baer Sorter, Fibrograph, Sheffield micronaire, Projection microscope, Pressley
strength tester and Shirley trash analyser were the instruments introduced in
this era in the market. As mentioned in the previous section, they involved
manual sampling followed by instrumental testing but the decision of end
point and recording as well as analysis of test data were still remained human
dependent. Even manual mode of sampling was based on three steps, viz;
preparation of bulk sample, base sample and laboratory sample.
(1) Bulk Sample: From a large number of bales arriving at mill only
representative few bales were chosen at random. The number of bales
selected for testing purpose purely depends on lot size and defined by
standards set up by research organisations [T 3.9].
Table 3.9 Standards for bulk sample size
Bulk size (Bales) Sample size (Bales)
Up to 50 2
51 to 100 4
101 to 150 7
151 to 300 13
301 to 500 20
501 to 1000 32
1001 and above 40
(2) Basic sample: Representative 1 kg of sample was prepared by

withdrawing randomly handful of cotton from various places of
selected bales of bulk sample and mixing them thoroughly.
(3) Laboratory sample: Final lab sample was prepared from basic sample
on random way for conducting the test. Here sample size was affected
by type of test to be conducted, e.g. 3.24 g for Sheffield micronaire but
100 g for trash analyser.
Hence this manual mode of sampling was suffering from major limitations
as appended below:
a. Requires more time in sample preparation.

b. Accuracy of sampling is low due to manual error associated in

preparation.
c. Very small sample (4%) represents large size bulk, e.g. 2 bales out of
50 total number of bales (Table 3.1) represents only 4% [=(2×100)/50]
cotton out of lot to be processed.
Hence all the exercises starting from the sample preparation to data
analysis, exclusive of testing were done manually in first generation testing
technique. So, those instrument used for evaluation of fiber parameters were
referred as non-automatic and suffered from following limitations in addition
to sampling barriers:
(1) Time consuming and tedious: Each fiber property was tested on
separate but non-automatic instrument, thus test speed was very slow.
They have made testing not only time consuming and tedious process
but also resulted in delayed decision-making.
(2) Low accuracy: Manual errors were involved in sampling, defining
end point of testing and statistical data analysis. Just for an example
short fibers were discarded during hand stapling done for length (Baer
sorter) as well as strength (Pressley /Stelometer), give bias sampling
and did not allow measurement of true sample values. Demanded
more number of tests to be performed to meet desired accuracy level.
(3) Higher testing cost: This is mainly attributed to the use of separate
instrument and requirement of higher number of tests in order to
measure each fiber property accurately. Use of separate instruments
has not only added to initial capital investment (for instrument cost
and land cost) but also proportionately hiked labour compliment. The
value of later one was attributed to work practices with this generation
testing instruments, like;
a. Skilled labour were employed for reliable testing and their number
was also high due to more number of instruments involved in the
course,
b. Testing speed was slow and conversely more number of
observations per test were required due to low accuracy level of
the instrument.
Apart from these inventories required and maintenance involved
were also high. All together has made first generation instrumental
testing overall costly.
(4) Poor correlation: Different versions of testing instruments have
different units of measurements and methods of test, e.g. Baer sorter
diagram based on full length measure whereas Fibrogram on partial
length scanning. So, effective length value cannot be compared
directly with 2.5% span length, always need mathematical conversion

[equations (3.113.12)] before comparison, likely to result in rounding
off error. Thus poor correlation found among different instrumental
evaluations, although deals with the measurement of same fiber
property and abide by manual statistical conversions for comparison.
This has further added boundary in availing smooth flow about fiber
information in the textile world.
3.4.3 Automatic mode of testing
HVI was developed for the U.S. Department of Agriculture (USDA) in 1969
to evaluate multiple fiber characteristics in a high volume of samples at a high
testing speed. It uses automated sampling techniques and measures following
fiber properties from a bundle of fibers.
(1) Micronaire
(2) Maturity index
(3) UHML – Upper half mean length,
(4) UI – Uniformity index,
(5) SFI – Short fiber index,
(6) Fiber Strength in g/tex,
(7) Elongation,
(8) Moisture content,
(9) Colour (Reflectance Rd, Yellowness +b) and colour grade (USDA
Upland, Pima, or regional customised colour chart),
(10) Trash (% area, trash count) & trash grade (USDA),
It utilises best working principles in deriving fiber quality values as
described in Table 3.10.
Table 3.10 Different principles involved in HVI
Fineness Air-flow principle

Maturity ratio Calculated using a sophisticated algorithm based on
several HVI™ measurements
Length Optical principle
Strength, elongation CRE principle
Moisture content Conductive moisture probe.
Colour Optical principle with photodiode
Trash Measured optically by utilising a digital camera and
converted to USDA trash grades or customised regional
trash standards.
The measured sample values are evaluated automatically with microprocessor

interfaced with the system and summarised in one measure SCI (Spinning
Consistency Index). It is an output of a multiple regression equation (eq. 4.23)
based on fibre group properties and providing valuable information to anticipate
the yarn parameters especially strength and spinning potential.
The system has undergone various technological changes during its 50 years of
development. The latest Uster HVI – 1000 model offers testing speed as high as 240
bales/h. Provision of high speed but accurate testing for all fiber properties at one
junction has now allowed hundred percent quality check by verify each bale arriving
at mill end. Labour compliment is not only drastically reduced but also trained
labour can do the job well in place of special skilled labour. Reduced operative
hours involved in testing has facilitated reduction in testing cost marginally. Apart
from that eight digit bale ID (bar code) is generated based on cotton SCI value
for each tested bale by HVI testing. It is useful in determining the technological
and economic value of cotton, being in direct relationship with cotton quality and
cotton cost. Cotton values can easily be retrieved by barcode reader at any point
of transaction. This has permitted to shift job of cotton testing on the shoulders of
ginners from the mill quality control personnel. However, it measures behaviour of
group of fibers, so unable to provide good correlation between process variables and
fiber properties in the absence of individual fiber measure.
The development of AFIS has overcome this limitation of HVI and it was
resultant of combined efforts took place between the USDA Agricultural
Research Service at Clemsons, SC and Schaffner Technologies. The primary
objective of this development is measuring trash and neps, followed by efforts
to measure fiber dimensions, number of short fibers, and eventually a complete
fiber length distribution based on single fiber full length measurement (Fig. 3.3).
Figure 3.3 AFIS measurements.

These properties were chosen because of their value in the fiber to yarn
engineering process. Even this basic information about the fiber is useful
for quality control and production efficiency in mill as well as for providing
information needed to improve product quality (Fig. 3.4).
Figure 3.4 Significance of AFIS measurements.

Example for optimizing blow room and card cleaning efficiency: AFIS
test is conducted for cotton at three steps in judging blow room and card
performance. They are testing bale cotton mix before feeding to the blow
room, cotton after processing through blow room and card sliver produced.
(1) The residual trash in blow room output (at chute feed) in comparison to
fed cotton mix and residual trash in card sliver in comparison to blow
room product defines cleaning efficiency earned at blow room and
card respectively. Similarly residual trash in card sliver in comparison
to fed cotton mix defines combined cleaning efficiency of blow room
and card as follows;
Trash ( % ) in feed-Trash ( %) in delivery
Cleaning Efficiency ( % ) = × 100 3.32
Trash ( %) in feed
If the combined cleaning efficiency of blow room and card is found as per
specifications but along with higher short fibers and neps in output card than
they are indicative of fiber damage. The reason can be verified from the AFIS
results of blow room output and card sliver.
(2) The increase in short fiber content (%) observed for feed cotton to
blow room and similarly for blow room product to card sliver are
indicative of fiber damage caused during respective opening and
cleaning processes. Just for an example, higher short fibers found in
blow room output are indicatives of excessive beating action given at
the blow room.
(3) The increase in nep content (%) observed for feed cotton as compared
to blow room output and similarly between blow room product and
card sliver are the indicative of excessive fiber rolling occurred
during production course. These demand review of production pattern
specially for opening points settings at respective junction.
Initiating corrective measures and repeating quality control cycle leads
towards optimisation of blow room and card cleaning efficiency as per
specifications set by consumer.
Thus use of AFIS at each intermittent yarn production stage permits in
availing insights of process with respect to material processed in gaining
desired output. Else operates as a tool for scientific correlation amongst fiber,
yarn and process.
Hence both the methods (HVI and AFIS) are using auto mode of sampling,
testing and evaluation of test data, categorised as automatic testing techniques.
They have offered following advantages over semiautomatic instrumental
measurement techniques:
1. Quicker measurement of all fiber-parameters.
HVI measures all fiber properties at quite a high testing speed of 240 bales/h by
using best possible working concepts. Thus time consumed per fiber parameter
test is significantly less in terms of few seconds instead of minutes. The results
are analysed by microprocessor interfaced with the system which also provides
summary of cotton values in terms of SCI, thus facilitates in determining
technological as well as reduction in time involved for statistical compatibility
checks.
2. Reduces testing error and evaluation time.
All fiber properties are verified at one station only, reduces time log for separate
tests. HVI operates with auto sampling, testing and analysis mode thereby
eliminates human involvement and reduces human associated errors as well
as delay in the testing. Better test accuracy is achieved owing to adoption of
precise working principle. This not only reduces number of tests demanded
per cotton but also reduces testing time (operative hours) required.
3. Reduced testing cost:

All properties are measured by one instrument (HVI) at higher accuracy
by performing quite a less number of tests which reduces drastically labour
compliment, space, storage, power and maintenance results in considerably
reduced overall testing cost.
4. Quicker response to change over and corrective measure:
SCI provides good degree of correlation between fiber and yarn properties as
well as cost, Just for an example fiber with higher SCI should be used for finer
and costly yarn. Two cottons with same SCI are said to be identical in terms
of their quality and cost and can be used in replacement of each other. Thus
it provides scientific ground for changeover from one cotton variety to other
when one or more cotton components in the mix are found unavailable due to
one or other reason.
5. Good fiber properties and process variables correlation
AFIS measures behaviour of individual fibers, viz; average property based
on individual fiber measure as well as variation in it. So, able to provide good
correlation amongst process variables, yarn variables and fiber properties. It
becomes a scientific tool in controlling the process and quicker actuation of
corrective measure, as it also performs testing at quicker rate with accuracy.
References
1. Arindam Basu 2001, Textile testing fiber, yarn &fabric, SITRA
publication.
2. Audiologist Training http://www.freelinksubmission.net/ livescores.
webeden.co.uk.
3. Belmasrour R., Cui L,. et al., 2011, Obtaining cotton fiber length
distributions from the beard test method part 2 a new approach through
PLS regression, Journal of Cotton Science, 73.
4. Booth J.E., Principles of textile testing, CBS Publishers and distributors.
5. Hertel K.L., 1940, A method of fibre-length analysis using the
fibrograph, Textile Research Journal, 510.
6. Iwona F., Mieczysława R., Zbigniew C., 2010, Measurement of cotton
fineness and maturity by different methods, FIBRES & TEXTILES in
Eastern Europe, 18 6 (83), 5459.
7. James R., et al., 2011, Rapid measurement of cotton fiber maturity and
fineness by image analysis microscopy using the cottonscope, Textile
Research Journal, 1–13, DOI: 0.1177/0040517511431317.
8. Krifa M., 2006, Fiber length distribution in cotton processing:

dominant features and interaction effects, Textile Research Journal,
426.
9. Krifa M., 2013, Cotton fiber length distribution modality alteration in
ginning and mill processing, Journal Textile Institute, 731.
10. Shaikh T.N. and Agrawal S.A. 2011, Formulation of cotton mix:
660665.
11. Tallant J.D., 1952, Use of a servo system for automatic operation of the
fibrograph. Textile Research Journal, 22, 617.
12. http://www.cottonguide.org/cotton-guide/cotton-value-addition/the-
impact-of-cotton-fire-properties-on-textile-processing-performance-
quality-and-costs/#sthash.ASZSrG58.dpuf.
13. http://nptel.ac.in/courses/116102029/68.
4
Statistical techniques used in yarn engineering


*
Abstract
Statistical techniques are the inherent feature of mathematics based
engineering. So, the field of yarn engineering cannot be left untouched
in this regard. Yarn engineering begins from the point of yarn testing,
usually generating test data, may be smaller or larger in size depending
on type of test and its degree of complication. Handling such disperses
data and deriving trend out of it requires mathematical analysis in terms of
central tendency and dispersion. The most accurate methods like mean and
standard deviation are used for further evaluation purpose. Compatibility
of individual fiber parameters can be realized by using statistical tools
like; significance testing or critical difference. Fast track testing methods
and need of higher precision demands use of over all fiber quality index
for defining technological value in one numerical figure. Various statistical
techniques were developed in the due course of time, like; Q, FQI, MFQI
SCI, PDI, MIAHP, IG have been accounted in this chapter. Computer
software based statistical technique that is linear programming (LP)
method has also extended support in optimising cotton mix quality and
cost on scientific ground.
Key Words: central tendency, dispersion, significance test, quality indices,
linear programming
4.1 Introduction
Scientific cotton mix formulation involves testing of raw material as well
as analysis and interpretation of such test data. Testing of cottons produces
raw data, which cannot be used directly for decision-making process
and need to be analysed in terms of its central tendency and dispersion

measures. Compatibility tests for central tendency and dispersion measures
are applied to investigate whether significant difference exists between
the given samples or sample from nominal value in order to check the
compatibility. A scientific approach is needed to maximise interpretation
from the data for deriving useful information in context to technological
value of cotton. Hence each individual fiber property holds different units
and also have varying contribution to targeted type of yarn properties. It
thereby becomes complicated to establish scientific correlation between
fiber, yarn and process parameters. This demands a mathematical
expression which is generated by merging different contributing fiber
properties as per their respective weightage into one unique numerical
figure, known as over all fiber quality indexes. Many such indexes have
been proposed by different researchers to reach a decision which may
include the maximum amalgation of properties of fibers for producing a
quality yarn. It is not enough for yarn engineering that the formulated mix
produces only desired yarn quality, but it should be able to produce desired
yarn quality at an optimal cost. So statistical technique for yarn quality and
cost optimisation is obligatory for the successful yarn engineering. Linear
programming technique is the answer to this. Thus, use of statistical or
mathematical based methods out of available ones in right spirit is indeed
needed to reach a proper perspective on the output material. This chapter
basically explores different statistical methods for input fiber test data
analysis, interpretation and output yarn optimisation.
4.2 Analysis of test data

Whenever specimen is subjected to quality checks results in the generation of
data. No doubt size of data set differs from very small to very large, purely
depends on degree of accuracy demanded and type of testing mode. Higher
accuracy level demands more number of tests to be performed and vice versa.
In fact, feasibility of performing more number of tests also matters here. For
example, if fiber length testing is done by using classical Baer sorter diagram
technique and same by latest high volume instrument (HVI) technique then it
is apparent that technical limitations restricts person to conduct higher number
of test with classical one although preferred.
Analysis of data is a process of inspecting, cleaning, transforming, and
modelling data with the goal of discovering useful information, suggesting
conclusions, and supporting decision-making. Data is collected considering
a population. Population is basically the whole bulk of material which is
considered to be tested. It is not possible to test each and every bite of
Statistical techniques used in yarn engineering 73
the allotted bulk, especially true for all the forms of textiles material. A
small quantum of material known as sample is always withdrawn from the
population, either on random or bias basis as per process or product need.
Hence selection of sampling method is made in such a way that such sample
represents every individual aspects of the population to be tested. Since base
material that is fiber and especially cotton fiber here is inherently variable.
Thereby cotton fiber test data need to be analysed in terms of central tendency
and dispersion values. The various measures used in this regard are well
summarised in Fig. 4.1.
Figure 4.1 Different measures of average and dispersion.
4.2.1 Measures of central tendency
The primary analysis of data deals with the measurement of central tendency
or average behaviour. Mean, median and mode are the usually adopted
measures to compute central tendency of measured data.
Mean or average ( X ): It is an arithmetic mean of a given observation and
given by either of equations 4.1 (i–ii) for cotton fiber test values.
i=n
X = Âxi 4.1(i)
i=0
i=n
X = Âwi xi 4.1(ii)
i=0
Where, xi = ith observation and wi = weight or proportion for ith observation in

the data.
Median: Median is the middle value of the observations arranged in the
order of magnitude; either ascending or descending. If data set is composed
off odd number (n) of readings then median will be given by eq. 4.2, but if it is
formed off even number (n) of observations then given by eq. 4.3.
( n + 1)th 4.2
Median = observation,when n is odd
2
ÈÊ n ˆ th Ên ˆ ˘
th
ÍÁË ˜¯ observation + ÁË + 1˜¯ ˙
Median = Î 2 2 ˚when n is even 4.3
2
Mode: Mode basically represents the way or manner in which something
occurs. Mathematically it is the number which appears most often in a given
set of data or other way round it is the highest frequency occurring value in
the dataset.
Location of these measures on graphical plot of test results differs as per
type of distribution. This can be elaborated in Fig. 4.2 (a - c). The distribution
which is symmetric about centre (normal distribution) all three central
tendency measures are lying right at the centre, viz; mean = median = mode.
The distribution with a deviation from the symmetricity it referred as skew
distribution. This tendency of skew can be positive or negative depending
upon the characteristics of observations. If it is more inclined towards
positive side referred as positive skewness where, mean > median > mode.
Contrarily the relationship takes about turn for negative skewness that is
mean < median < mode. The distribution with two modes is known as binomial
distribution usually it follows cause and effect method.
4.2.2 Measurement of dispersion
The variability of any material is given in terms of dispersion. There are several
methods to indicate the dispersion of the values from the central tendency
values. Few of them are discussed below.
Figure 4.2 Different types of distribution curves.

Range (R): It is simply the difference between the highest and the lowest
values of observations in data set. It is simple to compute and also useful
for evaluating the whole of a dataset. The range is useful for showing the
spread within a dataset and for comparing the spread between similar datasets.
However, only two out of all measured values can be used for this purpose, so
level of accuracy involved is less.
Range ( R ) = X max - X min 4.4
Where, Xmax= maximum value observation;

Xmin = minimum value observation.
Mean range (wn): The entire dataset is split up into small ‘n’ number of sub
groups; each of similar size. Then for each subgroup range (Ri) is calculated,
where ‘i’ ranges from 1 to n. Mean range wn is the mean of ranges of ‘n’
subgroups.
n
wn
Â Ri
= i =1 4.5
n�
Hence number of observations incorporated is more here, the extent of
accuracy earned is also high. But still all the observations are not getting the
chance in the course of dispersion analysis.
Apart from that it represents absolute value bound by test method and unit
of measure, so cannot be used for direct comparison purpose.
Percentage mean range (PMR): This is developed to overcome the prescribed
limitations of wn. It expresses the mean range wn as a percentage of the mean and
thus makes it independent of test method and measuring unit. Such measure can
be used for direct comparison purpose. But still it suffers from the limitation of
lack of equal opportunity for each observation in the analysis process.
wn *100
P.M.R. = 4.6
x
Mean deviation (MD): It is the mean of deviations for measured test values.
Here, deviation represents mode taken for difference in the observed value of
data set from the arithmetic mean. This value is computed by considering each
and every observed value of the dataset and is therefore a more accurate value
for dispersion as compared to the range. It can also be expressed as percentage
of the mean for comparison purpose (eq. 4.8). But still sum of the deviations
is calculated by ignoring the sign of the difference, so unable to represent truly
dispersion value.
Mean deviation =
Â x-x 4.7
n
Â x-x
Percentage Mean deviation = n ¥ 100 4.8
x
Where x = arithmetic mean, x = observed value and n = number of the

observations.
Standard deviation (s): It is defined as root mean square (RMS) deviation of
the observations from their averages and mathematically given as:
Â i=0 ( X i - x )
n 2
Standard Deviation (s ) = ± 4.9

n
Hence, it incorporates all the observations in their true sense, thereby much
more reliable as compared to rest of the measures. But still it is an absolute
value and cannot be used for direct comparison purpose.
Coefficient of variation (CV%): This is a percentage expression of standard
deviation (SD) with respect to its mean (eq. 4.10). It is mainly introduced
for carrying out not only comparison but also mathematical operations. No
doubt for that square of CV is used as it involves variance instead of SD in its
structure.
s
Coefficient of Variation (CV ) % = ¥ 100 4.10
X
Variance (V): It is given by the mean of deviations squares (eq. 4.11).
Variance =
Â( x - x 2 ) 4.11
n
Hence, it is equals to square of SD (eq. 4.9) becomes independent of
positive or negative value and thereby preferred for mathematical operations
like addition and subtraction.
Example: Ten fiber staple length measurement test data is evaluated for central
tendency and dispersion by different methods mentioned above. Results of
evaluation are given in Table 4.1.
Table 4.1 Measures of central tendency and dispersion of fiber length
Number Staple length (mm) Central tendency Dispersion
1 27.8 Mean = 28.38 Range = 2.2
2 27.9 Mean range = 0.4
3 28 PMR = 1.409
4 28 Median = 28.1 MD = 0.492
5 28 PMD = 1.734 = 0.64
6 28.2 CV% = 2.2
7 28.3 Mode = 28
8 28.6
9 29
10 30
It is apparent from the results that although test data set is same the central value
and dispersion values measured by different analysis techniques are not same.
Selection of suitable data analysis system is purely defined by degree of
accuracy required and it is attained well with the accommodation of entire
data set in root form only. According to definitions given for each system,
except average and SD (CV%) rests of the systems are dependent only on
few data or considering modifications like mode taken for MD, so accuracy
involved for these measures is quite low. This argument is well substantiated
by results shown in Table 4.1.
Regarding SD it is known that deviations from the average may be negative
as well as positive; squaring these deviations removes all negative values. The
mean puts the deviation on an average basis. The root brings the expression
back to linear basis from the power basis resulting from the squaring operation.
Thus, SD value is unaffected by positive or negative values. Thus it is more
accurate measure of dispersion as compared to others.
4.3 Compatibility test methods

Once test data is summarised in terms of central value and dispersion value by
using the most accurate technique, viz; mean and SD, they should be evaluated
for their compatibility in meeting specifications, the next step of yarn engineering.
Specification can be any; standard norms or the nominal values put forward by a
customer. There are two such techniques used classically in this regard, they are:
(1) Significance testing: This includes two types of tests: t-test for mean
and F-test for dispersion (Fig. 4.3 and 4.4). This method was introduced
by Ahmedabad Textile Research Association (ATIRA). It is mandatory
that mean value must be checked for its significance difference first
and if it is not differing significantly thence only F-test is conducted.
Apart from this, method of evaluation gets influenced by sample
size (small n < 50 or large n ≥ 50) as well as kind of comparison done
that is comparing single sample with standard (norm) or comparison
between two samples. These make system complicated and also there
is discrimination about number of observations (n) for large sample
size; some places n ≥ 30 considered in place of n ≥ 50. Altogether make
implementation of such technique complicated and time consuming.
(2) Critical difference (CD %): Critical difference technique was
developed by SITRA to check compatibility of either one cotton with
nominal value or two cottons at a time. It is a measure of the difference
between two values that arises solely due to natural or unavoidable
causes. When the difference between these two values exceeds the
critical difference, then the two values are said to be statistically

different. The method of evaluation is depicted in Fig. 4.5.
Critical difference depends upon:
a. Coefficient of variation (CV %)
b. Number of tests carried out to determine the quality characteristics (n).
Figure 4.3 T-test.
Figure 4.4 F-test.

Hence it is independent of sample size, undue complication of selecting
evaluation methodology as per sample size gets eliminated. In addition, it
has a provision to modify lay down standards as per actual number of tests
performed. All these make CD system more perfect and easy to operate.
However, similar to significance test, it can able to deal with one property
at a time for maximum two cottons in consideration. This makes statistical
evaluation voluminous and full of error for manual operating techniques.
Figure 4.5 Critical difference.
4.4 Statistical techniques for defining technological

value of cotton
It is always preferred to relate precisely raw material to end product in terms
of quality and cost. It is also equally important this should be based on the
scientific ground and done before starting with actual yarn production.
Simultaneously it is also a concern that production technology and foreseen
response of cotton used in this course must be well defined. This is important
to produce the yarn with required characteristics at an economical rate.
Thereby statistical techniques for deriving scientific correlation amongst three
variables: raw material, product and process should be established.
Classically this was done on judiciary ground that is by individual’s
expertise and past experiences based subjective decision-making. Thus in
absence of scientific statistical techniques such decision was always human
biased and many a times unable to meet precisely quality and cost demands
for the product yarn. These are mainly attributed to the use of unnecessarily
costly cotton or lack of fiber utilisation during conversion process.
Later stage development of testing technique has made measurement of

highly variable cotton characteristics possible. But, this middle age non-
automatic testing mode has conflicting standards in terms of different units of
measurements used for quality parameters. Apart from slow and tedious testing
approach in this time, it was bit difficult to check statistically compatibility
(significance test or critical difference) of two cottons for different quality
parameters. It is also a fact that response of different fiber characteristics at
the point of integration in yarn and during processing is different. This makes
the situation more complex, applying some quality measure which represents
the different properties of fiber in unique way can deliver a plausible solution.
In this regard ATIRA (Ahmedabad Textile Research Association) and SITRA
have put forward concept of quality index, Q index and FQI, respectively.
But they are partially relating fiber properties with one of the important yarn
property and that is also in broader sense.
The development of high speed and fully automated fiber testing instruments
such as the HVI and the advanced fiber information system (AFIS) has
revolutionised the concept of fiber testing. With the HVI, it is pragmatically
possible to determine most of the quality characteristics of a cotton bale within
two minutes. Data generated by these instruments can easily be utilised with
micro-computers and powerful software programs interfaced and also gave the
spinning technicians a single quality parameter SCI (Spinning Consistency
Index) incorporating the various fiber properties and a prediction expression
to estimate the yarn properties from the fiber properties. This has permitted
100% bales to be tested precisely and evaluating fiber characteristics right
from the gin to the point where formulation of mix takes place, thus, made
modern testing system much more reliable. This can play a pivotal role in an
engineered fiber selection programme.
Based on the HVI results, composite indexes such as the modified fiber
quality index (MFQI), premium discount index (PDI), multiplicative analytic
hierarchy process (multiplicative AHP) of the MCDM method and geometric
properties index (IG) are developed and used for determining the technological
value of cotton apart from SCI. Significance of AHP’s lies in the fact that it can
handle the objective as well as subjective factors, and the criteria weights and
alternative scores are elicited through the formation of a pair-wise comparison
matrix. All these indices are described in details in the forthcoming sections
along with their inbuilt advantages and disadvantages.
4.4.1 Drafting quality index (Q)
There are number of parameters to judge yarn quality but out of them two
major properties have given utmost importance, viz; strength and uniformity.
ATIRA has identified yarn uniformity as prime one which is greatly affected
by drafting parameters at ring frame. Process proficiency was judged on the

basis of difference between actual and theoretically predicted one.
The mathematical relationship between the drafting quality of cotton (Q)
and yarn unevenness (u) established and given by ATIRA was:
d - 1ˆ
u 2 = Q.Ne ÊÁ + a ( d - 1) + ur2
Ë d ˜¯ 4.12
Where, Ne = yarn English count, d = draft in ring frame, ur = roving

unevenness, a = contribution by ring frame to yarn unevenness and Q = drafting
quality defined as:
2
F
Q = 29.4 ÈÍ ˘˙ 4.13
ÎL˚
Here F = fineness/maturity coefficient, L = 50% span length.

Normally, “a” assumes a value of 3.5 under good working conditions.
4.4.2 Fiber quality index (FQI)
SITRA has proposed the measure known as the fiber quality index (FQI) by
combining the effects of fiber characteristics into a quality index. Important
fiber properties are combined according to their effects on a particular yarn
quality that is yarn strength (CSP). The main reason for this system to gain
popularity over quality index is attributed to the simplicity of the equation
used. Mathematical form for deriving FQI is as follows:
Lsm
Fiber quality index = FQI = 4.14
f
Where, L = 50% S.L. in mm (Digital Fibrograph), s = 3 mm G.L strength
in g/tex (Stelometer), f = fiber fineness (Micronaire value), m = maturity
coefficient (NaOH method)
Several variants of the FQI model are available as per testing mode. The
above mentioned equation is true for semi-automatic testing era but with the
introduction of HVI measures the same equation gets reshape as:
UHML.UI.FS
FQI HVI = 4.15
FF
Where, FQIHVI = HVI quality index, UHML = upper half mean length and
UI = uniformity index.
Mathematical correlation between index value and at least one of the

important yarn properties; that’s strength for FQI was established.
Lea CSP = 320 ( )

FQI + 1 - 13C ; Carded yarn
4.16
Or
W ˆ
Lea CSP = ÈÎ320 ( )
FQI + 1 - 13C ˘˚ ÊÁ1 +
Ë 100 ˜¯
; Combed yarn 4.17
Where, C = Count Spun and W = Percentage Noil

Similarly for HVI based measurement the relationships will take the
following shape:
Lea CSP = 280 ( )

FQI + 700 - 13C ; Carded yarn 4.18
W ˆ
Lea CSP = ÈÎ280 ( )
FQI + 700 - 13C ˘˚ ÊÁ1 +
Ë 100 ˜¯
; Combed yarn 4.19
This has also initiated practices for define process proficiency with respect
to only one of the important property of yarn. Thus, quality index has given
statistical tool to predict about process performance abided by actual yarn value
(strength) and theoretically calculated one from the known fiber parameters.
Actual yarn CSP

Process Proficiency (P%) = ¥ 100 4.20
Theoratically Calculated CSP
Process performance ratings were given on this ground accordingly;

P (%) ≥ 100; Good
= 95; Average
= 90; Poor
The fiber parameters required to spin known counts with different CSP
under good processing conditions (P = 100%) were defined in terms of FQI.
Fiber with higher FQI is regarded good in terms of quality parameters and
thereby costly, normally used to earn better yarn strength for a given count.
This was also a fact that the fiber with higher FQI was used for spinning
finer count to get equivalent strength to that of coarser yarn in the group
(Fig. 4.6).
Figure 4.6 SITRA estimates for FQI and yarn.

Thus with the introduction of fiber quality index concept a basis for
scientific evaluation of cotton was build up. It became a scientific tool for
defining quality and cost of the cotton used in the mix and also facilitated in
establishing statistical correlation amongst fiber- yarn-process, but no doubt
partially. Hence FQI was derived on the basis of only four fiber parameters
on linear relationship, which in practice is not fully true. In absence of any
other option, SITRA has suggested that the FQI and the yarn CSP estimations
should be taken as a broad guide line only by the mill to judge the cotton
quality as well as spinning performance as shown in Fig. 4.6.
4.4.3 Modified fiber quality index (MFQI)
In order to improve the possibility of predicting the yarn tenacity from the HVI
measure based fiber properties, the concept of modified fiber quality index (MFQI)
was put forward by Magdi et al. (2013). Mathematically it is represented as:
MFQI = [ UHM ¥ UI ¥ STR b ¥ (1 + EL) ¥ (1 - SF)] MIC] 4.21
Where, UHM = upper half mean UHM in mm, UI = % stands for the fiber
length uniformity index, STRb = cN/tex the bundle tenacity, EL = % the fiber
elongation at break, MIC = Micronaire value, SF = % for the short fiber content.
Since MFQI is taking into consideration the variability in the fiber length, it
represents the specific work of rupture or the ability of a material to withstand
a given energy. In order to compare different materials specific work of rupture

should be really used to calculate the amount of energy needed to break a
material of unit mass. Thus, MFQI is more consistent as well as reliable for
the comparison between different types of cotton as compared to FQI. Thereby
index MFQI is used for the prediction of yarn tenacity, which is a function of
not only the fiber tenacity but also its length, its short fiber content, its fineness
and its elongation at break, the properties should be taken into consideration
while predicting the tenacity of yarns. Magdi et al. (2013) while working with
different varieties of Egyptian cotton has found a high correlation factor of 0.96
between the MFQI and yarn tenacity. This high correlation has justified the above
mentioned argument. Thus MFQI is not only more effective in characterising
fiber quality but also in predicting yarn tenacity. The general regression equation
for the prediction of yarn tenacity (Sy) as a function of MFQI was found to be:
Sy = 0.041 MFQI + 5.798 4.22
4.4.4 Spinning consistency index (SCI)
Spinning consistency index (SCI) is another HVI measure based cotton fiber
index (eq. 4.23) proposed by Uster. But it is calculated based on a multivariate
regression equation and proven useful index for predicting the overall fiber
quality as well as spinnability. The higher precision found with SCI is attributed
towards the involvement of all HVI properties as per their contribution on the
yarn quality and spinnability. Hence, it is derived with large data collected
from a five-year crop average of U. S. Upland and Pima cotton.
SCI = -414.67 + 2.9FS + 49.17 UHML

4.23
+4.74UI - 9.32FF + 0.65Rd + 0.36 ( + b )
Where, Rd = reflectance degree and +b = yellowness of cotton fiber.

For HVI calibration mode:
SCI = - 414.67 + (2.9Strength) - (9.32Mic.)

+ 49.178Length in inch) + (4.74Uniformity Index) 4.24
+ 0.65Rd + 0.36 ( +b )
If no colour module is installed then SCI equation is as follows:
SCI = 322.98 + ( 2.89Strength ) - (9.02Mic.)

4.25
+ ( 49.17Length ) + (4.29Uniformity Index)
Normally cotton with higher SCI is treated to be of good quality and capable
of producing stronger yarn. Hence such cotton is having better spinnability
and thereby costly.
4.4.5 Premium discount index (PDI)
Premium discount index (PDI) was proposed by EI Mogazhy et al. (1990).

According to him, the actual value of the cotton should be determined based
on those inherent characteristics that contribute to the defined yarn quality.
The difference exist in this system as compared to SCI is that it deals only with
the desirable properties of the cotton or those that allow the production of the
best selected quality of yarn in a particular spinning system. Thus, PDI value
adheres not only to particular system but also yarn quality in consideration.
The PDI is based on the values of percentage relative contribution of the
fiber property (Ci %) and the respective difference factor (Di). The generalised
mathematical form of PDI can be given as:
k
PDI = Â(Ci Di ) 4.26
i =1
Where, k = the number of the HVI fiber properties.

But while considering it for specific yarn property say tenacity it will take
the shape as follows:
Yarn Tenacity = C1 + C2 FS + C3 FE + C4 UHML

+C5 UI + C6SFC + C7 FF 4.27
Where, C1, C2 … C7, = regression coefficients, FS = fiber strength, FE = fiber

breaking elongation in percentage, UHML = upper half mean length, UI =
uniformity index for fiber length, FF = fiber fineness and SFC = short fiber
content as measured by AFIS.
This generalised format was put forward by Majmudar et al. (2005) based
on tested fiber and yarn data.
The percentage relative contribution of the ith fiber property is determined
by the following equation:
Ê B ˆ
Ci = 100 Á k i ˜ R 2 4.28
ÁË Â Bi ˜¯
i =1
Where, Bi = standardise coefficient of ith variable, k= number of HVI fiber

properties and R2= coefficient of determination.
Hence relative contribution coefficients are dependent on the scales of

measurement, and therefore cannot be used as a measure of their importance.
To overcome this problem, coefficients of the variables are determined using
the standardised variables in the regression equation.
The reference set is used and expressed in terms of the average and SD of a
fiber property. Then the relative difference factor (Di) for each cotton fiber is
determined by the following equation:
( xi - mi )
Di = 4.29
si
Where, xi = ith fiber property of a cotton, μi= overall average and si = over all
standard deviation of all the cottons in the ith property.
EI Mogahzy et al. (1990) combined trash content and short fiber content,
and stated processing performance index I as:
1
I= 4.30
(TC ) ( SFC )
Where, TC = trash content (by weight) and SFC = short fiber content.
4.4.6 Multi criteria decision-making (MCDM)
Based on the HVI results, different indices like FQI and SCI are used in
spinning industries to determine the quality value of cotton. However, in
case of FQI, all the cotton fiber properties are considered to have equal
importance but practically this assumption is invalid. Conversely, SCI has
been developed by United States Department of Agriculture (USDA) based
on their research on Upland and Pima cotton. Although it is claimed that SCI
possesses good association with yarn strength, neps and end breakage rate
during spinning, the correlation between the SCI and yarn tenacity is not
found encouraging. This is mainly due to conflicting performance score of
cotton fibers in terms of different properties. For example, a variety of cotton
fiber may be stronger but coarser than another variety. Therefore, the grading
or ranking of cotton fibers in terms of different quality criteria will certainly
not be the same. This will make the situation complex and application of
multi criteria decision-making (MCDM) methods can probably deliver a
plausible solution. The solution must produce a comprehensive index of
quality by incorporating important cotton fiber properties. The weights of
the cotton fiber properties should be commensurate with their importance
on the final yarn quality.
MCDM is a branch of operations research, which deals with decision

problems involving a number of decision criteria and a finite number of
alternatives. In this problem, cotton fiber properties are the decision criteria
and types of cotton available for grading are the alternatives. It provides a
useful framework for measuring and evaluating sustainability performance.
Various MCDM techniques, such as the weighted sum model (WSM), the
weighted product model (WPM), the analytic hierarchy process (AHP), the
revised AHP, the technique for order preference by similarity to an ideal
solution (TOPSIS), and elimination and choice translating reality (ELECTRE),
can be used in depending upon the complexity of the situation .The analytic
hierarchy process (AHP), introduced by Saaty (1980) is one of the most
frequently discussed methods of MCDM.
Multiplicative analytic hierarchy process (MIAHP)
Barzilai & Lootsma have proposed a multiplicative variant of AHP.
According to the multiplicative AHP model, the equation to calculate the
technological value of cotton (MIAHP) is as follows:
FS0.27 FE 0.039 UHML0.291UI0.145

MIAHP = 4.31
FF0.11SFC0.145
Where, FS = fiber strength, FE = fiber elongation, UHML = upper half
mean length, UI = uniformity index, FF = fiber fineness, SFC = short fiber
content.
4.4.7 Geometric properties index (IG)
Korickij (1983) introduced the index based on HVI measures known as

geometric properties index (IG).
Ê 1 - SF ˆ
IG = 0.1 Lm ¥ UI ¥ Á ¥ MAT ¥ FI -0.5
Ë 100 ˜¯ 4.32
Where, Lm = cotton fibers weighted mean length, FI = fiber fineness, MAT

= maturity.
For HVI measured properties can be IG expressed as:
UHM ¥ UI ¥ (100 - SF)

IG a = 4.33
10000 ¥ MIC
Where, UHM = upper half mean length, UI = uniformity index, SF = short
fiber, MIC = micronaire value representing the fiber fineness and maturity.
4.5 Statistical techniques for defining proportion of

cotton constituents in mix
Once technological value of the cotton is defined, its suitability for the
particular product (yarn) as a raw material gets verified. Now the next step
for scientific decision-making is to determine proportion of suitable cotton
components in the mix. This should be decided in such a way that mix so
formulate must be able to produce yarn with desired quality at minimum
possible cost. The fundamental problem here is to produce a specified cotton
yarn by feeding an appropriate mixing of raw cottons into the opening-and-
blending machine at the beginning of the yarn production process. In order
to produce a least-cost mix, the producer must consider a complex variety
of factors including the prices, grades, and availabilities of a great many raw
cottons. The crucial interrelations among the several qualities of raw cotton
make it exceptionally difficult to determine a least-cost mix.
4.5.1 Judicious mixing
Classical spinning system was expert decision based system and had no
provision for such statistical techniques. The answer to the critical cotton
selection problem from large number of options was to formulate a relatively
simple mix ratio for a specified yarn, which is then empirically tested and
corrected. This was often resulted in the undue addition of expensive high-
quality grades. This method frequently results in production delays and
quality giveaway. Further, the initial difficulty of determining a proper blend
is vastly compounded by common fluctuations in the availability and price of
specific cottons, since the alteration of anyone component will alter all the
relationships required for least-cost production.
4.5.2 Linear programming
Computer data program was developed by ATIRA for resolving this complex
problem by the use of linear programming (LP) model. It is a numerical
computation of mixing quality and cost, in the form of linear equations or
inequalities known as constraints, of all known and estimated factors relevant
to the production of the specified mix. A constraint relates the value of some
particular quality factor (or other factor, such as cost) in the final blend to the
amounts of each raw material used in the blend, and expresses the final blend
specification for this factor.
This is a mathematical approach for maximising or minimising a linear
objective functions that specifies the benefits of quality or cost associated
with each decision variable. Hence cotton mix formulation involves many
decision variables and many constraints in this optimisation process, the
so called “simplex method” can be used to perform optimisation analysis.

This is an iterative procedure that progressively approaches and ultimately
reaches an optimal solution to linear programming problems. Computerised
linear programming routines automatically arrange the inputs of the problem
(objective functions and constraints), perform the iterative analysis and
produce an output of the LP program.
The following set of linear equations was put forward in order to set the
exact proportion of each component for the mix and with their respective
price. Mathematically, these are fiber characteristics constraints which are
expressed by equations or inequalities, depending on the nature of constraint.
In establishing these constraints two basic assumptions are made: linearity
and additively. These assumptions may be valid from physical knowledge or
they may be verified using modelling analysis.
n
Emix = Âei pi' ≥ Es 4.34
i =1
n
M mix = Âmi pi' ≥ M s 4.35
i =1
n
Smix = Âsi pi' ≥ Ss 4.36
i =1
n
X mix = Âxi pi' ≥ Xs 4.37
i =1
1 n
p' 1
=Â i ≥ 4.38
Fmix i =1 fi Fs
The obvious constraint here is that the sum of proportions is unity or 100%.
n
Â pi' = 1 4.39
i =1
It was based on four important fiber mix properties viz; effective length
(E), mean length (M), bundle fiber strength (S) and fineness (F). Additionally
maturity (X) was also considered as Mc is the output of fineness and maturity.
Where ei, mi, si, f i and xi are fiber values for ith type and Es, Ms, Ss, Xs
and Fs are the corresponding threshold (norms) values desired for these
characteristics. Quality parameters of mix have been calculated as a weighted
average of proportion (p in kg) of respective property of cotton in mix. Quality

value of mix given on right hand side should be either equal or higher than
respective standard value placed on left hand side of equation.
Similarly cost of mixing (C’mix) is given in terms of Rs/kg. Clean cotton
cost (C’) instead of raw cotton (bale cotton) cost (C) is used for the purpose.
This is important as % trash content of bale cotton varies from cotton to cotton
and cannot be realized to yarn. So, yarn cost is related with the cost of clean
cotton in the process in reality.
n
Âci' pi' ≥ C'mix 4.40
i =1
100 ˆ
Cost of clean cotton C ' = C ¥ ÊÁ
Ë 100 - t ˜¯ 4.41
Accordingly proportion of cotton in the mix must be related with clean

cotton rather then bale cottton, which can be evaluated as:
100 - t ˆ
Proportion of clean cotton p ' = p ¥ ÊÁ
Ë 100 ˜¯ 4.42
But procurement of the cotton is done in the bale form only. So, it is needed
to define number of bales used in mixing per unit time for given component in
the mix which can be calculated as:
Proportion of bale cotton for the given component in the mix
Proportion of component in the mix in bale cotton form
= 4.43
Totaal amount of cotton in mix in bale cotton form
Where, proportion of bale cotton for the given component in the mix P1 is
given by
100p1’
100 - t1
p1 = 4.44
100p1’ 100p’2 100p3’ 100p’n
+ + + .. +
100 - t1 100 - t 2 100 - t 3 100 - t n
Basic constraints of L.P. mix formulation:
(1) All fiber properties must be known, especially trash content.

(2) Standards/requirements for fiber quality parameters must be known
for type of yarn to be produced.
(3) Initially “Trial” run of L.P. mix up to fabric finishing stage must be
carried out to verify its significance.
(4) Price for each components and their availability must be known.
(5) Additional constrains based on technological judgment may be put on
the usage of special cotton in terms of their proportion in mix.
Example: p4’ ≤ 0.20; as it is difficult to control fibers during process,
adds to irregularity. Decided on the basis of past experience. p1’ ≤ 0.08;
as costly component, not economical for the on going count.
(6) Determine L.P. mix afresh once every month, because fiber properties,
trash content and prices of the fresh lots may not be identical to those
used for working out L.P. mix. Also not all of the cottons used earlier
would always be available.
Requirements for successful utilisation of L.P. technique:
(1) Since knowledge of all fiber properties is needed, increases testing

load. With non-automatic or even semi automatic testing technique
it becomes difficult, because testing was slow, tedious and time
consuming liable to add in testing cost, time as well as human
associated errors in L.P. mix. These restrictions can be well overcome
by the use of fully automated, speedy and precise testing techniques
like HVI and AFIS as explained in Chapter 3.
(2) The mix quality parameters for achieving this target are defined purely
on past 18–24 months experiences of mills or in absentia, norms are
adopted and then modified in the light of experience
(3) Quite a more number of trial are required for optimisation of mix
values for getting desired performance. Adds to the cost and wastage.
4.5.3 Fuzzy linear programming approach
Linear programming method earlier used the simplex method has been
mathematically stabilised and permits sensitivity and parametric analyses.
Today the fuzzy logic technique is also used for solving it. The concept of
fuzzy linear programming was first proposed by Bellman and Zadeh (1970)
thereafter, many researchers have made significant contributions in the
development of fuzzy linear programming.
The dominance of fuzzy linear programming for the cotton mixing problem
is attributable to its ability of handling the situation where there are no exact
defined boundaries for the inequality constraints. For example, a spinner
often uses the terms such as low and high to assess the fiber strength, length,
fineness and SFC etc., however, these terms do not constitute a well-defined
boundary. All these fiber properties have approximate boundaries rather

than exact boundaries. In addition, the membership functions for the various
constraints could be developed by deciphering the experience of a spinner
which renders a better utilisation of the fiber properties needed for a particular
mixing. Thus, fuzzy linear programming approach of selection of the cotton
mix has an enormous scope for industrial acceptance in future.
References
1. Bellman R.E., Zadeh L.A, 1970, Decision making in a fuzzy

environment. Management Science, 17, 141–164.
2. Belton V., Kang B.C., Park S.W., Koo Y.J., and Jeong S. H., 2000, A
simplified optimization in cotton bale selection and laydown. Fibers
and Polymers, 55–58.
3. Booth J.E., Principles of textile testing, CBS Publishers and distributors.
4. Chanas S., 1983, Fuzzy sets & Systems. 11, 243–251.
5. Das Subhasis, Ghosh Anindya, 2015, Optimization of cotton mixing
cost: a fuzzy linear programming approach. ACC Journal, 21(1), doi:
10.15240/tul/004/2015-1-003
6. Elamvazuthi I., Ganesan T., Vasant P., et al., 2009, International Journal
of Computer Science and Information Security, 238–243.
7. Garde A.R. and Subramanian T.A., Process Control in Spinning,
ATIRA publication, third edition, 1987.
8. Grover E.B., and Hamby D.S., Handbook of Textile Testing and Quality
Control. Wiley India Pvt. Ltd.
9. Korickij K.I., Technological economic estimation and design of textile
materials quality, Legkaja industria, Moscow (in Russian).
10. Lord E., 1961, Manual of cotton spinning: The characteristics of raw
cotton, The Textile Institute, Manchester and London, pp 310–311.
11. Magdi El Messiry et al., March 2013, Characterization of Egyptian
cotton fibers, Indian Journal of Fiber & Textile Research. 38,109–113.
12. Majmudar Abhijit, Majmudar Prabal Kumar & Sarkar Bijan, June
2005, Determination of the technological value of cotton fiber: a
comparative study of the traditional and multiple-criteria decision-
making approaches. AUTEX Research Journal, 5(2), http://www.
autexrj.org/No2-2005/0126.pdf 71
13. Majumdar Abhijit, Das Apurba, R., Alagirusamy, Kothari V.K., Process
Control in Textile Manufacturing, Woodhead Publishers.
14. Militky Jiri and Kremenakova Dana, Comparision of complex indices
for cotton fiber quality characterization.
15. Mogazhy EL Y.E., Broughton R., and Lynch W.K., 1990, A statistical
approach for determining the technological value of cotton using HVI
fiber properties, Textile Research Journal, 60(9), 495–500.
16. Mogazhy EL Y.E., and Gowayed Y., 1995, Theory and practice of
cotton fiber selection, Part I: Fiber selection techniques and bale
picking algorithms, Textile Research Journal, 65(1), 32–40.
17. Mogazhy EL Y.E., and Gowayed Y., 1995, Theory and practice of
cotton fiber selection, Part II: Sources of cotton mix variability and
critical factors affecting it, Textile Research Journal, 65(2), 75–84.
18. Norms for the spinning mills, 1995, The south Indian textile research
association, 1–17.
19. Saaty T.L., 1980, The Analytic Hierarchy Process, McGraw-Hill
International, New York.
20. Sreenivasa Murthy H.V., and Samanta S.K., 2000, A fresh look at fiber
quality index. The Indian Textile Journal, 111(3), 29–37.
21. Verdegay J.L. Fuzzy mathematical programming. In: Gupta M.M.
and Sanchez E. (eds.) Fuzzy Information and Decision Processes.
Amsterdam, 1982.
22. Verdegay J.L., 1984, Fuzzy Sets & Systems, 1, 131–141.
23. Werners B., 1987, European Journal of Operations Research, 31, 342–
349.
24. Zimmermann H.J., 1978, Fuzzy Sets & Systems, 1, 45–55.
25. http://www.statcan.gc.ca/edu/power-pouvoir/ch11/median-
mediane/5214872-eng.htm seen on 11 December 2015.
26. http://www.tekstilvekonfeksiyon.com/en/measurement-of-
sustainability-performance-in-textile-industry-by-using-a-multi-
criteria-decision-making-method-524.html as seen on 28/12/15.
27. http://www2.le.ac.uk/offices/ld/resources/numerical-data/variability
seen on 11 December 2015.
5
Artificial neural networking (ANN)


*
Abstract
Artificial neural networks (ANNs) are a family of statistical learning models
inspired by biological neural networks (BNN). The common thing about both
of them is that they are used to estimate or approximate functions that can
depend on a large number of inputs and are generally unknown. ANNs are
generally presented as systems of interconnected “neurons” which exchange
messages between each other just like central nerves system of animals. The
connections have numeric weights known as synaptic weights, that can be
tuned based on experience, making neural nets adaptive to inputs and capable
of learning. Such system can serve well when used as a replacement for human
brain in addressing multivariate complex problem of yarn engineering.
Key Words: artificial neural network, neuron, synaptic weight, brain, learning,
functional unit, transfer function
5.1 Introduction
In machine learning and cognitive science, artificial neural networks (ANNs)
are a family of statistical learning models inspired by biological neural
networks (BNN). The central nervous systems of animals, in particular
the brain, are used to estimate or approximate functions that can depend on
a large number of inputs and are generally unknown. ANNs are generally
presented as systems of interconnected “neurons” which exchange messages
between each other. The connections have numeric weights known as synaptic
weights, that can be tuned based on experience, making neural nets adaptive
to inputs and capable of learning.
The process of cotton fiber selection for formulating the mix and getting
desired quality and cost of yarn by processing it on a given set up of machinery
is a critical job. The things becomes further crucial with a natural cotton
fiber, having inbuilt inherent high variations (within and between the lots),
and processing such material in a controlled manner on production line also
having too many variables, liable to influence product quality and cost largely.
A variety of approaches have been proposed to provide effective solutions to
these problems, which can be broadly categorised into classical approaches
and artificial intelligence (AI) approaches. The classical approach here
refers the stage later on to art (experience) based decision-making, i.e. non-
automatic instrumental era. It involved the use of an appropriate mathematical
description of the decision-making problem, and also solved through a
traditional mathematic manually. This was full of human error as well as time
consuming and tedious process.
On the contrary, AI is the branch of computer science focusing on creating
machines that can engage in behaviour that human consider intelligent. AI
techniques exhibit heuristic and intelligent natures, which have the potential
of providing superior solutions over classical techniques. Thereby over the
last more than one decade, AI has received more and more response from
the researchers to gain scientific solution to the complex textile industrial
problems which were resolved earlier by judiciary decision (Fig. 5.1).
Figure 5.1 Summary of textile industry using ANNs for solving problems.
Today, ANN is being applied to increasing number of real world problems of
considerable complexity. The advantage of ANN lies in their resilience against
Artificial neural networking (ANN) 97
distortion in the input data and their capability of learning. They are often good
at solving problems which are too complex for conventional technologies e.g.
problems that do not have a conventional algorithmic solution.
5.2 Historical background for the development of ANN

Human brain is the most dynamic feedback system in the universe and one can
fail to understand the power of computations that the human brain performs in
our day-to-day life. The basic unit of such a powerful and massively parallel
brain is Neuron. A neuron is a complex biochemical and electrical signal
processing system that receives and combines the signal from many other
neurons through various paths and dendrites. A simulation between biological
and artificial neuron structure is shown in the Fig. 5.2.
Figure 5.2 Simulation of basic structure of biological and artificial neuron.

Human brain has some 100 billion neurons, averaging to 10,000 synapses
per neuron and thus brain consists of 1000,000,000,000,000 (or 1015) synapses.
Thus, one can imagine such a complex network of neurons makes various
computations which are of non-linear, linear and random type a possible
job. This very fact leads to genesis of a new model which makes it possible
to replicate such kind of complexity in artificial systems. This leads to the
development of ANN.
5.3 Basic concept of ANN (Artificial Neural Network)

The ANN is information processing system and inspired from biological
neurons. The function of ANN in simpler words is to give an output pattern
when presented with an input pattern. An ANN is a system of interconnected
neurons in parallel, whose primary job is to estimate or approximate a function
which is of non-linear and random type or has a large number of inputs. These
neurons mime the characteristic of biological neurons and are intended to
abstract and model some of the functionality of human nervous system.
Another definition for functionalities of interest for textile engineering
can be given as: ANNs are algorithmic structures derived from a simplified
concept of the human brain structure. They belong to the soft computing
family of methods, along with fuzzy logic/fuzzy control algorithms and
genetic algorithms. They all share an iterative, non-linear search for optimal
or suboptimal solutions to a given problem, without the presupposition of a
model of any type for the underlying system or process.
On generalising the definition “ANN is a set of processing units when
assembled in a closely interconnected network, offers a rich structures
exhibiting some features of the biological neural network.” The ANN is
a massively parallel distributed processor made of single processing units,
which has a natural propensity for storing experiential knowledge and make it
available for use. The procedure used to perform the learning process is called
as learning algorithm, the function of which is to modify the synaptic weights
of the network in an orderly fashion to attain a desired design objective. In
practice, the ANN cannot provide the solution by working individually; rather
it needs to be integrated into a consistent system engineering approach.
5.4 Types of neural network

As the name suggests neural network is composed of a net of neurons. These
neurons are organised in the form of layers in a network. An ANN consists of
two basic layers, viz; an input layer of neurons (nodes or units), and a final
output layer of neurons. The input neurons are designated to receive external
stimuli presented to the network. The input neurons perform no computation
but just transfer the signal. Outputs from the network are generated as signals
of output Neurons. Some artificial neural networks also possess one or more
hidden layer of neurons. Such hidden neurons compute intermediate function
and their state are not accessible to the external environment.
In general, the ANN architecture is characterised by:

1. A large number of simple, neuron-like processing elements. They
are organised in a way that defines the network topology. Based on
topology networks they are categorised into three basic groups:
a. Single layer feed forward networks,
b. Multilayer feed forward networks and
c. Recurrent networks [Figures 5.3 (i-iii)].
2. A large number of weighted connections are used between these
elements (neurons). The weights of these connections store the
“knowledge” of the network. The adaptive adjustment of the weight
values, while moving down an error surface to a minimum point,
constitutes the “learning process” of the network. According to mode
of learning, network can be bifurcated into two main groups;
a. Supervised learning
b. Unsupervised learning.
3. Highly parallel and distributed control;
4. An emphasis on learning internal system representations automatically.
Figure 5.3 Types of neural networks.

Apart from the topology and node characteristics (type of activation

function, etc.), an ANN model is specified by the training rules or training
algorithm employed to adapt its weights. These rules define an initial set of
values for the weights and indicate how weights should be iteratively adapted
to improve the performance of the network by minimising the error between
actual and ideal outputs, when the network is presented with a set of known
inputs. A variety of different network models have already been proposed and
used in practical applications, such as Perceptron, the Multilayer Perceptron, the
Hopfield network, the Carpenter-Grossberg classifier, the Back-propagation
network, the Self organizing Map, the Radial Basis Function Network, the
Probabilistic Network, etc.
5.4.1 Single layer feed forward network
This is a simplest form of a layered network. In this category an input layer of

source nodes projects onto an outer layer of neurons (Fig. 5.3 i). The network
is strictly feed forward, i.e. unidirectional operation, and thereby characterised
as single layer feed forward network.
5.4.2 Multilayer feed forward network
As per name suggests it consists of multilayer network. Apart from input

and output layers, there is always a presence of one or more hidden layers
whose nodes are called as hidden neurons (Fig. 5.3 ii). Their function is to
intervene between external input and network output in some useful manner.
By adding more hidden layers, the network is enabled to extract higher order
statistics. The source nodes in input layer supplies respective element of
activation pattern, which constitutes input signals applied to neurons in first
hidden layer. The output signals of the first hidden layer are used as input to
the second hidden layer and so on. The set of output signals of the neurons
in the final output layer of network constitutes the overall response of the
network to the activation pattern supplied by the source nodes in the first
input layer.
5.4.3 Recurrent network
The recurrent neural network is distinct in the sense that, it has at least one
feedback loop. A recurrent network may consist of a single layer of neurons,
with each neuron feeding its output signals back to inputs of all other neurons.
The recurrent network may or may not have hidden neurons. The presence of
feedback loops has a profound impact on learning capability of the network
and its performance.
5.4.4 Learning of a network
A network can be learnt with or without training. The objective of learning is

to mold the mapping surface according to desired response. Once the network
is structured for a particular application, that network is ready to be trained. To
start with, the initial weights are chosen randomly. There are two approaches
to learning: supervised and unsupervised.
5.4.4.1 Supervised learning
In these systems both input and output data are provided to network. Then it
processes the inputs and compares the resulting outputs against the desired
outputs. Errors are propagated backwards through the system, causing system
to adjust the weights. The same set of data is processed many times, as the
weights are refined. When the system has been fully trained, no further
learning is required and the weights can be frozen. A vast bulk of networks
utilises supervised learning.
5.4.4.2 Unsupervised learning
This type of network is provided only with inputs. The system itself decides
which features to be used to group the input data. It is also called as adaptive
learning. This learning technique is not commercially practiced yet.
5.5 Architecture of ANN

The structure or “architecture” of an ANN contains a number of nodes
(neurons), organised in a number of layers and interconnected to form a
network. Neurons are rather simple algorithmic structures that can perform
parallel computation for data processing and knowledge representation.
Weighted averaging followed by linear or non-linear threshold operations,
with the possibility of feedback between layers, constitute the main processing
operations in an ANN. The acquired knowledge is stored as the weight values
of the nodes. The real power of this model, therefore, lies in the mesh of
simple but highly interconnected nodes rather than the power or sophistication
of each node. Although inspired from the way neurons are interconnected in
the human brain and nervous system to transfer messages by chemo-electrical
procedures, no further analogy between ANNs and the human brain operation
is claimed as to the functionalities achieved.
Example: The architecture of a multi-layer perceptron (MLP) for a sample
ANN is shown in Fig. 5.4. The artificial neurons receive a set of “n” inputs
(nodes) xi (where i = 1 to k), each representing the output of another (previous
layer) neuron. The inputs are collectively referred to as X. Each input signal
is weighed by traversing through weighed pathways wij in order to generate an
activation function “∑”. In addition to this, artificial neuron has a bias term, a
threshold value θ (theta) that has to be reached or extended for the neuron to
produce a signal. The resulting scalar value ( ni1 ) (undergoes a transformation
by a generally non-linear activation function g ( ni1 ) to yield corresponding ith
output for all the input nodes. This resulting function ( ni1 ) will be the input
to the activation function g ( ni1 ) or a non-linear function for the multilayer
set up and again processed in the similar fashion for earning the next layer
output ( ni2 ) . The cycle goes on repeating depending on number of hidden
layers present in the structure to yield final output (yi) (Fig. 5.4). Normally
sigmoid, log-sigmoid, hard-limiter, hyperbolic tangent or even linear types
of functions are employed as an activation function, resulting in accordingly
varying properties of the network.
Figure 5.4 Architecture of multilayer perceptron neuron model.
5.6 Designing the network

For the known input and output data, finding the best MLP network is
formulated as data fitting problem. The parameters to be determined are
weights wij and threshold values θ. There exist various algorithms to find
out these parameters. The procedure adopted for designing the network is as
follows:
1. First of all the structure of the network is defined. This includes

selection of activation functions and the network parameters, weights
and biases, are initialised.
2. The parameters associated with the training algorithm like error, goal,
maximum number of epochs (iterations), etc., are defined.
3. The training algorithm is called learning process. There exist two
different learning processes, supervise learning and unsupervised
learning. As mentioned before, in supervise learning; comparison is
made between actual outputs and desirable outputs supplied by training
data, whereas there is no external trainer in unsupervised learning. The
training of MLP network proceeds in supervised manner also known
as back-propagation algorithm, as well as the generalised delta rule. It
is the most commonly used training method for the ANN model.
4. After the neural network design has been fully defined, its performance
is tested by simulating the output of the neural network with the
environment in which the network is embedded. Here actual output
data is compared with the measured outputs.
5. Final validation must be carried out with independent data.
All these procedures can be carried out in MATLAB, C++ and Java.
Discussing in depth the various algorithms and commands to be used for this
is outside the scope of this book.
5.7 Operational mode of ANN

Operation mode can be split up into following steps:
(1) Training the network

(2) Verification testing
5.7.1 Training the network
Initially once ANN is designed it required to undergo a training phase. The

step includes a rich set of “training set” presented to the network input.
Training sets are the empirically derived examples matched in pairs of input–
output values, normally obtained from the back stuff records of the respective
process. The weights in the nodes of the network are iteratively optimised by
a gradient type algorithm so as to produce correct outputs for all inputs in the
training set. As iterations proceed, the algorithm leads to sets of weights that
minimise the error between ideal and actual ANN outputs – usually, in the
least squares sense. On getting desired level of accuracy training of network
is ceased.
5.7.2 Verification testing
On getting trained, the network enters the second testing phase. Here, it is
asked to produce outputs for unknown inputs presented to it. Outputs are
known as decision of ANNs. Such decisions are made on the basis of the
learning of (experience earned by) the network over the training set. During
the test phase, the network is expected to generalise successfully. Generalise
means it should make use of the experience stored in its elements during the
training phase in order to make correct decisions on incoming data that are
similar but not identical to the training set data. Satisfactory generalisation
implies a correct choice of the network type in relation to the problem at
hand and a successful training phase. Successful training makes the sense that
the training set was rich and informative enough to represent adequately the
space of the input vectors, and the training algorithm was allowed an adequate
number of iteration to converge.
Major reasons for non-convergence of the training algorithm lies in
following attributes:
1. The choice of an inappropriate ANN structure (e.g. one with too many/
not enough nodes or too many/not enough layers of nodes), and
2. The unavailability of a rich training set, which means that more data
or measurements are necessary to solve the problem at hand with least
error.
The second situation is more crucial in practice, as it is not always
straight forward how more data are to be obtained or measured in real field
applications. Mainly resistance arises from the privacy policies of industries
in the field. The first attribute, on the other hand, can be efficiently addressed
by simulations during the design phase of the whole application.
5.8 Applications areas of ANN

Functionalities of interest for any engineering ANNs are algorithmic structures
derived from a simplified concept of the human brain structure. They belong
to the soft computing family of methods, along with fuzzy logic/fuzzy control
algorithms and genetic algorithms. They all share an iterative, non-linear
search for optimal or suboptimal solutions to a given problem, without the
presupposition of a model of any type for the underlying system or process.
Various different ANN types are listed below have already been successfully
employed in a wide variety of application fields.
1. Feed forward neural network

2. Radial basis function (RBF) network
3. Kohonen self-organising network

4. Learning vector quantisation
5. Recurrent neural network
6. Fully recurrent network
7. Hopfield network
8. Boltzmann machine
9. Simple recurrent networks
10. Echo state network
11. Long short term memory network
12. Bi-directional RNN
13. Hierarchical RNN
14. Stochastic neural networks
15. Modular neural networks
16. Committee of machines
17. Associative neural network (ASNN)
18. Physical neural network
19. Holographic associative memory
Broadly these application areas can be classified into five groups, they are:
(a) Function approximation or regression analysis: This includes time

series prediction, fitness approximation and modelling.
(b) Classification: This category includes pattern and sequence
recognition, novelty detection and sequential decision-making.
(c) Data processing: This category incorporates filtering, clustering,
blind source separation and compression.
(d) Robotics: This includes directing manipulators, prosthesis.
(e) Control: This includes computer numerical control.
5.9 ANN applications in the field of textile engineering

Since 1990 the applications of ANNs in the textile engineering field have
become more and more popular (Fig. 5.1). Gradually it was proven that they
can address successfully complex engineering problems. Many researchers
have turned to ANNs when they were in front of a multi-parameter and non-
linear problem, without an obvious or straightforward analytical solution.
The implementation of the ANNs presumes an initial phase of features

extraction, which will be used later to feed the ANN. It includes the processing
of the given data or measurements, typically in the form of a signal, an image
etc. It has allowed covering up wide application areas in all five categories to
ANN in textile field. A very brief overview of such uses of the ANNs in the
basic sectors of textile engineering, viz; fiber, yarn, fabric and garment are
given here as a ready reference.
5.9.1 Fibers
Some of the major ANN application in fiber sector are summarised below.
i. An ANN is used for the prediction of copolymer composition very

correctly, as a function of reaction conditions and conversions. Using
this technique, the free radical copolymerisation of acrylamide with
quaternary ammonium cationic co-monomer diethylaminoethyl
acrylate (DMAEA) has been investigated in inverse emulsion. The
copolymer composition is determined from residual monomer
concentration.
ii. Classification of the animal fibers is one of the most typical problems.
It has been resolved successfully using ANNs. It is used to classify
two kinds of animal fibers: Merino and Mohair. The model extracting
six scale parameters with image processing and other using an
unsupervised neural network to extract features automatically, which
are determined in accordance with the complexity of scale structure
and the accuracy of the model (Fig. 5.5).
iii. Cotton trades in India assess the commercial value of cotton on the
basis of “grade” assigned by hand and eye methods by professional
cotton classers. Thus in the case of cotton, unbiased measurement of
colour, a subjective entity is tough by classical approach. ANN has
offered a reliable solution by replacing the existing subjective grading
with the objective one. By this way, cotton grading can be made more
precise by avoiding human biasness. Research is going on in this
direction now a day.
The properties of cotton measured accurately by using ANN are colour
of cotton, trash present and ginning preparation. Thus, ANN is used
not only for the grading of the colour of the raw fibers but also for the
classification of the cotton lint.
Production point of view, ANN based method for the selection of cotton
bales for meeting certain quality and cost requirements criteria, was
established. Apart from this, ANN approach was used in developing a
system for the optimisation of the yarn production based on the blend
characteristics and the process parameters.
iv. ANNs have supported the identification of the production control

parameters and the prediction of the properties of the melt spun fibers
in the case of synthetic fibers.
v. ANNs have been used in conjunction with NIR` spectroscopy for the
identification of the textile fibers. Thereby the blend made out of two
different kinds of fibers, the exact proportion in which the mixing
has to be done to obtain the required properties of the yarn can be
predicted by using ANN.
Figure 5.5 Intelligent fiber classification system.

5.9.2 Yarn
Some of the major ANN Application in yarn sector are summarised below.
1. The examination of the image of the web produced by a carding

machine and the detection of faults has been made possible in an
automatic sense using ANNs. This can help in imparting the better
control on yarn quality.
2. The auto-levelling used in the draw frame for imparting desired linear
density control. Once produced there is no way to remedied faults
passing up to yarn stage after draw frame. Instead of manual decision
based system, ANN defined auto-levelling system at draw frame has
served excellent.
3. The optimisation of the top roller diameter as well as the study of
the spinning balloon in the main spinning phase is important for
controlling yarn quality. It has been well examined and controlled via
the use of ANNs.
4. The warp breakage rate during the weaving is a complex function of
the yarn properties. This can be regulated by an application field of an
ANN model.
5. The prediction of the yarn evenness and hairiness is of great practical
interest. ANNs have been used for the prediction of hairiness of
worsted wool yarns.
6. The spinning of the staple fibers for the production of the yarns is
a multistage procedure including many parameters, which influence
the characteristics of the end product, viz; the spun yarn. ANN is the
excellent method for predictors. Numerical simulation provides a useful
insight into a relationship between yarn tenacity and fiber parameters.
It is used for predicting the tensile properties of air jet spun yarns by
analysing the effect of input parameters on yarn tenacity with neural
network model. Similarly the tensile properties of cotton yarn from
fiber properties are well predicted by a yarn to fiber “reverse” model
using ANN. This approach is entirely different from the prevailing
forward models, which predict the properties of final yarn by using
the fiber properties as inputs, where spinning consistency index (SCI)
comprises most of cotton properties is used in replacement of fiber
properties. The cost minimisation of cotton fiber is also ensured by
using classical linear programming approach in combination with
ANN.
7. The prediction of the yarn evenness and hairiness is of great practical
interest. ANNs have been used for the prediction of hairiness of
worsted wool yarns and of cotton yarns. In a similar way, ANNs have
been used for the prediction of the evenness of ring spun worsted yarns
and cotton yarns or the evenness of blended rotor yarns.
8. When two yarn ends must be joined, instead of knotting they are
subjected in the splicing process. Splicing positions are of special
interest because they could affect heavily the mechanical performance
of the yarn in total. Evaluation and comparison of the properties of
the spliced yarns have been made based also on ANNs. Later studies
have used ANNs to predict the properties of the spliced yarns. Latest
studies have contributed to the prediction of the spliced yarns tensile
properties as well as to the prediction of the retained yarn diameter,
thus covering the mechanical and the visible results of the presence of
the splicing points in the yarn.
9. ANN is useful in defining relationship between process variables
and molecular structure for synthetic yarns. The thermo-mechanical
properties of PET yarns are determined using ANN accurately in this
regards.
10. A multilayer feed forward network with back propagation algorithm is
used to determine the optimised machine settings of a draw texturing
machine, with regard to desired yarn properties. The properties of yarn
like tenacity and elongation are well predicted by using regression
analysis and also the process parameters used; draw ratio, D/Y ratio,
speed and primary heater temperature.
11. Traditionally, the quality grades of false twist yarn packages are
classified by human inspection, but the result may be affected by
personal and subjective factors. The image processing technology is
interfaced with neural networks to extract the defects in yarn packages
and thereby used to classify the quality grades of the yarn packages.
12. ANNs have also been used for the appearance analysis of false twist
textured yarn packages, for the prediction of yarn shrinkage or for the
modelling of the relaxation behaviour of yarns.
5.9.3 Fabric
According to Basu et al. (2002), textile fabrics are often considered as the
final product of the textile process. Their properties must directly meet the
user requirements; obviously, the prediction of their properties and their
final behaviour is very important. If the properties of a fabric are predicted
accurately, they enable the support of the design phase. But the fabrics are
complex structures, if their micromechanical structure is considered. The
structural complexity in conjunction with the materials complexity do not
usually permit the development of computer aided engineering tools for

the support of the design phase, as it usually the case in other engineering
fields such as mechanical, structural, naval, electrical, etc. Therefore, a lot of
effort has been given towards the development of computational tools for the
prediction of the behaviour of the fabrics.
The major areas where researchers have employed ANN are listed below:
1. The inspection of the fabrics for the detection of faults is a very

important operation, traditionally carried out by skilled operators.
Many research teams have focused their interest on making it automatic
by using ANNs to support the fault detection task. It serves as a tool
for dynamic inspection of fabrics; the inspection sample is a piece of
plain white fabric. The four defects are holes, oil stains, warp lacking
and weft lacking. The image treatment employs a high resolution
linear scan digital camera. Fabric images are acquired first, and then
the images are transferred to a computer for analysis. Finally, the data
is adopted as input for neural network which is obtained from readings
after treating the images. In this system, there are three feed forward
networks: an input layer, one hidden layer and an outer layer. As it has
the ability to cope with the non-linear regression property, this method
can reinforce the effects of image identification.
The back propagation algorithm is used to study the missing ends
and picks, oily fabrics and broken fabric, all often found as a defect
in fabric. The correct selection of parameters for input layer plays
a great role in recognition rate. Once a defect occurs in a fabric, its
periodicity is changed so that corresponding intensities at specific
position of spectrum obviously changed. These intensions can act as
characteristics parameters and can be substituted in ANN for learning.
The efficiency and accuracy of a method of detecting fabric defects
that have been classified into different categories by neural network
in the study. Four fabric defects most likely found during weaving
were learned by network. The method used for processing image is
co-occurrence based method, by which six parameters are obtained.
The results shows that, the fabric defects inspected by means of
image reorganisation in accordance with ANN agrees well with initial
expectations.
2. The relationship between structure and properties of the fabric is
complex and inherently non-linear; to create a predictive model one
must resolve these complexities. The optimisation of input parameters
required to produce a fabric as per end use is called as fabric
engineering. Fabrics can be engineered either by weaving, knitting or
bonding. Neural networks are successfully implemented in all three
to optimise the input parameters. The yarn properties and machine

settings are used as inputs to predict the fabric properties as outputs.
The construction of neural network for fabric engineering is shown in
Fig. 5.6
3. The detection and recognition of the patterns on a fabric is of the same
complex category of problems and thus resolved by the implementation
of ANNs. Similarly automated stitch inspection is achieved. An auto
correlation function is used to determine one weave repeat of the fabric.
The reflected fabric image is captured by CCD camera and digitised
by computer system. The learning vector quantisation’s algorithm as
a learning rule of the artificial neural network enables recognition of
woven fabric types more effectively. The results demonstrated that, the
three fundamental weaves can be classified accurately and structural
parameters such as yarn spacing, its variance and the ratio of warp to
weft spacing can also be obtained. The system consists of a PC and
CCD (charged coupled device) camera equipped with an attachable
zoom lens. The fabric image is captured by a lightening device which
is mounted at the end of the camera and zooms in on the image,
which is also displayed on the coloured monitor. The image is then
digitised by the AC to DC convertor and processed using techniques
such as thresholding and auto-correlation to analyse the fabric pattern.
The original image is converted into a grey level image to improve
computer processing time and speed and the histograms are obtained.
4. Drapability is far the most complex mechanical property of the fabrics
and it is essential form for any applications of the textile fabrics. The
prediction of the drape has been made using ANNs. In parallel the
engineering of the drapability of the fabrics became possible though a
predictive tool.
5. Fabric hand is a property that combines the mechanical properties
of a fabric with the sensory perception of the fabric by the humans
when they touch it. It is difficult to give an objective description of the
fabric handle, because a subjective evaluation takes place in practice.
Thereby prediction of the fabric hand is a complex, highly non-linear
problem and therefore an early target for the application of ANNs.
The data from the FAST system were used to approach the hand of
the fabrics, while fuzzy logic has been combined with ANN for the
evaluation of the fabric hand.
6. ANNs in combination with fuzzy logic have been used in the case of
the prediction of the sensory properties of the fabrics.
7. The prediction of the mechanical properties of the textile fabrics is
an essential technical requirement. ANNs have been used for the
prediction of the tensile strength and for the initial stress-strain curve
of the fabrics. The same problem has been solved using an adaptive
neuro-fuzzy system.
8. The shear stiffness of the worsted fabrics and their compression
properties has been successfully modelled.
9. The prediction of bursting strength using ANNs for knitted fabrics as
well as for woven fabrics has been achieved with satisfactory results.
10. The permeability of the woven fabrics has been modelled using ANNs
as well as, the impact permeability has been studied and the quality of
the neural models has been assessed.
11. The pilling propensity of the fabrics has been predicted and the pilling
of the fabrics has been evaluated.
12. The presence of fuzz fibers has been modelled by ANN.
13. The evaluation of the wrinkle of the fabrics has been realized on
an objective basis with a system based on ANNs. The Uster fabric
scan system uses ANN in evaluating wrinkled fabrics with image
analysis. Wrinkles in cloth usually develop with deformation
during storage. It is not easy even for trained observers to judge
the wrinkles; thereby an objective method is being proposed. The
visual information stimulates their senses and makes them to judge
the grade of wrinkle. Since visual evaluation is complex and bias,
linear evaluation system cannot be used for the purpose being but
instead neural networks are widely used. Since networks are non-
linear, the training rule is regularised as non-linear optimisation
problem. Input parameters such as angular second monument,
contrast, correlation, entropy and fractal dimension are obtained
using image analysis and are fed to the neural network algorithm.
The mean sensory value presenting the grade of wrinkled fabric as
output is obtained with desired parameters. The parameters to be
perceived such as density, size, form and contrast can be inspected
using multi-input and multi-output concept of neural network. A
specific algorithm named “Kalman’s Filter Algorithm” is used for
this purpose and results are referred to be very much feasible for
evaluating wrinkled fabrics.
14. Prediction of the spirality of the relaxed knitted fabrics as well as knit
global quality and subjective assessment of the knit fabrics have been
implemented using ANNs.
15. Prediction of the thermal resistance and the thermal conductivity of
the textile fabrics have been realized with the help of ANNs.
Figure 5.6 Neural network structure for fabric engineering.
16. Moisture and heat transfer in knitted fabrics has been studied with
ANN modelling successfully.
17. Engineering of fabrics used in safety and protection applications is
supported by ANNs.
18. Prediction of the fabrics end use is also possible via ANN method.
19. Optimisation of the application of a repellent coating has also been
approached by the ANN model. The ANN is trained and validated
using a wide experimental database. The work is proved to be useful
for technicians in selecting the optimal processing parameters for
obtaining the desired coated fabric properties. The developed system
requires technicians to set the values of process parameters and to
stimulate the ANN based model to forecast the quality of coated fabric.
20. The colour measurement, comparison, evaluation and prediction
are major actions in the dyeing and finishing field of the textile
process. Although colour measurement is possible in the laboratory
with the help of specialised equipment like the spectrophotometers,
few capabilities exist for the prediction of the colour changes or
the final colour appearance, because the problem is multivariable.
The prediction of the colour and the colour solidity of a jigger dyed
cellulose based fabric has been achieved by using cascade ANNs.
In the field of printing, the colour recipe specification has been
made possible using radial basis function neural networks. The
pigment combinations for the textile printing can be determined,
the colour of the printed fabric images can be identified and the
colour separation can take place by using different ANN types. The
prediction of CIELAB values is possible for colour changes after
chemical processes, for nylon 6,6 and for stripped cotton fabrics.
The optimisation of the processing conditions and the prediction of
the dyeing quality of nylon and lycra fabrics and the classification
of dyeing defects have been carried out with the help of ANNs and
fuzzy neural networks, respectively.
5.9.4 Garment
The readymade garment industry (RGI) is characterised by low-

automation and labour intensive manufacturing, as well as fiercely
changing customer’s demands. These characteristics add to the complexity
of decision-making problems by bringing with them particular problem
features, such as uncertainty and flexibility in PPC (Production Planning
and Control), storage non-linearity, and seasonality in apparel retailing,
which are hard to solve by conventional techniques. A large variety of
ANN approaches for decision-making problems in the apparel industry
have been proposed in the last two decades, three major leading areas are
summarised below:
1. Resolving prediction problem: RGI involves various predictions

during production process, like; fabric manufacturing performance,
sewing thread consumption, fashion sensory comfort, cutting time,
apparel sales, etc.
2. Resolving classification problems: It involves classification at
various levels, just for an example; fabric online classification, fabric
defect classification, fabric end-use classification and seam feature
classification etc.
3. Model identification problems.
Hence clothing articles are the end product of the main stream of the textile
production flow. A short summary of various researched done in the area is
appended here;
1. Although precision of the prediction of properties is not that critical as

in technical applications, estimation of the final properties is essential
for the clothing design. Thereby, selection of raw materials and their
required properties, one of the most important factors affecting the
garment quality is related to the seam, the result of the sewing process
have been well defined with ANNs. Indeed, an application area purely
influenced by end use, e.g. prediction of the seam strength is very
important, for parachutes.
2. The thread consumption is predicted via an ANN model.
3. The seam puckering is evaluated and the sewing thread is optimised
through ANN models, respectively.
4. The prediction of the sewing performance is also possible using ANNs.
5. The human psychological perceptions of the clothing sensory comfort
and the analysis of the tactile perception of the textile materials can be
carried out using ANN approaches.
6. Prediction of the performance of the fabrics in garment manufacturing
and fit garment design has been realized based on ANN systems.
7. Cases of special interest, like the selection of the optimal interlinings,
or of broad interest, like the simulation of a textile supply chain, have
been successfully modelled by ANNs.
5.9.5 Non-woven
The non-woven is a specific category of fabrics, made directly of fibers and

not of yarns. The non-woven fabrics find many technical applications and
their role is essential. Some of the major areas researchers have identified for
ANN application are given below;
i. The non-woven fabrics undergo a process of inspection in order to

ensure quality of the delivered material. A visual inspection system
has been based on wavelet texture analysis and robust Bayesian ANNs,
or similarly wavelet transforms and ANNs.
ii. A neuro-fractal approach has been used for the recognition and
classification of non-woven web images.
iii. Many quality issues are addressed via ANN methods, like the structure-
properties relations of the non-woven fabrics, the construction of a
quality prediction system, the modelling of the compression properties
of needle-punched non-woven fabrics, the simulation of the drawing
of spun bonding non-woven process and also the objective evaluation
of the pilling on non-woven fabrics.
5.10 Connotation of ANN offered solutions over the

other methods
Various researchers have investigated a comparative study in order to relate
performance of mechanistic models, simple statistical models, and ANN
models in various textile applications. During the study conducted for
predicting ring-spun cotton yarn tenacity from fiber properties and process
parameters, it was seen that the yarn tenacity prediction error for an ANN
model was 6.9% against 9.3% and 9.9% for the mechanistic and statistical
models, respectively. The same trend was found more prominent for ring-
spun polyester yarns. The mechanistic model, with an average error of 8%,
fared far worse than the ANN model, with an average error of 1.1%. The
statistical model came close with an average error of 2.2%. Thus compared
to traditional analysis systems used to study the relationships characterising
the textile production line, the ANNs proved to have a valued potential along
the entire textile pipeline. However, the fact should be remembered that the
correct selection of the ANN parameters, of the ANN architecture, of the
training technique as well as of the number of input and output parameters
play a decisive role in the effectiveness of the ANN.
References
1. Admuthe L.S. & Apte S. 2010, Adaptive neuro-fuzzy inference

system with subtractive clustering: A model to predict fiber and yarn
relationship, Textile Research Journal, 80(9), 841–846.
2. Allan G., Yang R., Fotheringham A. & Mather R. 2001, Neural
modelling of polypropylene fibre processing: Predicting the structure
and properties and identifying the control parameters for specified
fibres. Journal of Materials Science, 36 3113–3118.
3. Allan G., Fotheringham A. & Weedall P. 2002, The use of plasma and
neural modelling to optimise the application of a repellent coating to
disposable surgical garments, AUTEX Research Journal, 2(2).
4. Babay A., Cheikhrouhou M., Vermeulen B., Rabenasolo B. & Castelain
J.M., 2005, Selecting the optimal neural network architecture for
predicting cotton yarn hairiness. Journal of the Textile Inst., 96(3),
185–192.
5. Backer S. 1951, The relationship between the structural geometry of
a textile fabric and its physical properties, Part IV: Intercise geometry
and air permeability. Textile Research Journal, 2, 703–714.
6. Balci O., Ogulata S.N., Sahin C. & Ogulata R.Y., 2008, An artificial
neural network approach to prediction of the colorimetric values of the
stripped cotton fabrics, fibers and polymers, 9(5), 604–614.
7. Balci O. & Ogulata R.T., 2009, Prediction of CIELab Values and
Color Changing Occurred After Chemical Finishing Applications by
Artificial Neural Networks on Dyed Fabrics. Tekstil ve Konfeksiyon,
1, 61–69.
8. Balci O. & Ogulata R.T., 2009, Prediction of the changes on the
CIELab values of fabric after chemical finishing using artificial neural
network and linear regression models. Fibers and Polymers, 10(3),
384–393.
9. Basu A., Chellamani, K.P. & Ramesh P.R., 2002, Fabric engineering
by means of an artificial neural network. Journal of the Textile Inst.,
93(3), 283–296.
10. Behera B.K. & Mani M.P., 2007, Characterization and classification of
fabric defects using discrete cosine transformation and artificial neural
network. Indian Journal of Fibre & Textile Research, 32,421–426.
11. Behera B.K. & Muttagi S.B., 2004, Performance of error back
propagation vis-á-vis radial basis function neural network: Part I:
prediction of properties for design engineering of woven suiting
fabrics. Journal of the Textile Industry, 95(1) 283–300.
12. Behera B.K. & Muttagi S.B., 2004, Performance of error back
propagation vis-à-vis radial basis function neural network: Part ii:
reverse engineering of woven fabrics. Journal of the Textile Insdutry,
95(1) 301–317.
13. Beltran R., Wang L. & Wang X., 2005, Predicting the pilling propensity
of fabrics through artificial neural network modeling. TextIle Research
Journal, 75(7), 557–561.
14. Bhattacharjee D. & Kothari V.K., 2007, A neural network system for
prediction of thermal resistance of textile fabrics. Textile Research
Journal, 77(1), 4–12.
15. Çay A., Vassiliadis S., Rangoussi M. & Tarakçıoğlu I., 2007, Prediction
of the air permeability of woven fabrics using neural networks. Int. J.
of Cloth. Sc. & Techn, 19(1), 18–35.
16. Chattopadhyay R. & Guha A., 2004, Artificial neural networks:
Applications to textiles. Textile Progress, 35(1), 1–46.
17. Chen T., Li L., Koehl L., Vroman P. & Zeng X., 2007, A soft computing
approach to model the structure – property relations of nonwoven
fabrics. Journal of Applied Polymer Science. 103, 442–450.
18. Chen T., Zhang C., Li L. & Chen, X., 2008, Simulating the drawing
of spunbonding nonwoven process using an artificial neural network
technique. Journal of the Textile Inst., 99(5) 479–488.
19. Chen X. & Huang X.B., 2004, Evaluating fabric pilling with light-
projected image analysis. Textile ResEarch Journal, 74(11), 977–981.
20. Chen Y., Zhao T. & Collier B.J., 2001, Prediction of fabric end-use
using a neural network technique. Journal of the Textile Inst., 92(2)
157–163.
21. Cheng K.P.S. & Lam H.L.I., 2003, Evaluating and comparing the
physical properties of spliced yarns by regression and neural network
techniques. Textile Research Journal, 73(2), 161–164.
22. Cheng L., Ghorashi H., Duckett K., Zapletalova T. & Watson M., 1999,
Color grading of cotton. Part ii: color grading with an expert system
and neural networks, Textile Research Journal, 69(12), 893–903.
23. Cherkassky A. & Weinberg A., 2010, Objective evaluation of textile
fabric appearance. Part 2: set opti-grade tester, grading algorithms and
testing. Textile Research Journal, 80(2), 135–144.
24. Chiu S.H., Chen H.M., Chen J.Y. & Wen C.Y., 2001, Appearance
analysis of false twist textured yarn packages using image processing
and neural network technology. Textile Research Journal, 71(4),313–
317.
25. Chiou Y.-C., Lin C.- S. & Chen G.-Z., 2009, Automatic texture
inspection in the classification of papers and cloths with neural
networks method. Sens. Review, 29(3), 250–259.
26. Choi H.T., Jeong S.H., Kim S.R., Jaung J.Y. & Kim S.H., 2001,
Detecting fabric defects with computer vision and fuzzy rule
generation. Part ii: Defect identification by a fuzzy expert system,
27. Dayik M., 2009. Prediction of yarn properties using evaluation
programing, Textile Research Journal, 79(11), 963–972.
28. Debnath S. & Madhusoothanan M., 2008, Modeling of compression
properties of needle- punched nonwoven fabrics using artificial neural
network. Indian Journal of Fibre & Textile Research, 33, 392–399.
29. Demiryurek O. & Koc E., 2009, Predicting the unevenness of polyester/
viscose blended open-end rotor spun yarns using artificial neural
network and statistical models. Fibers and Polymers, 10(2), 237–245.
30. Ertugrul S. & Ucar N., 2000, Predicting bursting strength of cotton
plain knitted fabrics using intelligent techniques, Textile Research
Journal, 70(10), 845–851.
31. Fan J., Newton E., Au R. & Chan S.C.F., 2001, Predicting garment
drape with a fuzzy- neural network, Textile Research Journal, 71(7),
605–608.
32. Farooq A. & Cherif C., 2008, Use of artificial neural networks for
determining the leveling action point at the auto-leveling draw frame.
33. Fayala F., Alibi H., Benltoufa S. & Jemni A., 2008, Neural network
for predicting thermal conductivity of knit materials, Journal of
Engineering Fibers and Fabrics, 3(4),pp. 53–60.
34. Furferi R. & Carfagni M., 2010, Prediction of the color and of the
color solidity of a jigger dyed cellulose-based fabric: A cascade neural
network approach. Textile Research Journal, 80.
35. Ghane M., Semnani D., Saghafi R. & Beigzadeh H., 2008, Optimization
of top roller diameter of ring machine to enhance yarn evenness by
using artificial intelligence. Indian Journal of Fibre & Textile Research,
33, 365–370.
36. Ghosh A. & Chatterjee P., 2010, Prediction of cotton yarn properties
using support vector machine. Fibers and Polymers, 11(1), 84–88.
37. Golob D., Osterman D.P. & Zupan J., 2008, Determination of pigment
combinations for textile printing using artificial neural networks. Fibr.
& Text. in East. Eur., 16(3), 93–98.
38. Gong R.H. & Chen, Y.,1999, Predicting the performance of fabrics
in garment manufacturing with artificial neural networks. Textile
39. Guha A., Chattopadhyay, R. & Jayadeva, 2001, Predicting yarn
tenacity: A comparison of mechanistic, statistical, and neural network
models, Journal of the Textile Institute, 92(2) 139–145.
40. Guo Z.X., Wong W.K., Leung S.Y.S., and Min Li, 2011, Application of
artificial intelligence in the apparel industry: a review. Textile Research
Journal, online 20 June 2011.
41. Gürkan Ünal P., Özdil N. & Taskin C., 2010, The effect of fiber
properties on the characteristics of spliced yarns part i: prediction of
spliced yarns tensile properties. Textile Research Journal, 80(5), 429–
438.
42. Gürkan Ünal P., Arikan C., Özdil N. & Taskin C., 2010, The effect
of fiber properties on the characteristics of spliced yarns: Part ii:
prediction of retained spliced diameter. Textile Research Journal, 80.
43. Gurumurthy B.R., 2007, Prediction of fabric compressive properties
using artificial neural networks. AUTEX Research Journal, 7(1).
44. Hadizadeh M., Jeddi A.A.A. & Tehran M.A., 2009, The prediction of
initial load-extension behavior of woven fabrics using artificial neural
network. Textile Research Journal, 79(17), 1599–1609.
45. Hadizadeh M., Tehran M. A. & Jeddi A.A.A., 2010, Application of
an adaptive neurofuzzy system for prediction of initial load/extension
behavior of plain-woven fabrics. Textile Research Journal, 80(10),
981–990.
46. Hornik K., Stinchcombe M. & White H., 1989, Multilayer feedforward
networks are universal approximators. Neural Networks, 2, 359–366.
47. Hu Z.H., Ding Y.S., Yu X. K., Zhang W.B. & Yan Q., 2009, A hybrid
neural network and immune algorithm approach for fit garment design.
48. Huang C.C. & Chen I.C., 2001, Neural-fuzzy classification for fabric
defects. Textile Research Journal, 71(3), 220–224.
49. Huang C.C. & Chang K.T., 2001, Fuzzy self-organizing and neural
network control of sliver linear density in a drawing frame. Textile
50. Huang C.C. & Lin T.F., 2008, Image inspection of nonwoven defects
using wavelet transforms and neural networks. Fibers and Polymers,
9(5),633–638
51. Huang C.C. & Yu W.H., 2001, Fuzzy neural network approach to
classifying dyeing defects. Textile Research Journal, 71(2), 100–104.
52. Hui C.L., Lau T.W., Ng S.F. & Chan K.C.C. (2004). Neural network
prediction of human psychological perceptions of fabric hand. Textile
53. Hui C.-L. & Ng S.-F., 2005, A new approach for prediction of sewing
performance of fabrics in apparel manufacturing using artificial neural
networks. Journal of the Textile Institute, 96(6) 401–405.
54. Hui P.C.L., Chan K.C.C., Yeung K.W. & Ng F.S.F., 2007, Application
of artificial neural networks to the prediction of sewing performance
of fabrics. Int. J. of Cloth. Sc. & Techn., 19(5), 291–318.
55. Hui C.L. & Ng S.F., 2009, Predicting seam performance of commercial
woven fabrics using multiple logarithm regression and artificial neural
networks. Textile Research Journal, 79(18), 1649–1657.
56. Islam M.A., Akhter S., Mursalin T.E. & Amin M.A., 2006, A Suitable
Neural Network to Detect Textile Defects, King et al. (Eds.): ICONIP
2006, Part II, LNCS 4233,430–438, Springer.
57. Jackowska-Strumiłło L., Cyniak D., Czekalski J. & Jackowski T.,
2008, Neural model of the spinning process dedicated to predicting
properties of cotton-polyester blended yarns on the basis of the
characteristics of feeding streams, Fibr. & Text. in East. Eur., 16(1),
66.
58. Jasper W.J. & Kovacs E., 1994, Using neural networks and NIR
spectro-photometry to identify fibers. Textile Research Journal, 64(8),
444–448.
59. Jeguirim S.E.G., Dhouib A.B., Sahnoun A., Cheikhrouhou M.,
Schacher L. & Adolphe D., 2009, The use of fuzzy logic and neural
networks models for sensory properties prediction from process and
structure parameters of knitted fabrics. J. Intell. Manuf., DOI 10.1007/
s10845-009-0362-y.
60. Jeon B.S., Bae J.H. & Suh M.W., 2003, Automatic recognition of
woven fabric patterns by an artificial neural network. Textile Research
Journal, 73(7), 645–650.
61. Jeong S.H., Kim J.H. & Hong C.J., 2000, Selecting optimal interlinings
with a neural network. Textile Research Journal, 70(11), 1005–1010.
62. Jianli, Liu & Baoqi, Zuo 2007, Identification of fabric defects based
on discrete wavelet transform and back-propagation neural network.
Journal Textile Institute, 98(4), 355–362.
63. Ju J. & Ryu H., 2006, A study on subjective assessment of knit fabric
by ANFIS. Fibers and Polymers, 7(2), 203–212.
64. Kang T.J. & Kim S.C., 2002, Objective evaluation of the trash and
color of raw cotton by image processing and neural network. Textile
65. Karthikeyan B. & Sztandera L.M., 2010, Analysis of tactile perceptions
of textile materials using artificial intelligence techniques, Part 2:
reverse engineering using genetic algorithm coupled neural network.
Int. J. of Cloth. Sc. & Techn., 22(2/3) 202–210.
66. Keeler, J., 1992, Vision of neural networks and fuzzy logic for prediction
and optimisation of manufacturing processes, in: applications of
artificial neural networks III, 1709, 447–456.
67. Keshavaraj R., Tock R.W. & Nusholtz G.S., 1995, A simple neural
network based model approach for nylon 66 fabrics used in safety
restraint systems: a comparison of two training algorithms. Journal of
Applied Polymer Science, 57, 1127–1144.
68. Kim E.H., 1999, Objective evaluation of wrinkle recovery. Textile
69. Kuo C.F.J. & Fang C.-C., 2006, Optimization of the processing
conditions and prediction of the quality for dyeing nylon and lycra
blended fabrics. Fibers and Polymers, 7(4), 344–351.
70. Kuo C.F.J., Hsiao K.I. & Wu Y.S., 2004, Using neural network theory
to predict the properties of melt spun fibers. Textile Research Journal,
74(9), 840–843.
71. Kuo C.F.J., Su T.L., Chang C.D. & Lee C.H., 2008, Intelligence control
of on-line dynamic gray cloth inspecting machine system module
design. ii. defects inspecting module design. Fibers and Polymers,
9(6), 768–775.
72. Kuo C.-F. J., Su T.L., Chiu C.-H. & Tsai C.-P., 2007, Analysis and
construction of a quality prediction system for needle-punched non-
woven fabrics. Fibers and Polymers, 8(1), 66–71.
73. Kuo C.-F.J., Su T.L. & Huang Y.J., 2007, Computerized color
separation system for printed fabrics by using backward-propagation
neural network. Fibers and Polymers, 8(5), 529–536.
74. Kuo C.F.J., Wang C.C. & Hsieh C.T., 1999, Theoretical control and
experimental verification of carded web density part iii: neural network
controller design. Textile Research Journal, 69(6), 401–406.
75. Lewandowski S. & Drobina R., 2008, Prediction of properties of
unknotted spliced ends of yarns using multiple regression and artificial
neural networks. part i: identification of spliced joints of combed wool
yarn by artificial neural networks and multiple regression. Fibr. &
Text. in East. Eur., 16, 5 (70), 33–39.
76. Lin T.H., 2004, Construction of predictive model on fabric and sewing
thread optimization. Journal of Textile Engineering, 50(1), 6–11.
77. Liu S., Wan O. & Zhang H., 2009, Fabric Weave Identification Based
on Cellular Neural Network, The Sixth Int. Symposium on Neural
Networks, AISC 56, pp. 563–569.
78. Majumdar A., Majumdar P.K. & Sarkar B., 2004, Selecting cotton
bales by spinning consistency index and micronaire using artificial
neural networks. AUTEX Research Journal, 4(1),2–8.
79. Majumdar P.K. & Majumdar A., 2004, Predicting the breaking
elongation of ring spun cotton yarns using mathematical, statistical,
and artificial neural network models. Textile Research Journal, 74(7),
652–655.
80. Majumdar A., Majumdar P.K. & Sarkar B., 2006, An investigation
on yarn engineering using artificial neural networks. Journal of the
Textile Institute, 97(5) 429–434.
81. Majumdar A., Ciocoiu M. & Blaga M., 2008, Modelling of ring yarn
unevenness by soft computing approach. Fibers and Polymers, 9(2),
210–216.
82. Majumdar A., Ghosh A., Saha S.S., Roy A., Barman S., Panigrahi D.
& Biswas A., 2008, Empirical modelling of tensile strength of woven
fabrics. Fibers and Polymers, 9(2), 240–245.
83. Majumdar A., 2010, Modeling of cotton yarn hairiness using adaptive
neuro-fuzzy inference system. Indian Journal of Fibre & Textile
Research, 35, 121–127.
84. Mak K.L. & Li W., 2007, Objective evaluation of Seam Pucker on
textiles by using self organizing map. IAENG International Journal of
Computer Science, 35(1), IJCS_35_1.
85. Mori T. & Komiyama J., 2002, Evaluating wrinkled fabrics with image
analysis and neural networks. Textile Research Journal, 72(5), 417–
422.
86. www.intechopen.com, Woven Fabric Engineering132.
87. Murrells C.M., Tao X.M., Xu B.G. & Cheng K.P.S., 2009, An artificial
neural network model for the prediction of spirality of fully relaxed
single jersey fabrics. Textile Research Journal, 79(3), 227–234.
88. Mursalin T.E., Eishita F.Z. & Islam A.R., 2008, Fabric defect
inspection system using neural network and microcontroller. Journal
of Theoretical and Applied Information Technology, l4(7),
89. Mwasiagi J.I., Huang X. & Wang X., 2008, Performance of neural
network algorithms during the prediction of yarn breaking elongation.
Fibers and Polymers, 9(1), 80–86.
90. Nurwaha D. & Wang X.H., 2010, Prediction of rotor spun yarn strength
from cotton fiber properties using adaptive neuro-fuzzy inference
system method. Fibers and Polymers, 11(1), 97–100.
91. Onal L., Zeydan M., Korkmaz M. & Meeran S., 2009, Predicting the
seam strength of notched webbings for parachute assemblies using the
Taguchi’s design of experiment and artificial neural networks. Textile

92. Park S.W., Hwang Y.G., Kang B.C. & Yeo S.W., 2000, Applying fuzzy
logic and neural networks to total hand evaluation of knitted fabrics.
93. Park S.W., Hwang Y. G., Kang B.C. & Yeo S.W., 2001, Total handle
evaluation from selected mechanical properties of knitted fabrics
using neural network. Int. J. of Cloth. Sc. & Techn., 13(2) 106–114.
94. Ramesh M.C., Rajamanickam R. & Jayaraman S., 1995, The prediction
of yarn tensile properties by using artificial neural networks. Journal
Textile Institute, 86(3) 459–469.
95. Savvas Vassiliadis, Maria Rangoussi, Ahmet Cay and Christopher
Provatidis, Artificial neural networks and their applications in the
engineering of fabrics, www.intechopen.com, 112–134.
96. Sentthil Kumar C.S., Dhinakaran M., 2015, Scope for artificial
neural network in textiles. IOSR Journal of Polymer and Textile
Engineering,2(1) 34–39.
97. Sette S. & van Langenhove L., 2002, Optimising the fibre-to-yarn
production process: Finding a blend of fibre qualities to create an
optimal price/quality yarn. AUTEX Research Journal, 2(2), 57–63.
98. Shady E., Gowayed Y., Abouiiana M., Youssef S. & Pastore C., 2006,
Detection and classification of defects in knitted fabric structures.
99. She F.H., Kong L.X., Nahavandi S. & Kouzani A.S., 2002, Intelligent
animal fiber classification with artificial neural networks. Textile
660–665.
101. Shiau Y.-R., Tsai I.-S. & Lin C.-S., 2000, Classifying web defects with
a back-propagation neural network by color image processing. Textile
102. Stylios G. & Parsons-Moore R., 1993, Seam pucker prediction using
neural computing. International Journal of Clothing Science and
Technology, 5(5), 24–27.
103. Stylios G.K. & Powell N.J., 2003, Engineering the drapability of textile
fabrics, International Journal of Clothing Science and Technology,
15(3/4), 211–217.
104. Subramanian S.N., Venkatachalam A. & Subramaniam V. 2007,
Prediction and optimization of yarn properties using genetic algorithm/
artificial neural network. Indian Journal of Fibre & Textile Research,
32, 409–413.
105. Tokarska M. & Gniotek K., 2009, Determination of woven fabric
impact permeability index. Indian Journal of Fibre & Textile Research,
34, 239–244.
106. Tran C.-D., Phillips D.G. & Fraser W.B., 2010, Stationary solution of
the ring-spinning balloon in zero air drag using a RBFN based mesh-
free method. Journal of the Textile Institute, 101(2) 101–110.
107. Üreyen M.E. & Gürkan P., 2008, Comparison of artificial neural
network and linear regression models for prediction of ring spun yarn
properties. i. prediction of yarn tensile properties. Fibers and Polymers,
9(1), 87–91.
108. Üreyen M.E. & Gürkan P., 2008, Comparison of artificial neural
network and linear regression models for prediction of ring spun yarn
properties. ii. prediction of yarn hairiness and unevenness. Fibers and
Polymers, 9(1), 92–96.
109. Xu B., Su J., Dale D.S. & Watson M.D., 2000, Cotton color grading
with a neural network. Textile Research Journal, 70(5), 430–436.
110. Yao G., Guo J. & Zhou Y., 2005, Predicting the warp breakage rate
in weaving by neural network Techniques, Textile Research Journal,
75(3), 274–278.
111. Yazdi M.M., Semnani D. & Sheikhzadeh M., 2009, Moisture and
heat transfer in hybrid weft knitted fabric with artificial intelligence.
Journal of Applied Polymer Science, 114, 1731–1737.
112. Yin X.G. & Yu W., 2007, Selection and evaluation of input parameters
of neural networks using grey superior analysis. Textile Research
Journal, 77(6), 377–386.
113. Zeng Y.C., Wang K.F. & Yu C.W., 2004, Predicting the tensile properties
of air-jet spun yarns. Textile Research Journal, 74(8), 689–694.
114. Zhang J., Wang X. & Palmer S., 2010, Performance of an objective
fabric pilling evaluation method, Textile Research Journal, 77(6), .
115. Zhang J., Wang X. & Palmer S., 2010, Objective pilling evaluation of
nonwoven fabrics, fibers and polymers. 11(1), 115–120.
6
Changes in mix formulation approach with the
technological developments


*
Abstract
Mix formulation process commences with the selection of suitable cotton
component. Suitability here defined in terms of quality and cost of the cotton
in terms of its ability to meet requirements of desired output yarn. The major
constraint of this production process is its irreversible nature that prevents
rectification of the mistakes even after having modern machines in the later
processes. Thus, it becomes mandatory for running the spinning mill properly
that the raw material in the back processes must be selected without any
mistake or within the acceptable range of selection. Art based classical system
has shifted to scientifically engineered selection system for mix formulation,
mainly due to revolutionary changes took place not only in production but
also in the associated techniques like agricultural, transport, testing, statistical,
mixing, etc.
Key Words: Mix formulation, Agriculture, textile testing, transport tools,
production process
6.1 Introduction
Raw cotton quality is the prime factor that influences yarn quality by 80–90%
in ring spinning. Determination of cotton fiber traits is based on complex
interactions among genetic, environmental and processing conditions. Fiber
properties thereby vary significantly at multiple junctions, i.e. between fields,
between individual plants within fields, and even within single plants and
within the same seed. Thus, the major challenge lies in cotton processing is to
convert a highly variable raw material into a uniform product with consistent
quality that remains true over long period of production time. One of the
common approach adopted by textile industry is to use cotton mix instead

of individual cotton variety. However, addressing this challenge successfully
demands adequate measurement of all the important fiber properties which
in itself a critical job, and also assure that all accepted cotton bale classing
based on those measurements only. If any mistakes happen in selecting the
raw material during the preparatory process before and within the blow room
then it becomes apparently impossible to rectify the mistakes even after
having modern machines in the later processes. Thus, it becomes mandatory
for running the spinning mill properly that the raw material in the back
processes must be selected without any mistake or within the acceptable range
of selection.
Cotton mixing has a significant impact on end product cost apart from
quality. Its share on yarn cost ranges between 60–65% and about 50–55% for
cloth cost. It is thereby obvious that savings incurred in purchasing the raw
material have a positive impact on production cost reduction of the spinning
mills. Thereby vital role of correct raw material selection in ensuring the
efficient and smooth running of the spinning mill at an economical rate cannot
be ignored. It is no more be exaggerating at this point to deduce that mixing
plays an important role both for end product quality and its cost.
Classical time selection of the raw material was art based and suffering
from many limitations. No doubt system was undergone drastic modifications
by the virtue of revolutionary changes took place not only in production but
also in the associated techniques like agricultural, transport, testing, statistical,
mixing, etc. The process of cotton fiber selection has thus passed through
a span changes from the traditional pure art based selection to a full proof
scientific technique. Achievement of foreseeable transition into a cotton fiber
mixing program must be routed through fiber selection thence only successful
optimisation of cotton fiber use with respect to cost and quality of end product
can be realized. This is important to address quality requirements imposed
by the rapidly developing technology and continuous change in customer
demand.
6.2 Basic objectives of mix formulation

The basic aim of mix formulation deals with optimisation of yarn quality at an
economical rate and retain its consistency for a long production period.
This basic aims of cotton mix formulation can be well attained by
minimising the effect of individual component in the mix. Use of large
number of components with optimised proportions in the mix is preferred for
this purpose. This can reduce share of individual’s in the mix and also reduces
adverse impact on mix quality at the verge of replacement required for one or
more cotton components in the case of unavailability or price hike. Apart from
Changes in mix formulation approach with the technological 129
that it facilitates in maintaining consistency of mix quality and cost on long

run. It is equally true that selection of undue costly cotton although of good
variety should be avoided; otherwise it will lead to yarn with undue higher
cost with negligible improvement in quality.
6.3 Constrains for accurate mixing

Accurate mix formation demands keen knowledge about:
1. Correlation between; fiber properties and yarn properties, fiber
properties and process parameters, process parameters and yarn
properties.
2. Cost of cotton; any specific constrains about cotton, like its price, its
behaviour during processing, etc. and
3. Trash of cotton and its impact on yarn quality, yarn realization and
waste levels.
This overwhelmed view specify that scientific mix formulation should be
filtered through following steps:
(a) Knowledge about all the fiber properties. This requires accurate,
quicker and scientific ways of sampling, fiber testing and analysis of
test data.
(b) Knowledge about production technology available at the end. This
facilitates in identifying required characteristics of cotton fiber mix
for desired end product.
(c) Knowledge about compatibility of cottons in terms of various
properties and cost. This helps in sorting out right components for mix
formulation.
(d) Knowledge about product yarn quality and cost requirements for a
given production set up. This can help in exactifying proportion of
different components in the mix.
(5) Scientific ground in selecting the right replacement for non-available
components in the mix.
Once these brainstorming exercises for fiber selection is completed, it is
now a challenge for the production technology to get the best and consistent
results over a longer run.
6.4 Different approaches of mix formulation

There are different practices of mix formulation prevailing right from classical
to the recent spinning technique. The change is arisen due to technological
changes took place with respect to time in all the areas of mix formulation
process, viz; testing, statistics and production. Accordingly the mixing
techniques used are broadly divided into three main groups;
1. Classical visual judgment approach
2. Mix formulation with non-automatic instrumental technology
3. Mix formulation with automatic instrumental technology.
In the forthcoming sections each of them has been discussed in terms of
available technology in the particular era for various mix formulation attributes
and their influence on mix homogeneity (quality) and cost.
6.4.1 Classical visual judgment approach
Classical approach refers time span when machine spinning has just replaced
manual spinning, near about time span of 1940s–1945s.
(1) Testing technique
Traditionally, three fiber parameters: grade, fiber length and fiber fineness
have been used to determine the quality value of cotton fiber. As mentioned
in chapter 3, these measures were purely manual judgment based, like grade
was decided on the visual classification of colour and trash content of hand
stapled fiber beard. The same beard was visually judged for length (staple
length) as well as fineness (coarse or fine) fibers. Judicious method of fiber
quality measurement in terms of three parameters only, done from a pinch
of cotton and was resulted in insufficient fiber information. Even it was
human bias means differs from expert to expert due to is subjective ground
of measurement. Thus it was not able to satisfy one of the basic needs of mix
homogeneity
(2) Agricultural, transport and communication technology:
Classical agricultural technique is a labour intensive and relies on regularity
of monsoon cycle. This has made cotton as a seasonal crop and its quality and
price were fluctuating with respect to climatic conditions, ginning conditions,
lot to lot and government policies. Infrastructural facilities like roadways
and railways were in their beginning phase and no scientific communication
techniques were risen by the time. Limitations of conventional agricultural
techniques, non-availability of speedy transport and communication facility
had rendered it difficult to use more varieties of cottons availed from far
away stations in the mix. Usually mix was formed out of one or at the most
two varieties of the cottons as per ease of availability of this seasonal crop
at a particular or nearby station due to restriction imposed by transportation
technology. Even no scientific ground was found in such selection and mainly
biased by individual’s expertise or experience. Many a times such work
practices were enforced to spin yarn with unnecessarily high value cotton.
High inventories was the regular feature and practiced normally to avoid
adverse impact on the consistency of the production pattern. The situation
arose due to replacement of existing cotton component with very high share
(50–100%) from the mix in the virtue of non-availability of crop, or price
hike due to one or other reasons. Working with such huge inventories required
higher capital investments and payment of undue high interest. It was also
adding headache of storage and maintenance, as being a cellulosic fiber more
prone to be attacked by mildew and also fire hazardous. All these factors has
made classically spun yarn unduly costly as well as highly variable.
(3) Mixing technique
The mix was formulated from single or maximum two varieties of cottons
available from a nearby one station only and thereby contribution arisen
from each component was very high. Apart from that each component had
wide lot to lot variation due to differences in growth, cultivation and ginning
techniques. Highly variable cotton fiber properties from lot to lot even though
belonging to same variety were equalized by one of the common approach
used that was huge blending before feeding to the spinning line, also known
as stack mixing and further extended by cross blending opportunities at each
storage point (Fig. 6.1). These cross blending were extended up to the ring
frame stage, as doubling was adopted at the speed frame and ring frames also
in classical spinning system. The increased blending delay so earned gave
an equal chance to the first processed tuft to that of the last at blow room to
be represented in ring spun yarn, no doubt major responsibility for longer
blending delay was born by stack mixing only.
Figure 6.1 Stack mixing concept.

Stack mixing/bale mixing

Vast quantities of bales arriving at mill were opened and spread horizontally
for mixing by grade or growth area in a big mixing room to reduce variability.
Such horizontally laid cottons were then vertically worked off (stacked), re-
baled and fed to the Hopper Bale Opener (HBO), the first machine in the
opening line in random order to further enhance the mixing effect. Since bales
are mixed together in this system, it is also known as bale blending technique.
The entire exercise was carried out manually and at later stage spreading only
was done with the help of mixing wheel. The blending delay added to this
point was quite high due to higher number of bales involved and cross mixing
done at the point of deposit and withdrawal. However, it was resulted in poor
working environment due to very high accumulation of fly and dust. Such
health hazardous working conditions for the operators were not allowed mill
to meet statutory requirements also. Apart from that higher labour compliment
was involved in the process; it was not only cost adding but also leading to
human errors. The rising cost of labour, storage, equipment and raw material
were made theoretically highly favourable concept for mix uniformity, but
practically a messy blending approach of a stack mixing largely impractical.
Cross mixing at the point of intermittent storage
The kind of storages involved in conventional line was gigantic storage trunks
at blow room, Scutcher lap storage, Card-can buffer stock, Drawframe-can
buffer stock and Roving bobbin buffer stock.
Storage trunks: Classical spinning system was dealing with five to six
gigantic machines in blow room line. Accordingly large capacity storage
trunks, operating on ON-OFF mode were provided at the entry point of each
machine, except at HBO in conventional blow room line. Cotton moving
under the influence of air current was filled vertically in these storage trunk’s
columns and withdrawn horizontally, giving a good degree of cross mixing
due to their higher capacity and more number of storage points in line for the
fibers.
Scutcher Lap storage: Laps were stacked horizontally and withdrawn
horizontally in the buffer stock room for the card (Fig. 6.2).
Mathematically, blending delay (D) at this point can be represented as the
time taken for processing at the blow-room an amount of cotton representing
the first and the last tuft going into a draw-frame feed (eq. 6.1).
N ¥ W ¥ 60
D ( minutes ) = 6.1
P ¥L
Figure 6.2 Cross blending at card.

Where,
P - Production (lb/h) of one lap end/Scutcher
L - Number of lap end fed by opening line
W - Lap weight (lb)
N - Number of cards feeding one draw-frames
D - Blending delay time
Buffer stock (Card Can, Draw frame cans, Roving Bobbins): Storage
after each intermittent point like card, draw frame, flyer frame were similarly
operated on cross mixing concept and further enhancing blending delay.
However, in such situation if the card production gets increased than less
number of laps will be represented in bale feed and blending delay time
decreases [eq. 6.1]. The phenomenon of blending delay due to cross mixing is
well explained with an example below.
Example: Blow room operates with 200 kg/h production per Scutcher and
provided with two lap ends in the line. Full lap weight is 20 kg and six
doublings are done at draw frame. Scutcher lap storage capacity is 36 laps.
Blending delay
(1) For consecutive lap feeding: Six slivers going to draw frame
represents six laps fed to six cards. So, time taken to process six laps
each of 20 kg for consecutive feed is 120 kg (= 6 × 20 kg).
N ¥ W ¥ 60 6 ¥ 20 ¥ 60
D ( minutes ) = = = 18 minutes
P¥L 200 ¥ 2
(2) For cross feeding: Six slivers going to draw frame represents six laps
fed to six cards. So, time taken to process six laps each of 20 kg for
cross feed with storage capacity of 36 is 720 kg (= 36 × 20 kg).
N ¥ W ¥ 60 36 ¥ 20 ¥ 60
P¥L 200 ¥ 2
• Thus with changing method of withdrawal, blending delay can
be increased six times. If storage capacity is further higher from
thirty six, delay can be accordingly higher.
Example: For the above mentioned data only if Scutcher lap storage capacity
is 54 laps. Blending delay for cross feed:
N ¥ W ¥ 60 54 ¥ 20 ¥ 60
P¥L 200 ¥ 2
Thus extent of mix homogeneity earned in classical spinning approach
was purely dependent on number of intermittent storage points, their working
capacity and method of withdrawal with respect to storage. However, poor
health hazardous working environment prevailing at stack mixing, increased
labor compliment in material handling and transport, human associated errors,
higher cost of large capacity storage, instrument and raw material cost made
the conventional blending approach largely impractical and obsolate.
Summarising the discussion, the major lagging observed with judiciary
system in meeting demands of homogeneous mix formulation were
(i) Absence of capable and efficient testing methods for all the fiber
properties. Subjective measures and that also for only three main fiber
properties, viz; length, fineness and grade from a very small quantity
of sample withdrawn from a large bulk. They did not allow measure of
true values of cotton and even these measures were human bias. As a
result for the given cotton measured values were differed unpredictably
from person to person and place to place.
(ii) Scarcity of more varieties of cottons in mixing: This was due to lack
of availability of speedy and cheaper transport facility and restricted
agricultural technique.
(iii) Absence of statistical techniques for correlation between fiber
and yarn in terms of quality and cost. Decision for fiber, yarn and
process performance was relied on person’s expertise and experience,
usually biased and having no scientific ground.
(iv) Untidy working environment: Lot of dirt and fly were accumulated
in the house due to massive manual blending of bales during stack
mixing.
6.4.2 Mix formulation with non–automatic instrumental

technology
The major changes in mix formulation techniques in middle of twentieth

century (1960–1970) were mainly due to grass root changes took place in
testing technique. Art based (visual judgment based) fiber evaluation was
replaced by instrumental technique, where human work was partially done
by equipment. The testing techniques were designed for measuring six main
fiber properties, viz; length, fineness, maturity, strength, trash and moisture.
Simultaneously transport, production and agricultural technology had also
undergone reformation and maintained pace with testing techniques. The
overall scenario of mix formulation technology was thereby got changed
remarkably.
Transport, communication and agricultural technologies
Enhanced infrastructure for the transportation means like roadways and
railways along with a communication mean was the notable change of this
time. Telephone has narrowed down the world and extended approachable
limits of spinner for acquiring cotton varieties from nearby stations too at
least within state or maximum within country in the middle age. This was a
good sign for the technological value of mix which was worked out of more
than one or two constituents and affected considerably less by the replacement
of one of the component from the mix.
Agriculture sector was also furnished with upgraded techniques for
ploughing, fertilizer synthesis and varieties of seed resulted in the much
better growth and cultivation of cotton crop. Simultaneously developments of
irrigation facilities had lifted up barrier for water supply in cotton growing.
Cotton was no more be remained as a seasonal crop, as its growth was not
abide by monsoon conditions as in earlier time. Ease of availability of cotton
crop throughout the calendar year had stopped practices of working with huge
inventories and associated problems.
Testing techniques
Earlier prediction about all fiber properties is a primary need to obtain
a unique blending result. However, it was mere difficult with art based
traditional testing technique. Non-automatic instrumental testing techniques
like Pressley strength tester, Baer sorter diagram, Fibrograph, Sheffield
micronaire, Shirley trash analyser, etc., were introduced in the market during
middle age. Comparatively scientific way measurements for all major fiber
characteristics; length, fineness, strength, maturity, trash and moisture, were
became possible. However, involvement of human compliment in sampling
and decision-making as well as analysis of test reports was compulsion and
thereby not independent of human limitations fully as mentioned in section
3.2.4. Later on modified semiautomatic versions of testing equipment like
Stelometer, Digital Fibrograph, etc., were launched in the market as their
replacement. But still problems were not fully resolved and suffered from the
limitations of higher testing time, cost and low accuracy even for them. The
slow testing speed had not allowed the system to do 100% quality check; i.e.
test each bale arriving at mill end. Only 4% of total bales randomly picked up
from the big input lot were checked for quality parameters and based on their
values entire lot characteristics were defined. Eventually that was not enough
to explore correctly within and between bales variations.
Statistical techniques
Test data obtained for different fiber characteristics were analysed for mean
(average) and standard deviation or coefficient of variation (dispersion) values.
i=n
X = Âxi
i=0
Â i=0 ( X i - x )
n 2
Standard Deviation (s ) = ±
n
And
s
Coefficient of Variation (CV ) % = ¥ 100
X
It is statistically proven that accuracy is inversely proportional to number of
observation, thereby in order to earn better test accuracy, higher observations
were taken. However, this need for statistical analysis had further added to
testing time and thereby cost.
Statistical techniques to find compatibility of two cottons used in the mix,
like significance test (ATIRA) and critical difference (SITRA) were also
established simultaneously.
Sample value1 - SampleValue2

Critical Difference ( %) = ¥ 100 6.2
Average value of two samples
And
Sample value1 - SampleValue2

t value = 6.3
Standard Error
They provide information about allowable variations in fiber properties
before mixing them together. Such table for SITRA standards is given here
(Table 6.1). If the difference exceeds the suggested value then creates mess on
the machine and adversely impact yarn quality and cost.
Table 6.1 Allowable critical difference (%) for given number of standard
Fiber properties Number of tests Allowable critical
difference (%)
2.5% Span length 4 combs/sample 4
Uniformity ratio 4 combs/sample 5
Micronaire value 4 plugs/sample 6
Fiber strength at 3 mm gauge length 10 breaks/sample 5
Maturity coefficient 600 fibers/sample 7
Trash content 8 tests/sample 7
Thus, fibers compatible in terms of their characteristics were sorted out

from the available varieties by significant testing or critical difference testing.
Operating with critical difference test method was found preference over
significance testing due to its simplicity. However, with either of these systems
compatibility check for one property of two cottons was possible only at a
time. Thus for six properties of two cottons in consideration, 26 (= 64) such
tests were required to be performed and those also manually. This has imposed
limitation against the use of more number of components in the mix, which
are desirable.
An individual fibers characteristic differs from the mix characteristics as
per their share in the mix and was calculated statistically. Manually calculated
weighted averages were used to predict mix characteristics, where weight
(W) represents proportion of fiber in the mix and numerical figure (1 to n)
represents number of components in the mix. For example effective length (L)
of the mix can be predicted as follows:
L = W1 L1 + W2 L2 + W3 L3 + º.. + Wn Ln 6.4
Such calculated mix values were then compared with standards to check
whether they are meeting to customer’s demand in terms of yarn properties
desired or not. These standards were derived by research institution based
on survey done on a large amount of empirical data, collected for at least
last eighteen months work practices followed in the mills for a particular
product. The apparent difference noted here was that these standards were
derived on the past experience basis only but not of an individual one, it was
of a group of experts in the field. This has reduced biasness and imparted
accuracy in decision making process of defining output yarn quality and
cost for fed cotton values. Other way round with this statistical technique,
first base was established for deriving scientific correlation between input
fiber parameters and output yarn parameters for fixed production set up. But
still struggle was not ended as optimized proportion of different components
in the mix was decided by working out all possible combinations by such
lengthy manual calculations. Length of formula and number for combinations
were dependent on number of components and mix parameters under
consideration. For example, six fiber mix parameters and six cotton varieties
used for mix (66 =46656) combinations need to be verified and that was
also manually, liable to introduce error due to human fatigue. Generalising
for “x” number of sorted out cotton varieties the number of combinations
to be judged are nx, where “n” represents number of levels considered for
proportion in mix. More the levels considered, more wide the range of
calculation spread about. Thus lengthy, slow and tedious statistical analysis
process although having scientific base had put barrier to work with higher
number of components in mix.
Another positive initiation noted for this era was that setting of process
parameters carried out on the basis of such calculated mix characteristics
instead of judiciary decision base of an expert, like; Beater speed and grid bar
settings at blow room were decided on the basis of trash content of mix; or
drafting zone setting = effective length of mix + allowance, etc. The process
was controlled on scientific way more and thereby reduced outgoing material
variations.
Later on with the introduction of semiautomatic testing technology,
statistical technique for defining technological value of the cotton was put
forward by research institutes. It was defined by synthesizing important fiber
characteristics in to one index; FQI (Fiber Quality Index) given by SITRA and
Q (Drafting Quality Index) put forward by ATIRA.
Lsm
Fiber quality index = FQI =
f
Where,
L = 50% span length in mm, measured by Digital Fibrograph,

S = bundle fiber strength measured on Stelometer at 3 mm Gauge
length,
m = maturity coefficient measured by using NaOH method and
f = fiber fineness in micronaire.
and
2
ÈF ˘
Q = 29.4 Í ˙
ÎL˚
Where, F = fineness/maturity coefficient, L = 50% span length.
Two fibers with similar or closer FQI were considered compatible. This has
eliminated need of compatibility test for an individual property. Even fiber
mix parameters required to spin known counts with different CSP under good
processing conditions (P = 100%) were defined in terms of FQI. Fiber with
higher FQI value was used to earn better yarn strength for a given count.
Similarly the fiber with higher FQI was used for spinning finer count to get
equivalent strength to that of coarser yarn in the group (Table 6.2). Such high
FQI fiber was also costly.
Table 6.2 Co-relationship between cotton FQI value, yarn count and CSP
Yarn count CSP FQI

30 K 1800 35
1900 38
2000 42
2100 46
60 K 21oo 65
2200 70
2300 75
2400 81
100 C 2400 95
2500 101
2600 106
2700 112
K = carded, C = combed
Thus with the introduction of fiber quality index concept, basis for scientific
evaluation of cotton technological value that is quality and cost was build up.
As mentioned in chapter 4, it has also facilitated in establishing statistical
correlation amongst fiber-yarn-process, but no doubt partially. As it has given
mathematical correlation between index value and at least one of the important
yarn property (that’s strength for FQI and uniformity for Q), as given below:
Lea CSP = 320 ( )

FQI + 1 - 13C ; Carded yarn
Or
Lea CSP = ÈÎ320 ( ) Ê

FQI + 1 - 13C ˘˚ Á1 +
W ˆ
Ë 100 ˜¯
; Combed yarn
Where,
C = count spun,
W = Percentage Noil
Ê d - 1ˆ
u 2 = Q.Ne Á + a ( d - 1) + ur2
Ë d ˜¯
Where, Ne = Yarn count, d = draft on the R.F., Ur = Roving Unevenness (U%),

a = Contribution of R.F. to yarn, U = yarn Unevenness and Q = Drafting
quality index.
Rating: U% if found 3.5; referred as spinning under good condition.
This has also initiated practices for define process performance with
respect to only one of the important property of yarn, viz; strength for FQI.
Thus, quality index concept has given statistical tool to predict about process
performance abide by actual yarn value (strength) and theoretically calculated
one from the known fiber parameters.
Actual yarn CSP

Process Proficiency (P%) = ¥ 100
Theoratically Calculated CSP equation 4.20
Process ratings were given on this ground accordingly;
P(%) ≥ 100; Good

= 95; Average
= 90; Poor
Fascinating concept of FQI was suffered from limitations also. It was
determined on the basis of only four fiber parameters on linear relationship,
which in practice is not fully true. More or less contribution from rest of the
properties also seen on material and process performance which is also not
linear in nature. In absence of any other option, SITRA has suggested that the
FQI and the yarn CSP estimations should be taken as a broad guide line only
by the mill to judge the cotton quality as well as spinning performance.
On summation the major limitations of first phase of statistical analysis are:
1. The entire statistics right from the analysis of test data to optimisation
of mix quality and cost was done manually. This lengthy and
tedious statistical calculation to be performed before drawing useful
conclusion, usually resulted in delayed decision-making and liable to
suffer from human associated errors.
2. Broader sense linear correlation was established between four
important fiber parameters and one of the yarn parameters by FQI.
Going a step ahead, process performance was judged on the basis of
this predicted yarn property in relation to actually achieved one. But all
these were done by giving consideration to four main fiber parameters
and ignoring rest. In fact it is needed that theoretical multivariate
polynomial index to be derived by including of all fiber parameters as
per their participation in meeting very closely actual yarn values. In
the absence of such accurate correlation amongst all fiber properties
wide deviation of theoretical value from the actual one is likely. This
was reflected in reduced reliability of correlation found between actual
and predicted one.
3. Still human brain has the domination in decision-making that means
defining relationship amongst yarn parameters - process parameters -
fiber parameters.
Blending technology
Although more cotton varieties were available and used for mix formulation
as compared to classical time. But still the main challenge to meet about was
to produce homogeneous blend out of highly variable inputted cotton fibers
belonging to more than one origin. This critical task of bringing regularity and
equilibrium for the bales carrying within as well as between bale variations
in the cotton fiber properties was done very well in classical practices by
the high volume bale blending or stack mixing performed in the earlier
production stages. However, this work practice was found highly impractical
and discontinued. Elimination of stack mixing has shifted the responsibility to
work with equivalent high blending delay on the shoulders of bale feed. The
desired degree of blending was achieved in this period by tuft blending; it was
done at Blending Bale Opener (BBO), incorporated in the initial stage of blow
room line in the replacement of HBO. But capacity of BBO was quite smaller
than massive mixing room used earlier and that was resulted in considerable
reduction in blending delay. Even the cross blending was also extended up to
draw frame only as doubling practices at flyer frame and ring frame were also
discontinued.
Compensation for this reduced blending delay was realized by the use of
bigger bale lay down, up to 32–40, instead of 8–10 possible with HBO in
classical blow room set up. Increased number of bales per lay (N), almost four
times than earlier, along with almost identical feed tuft size (W) has allowed
proportionate rise in blending delay [eq. 6.5].
Weigh of Equivalent Blending Delay = N ¥ W 6.5
However, in the absence of stack mixing, the number of bales represented

at this point would be determined by the number of bales fed to the opening
machine, which was comparatively quite less. For example, if stack mixing
has provision of mixing 1000 bales then each bale at HBO feed represents
1000 bales due to cross mixing.
Bigger tuft were fed manually to the opening machine (BBO) and this was
the major restriction in acquiring homogeneity of feed cycles. Although it is
assumed that manual feeding was uniform, still larger size of fed tuft/ bale and
the small capacity of the hopper of the blending bale opener were unable to
represent all the bales uniformly within and between feed cycles and liable to
introduce variations. There was a full chance that all the bales in the laydown
may not be either equally or at all been represented in each feeding round
and thereby in the lap (Figure 6.3). This can be well explained by an example
given below:
Example: Let the capacity of the hopper is 20 kg and the tuft size fed per bale
per round is 2 kg for 32 bales in lay down.
 20 Kg 
• Hopper gets full by the time feeding is done from 10th bale =  
 2 Kg 
and rest of the bales do not get chance to be represented in 1st round.
• Similarly second round incorporates 11th to 20th bales and
• Third round 21st to 30th bales.
• Thus each round represents partly the bales used in the laydown and
thus each round is differing from each other and within itself.
• Scutcher laps so produced do not able to represent all the bales
uniformly thereby.
Figure 6.3 Working concept of blending bale opener.

Opportunities of cross blending were also reduced considerably with the
increased production rates attained at card and draw frames on the adoption of
improved technology. It was further reduced severely with the introduction of
chute feed system to the card. The blending delay time in this age was thereby
very small, would be 1 or 2 minutes, as material going to six different cards
will be separated in time only by a few meters. It utmost became necessary
to complete blending of material before feeding to the card. Thus problem
of messy blending was sorted out on the elimination of stack mixing but at
the cost of reduced blend homogeneity. Since evening out of the variations in
fiber characteristics requires minimization of within round, between rounds,
between laydowns differences. So, the mixing mode of this period was needed
to rethink in these regards.
6.4.3 Mix formulation with automatic instrumental technology
Textile industry has undergone radical technological changes at every front in

the beginning of 1970. Fully automated bundle fiber testing technique HVI,
followed by individual fiber testing technique AFIS were introduced to the
market. This capable and efficient testing mode has created revolutionary
changes in the entire mix formulation process. These was further supported
by the remarkable changes took place in the theory and design of spinning
machinery.
Transport, communication and agricultural technologies
Connectivity of the world in terms of roadways, railways, airways and
waterways was reached to its peak of development and made the world
very small. Seeking the information from the any corner of the world
became possible with the invention of modern communication tools like
microprocessor, internet, mobile and various apps. These have made targeted
cotton varieties details available from electronic data bank with a single click
by a person seating in a room only. The same has also made procurement
of desired cotton variety from any corner of the world speedy and easy to
work with higher number of cotton components in mix. This has boosted up
scientific cotton selection process.
Scientific approach was instilled in agriculture sector also. Farmers were
well educated and trained to use hybrid seed to earn better quality and quantum
of fiber production with due support of modern sophisticated agricultural
tools. This green revolution has added more options in terms of cotton varieties
available in the market. Manual picking system was replaced with mechanical
one to earn higher output. Roller gin has taken up the charge from saw gin
to deal with more cohesive hybrid cotton at high speed for preventing fiber
damage. However, technological changes related with growth, cultivation and
ginning has changed cotton factor (C) as well as trash factor (T) of the cotton
supplied to the modern blow room line. Machine factor (M) of modern high
production technology has maintained pace to respond changes in supply
material characteristics. All together have made possible to work with better
cotton variety procured either at same or may be lesser cost; enhanced quality
as well as economy of output yarn.
Testing techniques
The selection of raw material during mixing was done on the basis of
four important characteristics, viz; fiber length, maturity, strength and
fineness during non-automatic testing or semiautomatic testing era. These
cotton values were determined by using separate equipment and one has
to wait for a pretty long time for the test results, time interval was further
extended for more number of cotton varieties in consideration. Automatic
mode of fiber testing equipment i.e. the high volume instrument (HVI)
and the Advanced Fiber Information System (AFIS) has improved this
scenario of fiber testing. The HVI pragmatically determines most of the
quality characteristics of cotton within 15 seconds nowadays with testing
speed of 240 bales/h. AFIS operates at comparatively slower speed
but quite higher than semiautomatic testers. Using these instruments,
hundreds and thousands of cotton bales can be tested for several fiber
properties accurately as well as speedily. Thus, 100% quality check for all
the bales arriving at ware house becomes possible. This has eliminated the
hindrance involved in the cotton selection by insufficient fiber information,
resulting from a lack of capable and efficient testing technique. Accurate
and timely information of fiber properties has allowed to engineer input
fiber profiling. A mere base required for controlled within and between
bale variation distributions to earn long term mix homogeneity.
HVI provides information about group of fibers and AFIS about individual
fiber characteristics. As mentioned earlier these have significance role in
assigning scientific correlation between fiber parameters and yarn parameters
and correlation between fiber parameters and process parameters, respectively.
This has replaced art and experience based decision-making by a solid
scientific based in the fiber selection and mixing processes.
Statistical techniques
High volume instruments (HVI) and the Advanced Fiber Information System
(AFIS) instruments can perform speedy and accurate testing for hundreds
and thousands of cotton bales at a rate of 240 bales/h. Data provided with
these instruments can easily be analysed in terms of average and dispersion
value with microprocessor interfaced software programs. These data can be
displayed on the screen for simultaneous study or printed for record or saved
as soft copy in hard disk, CD, etc., for future record. This has saved time,
labour and money involved in testing as well as first step statistical analysis
of test data also.
Depending on the HVI test results indexes such as the fiber quality
index (FQI), (in earlier versions) and spinning consistency index (SCI),
are calculated to determine technological value of the cotton. The complex
mathematical calculation for multivariate polynomial equation of SCI is done
by the inbuilt software in microprocessor interfaced with HVI only. This has
eliminated human associated errors involved in such derivations and also made
the process speedy. Apart from that such indexes have eliminated need for
voluminous statistical compatibility tests (significance or critical difference
check) for sorting out cottons from the available varieties, saved cost and
human hours. Fiber selection done on SCI base has shown good consistency
in fiber properties due to accurately defined compatibility. Practically SCI is
used in first priority followed by Micronaire in the selection of bales in modern
ware house. Even experimentally significance of SCI based bale management
over classical mixing is well proven by researchers in following regards:
1. Very good correlation exist between SCI and yarn strength and quality
parameters (Fig. 6.4 (i)).
2. Bale management based on SCI ensures mixing consistency, thereby
consistent yarn strength executes very good correlation between SCI
and spinning end breaks (Fig. 6.4 (ii)).
3. Controls within and between lay down mix variations as a result
executes good correlation between SCI and yarn strength variations
(Fig. 6.4 (iii)).
Figure 6.4 Significance of SCI based mix. (i) Relationship between SCI and
yarn strength. (ii) Relationship between SCI and end breakage
rate. (iii) Relationship between SCI and CV% yarn strength
control on within and between lay down variations.
Presence of good statistical correlation between SCI and yarn performance
has facilitated in replacing the error some judiciary based decision-making
technique with scientific technique with added accuracy.
Optimised proportion of such sorted out cottons in the mix is still defined
by weighted average based linear equations [equation 6. 4], but resolved by
computer program instead of manually done calculations. Computer software
in the form of linear programming (LP) was developed by ATIRA. Details of
all cotton parameters and standards for these parameters are given as input to
this software; it will work out different possible combination within a very
short time and display the results on the screen. The close fit program of this
software also permits in identifying optimum mix combination very closely
meeting customer’s requirement. Thus development of microprocessor based
statistical technique; LP has optimised proportion of selected fibers in mix
scientifically, speedily and accurately in automatic mode of mix formulation.
Blending techniques
More number of cotton varieties were made available for mixing very easily
and speedily in this age. Increased components in the mix formulation process
has reduced drastically their share. Reduced contribution coming from each
cotton in mix has put one step forward in earning mix homogeneity for
prolonged time interval. SCI based selection has facilitated in maintaining

consistency of inputted fiber characteristics even though one of the component
need to be replaced. But simultaneously it was also a fact that within and
between bale variations of each cotton variety arriving at ware house should
be addressed by dynamic blending.
Small capacity hopper based tuft blending done at BBO was not sufficient
to handle correctly more varieties of cottons in mix and hinders the associated
advantages. Automatic Bale Opener (ABO) has replaced BBO in this era.
It has large capacity hoppers working with bigger lay size (180–200 bales
mixing) and fed with very small flocks (2–3 mg) even though working at
higher production rate (Fig. 6.5).
Figure 6.5 Automatic bale opener.

Smaller the weight (W) fed from each bale of a very big bale laydown
(mixing), to the large capacity hopper gives equal chance to all bales
to be represented in each feed cycle and thereby output to the blow room
[equation 6.8]. Feeding very small flocks per bale demands more rounds to
be performed for desired input capacity of the machine, also nullifies between
round variations. Thus modern blow room line fulfills all three basic needs for
attaining blend homogeneity, viz; use of bigger bale laydown (large mixing)
and working with reduced within and between rounds variations with the
introduction of ABO at the feed end. However, the value of blending delay is
considerably less as compared to stack mixing for modern blow room. This is
attributed to feeding of very small flock weight (W mg) from comparatively
less number (N); multiple of hundred instead of thousands of bales in lay [eq.
6.5]. This can be well illustrated with example given below:
Example: Stack mixing is done by feeding 3 kg of lumps manually out of

2,000 bales in classical spinning line.
Weight equivalent to blending delay = 3 kg × 2,000 = 6,000 kg
Flock mixing is done for modern blow room from 200 bale lay down with the
feed flock size of 2 mg.
Weight equivalent to blending delay = 2 mg × 200 = 400 mg
Automation in doffing, donning and material handling for small capacity
but high production machinery in line has also eliminated provision of cross
mixing at and the later stages to blow room, adversely impacted to blending
delay value for modern yarn production system. Concluding the discussion,
reduction in blending delay for modern blow room is observed due to:
1. Elimination of massive (n × 1000) bale mixing.
2. Continuous working mode of machines instead of ON/OFF mode.
3. Shorten blow room length by accommodating less number (2 instead
of 6) of small capacity machines.
4. Machines are operating at high production rates.
5. Elimination of high capacity storage trunks before machine and
6. Adoption of machine linking like chute feed and auto transport of
material has eliminated need for intermittent storage and thereby
respective cross feed.
Under such circumstances, attaining desired mix uniformity required
special blending efforts. Special mixing machine (Auto mixer/Multi mixer)
is introduced between pre cleaner and fine cleaner in modern blow room line
for this purpose. This capable blending machine has resolved the problem in
conjunction with automatic bale opener.
It offers three point blending, first cross blending and rest two due to
asymmetrical path lengths (Figure 6.6) for very small but uniform cotton flocks
arriving from big bale laydown. These fiber flocks are subjected to voluminous
but efficient blending by large capacity 6–8 feed columns of machine, with the
production speed compatible with rest of the machines in line. Depending on
intensity of blending needed, number of feed columns and their size, many a
times two blenders are arranged in a sequence to earn C2 level of blending,
where C represents capacity of one blender. Fully automatic operational mode
not only ensures mix homogeneity but also provides ecofriendly working
environment for modern spinning preparatory. Thus critical task of translating
highly variable input material to long term consistent product becomes
possible due to three point technological changes in recent time, even though
operating at very low blending delay value. They are:
i. Accurate and speedy information about fiber properties on the basis of

100% quality check.
ii. Use of ABO, which has regularised within as well as between feed
cycles by feeding very small flocks from a big bale lay down.
iii. Use of high capacity and efficient automatic blending machine.
iv. Consistent input fiber profile due to scientifically selected large number
of components in optimised proportion for the mix formulation.
Figure 6.6 Mixing machine working concept.
References
1. Application Handbook of USTER HVI SPECTRUM, Zellweger Uster,

pp 1.1–1.9, 1999.
2. EI Mogazhy Y.E., Broughton R. and Lynch W.K., 1990, A statistical
fiber properties. Textile Research Journal, 60(9), 495–500.
3. EI Mogazhy Y. E. and Gowayed Y., 1995, Theory and practice of cotton
fiber selection, Part I: fiber selection techniques and bale picking
algorithems. Textile Research, Journal, 65(1), 32–40.
4. EI Mogazhy Y.E. and Gowayed Y., 1995, Theory and practice of cotton
fiber selection, Part II: Sources of cotton mix variability and critical
factors affecting it. Textile Research, Journal, 65(2), 75–84.
5. Hequet E., Wyatt B., Abidi N. and Thibodeaux D.P., 2006, Creation
of a set of reference material for cotton fiber maturity measurements.
Textile Research Journal, 76(7) 576–586.
6. May O.L. and Jividen G.M., 1999, Genetic modification of cotton fiber
properties as measured by single and high volume instruments. Crop
Science 39, 328–333.
7. May O.L. and Taylor R.A., 1998, Breeding cotton with higher yarn
tenacity. Textile Research Journal, 68(4), 302–307.
8. Meredith R., 1945, The tensile behavior of raw cotton and other textile
fibers. Journal Textile Institute, 36, T107-T-130.
9. Meredith W.R., Culp T.W., Robert K.Q., Ruppenicker G.F., Anthony
W.S. and Williford J.R., 1991, Determining future cotton variety fiber
quality objectives. Textile Research Journal, 61(12), 715–720.
10. Mohammed Rubaiyat Chowdhury, Md. Saiful Islam, Dr. Amir Hussain
Khan and Dr. Abul Kalam Azad. Importance of mixing for selecting of
cotton fibre.
11. Ratnam T.V., Chellamani K.P., 1999, Quality control in spinning; Third
Edition - Raw material, The South Indian Textile Research Association,
Coimbatore, pp 31–66.
660–665.
13. StatSoft, Inc. (2007). STATISTICA (data analysis software system),
version 8.0. www.statsoft.com.
14. Sreenivasa Murthy H.V. and Samanta S.K., 2000, A fresh look at fiber
quality index. The Indian Textile Journal, 111(3), 29–37.
15. Sunil Kumar Sharma, 2014, Cotton yarn quality depends on mixing
strategy. The Indian Textile Journal, 124(6), 1–19.
16. USDA. 1999. International Calibration Cotton Standards program.
Organization and operating procedures. USDA-AMS, Cotton Program,
Memphis, TN, March 1999.7.
17. USTER ®AFIS PRO Application Hand Book
7
Cotton fiber engineering


*
Abstract
The process of cotton fiber selection as mentioned in the previous chapter has
undergone an inevitable transition from the traditional pure art to a modern
sound scientific technique. The success of this transition requires integration
of fiber selection process with fiber engineering program, which is designed
to optimise cotton fiber use with respect to quality and cost of end product
(yarn/fabric/garment etc.). A cotton fiber engineering program should be
based on fiber information that meets quality requirements imposed by the
rapidly developing technology and frequently changing customer demands.
Purchasing strategies should also optimise cotton blend components under
inventory and quality control. Scientific procedures that use parametric linear
programming (mentioned in earlier chapter) can be the most powerful tools
for achieving this critical task.
KeyWords: bale management, EFS, BIAS, linear programming
7.1 Introduction
Raw material and technology upgradation are often recited as the major decisive
factors in defining the quality of product, performance and profitability of a
textile mill. However, there is always a limited scope available in regard to
raw material selection especially for natural fiber like cotton. Availability of
quality and quantity of cotton is highly influenced by crop availability and
price levels in the country, crop cultivation conditions, monsoon, lot to lot
and station to station variations in quality, ginning conditions and government
policies. Further restriction arises from the non-precise co-relationship
that exists between measurable characteristics of cotton and yarn quality,

as only 70–80% of yarn quality is being determined by measurable fiber
properties. Technology upgradation and modernisation on the other hand
are more rewarding in the long run for improving the mills’ product quality
and performance. However in actual practice, the gain of modernisation is
not fully realized (in some cases not even partially realized by mills). Even
returns from modernisation also differ markedly from mill to mill and always
conflicting views are prevailing among the mill executives and technicians
in regard to priority areas of modernisation and technology choice. The
identification of the relationship between cotton fiber and yarn parameters
with respect to production technology is therefore important. The approach
of applying statistical modelling, simulation and artificial neural network
techniques in prediction of yarn parameters form known fiber parameters and
process parameters can develop a new dimension in fiber-to-yarn modelling.
7.2 Importance of cotton fiber engineering

Cotton being a non-homogeneous natural fiber has high variability in
its characteristics. Conversion of cotton fibers having high variation in
characteristics into consistent yarn quality is a cherished goal sought by every
spinner. This objective can be achieved with the aid of bale management
system and hence it has become an integral part of quality management system
in most of the export oriented spinning units. The technique of appropriate
bale selection in a laydown for consistent end product demands the grading
of each and every bale in the population on the basis of fiber properties. The
engineered fiber selection (EFS®) system developed by Cotton Incorporated
in 1980 provides the tools and algorithms that allow computerised bale
management. EFS® system entails the procedures that make use of the
frequency distributions of HVI data of the cotton bale population to establish
a category or group scheme from which bales are selected for the laydown
using bale picking algorithms like random picking, proportional weight
category picking, optimum category picking, etc. The main drawback of this
system is that as the number of fiber criteria and their level increases, the
number of category combinations increase overwhelmingly. In an attempt to
simplify the bale management system, some effective indices such as fiber
quality index (FQI), the spinning consistency index (SCI), and the premium
discount index (PDI) have been developed based on the multivariate regression
model. These indices usually depend on the range of bales used to formulate
equations and seldom generalise to characterise the complex multivariate
nature of cotton fiber properties. For example, SCI regression equation was
derived based on fiber properties of Pima and Upland cotton and it may not
replicate a good fit with the Indian cotton. Moreover, indices like FQI consist
of linear combinations of the fiber properties; however, a highly non-linear

relationship actually exists. These drawbacks need to be considered at the
point of application.
7.3 Attributes of cotton fiber engineering

The process of cotton fiber selection as mentioned in previous chapter has
undergone an inevitable transition from the traditional pure art to a modern
sound scientific technique. The success of this transition requires integration
of fiber selection process with fiber engineering program, which is designed
to optimise cotton fiber use with respect to quality and cost of end product
(yarn/fabric/garment etc.). A cotton fiber engineering program should be
based on fiber information that meets quality requirements imposed by the
rapidly developing technology and frequently changing customer demand.
The proposed scheme for such program thereby includes four interactive
attributes demonstrated below:
1. Cotton purchasing strategy
2. Cotton testing
3. Bale management and
4. Cotton fiber selection (Fig. 7.1).
Figure 7.1 Structure of cotton fiber engineering.

7.3.1 Cotton purchasing strategy
Cotton fibers must be able to meet technological requirement primarily. This

requirement differs as per type of production process, end use requirement, yarn
size, etc. For example, cotton suitable for ring spinning may not necessarily be
right for rotor spinning, friction spinning or air jet spinning. This is due to the
fact that the relative importance of fiber properties are not same for different
spinning systems, like fiber length is the major factor for ring spinning, but
for rotor spinning fiber strength has major consideration. Similarly, for the
given set up of spinning system, process parameters like; yarn size, twist, etc.,
and end product specifications like strength, feel, cover, etc., also influence
the kind of cotton to be used. In fact, the cotton that may have a premium
market value is not necessarily the best for the particular process or end use
application. Thereby cotton fiber purchase strategy should only be based on an
evaluation of the technological value of cotton. Purchasing strategies should
also optimise cotton blend components under inventory and quality control.
Scientific procedures that use parametric linear programming (mentioned in
earlier chapter) can be the most powerful tools for achieving this critical task.
7.3.2 Cotton testing
Details about various fiber testing techniques are given in depth in earlier
chapter. Even their importance on product quality and cost is also briefly
summarised. It is well known fact that traditional fiber testing was a tedious
and slow process, requiring four to six hours on an average for testing one
cotton fiber sample for the various properties. Because of this, only a small
proportion of the bales in the lot were tested and cotton procurement was
based on nominal specifications only thereby consistency of fiber quality
could not be ensured.
With the introduction of HVI, the testing capabilities increased. By the
late 80s, the instrument was capable of testing 150 samples/h, which have
been increased up to 240 bales/h with latest version. The testing includes all
the traditionally used parameters and in addition elongation, colour and light
reflectance are also measured. Spinning consistency index (SCI) is obtained
by microprocessor interfaced with HVI and also generates 8 digit Bale ID (bar
code) based on this SCI value. Thus, each bale coming to the modern mill is
thoroughly tested and provided with unique 8 digit barcode (ID) for their easy
identification in terms of the variety, the station and all the important properties.
Hence HVI fiber testing is done on bundle of fibers and therefore fiber
parameter distribution is not available. Advance fiber information system
(AFIS) can assess the fiber individually, so now distribution of important fiber
properties have become possible. This has served as the best verification tool
for nominal values of fiber characteristics.
Thus, use of HVI and AFIS in fiber testing has not only made 100% testing
of fiber characteristics but also its variations are possible and will produce an
accurate evaluation of the fiber values and variability in the bale population.
Generated data can be used for bale management and for implementing
suitable fiber selection techniques.
7.3.3 Bale management
Bale management basically deals with the judicious selection of cotton bales
in the lay down in order to obtain consistent quality as well as economic
spinning performance. The fiber property variations are liable to give shade
variations in fabric on dyeing or finishing. Thus, in broader sense, objectives
of bale management program are;
1. To get uniform yarn quality and cost economy.

2. To minimise shade variation of the finished fabric.
3. To reduce or control fabric barre.
It is the third important step of cotton fiber engineering and in dealing with
storage and retrieval of cotton bales. Two main methods of bale acquisition can
be used: (i) Storage and retrieval by group and category or (ii) Based on bale
ID. selection of one out of two methods depends on the ware house structure
and the bale population size. A 100% bale testing by HVI is mandatory for
both the methods and based on these test results the bales are either categorised
as in the former case or allotted with auto generated eight digit bale ID.
The categorisation is done on the basis of different fiber properties,
usually on the basis of micronaire (fineness) and colour [chrome (+b)]
value. After categorizing by either method, the list of bales for each lay-
down of required number of bales is given. The average fiber properties
from lay-down to lay-down are always kept constant. The list is sent to the
warehouse. From there the store personnel procure the bales according to
the category or else send a list of the ID and numbers of the bale. Then
the quality assurance (QA) personnel again verify; whether these bales
are capable of fulfilling the requirements or not and once gets assured,
the lay-down plan is prepared. During preparation of lay-down plan, one
important thing always kept in mind is that bales of same categories should
not be placed right next to each other. Then according to the lay-down plan
the bales are arranged under the Bale Plucker or Automatic Bale Opener.
Thus, bale management deals with testing, sorting and mixing of bales
according to the properties of fiber for producing specific good quality of
yarn at minimum cost (Fig. 7.2).
Figure 7.2 Bale management.

Various definitions are given for bale management. Among them popularly
used are as follows:
Bale management is a sophisticated attempt to acquire the raw-material
with desired characteristics to parameters categorised and store bales
according to predetermined patterns and to select and blend bales in order to
attain a consistent and uniform product.
Bale management is a process to mix fiber homogeneously to get consistent
production, quality of yarn, inventory control and selection of fibers according
to its properties.
According to the fiber characteristics:
Bale management refers to a choice of cotton bales in order to achieve
acceptable and a constant yarn quality and economical processing conditions.
7.3.4 Cotton fiber selection
Fiber selection approach should meet two main objectives:
1. It should achieve a uniform profile of the characteristics of input fibers

and its corresponding end product.
2. It should maintain the average values of output characteristics at their
desired level.
From an economical point of view, a proper fiber selection strategy should
result in better bale management, improved bale acquisition, improved mill
efficiency and optimum cotton use.
A fiber selection program should involve four basic steps:

(1) Examining the population(s) distribution of fiber properties of cotton
bales.
(2) Implementing reliable bale picking schemes, based on the distribution
of fiber properties of the bale population(s) and bale management
methodology.
(3) Control average output characteristics by developing reliable fiber–
yarn relationships. Regression analysis is a common approach for
developing such relationship.
(4) Verify the effectiveness of the fiber selection program by monitoring
the uniformity of the fiber characteristics of bale lay downs and
corresponding yarn characteristics.
Accordingly, fiber selection technique for a cotton fiber engineering
program should involve broadly two main procedures: (i) Implementation
of suitable bale picking schemes and (ii) Fiber/yarn modelling. The first one
provides mill to work with uniform fiber characteristics on a mix-to-mix
basis without violating inventory constraints (steps 1–2), whereas second one
controls desired output characteristics (steps 3–4).
7.3.4.1 Different bale picking schemes
There are three different bale picking schemes, viz; (i) Random picking, (ii)
Proportional weight category (PWC) picking and (iii) Optimum category
picking (OPC). These schemes mainly aim at reducing between-mix variability
by producing a consistent average of fiber characteristics on a mix-to-mix
basis.
The random picking scheme resembles the traditional massive blending
and served as the basis for more advanced schemes. In proportionate
weight category picking, the distribution of fiber characteristics are divided
into number of classes and bales are picked from each class in quantities
proportional to the class relative frequency. In optimum category picking,
bales are selected on the basis of optimisation of factors contributing to blend
uniformity, including category variance, picking cost and category inventory.
EI Mogahzy et al. (1992, 1994) have done an extensive work in this direction
and according to them, proportionate weight category picking scheme can
ideally be used for normally distributed fiber characteristics. But for the
cases where category variances are substantially different, although following
normal distributions, they have highly recommended use of OPM.
The fundamental difference between random picking and category picking
lies in the method by which population variability is controlled in selected
mix. When ideal random picking is used, every bale mix selected from
finite bale proportion will have an average fiber characteristic value equal to
that of population. Between-mix variability should be less than population
variability by an amount dependent on the number of bales in the mix. Within-
mix variability will be more or less identical to that of the parent population.
Category picking will execute similar trends, except that within-mix variability
can be reduced by controlling the picking order.
In an idealised random picking scheme, each bale in the population has
an equal opportunity of being picked. When the population is large, “N”, or
the selected mix contains small number of bales “n”, random picking fails
to replicate population heterogeneity in the selected mix. The problem is
further pronounced in populations exhibiting high variability in their fiber
characteristics. In practice, ideal random picking is difficult to achieve, this
further adds to the deficiency of the system in comparison with category
picking.
In case of category picking scheme, the entire population is divided into
say “k” number of categories and then bales are picked randomly from each
category. Thus, it provides better replication of population variability by
ensuring representation in mix of different values in the population.
The category picking has further higher opportunities to enhance mix
uniformity by optimising a number of critical factors. They are:
i. Population variability,
ii. Location of category break points in the distribution of fiber
characteristics,
iii. Number of categories in the distribution of fiber characteristics and
iv. Number of bales per mix (or bale laydown size).
A key factor pertaining to all the above mentioned factors are the particular
fiber property used in the picking scheme. This factor involves the testing
method, the reliability of test data and testing time.
7.3.4.2 Different techniques for fiber/yarn modelling
The prerequisite for success of cotton textile manufacturers in today’s global

market is to get reliable and cost-effective quality product. The primary
mechanism for achieving this is (a) appropriate blending of cotton and (b)
proper setting of textile machinery. Hence both being based on accurate and
speedy measurements of critical fiber properties, done by HVI and AFIS.
Data generated by these instruments can easily be utilized with
microcomputers and powerful software programs. Since the prerequisite for
engineered cotton fiber selection and blending is shifted towards the

development of models that more accurately predicts the processing efficiency
and yarn quality by reliable fiber-yarn relationships based on objective fiber
properties and also verify their effectiveness. The following software based
techniques are available in this regard:
1. BIAS (Bale inventory analysis system)
2. EFS (Engineered fiber selection)
3. Linear programming technique
These revolutionary developments have led to the substantial rethinking
of cotton fiber selection, driven by the rising costs of both labour, raw
material and the more demanding quality requirements of end products.
7.4 Bale inventory analysis system (BIAS)

Bale inventory analysis system popularly known as BIAS is developed by M/S
Zellweger Uster. The basic aim involved in this development is formulating
fiber mix in such a way that bales taken from stock carries minimum variations
in quality parameters from bale to bale and day to day. The entire process can
be bifurcated into following steps:
1. 100% bale testing: All the bales received are tested for all fiber
characteristics by HVI. The test results are exported to BIAS either
directly (Electronic data transfer) or by devices like floppy, CD, etc.
2. Categorisation of bales: All the tested bales are numbered and divided
into categories on the basis of SCI (Spinning consistency index) and
Mic (Micronaire).
3. Mix formation: First of all, number of bale categories out of which
mix will be formed is decided. For example, mix is formed out of 21
bales by making following categories:
Categories for SCI Categories for MIC

<70 <3.0
70–80 3.0–3.2
80–90 3.3–3.5
90–100 3.6–3.8
100–110 3.9–4.1
˃110 >4.1
The decision of categories should be based on the minimum thousand

bales analysis done for a particular season. Since six categories are defined
SCI wise and exactly same for MIC in the above example, the total number
of categories become 6×6=36. So, all the bales tested are stocked into 36
categories in such a way that total of SCI category in any direction will be the
same. Now bales issued are subtracted out from the stock.
4. Allotment of bale number as per mix plan: Bale number is entered
to each bale issued and average of all quality parameters will
be printed.
Thus, BIAS helps in minimising variations in quality parameters on a day
–to–day basis for particular mixing. Fiber information from HVI and yarn
information from Tenso-rapid have been used to improve performance and
profitability of the yarn manufacturing process.
7.5 Engineered fiber selection (EFS)

The process that was primarily based on individuals (expert’s) instinctive and
subjective judgment, is translated to a total scientific engineering system of
fiber selection with the development of powerful tools like bale management
and fiber selection software programs. The former one is described above and
later one also commercially known as engineered fiber selection or EFS is
developed by Cotton Incorporated and a leading program in this regard.
The sequence followed for scientific fiber to yarn and later on fiber to fabric
engineering, by using bale management system software is as follows:
(1) Determination of cotton specifications
(2) Opening line configuration and availability
(3) In-house inventory management
(4) Mix profiles
(5) Bale selection
(6) Mix evaluation and performance verification.
Each of the above mentioned components (step) of EFS is briefly reviewed
in the appending paragraphs to clarify the importance of HVI data to cost
effective raw cotton management as well as mill and product performance.
7.5.1 Determination of cotton specifications
Over the years procurement of cotton was done solely on the basis of cotton
cost without considering need of mill and product, thereby could not able to
compete effectively in their markets for long run. But scenario is different
with the development of scientific fiber selection technique. Here selection
of cotton specifications are based on a multivariate function, where variables
include end-product (yarn/fabric) performance (quality and cost) expectations,
the machine components like process flow design and related maintenance,
settings, production rate, etc. and the management philosophy of a given mill.
Specifications of the end product performance are identified and assigned
in terms of desired cotton attributes, intermediate product and machinery
performance requirements. This aspect of cotton management is well realized
with the due utilisation of various quality control reports and charts provided
by the bale management system software. Simultaneously, desired end product,
i.e. yarn or fabric, quality parameters like appearance, tensile strength, tear
strength, bursting strength, etc., and associated each process parameters like
production rate, efficiency should be collected and analysed on a daily basis.
The same bale management (cotton engineering) system software can be used
to assist the appropriate manager in determining which process or product
variables are co-related to cotton HVI measured properties.
Once such correlations are understood, then cotton purchasing can be
conducted with a technical perspective that will likely ensure the purchase of
cotton using HVI data with a profile that leads to improved mill and product
performance.
7.5.2 Opening line configuration and availability
It is never been cost effective to spin multiple yarn count from the same
cotton mix on multiple spinning systems. But still such situation arises where
spinning needs to be carried out from large quantum of cotton. It can usually
be revealed that the quality of the laydown is geared with respect to the most
critical spun yarn among all varieties instead of giving preference to the
highest volume yarn in a group. The cost of directing such expensive cotton
into yarns that have less critical specifications will quickly be equal to the cost
of an additional opening line.
Consistency of yarn quality and cost for long production time is purely
affected by the homogeneity of mix earned at the spinning preparation step.
As mentioned earlier also technological changes at this front have changed
the capacity and efficiency of spinning preparatory set up. Opening line
configuration in terms of working principles adopted for opening, cleaning
and blending machines are the major amongst all, leaving remarkable
influence on blending machine efficiency and thereby consistent yarn quality.
Blending efficiency here demonstrate fiber mix produced with constant
variation throughout. The inherent blending efficiency of each mill’s opening
lines has a direct influence on the acceptable distribution (SD) about the mean
that can be processed predictably as well as efficiently into a consistent yarn

quality. Blending efficiency should not be confused with cleaning efficiency.
The cleaning efficiency has been demonstrated to be the best when the cotton
is cleaned with the least amount of work possible in order to reduce the chance
that the fiber will be nepped up and shorten.
Better opening and cleaning performance of the opening line drops
remarkably. The level of fiber damage occurs in terms of short fiber
content, neppiness and restricts acceptable distribution about the mean
for different fiber variants in mix, i.e. enhances blending efficiency. For
example, today’s fiber preparation lines have been substantially shortened
and it usually composed off four machines: Feeding machine (Bale
plucker), pre cleaner, blending machine and fine cleaner. Very small fiber
flocks detached by bale plucker are blended well by using two or three
fiber mixer (depending on capacity of machine and number of blending
variants), opened and cleaned by two cleaning units (pre cleaner and fine
cleaner) before feeding to carding process for final stage blending, opening
and cleaning. Throughout these sequential processes, the bulk of blended
fibers are gradually and rapidly reduced to the size of carded sliver. The
gentle opening and cleaning actions along with vigorous blending happens
at modern blow room and carding machine has allowed opening system
to attain higher blending efficiency for the very wide (variation) range of
critical fiber properties such as Micronaire.
Hence, the range of cotton properties to be purchased to form a mix is
dependent on blending efficiency of respective opening line. The less efficient
conventional line has thereby narrowed down this range, as compared to that
efficient modern line, which has widen up this range and thereby offered a
competitive advantage. Thus knowledge about opening line configuration
in terms of its blending efficiency is a must to start with yarn engineering.
The bale management system software program assist mill management in
determining the best fit fiber dispersion (CV%) values for different HVI based
fiber properties measured through the use of control charts. They serve as
statistical tools, used to plot averages and dispersion (CV%) values of fibers
to parameters influencing mill quality and efficiency; like end breakage rate,
yarn irregularities, etc., and thereby facilitate in establishing their correlation
trends.
In modern opening lines, the most efficient method to dramatically improve
blending efficiency is to install two 10-cell multi mixers in tandem. Such a
configuration has a doubling factor equals to 102 or 10 × 10 = 100. Apart from
the settings that affect flock size, line run time efficiencies and throughput per
hour also have influences on actual or effective blending efficiencies.
7.5.3 In-house inventory management
The mill inventory management may be divided into three basic approaches:
i. Mill-owned central warehouses
ii. Mill warehouses at each plant
iii. Just-in-time shipments from merchant warehouse
Mill-owned warehouse may be central or plant located. The bale
management schemes used by many mills can be divided into three basic
concepts mentioned above.
1. Mill-owned central warehouses: In this case warehouse is located at
individual plant to receive the cotton from multiple merchants as per
mixing requirements. These cottons are then processed at the receiving
plant only regardless of its quality.
2. Mill warehouses at each plant: As the name suggests in this case
warehouse is central or plant located to receive the cotton from
multiple merchants as per mixing requirements. These cottons are then
distributed in uniform lots or mixes to individual plants for processing.
3. Just-in-time shipments from merchant warehouse: In this case a
merchant in a Just-in-time relationship with the mill ships (supplies)
cotton directly to each plant site as per laydown or mixing planning.
Just in-time shipments generally refers to an agreement between
a mill and its cotton suppliers to ship the mixes, selected by cotton
fiber engineering program, to its various plants rather than normal
truckloads of cotton randomly drawn from the inventory purchased
from the supplier. The main advantages of such shipments are two-
fold, first the inventor held on site by the mill is minimised; and second
the mixes drawn from large inventories typically held by suppliers
that use software system and tends to be quite uniform from mix-
to-mix over time. Both of these factors have the potential to reduce
the costs. Disadvantages include higher shipment cost and distance
to the suppliers warehouse, which may result in the necessity to hold
excessively large number of bales on site. It is important to understand
that special just-in-time like services may generate mill savings while
adding to supplier cost. Therefore, they may command a premium
while still providing overall cost advantages.
The EFS® – MILLNET program includes an Electronic data information
(EDI) translation program to pass 863/856 ANSI X12 documents between
cotton merchants and mills. The use of this extensive information like weight,
bale numbers, HVI properties, price, etc., contained in this document can
be used to fill in most of the needed fields in the fiber engineering system
program to manage a cotton department. Thus eliminates tedious and error

some hand inputting of data. A highly advantageous feature of receiving such
electronic data information document is the verification of the HVI properties
of shipments before they are sent from the vendor. If there are any great
differences between averages of the current inventories HVI properties and
the arriving shipments, the mill manager can prepare in advance for necessary
changes required in mix selection.
In case shipments are not strictly managed for HVI properties from the
vendor, the advantage of the central warehouse is that the HVI data of each
bale can be reviewed and only bales most suited for a given plant or product
are shipped to a selected plant. This can be a major cost advantage to the mill.
The advantage of direct shipments to each plant is that the cost associated with
redistribution is eliminated. However, performance may be compromised by
either cost or performance factors that might be easily dealt with in a central
warehouse.
Mills that traditionally carry small cotton inventories, compared to their
consumption, generally report that additional costs associated with direct
shipment of cotton from merchants may be offset by the savings associated
with better machine efficiencies, product quality, reduced in-house warehouse
costs, and lower effective cotton costs. These savings ensue from the fact that
merchants can pull lay downs from inventories that are much larger than that
held by a typical mill. As larger the inventories can make it possible to select
mixes that are uniform over long time interval, thus, increasing improved mill
stability.
Inventory size is found to be related mainly with two factors, the first being
the number of bales required to cover any delayed delivery of cotton from the
mill’s suppliers. The second determines the minimum size of the inventory,
it should be large enough to ensure that no single replenishment can change
significantly the averages and distributions (CV%) of the important HVI-
controlled cotton properties of the on-site inventory. The degree of control
can achieve only when each bale HVI data gets available.
One of the common mill practices is storing the cotton by fiber groups in
their warehouses because it places bales with similar HVI properties together,
thus making retrieval of mixes easy. HVI categories are normally set within a
group, based on mill and end product requirements. Almost without exception,
mill finds it necessary to create 3, 5, or, in some cases 7 Micronaire categories.
It is difficult to overemphasise the importance of Micronaire management to
mill and end-product performance.
Other HVI properties that are likely to be considered for the purpose of
categorising cotton are length, length uniformity and strength and colour.
Normally, 18–45 categories are created as working with too few populated
categories will not allow to realize the expected degree of fiber control. On
the contrary, working with too many categories can make warehouse space
utilisation insufficient.
7.5.4 Mix profiles
Each yarn quality should be considered as a candidate for its own mix.
Practically this depends on the quality of yarn being spun and the number
of opening lines. In every event, the mix profile for a given mills opening
line should be selected to maintain appropriate control of those HVI-measure
properties which are deemed by the mill to be important to machinery
performance, efficiency and product quality.
The cotton fiber engineering software system provides users with
histograms, colour charts and also controls charts which can be used to
establish the best mix profile. This provides even running mixes over long
periods, mainly attributed to control distribution CV% of selected HVI-
measured fiber properties. The distribution CV% of fiber properties should be
no longer >10 and in many instances not >5. Anyway actual value is dependent
on mill, end product parameters and available inventories.
7.5.5 Bale selection
Bale selection considerations include the determination of the required run

time of the mix. More importantly, the bale selection process should include the
creation of mini-mixes within the mix to ensure that the cotton fed in opening
and cleaning line is always representative of the mix’s overall properties and
distributions. Waste cotton and/or other non-HVI measured bales should be
distributed uniformly within the mix and not grouped at one location.
For best utilisation of existing inventory and to avoid step changes in HVI
averages or distribution (CV%) values, bales are best selected each time a
mix is required. In other words, it is better for the process control and product
quality to let each mix correctly reflect the averages and distribution (CV%)
values of the existing inventory than to run exactly the same mix. This is true
because inventory variances make it virtually impossible to select a mix that
can be run over long periods without substantial changes in property averages
and/or distributions.
By definition, any category system that is created by a mill that has so
few categories that it seems possible to run the same exact mix for a long
period of time probably does not provide proper control of the cotton they
are processing. Such mix programs inevitably result in step changes in the
average or distribution (CV%) of one or more important HVI-measured
cotton properties.
7.5.6 Mix evaluation and performance verification
A great deal of useful information may be obtained by cotton fiber engineering

system program to provide mill management the averages and distribution
(CV%) of each mix plotted as control charts. The variations reflected on these
charts can quickly be correlated with mill production, efficiency and quality
considerations. In some mills, mixes are selected far enough in advance to
allow management the opportunity to request that the mix be reselected when
certain advantages and/or distributions appear to be out of line.
By closely monitoring mill process and product performance using various
observations and reports, cotton buyers typically can refine their purchasing
practices to lower costs while improving mill operations and lowering inventory
levels. In many cases, labour cost can be reduced as well throughout the company.
7.5.7 Benefits offered by EFS®
The benefits most often reported with the use of EFS® users are:
1. Use of HVI data eliminates cotton bale sample cutting and classing at
the mill.
2. Reduction of inventories carried by the mill.
3. Just-in-time delivery of cotton improves quality.
4. Yarn quality is improved including yarn count variation, strength and
Uster statistics.
5. Fewer fabric defects.
6. Elimination of cotton mix selection as a cause of barre.
7. Reduction of comber noils without loss of quality.
8. Improved efficiencies lead to lower labour costs.
9. Improved warehouse management.
10. Short fiber control gets improved.
11. Adjustment of mix averages and distribution (CV%) values based on
incoming but not yet received cotton.
12. Better contract management, reporting and improved communications
or understanding between textile mills and their cotton suppliers.
The system is capable of receiving all HVI data and related cotton
transaction information electronically, thus eliminating most of all routine
keypunching. Warehouses are managed using portable barcode terminals.
Both batch and Ring frame units made by symbol are supported.
7.6 Linear programming

The economical aspect of blending results from its ability to reduce cost
by providing an opportunity to substitute a less expensive cotton in a blend
without impairing quality or processing efficiency, by controlling the
percentage of fiber components in the blend and by permitting the use of
available cotton varieties efficiently throughout the entire crop year without
affecting processing performance or cost structure.
The recent transition to marketing of cotton on the basis of HVI
measurements calls for a scientific economical approach by which cost and
quality can be related and cost optimisation achieved. However, such approach
should meet two criteria: the approach should be fundamentally sound, and it
should be based on understanding the extent of cotton use in manufacturing.
Ram et al. and Bezdudnyi (1965) suggested the use of linear programming for
cotton cost optimisation, but the practicality of this approach was hindered by
a number of obstacles, including slow methods of fiber testing and the lack
of powerful computing systems. No doubt linear programing has undergone
several developments along with fiber testing techniques over the last few
years that have boosted up optimisation procedures. Powerful computers and
efficient software programs have rendered the mathematical complexity of
this analysis a minor issue. This has denoted practical merits of this method
and provided validation of its output and made this mathematical technique
extensively useful in engineered fiber selection for optimising cost and profits.
Linear programing in short form also known as LP technique and
described in detail in chapter 4. It is basically a mathematical technique for
maximising or minimising a linear objective function that specifies the benefit
or cost associated with each decision variable. In the case of cotton blending,
the decision variable is the proportion of a particular cotton to be used in the
blend. The objective function is a linear function of the cotton cost, which can
be represented by the following expression:
i=k
Minimize Z = Âai ci = a1c1 + a2 c2 + º ak ck 7.1
i =1
Where, Z is the total cost, ai is the proportion of the ith cotton type in the
mix and ci is the cost of this cotton.
Associated with this function is a set of constraints that may express
minimum quality requirements of raw material or inventory limitations.
Constraints may take the form of an equation or an inequality, depending on
the nature of constraint. In the case of cotton blending, an obvious constraints
is the sum of proportions is unity, or
i=k
Sum of Proportions = Âai = a1 + a2 + ºak = 1 7.2
i =1
The second constraint is based on quality characteristics that can be

expressed by a formula of the following linear forms:
i=k
Sum of Qulaity = Âai Qi = a1Q1 + a2 Q2 + ºak Qk ≶ Q * 7.3
i =1
Where, a1, a2….ak are the proportions of the components of the blend and Q1,
Q2…..Qk are their corresponding average HVI fiber characteristics values and
Q* represents the quality limit; minimum or maximum desirable value.
7.6.1 Assumptions of linear programming
Mathematical formulation of a linear programming model is based on

linearity, i.e. linear forms of both objective functions and constraints.
Linearity includes proportionality and additivity. In the case of cotton
blending, proportionality implies that the contribution of each cotton
proportion to the blend cost value is proportional to the value assumed by
this proportion. Similarly, additivity implies that the total contributions to
the cost value are obtained by an algebraic sum of the individual component
contributions. For example, if the proportion of cotton A is doubled then it
is assumed that the cost contribution of this cotton to the cost function will
also be doubled.
Another assumption of linear programming is certainty, which implies
that parameters such as costs and average levels of quality are known with
certainty. In other words, LP assumes that accurate data on the objective
function and constraints of the models are available, i.e. they are not subject to
fluctuation and they are deterministic and constant.
Existing physical theories or previous experimental results may be useful
in verifying the assumption of linearity. In addition, experiments using small
blends to examine linearity are strongly recommended.
7.6.2 Types of linear programming
Linear programming methods can generally be divided into two main types,
graphic and analytical. First type may be used in simple problems involving
only two decision variables and few constraints. More importantly, the
method can be used to illustrate the concept of linear programming without
getting into its mathematical complexity. Second type is used for more
complex problems involving several decision or structural variables and
many constraints.
Analytical linear programming handles the problems where graphical

method fails, i.e. problem with more decision variables and more constraints.
The most commonly used computerised linear programing in this category
is simplex method. It is an iterative procedure that progressively approaches
and ultimately reaches an optimal solution to linear programing problems.
Computerised linear programing routines automatically arrange the inputs
of the problem (objective functions and constraints), perform the iterative
procedure, and produce an output of the LP program.
The general form of problem solved in the simplex method is:
Minimize C ’x 7.4
Subject to : Ax bli £ xi £ ui
7.5
Where, A is an (m × n) matrix of technological coefficients (fiber properties
or yarn properties), b is an (m × 1) matrix of right-hand side (RHS) constants
or minimum requirements of fiber or yarn characteristics, C’ is an (n × 1)
matrix of cost coefficients, x is an (n × 1) matrix of structural or decision
variables (cotton proportions), li is a lower bound on xi and ui is an upper
bound on xi.
However, in this format of analysis, individual fiber characteristic has
assigned quality constraint, in which a desirable minimum level of this quality
characteristic was specified. This requires a high degree of certainty of their
effects on quality and processing performance. Another important aspect
raises reliability of such program lies in the fact that high variability within
each cotton component reduces degree of certainty about the level of quality.
In addition, even with characteristics that are known to be linearly additive,
high variability can lead to a significant departure from linearity, the basic
assumption of linear programming. This problem can be well addressed by
replacing individual fiber characteristics with combined fiber characteristics
into one parameter, i.e. a fiber quality index or a yarn quality parameter,
involves mutual effects of fiber properties. Higher the correlation of this
index with fiber properties, better will be the performance earned by linear
programming.
7.6.3 Effect of inventory constraints
As in classical mix formulation approach, cottons required for a years production

is purchased at once, then quality demands and processing conditions are
more or less fixed during the year. In such case, linear programing can be
used as a part of decision-making process of cotton purchase. This situation
imposes no inventory constraints, since each cotton available in the inventory
will be proportional to the amount of this cotton used in the blend. In fact in
actual practice, cottons are purchased at different times of the year because of
prohibited storage space and need for taking advantage of favourable market
conditions of price and availability of desired cotton quantities. In this case,
the introduction of an inventory constraint becomes a necessity, and may be
changed as the amount of cotton of a particular type changes. Hence addition
of an inventory constraint leads to a slight increase in the cost of blending.
References
1. Bezdudnyi F.F., 1965, Planning the most economical blend in cotton

spinning,Technology, Textile Industry (USSR), 5, 37.
2. Charles H. and Chewning C., 2004, Cotton bale management EFS®
system software: engineering cotton products from fiber to fabric, 18,
19.
3. Chewning C., Zeplin J. and Vodica S., 1994, Proceedings of Beltwide
Cotton Conferences, San Diego, USA.
4. Mogahzy EI Y.E. and Gowayed Y., 1994, Theory and practice of cotton
fiber selection. Part 1: fiber selection techniques and bale picking
algorithm. Texttile Research Journal, 64, 3240.
5. Mogahzy EI Y.E., Broughton R. and Lynch W.K., 1990, A statistical
fiber properties. Textile Research Journal, 60(9), 495500.
6. Mogahzy EI Y.E., 1992, Optimising cotton blend costs with respect to
quality using HVI and linear programming. Part I: fundamentals and
advanced techniques of linear programming. Textile Research Journal,
62, 1 8.
7. Mogahzy EI Y.E., 1992, Optimising cotton blend costs with respect to
quality using HVI and linear programming. Part II: combined effects
of fiber properties and variability constraints. Textile Research Journal,
62, 108 114.
8. Mogahzy EI, Y.E., Reed M. and Lynch W.K.,. Implementation of on-
line fiber analysis in the textile industry, in Ergonomics of hybrid
automated systems 1, Ch. IX, In: Karwoski W., .Parsaei H.R., and
Wilhelm M.R. (eds.) Elsevier science publishers, NY, 1988, 695 698.
9. Kang B.C., Park S.W., Koo H.J., et al., 2000, Fibers and Polymers,
1, 55.Lewis H.L. 1989, Proceedings of Beltwide Cotton Conferences,
Nashville, USA.
660665.
8
Yarn engineering by back propagation algorithm
concept of ANN


*
Abstract
Yarn engineering has attained an accurate model predicting yarn and fibre
properties status with the development of ANN. The studies in literature have
shown clearly that the relationship prevailing between yarn parameters and
fiber properties is non-linear and complex. The usual statistical technique still
not capable of deciding the exact fiber properties required for a particular yarn
property as well as how these properties affect the yarn properties. The advent
of high-speed fiber-testing machines and development of powerful modelling
tools such as artificial neural network (ANN) have provided a great stimulus
in the yarn engineering research. The feasibility of yarn engineering can be
verified by developing a yarn-to-fiber or reverse engineering model, using
ANN. This concept is entirely different from the prevailing forward models,
as it predicts exactly in reverse manner, viz; the properties of final yarn,
by using the fiber properties as inputs. This model could handle numerous
dependent and independent variables. Characteristics like generalisation,
vigour against changes, etc., have made it a superior choice for modelling than
others. A number of attempts are made by various researchers to engineer yarn
quality by utilising back-propagation algorithm of artificial neural network.
Microprocessor base classical linear programing approach was made use of
in combination with ANN ensuring cost minimisation of optimised quality
cotton fiber mix.
Key Words: artificial neural network, cotton fiber, linear programming,
spinning consistency index, yarn engineering
8.1 Introduction
Yarn engineering, which implies the production of yarns having customer-
defined properties, has often attracted the attention of textile researchers.
However, the progress of yarn engineering research has been dissatisfied
by the absence of an adequate model predicting yarn properties from fiber
properties and vice versa. Copious mathematical models developed in early
90s by researchers like Bogdan (1967), Subramanian et al. (1974) Zurec et al.
(1987), Aggarwal (1989), Frydrych (1992), Morris et al. (1999), and statistical
models developed by Hafez (1978), Smith and Waters (1985) and Hunter
(1988) etc., to relate fiber and yarn properties. These models have achieved
only a limited success in practice due to limitations of simple mathematical
functions in expressing the complex relationship between fiber and yarn.
These primary methods (mechanical, mathematical and statistical) require
highly experienced personnel as well as numerous of test steps, therefore, they
could not give accurate models with reasonable computational costs.
The evolution of ANN has ushered in a new era in the domain of yarn
engineering research. Artificial neural network is not only a very persuasive
modelling tool but also can handle a large number of dependent and independent
variables simultaneously. Artificial neural network (ANN) which is inspired
from the evolution of biological neurons of brain is a powerful method for
modelling of complex phenomena. Some of its characteristics such as the
ability of learning and generalisation, robustness against disturbances, and
parallel information processing have made ANN superior to other modelling
approaches. Nowadays, ANNs are widely used for solving many engineering
problems as mentioned in earlier chapter.
Influenced by these in late 90s, ANN based mathematical models has been
used by Cheng and Adams (1995), Zhu and Ethridge (1996, 1997), Majumdar
and Majumdar (2004), Rajamanickam et al. (1997) and Ramesh et al. (1995)
to predict spun yarn properties, such as strength, elongation, evenness and
hairiness from the fiber properties and process parameters. Most of these
researchers have valued the excellent prediction performance of ANN models.
However, a “reverse engineering” model that can predict the fiber properties
and process parameters from the properties of final yarn is more enviable from
a yarn engineering point of view. Unfortunately, there is lack of efforts seen
in the direction of “reverse engineering” modelling of spinning process. The
plausible reason is the limitations associated with the traditional modelling
methodologies.
Using ANN, Guha (2002) and Chattopadhyay et al. (2003) have developed
“reverse” models relating yarn properties and spinning process parameters.
However, fiber parameters are the fundamental determinant of yarn properties
and therefore the selection of suitable raw material should be given the
first priority in yarn engineering endeavours. After selecting the proper raw
material, process parameters can always be tuned for the final adjustment of
yarn properties.
8.2 Reverse yarn engineering

Reverse engineering in context to cotton yarn production system, viz; ring
spinning, rotor spinning, etc., is the process of discovering the technological
principles of a spinning system through analysis of its structure, function and
operation.
8.2.1 Importance
The engineering design of ring spun yarn is a complex task and makes use of
empirical knowledge accumulated over time. Only few, highly experienced
experts in the industry know thumb rules/principles and methods of complex
structure synthesis under performance constraints. Based on the knowledge
and expertise of the experts in the field, if a scientific database is prepared
then ANN based reverse modelling can be prepared more precisely. It is a
system that has been engineered to simulate a human expert’s reasoning and
decision-making.
8.2.2 Basic steps of networking
A number of attempts are made by various researchers to engineer yarn quality

by utilising back-propagation algorithm of artificial neural network (Section
8.1). The steps followed in this networking are:
(1) Initialisation: Assuming that no prior information is available, the

weights and thresholds are picked from a uniform distribution whose
mean is zero and whose variance is chosen to make the standard
deviation of the induced local field of the neurons lie at the transition
between the linear and saturated parts of the sigmoid activation
function.
(2) Presentation of training examples: The network is presented with an
epoch of training examples.
(3) Forward computation: In the forward pass the synaptic weights
remain unaltered throughout the network, and the function signals of
the network are computed on a neuron-by-neuron basis.
(4) Backward computation: The backward pass starts at the output layer
by passing the error signals leftward through the network layer by layer
and recursively computing the local gradient.
(5) Iteration: Iteration is done in forward and backward computations by

presenting new epochs of training examples to the network until the
stopping criterion is met. Here iteration is the act of repeating a process
with the aim of approaching a desired goal, target or result. Each
repetition of the process is also called an “iteration”, and the results of
one iteration are used as the starting point for the next iteration.
8.3 Procedure for cotton yarn engineering

Scientific mode of yarn engineering does not differ considerably in terms
of procedure to be followed from decision based classical yarn engineering
system. During yarn spinning, textile experts commonly control a series of
parameters for process as per known fiber parameters, keeping in mind the
required output in terms of yarn quality parameters and cost. Parameters
usually focused for fiber are strength, fineness, length, maturity, etc., and for
process are draft, twist, spindle speed, delivery speed, etc. However, their
sequence of importance differs as per targeted yarn quality parameter/s, viz;
count, strength, evenness, hairiness, etc., and cost. Indeed the sequence of
importance for yarn parameters is purely affected by yarn type in terms of its
material and production pattern, as well as its end use application.
The studies in literature have shown clearly that the relationship prevailing
between yarn parameters and fiber properties is non-linear and complex. So,
proposing optimum values of fiber parameters to accurately meet desired yarn
values is not possible by usual statistical techniques (mechanical or statistical
mathematical models based on the fundamental mechanics of ring spinning) and
often fails to reach satisfactory results. This is concerned with two constraints:
(1) Although fiber parameters influencing the yarn parameters are well
known, there is no accurate mathematical relationship available between
these two for clear determination. For example, how fiber parameters
affect the final yarn tenacity is yet not perfectly defined by usual statistic
technique.
(2) There is no statistical way to determine optimal input fiber values to
reach desirable yarn value.
In decision based engineering, they were defined by brainstorming exercise.
In fact, they are determined only by “trial and error” and former experiences
of experts only. However, use of scientifically trained artificial neurons instead
of biological one can address well to this complex and non-linear problem on
imparting due training. It is also equally important for the manufacturer to
follow the prediction trend in a reverse fashion also. It means system should
have provision to identify input fiber parameters for meeting the targeted yarn
values. Thus in broader sense ANN based yarn reverse engineering requires:
(1) A group of yarns produced using cotton or other fibers with known
fiber and process parameters.
(2) The reverse artificial neural network should be trained by using yarn
parameters as the inputs and fiber properties and/or process parameters
as outputs.
(3) After the training, the testing data set is presented to the neural network
and fiber and/or process parameters needed to achieve desired yarn
properties are predicted.
(4) Yarns are spun using the predicted combinations of fiber and/or
process parameters. Association or closeness between the desired and
achieved yarn properties is compared to appraise the accuracy of yarn
engineering.
They are well elaborated in the forthcoming sections.
8.3.1 Defining aim of yarn engineering
It refers identification of output in terms of yarn type, its important

parameter/s and cost to be engineered for a particular domain of application.
Hence, desired output can be met only when feed and processing elements are
competitive enough. Thus, aim of the yarn engineering can be set up in either
way as per need:
(1) Prediction of yarn quality and cost to be produced out of known cotton
fibers values (quality and cost) and weights (proportions) mixed
together and spun with well defined system (ring/rotor/air jet) with
fixed process parameters.
(2) Prediction of cotton fiber values (quality and cost) and weights
(proportions) mixed together, for spinning desired yarn value (quality
and cost) as an output on known production pattern (ring/rotor/air jet)
with fixed process parameters.
In the above list, the former one relates with forward feed yarn engineering
(neural network), whereas the later one back propagation neural network or
yarn reverse engineering. But irrespective of the network chosen, clarity about
(1) The technical definitions of inputs, outputs, hidden layer attributes and
(2) Effectiveness of input parameters on output parameters;
are most important for accurate engineering modelling.

Example: 50s Ne ring spun yarn of 100% cotton is popularly used for shirting
fabric. Yarn tenacity is one of the most significant parameters to be controlled
during yarn spinning process. So, a yarn engineering approach for predicting
the yarn tenacity can be proposed in two ways:
(1) Feed Forward Artificial Neural Network (FF ANNs) or

(2) Feed Forward Back-propagation Artificial Neural Network (FFBP
ANNs).
Also as mentioned above it requires clarity about:
(1) Technical definition: In the present example yarn tenacity is the

output and technically defined as:
The resistance of yarn against tensile forces is called yarn tenacity or
It is the minimum force which is needed to rupture the yarn.
(2) Effective input parameters on the selected yarn parameters
(output):
Several factors are influencing yarn tenacity, but the most important
out of them are considered.
Fiber parameters: The properties of the fibers influencing yarn tenacity
are: Upper half means length, length uniformity, short fiber content,
fibers strength and elongation at break, maturity ratio, yellowness
and linear density. But the most effective are considered, say five out
of listed, viz; Upper half mean length, length uniformity, short fiber
content, fibers strength and elongation at break.
Process parameters: Process variables like; yarn count, twist, spindle
speed, traveller count (spinning tension), package hardness (winding
tension) have greater influence on yarn tenacity.
However, the selection of input variables is done with reference to the
particular production pattern and yarn size. Thereby production variables are
set fixed many a times.
8.3.2 Database creation
This step deals with the generation of empirical data for the development
of desired ANN-based system, which can be used for training and testing of
proposed network.
This demands to perform a series of experimental tests under identical but
optimal processing pattern by covering up wide variations in inputted values.
The yarn so produced at the end of each trial is verified for its specified
parameters. Care must be taken at the testing end also for the precision of
testing; thereby always standard and precise test methods are used after due
conditioning at standard atmospheric conditions for both inputted as well as
outputted materials (variables). The record for testing set up used should be
preserved for future reference. Values for influencing process variables used
during each trail should also be measured and preserved as back stuff record.
The main intent of such an experimental approach is to create an authentic

databank with known values of inputs with their corresponding outputs and
hidden layer variables of an ANN-based algorithm. It is then divided into two
groups, first one composed of major part; known as training dataset and the
second smaller group as testing dataset. Training dataset serves as trainer
for proposed ANN model, as neural networks in the first step of modelling,
finds and learns non-linear patterns from such training data. Thereby training
data must cover the whole input range and need to have a suitable dispersion.
Generally, if more training data is used then the accuracy in prediction will
often be found more thereby many a times it holds 95% of share. However,
division line varies as per need of engineering.
The entire dataset is then translated in terms of two matrices of “n × N”, one
for creating inputting training set (P) and second as outputting target set (T). In
case of training set, ‘n” is representing size (number) of inputted variables but
in case of target set it represents outputted variable that is in present case yarn
parameter size, whereas “N” represents the number of experimental trials.
Example: In continuation with the above mentioned example (Section 8.3.1),
large number, say hundred different families of mixing are formed out of
different kinds of fibers and processed on fixed ring spinning mill set up
used to produce desired 100% cotton 50s Ne yarn. Each mix represents set
of five different fiber property values used for producing same type of yarn.
Accordingly output yarn tenacity value obtained for each trial does not remain
same although produced on a fixed process set up. This happens as fiber
mix characteristics get changed with contributing fiber parameters. Thereby
databank furnished with total hundred input and output elements gets created.
Standard test procedures need to be adopted for measuring fiber properties
and yarn tenacity.
At the end of experiments, the databank composed of hundred pairs of
inputs and corresponding outputs generated. It is converted into two groups;
ninety five percent data (in present example only) as training group and rest
five percent as testing group. Now the entire data is translated in terms of
pair of two matrices of “n × N” format, one for training set P = [5 × 100]
and second as its corresponding target set T = [1 × 100]. In present example,
for training set, n = 5, as it is representing size or number of fiber properties
inputted and for target set n = 1, as only one yarn parameter outputted i.e.
tenacity is for all 100 trails, thereby N = 100.
8.3.3 Construction of desired artificial neural network
Artificial neural network consists of a large number of interconnected groups

of artificial neurons (units, cells, nodes or processing elements) as explained
in Chapter 5. Each neuron receives connections from other neurons and/or

itself, each with an associated weight.
The inter connectivity defines the topology (type of graph) of the ANN.
An artificial neuron is basically a mathematical function representing an
abstraction of biological neurons. The ANN is composed by three layers:
input, hidden and output layer. A brief mention is given here also, just for ease
of understanding of modelling process.
The basic structure of a multi input neuron is given in Fig. 8.1. It has
Vector X = [X1........Xn]T as input and the scalar y as the output of the neuron.
The influence of Xi on y is determined by w1,j . Another input for neuron is bias
parameter (b) with its corresponding weight is 1. The output of the neuron is
computed by:
Figure 8.1 Model of multi input neuron.
y = f ( Z ) = f (Wx + b ) 8.1
n
Z = ÂX iWi + b = WX + b 8.2
i =1
W = W1,1 ºº.. W1,n 8.3
The hidden layer is made of logistic neurons followed by an output layer

of linear neurons. The number of hidden neurons of feed-forward neural
networks, generally decided on the basis of experience, is an important factor
for the training, in order to avoid over fitting in the function approximation.
From one point of view the number of hidden units may be stated a priori by
means of empirical equations provided by the literature. On the other hand it
is possible to select the best network by estimation, for a given problem, of the
network architecture and parameters within a set of candidate configurations.
The activation function ƒ could be linear or either non-linear. Usually, the

designer selects a suitable activation function with respect to the problem
features. Some of the widely used activation functions are summarised in
Table 8.1.
Table 8.1 Popular activation functions
Function Definition Range

Linear f (Z ) = Z (-∞, + ∞)
Logistic 1 (0, +1)

f (Z ) =
1 + e− az
a is a slope parameter
Hyperbolic e z − e− z (-1, +1)
f (Z ) =
e z + e− z
Threshold 0, | z < 0 {0, +1}
f (Z ) = 
1, | z ≥ 0
In comparison with single layer networks, multilayer neural networks

have more capabilities. Double layers feed forward neural networks (with
sigmoid functions in first layer and that of linear in second layer) can
estimate any continuous function with arbitrary precision. It is well known
fact that neural networks with one hidden layer are suitable for majority
of applications, and the second hidden layer can improve the performance
of the network if there is a complex relationship between the input and
output parameters. However, the number of hidden neurons and the number
of hidden layers are usually adjusted by trial and error because these are
problem-dependent parameters.
There are different kinds of topologies (its graphs) and training algorithms,
they are: Back Propagation, Hopfield Networks, Competitive Networks and
networks using Spiky Neurons. But the feed forward neural network with back
propagation learning algorithm is more popular for addressing yarn engineering
problems precisely. In this structure, the neurons are located in layers and
from one layer to another one connected with each other with links to carry
the signals between them. There is a weight for each connection link which
acts as a multiplication factor to the transmitted signal. The weights represent
information being used by the neural network model to solve a problem. One of
the central issues in neural network design is to utilise systematic procedures (a
training algorithm) to modify the weights directly from the training data without
any assumptions about the data’s statistical distribution.
In accordance to all these basic constrains, the first step of networking is

“initialisation”. Here assuming that no prior information is available, the weights
and thresholds are picked from a uniform distribution whose mean is zero and
whose variance is chosen to make the standard deviation of the induced local
field of the neurons lie at the transition between the linear and saturated parts of
the sigmoid activation function. Simple way the network is initialised by setting
up all the weights to be small random numbers, say between –1 and +1.
Example: Under the heading of initialisation for the above mentioned
problem; a feed forward neural network structure (Fig. 8.2) with two layers is
proposed, where the first layer is known as the hidden layer.
Figure 8.2 A feed forward neural network.

This structure is presented as: n: nh: o where n is the number of inputs, nh
is the number of hidden layer neurons, and o is the number of output layer
neurons. The output of network is computed by:
Ê nh ˆ
y = f 0 Á ÂOiWi 2 + b˜ 8.4
Ë i =1 ¯
Ê n ˆ
Oi = Á ÂX jWi1, j + bi ˜ , 8.5
Ë j =1 ¯
Where, i = 1,…nh.
where, y is the final output, ƒ0 is the activation function of output neuron, Wi2
is the weight of link between i-th output neuron in the hidden layer and the
final output, b is the bias of output neuron, Oi is the output of i-th neuron in
hidden layer, W1i, j is the weight of link between j-th input and i-th neuron in
hidden layer, bi is the bias of i-th neuron in hidden layer, and f is the activation
function in i-th neuron in hidden layer.
The weights (W1i, j, Wi2, bi, b) are adjustable parameters that need to be
tuned by training. In fact, the objective of network training is adjusting
these parameters in such a way that the network generates desired output for
different inputs.
8.3.4 Modelling
In order to model the yarn spinning process, it is necessary to work out a

proper neural network able to predict reliably the value of yarn and respective
inputted values. Artificial neural network (ANN), especially back propagation
method has been widely applied to solve problems in many areas of textile
manufacturing, e.g. for the purposes of forecasting, diagnostics, and pattern
recognition. An important part at ANN in determining the performance of the
network is training algorithm used. Because there are 12 training algorithms
that can be used at back propagation method, of course, it’s needed to be
selected the most optimal algorithm for the given problem in order to obtain
the best results. Training algorithm performance is said optimal in providing
solutions can be seen from the error generated. The smaller the error generated
the more optimal performance of the algorithm said to be. This becomes
possible only if the ANN is properly trained by pairs of the training and target
sets. Both structure and training of the ANN have been developed by using
the Artificial Neural Network Toolbox working in respective mathematical
modelling environment, e.g. Matlab.
Thus after initialisation that is choosing the weights of the network
randomly, the back propagation algorithm is used to compute necessary
corrections. This step is often regarded as network training. The training of a
neural network by back propagation involves following stages (Fig. 8.3):
(1) Feed forward computation of the input training pattern,

(2) The calculation of forward computation error
(3) Adjustment of input weight
(4) Backward pass: Back propagation of the associated error, and
(5) Adjustment of the weights for hidden layer.
The algorithm is stopped when the value of error function has become
sufficiently small.
Figure. 8.3 Steps of artificial neural networking back propagation algorithm.

(a) Initialisation. (b) Forward pass. (c) Backward pass. (d)
Backward pass for hidden layers.
8.3.4.1 The feed forward of the input training pattern
It is also known as forward computation or forward pass and deals with the
presentation of an epoch of training examples to the network. Since neural
networks find and learn patterns in the training data, in the first step of
modelling, training data set (P) is needed, which is generated empirically
by conducting experimental trial in the same production environment as
explained in Section 8.3.2. In other words, the feed forward trained ANN
is able to receive any vector of “n” elements composed by the inputted
parameters and to give, as output, the prediction of the yarn value of the
yarn produced with that feed material. Definitely the calculated (predicted)
output value of the network is completely different than expected on (target),
since all the weights are randomly selected at the point of initialisation.
Thus in the forward pass the synaptic weights remain unaltered throughout
the network, and the function signals of the network are computed on a
neuron-by-neuron basis.
8.3.4.2 The calculation of forward computation error
Hence all the weights are randomly selected at the point of initialisation, the
error of computation of each neuron is essential in second step to rectify it.
Thereby outputs of training data set are compared with respective target data
set to define success of training.
The mean square error (MSE) correspondent to deviation between the
actual elements of target set and predicted output of training set elements
(equation 8.6).
Â j =0Â i =0 (tij - � yij )

P N
MSE = 8.6
NP
Where, ti j and yi j are the target output and predicted output, respectively,
for ith training pattern at jth output neuron. P is the total number of output
neurons and N indicates the number of training patterns. Simply error is the
mathematical difference between targeted value and actual value of neuron
and for the entire network it is the sum of all such neurons involved in
prediction.
8.3.4.3 Adjustment of input weights
The calculated error vector done according to the calculated mean square
error (MSE) is then used to adjust the weights mathematically in such a
way that the error will get smaller. In other words output of each neuron
will get closer to its target; this process is known as reverse pass or
backward pass.
In the backward pass, the error signal is propagated backwards to the
neural network and the synaptic weights are adjusted in such a manner
that the error signal decreases with each iteration process. Thus, the neural
network model approaches closer and closer for producing the desired output
by minimising the error for each output neuron and the network as a whole.
The corrections necessary in the synaptic weights are carried out by a delta
rule, which is expressed by the following equation:
∂ ( MSE )
W ji ( n ) = -h 8.7
∂W ji ( n )
Where, Wji(n) is the weight connecting the neurons j and i at the nth
iteration; ΔWji(n) is the correction applied to Wji(n) at the nth iteration; η is a
constant known as learning rate.
8.3.4.4 Calculate the errors for the hidden layer neurons
Unlike the output layers for the hidden layers error cannot be calculated
directly in the absentia of target value. Thereby back propagation is started
from the output layer. This is done by taking the errors from the output neurons
and running them back through the weights to get the hidden layer errors. This
step is known as backward pass or reverse pass.
Thus, the backward pass starts at the output layer by passing the error
signals leftward through the network layer by layer and recursively computing
the local gradient. In terms of yarn production system network, the hidden
layers of artificial neural network are trained by using reverse pass that is by
using yarn parameters as the inputs and fiber and/or process parameters as
outputs.
8.3.4.5 Adjustment of hidden layer weights
Once the error for the hidden layer neurons is obtained, similar to forward
pass, the hidden layer weights are changed. Iteration is done in backward
computations, similar to forward computations by presenting new epochs of
training examples to the network until the stopping criterion is met.
8.3.4.6 Major problem faced during training
The major problem faced in training neural model is over fitting. The
possibility of over fitting exists because the criterion used for training the
model is not the same as the criterion used to judge the efficacy of a model.
Particularly, efficacy is not determined on the basis of its performance on the
training data, but by its ability to perform well on unseen data. A model is said
to be trained, by maximising its performance on some set of training data.
The problem begins, when a model begins to “memorise” training data rather
than “learning” to generalise from trend. Such a trained model will usually
execute poor predictive performance as it can exaggerate minor fluctuations in
the data. Either way, the value of the coefficient of determination will shrink
relative to the original training data.
The potential for over fitting depends on;
(1) The number of parameters and data (observations): A model which is

excessively complex or a model is having too many parameters relative
to the number of observations then over fitting occurs. For example,
if the number of parameters is the same as or greater than the number
of observations, a simple model or learning process can perfectly
predict the training data simply by memorising the training data in its
entirety. But such a model will typically fail drastically when making
predictions about new or unseen data, since the simple model has not
learned to generalise at all.
(2) The conformability of the model structure with the data shape, and
(3) The magnitude of model error compared to the expected level of noise
or error in the data.
Even when the fitted model does not have an excessive number of
parameters, it is to be expected that the fitted relationship will appear to
perform less well on a new data set than on the data set used for fitting.
8.3.4.7 Solution to training problem
Over fitting can be avoided by using additional techniques which can indicate
when further training is not resulting in better generalisation. The basis of
selection of such techniques is either;
(1) To plainly penalise overly complex models, or

(2) To test the model’s ability to generalise by evaluating its performance
on a set of data not used for training, this is assumed to approximate
the typical unseen data that a model will encounter.
The generally used techniques are cross validation, regularisation, early
stopping, pruning, Bayesian priors, on parameters or model comparison, etc.
But out of them cross validation is usually employed for yarn engineering
purpose.
Cross validation is often known as rotation estimation and it is a model
validation technique. As mentioned above, it is used for assessing how the
results of a statistical analysis will generalise to an independent data set.
It is normally used in settings where the goal is prediction, and one wants
to estimate how accurately a predictive model will perform in practice.
Such a model is usually given a training dataset of known data on which
training is run, and also testing dataset of unknown data (or first seen data)
against which the model is tested. The goal of cross validation is to define
a dataset to “test” the model in the training phase only with the validation
dataset, in order to limit problems like over-fitting. This practice gives an
insight on how the model will generalise to an independent dataset, or say
an unknown dataset picked up for instance from a real problem. Cross
validation is important in guarding against testing hypotheses suggested by
the data, often called “Type III errors”, especially where further samples are
hazardous, costly or impossible to collect.
There are basically two methods of validation: conventional validation and

multiple round cross validations.
(1) Conventional validation: It deals with partitioning the data set into
two sets of 70% for training and 30% for test and estimating the error
in terms of Root Mean Square Error on the training set. Hence, there
is not enough data available or there is not a good distribution and
spread of data to partition it into separate training and test sets in the
conventional validation method it cannot serve as a useful estimator
of model performance and thus the error on the test data set does not
properly represent the assessment of model performance.
(2) Cross validation: It is a powerful general technique used to
properly estimate model prediction performance. Here in order to
reduce variability, multiple rounds of cross validation are performed
using different partitions after each training epoch. Each time after
the network has been trained; a second set of patterns (data) which
are noisy versions of the training set, but aren’t used for training
themselves (called the validation set) is used to calculate an error.
Heuristically, it might be expected that the noise will “smear out” each
data point and make it difficult for the network to fit individual data
points precisely. Indeed, it has been demonstrated experimentally that
training with noise can lead to improvements in network generalisation.
Thus deliberate addition of some noisy versions of the patterns into
the training set to train the network (say one in five); can improve the
network’s performance in this respect. The training may also benefit
from applying the patterns in a random order to the network. Further
these calculated error of the validation results are averaged over the
rounds. Thus it combines (averages) measures of fit (prediction error)
to correct for the optimistic nature of training error and derive a more
accurate estimate of model prediction performance. This is one of the
main reasons for proving superiority of using cross validation instead
of classical technique.
When the network is fully trained, the validation set error reaches a
minimum (Fig. 8.4). Such trained network should be able to recognise not
just the perfect patterns, but also corrupted or noisy versions. But when the
network is overtraining (becoming too accurate) the validation set error starts
rising, the training will be stopped, and as the over trained network won’t be
able to handle noisy data so well.
Figure 8.4 Training error.
Summarising the entire discussion

The validation data set is used to minimise over fitting. The weights of the
network are not adjusted with this data set, but just verifying that any increase
in accuracy over the training data set actually yields an increase in accuracy
over a data set that has not been shown to the network before, or at least the
network hasn’t trained on it (i.e. validation data set). If the accuracy over the
training data set increases, but the accuracy over then validation data set stay
the same or decreases, then it indicates over fitting of neural network and
training should be stopped.
8.3.4.8 Generalising a neural network
Neural networks may produce reasonable answers for input patterns which
are not seen during training. This is done by memorising patterns in the data
during training; such behaviour of network is referred as generalisation.
Generalisation is particularly useful for classification of noisy data, the “what-
if ” analysis and prediction (e.g. time-series forecast).
There are also issues regarding generalising a neural network. Issues to
consider are problems associated with under-training and over-training data.
Under-training can occur when the neural network is not complex enough to
detect a pattern in a complicated data set. This is usually the result of networks
with so few hidden nodes that it cannot accurately represent the solution,
therefore under fitting the data.
On the other hand, over-training can result in a network that is too complex,
resulting in predictions that are far beyond the range of the training data.
Networks with too many hidden nodes will tend to over fit the solution. The
aim is to create a neural network with the right number of hidden nodes that
will lead to a good solution to the problem (Figures 8.5 (a-b)).
Figure 8.5 Generalising a neural network. (a) Under fitting data. (b) Over
fitting data. (c) Good fit of data.
8.3.5 Testing of neural network
The network behaviour is analogous to a human that is shown a set of data and
is asked to classify them into predefined classes. Like a human, it will come up
with theories about how the samples fit into the classes. These are then tested
against the correct outputs to see how accurate the guesses of the network
are. Radical changes in the latest theory are indicated by large changes in the
weights, and small changes may be seen as minor adjustments to the theory.
Once trained for forward pass as well as backward pass, the network is
ready to predict network elements on either side. Now the test data is needed
in the final step to assess performance of designed and trained reverse neural
model. Test data is the part of the databank (step 2), normally kept quite
small as compared to training data and it is not used for the training purpose.
Allocation of larger data share to training set from databank is mainly
attributed to acquire higher accuracy of prediction as mentioned before.
Similar to training set, test data is also split into two matrices one for feed
inputs and another for targeted output of test. Predicted yarn output values
are obtained by using trained feed forward network for input test data feed
elements (fiber data).
Validation set is different from test set. Validation set actually can be
regarded as a part of training set, because it is used to build ANN model,
neural networks or others. It is usually used for parameter selection and to
avoid over fitting. The non-linear model of ANN, if trained on a training set
only, it is very likely to get 100% accuracy and over fit, thus get very poor
performance on test set. This can be remedied by validation set as explained
in earlier section only. Thus validation set is used for tuning the parameters of
a model against that test set is used for performance evaluation.
In order to validate the tested reverse engineering approach, a new series
of experimental test (called “validation set”) has been carried out. It involves
actual manufacturing of engineered yarn to validate; whether the predicted
cotton mix combination of fiber properties is really capable of producing the
target yarn properties. Thereby this phase consists of collecting the predicted
parameters of new inputs (fiber mix values) from their respective fed outputs
(yarn value) to neural network after successful completion of testing. Actual
yarns are spun by using the suggested combinations of fiber on fixed process
parameters used for ANN prediction. The quality parameters of output yarns
are verified in the similar manner to that of done at the point of generation of
databank. These inputted and outputted parameters pairs are then used as a
validation set for the trained and tested ANN.
8.3.6 Computing prediction error
The predicted value of the yarn (YP) must be compared with the real value
(YR) in order to assess the reliability of the prediction in forward feed network.
The comparison may be evaluated, in percentage, by defining a coefficient h,
called “prediction error”, given by following expression:
YR � -YP
h= 8.8
YR
Smaller is the value of h and better is the prediction. Similarly the predicted
values of the fiber (FP) must be compared with the real values (FR) in order
to assess the reliability of the prediction in back propagation forward feed
network.
FR � - FP
h= 8.9
FR
Well trained network is able to correlate the testing set elements to the
target ones by giving a response closer to the real value of the really produced
yarn value. It means extent of association or closeness of yarn properties is
found on comparison between the desired and achieved from the predicted
inputted elements. Hence it is a reverse engineering network; testing and also
prediction error computation can be carried out in either way. Prediction error
serves as the best tool to appraise the accuracy of proposed yarn engineering
structure.
Regression analysis is performed to measure the correlation between
outputs and targets. If the training is perfect, the network outputs and the
targets would be exactly equal, but the relationship is rarely perfect in practice.
If in regression plot, the validation and test curves are very similar it indicates
that learning with back-propagation algorithm does not indicate any major
problems with the training. If the test curve had increased significantly before
the validation curve increased, then it is possible that some over fitting might
have occurred.
8.3.7 Impact analysis
Optimised quality and cost values for cotton mix obtained by the virtue of
linear programming are inputted to validate ANN models and respective
outputted attributes are obtained. Such suggested outputted yarn values by
ANN are then compared with targeted yarn values. These can define the
success rate of the ANN architecture used for the study.
References
1. Dashti M., Derhami V., and Ekhtiyari E., 2014, Yarn tenacity modeling
using artificial neural networks and development of a decision support
system based on genetic algorithms. Journal of AI and Data Mining,
2(1), 73–78.
2. Majumdar A., Majumdar P.K. and Sarkar B., 2004, Selecting cotton
bales by spinning consistency index and micronaire using artificial
neural networks. AUTEX Research Journal, 4(1), 1–8.
3. Majumdar P.K. and Majumdar, A., 2004, Predicting the breaking

652–655.
4. Pynckels F., Sette S., Van Langenhove L., Kiekens P. and Impe K.,
1995, Use of neural nets for detemining the spinnability of fibres.
Journal of the Textile Institute, 86(395), 425–437.
5. Rocco Furferi and Maurizio Gelli, 2010, Yarn strength prediction: A
practical model based on artificial neural networks, Hindawi Publishing
Corporation. Advances in Mechanical Engineering, Volume 2010,
Article ID 640103, 11 pages, doi:10.1155/2010/640103.
6. Sette S., Bouliart L., Van Langenhove L. and Kiekens P., 1997,
Optimizing the fiber to yarn production process with a combined
neural network/genetic algorithm approach. Textile Research Journal,
67(2), 84–92.
7. Simon Haykin, Kalman Filtering and Neural Networks Copyright
© 2001 John Wiley & Sons, Inc. ISBNs: 0-471-36998-5 (Hardback);
0-471-22154-6 (Electronic).
8. Tino P.C., Implementation of Feed Forward Backpropagation Network,
School of Computer Science, University of Birmingham.http://www.
cs.bham.ac.uk/~jxb/NN/nn.html, dated: 30/06/15.
9. www5.in.tum.de/JASS/JASS06/presentations/Minin_handout.doc,
dated: 22/4/15.
10. www.eis.mdx.ac.uk/staffpages/rvb/teaching/BIS3226/hand11.pdf,
dated: 15/4/15.
11. Backpropagation – Wikipedia, the free encyclopedia, https://en.wikipedia.
org/wiki/Backpropagation, dated: 12/06/15.
12. https://www4.rgu.ac.uk/files/chapter3%20-%20bp.pdf; dated: 20/03/15.
13. A Step by Step Back propagation Example; mattmazur.com/2015/03/17/
a-step-by-step-back propagation-example.
9
Optimisation of yarn quality, cost
and process parameters


*
Abstract
Fiber to yarn conversion process for staple natural fiber like cotton is complex
in nature. This is due to the interrelated multivariate relationship, which is non-
linear in nature present between highly variable characteristics natural fiber
and yarn properties. Thereby task of producing yarn with precise quality and
cost, becomes utmost difficult for the industry mainly rely on human expertise
based decision-making. Exercise for optimisation of yarn quality, cost and
process parameters starts with precise selection of raw material, a maximum
shareholder on yarn quality and cost and then remain focused on desired state
of process control to yield maximum beneficiary fiber properties into output
yarn till the end. Thus spinning technologist, a decision maker has to take care
of two major tasks for successful yarn engineering, viz; selection of suitable
quality raw materials at minimum cost and optimisation of process parameters.
Introduction of automated as well as precise testing and production techniques
and development of software for linear programing and EFS together have
realized a long-cherished dream of spinning technologists on real ground.
Key Words: cotton cost, cotton quality, commodity, process control, EFS,
linear programing
9.1 Introduction
Yarn engineering is a process of yarn quality planning and manufacturing of
yarn based on the fiber properties for fixed process set up. It is quite important
to know-how fiber material respond during spinning process. In the absence of
adequate information in this area will lead the cotton spinning mills liable to
buy unnecessarily too good and thereby too expensive cotton which adversely
impacts yarn economy. Thereby a detailed knowledge about raw material
to be processed and its reaction during the process is mandatory to achieve
optimum settings of production pattern and fully explore the quality potential
of raw material.
The raw material can be best utilised if an ANN based yarn engineering
inclusive of bale selection (as well as bale management systems and linear
programming) are made available. These accurate tools for precise yarn
quality parameter prediction for known raw material physical and mechanical
characteristics are needed not only to assure the effective use of cotton fibers
but also for optimising the fiber mixing. Hence it prevents the user from
undue utilisation of costly cotton components in the mix (which reduces
considerably raw material cost, the major yarn cost contributor). Apart from
that appropriate and consistent yarn quality also avails greater share of profit
and therefore provides the ground for further minimisation of the raw material
costs. If such artificial neural networking based engineering is extended to
next step of textile manufacturing that is fabric and/or garment production
pattern then it can be proved useful for their optimisation also.
In the spinning mill (either on-line, off-line or their combination) quality
assurance and data collection/processing systems are used, purely dependent
on type of technology employed; conventional or modern. This ensures the
earlier detection of process faults and thereby reduces amount of downgrade
material produced. Adaptive networking system of yarn engineering if
provided with modern technology with high degree of automation can serve
as best platform for fine and continuous tuning of the fiber properties in
the bale mix to the requirements of the end product. With such an extent of
fiber mix optimisation, lagging in yarn quality produced is attributed to the
machine failure only. Effective fiber testing systems, together with application
of modern computer hardware and software, enabled new approaches of yarn
engineering to specify clearly point of trouble during production with respect
to time and location. Thus, profound corrective measures taken will facilitates
in achieving process optimisation also.
9.2 Components for optimisation

Yarn is not always a final textile product, it needs to be converted to either
fabric form or modified by doubling (sewing thread) or finishing (embroidery
thread) to suit end use applications (Fig. 1.5). But irrespective of the form it
resumes in final product, its role can be justified only if it is fault free. On the
other hand, the good yarn properties can only be ascertained by;
(1) Characterisation of the raw material (fibers) in terms of cotton mix

technological value and then selecting mix constituents compatibly,
(2) Processing such cotton mix on optimised process set up by subjecting
to continuous monitoring of quality and process.
Thus, the maximum profits theoretically achievable can never be fully
realized if optimum conditions of the raw material and the spinning process
are not made available. For example, if a too good fiber quality is used then
yarn quality will be better but at the same time yarn cost will also be too high.
Conversely if a process set up is in poor condition then although correct fiber
quality is used, the product yarn will not be able to satisfy cost as well as
quality demands.
According to the analysis of experts, spinning mill performance is primarily
a reflection of a ratio observed for technological values between input fibers to
that of output yarn of the system. The term “technological value” represents
a combination of price and quality. In this regard, the best raw material value
situation arises when quality is at its highest possible level and price is at its
lowest possible level. On the other hand, the best yarn value is achieved when
quality is at its highest possible level and price is correspondingly high from
a spinner viewpoint. This is the only way to accomplish profit in the spinning
business. Yet, it is the most difficult way also to achieve profit particularly in
today’s competitive market.
Summarising the discussion, the best fruits in terms of output yarn
technological value can be rejoiced on optimisation of
(1) Technological value of cotton, and

(2) Production process is obligatory.
9.3 Technological value of cotton mix

Technological value of cotton has two attributes as mentioned above: quality
and cost.
9.3.1 Cotton fiber quality
The quality of final yarn is largely influenced (upto 80%) by the characteristics
of raw cotton. Moreover, cotton fiber is a cultured natural product; it will
always refer to a non-homogeneous raw material because the cotton fiber
properties are much more influenced by growing area, origin, harvesting
methods, year of growth and many other factors. Under such circumstances
controlled changes of fiber properties over a given length of time are very
difficult and demands formulation of cotton mix by incorporating as many
components as possible. This can help in maintaining consistency in terms

of both cotton quality and cost. Even cotton is having conflicting standards
in terms of different yarn quality criteria as the level to which various fiber
properties influence yarn quality is diverse, and also changes depending on
the yarn manufacturing technology (Table 2.1). Thus, it becomes necessary
to specify the potential target application of fibers before specifying about its
quality potential. The selection of mix components in terms of fiber quality
is thereby concerned with type of process, yarns characteristics desired and
mixing potency. Necessitate knowledge of each available cotton quality value.
The major problem arises with utilisation of various measures of cotton
fiber properties in quality characterisation is the multivariate character of
information, holding various measuring units and lack of transformation to
the utility scale. The solution offered in terms of overall quality index (FQI/
PDI/SCI etc.) of cotton fiber used in defining technological value of cotton
by incorporating all the important fiber parameters. The weights of the
fiber parameters should be commensurate with their importance and actual
(empirically measured) contribution on the final yarn quality.
Availability of the fully automated and revolutionised fiber testing
concept, such as the high volume instrument (HVI) and the Advanced Fiber
Information System (AFIS) has allowed speedy and precise computation of
cotton quality. Also based on the HVI results, composite indexes such as the
fiber quality index (FQI) and spinning consistency index (SCI) can be speedily
and precisely calculated by interfaced microprocessor. This can create a vast
data bank for cotton varieties produced all over the world and make this
information available at each junction in the network with the provision of
electronic data interchange system (EDI). Thus, HVI emerges out as one of
the key players in defining cotton quality value.
9.3.2 Cotton cost
Cotton prices are fluctuating tremendously thereby the cotton market feels
handicap in forecasting about cotton price from year to year and sometimes
from period to period within a year. Apart from these the market on getting
globalise, forecasting about cotton price becomes a further complex task due to
the many unknown controlling factors that are often politically oriented. In that,
a few of the major cotton producing nations can impact the price, particularly
markets that vary from importing to exporting. Ignoring political factors, still
there are number of factors that influence the global price of cotton as a raw
material but the biggest primary reason is the complex interaction existing
between supply (actual and potential) and demand (domestic consumption,
exports and changes in the annual and seasonal demand for cotton) followed
by changing trends in commodity price. Since cotton cost is a major regulator

for yarn cost, due to its highest (50–70%) share towards the total yarn cost; it
become indeed necessary to explore these elements bit in detail.
9.3.2.1 Supply factors
Supply factors basically associated with yield of cotton for particular time, it
can be related with month, season or year. This is important for cotton cost
point of view as it determines resultant impact of following factors:
(1) Total cost of lint production (land rent, planting seed, fertilizer,
harvesting and ginning).
(2) Rate of technological development at agricultural, ginning and picking
fronts.
(3) Estimated environmental changes on cotton quality and accordingly
defining cost against cotton quality produced.
(4) The price of cotton received by producers in relation with prices for
alternative crops.
The price received by the producer is the key factor in the supply aspect.
This price is subject to many factors including: global competition, exchange
rates, and economical gains resulting from developing technologies. In some
countries, these important factors are overshadowed by governments subsiding
producers, and distorting markets resulting from government policies.
9.3.2.2 Demand factors
This criterion is mainly affected by ease of availability of cotton variety as

per customers need in terms of both cost and quality. Competency of cotton
to address changing fashion demands in comparison to alternative fibers is
also equally important in today’s context. Hence end use application range
for cotton fiber is vast starting from costliest clothing to say cheapest wipes.
Traditional demand factors were least affected by fashion element and mainly
centred around customer’s buying capacity (income, price) and suitable options
(e.g. products of synthetic or other natural fibers) available in the market.
Consumer demand is often a reaction to successful promotional efforts and
mainly shuffled towards fashion more in today’s competitive market. This is
also a hard core fact that today’s consumers have immense choices of fibers and
end products to choose from along with increased capacity to buy. This places
new challenges on most suppliers from the fiber producers to the spinners
and weavers. More specifically, the spinner today must sell a yarn that is fully
described in terms of end product performance, quality as well as cost.
9.3.2.3 Effect of change in commodity price
Regarding cotton cost, it is commonly understood that a large crop of cotton

brings a low price and a small crop a high price. But the mathematical relation
between the size of the crop and the resulting price is not usually linear. For
example, what change in the average annual price would be likely to result if
the size of the crop were increased from 12,000,000 to 18,000,000 bales? To
answer this question, it is necessary first to take account of another important
factor which has been found to have an important influence on changes in price
from year to year i.e. the general level of commodity prices. Here a change in
the general commodity price level means a change in the value of money with
which cotton is purchased. That’s why when prices of other things go up there
is a tendency for cotton prices to go up with them; when other prices go down
there is a tendency for the price of cotton also to go down. If the influence of
this factor, the general level of commodity prices is removed from the price
of cotton, a fairly definite relationship can be established between the price of
cotton thus adjusted and the size of the supply.
Summarising the entire discussion, it is quite apparent that cotton
consumption at producer’s end plays a determinant role in defining cotton
cost. Where, cotton consumption at producer’s end depends on;
(1) Production rate,

(2) Varieties of cotton used and their proportion utilised in mix formation.
However, production rate remain constant for the given type of yarn produced
by using fixed technology set up. So, the major influencing factor for cotton cost
is type of cotton selected in mix and their quantities (proportions) used.
9.4 Optimisation of process parameters

According to a survey, the production costs can be reduced by 1–5% only
by increasing efficiency, whereas by using automating spinning machinery
another 5% can be reduced. Under such circumstances raw material selection
should also be concerned about reduction in the process disturbances such
as end breaks, act as an obstacle in attaining higher working efficiency at
optimum processing speeds although latest technology is implemented.
The survival and prosperity of any production house depends largely on
two major economical components: cost and revenue; this fact should not be
over looked by any industry. The overall financial status of a spinning company
depends largely on the profit that a company earns from selling the yarn. No
doubt the extent of profit is equal to the positive difference created between the
market yarn selling value and the cost encountered in manufacturing the yarn.
The key variables affecting yarn price and its manufacturing cost are executed in
Figure 9.1. However, the selling value gets influenced by yarn quality, cost and
image in the market; where image deals with type of yarn and its success rate
in meeting end product requirement. Conversely, quality and cost of yarn are
safeguarded by three basic elements: raw material, labour and technology. All the
three fronts have undergone drastic changes in today’s spinning mill operating
practices. As a result optimisation potential (not only in the field of raw material
costs but also in conversion cost) is estimated to be even higher today.
Figure 9.1 Variables affecting yarn cost.

While switching away from the conventional to modern production
technology, success of a spinning plant in earning higher profit margin likely
to be dependent on changing trends of yarn manufacturing. A few of them are
mentioned below:
(1) The speed of machinery is increased, which has resulted in increased
stress placed on the fiber during high speed production process.
Such stresses are well controlled by the improvement in design and
metallurgy of various machine parts, e.g. use of double apron at ring
frame drafting system.
(2) Size and number of machines involved in production course have
reduced, e.g. small size modern blow room (two beaters) has replaced
large capacity classical blow room (six beating points). Further the
changed characteristics of trash and cotton has made it difficult to
clean cotton, and thereby controlling waste level (no good fiber loss
in waste) to meet stringent quality and cost demands become further
tough. These are well addressed by blow room machines working with
gentle but efficient opening and cleaning action with precise control of
system like Varioset.
(3) As the raw material represents a larger amount of yarn price, the
demands for providing more options for the varieties and respective
quality of cotton are increasing. The cotton purchase market is widen
from particular station to the entire world.
(4) Automation has reduced drastically labour compliment and associated
errors but at the same time regular training of skilled labour for updating
about use of advanced computer-aided technologies is needed.
(5) To optimise the spinning process both off-line and on-line quality
controls of intermittent products of the spinning mill are used. The
collected data provide a direct selection of machine settings best
suited to the fiber characteristics, optimum machine settings by
means of comparison trials, optimum selection of machine elements
and optimum machine maintenance conditions. Thus made it free of
human bias decisions as well as restricted expertise.
(6) Addition to this on-line control for various process parameters
regulates various tensions prevailing during spinning. This helps in
minimising variations in outgoing yarn properties.
Thus, modern technology can able to attain higher productivity and
flexibility on process optimisation, as it has ensured the full utilisation of the
positive fiber characteristics at every processing stage from the bale opening
up to the yarn.
9.5 Optimisation of yarn technological value

Requirements for technological value of yarn (yarn quality and cost) are
usually fixed as per specifications put forward by targeted customer but the
spinner’s concern is dedicated towards actual realization of these values. The
major constraint faced in this course is to define cotton types to be considered
in meeting desired yarn technological value. Continuous efforts have been
put forward for scientific way of cotton selection; a major shareholder in yarn
quality and cost is standardised by cotton classing. But it is also a fact that
such selected cotton must be truly utilised in meeting output yarn’s quality and
quantity demands to survive in today’s global competitive market.
The second attribute for controlling yarn cost is its manufacturing cost
(Fig. 9.1). It involves basically two types of costs: fixed costs and variable
costs. The fixed costs are function of a number of factors that are unique for
each operation and appended below:
(1) The size and the capacity of the spinning mill: It defines two major
cost factors, viz; bale cotton consumption and number of spindles
(spinning machines in line).
(2) The standard of living or the economic status of the country in
which the spinning mill is operating. As second highest yarn cost
contributor is labour cost, which is quite high for industrialised and
densely populated developing countries. On the other hand it is quite
low for developed countries operating with modern technologies, but
simultaneously it is also true that they have to pay higher for capital
investment. However, these factors should be accounted together for
wage differentials with respect to production rate earned in labour cost
analysis.
(3) Taxes, utility, lease cost, transport, and regulation fees.
The variable cost is the cost that changes in accordance to production level
or other way round production technology.
On a long run, the fixed cost can be reduced through a substantial reduction
in manpower, redistribution of personnel assignments are provided in such
a way that it can provide maximum service to the spinning mill, and better
logistics and production control. In addition, utilisation of appropriate cotton
mixes can result in a long-term improvement in machinery performance.
Hence variable cost is mainly affected by production rate; the efforts should
be made at continually reducing manufacturing cost. This can be achieved by
addressing the following critical tasks:
(1) Maximising fiber-to-yarn yield efficiency

(2) Maximising fiber-to-yarn conversion efficiency
(3) Diagnosing quality problems
(4) Improving yarn productivity and efficiency
The EFS® system can assist in reducing both sources of cost on scientific
ground. The key factors taken care off are:
(i) End product added value: It primarily aims at selecting suitable

raw material for a given predetermined specifications of end product
performance.
(ii) Fiber quality utilisation efficiency: It basically deals with maximum
utilisation of bale cotton quality to yarn quality at an economical way.
(3) Fiber quantity utilisation efficiency: It basically deals with

maximising the quantum of production from the supplied bale cotton
by minimising (controlling) cotton waste. Other way round deals with
optimisation of yarn realization for a given yarn type.
However, the real bottle neck for cotton yarn is that the input contributes
to the output in a very imbalanced and ostensibly unfair way. Under such
circumstances the spinner is always struggling in addressing above mentioned
three factors satisfactorily.
9.5.1 End product added value
Artificial neural network based various back propagation algorithm have been
developed and verified for their significance in scientific selection of desirable
cotton constituents as per yarn quality requirements. The collective use of
HVI, EDI and artificial neural network based back propagation algorithm can
serve as a pioneer tool in accurate selection of suitable cotton components in
mix or engineered fiber selection (EFS®).
9.5.1.1 Trait for EFS®-MillNet
The basic implementation set up used for engineered fiber selection (EFS®) is
schematically illustrated in Figure 9.2. The set up is mainly consisting of bale
management, warehouse management and HVI measured EDI data handling.
(1) Bale management: This attribute is well defined in earlier chapter.

It is based on the categorising of cotton bales according to their fiber
quality properties. The measurement of the fiber characteristics with
reference to each individual bale, the separation of bales into classes
and the determination of the mixture proportions and the laying down
of balanced bale mixes based on classes are included tasks under this
heading. Thus under this heading two main tasks are furnished and
they are:
(i) Consignment management: It basically defines path for the
movement of bales from the field to the textile mill on clearing
the essential channels in the market. Typically, this plan may vary
from one company to another depending on the domestic market
at which the company is located and the different market channels
that the company deals with.
(ii) Warehouse management: It deals with making a distinction
amongst bales and stores them according to pre specified categories
and groups assigned. Bale identification is done by a number of
criteria like cotton type, cotton source, gin identification and
fiber properties, etc. On the other hand, storage of bales should

be carried out in such a way that it allows easy identification and
retrieval.
(2) HVI data: It gives the provision for testing each bale accurately and
precisely as well as evaluation of fiber overall quality index. Data set
so generated is utilised by ANN based networking system algorithms
for selecting cotton bale lay down suitable for the spinning mill in
consideration.
(3) EDI data: HVI fiber properties and bale information of each
consignment sent electronically by the merchants to the various mills
in network system.
Figure 9.2 Basic EFS®-MillNet implementation set up for end product

added value.
9.5.1.2 Operating mode of basic implementation set up
It can be divided in four steps as below:
(1) First step include creation of cotton fiber data bank by carrying out HVI
testing for each bale and also collecting other associated information
like; cotton type, cost, location, etc. Based on these eight or twelve
digits barcode is generated for bale identification as per system. In
EFS®-MillNet system the testing end gets shifted from mill to the field.
So, each outgoing cotton bale from the field has its clear cut identity in
terms of barcode.
(2) Second step involves uploading of cotton details to EFS®-MillNet
system, which is accessed by respective producer (spinning mill), part
of EFS®-MillNet system via EDI.
(3) Producer identifies the suitable cotton components as per targeted
product requirements by using suitable ANN (networking) algorithm
and according place order electronically back to EFS®-MillNet system.
(4) Consignment management system transfers the lot to warehouse
management at producer’s end as per the deal. The warehouse
management now stores the bales arrived in the lot as per pre specified
category to meet the requirements of bale lay down.
9.5.2 Optimum fiber quality utilisation for target yarn by the use
of EFS® system
Spinning mill always seek efficient and effective ways to utilise maximum
technological value of cotton (minimum price/maximum quality) for getting
desired yarn (maximum price gain/maximum quality) at the least possible
cost. Here cost optimisation is the sum of all the measures, which results in
reduced manufacturing costs (Fig. 9.1). Hence the raw material represents the
greatest share not only to the total material costs (50–70%) but also quality
(80–90%), it is considered as a major factor for improving the competitiveness
of a spinning plant. That’s why one should use the most reasonable priced
cotton mix composed of as many constituents as possible, so as to meet yarn
quality requirements along with cost optimisation object. Such cotton mix
is always referred as optimum. Indeed one of the key aspects of producing
a high quality yarn is to achieve an optimum conversion of such cotton mix
fiber properties into yarn characteristics. This can be explained with simple
example of fiber strength to yarn strength conversion efficiency. The value
may range from 60–90%, as per yarn type, yarn properties like count and
twist, cotton type and fiber mix properties: length, fineness, strength, and
elongation. It is a well-known fact that longer, finer, and stronger fibers can
produce higher strength efficiency than shorter, coarser, and weaker fibers.
Thereby an appropriate selection of optimum combinations of the fiber
properties in the mix is the only way to produce optimum strength efficiency,
as a deficiency in the value of one fiber characteristic can be compensated for
by an improvement in another fiber characteristic. The EFS® system provides
many ways to achieve required optimum combination (as per application area,
yarn type) and proven to be both economically and technologically sound.
The issue of fiber-to-yarn conversion efficiency is an exhaustive one, and
it has to be dealt with in view of the application in hand. This is because, yarn
quality is typically defined in view of the type of processing used and the end
product intended. In general, the spinner may define yarn quality as an index
of appearance, strength, uniformity, and level of imperfections. The knitter
may have more detailed criteria of yarn quality. For example in continuation
with previously mentioned one in this section; strength criteria for weaver and
knitter are different. During weaving, higher end breaks should be avoided
for controlled input of yarn faults due to splicing or knotting. It is a known
fact that the yarn does not break until the tension exceeds its average strength.
Normally yarn breaks when the tension exceeds the strength of the weakest
point of the yarn. It is important therefore to achieve consistent yarn strength
with minimum weak points more as compared to further stronger yarn. So,
mix component selection is accomplished accordingly.
Conversely for knitted yarn, a combination of strength and flexibility or
pliability is required. In this case, the cotton mix should be selected in such
a way that minimum acceptable strength associated with high yarn flexibility
is obtained. Advanced implementation of the EFS® system can facilitate in
accomplishing this goal.
In light of these different and often conflicting views of yarn quality, the
spinner must customize the yarn to meet its intended purpose. In this regard,
the EFS® system represents a powerful tool to meet different end product and
processing requirements through appropriate selections of the combinations of
fiber attributes suitable for each process. Through the use of the EFS® system
quality logistics can be accomplished so that different fibers can be used for
different purposes and their proportions corresponding to fiber capability.
The basic EFS®-MillNet system provides a scientific ground for the
selection of suitable cotton constituents but not exactly their proportions in
mix. Judiciary decision-making demands a lot of complicated and tedious
statistical evaluation before drawing useful conclusions and even they are liable
to get influenced by human fatigue. Advanced implementation methodologies
of the EFS® system can serve as a scientific tool in meeting this complex
objective smoothly and reliably. One such model put forward by EI Mogazhy
et al. (2002) is incorporated here for justification.
9.5.2.1 Advanced implementation methodologies of the EFS®

system for optimum utilisation of cotton technological
value
Motive of attaining optimum yarn value vide yarn engineering concept pass
through a number of tasks (Fig. 9.3).
Figure 9.3 Advanced implementation of the EFS® system for optimum

utilisation of cotton technological value.
Step I - Determining desired cotton type

It deals with the decision made in reference to cotton types suitable
for production process. Various attributes considered for the purpose are
type of spinning system, yarn fineness and level of twist and end product.
One such example is illustrated in Figure 9.3 (i). It deals with decision
to be made for type of spinning system to be opted for meeting desired
yarn quality requirements in the best possible way. The percentile for yarn
property values observed for different spinning systems under consideration
are presented graphically. The highest values of the quality parameters are
located at the circular line of graph and value gets diminish as they approach
the circle centre.
Similarly for the well-defined production system; among the different

cotton types accessible in the market, the cotton types capable of meeting in
part or in full the technological and quality requirements of particular process
are selected vide the same graphical method. Linking with EFS®-MillNet
system, the above mentioned selection can be made from cotton data received
via EDI. The base for an appropriate selection is to achieve a cotton mix
exhibiting the average desired fiber properties with respect to the process, and
consisting of cotton that exhibit values of fiber noise that are below spinning
units maximum acceptable limits. Fiber noise here refers short fiber content,
neps, trash, and contamination.
Step II - Cotton cost optimisation

Hence cotton selected in first step holds different price as per their type,
quality, accessibility and expenditures caused. It thereby becomes utmost
important to define optimum mix value in terms of mix cost. As discussed
before it relies on proportion of each cotton type used in the mix. If these
proportions can be optimised then the resultant cotton mix formulated exhibits
the desirable fiber quality attributes, the least fiber noise and at the lowest cost
possible.
Computer based linear programming (LP) technique serves as a remedial
tool in this regard. It is the simplest optimisation technique which attempts
to maximise or minimise a linear function of decisive variables. The values
of decision variables are chosen such that a set of restricting conditions are
satisfied. It involves only two decision variables those can be solved by using
graphical method. More complex problems involving several decision or
structural variables and many constraints are resolved by analytical method
of this program. The software prepares and analyses various possible
combinations on iterative based method for selected cotton components in
mix within a very short time and executing best possible combination with
respect to lowest cost value. Accordingly respective cotton proportions are
defined in the mix.
Step III - Impact analysis

The basic aim of this entire process is to define the cotton mix which can
produce yarn with not only optimum cost but also quality. Thereby third step
is designed to determine the impact of the above two tasks on the actual values
of yarn quality. That means in third step with the virtue of linear programming
for defined cotton components and their proportions in mix, yarn quality
values are predicted and they are verified with standard or specified values. The
difference observed between actual and standard yarn values are expressed in
terms of premium/discount (eq. 9.1).
Actual value-Standard value

Premium/Discount (%) = ¥ 100 9.1
Standard value
The interpretation of impact analysis involves:
(1) Performance is as per desire: Zero discount value,

(2) Performance is better than desire: Positive discount value,
(3) Performance is poorer than desire: Negative discount value,
Some important points to be noted here for the successful implementation,
they are as follows:
(1) Different yarn types will be different in their desired levels of yarn
quality, so standards must be inputted accordingly to software.
(2) The above three tasks represent one cycle and this cycle needs to be
repeated for different processes.
(3) The analysis can be made systematic so that if the impact is not
satisfactory, the cycle can be repeated until an optimum raw material
cost can be achieved.
It should be pointed out that the above tasks can be made largely systematic
through developing simple software programs that can perform the tasks in a
user-friendly atmosphere and allow the user a great input and manipulation
depending on its specific applications.
9.5.2.2 Diagnosing quality problems
Whenever yarn quality problems occur, a logical question usually asked;

whether the problem is associated with lagging conversion efficiency of
fiber quality or machine parameters. In reality the major difficulty faced
in answering such question stems from the fact that there is a great deal of
interaction prevailing between fiber and machine that makes it difficult to
isolate each effect efficiently. This point was clearly seen in the early generation
of the EFS® system designed by keeping only fiber quality efficiency in mind.
Optimum fiber quality to yarn quality conversion was realized theoretically
with such set up, but unable to distinguish causes for quality problems, which
were arisen during actual production course. So, later on system is modified to
diagnose quality associated problems. They are addressed through developing
a prediction (regression) equation relating lay-down fiber properties to some
major quality parameters, e.g. yarn strength or yarn irregularity. A reliable
prediction equation is derived by using artificial neural networking, having
well-trained neurons on the basis of sound mill data, so as to provide good
prediction of the quality parameter. The predicted profile of the quality
parameter is then compared with the actual profile. The predicted profile
should expectedly agree (at least in trends) with the actual fiber profile as
shown in Figure 9.4, if everything goes well. But if the actual yarn quality
profile deviates from the predicted profile as shown by the dotted circle areas
of Figure 9.4, although optimum fiber mix is used (as suggested by EFS®
system), the difference is typically arisen due to processing-related problems.
Figure 9.4 Diagnoses process related quality problem by EFS®MILLNET

system.
It should be noted that the issue of FYC efficiency has to be dealt with through
advanced implementation of the EFS® system in which mill management is
committed to take appropriate actions and make intelligent decisions. It is
equally important that such implementation can be achieved at minimum or
no disturbance to mill production. Thus, in today’s competitive market FYC
represents one of the secret recipes of success that can differentiate between
the performance of one spinning mill and another.
9.5.3 Optimum fiber quantity utilisation efficiency for target yarn

by the use of EFS® system
It is also known as, fiber yield efficiency and basically deals with maximizing
the quantum of production for a given yarn type from the supplied bale cotton
by minimising (controlling) cotton waste, i.e. optimisation of yarn realization.
The expression for yarn realization is as follows:
Yarn produced per month (Kg)

Yarn Realization (%) = ¥ 100 9.2
Bale cotton consumed per month (Kg)
The above expression has two virtual extremes: 100% efficiency at which
all cotton in the bales is converted into yarn (no waste) and zero efficiency at
which no yarn was produced (all waste). Usually yarn realization values for
carded yarns are higher than combed yarns of the same size. This is mainly
attributed to the additional waste removed at comber (Noil). The key to
determine a reliable fiber-to-yarn yield efficiency is to determine the various
components influencing it. Figure 9.5 shows these components.
Figure 9.5 Various components of fiber-to-yarn yield efficiency.

This monthly exercise demands keeping an account for vast records
of various factors participating in defining fiber-to-yarn yield efficiency.
Apart from these lots of tedious and time consuming statistical evaluation
is involved for calculating all three attributes of yarn realization; bale cotton
consumed, yarn produced and different types of waste produced as well as
their interpretation. No doubt their mention is beyond the scope of this book.
These exercises were carried out manually with classical system liable to error
some due to human fatigue and wrong work practices. Further poor working
environment caused on high fly generation at the point of waste collection
deviate recorded data from actual one considerably. Automization has
relaxed this stress partially by the use of auto waste collection and recording
system as well as computer based accurate and speedy statistical evaluation.
But interpretation of such statistically derived measures and based on these
initiating corrective measures, depends still on individual’s decisions and
expertise, liable to impact adversely optimization drive. The EFS® system
incorporated with empirical data based trained neurons of ANN system can
assist in reaching scientifically the required optimum level. It facilitates in

maximising the yield through appropriate selection of cotton types that have
minimum potential waste. Critical fiber properties; cotton grade, fiber fineness
(Micronaire), fiber length, length uniformity, short fiber content, and fiber
maturity are considered in this regards. Their influence on production process
is described in details in earlier chapter.
Productivity and efficiency of spinning system are the two measure
controller for the yield of yarn. They are primarily depending on three main
factors: spinning system, yarn count and yarn twist. The ideal production rates
differs considerably for ring spinning and rotor spinning at different spindle
or rotor speeds, for different yarn counts, and different twist multiplier. They
are ideal in the sense that they assume no loss in productivity and efficiency
that means no ends down and a smooth flow of material through the spinning
process. Any deviations from these ideal rates will thereby immediately
highlighted by loss in yield. Assuming that the spinning machine is operating
under optimum conditions, the primary interruption will result from a variation
in the incoming material. Such variations can be resulted from between-
machine variation in the spinning preparation process or cotton mix variations.
Response of different spinning systems in this regards are different based on
their sensitivity to particular fiber attributes. Thereby all the material-related
causes of ends down or loss of production efficiency can be handled through
appropriate fiber selection and consistent blending. Apart from that spinning
mills should utilize the EFS® system based monitoring for ends down and
production efficiency. It is important to point out that ends down do not occur
due to particular average levels of fiber properties but occur due to variation
in the values of fiber properties, or excessive levels of fiber noise (short fibers,
trash, and contamination). Overcoming these problems represent primary
tasks that can be performed by the EFS® system to enhance yield of the yarn.
Specific spinning situations, where waste materials are recycled by mixing
with the primary material, especially in case of coarser yarns spun on either
ring-spinning or rotor-spinning, the EFS® system can assist in the addition of
careful and pre-calculated proportions of waste fibers to the primary material.
Unfortunately, there are no standard procedures to implement these tasks and
they have to be performed in view of the given conditions of the spinning
mill in question based on machinery type, cotton types used, cotton fiber
properties, yarn type, and end use application.
Differentiation in spinning mills performance is done on the basis of their
yarn realization values (FYY efficiencies) and thus regarded as one of the key
factors in defining mill profit. As it is directly related to variable cost reduction
since each 1% increase in yarn realization could mean a major reduction in the
manufacturing cost and additional thousands of kilograms of the yarn produced.
9.6 Summary
Summarising the entire discussion on advance implementation phase of
EFS system, it is quite apparent that it counterparts the basic phase in many
ways, especially in relation to cost related aspects. The basic utilisation of the
EFS® system can yield immediate economical and technological gains to the
spinner. However, optimum utilisation of cotton technological value in terms
of yarn quality and quantity can be earned by advance implementation phase
of EFS® system. It also extends help in identifying loose poles of process and
thus serves as a reliable tool for process control. The main desired features of
advance implementation phase of EFS® system are summarised in Figure 9.6.
Figure 9.6 Elements of advance implementation phase of EFS® system.

It deals with five primary critical tasks. They are as follows:
(1) Using the EFS system, determine the technological value of the cotton
and optimised cotton mix.
(2) Exploring areas of material related manufacturing cost reduction
based on the diagnostic measures inherent in the EFS system.
(3) Producing yarns that are EFS stamped for uncontested quality and low
cost.
(4) Characterising the EFS fiber profiles with respect to end product, viz;
fabric or garment, quality.
(5) Establishing an overall systematic approach of measuring cost saving
resulting from the use of the EFS system.
References
1. Garde A.R. and Subramanian T.A., Process Control in Spinning,

ATIRA publication, third edition, 1987.
2. El Mogazhy & Chewning, Cotton Fiber to Yarn Manufacturing
Technology Published by Cotton Incorporated, 2002.
3. EI Mogahzy Y.E. and Gowayed Y., 1994, Theory and practice of cotton
fiber selection. Part 1: fiber selection techniques and bale picking
algorithm. Textile Research Journal, 64, 32–40.
4. EI Mogahzy Y.E., Broughton R. and Lynch W.K.A., 1990, Statistical
fiber properties. Textile Research Journal, 60(9), 495–500.
5. EI Mogahzy Y.E., 1992, Optimizing cotton blend costs with respect to
quality using HVI and linear programmng. Part I: fundamentals and
advanced techniques of linear programming. Textile Research Journal,
62, 1–8.
6. EI Mogahzy Y.E., 1992, Optimizing cotton blend costs with respect to
quality using HVI and linear programmng. Part II: combined effects of
fiber properties and variability constraints. Textile Research Journal,
62, 108–114.
7. EI Mogahzy Y.E., Reed M. and Lynch W.K., 1998, Implemetation of
on-line fiber analysis in the textile industry, in Ergonomics of Hybrid
Automated systems 1, Ch. IX, W. Karwoski, H.R. Parsaei, and M.R.
Wilhelm, Eds., Elsevier Science Publishers, NY, pp. 695–698.
8. Shaikh T.N. and Agrawal S.A., 2011, Formulation of cotton mix: Devel-
opment from indecisive to decision support systems. International Jour-
nal of Engineering Research and Applications (IJERA), 1(3), 660–665.
9. Ratnam T.V., Chellamani K.P., 1999, Quality control in spinning; Third
Edition - Raw material, The South Indian Textile Research Association,
Coimbatore, pp 31–66.
10
Case study


*
Abstract
Product engineering for ring spun yarn, which involves the designing of yarn
parameters, scientific raw material selection, production process optimisation
and product quality evaluation; it is a very tricky aspect for the manufacturing
industries and cannot be justify by manual mode. The complexity of this
decision-making chain needs to replace human brain by a powerful data-
modelling tool that is able to capture and represent any kind of input–output
relationships, viz; Artificial Neural Network (ANN). Understanding the
structure and mode of operation of this mathematic based decision-making
system can be done well by actual practice examples. So, this chapter is
strictly designed for discussing different case studies conducted by various
researchers for addressing three main variants of cotton spinning, viz; cotton
fiber characteristics, yarn characteristics and process parameters.
Key Words: ANN, SCI, validation test, impact analysis, MSE, over fitting
10.1 Introduction
Product engineering for ring spun yarn, which involves the designing of yarn
parameters, scientific raw material selection, production process optimisation
and product quality evaluation; it is a very tricky aspect for the manufacturing
industries and cannot be justify by manual mode. The complexity of this
decision-making chain need to replace human brain by a powerful data-
modelling tool that is able to capture and represent any kind of input– output
relationships, viz; Artificial Neural Network (ANN). Understanding the

structure and mode of operation of this mathematic based decision-making
system can be done well by actual practice examples. So, this chapter is strictly
designed for discussing different case studies only.
10.2 Case study I

Majmudar A, Majmudar P.K. and Sarkar, B.(2006) have investigated on the
feasibility cotton ring spun yarn engineering using reverse modelling concept
of artificial neural network (ANN). They had used linear programming
approach to ensure that the yarn engineering is achieved at the minimum raw
material cost. Raw materials in the form of cotton fibres were selected on
the basis of only two quality parameters, viz; SCI and micronaire, in order
to reduce the complexity of lay-down formation. The experimental results
have demonstrated that the engineered yarns were having reasonably good
agreement with target yarns in terms of most of the properties.
So, their experimentation has been represented over here as one of the case
study to explain significance of this tool.
10.2.1 Basic attributes used for ANN reverse yarn engineering

model
Input: Yarn tenacity (cN/tex), breaking elongation (%), unevenness (CV %)

and hairiness index were chosen as the inputs of ANN. The basis of selection
was confined as, these four yarn properties are commonly tested in the textile
industries and also prescribed by the customer while ordering the consignment.
Output: The characteristics of cotton fiber should be the expected output
of reverse yarn engineering model. But spinning consistency index (SCI) and
micronaire were used for this purpose in the study.
SCI, a comprehensive quality index of all the cotton fiber properties
measured by high-volume instrument (HVI) was used in replacement of
individuals. The reason behind was to avoid undue structural complexity of
the ANN model and meeting ease of practical implementation. As in actual
practices it becomes impractical to formulate a cotton mix that would fulfil all
the cotton properties predicted by ANN.
Second output parameter added in network was micronaire to exert
additional control especially during ring spinning. Its impact on the process is
well defined in earlier chapter also.
ANN model: The schematic representation of ANN model used for the
study is given in Figure 10.1.
Figure 10.1 Structure of artificial neural network (ANN) model for yarn
engineering.
10.2.2 Sample preparation
Twenty-five controlled samples of ring-spun cotton yarn having linear densities

ranging from 34s Ne to 90 s Ne were spun in a spinning mill by using optimised
process parameters. The summary statistics of cotton fiber properties in the
mix based on HVI testing are given in Table 10.1. It can be seen that the range
of SCI and fiber fineness used was 67 and 1.26, respectively.
Table 10.1 Summary statistics of cotton fiber properties
Fiber properties Minimum Maximum Average
Fiber bundle tenacity (cN/tex) 29.48 40.80 34.73
Fiber bundle elongation (%) 6.44 7.95 6.97
UHML (inches) 1.11 1.40 1.25
Uniformity index 81.80 86.37 84.04
Micronaire 3.1 4.35 3.92
Reflectance degree 68.38 77.04 75.16
Yellowness 8.27 12.56 9.31
SCI 126 193 161.72
Yarns so manufactured were tested for above mentioned four properties.
(1) Tensile properties: USTER Tenso-jet operating at a constant rate of

elongation of 400 m/min. For each sample, 10,000 individual tests
were conducted to obtain a consistent mean value.
(2) Unevenness CV% and hairiness index: They were measured using
USTER Tester III at a speed of 400 m/min for 2 min. For each sample,
five ring bobbins were tested and the average unevenness CV% and
hairiness index were calculated.
10.2.3 Optimisation of ANN parameters
Artificial neural network architecture: Only one hidden layer was used for
“reverse” modelling architecture. The number of nodes in the hidden layer and
learning parameters, i.e. learning rate and momentum, were optimised by trial
and error method, so that minimum prediction error could be attained in the
unseen test data set. Log sigmoid transfer function was chosen in the hidden
and output layers so that the non-linear relationship between fibre and yarn
properties could be captured.
Training of ANN: It was done using the back-propagation algorithm
developed by Rumelhart et al. (1986). From the available 25 data sets, 20
data sets were used for the training of ANN model. The remaining five data
sets were used for testing, which means that these yarns were attempted for
engineering.
Mathematical functions used during training: The calculation of error
vector is done from the difference between actual output and predicted output.
The error function (E) is defined as the sum of errors due to each pattern as
shown in equation 10.1 as follows:
m
E = ÂE j
j =1
10.1
Where, Ej is the error associated with the jth pattern and m is total number
of training patterns. The expression of Ej is given as shown in equation 10.2.
1 s
Ej = Â(Tr - Or )2
2 r =1 10.2
Where, Tr and Or denote the target output and predicted output, respectively,
at output node r, and s is the total number of output nodes. The corrections
necessary in the synaptic weight between output and hidden layer are carried
out by a delta rule as shown below:
W jk = -h ÈÎ∂E / ∂W jk ˘˚
= h [(Tk - Out k ) Out k (1 - Out k )] Out j ; 10.3
= hd k Out j
Where, Wjk is the weight connecting the neurons j of hidden layer and
neuron k of the output layer, Wjk the correction applied to Wjk at a particular
iteration, η is a constant known as learning rate and out j the output of neuron
j. The weight change of the hidden layer to the input layer is calculated as
shown in equation 10.4 as below:
W jk = h f ' (net j ) X i Âd kW jk 10.4
Where, Wij is the correction applied to the weight connecting input neuron
i and hidden neuron j, and Xi is the input received by the neuron i. As the
dimensions of two outputs, namely SCI and micronaire were different, therefore,
normalised error corresponding to each output was calculated as shown below:
Absolute error
Normalized error =
Range
Actual value - Predictedvalue 10.5
=
Maximumvalue - Minimumvalue
Testing of ANN: Training was ceased when the sum of mean squared
normalised error of two outputs reached to the minimum level in the testing
set. On testing it was found that six nodes in the hidden layer after 5,000
iteration are giving the best prediction results. The detailed prediction results
are shown in Table 10.2.
Table 10.2 Prediction results of SCI and micronaire in testing data set
Test Actual Predicted Error (%) Normalised error
sample combination combination
number SCI Micronaire SCI Micronaire SCI Micronaire SCI Micronaire
and
count
40s Ne 130 4.00 139 4.19 6.92 4.75 0.134 0.151
50s Ne 165 4.20 147 4.16 10.91 0.95 0.269 0.032
50s Ne 154 4.14 155 4.14 0.65 0.00 0.015 0.000
80s Ne 188 3.10 192 3.16 2.13 1.94 0.060 0.048
60s Ne 155 4.08 157 4.16 1.29 1.96 0.030 0.063
Mean 4.38 1.92 0.102 0.059
Accuracy of testing: It is observed that the mean absolute error of

prediction is 4.38% and 1.92% for SCI and micronaire, respectively (Table
10.2). Furthermore, the entire individual prediction error is <10% except in
one case of SCI. The correlation coefficient between the actual value and
predicted value was 0.88 for SCI and 0.98 for fiber fineness.
10.2.4 Validation of ANN prediction
It involves manufacturing of engineered yarn to validate whether the cotton

mix having predicted combination of SCI and micronaire is really capable of
producing the target yarn properties. Three fresh samples of yarns (test sample
nos. 1, 3 and 5 of Table 10.2) were spun for the purpose. The properties and
price of cotton fibers used for the production of engineered yarn are given in
Table 10.3. Only the binary blend of cotton varieties was tried.
Table 10.3 Properties and price of cotton fibers used for yarn engineering
Cotton fiber type Spinning Fineness Price Yarn count
consistency index spun (Ne)
(code) (µ) (Rs/kg)
A 160 4.01 65 40 and 50
B 126 4.35 52 40
C 154 4.20 59 50
D 165 4.20 66 60
E 150 4.06 62 60
Classical linear programming approach was adopted in order to minimise

the raw material cost. This has also ensured that the minimum requirement of
the product mix in terms of SCI and fibre fineness is fulfilled.
10.2.5 Linear programming
The values for fiber constraints, viz; SCI and micronaire have already been
determined by ANN (Table 10.2), so accordingly cotton fiber types were
chosen for engineered yarn production. For example, the 40s Ne yarn was spun
using cotton fibre type A and B (Table 10.3). Therefore, linear programming
was designed to meet these objective function and constraint as described in
equations (10.6–10.8) for 40 s Ne yarn in terms of cost (Z), SCI and fineness,
respectively.
Minimize Z = PA .65 + PB .52 10.6
PA .160 + PB .126 ≥ 139 10.7
1 1 1
PA . + PB . ≥ 10.8
4.01 4.35 4.19
Where,
PA + PB = 1, and PA ≥ 0, PB ≥ 0
Similar objective functions and constraint equations were formulated for

rest of the engineered yarns, viz; 50 s Ne and 60 s Ne yarns also. The linear
programming problems were solved by simplex method.
10.2.6 Impact analysis
This involves analysis of both inputted and outputted attributes of ANN

modelling. Cotton mix quality and cost values obtained by the virtue of linear
programming were compared with suggested values of ANN. Also engineered

yarn quality parameters were compared with targeted yarn values. These can
define the success of the ANN architecture used for the study.
10.2.6.1 Cotton mix properties
The basic objective of linear programming is to define proportion of various

types of cotton required to fulfil the average SCI and micronaire value of the
cotton mix at minimum possible cost. Test results obtained on implementing
above mentioned linear programming for selected engineered yarns are
reported in Table 10.4.
Table 10.4 Proportion of cotton fiber used and mix cost
Yarn Cotton fiber type Product mix Mix cost

count properties (Rs/kg)
(Ne)
A B C D E SCI Fineness
(µ)
40 45.40 54.60 ---- --- ---- 141.4 4.19 57.90
50 30.10 ---- 69.90 --- ---- 155.8 4.14 60.81
60 --- ---- ---- 46.66 53.33 157.0 4.12 63.87
It is observed that for 40sNe yarn, 45.40% of cotton A and 54.60% of

cotton B have to be used in the mix to fulfil the SCI and fibre fineness
requirement of 139 and 4.19, respectively. It will be relatable here to
mention that for this particular yarn, SCI becomes a slack variable, as the
achieved value in the mix (141.4) is higher than the minimum suggested one
(139.0). The mixing cost per kilogram becomes Rs 57.90 (= Rs 65 × 0.454 +
Rs 52 × 0.546).
Hence fiber mix cost is always higher for finer count, going in agreement
to this perception comparatively higher mixing cost values were observed for
50sNe and 60sNe yarns, Rs 60.81/kg and Rs 63.86/kg, respectively.
10.2.6.2 Properties of engineered yarns

Comparative status for the properties of three target yarns and corresponding
properties of engineered yarns is given in Table 10.5 and also illustrated
graphically in Figure 10.2. The deviation (%) between the target and engineered
yarn properties is given in Table 10.6.
Table 10.5 Properties of target and engineered yarns

Test sample Yarn count Tenacity Elongation Unevenness Hairiness
number (%)
(Ne) (cN/tex) (CV%)
Target yarn properties
1 40 18.70 3.30 9.43 4.67
2 50 20.51 3.64 10.16 3.62
3 60 21.80 3.10 10.90 3.85
Engineered yarn properties
1 40 19.21 3.35 9.80 5.22
2 50 21.23 3.66 10.14 3.70
3 60 21.15 2.67 10.61 4.06
Figure 10.2 Properties of target and engineered yarns.
Table 10.6 Deviation between the target yarn and engineered yarn
Test Yarn Absolute deviation (%) Average

sample count Tenacity Elongation Unevenness Hairiness deviation
number (Ne) (%) (%)
1 40 2.727 1.515 3.924 11.777 4.986

2 50 3.510 0.549 0.197 2.210 1.617
3 60 2.982 13.871 2.661 5.455 6.242
It is quite clear from results (Tables (10.5, 10. 6)) that the tenacity and
unevenness CV% values of target and engineered yarns have gone in reasonably
good agreement. All the individual deviations of tenacity and unevenness CV%
are well within the limit of 5%, which is quite encouraging. However, in case
of elongation and hairiness, there is one reading in each case with deviation
>10%. For elongation, 60 s Ne yarn is showing deviation of 13.871%. It should
be noted here that SCI regression equation does not include fibre elongation,
which is a strong determinant of yarn elongation (Majmudar and Majmudar,
2004). This may be the most plausible reason behind the high deviation (%) of
elongation found for 60 s Ne yarn. The hairiness of 40 s Ne yarn is also having a
high deviation of 11.777%. The maximum average deviation of 6.242% is gone
in the account of 60 s Ne yarn whereas 50 s Ne yarn demonstrates the minimum
average deviation of 1.617%, with respect to their targeted counterpart.
10.3 Case study II

Mohsen Shanbeh, Hossein Hasani and Somayeh Akhavan Tabatabael (2011)
have worked on modelling and predicting the breaking strength and mass
irregularity of cotton rotor-spun yarns containing cotton fiber recovered
from ginning process by using artificial neural network trained with back
propagation algorithm. They have met with satisfactory results of prediction
for breaking strength and mass irregularities of rotor spun yarn. Their work has
been incorporated here for understanding insight strengths of yarn engineering
techniques.
10.3.1 Basic attributes of artificial neural network
Inputted parameters: Process parameters of rotor spinning, viz; rotor

diameter (mm), naval type, rotor speed (rpm), opener roller speed (rpm), yarn
count (Ne) and cotton ginning waste percentage.
All these parameters were expressed in a vector form and qualitative
parameters like naval type with encoding 0, 1 and 2.
Outputted parameters: Breaking strength and mass irregularity (CVm%) of
rotor spun yarn Artificial Neural Networking Parameters:
Types of network: Two feed forward multilayer neural network models, based
on back propagation learning algorithm with one neuron as output, and six-input
unit in input layer were designed. Over fitting of network was prevented by using
the weight decay method for a small sample data. Mean square error regularisation
(MSEREG) performance function (equation 10.9) was used instead of common
mean square error function for modifying the performance function.
msereg = g ( mse) + (1 - g ) msw 10.9

Where, g is the performance ratio and
1 n 2
msw = ÂW j
n j =1 10.10
Here n is the number of weighted connections.
10.3.2 Sample preparation
Cotton fibers and fibers recovered from ginning process were used in three
different proportions (50–50%, 35–65% and 65–35%) and fiber properties for
each variable summarised in Table 10.7.
Table 10.7 Summary of fiber properties
Fiber characteristics Cotton Ginning waste
Tenacity (cN/tex) 27.1 23.2
Breaking elongation (%) 6.7 6.6
Mean length by weight (mm) 23.90 21.77
Short fiber content by weight (%) 7 11.1
Maturity index 0.82 0.81
Micronaire 4.38 4.29
UQL (mm) 28.89 27.19
Length CVw (%) 1.91 1.94
The fibers were blended after carding at the first passage in the draw frame
and second draw frame passage was used to improve homogeneity for better
rotor spinning. Such slivers were used to produce yarns with three linear
densities 12, 16 and 20s Ne. Eighty-one yarn samples were produced with
different machine, yarn and fiber parameters (Table 10.8). Samples were
conditioned for 24 hours at standard atmospheric conditions prior to testing.
ASTM-D2256 and ASTM-D 1425 were the standard test methods used for
measuring breaking strength and irregularities of yarns, respectively.
Table 10.8 Rotor spinning machine parameters
Machine parameter Setting values
Rotor diameter (mm) 48, 54, 66
Rotor speed (rpm) 41,000; 47, 000; 53, 000
Opening roller type OK40 (for cotton fibers)
Opening roller speed (rpm) 9800, 8400, 7350
Naval type (Steel); 8, 4 and without flutes
Delivery speed (m/min) 52.4 m/min
Samples were divided randomly in training and testing sets. Seventy-two and
nine sets of data were selected for training the neural network models and testing
the predictive power of developed models, respectively. To eliminate the effect of
different units of input and output parameters, data normalising was carried out
in such a way that they got zero mean and unit standard deviation. The sigmoid
and linear transfer functions were used for hidden neurons and output neurons,
respectively. Neural networks with one hidden layer are suitable for majority of
applications, and the second hidden layer can improve the performance of the
network if there is a complex relationship between the input and output parameters.
Therefore, number of hidden neurons and the number of hidden layers were
adjusted by trial and error for problem dependent parameters. Fifteen topology
with one or two hidden layers and 6–12 neurons in hidden layers were tested for
the purpose. The mean square error of testing sets was considered for getting the
best topology. The schematic of topologies of ANN models has been illustrated
in Figure 10.3.
Figure 10.3 Types of ANNs.
10.3.3 Experimentation
It was carried out in to two phases.

Phase I: Capability of different ANN models under study to predict the
mass irregularity and breaking strength of cotton/ginning waste rotor spun
yarns were realized and the best one was sorted out.
Phase II: Using the best capacity ANN topology for the selected yarn property,
impact analysis of the rotor spun yarn production parameters was conducted.
Phase I: Analysis of artificial neural network models capability: Initially
trials were taken with 6-8-6-1 topology in the 2,000 epoch in order to identify
the optimum value of learning and momentum rate. The suitable learning rates
found were 0.15 and 0.45 for models of breaking strength and mass irregularities,
respectively. Using these values optimum values of momentum rates 0.9 and 0.7
were found for two models, respectively. Optimised values of learning rate and
momentum rate were thence after applied in training process for both the models.
The ANN model with two hidden layers and eight processing elements into
the first and second hidden layers (6-8-8-1 topology) gave the best performance
and the least MSE on testing data for predicting the breaking strength after
2,000 epoch. On the other hand 6-10-8-1 topology gave the best performance
for predicting the mass irregularity.
Phase II: Impact analysis: Four process related and two material related
parameters having considerable influence on the two major rotor spun yarn
properties. Breaking strength (Figure 10.4) and mass irregularities (Figure
10.5) were inputted to verify their influence.
Figure 10.4 Impact analysis for breaking strength.
Figure 10.5 Impact analysis for mass irregularity.

It could be observed that opening roller speed had the least impact on
breaking strength and irregularities. Waste percentage and yarn count were
found major parameters influencing rotor spun yarn mass irregularity.
Whereas rotor speed and waste percentage had shown greater impact on the
breaking strength.
Thus with the implementation of ANN, influence of process parameters and
material parameters on the performance of rotor spun yarn breaking strength
and mass irregularity can be done more accurately.
10.4 Case study III

Area of application of ANN is not only restricted in terms of scientific selection
of cotton mix components by yarn reverse engineering as studied in case I or
prediction about impact of process variables on yarn quality as per case study
II. This can be further extended up to fabric as well as garment stage, even
applicable to other domains of textile applications like technical textiles equally.
Linking up yarn engineering with fabric engineering can also facilitate in
selecting suitable cotton components in meeting desired fabric quality. Such
case study will be too lengthen and hence yarn reverse engineering is well
mentioned in case study I, this study is restricted for fabric engineering part only.
Aesthetic properties of fabrics have been considered as the most important
fabric attribute for years. However, recently there has been a paradigm shift in
the domain of textile material applications and consequently more emphasis
is now being given on the mechanical and functional properties of fabrics
rather than its aesthetic appeal. Moreover, in certain woven fabrics used for
technical applications, strength is a decisive quality parameter. Thereby, a case
study has been put up here for the prediction of tensile strength of plain woven
fabrics; a work done by Abhijit Majmudar, Anindya Ghose, Shib Sankar Saha,
Ayan Roy, Subir Burman, Dhrubajyoti Panigrahi and Anjan Biswas (2008) by
using empirical modelling method namely artificial neural network (ANN).
10.4.1 Artificial neural networking
Back-propagation algorithm was employed to predict the fabric strength, as it

is suitable for the prediction modelling. The algorithm uses the input–output
data during training to optimise the synaptic weights, so that the error signal
between actual and predicted output becomes minimum. According to this
algorithm, training occurs in two phases, namely forward pass and backward
pass. In the forward pass, a set of data is presented to the network as input
and its effect is propagated, in stages, through different layers of the network.
Finally, a set of outputs is produced and the error vector is calculated according
to the following equation:
p
E = ÂE j
j =1
10.11
Where, E is the error vector, Ej is the error associated with the jth pattern and
P is the total number of training patterns. The expression of Ej is given in the
following equation:
1 s
Ej = Â (Tk - Outk )2
2 K =1 10.12
Where, Tk and outk, the target output and predicted output, respectively at output
node k and S is the total number of output nodes. In the backward pass, this
error signal is propagated backwards to the network and the synaptic weights
are adjusted in such a manner that the error signal decreases in each iteration
process. The corrections necessary in the synaptic weights between output and
hidden layer are carried out by a delta rule. Once the training is completed the
developed model can predict the output for any set of input pattern.
10.4.2 ANN parameters
The prediction performance of ANN model is largely influenced by

construction of a proper network structure and optimisation of learning
parameters. The number of hidden layers and the number of neurons in each
hidden layer are the important structural parameters to be determined. Since
one hidden layer was capable of handling most of the situations in absentia
of highly complex input–output functional relationship, only one hidden layer
was used in this investigation. Transfer function in the hidden layer was log-
sigmoid as shown below:
1
f (Z ) =
(1 + � e z ) 10.13
Where, Z is the weighted sum of inputs to a neuron and f (Z) is the transformed
output from that neuron.
Input parameters: There was five input parameters namely warp yarn
strength (N), warp yarn elongation (mm), ends per inch (EPI), picks per inch
(PPI) and weft count (Ne). So, the number of nodes in the input layer of ANN
was five.
Output parameters: There was only one output; fabric strip strength in
warp direction. So, only one node was kept in the output layer of ANN to
predict it.
Training of ANN: It was done using standard back-propagation algorithm.

The number of nodes in the hidden layer and learning parameters (learning
rate and momentum) was optimised by trial and error method. Training was
ceased when the error in the unseen testing data reaches the minimum level.
For ANN modelling, Easy NN plus software was used. Statistical software
(version 5.0) was used to develop the multiple linear regression model and to
determine the relative contribution of input variables.
10.4.3 Experimental
Thirty-three (33) plain-woven fabric samples having wide range of yarn count
and cloth setting were collected. They had been tested for inputted as well
as outputted parameters, viz; ends per inch (EPI), picks per inch (PPI) and
weft count (Ne), tensile properties of warp yarns and fabric strength in warp
direction by using standard testing methods. These had created authentic
databank of 33 input–output data sets. Out of those available, 27 data sets
were randomly chosen as training data and were used for the development
of ANN and linear regression models. Remaining six data sets were kept as
testing data to evaluate the prediction performance of the models. Statistical
parameters like mean absolute error % and correlation coefficient were used
to judge the prediction performance of the developed models. Absolute error
% was calculated by using the following expression:
Actual value - Predicted value

Absolute error (%) = ¥ 100 10.14
Actual value
10.4.4 Prediction performance of ANN
The prediction performance of selected and trained ANN was evaluated. The
summary of prediction performance of ANN is shown in Table 10.9. As per
expected in case of training data, the prediction performance of ANN model
was found much better.
Table 10.9 Prediction performance of ANN
Statistical parameters Training data Testing data

Correlation coefficient (R) 0.92 0.95
Mean absolute error (%) 7.83 7.78
The actual and predicted fabric strength values in the unseen or testing data
sets are summarised in Table 10.10.
Table 10.10 Details of fabric strip strength for testing data set
Fabric strength (N) Absolute error (%)
Actual Predicted
461.45 539.93 17.01
754.63 689.09 8.69
813.22 707.01 13.06
561.60 558.73 0.51
447.05 427.39 4.40
574.50 557.23 3.01
Mean error (%) 7.78
The mean error of prediction for ANN model in the unseen testing data is
7.78%, which is almost equal to the mean error of ANN model for the training
data (7.83%). Comparable prediction error in the training and testing data
strengthen the fact that the ANN model was adequately trained and was able
to achieve generalisation.
10.4.5 Analysis of input parameters influence

The influence of various input parameters on the fabric strip strength was
analysed with the developed and trained ANN model in second phase
of experimentation. Sensitivity analysis was done in order to judge the
importance of various input parameters in accordance with the ANN model.
All the parameters were kept at their middle value, except for the parameter
whose sensitivity is being estimated. Then, the parameter, whose importance
or sensitivity was evaluated, varied from its minimum to maximum level and
the corresponding change in the dependent variable or fabric strength was
noted. This change was expressed as a fraction of the range of dependent
variable. The results of sensitivity analysis of ANN have been shown in
Table 10.11.
Table 10.11 Influence of input parameters on fabric strip strength

Input parameters Sensitivity value
Warp yarn strength (N) 0.631
Warp elongation (mm) 0.038
EPI 0.420
PPI 0.235
Weft count (Ne) 0.018
It is clear from the results that fabric strip strength in warp direction was
mainly influenced by the warp yarn strength (0.631) and EPI (0.420). Weft
count and PPI are the factors along with yarn to yarn friction, which influence
fabric assistance and thereby fabric strength. However, the lowest sensitivity
value was measured for weft count (Ne). When all other fabric parameters
remain constant, coarse weft yarns will lead to higher fabric strength in
warp direction, as it can cover more length of warp yarns. Therefore, the
weak spots of warp yarns are expected to be more effectively buried by
the coarser weft, leading to greater fabric assistance and fabric strength in
warp direction. PPI ranked third in the grading of input parameters. The
contributions of other two parameters namely warp elongation and weft
count were found negligible.
10.4.6 Trend analysis by ANN model
Trend analysis was performed with the developed ANN model in order to
understand the effect of various input parameters on the fabric strength.
The effect on the output was noted by changing only one input parameter
at a time and keeping all other inputs constant at their middle value. The
impact of selected input parameters was studied at five points by dividing
the entire range of all the input parameters. For example, the minimum
and maximum values of EPI in this study were 60 and 156, respectively.
Therefore, the fabric strength was noted at five EPI values of 60, 84, 108,
132 and 156 keeping other inputs such as yarn strength, yarn elongation,
PPI and weft count at the middle values of their respective ranges. The
results of analysis are illustrated graphically in Figure 10.6. A linear
relationship was observed for the strength of warp yarns and EPI increase
with the fabric strip strength in warp direction, no doubt the trend became
slightly non-linear in case of the latter. During the tensile rupture of fabric
in warp direction, most of the warp threads break and thereby contribute
towards the fabric strength. Moreover, if the EPI value is higher in that
case more warp threads contribute towards the fabric strength and therefore
associated increase in the fabric strength was observed. It depicts from
graphical relationship that when the PPI value goes beyond 70, the strength
of fabric increases with the increase of PPI value. This may be attributed
to the enhancement of fabric assistance arising from more number of cross
threads. Trend analysis demonstrates that warp elongation and weft count
have negligible influence on the fabric strength.
Figure 10.6 Graphical presentation of trends.

Most of the trends observed are going in close agreement with the
fundamental perception of fabric mechanics. Therefore, it could be
inferred that ANN model is giving a clear insight about the role of various
input parameters on the fabric strength and provides immense help for the
engineering of woven fabrics.
Summary
It can be seen from above three case studies successful implementation of
ANN instead of BNN (Biological Neural Network) in different areas of
textile manufacturing. The pattern of study and kind of functions as well as
commands used were differed as per problem type studied.
References
1. Abhijit Majmudar, Anindya Ghose, Shib Sankar Saha, Ayan Roy, Subir
Burman, Dhrubajyoti Panigrahi and Anjan Biswas, 2008, Empirical
modelling of tensile strength of woven fabrics. Fibers and Polymers,
9(2), 240–245.
2. Aggarwal S.K., 1989, A model to estimate the breaking elongation of

high twist ring spun cotton yarns, Part 1: Derivation of the model for
yarns from single cotton varieties. Textile Research Journal, 59(11),
691–695.
3. Bodan J.F., 1967, The prediction of cotton yarn strengths. Textile
4. Chattopadhyay R., Guha A. and Mahapatra U., 2003, Simulation of
spinning process: A comparative study. In: Proceedings of the 7th
Asian Textile Conference, December 1–3, New Delhi.
5. Cheng L. and Adams D.L., 1995, Yarn strength prediction using neural
networks. Textile Research Journal, 65(9), 495–500.
6. El Moghazy Y.E., 1992, Optimizing cotton blend cost with respect
to quality using HVI fibre properties and linear programming, Part
1: Fundamentals and advanced techniques of linear programming.
7. Fiori L.A., 1965, Blending. In: Hamby, D. S. (ed.), The American
Cotton Handbook, Vol. 1, Interscience Publications, New York, pp.
132–206.
8. Frydrych I., 1992, A new approach for predicting strength properties
of yarns. Textile Research Journal, 62(6), 340–348.
9. Guha A., 2002, Application of Artificial Neural Networks for Predicting
Yarn Properties and Process Parameters, PhD thesis, Indian Institute of
Technology, New Delhi.
10. Hafez O.M.A., 1978, Yarn strength prediction of American cottons.
11. Hunter L., 1988, Prediction of cotton processing performance and yarn
properties from HVI test results. Melliend Textilber, 69, E123–E124.
12. Kothari V.K., 1999, Testing and Quality Management, IAFL
Publications, New Delhi.
13. Majmudar P.K. and Majmudar A., 2004, Predicting the breaking
652–655.
14. Mohsen Shanbeh, Hossein Hasani and Somayeh Akhavan Tabatabael,
2011, Modelling and predicting the breaking strength and mass
irregularity of cotton rotor-spun yarns containing cotton fiber recovered

from ginning process by using Artificial Neural Network algorithm.
Modelling and Stimulation in Engineering, Hindwai publishing
corporation, 4, 1–8.
15. Morris P.J., Merkin J.H. and Rennil R.W., 1999, Modelling of yarn
properties from fibre properties. Journal Textile Institute, 90, 322–335.
Appendix
Chapter 2
YarnFineness (Tex y )
NumberofFibers N f = ( ) Fiberfineness (Tex f )
≥ 33; forcombedyarn;
≥ 75; forcardedyarn ……… Equation 2.1
Chapter 3
∑ L ;…..equation 3.1
n
i =1 i
LN =
∑i
Where, L= fiber length, i = number of fiber
∑ W L ;…….. equation 3.2

n
i =1 i i
MLW =
∑W i
Where, L = Fiber length and W = Weight of fiber length group.

Upper quartile length = length exceeded by 25% of the fibers ………equation 3.3
σ
CV % = ¥ 100 ºº..equation 3.4
ML
InterQuartileLengthRange
Dispersion ( % ) = × 100 …..equation 3.5
EffectiveLength
238 Appendix
50% SpanLength
Uniformityratio = × 100 …………….. equation 3.6
2.5% SpanLength
MLN × 100
UI N = ………..equation 3.7
UHML N
MLW × 100
UI W = ………..equation 3.8
UHML W
Short Fibre Index (SFI ) = 122.56 − (12.87 × UHML ) − (1.22 × UI )…..equation 3.9
Where UHML = Upper Half Mean Length (inches), UI = Uniformity Index
Short Fibre Index (SFI ) = 90.34 – (37.47 xS.L.2.5% )

– (0.90 x UR) ºººººequation 3.10
Area of fiber array covered by short fibers

SFC W % = × 100 …..equation 3.11
Total Area of fiber array
Base Length of fiber array covered by short fibers

SFC N % =
Total Base Length of fiber array
¥100 º..equation 3.12
EffectiveLength = 1.013 ×2.5% SpanLength +4.39 …… equation 3.13
MeanLength = 1.242 ×50% SpanLength +9.78…… equation 3.14
weight in mg
mtex = ………… Equation 3.15
Lengthin Km
weight in µ g
Micron = ……..Equation 3.16
Lengthininches
weightinmg × 106
Fiberfineness ( mtex ) = ………..equation 3.17
100 × fiberlength ( mm )
Or
weightinmicrogram
Fiberfineness ( Micronaire ) = ………..equation 3.18
100 × fiberlength (inch )
Appendix 239
π dl 1
Specific surface area ( s ) = 2
∞ ……..equation 3.19
π × d 4 ×l d
Normal − Dead
Maturity ratio = + 0.70……… equation 3.21
200
M + 0.6 H + 0.4 I
Maturity Coefficient = ………..equation 3.22
100
Breaking Load (lbs )

Pressley Index ( PI ) = ……..equation 3.23
Fiber Beard weight (mg )
 g 
Breaking LengthORTenacity   = 5.36 × PI ……..equation 3.24
 tex 
 lbs 
Tensile Strength  .inch = (10.8116 × PI ) − 0.12 × 103 …… equation 3.25
 sq 
Similarly for comparison purpose physical measures of Stelometer are con-

verted to tensile strength (equation 3.25).
 g  BreakingLoad ( Kg ) × 1.5 ×10
Tensilestrength   = … equation 3.26
 tex  FiberBeardWeight (mg )
Trash weight × 100

TrashContent ( T% ) = ……….Equation 3.27
Lint weight × 100

Lint Content ( L% ) = ……….Equation 3.28
Dust weight × 100

Dust Content ( D% ) = ……….Equation 3.29
W
R% = × 100 ………..equation3.30
D
240 Appendix
W
C%= × 100 ………..equation3.31
D +W
Trash ( %) in feed - Trash ( %) in delivery

Cleaning Efficiency ( %) =
Trash ( % ) in feed
¥100 º.equation 3.32
Chapter 4
i=n
X = Âxi ººº equation 4.1(i )
i=0
i=n
X = ∑wi xi ……… equation 4.1(ii )
i=0
(n + 1)th
Median = observations, when nis odd ………equation 4.2
2
th
 n   n  
 2  +  2 + 1 
Median =   observations, when nis even ………equation 4.3
2
Range ( R ) = X max − X min ……… equation 4.4
Where, Xmax= maximum value observation;
Xmin =minimum value observation.
∑
n
R
i =1 i
ωn = ……… equation 4.5
n
wn *100
P.M .R. = ……… equation 4.6
x
Mean deviation =
∑ x − x ………equation
4.7
n
∑ x−x
Percentage Mean deviation = n *100 ……… equation 4.8

x
Appendix 241
Where x arithmetic mean, x = observed value and n = number of the obser-

vations.
( X – x)
2
∑
n
Standard Deviation (σ ) = ± i=0 i

……… equation 4.9
n
σ
Coefficient of Variation (CV ) % = ×100 ……….equation 4.10
x
Variance =
∑(x − x) 2
……… equation
n 4.11
 d − 1
u 2 = Q.Ne  + a ( d − 1) + ur2 ………equation 4.12
 d 
F
Q = 29.4 [ ]2 ……… equation 4.13
L
Lsm
Fiber quality index = FQI = ……… equation 4.14
f
. .FS
UHMLUI
FQI HVI = ……… equation 4.15
FF
Lea CSP = 320 ( )

FQI + 1 − 13C ; Carded yarn ………equation 4.16
Or
Lea CSP = 320

 ( )
FQI + 1 − 13C  1 +
W 
  100 
; Combed yarn ……….equation 4.17
Where; C= Count Spun & W = Percentage Noil
Similarly for HVI based measurement the relationships will take the follow-
ing shape:
Lea CSP = 280 ( )

242 Appendix
LeaCSP = ÈÎ 280 ( ) Ê
FQI + 700 - 13C ˘˚ Á1 +
W ˆ
Ë 100 ˜¯
;
Combed yarnººº.equation 4.19
Actual yarnCSP
Process Proficiency ( P %) =
Theoretically Calculated CSP
¥100 ºººequation 4.20
MFQI = [UHM *UI * STRb * (1 + EL) * (1 − SF ) ] / MIC ]………equation 4.21
Sy = 0.041 MFQI + 5.798……… equation 4.22
SCI = -414.67 + 2.9 FS + 49.17UHML + 4.74UI - 9.32 FF

+ 0.65Rd + 0.36 ( + b )ºººequation 4.23
Where; Rd = reflectance degree and +b = yellowness of cotton fiber.
For HVI Calibration mode:
SCI = -414.67 +(2.9Strength) - (9.32 Mic.) + (49.178Lengthininch)

+ ( 4.74Uniformity Index ) + 0.65 Rd +0.36 ( + b )ºequation 4.24
If no color module is installed then SCI equation is as follows:
SCI = 322.98 + ( 2.89Strength) - ( 9.02 Mic.) + ( 49.17 Length)

+(4.29Uniformity Index )ºequation 4.25
k
PDI = ∑ (Ci Di ) ……… equation 4.26
i =1
YarnTenacity = C 1 +C 2 FS +C3 FE +C4 UHML +C 5UI

+C6 SFC +C7 FF ºººequation 4.27
 B 
Ci = 100  k i  R 2 ……… equation 4.28
 ∑ Bi 
i =1
( xi − µi )
Di = ……… equation 4.29
σi
Appendix 243
Where xi= i th fiber property of a cotton, μi= overall average and σi= over all
standard deviation of all the cottons in the ith property.
Mogahzy at el (1990) combined Trash content and Short fiber content and
stated Processing Performance Index I as:
1
I= ……… equation 4.30
(TC ) ( SFC )
Where; TC = Trash Content (by weight) and SFC = Short Fiber Content
FS 0.27 FE 0.039UHML0.291UI 0.145

MI AHP = ……… equation 4.31
FF 0.11 SFC 0.145
 1 − SF 
IG = 0.1 Lm *UI *  * MAT * FI −0.5 ……… equation 4.32
 100 
UHM*UI*(100 − SF)
IG a = ……… equation 4.33
10000* MIC
n
Emix = ∑ei pi' ≥ E s ……… equation 4.34
i =1
n
M mix = ∑mi pi' ≥ M s ……… equation4.35
i =1
n
S mix = ∑si pi' ≥ Ss ……… equation4.36
i =1
n
X mix = ∑xi pi'1 ≥ X s ……… equation4.37
i =1
1 1
= ∑ ≥ ……… equation4.38
Fmix fi
The obvious constraint here is that the sum of proportions is unity or 100 %.
n
∑p
i =1
'
i = 1……… equation 4.39
244 Appendix
∑c p
i =1
'
i
'
i ≥ C'mix ……… equation4.40
100
Cost of clean cottonC ' = C *( ) ……… equation4.41
100 − t
100 − t
Proportion ofclean cotton p ' = p *( ) ……… equation4.42
100
Proportion of bale cotton for the given component in the mix
Proportion of component in the mix in bale cotton form

= ……… equation 4.43
Total amount of cotton in mix in bale cotton form
Where, Proportion of Bale cotton for the given component in the mix P1 is
given by
100p1'
100 − t1
p1 = ………equation 4.44
100p1' 100p '2 100p3' 100p n'
+ + + .. +
100 − t1 100 − t 2 100 − t 3 100 − t n
Chapter 6
N ×W ×60
D ( minutes ) = …………..equation 6.1
P ×L
n
Mean ( x) = ∑ X i …… equation 4.1(i )
i =1
(X )
2
∑
n
−x
Standard Deviation (σ ) = i=0 i
…..equation 4.9
n
And
σ
Coefficient of Variation (CV ) % = ×100 ……..equation 4.10
x
Sample value1 − SampleValue2

Critical Difference ( % ) = × 100 …..equation 6.2
Average value of two samples
Appendix 245
And
Sample value1 − SampleValue2

t value = …..equation 6.3
Standard Error
L = W1 L1 + W2 L2 + W3 L3 + ….. + Wn Ln …….equation 6.4
L.S .m
FQI = ………..equation 4.14
f
F
Q = 29.4 [ ]2 ………….equation 4.13
L
Lea CSP = 320 ( )

Lea CSP = 320

 ( )

FQI + 1 − 13C  1 +
W 
  100 
; Combed yarn ……….equation 4.17
d −1
U 2 = Q.N e + a ( d − 1) + U r2 ………… equation 4.12
d
Actual yarnCSP
Process Proficiency ( P %) =
Theoratically Calculated CSP
¥100 ºººequation 4.20
Weigh of Equivalent Blending Delay = N × W ………..equation 6.5
Chapter 7
i=k
Minimize Z = ∑ai ci = a1c1 + a2 c2 + … ak ck ………..equation 7.1
i =1
i=k
Sum of Proportions = ∑ai = a1 + a2 + … ak = 1………..equation 7.2
i =1
i=k
Sum of Qulaity = ∑ai Qi = a1Q1 + a2 Q2 + … ak Qk ≶ Q* ………..equation 7.3
i =1
MinimizeC ' x ……… equation 7.4

246 Appendix
Subject to : Ax bli ≤ xi ≤ ui ……..equation 7.5
Chapter 8
y = ƒ ( Z ) = ƒ (Wx + b )…… equation 8.1
n
Z = ∑ X iWi + b = WX + b ……..equation 8.2
i =1
W = W1,1 ……..W1, n …………..equation 8.3
 nh 
y = ƒ 0  ∑OiWi 2 + b …… equation 8.4
 i =1 
n
Oi = (∑ X jWi1, j + bi ), ……..equation 8.5
j =1
∑ ∑ (t ) …….equation 8.6
P N
j=0 i=0 ij − yij
MSE =
NP
∂ ( MSE )
∆W ji ( n ) = −η ………… equation 8.7
∂W ji ( n )
YR −YP
η= ……..equation 8.8
YR
FR − FP
η= ……..equation 8.9
FR
Chapter 9
Actual value − Standard value

Premium / Discount ( % ) = ×100 …… equation 9.1
Standard value
Yarn produced per month ( Kg )

Yarn Realization ( % ) = ×100 ….equation 9.2
Bale cotton consumed per month ( Kg )
Appendix 247
Chapter 10
m
E = ∑E j ………… equation10.1
j =1
1 s
Ej = ∑(Tr − Or )2 …………equation10.2
2 r =1
∆W jk = -h ÈÎ∂E / ∂W jk ˘˚
= η ( Tk − Out k ) Out k ( 1 − Out k ) Out j ;
= ηδ k Out j ………………….equation10.3
∆W jk = η f ' (net j ) X i ∑δ kW jk ………… equation10.4
Absolute error
Normalized error =
Range
Actualvalue − Predictedvalue
=
Maximumvalue − Minimumvalue ……equation 10.5
Minimize Z = PA .65 + PB .52…….equation10.6
PA .160 + PB .126 ≥ 139 …..equation10.7
1 1 1
PA . + PB . ≥ …..equation10.8
4.01 4.35 4.19
Where,
PA + PB = 1, and PA ≥ 0, PB ≥ 0
msereg = γ ( mse ) + (1 − γ ) msw………..equation10.9
Where γ is the performance ratio and
1 n 2
msw = ∑W j ……….equation10. 10
n j =1
248 Appendix
Here n is the number of weighted connections.

p
E = ∑E j …………..equation10.11
j =1
Where E is the error vector, Ej is the error associated with the jth pattern and
P is the total number of training patterns. The expression of Ej is given in the
following equation.
1 s
Ej = ∑
2 K =1
(Tk − Outk ) ……… equation10.12
2
1
f (Z ) = ……… equation10.13
(1 + e )
z
Where, Z is the weighted sum of inputs to a neuron and f (Z) is the trans-
formed output from that neuron.
Abbrevations
AFIS Advanced Fiber Information System

AI Artificial Intelligence
ANN Artificial Neutral Network
ASTM American Standard Testing Method
ATIRA Ahmedabad Textile Research Association
BIAS Bale Inventory Analysis System
BNN Biological Neutral Network
CRE Constant Rate Of Extension
CRL Constant Rate Of Loading
CRT Constant Rate Of Traverse
CV Coefficient Of Variation
EDI Electronic data information
EFS Engineered Fiber Selection
EYC Electronic Yarn Clearer
FDA Fiber Diameter Analyzer
FF ANNs Feed Forward Artificial Neural Network
FFBP ANNs Feed Forward Back-Propagation Artificial Neural Network
FQI Fiber Quality Index
GDP Gross Domestic Product
GI Geometric Properties Index
HVI High Volume Instrument
MAHP Multiplicative Analytic Hierarchy Process
MCDM Multi Criteria Decision Making
MDTA Micro-dust Trash Analyser
250 Abbrevations
MFQI Modified Fiber Quality Index

MI Maturity Index
MIAHP Multiplicative Analytic Hierarchy Process
ML Mean Length
MLP Multi-Layer Perceptron
MSE Mean Square Error
OFDA Optical Fiber Diameter Analysis
PDI Premium Discount Index
PI Pressley Index
PPC Production Planning And Control
RGI Readymade Garment Industry
SCI Spinning Consistency Index
SFC% Short Fiber Content %
SITRA South India Textile Research Association
SL Span length
UHML Upper Half Mean Length
UI Uniformity Index
UQL Upper Quartile Length
UR Uniformity Ratio
WTO World Trade Organisation
Index
A Cotton 10
Critical difference 78, 80, 137
AFIS 37, 41-43, 57-58, 62, 66-69,
Cross blending 131
81, 86, 144-145, 155,
CSP 83-84
AI 95-96
ANN 95-103, 106-116, 173-174, D
176-178, 183, 191, 218-220,
Deviation 73
230-234
Digital Fibrograph 41
B Dispersion 39, 71, 73-74
Drafting quality index 81 , 139
+b value 60
Back propagation 173 E
Bale management 151-153, 155-
Effective length 36, 45
157, 160-163
EFS 151, 160, 163, 166, 204-208
BIAS 151, 159
Blending 14, 141 F
Blending delay 131-134, 141-143,
Fabric engineering 110-111, 113,
147-148
229
BNN 95
Fiber Fineness 23, 44, 45
Brain 95, 97
Fiber length 36
Breaking length 52-53
Fiber length uniformity 36
Buffer stock 133
Fiber strength 51
Bundle fiber strength 27, 53-55, 62,
Fibrograph 41, 43, 82, 139
139, 143
Filament 2, 5
C Fingering 9
FQI 81-82, 84, 138
Cabled yarn 7
f-test 79
Central tendency 71, 73
Functional unit 95
Cleaning efficiency 67
Fuzzy linear programming 92-93
Coefficient of Variation % 36, 39,
Fuzzy logic 98, 104, 111,
77, 79, 136
Colour grade 29, 60 G
Comb sorter technique 43
Geometric index 81, 88
252 Index
Gravimetric 45 Mix formulation 127, 128

Mix homogeneity 30
H
Mode 73
Hand knitting 9 Modelling 183
Hand stapling 41 Moisture 28
HVI 42-43, 144-145, 154-155, 165, Moisture content 58
204 Moisture regain 58
multi-layer Perceptron 101
I
N
Impact analysis 217
Intelligent fiber 107 Natural fibers 3
Inventory management 163 Neuron 95, 98
Iteration 98, 176 Neutral network 98-100, 102-105
L O
Learning algorithm 95, 99, 101 Over fitting 186-187, 217, 225
Length and length variations 21
P
Linear programming 89-92, 146,
167-170 PDI 81, 86, 210
Lint 56 Percentage mean range 76
Load 51 Plying 7-8
PPC (Production Planning
M
and Control) 114
MAHP 81 Prediction error 191
Manmade fiber 3 Pressley strength 53
Manufactured fiber 3 Process control 214
Mass Evenness 30 Process proficiency 83, 140
Maturity 25 Projection microscope 47
Maturity coefficient 49
Q
Maturity index 26
Maturity ratio 49 Q index 81
MCDM 81 Quality indices 71, 81-88
MCDM 87
R
MDTA 56
Mean 73, 136 Range 73
Mean length 36, 37, 43 Rd value 60
Mean square error 185 Reverse yarn engineering 175, 218
Median 73
S
MFQI 81, 84
MIAHP 88 SCI 69, 85, 87, 145-146
Micronaire 45 Shirley trash anlayser 56
Millitex 45 short fiber content % 36, 40
Mineral matter 28 Significance test 71, 78, 80
Mix formulation 17 Significance testing 78
Index 253
SITRA 84 U
Span length 36, 38-39
Uniformity index (ui %) 36, 39
Specific surface area 47
Upper half mean length 36, 37, 85
Spinning limit 23
Upper quartile length 36, 37
Stack mixing 132
Upper quartile mean length 36, 38
Standard deviation 136
Staple length 36 V
Strength 27
Validation test 188-189, 217
Stress 51
Variance 77
Synaptic weight 95, 98, 175, 184-
Vegetable matter 28
185, 230
Y
T
Yarn engineering 95
Tenacity 51, 53
Yarn realization 17, 33-35, 71, 173,
Tensile strength 53
176-177, 195-196, 208
Textile fiber classification 1-4
Transfer function 95, 220, 227, 230
Trash 27, 56
-test 79

(Woodhead Publishing India in Textiles) Agrawal, Sweety A. - Shaikh, Tasnim N - Engineering Cotton Yarns With Artificial Neural Networking (ANN) - Woodhead Publishing India PVT LTD (2017)

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Woodhead Publishing India in Textiles) Agrawal, Sweety A. - Shaikh, Tasnim N - Engineering Cotton Yarns With Artificial Neural Networking (ANN) - Woodhead Publishing India PVT LTD (2017)

Uploaded by

Copyright:

Available Formats

Engineering Cotton Yarns with Artificial Neural

Dr. (Mrs.) Tasnim N. Shaikh

WOODHEAD PUBLISHING INDIA PVT LTD

First published 2017, Woodhead Publishing India Pvt. Ltd.

Trademark notice: Product or corporate names may be trademarks or registered

Woodhead Publishing India Pvt. Ltd. ISBN: 978-9-38505-920-9

Typeset by Third EyeQ Technologies Pvt Ltd, New Delhi

3 Testing techniques used in yarn engineering 33

4.4.7 Geometric properties index (IG) 88

6 Changes in mix formulation approach with the

7.6.2 Types of linear programming 168

Cotton fiber is holding an unbeaten position in the apparel sector especially in

Cotton fiber is holding an unbeaten position in the apparel sector especially in

attributes of cotton mixing: testing techniques, correlation amongst fiber

Vadodara, Gujarat (India) Dr. T.N. Shaikh

Dr. (Mrs.) Tasnim N. Shaikh

Department of Textile Engineering, Faculty of Technology & Engineering, The Maharaja

1.2 Types of textile fibers

Different types of textile fibers.

Conversely, high performance fibers are driven by special technical

1.3 Types of textile yarn

Figure 1.2 Basic steps of spun yarn production.

Figure 1.3 Types of yarn spinning.

Figure 1.4 Different types of ring spun.

1. Knitted worsted: The four-ply all-around yarn is used for accessories,

1.4 Significance of cotton yarn in textile industry

1.5 Production pattern in brief for most significant

Figure 1.5 Cotton textile production sequence for apparel.

The manufacturing process commences with the arrival of raw materials

Basic tasks of cotton spinning.

(Figs. 1.2–1.6). As per the quality requirements, sliver is subjected to the

Spinning starts with drawing process mainly consists of doubling and

1.6 Impact of natural fiber variations on yarn

spinner because the consistency of yarn quality depends on the consistency of

1. Banerjee P.K., Gupta Kaushal K., et al., 1991, Publication of Two

2. Gowayed Yasser, 1995, Theory and practices of cotton fiber selection.

Dr. (Mrs.) Tasnim N. Shaikh

Department of Textile Engineering, Faculty of Technology & Engineering, The Maharaja

2.1 Need for mix formulation

in processing fiber material of first lot is compared with the difficulties,

1. Number of fiber per cross section should be uniform throughout the

2.2 Interrelationship between fiber characteristics

characteristics for a particular yarn property varies as per type of spinning

Fiber properties % Contribution

Fiber parameters Yarn parameters

2.3 Contribution of fiber parameters on ring spun yarn

1. Length & length variations

2.3.1 Length & length variations

Count range (Ne) Fiber length Micronaire Strength Maturity index

However, length of cotton is a variable quantity as in natural fibers vary

yarn realization, lowers down machine efficiency with more tendency of

Fiber fineness is one amongst three important fiber quality characteristic

2.3.2.1 Spinning limit

2.3.2.2 Performance of product

2.3.2.3 Performance of process

Fiber maturity is an important character of cotton and is an index of

Figure 2.1 Fiber maturity.

Cotton fiber is hygroscopic in nature, thereby absorbs moisture easily from

sampling cotton bales as well as smoothening trade resistance. Excessive

2.4 Importance of mix homogeneity

(1) Radial homogeneity: Uniform arrangement of fibers in the cross

2.5 Impact of technological changes on homogeneity