Application of Soft Computing Techniques for Cell Formation Considering Operational Time and Sequence

By: Sudhakarapandian, RContributor(s): Mohapatra, Siba Sankar [Supervisor] | Department of Mechanical EngineeringMaterial type: TextTextLanguage: English Publisher: 2007Description: 186 pSubject(s): Engineering and Technology | Mechanical Engineering | Mechanical EngineeringOnline resources: Click here to access online Dissertation note: Thesis (Ph.D)- National Institute of Technology, Rourkela Summary: In response to demand in market place, discrete manufact uring firms need to adopt batch type manufacturing for incorporating con tinuous and rapid changes in manufacturing to gain edge over competitors. In addition, there is an increasing trend toward achieving higher level of integ ration between design and manufacturing functions in industries to make batch manuf acturing more efficient and productive. In batch shop production environment, the cost of manufacturing is inversely proportional to batch size and the batch si ze determines the productivity. In real time environment, the batch size of the components is often small leading to frequent changeovers, larger machine i dleness and so lesser productivity. To alleviate these problems, “Cellular M anufacturing Systems” (CMS) can be implemented to accommodate small batches wi thout loosing much of production run time. Cellular manufacturing is an application of group technology (GT) in which similar parts are identified a nd grouped together to take advantage of their similarities in design and productio n. Similar parts are arranged into part families and each part family proce sses similar design and manufacturing characteristics. Cellular manufacturing i s a good example of mixed model production and needs to resolve two tasks wh ile implementing cellular manufacturing. The first task is to identify the part families and the next task is to cluster the production machines into machine cell s known as cell formation (CF). GT ideas were first systematically prese nted by Burbidge following the pioneering work of Mitrofanov in U.S.S .R. Burbidge developed the concept of production flow analysis and successfully implem ented in industries. After this, many countries started following GT concepts in their manufacturing lines. Researchers initiated to develop various methods l ike similarity coefficient method, graph theoretic approaches and array based meth ods in this field. In this trend, modeling of CMS through mathematical programm ing was started to incorporate more real life constraints on the problem. Later researchers started developing heuristics and meta-heuristics to explore the best optimal solutions for the CF problems. Since soft computing techniques nowa days expand their applications to various fields like telecommunications, net working, design and ii manufacturing, current research in CMS is being carried out using soft computing techniques. As for as representation of the cell formation problem is concerned, most of the researchers use zero-one binary machine part inci dence matrix (MPIM) that is obtained from the route sheet of the manufact uring flow shop. The 1’s in the binary matrix represent the visit of the parts to the corresponding machines and 0’s represent the non-visit. The final output is a block diagonal structure from which the part families and corresponding machine cells w here the part families are to be manufactured can be identified. In such an in put representation, the process of clustering machines into machine cells and parts in to part families is done without using real life information which may le ad to inferior manufacturing plans. Therefore, there is a need to make use of as m any as real life production information in the input matrix for representing th e CF problem. In this research work, the real life production factors like, operational time of the parts in the machines known as workload data or ra tio level data, operational sequence of the parts known as ordinal le vel data and batch size are considered for the problem representation. The methodo logy uses soft computing techniques like genetic algorithm (GA) and n eural network to tackle the CF problem. In recent years, soft computing techniq ues have fascinated scientists and engineers all over the world because such te chniques possess the ability to learn and recall as similar to the main fun ctions of the human brain. They find better approaches to real world problems since soft computing incorporates human knowledge effectively. It deals with i mprecision and uncertainty and learn to adapt to unknown or changing environment for better performance. In neural network, adaptive resonance the ory (ART1) gives good results for binary MPIM CF problem. ART1 is not suitabl e for non-binary input pattern. Hence, in this work, suitable modification i s included in the basic ART1 to incorporate the operational time of the parts, a r atio level non-binary data. For dealing with sequence of operations of the parts, an or dinal level non-binary data, a supplementary procedure is first implemented to convert the non-binary data into a suitable binary data and subsequently by feeding to the basic ART1 networks to solve the CF problem. Finally both operati onal time and operational iii sequence are combined and represented in a single matrix . The modified ART1 used for solving CF problem with operational time is a pplied to solve the problem with combination of operational time and sequence. T he CF problem without any objective function is solved effectively by ART1 appro ach. For solving the CF problem with objective functions like total cell load variation (CLV) and exceptional elements, GA is propose d in this research work. CLV is calculated as the difference between the workload on the machine and the average load on the cell. Exceptional elements are the number of non-zero elements present in off diagonal blocks of the output m atrix. Both the objective functions are combined to get a multi objective CF prob lem and solved by using GA. In the past, several performance measures like group ing efficiency and grouping efficacy have been proposed to find out the g oodness of the output clusters. But most of them are applicable only for binar y data representation. In this research work, suitable performance measures are propo sed to measure the goodness of the block diagonal structure of the output ma trix with ratio level data, ordinal level data and combination of both data. The algorithms are designed to handle problem of any size and they are coded with C ++ and run on Pentium IV PC. Computational experience with the proposed techniq ues is presented and the results are compared with the problems available in open literature. The results are encouraging and the methodologies are found more appropriate for large scale production industries. Computational results suggest that the proposed approaches are reliable and efficient both in terms of quality and in speed in solving CF problems. Several directions for fut ure studies are also addressed in this research.
Tags from this library: No tags from this library for this title. Log in to add tags.
    Average rating: 0.0 (0 votes)
Item type Current location Collection Call number Status Date due Barcode
Thesis (Ph.D/M.Tech R) Thesis (Ph.D/M.Tech R) BP Central Library
Thesis Section
Reference Not for loan T28

Thesis (Ph.D)- National Institute of Technology, Rourkela

In response to demand in market place, discrete manufact
uring firms need
to adopt batch type manufacturing for incorporating con
tinuous and rapid
changes in manufacturing to gain edge over competitors.
In addition, there is an
increasing trend toward achieving higher level of integ
ration between design and
manufacturing functions in industries to make batch manuf
acturing more efficient
and productive. In batch shop production environment,
the cost of manufacturing
is inversely proportional to batch size and the batch si
ze determines the
productivity. In real time environment, the batch size
of the components is often
small leading to frequent changeovers, larger machine i
dleness and so lesser
productivity. To alleviate these problems, “Cellular M
anufacturing Systems”
(CMS) can be implemented to accommodate small batches wi
thout loosing much
of production run time. Cellular manufacturing is an
application of group
technology (GT) in which similar parts are identified a
nd grouped together to take
advantage of their similarities in design and productio
n. Similar parts are
arranged into part families and each part family proce
sses similar design and
manufacturing characteristics. Cellular manufacturing i
s a good example of
mixed model production and needs to resolve two tasks wh
ile implementing
cellular manufacturing. The first task is to identify the
part families and the next
task is to cluster the production machines into machine cell
s known as cell
formation (CF). GT ideas were first systematically prese
nted by Burbidge
following the pioneering work of Mitrofanov in U.S.S
.R. Burbidge developed the
concept of production flow analysis and successfully implem
ented in industries.
After this, many countries started following GT concepts
in their manufacturing
lines. Researchers initiated to develop various methods l
ike similarity coefficient
method, graph theoretic approaches and array based meth
ods in this field. In this
trend, modeling of CMS through mathematical programm
ing was started to
incorporate more real life constraints on the problem.
Later researchers started
developing heuristics and meta-heuristics to explore the
best optimal solutions
for the CF problems. Since soft computing techniques nowa
days expand their
applications to various fields like telecommunications, net
working, design and
ii
manufacturing, current research in CMS is being carried
out using soft computing
techniques.
As for as representation of the cell formation problem
is concerned, most
of the researchers use zero-one binary machine part inci
dence matrix (MPIM)
that is obtained from the route sheet of the manufact
uring flow shop. The 1’s in
the binary matrix represent the visit of the parts to
the corresponding machines
and 0’s represent the non-visit. The final output is a
block diagonal structure from
which the part families and corresponding machine cells w
here the part families
are to be manufactured can be identified. In such an in
put representation, the
process of clustering machines into machine cells and parts in
to part families is
done without using real life information which may le
ad to inferior manufacturing
plans. Therefore, there is a need to make use of as m
any as real life production
information in the input matrix for representing th
e CF problem.
In this research work, the real life production factors
like, operational time
of the parts in the machines known as workload data or ra
tio level data,
operational sequence of the parts known as ordinal le
vel data and batch size are
considered for the problem representation. The methodo
logy uses soft
computing techniques like genetic algorithm (GA) and n
eural network to tackle
the CF problem. In recent years, soft computing techniq
ues have fascinated
scientists and engineers all over the world because such te
chniques possess the
ability to learn and recall as similar to the main fun
ctions of the human brain.
They find better approaches to real world problems since
soft computing
incorporates human knowledge effectively. It deals with i
mprecision and
uncertainty and learn to adapt to unknown or changing
environment for better
performance. In neural network, adaptive resonance the
ory (ART1) gives good
results for binary MPIM CF problem. ART1 is not suitabl
e for non-binary input
pattern. Hence, in this work, suitable modification i
s included in the basic ART1
to incorporate the operational time of the parts, a r
atio level non-binary data. For
dealing with sequence of operations of the parts, an or
dinal level non-binary
data, a supplementary procedure is first implemented to
convert the non-binary
data into a suitable binary data and subsequently by
feeding to the basic ART1
networks to solve the CF problem. Finally both operati
onal time and operational
iii
sequence are combined and represented in a single matrix
. The modified ART1
used for solving CF problem with operational time is a
pplied to solve the problem
with combination of operational time and sequence. T
he CF problem without any
objective function is solved effectively by ART1 appro
ach.
For solving the CF problem with objective functions like
total cell load
variation (CLV) and exceptional elements, GA is propose
d in this research work.
CLV is calculated as the difference between the workload
on the machine and
the average load on the cell. Exceptional elements are
the number of non-zero
elements present in off diagonal blocks of the output m
atrix. Both the objective
functions are combined to get a multi objective CF prob
lem and solved by using
GA. In the past, several performance measures like group
ing efficiency and
grouping efficacy have been proposed to find out the g
oodness of the output
clusters. But most of them are applicable only for binar
y data representation. In
this research work, suitable performance measures are propo
sed to measure the
goodness of the block diagonal structure of the output ma
trix with ratio level data,
ordinal level data and combination of both data. The
algorithms are designed to
handle problem of any size and they are coded with C
++
and run on Pentium IV
PC. Computational experience with the proposed techniq
ues is presented and
the results are compared with the problems available in
open literature. The
results are encouraging and the methodologies are found
more appropriate for
large scale production industries. Computational results
suggest that the
proposed approaches are reliable and efficient both
in terms of quality and in
speed in solving CF problems. Several directions for fut
ure studies are also
addressed in this research.

There are no comments on this title.

to post a comment.

Implemented and Maintained by Biju Patnaik Central Library.
For any Suggestions/Query Contact to library or Email: library@nitrkl.ac.in OR bpcl-cir@nitrkl.ac.in. Ph:91+6612462103
Website/OPAC best viewed in Mozilla Browser in 1366X768 Resolution.

Powered by Koha