See discussions, stats, and author profiles for this publication at: [Link]

net/publication/305670368

Modified Rank Order Clustering Algorithm Approach by Including

Manufacturing Data

Conference Paper · December 2016

DOI: 10.1016/[Link].2016.07.103

CITATIONS READS

9 795

2 authors:

Nagdev Amruthnath Tarun Gupta

Western Michigan University Western Michigan University
14 PUBLICATIONS 21 CITATIONS 33 PUBLICATIONS 693 CITATIONS

SEE PROFILE SEE PROFILE

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

Wireless Sensor Network for Predictive Maintenance View project

Concept paper View project

All content following this page was uploaded by Nagdev Amruthnath on 25 October 2016.

The user has requested enhancement of the downloaded file.

4th IFAC International Conference on
Intelligent
4th
4th Control andConference
IFAC International Automationon
Sciences
4th IFAC
IFAC International
International Conference
Conference on
onAvailable online at [Link]
June 1-3,
Intelligent 2016. Reims,
Control
Intelligent Control and
Control and France
Automation
Automation Sciences
and Automation Sciences
Intelligent Sciences
June
June 1-3, 2016. Reims, France
June 1-3, 2016. Reims, France
1-3, 2016. Reims, France
ScienceDirect
IFAC-PapersOnLine 49-5 (2016) 138–142
Modified Rank Order Clustering Algorithm Approach by Including Manufacturing
Modified
Modified Rank
Rank Order
Order Clustering
Clustering Algorithm
Data Approach
Algorithm Approach by
by Including
Including Manufacturing
Manufacturing
Data
Data
Nagdev Amruthnath Tarun Gupta
Nagdev
Nagdev Amruthnath
Nagdev Amruthnath
Amruthnath Tarun
Tarun Gupta
Tarun Gupta
Gupta
IEEEM Department, Western Michigan University, MI 49009 USA (e-mail: [Link]@[Link])
IEEEM
IEEEM
IEEEM Department,
IEEEM
Department,
Department, Western
Department,
Western
Western Michigan
Western
Michigan
Michigan University,
Michigan
University,
University, MI
University,
MI 49009
MI 49009
49009 USA
MI 49009
USA (e-mail:
USAUSA
(e-mail:
(e-mail: [Link]@[Link])
(e-mail: [Link]@[Link])
[Link]@[Link])
[Link]@[Link])
IEEEM
IEEEM Department, Western Michigan University, MI 49009 USA (e-mail: [Link]@[Link])
IEEEM Department, Western Michigan University, MI 49009 USA (e-mail: [Link]@[Link])
Department, Western Michigan University, MI 49009 USA (e-mail: [Link]@[Link])
Abstract: A modified rank order clustering (MROC) method based on weight and data reorganization
Abstract:
Abstract:
Abstract: A
A modified
has been developed
A modified
modified rank
rank order
to facilitate
rank theclustering
order
order clustering
clustering (MROC)
needs of real worldmethod
(MROC)
(MROC) method
method based
based on
manufacturing
based on
on weight
weight and
environment.
weight and data
data reorganization
and MROC
data is designed
reorganization
reorganization
has
has been developed to facilitate the needs of real world manufacturing environment. MROC is designed
to
has been
been developed
optimize the
developed to
to facilitate
manufacturing
facilitate the
the needs
process
needs of
of real
based
realonworld manufacturing
important
world environment.
independent
manufacturing variables
environment. MROC
with
MROC is
weights
is and
designed
designed
to
to
to optimize
optimizethethe
reorganize
optimize the manufacturing
manufacturing process
themachine-component
manufacturing process
process based
data that
based
based on
helps
on important
onform
important
important independent
cells where each cellvariables
independent
independent variables
variables with
would have weights
weights and
withapproximately
with weights and
and
reorganize
the same work
reorganize
reorganize the
the machine-component
the load. The developed data
machine-component
machine-component data
data that
that helps
algorithm
that usingform
helps
helps form
form cells
cells where
a heuristics
cells where
where each
each cell
minimizes
each cell
cell would
wouldofhave
number
would have
have approximately
bottlenecks for the
approximately
approximately
the
the same
cellular
the same work
solution
same work load.
work load. The
without
load. The developed
human
The developed algorithm
input
developed algorithm using
(necessary
algorithm using in a heuristics
King minimizes
{1980)),
using aa heuristics while
heuristics minimizes number
ensuring
minimizes number
number ofof bottlenecks
comparable
of bottlenecks for
for the
machine
bottlenecks for the
the
cellular
cellular solution
utilizations in
cellular solution without
each cell.
solution without human
This
without human input
paper
human input (necessary
describes our
input (necessary
(necessary inin King
proposed
in King {1980)), while
algorithm
King {1980)), and
{1980)), while ensuring
a solution
while ensuring comparable
to the
ensuring comparable machine
machine
comparable machine cell
machine
utilizations
design process
utilizations
utilizations in
in each
in each
each cell.
for the
cell.
cell. This
realThis
This paper
world
paper
paper describes
manufacturing
describes our proposed
environment.
describes our
our proposed algorithm
proposed algorithm and
and aaa solution
algorithm and solution to
solution to the
to the machine
the machine cell
machine cell
cell
design
design process
process for
for the
the real
real world
world manufacturing
manufacturing environment.
environment.
© 2016, IFAC
design process
Keywords: Rank(International
for the real Federation
world
order clustering, of Automatic
manufacturing
Production Control) Hosting
environment.
Flow Analysis, by Elsevier Ltd. All rights reserved.
Data Reorganization
Keywords:
Keywords: Rank
Rank order
order clustering,
clustering, Production
Production Flow
Flow Analysis,
Analysis, Data
Data Reorganization
Reorganization
Keywords: Rank order clustering, Production Flow Analysis, Data Reorganization
McCauley (1972) followed by McCormick et. al. (1976).
1. INTRODUCTION McCauley
1. The SCM is(1972)
McCauley
McCauley basedfollowed
(1972)
(1972) on establishing
followed
followed by
by McCormick
by McCormick
McCormick et.
et. al.
similarity coefficient
et. al. (1976).
al. for
(1976).
(1976).
Over fifty years rank-order 1. INTRODUCTION
1. INTRODUCTION
INTRODUCTION
clustering (ROC) algorithm has The
each
The SCM
pair is
of based
machines on establishing
derived from similarity
the same coefficient
m x n for
(m-
The SCM is based on establishing similarity coefficient for
SCM is based on establishing similarity coefficient for
Over
matured fifty years
with its rank-order
implementation clustering (ROC)
in many algorithm has each pair
machines &of machines
n-parts) derived
machine-component from the same
matrix m
also x n
used (m-
in
Over
Over fifty years rank-order clustering
fifty years rank-order clustering (ROC)domains.
(ROC) algorithmKing
algorithm has
has each pair of machines derived
each pair of machines derived from the same m x n (m- from the same m x n (m-
matured
(1980)
matured with
first
with its
introduced
its implementation
the
implementationnotion in
of
in many
ROC
many indomains.
the
[Link] King
Kingof machines
machines & n-parts) machine-component matrix also used in
ROC
machinesand &
& n-parts)
MODROC
n-parts) machine-component
methods.
machine-component These matrix
similarity
matrix also
also used
coefficients
used in
in
matured with its implementation in many domains. King
(1980) ROC and MODROC methods. These similarity coefficients
(1980) first introduced the notion of ROC in the domain of
(1980) first
manufacturing
first introduced
for
introduced the
clustering
the notion of
machines
notion of ROC
ROC in
into
in the domain
machine
the domain cells
of
of are
ROC arranged
and in
MODROC m x n similarity
methods.
ROC and MODROC methods. These similarity coefficients matrix
These with
similarity one half of
coefficientsthe
manufacturing
simultaneous
manufacturing for
formation
for clustering
clusteringof machines
part families
machines into
that
into machine
need
machine to cells
be
cells are
are arranged
matrix
are be thein
arranged
arranged in m
mx
inmirror
m xn
x n similarity
nimage
similarity
similarityof the matrix
otherwith
matrix
matrix [Link]
with
with one
one half
half of
Guided
half by the
of
of an
the
the
manufacturing for clustering machines into machine cells
simultaneous to oneformation of matrix be the mirror image of the other half. Guided
Guided by an
assigned
simultaneous
simultaneous of the machine
formation
formation of part
of part
part families
clusters.
families
families that
Ideally
that need
that each
needpart
need to
to and
to be
be
be overall
matrix
matrix be measure
be the the mirrorof similarity
mirror image
image of among
of the machines
the other
other half. in each
half. Guided cell
by the
by an
an
assigned
its part
assigned to
familyone of
has the a machine
unique clusters.
assignment Ideally
when each
it part
comes and
to overall
clustering
overall measure
overall measure process
measure of of similarity
continues
of similarity
similarity among among
until all machines
among machinesmachines
machines in in each
cells
in each cell
converge
each cell the
cell the
assigned to to one
one ofof the
the machine
machine clusters.
clusters. Ideally
Ideally each
each partpart and and the
its part family has a unique assignment when it comes clustering process continueswhichuntil all
all machines cells converge
assigning
its
its part to
part family one
family has and only
has aa unique one cell
unique assignment while
assignment when getting
when it comes to
completely
it comes to
to into one large
clustering
clustering process
process cluster
continues
continues until
untilalmost
all always we
machines
machines cellswould
cells converge
convergenot
assigning
processed
assigning to one
inside
to one and
the
and only
assigned
only one
one cell
machine
cell while
cell.
while getting
A completely
machine
getting completely cell is into
want
into one
for large
obvious cluster which
reasons. The almost
into one large cluster which almost always we would not
one large cluster which almost always
solution
always to we
the
we would
would not
clustering
not
assigning to one and only one cell while getting completely
processed
constituted
processed inside the
of all thethoseassigned
machines machine
that arecell. A
part machine
of the machine cell is want
process
want for
for is obvious
identified
obvious reasons.
at
reasons.a The
carefully
The solution
selected
solution to
to the
threshold
the clustering
value
clustering
processed insideinside the assigned
assigned machine
machine cell.
cell. A
A machine
machine cell
cell isis want for obvious reasons. The solution to the clustering
constituted
cell. However,
constituted of
of all
all those
this
those machines
historic
machines ROC that
that are part
algorithmic
are part of
of the
method
the machine
machine has process
that varies
process
process is
is identified
is between 0at
identified
identified atand
at aaa carefully
1 (same selected
carefully
carefully as similarity
selected
selected threshold
threshold
threshold value
coefficient
value
value
constituted of all those machines that are part of the machine
cell. that varies between 0 and 1 (same as similarity coefficient
cell. However,
since
cell. been applied
However,
However, this
this historic
this in various
historic
historic ROC
ROC
ROC algorithmic
algorithmic method
domains–image
algorithmic processing
method
method has
has
has value
that for
varies a pair
between of machines).
0 and 1 Often
(same
that varies between 0 and 1 (same as similarity coefficient the
as selection
similarity of suitable
coefficient
since
(Patel
since been
&
been applied
Stonhem
applied in
1992),
in various
Bromley
various domains–image
(1966)
domains–image Jiang et processing
al. (2004)
processing value
value for
value for aaavalue
threshold
for pair
pair of
of machines).
pair waits
of machines).
machines). Often
Often the
until the similarity
Often the selection
the matrix
selection
selection of
of suitable
is established
of suitable
suitable
since been applied in various domains–image processing
(Patel & threshold value waits until the similarity matrix is established
to name
(Patel
(Patel & aStonhem
& Stonhem
Stonhem 1992),
1992), Bromley
few. Chandrasekheran
1992), Bromley& (1966)
Bromley Rajgopalan
(1966)
(1966) Jiang
Jiang et
et al.
Jiang (1986)
et al. (2004)
al. found
(2004)
(2004) or machine
threshold
threshold value cluster
value waits formation
waits until
until the process
the similarity has
similarity matrix begun.
matrix is Needless
is established
established to
to name a few. Chandrasekheran & Rajgopalan (1986) or machine cluster formation process has begun. Needless
Needless to
limitations
to
to name
name aa few. with ROC presented
few. Chandrasekheran
Chandrasekheran & originally
& Rajgopalan in King(1986) found
Rajgopalan (1986) (1980) and
found
found say
or decisions
or machine
machine cluster regarding
cluster formation cell
formation process size
process hasand number
has begun. of
begun. Needless cells is
to
to
limitations with ROC presented originally in King (1980) say decisions regarding cell size and number of cells
cells is
proposed
limitations
limitations with a MODROC.
with ROCROC presented Regardless
presented originally
originally inwhether
in King (1980) and
ROC
King (1980) or
and
and required
say
say decisions to be made
decisions regarding with
regarding cell experts’
cell size input
size and and
and number that
number of too without
of cells is
is
proposed a MODROC. Regardless whether ROC required to be made with experts’ input and that too without
MODROC
proposed is
proposed aa MODROC. used to design
MODROC. Regardless cellular manufacturing
Regardless whether
whether ROC ROC or
systems or
or any notion
required of cell performance. Often
required to be made with experts’ input and that too without
to be made with experts’ input the
and only
that performance
too without
MODROC
there
MODROC are is used
several to design
decisions cellular
that have manufacturing
needed human systemsinput any
any notion
criteria
any notion thatof of cell
ofiscell performance.
usedperformance.
is number ofOften Often the
intercellular
the only performance
only moves
the only that a
performance
MODROC is used to design cellular manufacturing systems
is used to design cellular manufacturing systems notion cell performance. Often performance
there are
requiring
there several decisions
expert’sdecisionsknowledge that have needed
and needed human
thus elements input
of criteria
criteria that is used is number of intercellular moves that
particular
criteria that
that is
solution
is used
used is
will
is number
result
number of
into;
of intercellular
this holds
intercellular good moves
movesas thatasaaa
long
that
there are
are several
several decisions that
that have
have needed human
human input input
requiring particular solution will result into; this holds good as long as
requiring expert’s knowledge and thus elements of
subjectivity
requiring expert’s
remains
expert’s knowledge
in the
knowledgeprocess and
of
and thus
arriving
thus elements
at a
elements feasible of
of product
particularmix
particular as well
solution
solution willasresult
will process
result into;routing
into; this
this holdsinformation
holds good
good as as did
longnot
long as
as
subjectivity
cellular
subjectivity remains
solution.
remains in
Some
in the
theofprocess
these
process of arriving
decisions
of arriving at
are,
at a
a feasible
desired
feasible product
change.
product
product mix mix
Moreover,
mix as
as well
wellthe
as well as
as process
as standard
process routing
binary matrix
process routing
routing information
information
information did
(or machine-
did not
did not
not
subjectivity remains in the process of arriving at a feasible
cellular change.
change. Moreover, the standard binary matrix (or machine-
cellular solution.
maximum
cellular cell size,Some
solution.
solution. Some
Somemaximumof
of these
of number
these
these decisions
of parts
decisions
decisions are,
are,in desired
are, a part
desired
desired component
change. matrix) did
Moreover,
Moreover, the not
the standard
standardaccount for neither
binary
binary matrix the
matrix (or frequency
(or machine-
machine-
maximum cell size, maximum number of parts component matrix) did not account for neither the frequency
family;
maximum
maximum celland also
cell size,selecting
size, maximum one
maximum numberof several
number of parts in
solutions
of parts infrom
in aaa part
the
part
part of visits
component to
component matrix) a specific
matrix) did did not machine
not account
account fornor the
for neither order.
neither the Gupta
the frequency
frequency &
family; and also selecting one of several solutions from of visits to a specific machine nor the order. Gupta
Gupta &
final
family;matrix
family; and and alsoof the iterative
also selecting
selecting one process.
one of of severalAs mentioned
several solutions
solutions from from the
earlier,the
the Seifoddini
of
of visits
visits to (1990)
to aa specificdeveloped
specific machine a
machine nor more comprehensive
nor the
the order.
order. Gupta SCM to
&
&
final
ROC matrix
is one of the
severaliterative process.
approaches to As mentioned
machine earlier, Seifoddini
incorporate (1990) developed a
Seifoddini (1990) developed a more comprehensive also
part routing sheet more comprehensive
information and SCM
SCMthe to
final
final matrix
matrix of of the
the iterative
iterative process.
process. As mentionedand
As mentionedcell part
earlier,
earlier, Seifoddini (1990) developed a more comprehensive SCM to
to
ROC
family is one of several approaches to machine cell and part incorporate
projected
incorporateproductionpart routing volume sheet information
sheet information
information toand and
andhelp also the
alsoassist
ROC
ROC is isformation
one
one of for cellular
of several
several manufacturing.
approaches
approaches to
to machine
machine Burbidge
cell
cell andand (1963,part
part incorporate part
part routing
routing sheet information also the
the
family
1975)
family formation
developed
formation for cellular
machine
for cellular manufacturing.
cell formation
manufacturing. Burbidge
methodology
Burbidge (1963,
based
(1963, projected
estimating
projected production
true work
production volume
load
volume on
projected production volume information to help assist information
each of
information the to
machine
to help
help assist
cell to
assist
family formation for cellular manufacturing. Burbidge (1963,
1975) developed machine cell estimating
estimating true work load on each of the machine cell to
true work load on each of the machine cell
on
1975)
1975) production
developedflow
developed machine
machine cell formation
analysis
cell (PFA). As
formation
formation methodology
interest among
methodology
methodology based
based
based provide
estimating a measure
true work of cell
load performance
on each of thefor manufacturing
machine cell to
to
on
on production
manufacturing
production flow
systems
flow analysis
research
analysis (PFA).
community
(PFA). As
As interest
grew
interest among
due
among to provide
system
provide aaa measure
provide reporting measure
measure of
of cell
needs.
of cell
cellYet,performance
the decisionfor
performance
performance for manufacturing
for
for forming cells
manufacturing
manufacturing
on production flow analysis (PFA). As interest among
manufacturing systems research community grew systemstillreporting
based onneeds. Yet, the
the decision
about thefor forming cells
rapidly changing
manufacturing
manufacturing industrial
systems
systems environment
research
research community
community grew due
primarily
grew due to
due to
to was
system
system reporting
reporting subjective
needs.
needs. Yet,input
Yet, the decision
decision threshold
for
for formingvalue,
forming cells
cells
rapidly was still based on subjective input about the threshold value,
rapidly changing
transition
rapidly from mass
changing
changing industrial
production
industrial
industrial environment
to small volume
environment
environment primarily
large due
primarily
primarily variety
due
due to
to
to cell
was size
was still and
still based
based on number of
on subjective cells
subjective input in the
input about desired
about the solution.
the threshold
threshold value, This
value,
transition cell size and number of cells in the desired solution.
solution. This
transition from
production,
transition from
from mass
massofproduction
more
mass cellular to
production
production to
to small
small volume
manufacturing
small volume was
volume large
large variety
largegaining
variety
variety leads
cell us
cell size
size and to infer
and number that
number of as
of cellsfar
cells in as machine
in the
the desired cell formation
desired solution. This
Thisis
production,
ground.
production, One more
of the of cellular
simpler methodmanufacturing
for forming was
cells gaining
was the leads
leads us
concerned to infer
many that
decisions as far
in as
the machine
process
leads us to infer that as far as machine cell formation is
us to infer that as far as machine cell
still
cell formation
remain
formationfor is
the
production, more more of of cellular
cellular manufacturing
manufacturing was was gaining
gaining is
ground.
result
ground. of One of
McCauley’sthe simpler method
Similarity for forming
Coefficient cells
Method was
(SCM), the concerned
experts
concerned to many
determine
many decisions
after
decisions thein
in the
adopted
the process
machine
process still
still remain
cell
remain for
formation
for the
the
ground. One of the simpler method for forming cells was the
One of the simpler method for forming cells was the concerned many decisions in the process still remain for the
result
result of
of McCauley’s
McCauley’s Similarity
Similarity Coefficient
Coefficient Method
Method (SCM),
(SCM), experts
experts to
experts to determine
to determine after
determine after the
after the adopted
the adopted machine
adopted machine cell
machine cell formation
cell formation
formation
result of McCauley’s Similarity Coefficient Method (SCM),
2405-8963 ©
Copyright © 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
2016 IFAC
Peer review©under
Copyright 2016 responsibility
IFAC of International Federation of Automatic Control.
Copyright
Copyright © 2016
© 2016 IFAC
IFAC
10.1016/[Link].2016.07.103
IFAC ICONS 2016
June 1-3, 2016. Reims, France Nagdev Amruthnath et al. / IFAC-PapersOnLine 49-5 (2016) 138–142 139

procedure will present a solution, typically more than one; machines behave exactly same. In this context, the value of
and again the decision maker is to deploy his/her expertise to the product refers to cycle time of the product or monthly
choose one that best meets the objectives. volume of the product likewise the value of the machine
refers to cycle time of the machine, reliability of the machine
The objective behind weight-based rank order clustering
and setup time of the machine. These values play an
(ROC) algorithm is to create workload balanced machine
important role while creating cells. In our research we found
cells and associated part numbers. Traditional ROC algorithm
that, in ROC method there is no consideration of real time
forms machine cells purely based on machine-component
data of either machines or part numbers such as cycle time,
matrix solely. Our algorithm generates machine cell solution
volume, set up time considered in creating the cells.
with balanced workload identified with similar weight ratio.
This task is accomplished by assigning weight to part
numbers and machines and then rearranging the data based 3. LITERATURE REVIEW
on these weights by concepts of Bond Energy algorithm. The There are several other approaches to clustering and machine
concept of weights will be discussed in detail in later cell formation. One commonly used approach is Similarity
sections. ROC algorithm is performed on this reorganized Coefficient Method (SCM). SCM is one of the methods used
data. The cells are formed when no more iterations can be to form the machine cells in group technology applications.
conducted satisfying the stopping rule. Compared to the other methods, SCM incorporates more
In the next section of this paper we define rank order flexibility into the machine-component grouping process and
clustering algorithm with its pros and cons. The literature more easily lends itself to the computer application,
review in support of this research is presented in section III. Rajagopalan & Batra (1975). The new model improves the
In section IV and V we present our weight based rank order existing models based on SCM by dealing with the
clustering algorithm and the results obtained. In section IV duplication of bottleneck machines and by employing special
we present our weight and data reorganization approach with data storage and analysis techniques which greatly simplify
modified rank ordering clustering algorithm. An analysis the machine-component grouping process, Krishnanada &
with important results is included in section V. We have also Chincholkar (2004). The duplication process in the new
performed sensitivity analysis to verify the results and model is based on the number of inter-cellular moves.
analyse the robustness of our new approach that is included Duplication starts with the machine generating the largest
in section VI. number of inter-cellular moves and continues until no
machine generates more inter-cellular moves than specified
by a threshold value. By changing the threshold value,
2. RANK ORDER CLUSTERING ALGORITHM alternative solutions can be examined. The new model
employs the bit-level data storage technique to reduce the
Rank order clustering algorithm is also called as production storage and computational requirements of the machine-
flow algorithm is used to create cells to accommodate part component grouping process.
numbers to specific machines. Although in manufacturing, Manufacturing has always been an area where having a
machines are capable of running different part numbers, it is competitive edge in the market has a strong foot in the
important to route them to create a specific flow of part market. To have this competitive edge it’s important to have
numbers through assigned machines. This also improves low product cost, on time delivery and quality of the product.
productivity and eliminates cross line flow. Rank order To achieve this competitive edge it is important to design a
clustering algorithm functions as follows reliable, lower lead time and cost effective manufacturing
process. One such process is directing right products through
2.1 Algorithm right machines.
Step 1: Create an n*m matrix bij (binary number for part and
machine). Where, n is parts and m is machines There has been various research based on rank order
m-j clustering in incorporating an algorithm within an algorithm.
Step 2: For each row of i compute, ij *2
Some of these approaches are distance measure based
Step 3: Rearrange the rows in descending order based on the approach, graphical approach, direct clustering, hierarchical
computed numbers clustering, data reorganization approach. All the research
n-i
Step 4: For each row of j compute, ij *2 from the past is more concentrated towards optimizing the
Step 5: Rearrange the columns in descending order based on rank order clustering process and unassigned part or
machines, reducing the complexity of the iteration process by
the computed numbers
grouping the parts and machines as the hierarchy increases,
Step 6: Repeat step 1 until there is no change is observed in reorganizing the rows and columns after the iteration to cater
step 3 and 5 the needs of respective environment and towards specific
Step 7: Stop environment. Specific environment based research such as
distance measure based approach which highlights more
This algorithm works well in an ideal manufacturing about conveyance of product between processes. Another
environment where all the products have same value and all such research is Problem decomposition and data
machines run exactly the same. In real world it highly reorganization where the data is organized by clustering the
unlikely where the entire product have same weight or all

139
IFAC ICONS 2016
140
June 1-3, 2016. Reims, France Nagdev Amruthnath et al. / IFAC-PapersOnLine 49-5 (2016) 138–142

data based on similar values which is again concise to 4.2 Weight Assignment
transportation or conveyance environment. The weights are assigned to each part based on its total cycle
time and the production volume. Since, these weights have
4. MODIFIED RANK ORDER CLUSTERING
different units, they are normalized by converting the weights
ALGORITHM
into percentage. The formula for this normalization is as
In our approach, we take rank order clustering (ROC) below.
algorithm to next level by incorporating manufacturing data
in the process for creation of cells. In a generic Raw Part Weight Data Sum
Cycle Time (sec) 12 10 15 16 17 10 11 15 10 13 129
understanding, this approach can be used to create a balanced
Volume (pcs) 6489 11527 13720 413 276 4152 1958 2109 1397 1938 43979
flow structure involving man, machine, method and money.
In our weight based approach we provide the best optimal Normalizing Data into Percentage
solution in cell creation irrespective on environment, and any Cycle Time (%) 0.093023 0.077519 0.116279 0.124031 0.131783 0.077519 0.085271 0.116279 0.077519 0.100775 1
number of dependent variables (also called as dependent Volume (%) 0.14755 0.262093 0.311956 0.009401 0.006278 0.094405 0.044523 0.047957 0.031773 0.044065 1

factors) affecting the solution of the problem. Normalized Part Weight

Part weight 0.120286 0.169806 0.214117 0.066716 0.069031 0.085962 0.064897 0.082118 0.054646 0.07242 1
Weight and data reorganization based rank clustering
algorithm is an approach where the weights are assigned to
either part numbers or machines or both and the data is Table 1. Normalized cycle time and normalized volume data
reorganized in descending order of their weights. This data is
iterated using rank order clustering algorithm and then the 4.3 Equations
cells are formed. The weights assigned in our approach based The above data is normalized using the formula
on the most common factors of a manufacturing process.
These factors are cycle time, monthly volume, design, nwi` = [CTi / ∑CTi] + [Volumei / ∑Volumei]
availability, ranking for a part structure; reliability, setup
time, utilization, manpower, inter-process distance for a Where, i = part numbers
machine and other generic factors. These weights are later CT is cycle time
normalized. This process is discussed in detail later section. Volume is the part volume processes through that
Based on the results in section V our approach is more in line machine
with real world scenario and can be modified based on nwi is normalized part weight
operational environment. The algorithm for Weight and data
reorganization based rank clustering is as follows. nwi = [CTi / ∑CTi] + [Volumei / ∑Volumei]
4.1 Algorithm nwi`
Step 1: Develop weight factors for part and machines i w and
jw
Step 2: If there are more than weight factors then, convert 5. RESULTS AND ANAYSIS
each weight factor into percentage and then sum it. 5.1 Matrix solutions
For the data analysis, Rank Order clustering method as well
Else assign the weight to the part numbers
as Modified Rank Order Clustering was solved for 3 different
Step 3: Create an n*m matrix bij (binary number for part and levels. The levels are as follows
machine). Where, n is parts and m is machines
Step 4: Rearrange the parts and machines in descending order M=N
based on weights M>N
m-j
Step 5: For each row of i compute, ij *2 M < N where, M is number of machines
Step 6: Rearrange the rows in descending order based on the N is number of parts
computed numbers
n-i This was performed to demonstrate how these two
Step 7: For each row of j compute, ij *2 procedures perform differently in the same environment; the
Step 8: Rearrange the columns in descending order based on difference is attributed to the additional data used and the
the computed numbers modified ROC algorithm. In real world, there is high
Step 9: Repeat step 1 until there is no change is observed in possibility that reflect the scenario and catering these needs is
step 3 and 5 one of the most important focus of our research. After the
final iteration, the weights are used in creation of balanced
Step 10: Stop
cells with equal or similar weights. Table 2 below represents
Since the data is reorganized based on the operating results of weight distribution of two different methods
environment before applying rank order clustering algorithm, applied to thirty (30) randomly generated problem data sets.
the cells formed from matrix obtained in the final iteration is
of same or similar weight. Because of this, the bottlenecks on In table 2 above, M represents number of machines and N is
machines are minimized as the product flows to machines are number of parts for each of the thirty data sets. Cell weight
equally divided among cells. ratio represents how well the cells are balanced. If the cell

140
IFAC ICONS 2016
June 1-3, 2016. Reims, France Nagdev Amruthnath et al. / IFAC-PapersOnLine 49-5 (2016) 138–142 141

balance ratio is 1, then the cells are balanced equally. For Sensitivity analysis is performed to understand the robustness
example, if the cell balance ratio is 0.70, then the cell is of the above particular model as well as to understand the
balanced 70% and there is an imbalance of 30%. Iterations relationship between input and output of our mathematical
represent the number of iterations required to arrive at final model. Since, our data is variable sensitivity analysis would
solution. Cell 1, Cell 2 and Cell 3 represents their cell be used at 3 different levels by altering the part weight to
weights after creation of their cells. understand the behavior of our model. The 3 different levels
are as follows
Mach Parts ROC MROC

# M N Cell Weight %Itirations Cell 1 Cell 2 Cell 3 Cell Weight %Itirations Cell 1 Cell 2 Cell 3 1. Use standard defined weight defined above
1 10 10 0.83 3 0.24 0.4 0.36 0.90 1 0.37 0.29 0.34
2 10 10 0.64 2 0.26 0.52 0.22 0.83 1 0.4 0.33 0.26
2. Changing the volume of the parts
3 10 10 0.81 2 0.28 0.41 0.31 0.89 1 0.37 0.38 0.26 3. Changing the cycle time of the part
4 10 10 0.83 2 0.4 0.37 0.23 0.83 1 0.4 0.38 0.22
5 10 10 0.66 1 0.5 0.29 0.206 0.66 1 0.5 0.28 0.214
The analysis set up is as follows
6 10 10 0.88 2 0.37 0.26 0.35 0.93 1 0.36 0.34 0.3
7 10 10 0.77 2 0.37 0.43 0.19 0.83 1 0.4 0.29 0.3 Level 1 Level 2 Level 3
8 10 10 0.72 1 0.46 0.34 0.194 0.79 1 0.4 0.34 0.206 #
CT (Sec) Volume(Pcs) Part weight CT (Sec) Volume(Pcs) Part weight CT (Sec) Volume(Pcs) Part weight
9 10 10 0.66 1 0.504 0.292 0.203 0.66 1 0.504 0.292 0.203
10 10 10 0.81 2 0.37 0.41 0.214 0.80 1 0.369 0.417 0.214 10 12 6489 0.12028547 12 600 0.05333307 6 6489 0.09702966
11 10 8 0.76 1 0.65 0.34 0.76 1 0.65 0.34 9 10 11527 0.16981087 10 500 0.04444422 12 11527 0.17756281
12 10 8 0.75 2 0.67 0.33 0.76 1 0.65 0.34 8 15 13720 0.21412307 15 12000 0.19456828 5 13720 0.17536338
13 10 8 0.81 1 0.61 0.38 0.89 1 0.44 0.56 7 16 413 0.06671093 16 413 0.06671093 12 413 0.05120705
14 10 8 0.88 1 0.56 0.43 0.99 1 0.5 0.49
6 17 276 0.06902933 17 5000 0.12273679 12 276 0.04964949
15 10 8 0.88 2 0.56 0.43 0.90 1 0.552 0.44
16 10 8 0.85 1 0.58 0.41 0.89 1 0.56 0.44 5 10 4152 0.08596404 10 3000 0.07286688 19 4152 0.12084776
17 10 8 0.87 2 0.57 0.42 0.93 1 0.54 0.46 4 11 1958 0.06489628 11 1700 0.06196307 8 1958 0.05326838
18 10 8 0.84 1 0.59 0.4 0.81 1 0.61 0.38 3 15 2109 0.08211689 15 900 0.06837169 7 2109 0.05110914
19 10 8 0.87 1 0.57 0.42 0.83 1 0.66 0.44 2 10 1397 0.05464227 10 100 0.0398966 15 1397 0.07402212
20 10 8 0.83 2 0.6 0.39 0.98 1 0.49 0.51
1 13 1938 0.07242084 13 12000 0.18681635 13 1938 0.07242084
21 8 10 0.77 1 0.64 0.35 0.77 1 0.64 0.35
22 8 10 0.98 2 0.51 0.49 0.76 1 0.65 0.34
23 8 10 0.96 1 0.52 0.48 0.96 1 0.52 0.48
24 8 10 0.84 1 0.59 0.4 0.91 1 0.55 0.45
Table 4. Normailized raw data that is varied for three levels
25 8 10 0.99 1 0.49 0.5 0.98 1 0.49 0.51
26 8 10 0.97 1 0.51 0.48 0.93 1 0.54 0.46
27 8 10 0.97 1 0.48 0.51 0.99 1 0.49 0.5
28 8 10 0.92 1 0.54 0.45 0.93 1 0.46 0.54 Table 5 is the initial matrix setup or data setup for Sensitivity
29 8 10 0.93 1 0.54 0.46 0.98 1 0.51 0.49 analysis. This is used at 3 different levels in performing the
30 8 10 0.70 1 0.71 0.29 0.90 1 0.55 0.44
analysis. In table 6 we can compare the behavior of MROC
approach and the data obtained at setup 3 levels. The data
Table 2. Cell balance data collection for the 30 problems obtained are positive at level 1 and level 3 but, there is a
obtained from both ROC and MROC process. negative difference at level 2 although, the difference is only
-4%.
Parts
5.2 Weight distribution Machine
512 256 128 64 32 16 8 4 2 1
Performing a weight distribution analysis for our data at their s
10 9 8 7 6 5 4 3 2 1
3 different levels is very important as to understand its 512 10 1 0 1 1 0 1 0 1 1 0 726
behavior at these levels. This also provides a clear view of 256 9 0 1 1 1 1 0 0 0 0 0 480
their performance which is one of the main scope of our 128 8 0 1 0 0 0 1 1 0 0 1 281
research. The results at 3 different levels are as follows 64 7 0 0 0 1 0 1 0 0 0 1 81
32 6 0 0 1 1 0 0 1 0 1 1 203
16 5 0 0 1 1 0 0 1 0 1 1 203
Level ROC MROC CBD 8 4 0 1 1 1 0 0 0 0 0 0 448
4 3 1 0 1 1 1 0 0 0 1 0 738
M=N 76% 81% 5%
2 2 1 0 1 0 1 1 1 1 1 0 702
M>N 83% 87% 4% 1 1 1 0 1 0 1 1 0 1 0 1 693
90% 91% 1% 519 392 831 892 263 707 178 515 566 241
M<N

Table 3. Weight distribution of final matrix obtained using Table 5. Initial matrix used for performing sensitivity
three different levels analysis

From the above table 3, we can analyze that the load balance From sensitivity analysis using 3 different levels, we can
at their 3 difference levels. CBD is the Cell Balance conclude that the MROC approach is more efficient in
difference between ROC approach and MROC approach. It is creation machine cells with balanced loads. Even when the
important to notice that, our new approach presented has weights are altered between the ranges of 5% to 25% the
proved to provide higher cell balance at all three levels with overall cell weight ratio for MROC approach is higher than
inclusion of real world manufacturing data in creation of ROC approach. Sensitivity analysis also provides enough
cells. evidence to conclude, MROC approach is more efficient and
robust in changing environments.
5.3 Sensitivity analysis

141
IFAC ICONS 2016
142
June 1-3, 2016. Reims, France Nagdev Amruthnath et al. / IFAC-PapersOnLine 49-5 (2016) 138–142

6. CONCLUSIONS Chan, H.M. and Milner, D. A.,“direct clustering algorithm for

group formation in Cellular manufacturing”, Journal of
In our research, we were able to include real time
Manufacturing Systems, Vol. 1, pp.65-75, 1982.
manufacturing data in creation of product flow for cellular
manufacturing based on weight assignment for parts and data R, Rajagopalan, J. L. Batra, “Design of cellular production
reorganization approach has proved not only robust and system – A graph theoretic approach”, International
efficient but also a methodology for creation of load balanced Journal of Production Research, Vol. 13, No.6, pp.567-
cells while minimizing bottlenecks. In conclusion based on 579, 1975
the results presented above MROC provided more efficient
P. Krishnananda Rao, and A. M. Chincholkar, A distance
cell formation by overcoming the deficiencies of ROC
measure based approach for solving group technology
approach.
cell, Formation problem”, (2004)

Level
ROC MROC Patel & T. J. Stonhem, Texture Image Classification and
Cells Cell Weight CWR Cells Cell Weight CWR Segmentation using RANK-order Clustering, IEEE,
10 8
1992 0-8186-2920-7/92
7 2
6 0.639 9 0.58 Bromley, D. B., “Rank Order Cluster Analysis”, The British
9 3 Journal of Mathematical and Statistical Psychology,
8 7 1966, Vol. 19, Part 1 105-123.
1 78% 85%
2 6
3 10 Jiang, D., Tang, C. & Zhang, A., “Cluster Analysis for Gene
1 0.36 1 0.41 Expression Data: A Survey”, IEEE Transactions on
5 4 Knowledge and Data Engineering, Vol. 16, No 11,
4 5
November 2004.
10 8
7 2
0.43
Chandrasekharan, B.S., “MODROC – Modified ROC for
6 0.48 7 Group Technology” IJPR, Vol 24, Issue 5, 1986, 1221-
9 6 1233.
8 1
2 95% 91%
2 9 King, J.R., “Machine-component grouping in production
3 3
0.47
flow analysis: an approach using a rank order clustering
1 0.42 10 algorithm”, IJPR Vol. 18 Issue 2, 1980, 213-232.
5 4
4 5 McCauley, J., “Machine Grouping for efficient production”,
10 9 Production Engineering, 1972, 51, 53-60.
7 2
6 0.55 8 0.52
McCormick, W. T., Schweitzer, P.J., and White, T. E.,
9 3 “Problem decomposition and data reorganization by
8 7 clustering technique”, Operation Research, 20, 993-
3 84% 87%
2 6 1008.
3 10
1 0.37 1 0.393 Gupta, T. and Seifoddini, “Production data based Similarity
5 5 Coefficient Method for Machine Cells and Part Family
4 4 Formation Heuristics”, IJPR 1990, No 7, 1247-1269

Table 6. Cell weight distribution for ROC and MROC

process for three levels set for performing sensitivity analysis

REFERENCES
Burbidge, J.L. (1975). “Production Flow Analysis”,
Production Engineering, 1963, 42, 742
W. T. McCornick, J. P. J. Schweitzer, and T. W. White,
“Problem decomposition and data reorganization by a
clustering technique”, Operations Research, Vol.20, No.
5, pp. 993-1009, 1972
C. Dimopoulos, and N. Mort, “A hierarchical clustering
methodology based on genetic programming for the
solution of simple cell formation problems”,
International Journal of Production research, Vol. 39,
No. 1, pp.1-19, 2001.

142
View publication stats