National Institute of Technology Calicut Department of Mechanical Engineering

CELLULAR MANUFACTURING SYSTEM

• A cellular manufacturing system (CMS) is one in which production shop is
partitioned into production cells, each dedicated to the production of part families
with similar processing requirements
• By dedicating a machine cell to the production of part family, many of the
efficiencies of mass production can be realized in a less repetitive batch environment
• Group of machines used to produce the family of parts forms a cell
• Generic fixtures are possible as a cell process a family of parts
• Tools can be stored locally since a part will always be processed through a given
machine
• Set-up-time reduction by using part family tooling and sequencing
Setup Time Reduction Results In:
saving in machine and labour time,
reduced inventory cost and storage space – reduced WIP and finished goods
inventory,
reduction in throughput time – waiting time reduction
batch size become small
better material coordination
Numerical Demonstration of Effect of Setup Time Reduction on WIP and Waiting
Time
• Consider a system where jobs arrive in batches of 50 units. Assume that the system
process two jobs (A and B) intermittently. Setup time for each job = 1 hour,
Processing time of 50 units = 3 hours. A job is sent out after processing all units of a
batch. That is, transfer batch size is equal to production batch size. Assume that both
jobs have same setup and processing time.
• After 4 hours of arrival of job A, the next job B arrives. If the jobs are arriving in the
sequence A first and then B, in a duration of 8 hours two jobs (two batches) are out.
• WIP = 50 units, Waiting time in system = 4 hours
• If the setup time is reduced by half, and the jobs arrives in batches of 25, for the same
duration
– WIP = 25 units, Waiting time in system = 2 hours
• Initially the system process two batches of jobs within the 8 hours duration. Later, in
the same duration 4 batches of jobs are processed. i.e. production rate of batches of
jobs increased (doubled) but the actual production rate in terms of units manufactured
is not changed

Cellular Manufacturing System 19 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

Demonstration Using Mathematical Model

• Assume that set up is cut in half
• For the same amount of production the batch size can be cut into half
• If batch interval arrival time and service time are exponentially distributed, then the
workcell can be modelled as M/M/1 queue
Arrival rate λ
Service rate µ
λ
ρ=
Utilisation
µ
ρ
W =
Waiting time of job in a system
[λ (1 − ρ )]
• Assume that above are the parameters before setup reduction and after the setup time
reduction the parameters be λ ′ = 2λ and µ ′ = 2µ
• Utilisation is not changed
ρ 1
• Waiting time W ′ = = W
[λ ′(1 − ρ )] 2
An Illustration of inventory reduction
• Make-to-stock environment can now become just-in-time environment as a result of
set up time reduction
For instance,
 The monthly requirement of an item were produced earlier
 After set up time reduction it can be produced on weekly basis.
 This reduces finished goods inventory by 75 percent and is likely to reduce
the safety stock by 50 percent
• Let us assume that monthly requirement be 2000 units
• Average cycle inventory is 1000 units
• (Cycle inventory is the average on-hand inventory due to production batches of more
than one unit)
• After set up reduction the production batch size is 500 units
• Average cycle inventory is 250 units
• A reduction in cycle inventory of 75% is obtained

Cellular Manufacturing System 20 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

Safety stock calculation

• Safety stock is often taken as a multiple of the standard deviation of demand in the
uncertainty period
• Assume that the uncertainty period is one month before setup reduction and after the
setup reduction the uncertainty period is reduced to one week
• Assume independent weekly demand, then monthly demand variance would be four
times weekly variance
• Monthly standard deviation is twice the weekly standard deviation
• For instance weekly standard deviation is 50 units, then monthly standard deviation is
100 units
• If safety stock is 1.26 times the standard deviation, the safety stock before and after
set up reduction are 126 and 63 respectively
• Safety stock is reduced by 50% after set up reduction
Some More Features of CMS
• In a cell machines or processes are located in close proximity, and it assists in
simplifying production schedule and control, and implementation of visual production
control procedure like kanban
• Cellular configuration is suitable for a repetitive production environment
• All these indicate that cellular manufacturing and JIT go hand in hand
• In a JIT system of manufacturing, cell uses pull production methods
• Pull production requires repetitive manufacturing, that is, fairly smooth continuous
production of somewhat standardised items
• Cellular manufacturing system create conducive environment for team work
• Allows each work group to deal with problems that arise within its boundaries
• Permits different role for the worker
The cellular manufacturing system design problems are concerned with:
1) Grouping of parts or Part family formation.
2) Design of a cellular layout
3) Design of a machine layout.
Grouping of parts or Part family formation
Grouping can be achieved by:
1. Visual inspection - using best judgment to group parts into appropriate families, based
on the parts or photos of the parts
2. Production flow analysis - using information contained on route sheets to classify
parts

Cellular Manufacturing System 21 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

3. Parts classification and coding – identifying similarities and differences among parts
and relating them by means of a coding scheme
Production Flow Analysis
• Burbridge proposed the Production Flow Analysis (PFA)
• It is a technique for simplifying material flow systems
• PFA consists of five sub-techniques used progressively to simplify the material flow
system in an enterprise
• The sub-techniques are
Company flow Analysis (CFA)
Factory Flow Analysis (FFA)
Group Analysis (GA)
Line Analysis (LA) and
Tooling Analysis (TA)
• Company flow Analysis (CFA): analyses the existing flow of materials between the
different factories in a large company and develops a new, simpler and therefore more
efficient system in which each factory completes all the parts it makes.
• Factory Flow Analysis (FFA): studies each factory in turn. It plans the division of the
factory into major groups or departments each of which completes all the parts it
makes, and it plans a simple unidirectional flow system joining these departments.
• Group analysis (GA): uses matrix resolution to divide each department in turn into
groups, each of which completes all the parts it makes.
• Line analysis (LA): analyses the flow of materials between the machines in each
group to find the information needed for plant layout.
• Tooling analysis (TA): studies each machine in each group in turn, in order to find
“tooling families” of parts which can all be made on the machine with the same set of
tools at the same setup and also to find the sequence of loading which will minimize
setup times.
• PFA is a systematic procedure for dividing the complete organization
• Identification of part families and machine groups is one of the steps in PFA
• The identification of part families and machine groups is commonly referred to as cell
formation
• Consider cell formation as a reorganization of an existing job shop into GT shops
using information given about the processing requirements of parts

Cellular Manufacturing System 22 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

• The processing requirement information used are generally two types:

Machine-part matrix or process matrix and
Route matrix
Machine-Part Matrix
• Machine-part matrix contain zero, one data
• If a part visit a machine, the part machine interaction element is 1, otherwise 0
Part

1 2 3 4 5 6 7 8

1 1 1 0 0 1 0 0 1
Machine
2 0 0 1 1 0 1 1 0

3 1 1 1 0 1 1 0 1

4 0 0 0 1 0 0 1 1

Fig. 1 Machine-part matrix

Route Matrix
• This matrix shows the machine required for a particular operation of a part
• For example, the operation 3 of part 1 is carried out in machine 5

Machine

1 2 3 4 5

1 - 2 1 4 3
Part

2 2 4 3 1 -

3 3 2 1 - -

4 4 1 - 2 3

Fig. 2 Route matrix

Cellular Manufacturing System 23 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

QUESTIONS:
1. Define a cellular manufacturing system.
2. How does the setup reduction affect batch size, cycle inventory and safety stock?
3. Jobs arrive to a production system at the rate of one job per two hours and the jobs are
serviced at the rate of two jobs per three hours. A job is a batch of 10 units and the
setup time is 30 minutes. The company undertakes setup time reduction measures and
as a result, the setup time is reduced by one third.
a) What is the saving in waiting time if the jobs arrive at the same rate?
b) How many jobs waiting in the system are reduced as a result of setup time
reduction?
c) Find the improvement in waiting time as a result of setup time reduction?
d) What is the utilization of the production facility?
4. Illustrate using queuing theory the effect of setup time reduction in waiting time of
jobs in a production shop.
5. How does a cellular manufacturing system maintain variety in the customers point
view, and reduces variety in the production point view?
6. What are the ways the setup reduction possible when a batch production system is
converted into cellular manufacturing system?
7. What is a cellular layout? What is its advantage over other types of layout?
8. List the major steps involved in the production flow analysis

Cellular Manufacturing System 24 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

ROLE OF SIMILARITY IN GT
• In GT similar parts are identified and grouped together to take advantage of their
similarities in manufacturing and design
• Similarities among parts permit them to be classified into part families
• Part families are a central feature of group technology
• A Part Family is a collection of parts which are similar because they possess
similarities in geometric shape and size, and/or similar processing steps are used in
their manufacture
• Purpose of the family determines the attributes to be considered
– If parts design advantages are to be gained then parts of identical shape, size, etc
which are based on design attributes are in one family
– This allows design engineers to retrieve existing drawings to support new part
– Standardization of attributes, prevent part variety proliferation and provide
accurate planning and cost estimation
– For the purpose of manufacturing attributes related with it included in part family
formation
• Similarities are recognized and exploited in three different ways:
– By performing like activities together
– By standardizing similar tasks
– By efficiently storing and retrieving information about recurring problems
Similarity Attributes in Design and Manufacturing
Design Attributes:
– Part configuration (round, prismatic)
– Dimensional envelope (length/diameter ratio, aspect ratio)
– Surface integrity (roughness, tolerance, etc)
– Material type
– Part function
– Raw material state (casting, bar stock, etc)
Manufacturing Features:
– Operations (turning, milling, etc)
– Sequences
– Batch sizes

Role of Similarity in GT

25 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

– Machine and cutting tools

– Processing times
– Production volumes,
– Surface finish, etc
• There are always differences among parts in a family (see figures given below)
• But the similarities are close enough that the parts can be grouped into the same
family
Grouping can be achieved by:
1. Visual inspection - using best judgment to group parts into appropriate families, based
on the parts or photos of the parts
2. Production flow analysis - using information contained on route sheets to classify
parts
3. Parts classification and coding – identifying similarities and differences among parts
and relating them by means of a coding scheme
Parts belong to part families: differences and similarities

(a) Part family1: Prismatic parts

(b) Part family 2: Rotational parts

Role of Similarity in GT

26 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

QUESTIONS:
1. What are the design and manufacturing attributes considered for identifying
similarities among parts?

Role of Similarity in GT

27 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

2. What are the manufacturing attributes considered to exploit manufacturing similarity?

3. What are the purposes, the similarity features of the parts produced in a
manufacturing system used?
4. What are the purposes of identifying the attributes for the creation of a part family?

Role of Similarity in GT

28 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

COMPOSITE PART
• A composite part for a given family is a hypothetical part
• It includes all of the design and manufacturing attributes of the family
• An individual part in the family will have some of the features that characterize the
family but not all of them
• A production cell designed for the part family would include those machines required
to make the composite part
• The cell would be designed to allow for size variations (Within allowable range) as
well as feature variations
• Such a cell would be able to produce any family member, by omitting operations
corresponding to features not possessed by that part
• Definition of composite parts simplifies the identification of groups and also provides
a basis for the design of group tooling, standard machine setups, generic fixtures, etc.
An example for a composite part with its features in parts belong to this composite
part

Composite Part
29 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

Another example of composite part

QUESTIONS:
1. What is a composite part? Explain with an example.
2. What is the use of a composite part? How is it useful for machine cell formation?

Composite Part
30 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

CODING AND CLASSIFICATION SCHEMES

• Design and manufacturing are principal functional areas considered
• Coding refers to assigning a multidigit alphanumeric code to an item
• Each digits and letters in a code represents a feature or attribute about the item like its
dimensions, materials, and machining requirement
• The intent is to compactly describe those part characteristics that will facilitate
determination and retrieval of similar parts
• Codes should reflect how activities could or should be performed
• When constructing a coding scheme avoid institutionalizing existing practices
• The code should inform us as to what machines or processes could be used
• Most classification and coding systems are one of the following:
– Systems based on part design attributes
– Systems based on part manufacturing attributes
– Systems based on both design and manufacturing attributes
Codes exist for:
– Design retrieval
– Machining
– Casting
– Parts feeding
– Robot end effector selection
Construction of Coding System
• Four major issues guide the construction of a coding system
– Part (component) population
– Code detail
– Code structure and
– Digital representation
Part Population
• Designed to cover the entire class or population of parts to be coded
• Scope of part type to be included must be known – rotational, prismatic, sheet metal
or some substance
• Flexible to handle future as well as current parts

Coding and Classification Schemes

31 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

• Discriminate between parts with different values for key attributes – part
characteristics, choice machining process, machines, tooling, tolerances, routings, etc
Code Detail
• Code has to be short so, that could uniquely identify each part and fully describe the
part from design and manufacturing viewpoint
• During design of code trade of between extra details and efficiency may be
considered
• An alternative to a too long and complex code is composite part
• As a general rule all information necessary for grouping the part for manufacturing
should be included whenever possible in the code
• For design, if a feature is such that the design for existing parts with this feature
forms a useful starting point, then the feature should be included in the code
• Primary and secondary shapes feature inclusion depends on the code complexity
Code Structure
• Depending on how the digits of a code are linked, there are three coding system:
Monocode (hierarchical code)
Polycode (attribute code or chain code)
Mixed code (hybrid code)
Monocode
• Inverted-tree hierarchy
• Meaning of a digit in the code depends on the values of preceding digits
• Efficient - relevant information need be considered at each digit and fewer digits
needed
• Difficult to learn because of the large number of conditional inference
• Preferred by design people
Polycode
• Each digit position in a polycode has the same attribute or feature of a part (consistent
meaning)
• Each digit has values 0-9 or A-Z, and each value has a particular meaning, which is
maintained in a reference table
• Easy to learn, use and alter
• Length of code may become excessive of its limited combinatorial features
• Preferred by manufacturing department
• Simple items and geometrically complex items have same length of code
Coding and Classification Schemes
32 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

Hybrid Structure
• A section of code is chain in nature and then switch to several hierarchical digits
• Several such chain/hierarchical sections may exists
Code Representation
• Individual digits should be alphabetic or numeric
• A code always produces numeric/alphabetic/numeric values is easier to remember
and verify (eg: 3B2, 5A6)
• Humans comfortable with characters that convey meaning such as ‘S’ for smooth or
‘T’ for thread
• Adhering to standard industrial terminology will facilitate comprehension of the code
and interpersonal communication

Coding and Classification Schemes

33 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

PART FAMILY FORMATION WHEN CODES AVAILABLE

• Classification or part family formation is the process of grouping similar parts or
separating dissimilar parts based on predetermined attributes
• If a mono code is used, a family is defined as a collection of ‘end twigs’ and their
common node
• To increase the family size, go up to the next node and include all branches attached
at that point
Mono Code System of Classification and Part Family Formation

3xxx Main category

31xx 32xx 33xx

Sub category

321x
322x 323x Special features

3211 3212 3213 3221 3222 3231 3232 3213 Values

3223

Family of parts
Fig: Mono code system of classification and part family formation

Part Family Formation when Polycode is Available

Part Family Formation When Codes Available

34 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

• Objective function is defined using distance function

f – Number of part family required
P – Number of parts
Xp – Code of the item
Xpk – kth attribute value of part p, K is the number of digits of the code,
k = 1, 2…K
dpq – Distance between codes Xp and Xq
Three distance measures are
 Minkowski distance metric
 Weighted Minkowski distance metric
 Hamming distance metric
Minkowski distance metric
1
 K
 r
d pq = ∑ X pk − X qk 
r

 k =1 
Where, r is a positive integer.
Two special cases of the above metric are
• Absolute metric (for r = 1)
• Euclidean metric (for r = 2)
Weighted Minkowski distance metric
1
K r r
d pq = ∑Wk X pk − X qk 
 k =1 
Hamming distance metric

d pq = ∑ δ (X pk . X qk )
K

k =1

1, ifX pk ≠ X qk
Where, δ (X pk , X qk ) = 
0, otherwise

Part Family Formation When Codes Available

35 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

Cluster Analysis
• Objective of cluster analysis is to assign P parts to f part families while minimizing
some measure of distance
Hierarchical Clustering
• Parts are first grouped into a few broad families
• Each of which is then partitioned into smaller part families and so on until the final
families are generated
• Parts are clustered at each step by lowering the amount of interaction between each
part and a part family, to develop a tree-like structure called a dendogram
Example:
A company uses an eight digit poly code to distinguish part types. Each code digit is
assigned a numeric value between 0 and 9. The six part types thus coded are given below.
a) Find the Minkowski absolute distance metric between the parts.
b) Determine the Hamming distance metric between parts.
c) Using the hierarchical clustering algorithm for the nearest neighboring approach
(with Minkowski absolute distance) construct the dendogram.
d) Find part families, which are 6 or more Minkowski absolute distance apart
Digits
Parts 1 2 3 4 5 6 7 8
1 3 1 1 6 3 8 0 7
2 4 3 1 5 1 8 1 4
3 4 2 1 5 1 8 0 4
4 5 1 1 6 3 7 0 7
5 4 2 1 5 1 5 1 4
6 3 1 1 6 3 6 2 7
Classification codes of parts

Solution
1 2 3 4 5 6
1 - 10 8 3 12 4
2 10 - 2 11 4 12
3 8 2 - 9 4 12
4 3 11 9 - 11 5
5 12 4 4 11 - 10
6 4 12 12 5 10 -
Minkowski absolute distance metric between parts

Part Family Formation When Codes Available

36 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

1 2 3 4 5 6
1 - 6 5 2 7 2
2 6 - 2 7 2 7
3 5 2 - 6 2 7
4 2 7 6 - 7 3
5 7 2 2 7 - 7
6 2 7 7 3 7 -
Hamming distance between parts

Hierarchical clustering steps

Iteration 1
• Group parts with minimum distance – {2,3}
• Revise the distance matrix

1 (2,3) 4 5 6
1 - 8 3 12 4
(2,3) - 9 4 12
4 - 11 5
5 - 10
6 -
Iteration 2
• Consider lowest distance and group – {1,4}
• Revise the distance matrix
(1,4) (2,3) 5 6
(1,4) - 8 11 4
(2,3) - 4 12
5 - 10
6 -
Iteration 3
• Group – {2,3,5} or {1,4,6}
take any one group
• Revise the distance matrix
(1,4) (2,3,5) 6
(1,4) - 8 4
(2,3,5) - 10
6 -

Part Family Formation When Codes Available

37 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

Iteration 4
• Group – {1,4,6}
• Revise the distance matrix
(1,4,6) (2,3,5)
(1,4,6) - 8
(2,3,5) -

Iteration 5
• Merge the two part families with a distance measure of 8
• Corresponding dendogram is shown below

Distance
0 2 4 6 8

Parts
5
1

6
Dendogram showing the distance of parts

• When the dendogram is cut at a distance of 6, two part families are formed
• The part families are {2,3,5} and {1,4,6}

Part Family Formation When Codes Available

38 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

QUESTIONS
1. Explain the part family formation (classification) using various coding schemes.
2. Distinguish between the part-family formations methods under poly codes and mono
codes.
3. What are the purposes of identifying the attributes for the creation of a part family?
4. Identify various code structures; Show diagrammatically the hierarchical code
structure.
5. Identify various code structures. Describe the characteristics of chain code.
6. What are the purposes coding and classification used? Explain how it can be used for
cellular manufacturing system design?
7. What is Minkowski distance metric? How is it useful for cellular manufacturing
system design?
8. A company has established a nine-digit coding scheme to distinguish between various
types of parts. The six part types coded are given below. Each code digit is assigned a
numeric value between 0 and 9:
Part 1: 112171213
Part 2: 112175427
Part 3: 112174327
Part 4: 102173203
Part 5: 112175327
Part 6: 412174453
a) Find the Minkowski absolute distance between parts.
b) Using the hierarchical clustering algorithm construct the dendogram for parts and
identify two part families.
c) Find the Hamming distance metrics between the six part types and identify two
part families by constructing the dendogram.
9. A company has established a mono code scheme to distinguish between various types
of parts. The four parts coded are given below.
Part 1: 53457
Part 2: 33457
Part 3: 53451
Part 4: 33456
Group the above parts such that dissimilar parts should be in different groups.

Part Family Formation When Codes Available

39 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

10. A XYZ company has established a nine-digit coding scheme to distinguish between
various types of parts. Polycode of two part types is given below. Determine the
Minkowski distance (for r = 2) and Hamming distance between the parts.
Part 1: 112171213
Part 2: 112175427
11. An eight-digit poly code is used in a company to distinguish part types. Each code
digit is assigned a numeric value between 0 and 9. The codes for the six part types are
given below. Using the hierarchical clustering algorithm for the nearest neighboring
approach (use Minkowski absolute distance) construct the dendogram.
Digits
Parts 1 2 3 4 5 6 7 8
1 2 1 1 6 3 8 0 7
2 4 3 1 5 1 8 1 4
3 3 2 1 5 1 8 0 6
4 5 1 1 6 3 7 1 7
5 2 2 1 6 1 5 1 4
6 3 1 1 6 3 6 2 7
Classification codes of parts

Part Family Formation When Codes Available

40 March 2013
 National Institute of Technology Calicut Department of Mechanical Engineering

SIMILARITY COEFFICIENT-BASED CLUSTERING

Similarity Coefficient
• Similarity coefficient between two machines is defined in terms of the number of
parts that visit each machine
• Input for this analysis is usually part-machine matrix or process matrix
• Part-machine matrix contain zero, one data
• If a part visit a machine, the part machine interaction element is 1, otherwise 0
Similarity Coefficient Determination

Machine n
1 0
1 a b
Machine m
0 c d

Where, a is the number of parts visiting both machines

b is the number of parts visiting machine m and not n
c is the number of parts visiting machine n and not m
d is the number of parts not visiting either machine
Let S mn denote the similarity between machines m and n
a
S mn = , 0.0 ≤ S mn ≤ 1.0
(a + b + c )
This coefficient is called Jacards similarity coefficient
This similarity coefficient can also be defined as follows:
p mn - number of parts visiting machines m and n
p m - number of parts visiting machines m

pn - number of parts visiting machines n

pmn
Smn =
(pm + pn − pmn )
Similarity between groups of machines is determined using any of the following method

Similarity Coefficient-Based Clustering 41 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

Let t and v are two groups of machines

m∈t
n∈v

{Smn } - Single linkage

Stv = Max
m∈t
n∈v

Stv = Min
 S mn { } - Complete linkage
m∈t
n∈v

∑ ∑ S mn
Stv = m ∈ t n ∈ v - Average linkage
Nt N v
Where, Nt – Number of machines in group t
Nv - Number of machines in group v
Clustering Algorithm
Step 1 Compute the similarity coefficient between machine groups in matrix form
Initially assume each machine is in a separate machine group; use Jacards
similarity
Step 2 Find the maximum value in the similarity coefficient matrix and join the two
machine groups. Update the similarity coefficient matrix (Use appropriate
similarity coefficient based on the linkage planned)
Step 3 when the similarity coefficient matrix consists of one machine group, stop;
otherwise go to step 2.
Step 4 Draw the dendogram and decide the part families
• In this case, first machines are grouped and then part families are formed (sequential
approach)

Similarity Coefficient-Based Clustering 42 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

Example
Apply single linkage clustering algorithm for the data given below.
Part
1 2 3 4 5 6 7 8

1 1 1 0 0 1 0 0 1

2 0 0 1 1 0 1 1 0

Machine 3
1 1 1 0 1 1 0 1

4 0 0 0 1 0 0 1 1

5 0 0 1 1 0 0 1 0

6 1 1 0 0 0 1 0 0

Initial Part-Machine Matrix

Solution
Step 1
Construct Jacards similarity coefficient matrix
1 2 3 4 5 6
1 0 0 0. 67 0.17 0 0.40
2 0 0.25 0.40 0.75 0.17
3 0 0.125 0.125 0.50
4 0 0.50 0
5 0 0
6 0
Step 2.
Iteration1
Group – {2,5}
1 (2,5) 3 4 6
1 0 0 0.67 0.17 0.4
(2,5) 0 0.25 0.50 0.17
3 0 0.125 0.5
4 0 0
6 0

Similarity Coefficient-Based Clustering 43 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

Iteration 2
Group – {1,3}
(1,3) (2,5) 4 6
(1,3) 0 0.25 0.17 0.5
(2,5) 0 0.50 0.17
4 0 0
6 0
Iteration 3
Group – {1,3,6}
(1,3,6) (2,5) 4
(1,3,6) 0 0.25 0.17
(2,5) 0 0.50
4 0
Iteration 4
Group – {2,5,4}
(1,3,6) (2,5,4)
(1,3,6) 0 0.25
(2,5,4) 0

QUESTIONS:

1. A manufacturing system is producing 8 products (A, B… H). The route sheet of the
products is given below. For the data given,
(i) Develop the part-machine incidence matrix.
(ii) Determine the Jacard similarity between machines (processes).
(iii) Develop a dendogram and suggest possible groups of machines. Use average
linkage clustering procedure.

A----- mill – turn – drill – cut – chamfer

B----- plane – grind – weld – file
C----- turn – drill – cut – chamfer
D----- mill – turn – cut – chamfer
E----- plane – weld – file – grind
F----- weld – file – grind
G---- mill – turn – grind – file
H---- turn – mill – grind – file
Route sheet
2. There are seven machines (A through G) to process 6 components. The routing of
component is provided in the table given below. Calculate the Jacards similarity. If

Similarity Coefficient-Based Clustering 44 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

two machines groups {B, D} and {A, E} exist, then calculate the similarity between
these two groups using average linkage.

Component Machine Required for Operation

1 2 3 4
3B15 B D A E
3B27 C F G -
4C18 F C G -
5CA1 B A E D
1FB1 B D C -
1FB3 B D C E

3. The following table lists the weekly quantities (demand) and routings of ten parts that
are being considered for cellular manufacturing in a machine shop. Parts are
identified by letters, and machines are identified numerically. For the data given,
(i) Develop the part-machine incidence matrix, and
(ii) Determine the Jacards similarity between machines
Part Weekly Machine Routing Part Weekly Machine Routing
Quantity Quantity
A 50 3 2 4 F 60 5 1
B 20 6 1 G 5 3 7 4
C 75 1 5 H 100 3 2 4 7
D 10 6 5 1 I 40 3 4 7
E 12 3 2 7 4 J 15 5 6 1

(iii) If machines 3 and 2 are grouped, and machines 7 and 4 are grouped, calculate the
similarity between these two groups based on complete linkage.
Can these machines form a group as per cellular manufacturing concept, why?

4. Develop machine groups using complete linkage clustering method and average
linkage clustering and draw the dendogram for the following machine-part matrix.

Part
1 2 3 4 5 6 7
1 1 1 0 0 1 0 1
2 0 0 1 1 0 1 0
Machine 3 1 1 1 0 1 1 1
4 0 0 0 1 0 0 1
5 0 0 1 1 0 0 0
6 1 1 0 0 0 1 0

Similarity Coefficient-Based Clustering 45 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

5. A production system has 8 machines (Namely A to H) which are used to produce 8

parts (1, 2... 8). The route matrix of this production system is given in figure below
which shows the operation number. For example, the first operation of part 1 is in
machine C and its second operation is in machine D and so on. The machine cells
formed for this system is as follows: Cell 1 – (F C H D), and Cell 2 – (G B A E).
Determine the part family corresponding to these machine cells. Describe your logic
of assigning a part to a part family.
Part
Machine 1 2 3 4 5 6 7 8
A 2 2 2
B 5 3 3 2
C 1 3 6 2
D 2 2 3
E 1 4
F 4 1 1 3
G 4 4 1 1 1
H 3 5 1
Fig. Route matrix

6. The machine-part incidence matrix of a production system is given below.

Part
1 2 3 4 5 6 7
1 1 1 0 0 1 0 1
2 0 0 1 1 0 1 0
Machine 3 1 1 1 0 1 1 1
4 0 0 0 1 0 0 1
5 0 0 1 1 0 0 0
6 1 1 0 0 0 1 0
Application of hierarchical clustering algorithm to this production system along with
average linkage clustering results in grouping of machines in the following order.
Step Groups formed Similarity coefficient
1 {1}&{3} 2/3
2 {2}&{5} 2/3
3 {1, 3}&{6} 9/20
4 {2,5}&{4} 7/24
5 {1,3,6} & {2,5,4} 137/1400
Develop a dendogram for the groups formed. If two machine groups are identified
from the dendogram, develop the corresponding part family.

Similarity Coefficient-Based Clustering 46 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

BINARY ORDERING ALGORITHM (RANK ORDER

CLUSTERING)
Block Diagonal Structure
• Provides an efficient routine for taking an arbitrary 0 – 1 machine-part matrix and
reordering the machine rows and part columns to obtain a nearly block diagonal
structure
Part
Machine 1 2 3 4 5 6 7 8
A 1 1 1
B 1 1 1
C 1 1 1
D 1 1 1
E 1 1
F 1 1
Example of Dense Block Diagonal Matrix
• Block diagonal means that a machine–part matrix can be partitioned such that the
main diagonal boxes contain mainly 1s and less number of 0s but the off-diagonal
boxes contain only 0s.
• To get G X G partitioned matrix, we have to draw (G – 1) horizontal lines and (G – 1)
vertical lines.
• The on-diagonal boxes show the natural group for the manufacturing plant
• However, groups may not be totally independent

Part
Machine 1 2 3 4 5 6 7 8
A 1 1 1
B 1 1 1
C 1 1 1 1
D 1 1 1 1
E 1 1
F 1 1
Example for groups not independent
• Elements C3 and D7 fall outside the diagonal blocks. These are called exceptional
elements
• Alternatives to handle exceptional elements
– Try to redesign parts 3 and 7 so as not to require machines C and D
respectively

Binary Ordering Algorithm 47 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

– Acquire extra machines of type C and D, placing a C in groups 1 and 2 and a

D in groups 2 and 3 (Utilisation should be considered)
– Route parts 3 and 7 to two groups
– Combine these nearly independent groups into one large group
Binary Ordering Algorithm
• Consider the rows and columns as binary strings
• To get block diagonal form, the similar rows should be brought together and similarly
similar columns
• A row is a binary number, then similar rows have similar values; similarly we can
envision columns
• That is, we can reorder rows or columns in the descending order of their binary value
Algorithm
Step 1. For row m = 1,2…M, compute the decimal equivalent of binary string cm

P
cm = ∑ 2 P − p a pm ; a pm = 0 or 1
p =1
Where, P – number of parts (columns), p – index for column
Reorder the rows in decreasing order of cm . In case of tie, keep the original order

Step 2. Compute the decimal equivalent of binary string r p corresponding to column p

r p = ∑ 2 M − m a pm ; a pm = 0 or 1

Reorder the columns in decreasing order rp . In case of tie, keep the original order
Step 3. If the new part-machine matrix is unchanged, then stop, else go to step 1.

Example:
Part
Machine 1 2 3 4 5 6 7 8
A 1 1 1
B 1 1
C 1 1 1 1
D 1 1
E 1 1 1 1
F 1 1 1

Binary Ordering Algorithm 48 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

Solution:

Step 1.
Part
Machine 1 2 3 4 5 6 7 8 Value
A 1 1 1 200
B 1 1 33
C 1 1 1 1 102
D 1 1 17
E 1 1 1 1 54
F 1 1 1 200
P− p 128 64 32 16 8 4 2 1
2
Order of rows – A, F, C, E, B and D

Step 2.
Part
Machine 1 2 3 4 5 6 7 8
2 M −m
A 1 1 1 32
F 1 1 1 16
C 1 1 1 1 8
E 1 1 1 1 4
B 1 1 2
D 1 1 1
Value 48 56 14 5 48 12 12 3

Order of columns – 2, 1, 5, 3, 6, 7, 4 and 8

Step 1.
Part
Machine 2 1 5 3 6 7 4 8 Value
A 1 1 1 224
F 1 1 1 224
C 1 1 1 1 156
E 1 1 1 1 30
B 1 1 17
D 1 1 3
P− p 128 64 32 16 8 4 2 1
2
• Ending orderings are not unique for a given data set.

Binary Ordering Algorithm 49 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

• Different starting orderings may yield different ending orderings.

• It provides starting point for most of the grouping procedure.

QUESTIONS:
1. Describe the use of block diagonal structure of process matrix. What is exceptional
element? Is it desirable? Give reason for your answer.
2. What is the basic principle of binary ordering algorithm?
3. How are the exceptional elements managed in the cellular manufacturing system?
4. Six manufactured components and their machine sequence are given below.

Component P1 P2 P3 P4 P5 P6

Machine sequence Q, S, T X, Q, S X, M, Q, S X, M Q, T T

Create a process matrix. Form work cell grouping using the binary ordering
algorithm. Discuss the result.
The production volume analysis shows a requirement of two machines of Q and T
and all other machines required are one each. Currently there are two machines of X
and M. So the number of machines made available is given below:

Machines M Q S T X

Number of machines available 2 2 1 2 2

Discuss the cell grouping under the above given machine scenario.
5. The following table lists the weekly quantities and routings of ten parts that are being
considered for cellular manufacturing in machine shop. Parts are identified by letters,
and machines are identified numerically. For the data given, (i) develop the part-
machine incidence matrix, and (ii) apply the rank order clustering techniques to the
part-machine incidence matrix to identify logical part families and machine groups.

Part Weekly Machine Routing Part Weekly Machine Routing

Quantity Quantity

A 50 3 2 7 F 60 5 1

B 20 6 1 G 5 3 2 4

C 75 6 5 H 100 3 2 4 7

D 10 6 5 1 I 40 2 4 7

E 12 3 2 7 4 J 15 5 6 1

Binary Ordering Algorithm 50 March 2013

 National Institute of Technology Calicut Department of Mechanical Engineering

6. What is an ideal cell? Discuss the implications of exceptional elements and voids in
the context of an Ideal cell.
7. Form a block diagonal structure and identify exceptional elements for the suggested
grouping of the given machine-part incidence matrix.
Cell 1 = {M1, M3, M6} and Cell 2 = {M2, M4, M5}
Part family 1 = {P1, P2, P5, P6, P8} and Part family 2 = {P3, P4, P7}
Part

1 2 3 4 5 6 7 8

1 1 1 0 0 1 0 0 1

2 0 0 1 1 0 1 1 0
Machine 3 1 1 1 0 1 1 0 1

4 0 0 0 1 0 0 1 1

5 0 0 1 1 0 0 1 0

6 1 1 0 0 0 1 0 0

Binary Ordering Algorithm 51 March 2013