You are on page 1of 13

This article was downloaded by: [University of York]

On: 15 January 2015, At: 19:49


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Production Research


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tprs20

Designing flexible manufacturing cells using a branch


and bound method
a
IBRAHIM AL-QATTAN
a
Tennessee Technological University , Box 5002, Cookeville, Tennessee, 38505, USA
Published online: 24 Oct 2007.

To cite this article: IBRAHIM AL-QATTAN (1990) Designing flexible manufacturing cells using a branch and bound method,
International Journal of Production Research, 28:2, 325-336, DOI: 10.1080/00207549008942714

To link to this article: http://dx.doi.org/10.1080/00207549008942714

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
INT. J. PROD. RES., 1990, VOL. 28, No.2, 325-336

Designing flexible manufacturing cells usinga branch and bound method

IBRAHIM AL-QAlTANt
The grouping of parts into families and machines into cells poses an important
problem in the designand planning ofthe flexible manufacturing cells(FMC).This
paper presents a new method of forming flexible manufacturing cells, based on
branching from seed machine and bounding on a completed part. This method
creates a number of alternative solutions based on the presenceofa seed part which
will enhance the flexibility of the manufacturingcells'design, and which providesan
opportunity to evaluatedifferent options and to selectthe one whichis the most cost
effective.
Downloaded by [University of York] at 19:49 15 January 2015

I. Introduction
The manufacture of batch/discrete lot of component parts has traditionally taken
place in a functional layout where similar machines are grouped together in one area of
the production plant. Thus, during the manufacturing process, batches move through
various work-centres according to specified machining sequences. The main step of
designing an FMC is grouping parts into families and the corresponding machines into
manufacturing cells, which may be used to convert the functional layout into a group
product layout. This layout concerns a group of machine tools and equipment that are
responsible for manufacturing a set of family parts.
Group technology provides a coding and classification scheme which provide tools
for developing methods for grouping various parts and products with similar design
and/or machining processes into a family of parts (FP) and corresponding machines
into machine cells (MC). Therefore, the family of parts is a collection of parts which are
similar either because of the geometric shape and size and/or because similar
processing steps are required in their manufacturing. Dealing with a number of flexible
manufacturing cells as an independent subsystem will insure smooth production flow
of the information and reduce the complexity of the scheduling problems. However,
moving from process or functional layout to group layout compels duplication of some
of the machines that do not have enough capacity to process all the job, and/or the
machine that represents a congestion point in the production flow, the so-called
'bottleneck' machine.
Burbidge (1988) has defined bottleneck machines as those which are heavily loaded
to above or near their available capacity. One way of solving this problem, as Burbidge
suggested, is by adding new machines. The criteria for adding new machines employed
in the proposed method is the number of jobs per machine, assuming that the set-up
time for a given job is much higher than the processing time of the job. Furthermore,
this method addresses more alternative solutions which increase the flexibility of
designing machine cells. The choice among the alternatives will take into account the
cost of machines. However, knowing the real processing and set-up time, the batch size
of each component and the sequence of machining operation may assist the selection
processes. In the following sections some machine-component grouping methods are
reviewed, and a new method is presented.

Revision received April 1989.


t Tennessee Technological University, Box 5002, Cookeville, Tennessee 38505, U.S.A.
002()-7543/90 ssoo © 1990 Taylor & Francis Ltd.
326 I. AI-Qattan

2. Review of literature
Production flow analysis (PFA) due to Burbidge (1975)has been most widely used
for industrial applications. Burbidge's pioneering work in PFA is one of the first
systematic and manual approaches to machine-component grouping. It consists of an
exhaustive analysis of production flow with respect to plant, group and production line.
Based on the concept of component flow analysis, EI-Essawy and Torrance (1972)
introduced similar techniques. McAuley (1972), King (1980), King and Nakornchai
(1982) and many others have proposed different approaches using cluster analysis for
cell and family formation. The bond energy algorithm (BEA) based on a general
clustering algorithm was discussed by McCormick et al. (1972). Rajagopalan and Batra
(1975) have introduced the concept of graph theory for the design of cellular
manufacturing system. Cell formation using the Monte Carlo simulation technique
was discussed by Crookall and Baldwin (1972). The combinatorial method 'hosts
combinations and guest combinations' and other mathematical and heuristic
Downloaded by [University of York] at 19:49 15 January 2015

approaches were developed by Purcheck (1975, 1985). Other investigators, Kumar


et al. (1986) and Vannelli and Kumar (1986), introduced a different algorithm. Hence,
there is a need to develop approaches capable of obtaining results with the minimal
number of duplicated machines. The method presented in this paper is an attempt in
this direction, while also providing more flexible MC design.

3. The proposed method


The proposed method employs network analysis to form the machine cells and
family of parts. The method is based on branching from seed machine and bounding on
a completed part (BMBP). The seed machine represents the starting node for the
network system. Selecting a seed machine with the smallest number of jobs will help to
reduce the size of the network tree and obtain more alternative solutions. Determine
the set of machines which are candidates for duplication either due to the lack of
capacity to process all of the job and/or the machine that represents a congestion point
in the production flow. Burbidge (1988) has defined the bottleneck machines as those
which are heavily loaded to above or very near their available capacity. This problem
may be solved by adding new machines (one of the criteria suggested by Burbidge). The
criterion for adding new machines employed in the proposed method is the number of
jobs per machine, considering the set-up time for a given job is much higher than the
processing time ofthejob. Therefore, the machines which have twice or more jobs that
the average number ofjob per machine will be considered as candidates for duplication.
Most machine-part grouping methods use a chart of M by N matrix with zero or
one entities, where there are M machines and N parts. A 'one' entry in row j and column
i of the matrix indicates that part i has an operation on machine j; a zero entry indicates
it does not.

3.1. Notation
Pi the ith part, where i = 1,2, ... , N
M J the jth machine where j = 1,2, ... , M
ar = { I If part i has an operation on machine j
J 0 Otherwise
1j the number of parts that have an operation on machine j
N
1j= ~:aij for all j= 1,2, ... , M
i= 1
K the number of family parts or machine cells
Designing FM Cs 327

C(K) represent the Kth machine cells


F(K) represent the Kth family parts
r represent the average number of parts/machine
M
['1) = L ~/M for allj= 1,2, ... , M
j=l
{AM} the set of all machines
{AP} the set of all parts
{B} the set of all bottle-neck machines which are candidates for
duplications.
b j = ~rt criteria for adding a new machine.
Set B contains all machines for which b j>2.
For 2.,;b j < 3 the possibility exists of adding one machine ofj. For 3.,;b j<4 the
possibility exists of adding two machines of j.
Downloaded by [University of York] at 19:49 15 January 2015

3.2. Algorithm
Step O. Obtain the bottleneck machine set {B}. Find all machines that b,» 2,
insert them into {B}. Insert all machines into a set {AM}. Insert all parts
into a set {AP].

Stage I-set K = 1
Step l. 'Initialize the Kth group'
Set C(K)= {O} and set F(K) = {O}
Step 2. Select the smallest value of T, as an initial network node for the Kth
group, called the seed machine (Ms).
Insert Ms into C(K) set.
Step 3. Branch all parts which visit seed machine Ms. Insert all parts into F(K)
set.
Step 4. Bound every part which does not require another machine. Otherwise,
branch to a new machine required to be visited. Insert the new m/cs into
C(K) set.
Step 5. Bound every machine which belongs to machine set {B}.
Branch all parts which visit all the remaining machines in set C(K).
Step 6. Criteria for alternative solutions: declare any single part which does
require one or more machines as an exceptional part or a seed part (SP).
The network tree at an SP node may be broken down into subtrees, thus
creating more alternative cells.
Step 7. Do Steps 4 through 6 until all nodes (machines and parts) are bounded
(i.e. the network is closed).
Step 8. Form the Kth group.
C(K) contains all machines in the closed network.
Insert any other parts that can be completed with C(K) machines.
F(K) contains all parts in the closed network.
Step 9. Delete all machines formed in CC(K) from set {AM} where
CC(K) = C(K) - {B}.
CC(K) denotes to machines in cell K with exception of those in set {B}
machines. And delete all parts in F(K) from set {AP}.
Step 10. Increment K by one.
328 I. Al-Qattan

FIND~M~
CCK). M
Downloaded by [University of York] at 19:49 15 January 2015

BRAN A L PARTS
INSERT THEM INTO FCK)
BRANCH ALL M/CS FROM FCK) •
INSERT THEM INTO ClK)

FORM THE KTH MACHINE CELLS CCK)


AND THE KTH FAMILY PART FCK)
DELETE CCK)FROM AM SET EXCEPT
THOSE M/cs E{B}. DELETE FCK)FROM
AP SET

Figure I. Algorithmic flow chart for the model.


Designing FMCs 329

Stage K Repeat Steps I through 10 until all machines and parts are allocated into
machine cells and family parts.
Stop criteria The set {AP} becomes empty.
The algorithmic flow chart for this model is illustrated in Fig. I.

3.3. Example
To illustrate the way the proposed method works and to present the results, a
machine-parts grouping problem involving 16machines and 43 parts has been selected
as a test problem. Burbidge (1975) has solved this problem using a manual method.
Most investigators used his solution to verify the results of the proposed method. The
initial machine-part chart of the problem is presented in Fig. 2.

Algorithm
Downloaded by [University of York] at 19:49 15 January 2015

Step O. Obtain the bottleneck machine set {B}


Find all machines where b,» 2
b 6 and bs are greater than two.
Therefore, B={M6, M8}
Machines 6 and 8 are candidates for duplication.
Stage I-set K = I
Step 1. Initialize the Kth group
C(I)= {O} and F(1)= {o}
Step 2. Select the smallest value of 1] for seed machine.
Obtain an initial network node for the first group.
Seed machine is MI with T1 =2
C(1)={MI}
Step 3. Branch all parts which visit MI.
Insert P37 and P42 in F(K)
F(I)={P37, P42}
Step 4. Branch all new machines required to process P37 and P42.
They are M2, M6, M8, M9 and M16, insert them into C(I).
C(I)={MI, M2, M6, M8, M9, Mll}
Step 5. Bound every machine which belongs to set {B}. M6 and M8
Branch all parts from C(l)-{B} machines.
F(I)={P37, P42, ri. PIO, P28, P32, P4, P38, P40, P18}
Step 6. Criteria for alternative: check for SP?
There is not an individual part that requires one or more machines; no
alternative.
Step 7. The following parts are completely 'bounded':
{PlO, Pl8, P32, P4, P38, P40, P18, P37, P42}
Part 2 needs M 14, and part 7 needs M3.
Insert Ml4 and M3 in C(K).
Parts (P6, P17, P35, P34, P36) will branch from MI4 and M3.
The cycle of network is closed.
Step 8. Forming the first group
C(I)={MI, M2, M3, M6, M8, M9, M14, M16}
F(1) = {P2, P4, P6, P7, PIO, P17, P18, P28, P32, P34, P35, P36, P37, P38,
P40, P42}
w
w
o
Downloaded by [University of York] at 19:49 15 January 2015

~, 2 J .. 5
PARTS
6 7 8 9101112131415 16171819202122232425262728293031 323334353637383940.'4243
TJ
1
2
, I II I
I
I I I
I
II
2
8
3 I I I I I 5

• I
I I I
I I
I I I I
I I I
I
J
I I I I
7
13
5
I I
I
I I I I I I I I I 19
:--
6 r I I I I I I I ;l.
7 I I I 3 -;-
V> I I I J II I I I I I I I I I I I to
-
w 8 I I I I 20
I s:>
Z
9 I I I I I I I I I 10
J:
0
-c
10 I I I ,I J I I 7 §
I I I I I I 6
::;;
"
12 I I I I I 5
13 I I 2
I I I I •
,.
14
15 I I I I J I I 7
I I I I I I I J 8

Figure 2. The original machine-parts matrix for 16 machines and 43 parts.


Designing FMCs 331

BOUND

COMPLETED
NO MORE
MACHINE
REQUIRED
"BOUND"
Downloaded by [University of York] at 19:49 15 January 2015

BOUND

Figure 3. Network diagram of the first group of FM cells.

Figure 3 shows the network diagram of first group of FM cell.


Step 9. Remove all mlcs belong to set CC(I)
CC(I)= C(I)- {B}
CC(I) denotes to all mlcs in cell I excluding those in set {B}.
Step 10. Increment K by 1.

Stage II-set K = I
Step 1. Initialize the second group.
Set C(2)= {O} and set F(2)= {O}
Step 2. The smallest value of ~ is T1 3 = 2.
Seed machine for this group is M13,
C(2)={M13}
Step 3. Branch all parts which visit M13.
P3 and P24 insert the new parts into F(2).
F(2)= {P3, P24}
Step 4. Branch all new machines from F(2).
Insert MS, Mil and MI2 into C(2).
Step 5. Bound every machine which belong to set {B}.
No branching from MS.
Branch from Mil and M12.
Parts (P9, P20, P27, P30, PI I, P22) will be inserted into F(I).
The following parts are completely 'bounded':
{P3, Pll, P20, P22, P24, P27, P30}
332 I. Al-Qattan

Step 6. Criteria for alternative: check for SP?


P9 is incomplete needs two additional mlcs M4 and M5.
Hence P9 is SP; at this node it is possible to generate a subtree (
alterna tive cells).
Insert M4 and M5 into C(2).
C(2)={M13, M8, Mll, MI2, M4, M5}
Part P9 only uses Mil from C(2) while it requires M4 and M5. Hence
may be more cost effectiveto duplicate M II and create another cell, Ot
purchase the part P9.
Step 7. Repeat Steps 4-6. Branch from M4 and M5 all parts
{P5, P9, P14, P19, P21, P23, P29, P8, P15, P16, P33, P4I, P43}
Insert M15, M6 in order to complete the new parts.The cycle is closed'
parts are completed'.
Step 8. Forming the second group:
Downloaded by [University of York] at 19:49 15 January 2015

C(2)={M4, M5, M6, MS, Mll, M12, M13, M15}


F(2) = {P5, P8, P9, P14, P15, P16, P19, P20, P21, P22, P23, P24, P
P29, P30, P33, P4I, P43}
Step 9. Remove all mics belong to set C(2)and not in set {B} from set (AM) (
CC(2) = C(2)- {B}).
Increment K by 1.
Figure 4. shows the network diagram of the second group of FM cells.
Stage II I-set K = 3
Step 1. Initialize the Kth group.
Set C(3)= {O} and set F(3) = {O}.
The smallest value of 1) is T, = 3.
Step 2. Seed machine for this group is M7, set C(3)={M7}.
Step 3. Branch all parts which visit M7, the parts are:
PI, PI3 and P25, set F(3) = {PI, P13, P25}
Figure 4 shows the network diagram of the second group of FM cells.

COMPLETED COMPLETED

Figure 4. Network diagram of the second group of FM cells.


Designing FM Cs 333

Figure 5. Network diagram of the third group of FM cells.

Step 4. Branch all new m/cs from F(3), the machines are M6, M8, and MIO.
Hence C(3)={M7, M6, M8, MIO}.
Step 5. Bound every machine which belong to set {B}.
Downloaded by [University of York] at 19:49 15 January 2015

Branch all parts from M10 which are parts:


{P12, P26, P31, P39}
Step 6. Criteria for alternative: check for SP?
There is not an individual part that requires one or more machines; no
alternative.
Step 7. The following parts are completed:
PI, P13, P25, P12, P26, P31, P39
Therefore, the cycle is closed.
Step 8: Form the third group:
C(3)={M7, M6, M8, MIO}
F(3) = {PI, P12, P13, P25, P26, P31, P39}
Figure 5 shows the network diagram of the third group of the FM cells.
Stop All machines and parts are allocated, hence, set {AM} and {AP} are empty.
Figure 6 shows the final machine-part groups.

3.4. Analysis of the alternatives


The following alternatives may be developed from the network diagrams Figs 3, 4
and 5. The cost of each machine, set-up time, processing time, and the quantity of each
part to be produced may affect the machine cells and family part formation. Therefore it
is important to have alternatives: for example, in Fig. 4, branching from machine (M II)
to part 9, and so on to the other parts. Part 9 requires in addition to M II two other
machines M4, M5 in order to be completed, while all other parts do not. Thus, it may be
more cost effective to either duplicate (MIl) and divide that big group into two groups,
or to buy part 9 from subcontractor without duplicating MIl. Therefore, two
alternative cells may be considered. Figure 7 shows the final machine-part matrix
including the alternative groups, and Table 1 shows the alternative of all FMC cells.
To demonstrate the way the proposed method works and to show how the cost
information plays a role in machine cell formation, another practical application
including cost factors is taken from Purcheck (1985). A facility-parts grouping problem
involving 19 facilities and 89 parts has been solved using the proposed method. The
result of employing the proposed method is to form two facility cells. The first cell
requires 7 facilities (a, b, d, k, n, q, r) and the second cell requires 17 facilities (a, b, c, d, e, f,
334 I. A [-Qattan

PARTS
~2 4 8 7'017182832343538373840423 58 911141518192021222324272930334143112132$283139
1 I I
2 I I I I I I I J
3 I I I I I
61 I I I I I I I I
8 I I I I
9 I I I I J I I I I I
14 I I I
16 I I I I I I I I
4 I I I I I I I
5 J I I I I I I I I I I I I
6 I I I I I I I I
en 8 I I I I I I I I I I I J I I
Downloaded by [University of York] at 19:49 15 January 2015

'"
z 11 I I I I III I
12 I I I
is 13
~ 15 I I I I I I I I
6 I III II
7 I I I
8 I I
,0 I I I I I I I

Figure 6. The final machine-parts matrix (no exceptional elements).

I~
PARTS
wle 2 4 8 7 to 1718 2832343538373840 42 11213 25283139:5 8 91415181921232933414331'2022242730
1
- I I
2 I I I I I I I I
3 I I I I I
61 I I I I I I I I
8 I I I I
91 I I I I I I I I I
14 I I I
16 I I I I I I I I .-
7 I I I
10 I I I I I I
6 I I I I
en 8 .. I I
'":::
4 I I I I I I I
5 I I I I I I I I 1 I I I I
is 6~ I I f I I I
~ 15 I I I I I I I
8 I I I I I I I I
11 I
12 I I I I I
13 I I
8 I I I I I
11 I 1 I III

Figure 7. The final machine-parts matrix for the alternatives, without exceptional elements.
Designing FM Cs 335

Machine--<:omponent M/cs used in each Total m/cs


Alternatives grouping group {m/c cell} required

One group, all 16 m/cs in


all m/c in one one cell 16

\ cell, all parts in


one family
Two groups
11 { CI 10m/cs
C2 Bm/cs
{I,2,3,6,7,8,9, 10, 14,16}
{4,5,6, 8,ll,12,13,15}
18

Three groups
111
\
CI 8m/cs
C2 4m/cs
C3 8m/cs
r
{1,2,3,6,8,9, 14,16}
IO,6,8}
4,5,6,8,11, 12, 13,15}
20
Downloaded by [University of York] at 19:49 15 January 2015

Four groups
CI 8m/cs fl,2,3,6,8,9, 14,16}
IV
j C2
C3
C4
4m/cs
6m/cs
4m/cs
7,IO,6,8}
f4,5,6, i5,8,ll}
12, 13,8, 11}
22

Two groups
V { CI 8m/cs
C2 IOm/cs
{1,2,3,6,8,9, 14,16}
{4,5,6,7,8,10,11,12, 13,15} 18

Table 1. Alternatives of machine-part groups.

g, h, i, j, k, I, m, n, p, s, t, u). The total cost of those 24 facilities is 274090. The optimal


solution for this problem as cited by Purcheck is two cells with 30 facilities, with total
cost of325 190.Thus, the proposed method will save 51100. For more information the
reader should refer to Purcheck (1985).

4. Conclusion
The proposed method forms machine cells and family parts by using network
analysis. This method is efficient and performs a better FM cells of 16 machines and 43
parts given in the example than the clustering algorithm proposed by King (1980), King
and Nakornchai (1982), and the similarity coefficient presented by Seifoddini and
Wolfe (1986). Vannelli and Kumar (1986) have obtained semilar FM cells for this
example, using graph theory, but the method proposed in this paper gives more
alternative solutions. Therefore, the method developed here will serve as a decision aid
to generate possible grouping and identify the minimal duplicated machines. The
generated alternative solutions provide more flexibility and the opportunity to
evaluate different options and choose the one which is the most cost effective.

5. Summary
The grouping of parts into families and machines into cells poses an important
problem in the design and planning of the flexible manufacturing cells. This paper
presents a new method of forming flexible manufacturing cells, based on branching
from seed machine and bound on a completed part. Also, this method creates a number
336 Designing FMCs

of alternative solutions based on the seed part node which will enhance the flexibility of
the manufacturing cell design, and opportunity to evaluate different options and
choose the one which is the most cost effective.

References
BURBIOGE, J. L., 1975, The Introduction of Group Technology (New York: John Wiley).
BURBIOGE, J. L., 1988, Operation scheduling with GT and PBe. International Journal of
Production Research, 26, 424-442.
CHANORASEKHARAN, M. P., and RAJAGOPALAN, R., 1986, MODROC: an extension of rank order
clustering for group technology. International Journal of Production Research, 24,1221-
1234.
CROOKALL, 1. R., and BALDWIN, K. I., 1972, An investigation into application of grouping
principles and cellular manufacturing using Monte Carlo simulation. CI RP, I, 3.
EL-EssAWV, I. G., and TORRANCE, J., 1972, Component flow analysis-an effective approach to
production systems design. Production Engineers, May, 165.
KING,J. R., 1980, Machine-component grouping in production flow analysis: an approach using
Downloaded by [University of York] at 19:49 15 January 2015

a rank order clustering algorithm. International Journal of Production Research, 18, 213-
237.
KING, J. R., and NAKORNCHAI, V., 1982, Machine-component grouping formation in
group technology: review and extension. International Journal ofProduction Research, 20,
117-133.
KUMAR, K. R., KUSIAK, A., and VANNELLI, A., 1986, Grouping ofparts and components in flexible
manufacturing system. European Journal of Operational Research, 24, 387-397.
McAULEY, J., 1972, Machine grouping for efficient production. Production Engineers, 52,
February, 53-57.
MCCORMICK, W. T., SCHWEITZER, P. J., and WHITE, T. E., 1972, Problem decomposition and data
recognition by a clustering technique. Operations Research, 20, 993.
PURCHECK, G., 1975, A linear-programming method for the combinatorial grouping of an
incomplete power set. Journal of Cybernetics, 5, 51.
PURCHECK, G., 1985, Machine-component group formation: an heuristic method for flexible
production cells and flexible manufacturing systems. International Journal of Production
Research, 23, 911-943.
SEIFODDlNI, H., and WOLFE, P. M.,1986, Application of the similarity coefficient method in group
technology. lIE Transactions, 18, 271-277.
RAJAGOPLAN, R., and BATRA, 1. L., 1975, Design ofcellular production systems: a graph-theoretic
approach. International Journal of Production Research, 13, 567.
VANNELLI, A., and KUMAR, R., 1986, A method for finding minimal bottle-neck cells for grouping
part-machine families. International Journal of Production Research, 24, 387-400.
WAGHODEKAR, P. H., and SAHU, S., 1984, Machine-component cell formation in group
technology: MACE. International Journal of Production Research, 22, 937-948.

You might also like