Documente Academic
Documente Profesional
Documente Cultură
Setup operators tried to run parts with similar grouping of parts and clustering of cells
operational requirements “back to back” but simultaneously.
experienced difficulties due to the confusion Some of the well-known model formula-
associated with 26 machines and 1,000 plus tion algorithms include: Clustering analysis
parts. Indeed, this resulted in an increase in (Kusiak et al., 1986) and similarity coefficient
lead-time because of the secondary queuing (Tam, 1990) for part families formations;
that this generated. assignment model (Offodile, 1993); branch-
Based on numerous success stories, the and-bound (Cheng et al., 1991), tree search
management was convinced that a change of (Cheng, 1993), integer programming
configuration to cellular would not only (Gunasingh and Lashkari, 1989a; Rajamani et
reduce the lead-time substantially but also al., 1992), graph theory (Askin and Chiu,
bring in the “focus” needed (and advocated 1990; Rajgopalan and Batra, 1975), and
by such eminent strategists as Hayes and simulation annealing (Venugopal and
Wheelright (1979), Skinner (1969; 1974) and Narendran, 1992) for solving the machine
others). This, in turn was expected to provide, allocation and grouping problems; cost-based
in somewhat longer term, additional advan- (Askin and Subramanian, 1987), sequential
tages in a market that sought quality, speed, modeling (Gunasingh and Lashkari, 1989b),
and customization to the maximum extent and mathematical programming (Sharer et al.,
possible. Finally, it was felt that such a focus 1992; Wei and Gaither, 1990) for cell forma-
would constitute an excellent precursor to tion in cellular manufacturing.
dovetail this effort into the long-term plans of
the company to go “JIT”. Clustering algorithms for solving a cell
formulation problem
Literature survey The clustering algorithms developed as solu-
tion techniques for cell formulation problems
Over the years, numerous studies have sur-
can also be classified into three categories:
faced that have proposed new methodologies
(1) array-dependent clustering methods,
and algorithms dealing with the issue of cellu-
(2) mathematical programming, and
larization of process layouts. Cheng et al.
(3) others.
(1995) present a comparison of several tech-
niques developed in cellular layout starting The array-dependent clustering methods rely
from a functional layout. In general, the on the spatial characteristics of the initial
literature on cellular manufacturing can be matrix to arrive at the final row and column
classified into two broad groups: algorithms permutations that define the block diagonal
that are used for generalized model formula- form (Shargal et al., 1995). The rank order
tion, and clustering algorithms for solving a clustering algorithm (King, 1980), the direct
cell formulation problem. clustering algorithm (Chan and Milner,
1982), the bond energy algorithm
Algorithms for generalized model (McCormick et al., 1972) and the shortest
formulation spanning path (Slagle et al., 1975) are array-
The algorithms for generalized model formu- dependent methods.
lation address the design issues in the cellular The mathematical programming models
manufacturing layout. These algorithms can are used to develop algorithms that use math-
be further classified as: ematical programming approaches to find the
(1) part family models; most desirable solution to a problem. Some of
(2) machine assignment models; and the well-known mathematical programming
(3) cell formation models. models include: close neighboring (Boe and
A part family model groups n parts into p Cheng, 1991), simulated annealing (Boctor,
families based on similar geometric character- 1991 ), A*-based approaches (Kusiak et al.,
istics or processing requirements to take 1993), branch-and-bound (Kusiak and
advantage of the similarities in design and Cheng, 1990), branching (Kusiak, 1991),
manufacturing processing of parts. A machine similarity coefficient (Waghodekar and Sahu,
assignment model assigns machines to 1984), single linkage (McAuley, 1972), aver-
machine cells to process part families effi- age linkage (Seifoddini and Wolfe, 1986), and
ciently. A cell formation model decides the zero-one formulation (Zhu et al., 1995).
245
Reconfiguring a manufacturing system for strategic advantage Logistics Information Management
Ashok Kumar and Jaideep Motwani Volume 11 · Number 4 · 1998 · 244–256
Other approaches used for cell formation • Work centers/machines that were too big to
include: neural networks (Kaparth and move (e.g. roll-form and tube mill) will not
Suresh, 1992; Kusiak and Chung, 1991), be considered for reconfigured layout.
knowledge-based system (Chow and • Work centers/machines which were preset
Hawaleshka, 1993), graph-based approaches to match with the non-mobile workstations
(Askin et al., 1991; Vannelli and Ravi Kumar, will not be considered for movement.
1986), production flow analysis (Burbidge, • Work centers/machines that were allocated
1971; Ganesh and Srinivasan, 1994), seed to certain projects that were assigned as
based (Chandrasekharan and Rajagopalan, stand-alone cost- or profit-centers will be
1986; 1987), simulation (Durmusoglu, 1993; excluded from configuration change.
Massay et al., 1995), data envelopment analy- • Work centers/machines that were posi-
sis (Shafer and Bradford, 1995), fuzzy adap- tioned near certain service areas (e.g.
tive resonance theory (Burke and Kamal, material test), due to frequent mutual
1995), occupancy value based (Khator and interaction will not be considered.
Irani, 1987), and genetic algorithms with
• Work centers needed for large volume
multiple objective (Venugopal and
production and part of other dedicated
Narendran, 1992).
systems or transfer lines were excluded
Even though there are many approaches to
from consideration.
designing and solving cell formulation prob-
• Finally, work centers/machines that were
lems, we believe based on experience and
perceived by the team not to “fit” the idea
analysis (Cheng et al., 1995) that the partially
of a “focused” cell will not be moved.
linearized King’s algorithm is one of the
simplest and perhaps the most efficient This process eliminated a bulk of machines/
method for generating cells that constitute the work centers from consideration and the team
basis for further discussion. was left with 38 work centers. Figure 1 shows
a conceptual layout (functional or process
Implementation of configuration change type) before the change was implemented.
Notice that the machines performing similar
The management began by forming a com- operations are grouped together at this stage.
mittee headed by the manufacturing superin- A further fine-tuning was done based on type
tendent, and one member each from engi- of products that were produced on this
neering, methods, process planning, quality machines and those involved in the produc-
control, and maintenance. The task assigned
tion of grossly dissimilar parts were also elimi-
was to be completed in two stages. First stage
nated. The residual 26 work centers were
involved conducting a feasibility study to
selected as candidates for reconfiguration.
bring about a layout transformation from
functional to cellular in a “suitable” area of
Creation of database
the company, and the second stage involved
Using intellect software of AIC, a compound
actual transformation and evaluation of bene-
database query was created using the work
fits of the switch.
centers as main field and the following infor-
The team devoted its first two meetings to
mation was downloaded for each work center:
fine-tune the objective of the study and identi-
fy the “pocket” (as one of the members pre- part numbers processed, standard hours,
ferred to call it) in the company that is best setup rate, minimum order quantity, and
suited for cellular implementation (we will use average weekly usage (Table I). It turned out
the term GT or group technology or cellular that a total of 1,054 parts would be affected
interchangeably). by the proposed reconfiguration. Unfortu-
nately, the standard code for parts at this
Identification of “candidate” areas for company had no relation with the design,
GT implementation shape or manufacturing process used for the
With over 30,000 parts and over 200 part and was therefore irrelevant in terms of
machines in the Panel plant, it was apparent forming a logical basis for classification. It was
that a smaller “focused” area will have to be necessary to develop a part-process matrix
identified that will be the first candidate for which is a (0, 1) matrix.
the change. The team converged on such an C = [Cij] where Cij = 1 if part i is processed
area by applying the following rules: on machine j, otherwise Cij = 0. This was
246
Reconfiguring a manufacturing system for strategic advantage Logistics Information Management
Ashok Kumar and Jaideep Motwani Volume 11 · Number 4 · 1998 · 244–256
developed using AIC download information Step 4: If all Wis are in descending order from
in Table I. top to bottom, go to Step 5. Else,
rearrange rows so that Wis are in
Application of grouping algorithm ascending order.
King’s algorithm Step 5: If all Wjs are in ascending order from
Step 0: Assign row numbers 1, ..., m from top left to right, and all Wis are in
to bottom, and column numbers 1, ..., descending order from top to bottom,
n, from left to right. stop. The clusters of ones form the
Step 1: For each column j, compute total desired clusters. Else go to Step 1.
weight of column j, j = 1, ..., n: Of numerous methods available in literature
(1) described previously, King’s algorithm (King,
1980) was identified to be the best one for the
purpose of forming focused cells. The original
Step 2: If all Wjs are in ascending order from
algorithm by King is described above. Notice
left to right go to Step 3. Else,
that multiplying by increasing powers of 2
rearrange columns so that Wjs are in
have the effect of moving ls towards the right
ascending order. and towards the bottom of the matrix, hence
Step 3: For each row i compute total weight of providing clustering of parts with similar
row i: processes.
It soon became clear that the algorithm is
(2)
inappropriate for the size of the configuration
change being contemplated. 21054 was too
247
Reconfiguring a manufacturing system for strategic advantage Logistics Information Management
Ashok Kumar and Jaideep Motwani Volume 11 · Number 4 · 1998 · 244–256
Note:
a = Suppressed for confidentiality
248
Reconfiguring a manufacturing system for strategic advantage Logistics Information Management
Ashok Kumar and Jaideep Motwani Volume 11 · Number 4 · 1998 · 244–256
(3) (8)
Step 2: If all Wjs are in ascending order from Step 4: If all Wis are in descending order from
left to right go to Step 3. Else, top to bottom, go to Step 5. Else,
rearrange columns so that Wjs are in rearrange rows so that Wis are in
ascending order. ascending order.
Step 3: For each row i, compute total weight
One of the 16 pages of the final clusters devel-
of row i:
oped is placed in Figure 2. A quick review of
(4) the clustered data laid out with remarkable
clarity many opportunities of cell formation.
The machines and parts included in the first
Step 4: If all Wis are in descending order from cell can be seen in Table II. The data were
top to bottom, go to Step 5. Else, beginning to look like information. Based on
rearrange rows so that Wis are in the output scan, the team developed three
ascending order.
alternate plans for reconfiguration of the
This allowed the cells to be made; however, existing plant. These were presented to the
the resolution between successive iterations CEO, shop superintendent, and many other
was not adequate and the process was taking personnel who would be impacted by the
too long. At this point King’s algorithm was change. A significant brainstorming ensued
modified one more time, to part linearize and and finally a plan that involved formation of
part exponentialize, to get the “best of both five cells was adopted.
worlds” – speed as well as resolution between
successive iterations of the algorithm. The Loading and capacity analysis
difference in the original and the modified Next, the issue of capacity planning for these
algorithm is the following: instead of the 1,054 parts was addressed. All members of
weights growing exponentially from left to the team were quite apprehensive about the
right and top to bottom, they grow linearly. ability of cells to accommodate the total load
This slows down the formation of the clusters that was normally processed on these
a little bit, but allows for the opportunity of machines. Load for each machine (in each of
making the clusters when the matrix is big. the five cells) was computed by taking into
With an exponential growth of weights as account the weekly usage, part processing
proposed by King, the cluster formation times, and setup times. To the team’s sur-
becomes impossible, even for the problems of prise, the loading on all machines including
decent practical size. The modified algorithm the one that would process the largest group
no. 2 shows the effect on steps 1 to 4. of parts was far less than 40 hours per week!
The team attributed this to low-medium
The modified King’s algorithm no. 2 volume of production and gross inefficiencies
Step 1: For each column j, compute total prevalent in the current system. However,
weight of column j, j = 1, ..., n: idling could be politically the “third rail” and
249
Reconfiguring a manufacturing system for strategic advantage Logistics Information Management
Ashok Kumar and Jaideep Motwani Volume 11 · Number 4 · 1998 · 244–256
Std hrs per 100 Min. STD Avg. wkly Avg. wkly hours Accum. wkly hrs
051 940 271 225 order code usage 051 940 271 225 Total 051 940 271 225 Total
0.65 1 97422672
0.65 1 97422682
0.65 1 97422692
0.65 1 97422702
0.65 1 97422712
0.65 1 97422722
0.65 1 97422732
0.65 1 97422742
0.22 500 9701360 22 0.05 0.05 0.05 0.05
0.27 2,000 9701361 59 0.16 0.16 0.21 0.21
0.27 100 9701362 24 0.06 0.06 0.27 0.27
0.27 500 9701363 39 0.11 0.11 0.38 0.38
0.12 0.11 400 9701114 8 0.01 0.01 0.02 0.39 0.01 0.40
0.12 0.11 700 9701115 3 0.00 0.00 0.01 0.39 0.01 0.40
0.12 0.11 1,000 9701116 0.39 0.01 0.40
0.12 0.11 500 9701117 0.39 0.01 0.40
0.2 0.13 800 9701119 0.39 0.01 0.40
0.12 0.26 25 9773037 20 0.02 0.05 0.07 0.42 0.06 0.48
0.1 0.11 25 1780004 0.42 0.06 0.48
0.11 0.17 25 1780005 2 0.00 0.00 0.01 0.00 0.00 0.42 0.06 0.48
0.11 0.15 25 1780006 16 0.02 0.02 0.04 0.02 0.03 0.42 0.06 0.53
0.12 0.18 25 1780007 0.02 0.03 0.42 0.06 0.53
0.11 0.11 20 9700203 49 0.05 0.05 0.11 0.07 0.08 0.42 0.06 0.64
0.12 0.15 20 9700205 0.07 0.08 0.42 0.06 0.64
0.11 0.15 200 9700206 14 0.02 0.02 0.04 0.09 0.10 0.42 0.06 0.67
0.13 0.18 20 9700207 98 0.12 0.18 0.30 0.21 0.28 0.42 0.06 0.97
0.16 0.29 30 97421301 3 0.00 0.01 0.01 0.22 0.29 0.42 0.06 0.99
0.16 0.29 50 97421311 9 0.01 0.03 0.04 0.23 0.32 0.42 0.06 1.03
0.16 0.18 100 97421321 15 0.02 0.03 0.05 0.26 0.34 0.42 0.06 1.08
0.16 0.15 50 97421331 7 0.01 0.01 0.02 0.27 0.35 0.42 0.06 1.10
0.15 0.15 300 97421341 114 0.17 0.17 0.34 0.44 0.53 0.42 0.06 1.44
0.16 0.15 50 97421351 5 0.01 0.01 0.02 0.45 0.53 0.42 0.06 1.46
0.19 0.29 50 97421851 2 0.00 0.01 0.01 0.45 0.54 0.42 0.06 1.47
0.19 0.29 50 97421861 3 0.01 0.01 0.01 0.46 0.55 0.42 0.06 1.48
0.19 0.18 100 97421871 78 0.15 0.14 0.29 0.60 0.69 0.42 0.06 1.77
0.18 0.15 100 97421881 1 0.00 0.00 0.00 0.60 0.69 0.42 0.06 1.77
0.18 0.15 250 97421891 46 0.08 0.07 0.15 0.69 0.76 0.42 0.06 1.92
0.19 0.15 50 97421901 8 0.01 0.01 0.03 0.70 0.77 0.42 0.06 1.95
0.16 0.11 50 9742192 12 0.02 0.01 0.03 0.72 0.79 0.42 0.06 1.98
0.17 0.15 50 9742194 98 0.17 0.15 0.32 0.89 0.93 0.42 0.06 2.30
0.18 0.15 50 9742195 51 0.09 0.08 0.17 0.98 1.01 0.42 0.06 2.47
0.18 0.18 50 9742196 10 0.02 0.02 0.04 1.00 1.03 0.42 0.06 2.51
0.15 0.11 300 9742392 86 0.13 0.10 0.23 1.13 1.12 0.42 0.06 2.73
0.15 0.12 500 9742393 80 0.12 0.10 0.22 1.25 1.22 0.42 0.06 2.95
0.15 0.15 100 9742394 17 0.03 0.03 0.05 1.28 1.25 0.42 0.06 3.01
0.15 0.15 900 9742395 371 0.57 0.56 1.13 1.85 1.81 0.42 0.06 4.13
0.15 0.18 100 9742396 29 0.04 0.05 0.10 1.89 1.85 0.42 0.06 4.23
0.15 0.11 50 9742397 14 0.02 0.02 0.04 1.91 1.83 0.42 0.06 4.27
251
Reconfiguring a manufacturing system for strategic advantage Logistics Information Management
Ashok Kumar and Jaideep Motwani Volume 11 · Number 4 · 1998 · 244–256
• Routing 3. Shear → Deburr → Small brake floor area reduction, WIP reduction and lead
press. time reduction. As Table III demonstrates,
the changed configuration was projected to
At this point, the team proceeded to develop save $3154.48 per week or approximately
the new layout (Figure 2) from the existing $164,000 a year from this cell alone. Indeed, a
layout (Figure 3). total change for the entire 26 workstations was
estimated at $720,000 approximately. Notice-
ably, this figure does not include a significant
Cost benefit analysis strategic advantage to be rendered by reduc-
tion in the manufacturing cycle time of the
The analysis for the costs and benefits of the panels. The industrial engineering estimates
configuration change in the pilot cell is placed of cycle time reductions were quite dramatic –
in Table III. It will be noticed that only certain from eight to ten days to one to two days!
quantifiable factors were considered for the
purpose. The cost factors primarily arose
from movement of machines to the newly Concluding remarks
configured place. The benefits stemmed from In this paper, we described a real-life experi-
reduced material movement, setup reduction, ence in reconfiguring a manufacturing
252
Reconfiguring a manufacturing system for strategic advantage Logistics Information Management
Ashok Kumar and Jaideep Motwani Volume 11 · Number 4 · 1998 · 244–256
Cost Savings
Item calculation calculation Rationale
Costsa
Move small
brake press $3,500 Estimate by plant eng.
Move Deburr
machine $600 Estimate by plant eng.
Project management 15 hrs x $25/hr Methods eng. estimate
Additional labor
expense $375
Total cost $4,475
Benefits
Material
movement (339.69)*2 Saved distance in feet/hour*
*0.035/60*40*9.03 = 2 (each way) *rate*
$114 per week hours/week*labor rate
Hilo moves 13*/13.5*28 = Saved hilos/week/no. of hour
$27 per week per hilo* standard labor
charge/hour
Setup reduction 57* 15/60* 28 = no. of setups reductions
$399 (estimates b y plant eng.)*
average time reduction per
setup* indirect staff rate
Floor reduction 1,042*$4.50 = $639 Savings in space/sq.ft*cost
per savings/sq.ft of operation
week (corporate guidelines)
WIP reduction 30%*68*1,054*0.23 = Percent batch size reduction*
$6,450.48 average batch size*No. of
parts *average inventory carrying cost per
week
Lead time reduction – Strategic. No appropriate
formula available
Total benefits $7,629.48 per week
Note:
a = One-time expenditures
Boe, J.W. and Cheng, C.H. (1991 ), “A close neighbor Kaparth, S. and Suresh, N.C. (1992), “Machine-component
algorithm for designing cellular manufacturing cell formulation in group technology: a neural
systems”, International Journal of Production network approach”, International Journal of
Research, Vol. 29 No. 10, pp. 2,097-116. Production Research, Vol. 30 No. 6, pp. 1,353-67.
Burbidge, J.L. (1971), “Production flow analysis”, Produc- Khator, S.K. and Irani, S.A. (1987), “Cell formation in group
tion Engineering, Vol. 4, pp. 139-52. technology: a new approach”, Computers and
Burke, L. and Kamal, S. (1995), “Neural networks and the Industrial Engineering, Vol. 12 No. 2, pp. 131-42.
part family/machine group formation problem in King, J.R. (1980), “Machine-component grouping in
cellular manufacturing: a framework using fuzzy production flow analysis: an approach using the
ART”, Journal of Manufacturing Systems, Vol. 14 rank order clustering algorithm”, International
No. 3, pp. 148-59. Journal of Production Research, Vol. 18 No. 2,
Chan, H.M. and Milner, D.A. (1982), “Direct clustering pp. 213-32.
algorithm for group formulation in cellular manufac- Kusiak, A. (1991), “Branching algorithms for solving the
turing”, Journal of Manufacturing Systems, Vol. 1 group technology problem”, Journal of Manufactur-
No. 1, pp. 65-74.
ing Systems, Vol. 10 No. 4, pp. 332-43.
Chandrasekharan, M.P. and Rajagopalan, R. (1986), “An
Kusiak, A. and Cheng, C.H. (1990), “A branch-and-bound
ideal seed non-hierarchical clustering algorithm for
algorithm for solving the group technology prob-
group technology”, International Journal of Produc-
lem”, Annals of Operations Research, Vol. 26 pp.
tion Research, Vol. 11 No. 6, pp. 835-50.
415-31.
Chandrasekharan, M.P. and Rajagopalan, R. (1987),
Kusiak, A. and Chung, Y.K. (1991), “GT/ART: using neural
“ZODIAC – an algorithm for concurrent formation of
networks to form machine cells.” Manufacturing
part-families and machine-cells”, International
Journal of Production Research, Vol. 25 No. 6, Review, Vol. 4 No. 4, pp. 293-301.
pp. 835-50. Kusiak, A., Boe, J. and Cheng, C.H. (1993), “Designing
Cheng, C.H. (1993), “A tree search algorithm for designing cellular manufacturing systems: branch and bound
a cellular manufacturing system”, Omega, Vol. 21 and A* approaches”, IIE Transactions, Vol. 25 No. 4,
No. 4, pp. 489-96. pp. 46-56.
Cheng, C.H., Kumar, A. and Motwani, J. (1995), “A compar- Kusiak, A., Vannelli, A. and Kumar, R.K. (1986), “Clustering
ative examination of selected cellular manufacturing analysis: models and algorithms”, Cybernetics,
clustering algorithms”, International Journal of Vol. 15 No. 2, pp. 139-54.
Operations and Production Management, Vol. 15 McAuley, A. (1972), “Machine grouping for efficient
No. 12, pp. 86-97. production”, Production Engineering, Vol. 2, pp. 53-7.
Cheng, C.H., Kusiak, A. and Boe, J. (1991), “A branch-and- McCormick, W.T., Schweitzer, P.J. and White, T.W. (1972),
bound algorithm for solving the machine allocation “Problem decomposition and data organization by
problem”, in Leondes, C.T. (Ed.), Manufacturing and clustering technique”, Operations Research, Vol. 20,
Automated Systems: Techniques and Technologies, pp. 992-1,009.
Academic Press, New York, NY.
Massay, L., Udoka, S., Silvanus, J. and Nbenjamin, C.
Chow, W. and Hawaleshka, O. (1993), “A novel machine (1995), “A simulator-based approached to cellular
grouping and knowledge-based approach for manufacturing system design”, Computers and
cellular manufacturing”, European Journal of Industrial Engineering, Vol. 29 No. 1-4, pp. 327-31.
Operational Research, Vol. 69 No. 3, pp. 357-72.
Offodile, O.F. (1993), “Assignment model formulation of
Durmusoglu, M. (1993), “Analysis of the conversion from a
the machine cell formation problem in cellular
job shop system to a cellular manufacturing
manufacturing”, International Journal of Operations
system”, International Journal of Production
and Production Management, Vol. 13 No. 10,
Economics, Vol. 30 No. 31, pp. 427-36.
pp. 49-59.
Ganesh, M.V. and Srinivasan, G. (1994), “A heuristic
Rajamani, D., Singh, N. and Aneja, Y. (1992), “Selection of
approach for the cell formation problem”, Comput-
parts and machines for cellularization: a mathemati-
ers and Industrial Engineering, Vol. 26 No. 1,
pp. 193-201. cal programming approach”, European Journal of
Operational Research, Vol. 62 No. 1, pp. 47-54.
Gunasingh, R.K. and Lashkari, R.S. (1989a), “Machine
grouping problem in cellular manufacturing systems Rajgopalan, R. and Batra, J.L. (1975), “Design for cellular
– an integer programming approach”, International production systems: a graph theoretic approach”,
Journal of Production Research, Vol. 27 No. 9, International Journal of Production Research,
pp. 1,465-73. Vol. 13, pp. 567-79.
Gunasingh, R.K. and Lashkari, R.S. (1989b), “The cell Seifoddini, H. and Wolfe, P.P. (1986), “Application of the
formation problem in cellular manufacturing similarity coefficient method in group technology”,
systems: a sequential modelling approach”, Com- lIE Transactions, Vol. 18 No. 3, pp. 271-7.
puters and Industrial Engineering, Vol. 16, Shafer, S.M. and Bradford (1995), “Efficient measurement
pp. 469-76. of alternative machine component grouping solu-
Hayes, R.H. and Wheelwright, S. (1979), “Link manufactur- tions via data envelopment analysis”, IEEE Transac-
ing process and product life cycles”, Harvard tions on Engineering Management, Vol. 42 No. 2,
Business Review, January-February, pp. 133-40. pp. 159-66.
255
Reconfiguring a manufacturing system for strategic advantage Logistics Information Management
Ashok Kumar and Jaideep Motwani Volume 11 · Number 4 · 1998 · 244–256
Shafer, S.M., Kern, G.M. and Wei, J.C. (1992), “A mathe- Vannelli, A. and Ravi Kumar, K. (1986), “A method for
matical programming approach for dealing with finding minimal bottleneck cells for grouping
exceptional elements in cellular manufacturing”, partmachine families”, International Journal of
International Journal of Production Research, Vol. 30 Production Research, Vol. 24 No. 2, pp. 387-400.
No. 5, pp. 1,029-36.
Venugopal, V. and Narendran, T.T. (1992), “Cell forma-
Shargal, M., Shekhar, S. and Irani, S. (1995), “Evaluation of tion in manufacturing systems through simulated
search algorithms and clustering efficiency mea- annealiing: an experimental evaluation”, Euro-
sures for machine-part matrix clustering”, IIE
pean Journal of Operational Research, Vol. 63,
Transactions, Vol. 27 No. 1, pp. 43-59.
pp. 409-22.
Skinner, W. (1969), “Manufacturing – missing link in corpo-
rate strategy”, Harvard Business Review, May-June, Waghodekar, P.H. and Sahu, S. (1984), “Machine-compo-
pp. 136-45. nent cell formation in group techology: MACE”,
International Journal of Production Research, Vol. 22
Skinner, W. (1974), “The focused factory”, Harvard Business
No. 6, pp. 937-48.
Review, May-June, pp. 113-21.
Slagle, J.L., Chang, C.L. and Heller, S.R. (1975), “A cluster- Wei, J.C. and Gaither, N. (1990), “An optimal model for cell
ing and data reorganization algorithm”, IEEE formation decisions”, Decision Sciences, Vol. 21
Transactions on Systems, Man, and Cybernetics, No. 2, pp. 416-33.
Vol. SMC-5 No. 2, pp. 125-8. Zhu, Z., Heady, R. and Reiner, S. (1995), “An efficient zero-
Tam, K.Y. (1990), “An operation sequence based similarity one formulation of the cell formation problem”,
coefficient for part families formations”, Journal of Computers and Industrial Engineering, Vol. 28 No. 4,
Manufacturing Systems, Vol. 9 No. 1, pp. 55-68. pp. 911-16.
256