Reduction of Handing Energy and Waste of Materials Through Optimized Manufacturing Cell

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Reduction of Handing Energy and Waste of

Materials Through optimized Manufacturing Cell


Tamal Ghosh and Sourav Sengupta

Department of Industrial Engineering & Management, West Bengal University of Technology, Kolkata, India.
BF 142, Salt Lake City, Kolkata 700064, India. Tel. +91-33-2334-1014.
tamal.31@gmail.com, sengupta.sourav86@gmail.com

Abstract— The present era is more concerned about sustainable manufacturing, which can
be facilitated by means of lean production process. Cellular manufacturing (CM) has
acquired an upward interest in recent lean production approaches. By implementing cellular
manufacturing structure on shop-floor can eventually reduce the overall production cost and
enhance the throughput of the system which is the most important part of today’s lean
manufacturing approach. The most important problem in designing cellular manufacturing
systems is the cell formation, which is concerned with formation of part families with similar
processing requirements and designated machines into machine cells in order to optimize the
production process. This approach further reduces the handling energy of materials and also
reduces the waste generates in the process. This paper presents a hybrid approach to the cell
formation problem in CM. The proposed hybrid approach is based on Sorenson’s similarity
coefficient and weighted average linkage clustering (WALC) technique to construct optimized
machine cells. Further a modified part clustering heuristic is embedded with the proposed
technique to facilitate the cell formation procedure. The proposed technique is tested on
published datasets of literature and the experimental results presented in this study signify
that the proposed method is quite effective for small to medium size problems. This article is
a substantial addition in the on-going research in CM and lean manufacturing and could be
cooperative to the future researchers and practitioners in the aforementioned domain.

Keywords—cellular manufacturing, lean production, machine-part grouping, hybrid


clustering analysis, similarity coefficient

I. INTRODUCTION
Manufacturing business is under extreme pressure from the modest global marketplace.
Briefer product life-cycles, impulsive demands, and varied customer needs have compelled
manufacturing firms to function more competently and realistically in order to familiarize to
varying necessities. Traditional manufacturing systems, such as job shops and flow shops,
cannot manage such environments. Lean production process is essential in this context which
takes the advantages of cellular manufacturing (CM), which further incorporates the
tractability of job shops and the high production rate of flow shops, has materialized as a
viable replacement for such cases [1]. Reducing handling cost of materials, i.e. minimizing
handling energy and waste are significant to reduce the throughput time and overall cost
involved in entire production process [2, 3]. CM is the application of the concept of group
technology (GT) in manufacturing systems. GT is a manufacturing philosophy that utilizes
similarities in product design and production processes. A primitive concern in CM is the
determination of the part families and machine cells. This problem is known as the cell
formation (CF) problem. The CF problem involves the breakdown of a manufacturing system
into cells. Part families are recognized such that they could be processed within a cell. The
cells are formed to take the advantages of GT such as reduced setup times, reduced in-process
inventories, improved product quality, shorter lead times, reduced tool requirements,
improved productivity, and better overall control of operations [4]. The CF problem has long
been identified as the most tricky problem in grasping the concept of cellular manufacturing.
It fits in the class of NP-complete problems [5], which means that an intensification in the
problem size will cause an exponential expansion in the computational time for all
established optimization techniques. Due to such nature of the problem, many computational
techniques are heavily practised for improved solution to the CFP, a thorough discussion can
be found in literature [6].
Over the last few decades many clustering methods are proposed by researchers to solve
CF problem, such as ZODIAC [7], GRAFICS [8], minimum spanning tree (MST) [9], K-
Harmonic Mean algorithm [5], Fuzzy C-Means algorithm [10] etc.
In this article a new clustering approach is proposed to solve CF problem by utilizing
weighted average linkage clustering method and modified part grouping heuristic technique
which are hybridized to form efficient cell formation algorithm.

II. SURVEY OF LITERATURE


Various techniques are developed to solve manufacturing cell formation problems since
last four decades, these include similarity coefficient methods, clustering analysis, array
based techniques, graph partitioning methods etc. The similarity coefficient approach was
first implemented by McAuley [11]. The foundation of similarity coefficient methods is to
measure the similarity between each pair of machines and subsequently group the machines
into cells based on their similarity measurements. Most similarity based methods utilize
machine–part mapping matrices. Some of the methods, which practiced this approach, are
Single linkage clustering algorithm [11], Average linkage clustering algorithm [12].
Clustering methods are categorized as hierarchical and non-hierarchical methods. Orthodox
or particularly designed clustering techniques could be utilized to develop clusters of either
components or machines. Machine–part grouping analysis is based on production flow
analysis, in which the machine-part groups are formed by rearranging rows and columns of
the machine-part mapping matrix in the form of a {0-1} incidence matrix. Some of the
methods are Rank order clustering [13], Bond energy algorithm [14] etc. Dimopoulos and
Mort has stated a hierarchical algorithm combined with genetic programming for a simple
cell formation problem [15]. Array based methods consider the rows and columns of the
machine-part incidence matrix as binary patterns and reconfigure them to obtain a block
diagonal cluster formation. The rank order clustering algorithm is the most familiar array-
based technique for cell formation [13]. Substantial alterations and enhancements over rank
order clustering algorithm have been described in [16, 17]. The direct clustering analysis has
been developed in [18] and bond energy analysis is proposed in [14]. Graph Theoretic
Approach depicts the machines as vertices and the similarity between machines as the
weights on the arcs. [19] proposed the use of graph theory to form machine cells.
Chandrasekharan and Rajagopalan proposed an ideal seed non-hierarchical clustering
algorithm for cellular manufacturing [17]. Graph searching algorithms was demonstrated in
[20] which select a key machine or part according to a pre-fixed criterion. A non-heuristic
network approach was stated in [21] to construct manufacturing cells with minimum inter-
cellular moves. Srinivasan implemented a method using minimum spanning tree for the
machine-part cell formation problem [9]. A polynomial-time algorithm based on a graph
theoretic approach for optimal cell formation was shown in [22], named as vertex-tree
graphic matrices.
III. CELL FORMATION PROBLEM
The cell formation problem in group technology begins with two fundamental tasks,
namely, machine-cell formation and part-family identification. To form machine-cell
dissimilar machines are grouped and they are dedicated for the manufacture of one or more
part-families. In part-family formation, parts with similar design features, attributes, shapes
are grouped, so that the group of parts can be manufactured within a cell. Generally, the cell
formation problems are represented in a matrix namly ‘machine-part incident matrix’ in
which all the elements are presented as either 0 or 1. Parts are arranged in cloumns and
machines are in row in the incidence matrix. An example matrix is presented in Fig. 1. It
depicts that part 2 is processed through machine 1, part 1, 4 are processed through machine 2,
part 3 is processed through machine 1, 3 and part 5 is processed through machine 1, 2, 3. In
this matrix a 0 indicates no mapping or no processing and an 1 indicates mapping or
processing.

Fig. 1 Machine-part incidence matrix (3×5)


From Fig. 1 solution matrix can be obtained after applying machine-part clustering technique
to form block diagonal structure as square boxes (Fig. 5). An 1 outside the block means a part
which is processed through some machine which does not belong to the corresponding
machine cell, i.e. bottleneck machine, therefore the intercellular move energy and cost will be
added. This element is known as an exceptional element (EE) and a 0 inside a cell means an
unutilized space in cell, therefore lesser utilization of space an increase in intracell handling
cost and energy. It is known as ‘void’. The objective of cell formation is to minimize the EEs
and voids in order to reduce the handling energy and material waste to maximize the overall
profit of the firm.

IV. SOLUTION METHODOLOGY


A. Similarity Coefficient Method
Similarity coefficient based techniques are massively practiced in formation of
manufacturing cells and a comprehensive study can be accomplished in [23]. In this article
the similarity measure method is utilized namely Sorenson’s similarity coefficient [23].

2𝑎𝑖𝑗
𝑆𝑖𝑗 = (1)
2𝑎𝑖𝑗 + 𝑏𝑖𝑗 + 𝑐𝑖𝑗

Sij = Similarity between machine i and machine j,


aij = the number of parts processed by both machines i and j,
bij = the number of parts processed by machine i but not by machine j,
cij = the number of parts processed by machine j but not by machine i.

This abovementioned similarity metric technique is utilized in this study to calculate the
similarity coefficient value between pair of machines presented as rows of machine-part
incidence matrix of Fig. 1. To minimize the computation in Matlab the similarity matrix is
transformed into distance matrix by using equation (2),
𝑑𝑖𝑗 = 1 − 𝑆𝑖𝑗 (2)

Fig. 2 shows the dissimilarity/distance matrix obtained from the example problem given in
Fig. 1.

Fig. 2 Similarity matrix of example problem (3×5)


B. Proposed Machine clustering technique
Hierarchical linkage clustering techniques have an affinity to group machines which are
basically dissimilar in nature due to the chaining problem they suffer from [11, 15]. That is
the reason why researchers are interested in improving the linkage technique. To overcome
this flaw McAulay proposed average linkage clustering algorithm for CF problem [15]. In
this research an Weighted Average Linkage Clustering model is introduced in the next stage
of the algorithm as a cure to this problem. Weighted Average Linkage Clustering Algorithm
(WALCA) is conceptually and mathematically simple algorithm practiced in hierarchical
clustering analysis of data [26]. It delivers informative descriptions and visualization of
potential data clustering structures. When there exists hierarchical relationship in data this
approach can be more competent. In this article a hybrid WALCA approach is adopted,
which is combined with Sorenson’s correlation coefficient to facilitate efficient cell
formation. Weighted average linkage uses a recursive definition for the distance between two
clusters. If cluster r is to be created by combining clusters p and q, the distance between r and
another cluster s is defined as the average of the distance between p and s and the distance
between q and s and it is given as:

(𝑑(𝑝, 𝑠) + 𝑑(𝑞, 𝑠))


𝑑(𝑟, 𝑠) = (3)
2

A matrix is generated using (3), which is a (m-1)×3 matrix, where m is the number of
machines in the original dataset.
machine-machine dendrogram

0.6

0.5
distances

0.4

0.3

0.2
1 3 2
machines
Fig. 3 Dendrogram of machine grouping of example problem (3×5)

Columns of the matrix contain cluster indices linked in pairs to form a binary tree. The leaf
nodes are numbered from 1 to m. Leaf nodes are the singleton clusters from which all higher
clusters are built. Further The dendrogram can be obtained from the matrix which indicates a
tree of potential solutions. Therefore it’s a decision making job to decide how to obtain a
particular group of machines based on pre-selected similarity threshold. Dendrogram
structure obtained for the machines is shown in Fig. 3, which clearly states that two cells are
formed, cell 1 contains machine 1 and 3 and cell 2 contains machine 2. Initial machine cell
formation is depicted in Fig. 4.

Fig. 4 Machine cluster obtained for (3×5)


C. Proposed Part Family Formation Approach
In order to construct efficient part families, a novel technique is proposed in this study
which is primarily inspired from past literature [38] and modified substantially. This method
is helpful to evaluate the cell formation obtained via the proposed hybrid method. This
phenomena is based on identifying a machine cell which processes the part for a maximum
number of operations than any other machine cell and assigning the corresponding part into
that cell. Therefore parts are assigned to the cells which further form tangible part families
using membership index given as,

𝑚𝑐𝑗 𝑚𝑐𝑗 1
𝐷𝑐𝑗 = × × (4)
𝑘𝑐 𝑛𝑗 𝑣

Dcj = Membership index of part j to cell c


mcj = Number of machines in cell c which process part j
kc = Total number of machines in cell c
nj = Total number of machines required by part j
v = Total number of zeros (voids) in the cells in obtained block diagonal matrix

In the above mathematical formula the count of voids has been introduced by us, which
implies the number of zeros in the cells. Using (4) the membership index value of each part
can be computed. Larger the membership index value of a part for a particular cell, will
subsequently assign it to that cell to obtain the final part families. The computed membership
index values for Figure 4, are depicted in Table 1.

TABLE I
COMPUTING THE MEMBERSHIP INDEX VALUES FOR PARTS

Parts Membership Index Values


Cell 1 Cell 2
P1 0 1
P2 0.5 0
P3 1 0
P4 0 1
P5 0.667 0.333

The above analysis illustrates the final cell formation, p1, p3, p5 are grouped into part family
1 and p2, p4 are grouped into part family 2 and they are assigned to cell 1 and cell 2
respectively. Therefore the final block diagonal structure is obtained and presented in Fig. 5.
The flow chart of the proposed technique is presented in Fig. 6.

Fig. 5 Final block diagonal structure (3×5)

START

Input the machine-


part matrix

Similarity coefficient method


Apply Sorenson’s
coefficient method

Form similarity matrix using


similarity values

Utilize weighted ALC


method
Machine cell formation

Generate Dendrogram

Obtain machine
groups/ cells

Compute the membership index


values of parts
Part Family Formation

Assign the correct part to the


correct cell according to the
membership index value

Obtain the final block


diagonal matrix

END

Fig. 6 Flow chart of the proposed methodology


D. Performance Measure Used
Two widely accepted performance measures to assess the goodness of CFP solutions are
grouping efficiency [17] which incorporates machine utilization and intercell moves, and the
grouping efficacy [25], which intends to minimize the number of exceptional elements and
the number of voids in the diagonal blocks. A detailed description regarding various
performance measure could be obtained from a critical survey of reference [24]. In this study
grouping efficacy measure is used as the evaluation criterion to test the effectiveness of the
proposed algorithm. Grouping efficacy measure is stated as,

E − E𝑒 E𝑣 + E𝑒
τ = =1− (5)
E + E𝑣 E + E𝑣

Where
E = Total number of 1s in incidence matrix
Ee = Total number of exceptional elements
Ev = Total number of voids

Therefore the final machine-part block diagonal structure of Fig. 5 depicts that,
Cell 1= {Machine 1, 3 || Parts 2, 3, 5} and Cell 2= {Machine 2 || Parts 1, 4}
Number of Exceptional elements = 1; Number of voids = 1
Grouping efficacy measure = (8-1) / (8+1) = 7/9 = .7778 Which is 77.78%.

V. RESULT AND DISCUSSION


The proposed hybrid technique is tested with a set of 12 problems that have been published
in the literature. All the data sets were transcribed from the original articles to avoid the
inconsistencies in data. The proposed hybrid technique is simulated with Matlab 7.0 and
tested on a laptop with a 2.1GHz processor and 2GB of RAM. Comparisons of the proposed
hybrid technique against the best results obtained from the literature are given in Table II.
Past Results are obtained from CM literature [5, 27].
TABLE III
COMPARISON AMONG THE RESULTS OBTAINED FROM HYBRID TECHNIQUE WITH PUBLISHED RESULTS
OBTAINED FROM LITERATURE
# Dataset size Grouping Efficacy Value cell EE void
Best result found Hybrid method
in literature
1 Waghodekar and Sahu (1984) [28] 5×7 62.5 69.56** 2 6 2
2 King and Nakornchai (1982) [16] 5×7 73.68 73.68 2 2 3
3 Seifoddini (1989) [29] 5×18 79.59 79.59 2 7 3
4 Kusiak and Cho (1992) [30] 6×8 76.92 76.92 2 2 4
5 Boctor (1991) [31] 7×11 70.37 70.37 3 2 6
6 Kusiak and Chow (1987) [32] 7×11 53.13 59.26 4 7 4
7 car and Mikac (2006) [33] 8×10 66.67 68.96 4 7 2
8 Chandrasekharan and Rajagopalan (1986) [17] 8×20 85.24 86.67 3 8 0
9 Chu and Hayya (1991) [34] 9×9 73.53 74.28 3 6 3
10 Goncalves and Resend (2004) [35] 15×12 86.67 86.67 4 0 6
11 Jayakrishnan and Narendran (1998) [36] 20×8 83.87* 82.25 3 9 1
12 Masnata and settineri (1997) [37] 25×10 63.93 70.27 4 9 13
* inconsistent result shown in [5], actual computed value is 82.25
** better result shown in boldface

For the problems solved with proposed hybrid technique to obtain optimal solution, the
grouping efficacy value is better or equal in all instances. All the outputs are obtained with
negligible computational time (< 15 sec.). This observation indicates that this hybrid
technique is very efficient and less complex because of its simplicity in simulation. The
improvement shown by the proposed method is presented in Fig. 7. The proposed hybrid
technique is shown to outperform the standard techniques in 6 instances, and equal in 6
instances, which further depicts 50% improved result than published result which is
significant in terms of solution quality and time and space complexities.

Fig. 7 improvement shown by proposed method over published result

VI. CONCLUSIONS
This paper proposes a hybrid clustering technique that combines Sorenson’s similarity
coefficient method with weighted average linkage clustering technique and modified part
grouping method. Experimental results presented in Section 5 demonstrate that the proposed
hybrid technique outperforms the other techniques, and delivers improved results in
comparison with the published results. This article states that, by combining the similarity
coefficient approach into a traditional hierarchical clustering technique can improve the
solution quality substantially, and inclusion of modified part grouping method would enhance
the clustering result but it also reduces the variability of the solutions obtained. The proposed
hybrid technique obtains better quality solutions by consuming lesser computational time and
resources than that of the traditional complex soft computing based methodologies. It is also
shown that the proposed hybrid technique performs at least as well as, and often better than
the available algorithms for the cell formation on all problems tested. Therefore it is verified
as a promising method in aforestated area. Further work can be done by utilizing this
technique in large-scale and complex cell formation problem which deals with ratio data of
production volume, operational time, worker assignment by considering multi-objective
factors. Future work can also be done by combining soft computing techniques in the
proposed hybrid technique, which would be the further extension of present research.

REFERENCES
[1] Burbidge, J.L., “Production flow Analysis”, Production Engineer, Vol.42, No.12, 1963,
pp. 742-752.
[2] Burbidge, J.L., “Production flow Analysis”, Production Engineer, Vol.50, . 1971, pp.
139-152.
[3] Burbidge, J.L., Production Flow Analysis For Planning Group Technology, Oxford
University Press, USA, 1996.
[4] Wemmerlöv, U., & Hyer, N.L., “Research issues in cellular manufacturing”,
International Journal of Production Research, Vol. 25, 1987, pp. 413–431.
[5] Unler, A., & Gungor, Z., “Applying K-harmonic means clustering to the part-machine
classification problem”, Expert Systems with Applications, Vol.36, 2009, pp. 1179–1194.
[6] Ghosh, T., Sengupta, S., Chattopadhyay, M., & Dan, P.K., “Meta-heuristics in cellular
manufacturing: A state-of-the-art review”, International Journal of Industrial Engineering
Computations, Vol.2, No.1, pp.87-122.
[7] Chandrasekharan, M.P., & Rajagopalan, R., “ZODIAC: an algorithm for concurrent
formation of part-families and machine-cells”, International Journal of Production
Research, Vol.25, No.6, 1987, pp. 835–50.
[8] Srinivasan, G. & Narendran, T.T., “GRAFICS: a non-hierarchical clustering algorithm
for group technology”, International Journal of Production Research, Vol.29, No.3, 1991,
pp. 463-478.
[9] Srinivasan, G., “A clustering algorithm for machine cell formation in group technology
using minimum spanning trees”, International Journal of Production Research, Vol.32,
1994, pp. 2149–58.
[10] Lozano, S., Dobado, D., Larraneta, J., Onieva L., “Modified fuzzy C-means algorithm
for cellular manufacturing”, Fuzzy Sets and Systems, Vol.126, 2002, pp. 23–32.
[11] McAuley, J., “Machine grouping for efficient production”, Production Engineer, Vol.51,
No.2, 1972, pp. 53-57.
[12] Seifoddini, H., & Wolfe, P.M., “Application of the similarity coefficient method in group
technology”, IIE Transactions, Vol.183, 1986, pp. 271-277.
[13] King, J.R., “Machine-component grouping in production flow analysis: an approach
using a rank order-clustering algorithm”, International Journal of Production Research,
Vol.18, 1980, pp. 213-232.
[14] McCornick, W. T., Schweitzer, J.P.J., & White, T.W., “Problem decomposition and data
reorganization by a clustering technique”, Operations Research, Vol.20, No.5, 1972, pp.
993-1009.
[15] Dimopoulos, C., & Mort, N., “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, 2001, pp. 1-19.
[16] King, J.R., & Nakornchai, V., “Machine-component group formation in group
technology: review and extension”, International Journal of Production Research, Vol.20,
1982, pp. 117-133.
[17] Chandrasekharan, M.P., & Rajagopalan, R., “An ideal seed non-hierarchical clustering
algorithm for cellular manufacturing”, International Journal of Production Research,
Vol.24, No.2, 1986, pp. 451-464.
[18] Chan, H.M., & Milner, D.A., “Direct Clustering algorithm for group formation in
Cellular Manufacturing”, Journal of Manufacturing Systems, Vol.1, 1982, pp. 65-75.
[19] Rajagopalan, R., & Batra, J.L., “Design of Cellular production system –A graph theoretic
approach”, International Journal of Production Research, Vol.13, No.6, 1975, pp. 567-
579.
[20] Ballakur, A., & Steudel, H.J., “A with-in cell utilization based heuristic for designing
cellular manufacturing systems”, International Journal of Production Research, Vol.25,
No.5, 1987, pp. 639-655.
[21] Vohra, T., Cheng, D., & Chang, J.L., “A network model for cell formation in CMS”,
International Journal of Production Research, Vol.28, No.11, 1990, pp. 2075-2084.
[22] Veeramani, D., & Mani, K., “A polynomial-time algorithm for optimal clustering in a
special class of {0, 1}-matrices”, International journal of production research, Vol.34,
No.9, 1996, pp. 2587-2611.
[23] Yin, Y. & Yasuda, K., “Similarity coefficient methods applied to the cell formation
problem: a comparative investigation”, Computers & Industrial Engineering, Vol. 48,
2005, pp. 471–489.
[24] Sarker, B.R., Mondal, S., “Grouping efficiency measures in cellular manufacturing: a
survey and critical review”, International Journal of Production Research, Vol.37, No.2,
1999, pp. 285-314.
[25] Kumar, S.C., Chandrasekharan, M.P., “Grouping efficacy: a quantitative criterion for
goodness of block diagonal forms of binary matrices in group technology”, International
Journal of Production Research, Vol.28, No.2, 1990, pp. 233-243.
[26] Anderberg, M.R., Cluster Analysis for Applications, Academic Press Inc., New York,
1973.
[27] James, T.L., Brown, E.C., & Keeling, K.B., “A hybrid grouping genetic algorithm for the
cell formation problem”, Computers & Operations Research, Vol.34, 2007, pp. 2059–
2079.
[28] Waghodekar, P.H., & Sahu, S., “Machine-component cell formation in group
technology”, International Journal of Production Research, Vol.22, No.6, 1984, pp. 937-
948.
[29] Seifoddini, H., “Duplication process in machine cells formation in group technology”,
IIE Transactions, Vol.21, 1989, pp. 382-388.
[30] Kusiak, A., & Cho, M., “Similarity coefficient algorithm for solving the group
technology problem”, International Journal of Production Research, Vol.30, No.11, 1992,
pp. 2633–2646.
[31] Boctor, F.F., “A linear formulation of the machine-part cell formation problem”,
International Journal of Production Research, Vol.29, No.2, 1991, pp. 343–356.
[32] Kusiak, A., & Chow, W.S., “Efficient solving of the group technology problem, Journal
of Manufacturing Systems, Vol.6, No,2, 1987, pp. 117–124.
[33] Car, Z., & Mikac, T., “Evolutionary approach for cell formation problem in cell
manufacturing”, Advanced Engineering Informatics, Vol.20, 2006, pp. 227–232.
[34] Chu, A.H., Hayya, J.C., “A fuzzy clustering approach to manufacturing cell formation”,
International Journal of Production Research, Vol.29, No.7, 1991, pp. 1475–1487.
[35] Goncalves, J., Resende, M. “An evolutionary algorithm for manufacturing cell
formation”, Computers & Industrial Engineering, Vol.47, 2004, pp. 247–73.
[36] Jayakrishnan, G.N., & Narendran, T.T., “CASE: A clustering algorithm for cell
formation with sequence data”, International Journal of Production Research, Vol.36,
No.1, 1998, pp. 157–179.
[37] Masnata, A., & Settineri, L., “An application of fuzzy clustering to cellular
manufacturing”, International Journal of Production Research, Vol. 35, No.4, 1997, pp.
1077–1094.
[38] Zolfaghari, S., & Liang, M., “A new genetic algorithm for the machine/part grouping
problem involving processing times and lot sizes”, Computers and Industrial engineering,
Vol. 45, 2003, pp. 713-731.

You might also like