Application of Cluster Analysis to Cellular Manufacturing

Abstract

The primary motivation for adopting cellular manufacturing is the globalization and intense competition in the current marketplace. The initial step in the design of a cellular manufacturing system is the identification of part families and machine groups and forming manufacturing cells so as to process each part family within a machine group with minimum intercellular movements of parts. One methodology to manufacturing cells is the use of similarity coefficients in conjunction with clustering procedures. In this chapter, we give a comprehensive overview and discussion for similarity coefficients developed to date for use in solving the cell formation problem. Despite previous studies indicated that the similarity coefficients-based method (SCM) is more flexible than other cell formation methods, none of the studies has explained the reason why SCM is more flexible. This chapter tries to explain the reason explicitly. To summarize various similarity coefficients, we develop a taxonomy to clarify the definition and usage of various similarity coefficients in designing cellular manufacturing systems. Existing similarity (dissimilarity) coefficients developed so far are mapped onto the taxonomy. Additionally, production information based similarity coefficients are discussed and a historical evolution of these similarity coefficients is outlined. Although many similarity coefficients have been proposed, very fewer comparative studies have been done to evaluate the performance of various similarity coefficients. In this chapter, we compare the performance of twenty well-known similarity coefficients. More than two hundred numerical cell formation problems, which are selected fromthe literature or generated deliberately, are used for the comparative study. Nine performance measures are used for evaluating the goodness of cell formation solutions. Two characteristics, discriminability and stability of the similarity coefficients are tested under different data conditions. From the results, three similarity coefficients are found to be more discriminable; Jaccard is found to be the most stable similarity coefficient. Four similarity coefficients are not recommendable due to their poor performances.