Assessment of clustering tendency for the design of cellular manufacturing systems

Purpose - Without reliance on results obtained from applying a cell formation method, this paper aims to describe a simple quantitative approach to testing whether an underlying pattern of relationships exists between machines of a given system, such that the machines may be rearranged into manufacturing cells. It also aims to support the approach by an index for measuring the clustering tendency. Design/methodology/approach - The eigenvalues of the similarity coefficient matrix and Kaiser's rule are used to: detect the number of clusters existing in the part-machine matrix, and derive an index for predicting the goodness of the best possible obtainable cell formation. Findings - The results of applying the proposed approach and the clustering tendency index to problems of different sizes taken from the literature have proven that both the approach and the clustering tendency index are powerful in performing the feasibility assessment and in predicting the right number of manufacturing cell to be formed. Practical implications - This study is of considerable value to practitioners because it provides them with a powerful yet very easy to apply approach for assessing the feasibility of adopting cellular manufacturing in early stages of design. Another characteristic of this approach is the possibility of using it as a decision support tool for practitioners who opt to use a cell formation method which requires specifying the number of cells in advance. Moreover, the approach does not require any special software package, since it can be easily performed using several available software packages such as MATLAB and Mathematica. Originality/value - A methodology for evaluating the adaptability of a system to cellular manufacturing has been proposed in a previous study. However, the methodology used is complex and uses a certain degree of subjectivity. In contrast, the proposed approach is simple and completely quantitative. Furthermore, a new index for measuring the clustering tendency is presented.