Machine-part grouping and cluster analysis: similarities, distances and grouping criteria

Abstract

Machine-part grouping and cluster analysis: similarities, distances and grouping criteria The paper considers the machine-part grouping problem, as equivalent to partitioning the set of machines and operations into subsets, corresponding to block diagonalisation with constraints. The attempts to solve the problem with clustering methods are outlined. The difficulties encountered are presented, related to (i) ambiguity of formulations; (ii) selection of criteria; and (iii) lack of effective algorithms. These are illustrated in more detail with a limited survey of similarity and distance definitions, and of criteria used, constituting the main body of the paper. The return is proposed to the basic paradigm of cluster analysis, as providing simple and fast algorithms, which, even if not yielding optimal solutions, can be controlled in a simple manner, and their solutions improved.