Two-mode modularity clustering of parts and activities for cell formation problems

Cell formation in cellular manufacturing is a critical step to improving productivity by grouping parts and machines. Numerous heuristic search algorithms and several performance measures have been used in finding an effective cell formation solution. It is still a challenging task to find a good cell formation that satisfies several performance measures. Clustering approaches aim to find good clusters of parts and machines according to their own similarity measures. We propose a two-mode modularity clustering method with new similarity measures for parts and machines using an ordinal part-machine matrix. The proposed method considers both incidence and transition among parts and machines and can find an optimal number of clusters. We demonstrate the effectiveness of the proposed method using cell formation problems in comparison with a few existing ones. The result shows that the proposed method produces good cell formation solutions in terms of several performance measures. In addition, we show a possible application area of the proposed method in process mining, using it to find interpretable clusters of processes and activities from real-life event log data.