Enhancing operational research in mechatronic systems via modularization: comparative analysis of four clustering algorithms using validation indices

• Published in: Operational research Vol. 24; no. 4; p. 63
• Main Authors: Mikrou, Ioannis, Sapidis, Nickolas S.
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.12.2024; Springer Nature B.V
• Subjects: Algorithms; Business and Management; Clustering; Comparative analysis; Complex systems; Complexity; Computational Intelligence; Cost benefit analysis; Density; Management Science; Mechatronics; Modularization; Operations Research; Operations Research/Decision Theory; Original Paper; Validation studies; Operation research software; Clustering validation techniques; Mechatronics; DSM models; Clustering
• ISSN: 1109-2858; 1866-1505
• DOI: 10.1007/s12351-024-00872-3

Modularization is one of the most robust methods that industries use to profit. This technique allows Operational Research to manage complex systems by efficiently dividing them into smaller ones and thus lowering the affiliated risks and costs. Mechatronic products are complex systems associated with diverse disciplines, laborious to compose and decompose, and can benefit from modularization. In this research, Partitioning Around Medoids (PAM), Ward’s method, Divisive ANAlysis (DIANA), and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithms are utilized in combination with Design Structure Matrices (DSM) to cluster 175 test subjects, and their results are compared using four validation techniques. Agglomerative Coefficient (AC), Divisive Coefficient (DC), Silhouette Coefficient (SC), Composed Density between and within clusters (CDbw), and the visual inspection of two-dimensional representations of each algorithm's clustering results are the validation techniques used in this research to find the most suitable algorithm for clustering such intricate systems. Additionally, other data that emerged from this research, such as time complexity, total execution time, and average RAM usage, are also used to evaluate the overall performance of each clustering algorithm.