An Evolutionary Clustering-Based Optimization to Minimize Total Weighted Completion Time Variance in a Multiple Machine Manufacturing System

• Published in: International journal of information technology & decision making Vol. 14; no. 5; pp. 971 - 991
• Main Authors: Mokhtari, Hadi, Salmasnia, Ali
• Format: Journal Article
• Language: English
• Published: Singapore World Scientific Publishing Company 01.09.2015; World Scientific Publishing Co. Pte., Ltd
• Subjects: Algorithms; Clustering; Completion time; Evolutionary; Evolutionary algorithms; Iterative methods; Optimization; Production scheduling; Search algorithms; Search methods; Searching; Strategy; Subspaces; parallel system; Evolutionary clustering search; neighborhood search algorithm; reducible processing times
• ISSN: 0219-6220; 1793-6845
• DOI: 10.1142/S0219622015500200

This paper discusses clustering as a new paradigm of optimization and devises an integration of clustering and an evolutionary algorithm, neighborhood search algorithm (NSA), for a multiple machine system with the case of reducible processing times (RPT). After the problem is formulated mathematically, evolutionary clustering search (ECS) is devised to reach the near-optimal solutions. It is a way of detecting interesting search areas based on clustering. In this approach, an iterative clustering is carried out which is integrated to evolutionary mechanism NSA to identify which subspace is promising, and then the search strategy becomes more aggressive in detected areas. It is interesting to find out such subspaces as soon as possible to increase the algorithm's efficiency by changing the search strategy over possible promising regions. Once relevant search regions are discovered by clustering they can be treated with special intensification by the NSA algorithm. Furthermore, different neighborhood mechanisms are designed to be embedded within the main NSA algorithm so as to enhance its performance. The applicability of the proposed model and the performance of the NSA approach are demonstrated via computational experiments.