Tri-objective based scalable balancing of reconfigurable manufacturing systems subject to resource constraints and task processing time uncertainty

This article studies reconfigurable manufacturing systems and their ability to efficiently evolve (or scale) through the use of a limited number of additional resources to improve two analysed system characteristics: productivity and robustness against task processing time uncertainty. The goal is to find an initial system configuration (or balancing) representing the best potential for scalability. To this end, we propose a multi-step based multi-objective approach, where productivity and robustness are considered as two objectives to be maximised and the number of used additional resources as the third objective to be minimised. Thus, for a given balancing, we first explore the set of its non-dominated (in the sense of Pareto) configurations, reflecting relevant possibilities for additional use of resources. Then, we calculate its corresponding tri-objective hyper-volume and suggest to consider it as a scalability metric. Finally, to find the most scalable balancing, we develop an efficient simulated annealing algorithm, where the proposed metric is used as a fitness function. The algorithm is tested and analysed on small- and medium-sized instances, extracted and adapted from the well-known benchmark dataset. The numerical results obtained demonstrate the stability and efficiency of the developed algorithm and confirm the relevance of the proposed scalability metric.