Multi-objective Robust Optimization Over Time for Dynamic Disassembly Sequence Planning

Disassembly sequence planning aims to optimize disassembly sequences of end-of-life (EOL) products in order to minimize the cost and environmental pollutant emission. Various unpredictable factors in the disassembly environment (e.g., EOL product status and capabilities of operators) lead to significant uncertainty making the optimal disassembly sequence change over time. Considering existing multiple objectives and dynamic environment, this problem is indeed dynamic multi-objective optimization. As deploying a new solution (i.e., disassembly sequence) is costly in this problem, it is undesirable to change the deployed solution after each environmental change. In this paper, we first propose a model for disassembly sequence planning problem in which several factors including the environmental changes, deployed solution switching cost, constraints, and multiple objectives are taken into account. To solve this problem where frequently changing the deployed solution must be avoided, we propose a new multi-objective robust optimization over time (ROOT) framework to find robust solutions based on two new robustness definitions: average performance and stability. The proposed framework benefits from a novel accurate online predictor and the knee-oriented dominance which is applied to select the naturally preferred tradeoff solution to meet the application requirements of ROOT. Computational experiments demonstrate the effectiveness of the proposed ROOT framework.