• CN: 11-2187/TH
  • ISSN: 0577-6686

Journal of Mechanical Engineering ›› 2023, Vol. 59 ›› Issue (7): 355-366.doi: 10.3901/JME.2023.07.355

Previous Articles     Next Articles

Reinforcement Learning-based Swarm Evolutionary Algorithm To Solve Two-sided Multi-objective Synchronous Parallel Disassembly Line Balancing Problem

GUO Hongfei1,2, LU Xinyu3, REN Yaping1,2, ZHANG Chaoyong4, LI Jianqing5   

  1. 1. School of Intelligent Systems Science and Engineering, Jinan University, Zhuhai 519070;
    2. Institute of Physical Internet, Jinan University, Zhuhai 519070;
    3. School of Management, Jinan University, Guangzhou 510632;
    4. State Key Lab of Digital Manufacturing Equipment and Technology, Huazhong University of Science and Technology, Wuhan 430074;
    5. School of Computer Science and Engineering, Macau University of Science and Technology, Macau 999078
  • Received:2022-05-23 Revised:2022-10-11 Online:2023-04-05 Published:2023-06-16

Abstract: Under the background of dual carbon, the industrial manufacturing field is transforming in the direction of green and energy-saving. The recycling, disassembling and remanufacturing of waste products is conducive to promoting high-quality development. Aiming at the problem that it is difficult to change the disassembly direction during the disassembly process of large and complex products, and comprehensively considering the various constraints that exist between tasks in the actual disassembly process, the paper studies the two-sided disassembly line balancing problem (TDLBP) in the synchronous parallel mode. Firstly, the disassembly line mode of two-sided layout is introduced, and the relationship of and priority and or priority is defined, and the mathematical model of TDLBP is established to optimize the layout of the production line, economic benefits, safety and environmental protection with a total of six indicators. Then, a swarm evolutionary algorithm based on reinforcement learning is proposed. Q-learning is used to select the best operator in the iteration by using the knowledge learned, the Pareto solution set is screened by the crowding distance, and the elite retention strategy is used to accelerate the convergence of the algorithm, so as to efficiently obtain the approximate optimal disassembly scheme. Finally, the effectiveness and superiority of the proposed algorithm are verified by solving small-scale cases and comparative analysis, and the application of large-scale case is carried out.

Key words: two-sided disassembly line, synchronous parallel, reinforcement learning, swarm evolutionary algorithm, Pareto

CLC Number: