An improved MOEA/D with reinforcement learning for flexible job shops incorporating manual operations and fatigue effects

Evidence shows that the manufacturing industry is transitioning towards “automation” at the present stage, workpieces will be processed by machines automatically, and workers only need to complete the “pre-processing” and “post-processing” tasks. In the meantime, worker fatigue also needs to be take...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Swarm and evolutionary computation Jg. 99; S. 102200
Hauptverfasser: Li, Yibing, Tong, Xin, Zou, Haiqi, Guo, Jun, Wang, Lei, Tang, Hongtao, Wang, Kaipu, Li, Xinyu
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier B.V 01.12.2025
Schlagworte:
Flexible job-shop scheduling
Manual task
Multi-objective evolutionary algorithm based on decomposition (MOEA/D)
Reinforcement learning
Worker fatigue
Manual task
Flexible job-shop scheduling
Multi-objective evolutionary algorithm based on decomposition (MOEA/D)
Worker fatigue
Reinforcement learning
ISSN:2210-6502
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Evidence shows that the manufacturing industry is transitioning towards “automation” at the present stage, workpieces will be processed by machines automatically, and workers only need to complete the “pre-processing” and “post-processing” tasks. In the meantime, worker fatigue also needs to be taken into consideration due to its influence on workers' health and efficiency. However, the majority of current research on job-shop scheduling overlooks manual operations and the impact of worker fatigue. To bridge this gap, we introduce manual operations and fatigue effects to flexible job-shop problems and formulate a mathematical model. Since the exact solver is inefficient in solving the problem proposed, a problem-specific improved multi-objective evolutionary algorithm based on decomposition with reinforcement learning is developed. Specifically, we design a local search strategy based on the online measurement of subproblem hardness, comprising three tailored local search methods. Considering the complex solution space of the proposed problem, a hybrid crossover method is designed to enhance the algorithm's performance. A dynamic parameter selection strategy based on reinforcement learning is proposed to intelligently select the appropriate parameters according to the iteration situation. The effectiveness of the improvement strategies is verified through numerical experiments, and the proposed algorithm is compared with five other multi-objective optimization algorithms on 18 random test instances. Then the algorithm is applied to a real engineering case. The results indicate reliable performance of the proposed algorithm. [Display omitted]
ISSN:2210-6502
DOI:10.1016/j.swevo.2025.102200