Memetic algorithm based on non-dominated levels for flexible job shop scheduling problem with learn-forgetting effect and worker cooperation

Abstract

Traditional flexible job shop scheduling problems (FJSP) often focus on the flexibility of machines, neglecting the effectiveness and flexibility of workers. In real production environments, workers’ processing proficiency is influenced by the learn-forgetting effect, and they tend to cooperate when handling complex tasks to reduce difficulties. The impact and interests of workers are increasingly becoming indispensable factors in modern manufacturing systems. Therefore, this paper investigates a FJSP with learn-forgetting effect and worker cooperation (FJSP-LFWC) to simultaneously optimize makespan and maximum worker workload. A mathematical model is established for this problem, and a memetic algorithm based on non-dominated levels (MANL) is proposed to efficiently solve it. MANL addresses the problem in several key ways. Firstly, it generates a high-quality initial population through a meticulously designed hybrid initialization strategy. Secondly, it applies a novel decoding method to improve solution quality. Thirdly, it adjusts the selection strategy based on the convergence of the population. Additionally, a tailored local search strategy incorporating five local search operators is utilized for three types of candidate solutions to accelerate convergence and fully utilize the solution space. Extensive experiments are conducted based on 28 newly formulated instances. The experimental results demonstrate that MANL significantly outperforms five well-known comparison algorithms, showcasing its efficiency in solving FJSP-LFWC.