Multi-objective design of the material handling system in zone-based job shop manufacturing environments using supervised machine learning
研究了区域式作业车间中物料搬运系统的后续设计步骤，提出集成实验设计、仿真、监督深度学习和多目标优化的级联方法，案例显示可大幅减少搬运时间和在制品。
Zone-based and flexible bay layout concepts are essential tools for facility designers, particularly in job shop manufacturing environments characterised by medium-to-high product variety and flexible routing. While traditional approaches focus primarily on optimising zone dimensions and allocating departments within zones, this paper advances the design process by investigating the subsequent steps required to finalise layout configurations in zone-based job shop systems. Specifically, these steps include identifying material-handling input/output points along cell boundaries, determining appropriate aisle dimensions, and selecting suitable material-handling transporters. To address these interrelated decisions, the study proposes a cascading methodology that integrates design of experiments (DOE), discrete-event simulation, supervised deep learning, and multi-objective optimisation. The optimisation process employs the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) in conjunction with an adaptive direct-search strategy implemented through NOMAD, a black-box optimisation solver. A numerical case study of a zone-based job shop system is presented. Computational results show that the optimised layouts achieve substantial reductions in material-handling time and work-in-process (WIP); for instance, one solution reduces average transport time by approximately 25% and WIP by approximately 35%, with only a modest increase in handling cost. These findings highlight the practical value of the proposed multi-objective approach.