Solution method for a large-scale loom scheduling problem with machine eligibility and splitting property

In this paper, we focused on a real-life, large-scale problem of loom scheduling, which has 1100 independent jobs and 133 unrelated parallel machines with machine eligibility constraint. The authors focused on the scheduling problems that includes sequence-dependent setup times and the job splitting property to minimize the makespan. An adapted version of hybrid genetic algorithm in this study, incorporated a machine eligibility constraint. Adding this constraint complicated the problem, but doing so is compulsory to solve a real-life problem. To simplify the problem, we used the typesetting structure of the weaving industry. Utilizing random key numbers provided feasible chromosomes for each generation. Flexible chromosome structures and local search adaptation into the genetic algorithm were some of the other factors that allowed us to improve the makespan by up to 14% in this application.

本研究聚焦于一类真实场景下的大规模织机调度问题，该问题包含1100个独立作业与133台非关联并行机器，且带有机器资格约束（machine eligibility constraint）。本文所研究的调度问题涵盖序列相关准备时间与作业可拆分特性，目标为最小化最大完工时间（makespan）。本研究采用了改进版的混合遗传算法，该算法融入了机器资格约束。增设该约束虽会提升问题复杂度，但为求解真实场景下的实际问题，该约束是必不可少的。为简化问题建模，我们采用了纺织行业的排版结构。通过随机键编码技术，可为每一代种群生成可行的染色体（chromosomes）。灵活的染色体结构与适配遗传算法的局部搜索策略，亦是本研究可将该应用场景下的最大完工时间最高优化14%的关键因素之一。