Robust scheduling based on extreme learning machine for bi-objective flexible job-shop problems with machine breakdowns

In modern manufacturing systems, a flexible job-shop schedule problem (FJSP) with random machine breakdown has been widely studied. Two objectives, namely makespan and robustness, were simultaneously considered in this study. Maximizing the workload and float time of each operation and the machine breakdowns, one surrogate measure named RMc was developed via an extreme learning machine (ELM) to evaluate robustness. Specifically, this measure determines the impact of float time on the robustness by the probability of machine breakdown and the location of float time. Simultaneously, the impact was automatically adjusted by the ELM. Then, a method combining an improved version of nondominated sorting genetic algorithm II and RMc was proposed to address the bi-objective FJSP. Computational results on the benchmarks show that RMc accurately evaluates the robustness of the schedules with a small amount of computation cost. (C) 2020 Elsevier Ltd. All rights reserved.