Flexible job-shop scheduling problem based on Markov state decision process

Abstract For flexible job shop scheduling (FJSP), this paper aims to minimize the maximum completion time and maintenance cost. Therefore, a flexible job shop production model based on a degraded state is proposed to realize the joint decision of scheduling and machine preventive maintenance. The machine state is modeled as a discrete multi-state degradation process through a continuous-time Markov chain, and a preventive maintenance strategy is developed based on the predicted state. To improve the solving efficiency, the MTSNSGA-II algorithm is designed, which integrates NSGA-II and hybrid tabu search and significantly improves the convergence speed of the algorithm. Finally, we use ten benchmark examples of the Brandimarte series for simulation experiments and analyze them using line and box charts. The experimental results show that the MTSNSGA-II algorithm performs well in both the IGD and HV indexes, verifying its effectiveness and superiority in the FJSP problem.