A SIMULATION-OPTIMIZATION APPROACH FOR THE ROBUST JOB SHOP SCHEDULING PROBLEM WITH CONDITION-BASED MAINTENANCE AND RANDOM BREAKDOWNS

Abstract

The integration of scheduling and maintenance activities is important in shop floor decision-making. Job shop scheduling with maintenance activities requires proper planning due to its complex nature. A simulation-optimization approach is proposed in this thesis to solve the Job Shop Scheduling Problem (JSSP) under both planned and unplanned machine unavailability, considered based on Condition-Based Maintenance (CBM) and random breakdowns, respectively. Furthermore, the proposed approach accounts for uncertainty in the machine degradation rate and the duration of CBM and Corrective Maintenance (CM) activities, and aims to generate robust schedules that have good performance both on average and under adverse scenarios. The objective function of the original problem is first approximated using multiple surrogate functions that are independently optimized using Genetic Algorithm (GA). The first surrogate function considers the actual jobs only, the second one adds up deterministic CBM, the third one adds up deterministic breakdowns happening at the Mean-Time-To-Failure (MTTF), and the remaining surrogate functions consider jobs with deterministic CBM and different breakdown scenarios. These random breakdown scenarios are determined by solving a diversity maximization problem such that the breakdown time differences between the surrogate functions are maximized. The approximated and optimized schedule is then evaluated through simulation with the original objective function considering stochastic degradation of machines, random breakdowns, and uncertain CBM and CM duration. A weighted average of the expected makespan and its 90th percentile is used as the objective function to ensure schedule robustness. To prevent a premature conversion, a stopping rule motivated by Simulated Annealing (SA) is employed in the outer loop of the proposed approach. The proposed approach performed well with a maximum average improvement of 8.68% for 15-job and 15-machine and a minimum average improvement of 5.46% for 50-job and 15-machine over the initial solution with an average run time of 1.88 hours and 6.01 hours, respectively. Numerical experimentation demonstrated that the proposed approach can effectively generate high-quality schedules while considering CBM, random breakdowns, and parameter uncertainty.