Workforce-constrained Preventive Maintenance Scheduling Using Evolution Strategies

Heavy equipment overhaul facilities such as aircraft service centers and rail-yards face the challenge of minimizing the makespan for a set of preventive maintenance (PM) tasks, requiring singleor multiple-skills, within workforce availability constraints. In this paper, we examine the utility of evolution strategies (ES) to this problem. Comparison of the computational efforts of ES with exhaustive enumeration (EE) to reach optimal solutions for sixty small problems illustrated the ability of ES to yield optimal solutions increasingly efficiently with increasing problem size. A set of 852 large-scale problems was solved using ES to examine the effects of task-related problem characteristics (number of tasks N, skills required per task SPT, and workers required per task WPT), workforce-related variables (number of workers W, skills per worker SPW), and ES population size (μ) on CPU time. Due to unavailability of optimal solution for comparison, each experiment was stopped at a quasi-optimal solution using a convergence criterion. In the estimated multiplicative power expression, the CPU time requirements using ES were proportional to N (versus e for the EE approach). The maximum CPU time was 55 minutes for a 1000-task problem. The results empirically supported practical utility of ES to solve large-scale, complex PM problems involving single and multiple-skilled workforce. Finally, comparison of convergence speeds of ES and simulated annealing (SA) for the 852 experiments indicated that, on the average, SA needed to search about 12 times the number of solutions searched by ES to converge to the quasi-optimal solution, demonstrating the superiority of the ES approach to the SA approach for this class of PM problems. Subject Area: Preventive Maintenance, Evolution Strategies, Workforce Scheduling