Solving a Single-Server Parallel Machine Problem with Preventive Maintenances Using a Hyper-Heuristic Algorithm

This paper considers a parallel machine problem subject to a single server, machine unavailability due to preventive maintenances, setup times and release dates. The goal is to minimize the makespan and the total idle times, simultaneously. Since this special case of parallel machine is known to be NP-hard, a hyper-heuristic method is proposed to solve the model. The hyper-heuristic has two levels; on the lower level, a set of well-known algorithms in scheduling is defined. The selection policy is presented in the upper level where, considering the convergence and diversity of the current solutions, the most appropriate algorithm is chosen from the toolbox. The advantage of a hyper-heuristic lies in its flexibility in switching between different methods to improve the exploration and exploitation of the overall procedure. The proposed hyper-heuristic is validated and tested on different problem scales and the results emphasize its high precision and rapid convergence.