A hybrid learning-based meta-heuristic algorithm for scheduling of an additive manufacturing system consisting of parallel SLM machines

Additive manufacturing (AM) has been recognised as a promising technology under the context of Industry 4.0, which is reshaping manufacturing paradigms. A prominent type of AM machine is the selective laser melting (SLM) machine, in which several parts may form a job and be produced concurrently. This paper aims to investigate a scheduling problem in an AM system with non-identical parallel SLM machines. Since, in this system, there might be differences in the material types of parts, the required setup time between two consecutive jobs on the relevant machine is dependent on their material types. Accordingly, a bi-objective mathematical model is extended for the problem, considering the makespan and the total tardiness penalty as two objective functions. Due to the high complexity of the problem, an efficient hybrid meta-heuristic algorithm is developed by combining the non-dominated sorting genetic algorithm (NSGA-II) with a novel learning-based local search founded on the k-means clustering algorithm and a regression neural network. The local search enhances the exploitation ability of the NSGA-II while intelligently being taught during the solving procedure. Finally, the superiority of the proposed hybrid algorithm is demonstrated through a computational experiment.