Minimizing makespan in two-stage assembly additive manufacturing: A reinforcement learning iterated greedy algorithm

Additive manufacturing (AM) is becoming increasingly important for producing mass-customized, small-quantity products with relatively low geometric constraints. Although some AM machine scheduling problems have been proposed in recent years, no research has addressed the parallel AM machine scheduling problem with an integrated assembly stage. In this study, a two-stage assembly additive manufacturing scheduling problem is considered, in which multiple parts are produced in job batches using identical parallel AM machines in the first stage and then assembled into the desired products in the second stage. Further, a mixed-integer linear programming model and an innovative reinforcement learning metaheuristic, called the iterated epsilon-greedy algorithm, are proposed to minimize the makespan of this significant scheduling extension. The computational results based on 810 test instances show that the developed approaches are highly effective, efficient, and robust in solving the addressed problem. Notably, the research results can effectively reduce the gap between the theory and practice of AM production planning by integrating the production stage with the assembly stage.& COPY; 2023 Elsevier B.V. All rights reserved.

Automated summary

This study addresses a two-stage assembly additive manufacturing scheduling problem, where multiple parts are produced in job batches using identical parallel AM machines in the first stage and then assembled into the desired products in the second stage. A mixed-integer linear programming model and an innovative reinforcement learning metaheuristic called the iterated epsilon-greedy algorithm are proposed to minimize the makespan of this significant scheduling extension. The computational results based on 810 test instances demonstrate that the developed approaches are highly effective, efficient, and robust in solving the addressed problem. Notably, the research results effectively bridge the gap between theory and practice in AM production planning by integrating the production stage with the assembly stage.