Energy-aware remanufacturing process planning and scheduling problem using reinforcement learning-based particle swarm optimization algorithm

Abstract

Solving remanufacturing process planning and scheduling problem collaboratively and leveraging the complementary attributes of process planning and shop scheduling to attain improved production flow and process routes, are crucial for further enhancing the environmental and economic benefits of remanufacturing. Most of the existing works regard these two segments as independent and solve them separately, which hinder the further improvements of remanufacturing system performance. Besides, studies on energy-aware remanufacturing scheduling have employed machine turn on/off strategy to achieve energy reductions. However, not all machines are suitable for the turn on/off strategy. Therefore, a new energy-aware remanufacturing process planning and scheduling model with process sequence flexibility is proposed. This model not only simultaneously solves the remanufacturing process planning and scheduling problem, but also employs machine speed-switching strategy to reduce energy consumption. To solve this model, a reinforcement learning-based particle swarm optimization algorithm with an efficient multi-dimensional encoding scheme is proposed, in which, a hybrid population initialization strategy, a novel reinforcement learning-based multi-directional guide position-updating mechanism, a local search strategy, and a restart mechanism are devised to enhance the performance. Simulation experiments were conducted on 18 sets of instances with different scales to compare the proposed algorithm with other advanced algorithms. The experimental results confirmed the superiority of the proposed algorithm.