Abstract
Distributed manufacturing is emerging as the mainstream production paradigm within contemporary industrial systems. The distributed flexible job shop scheduling problem (DFJSP) is an NP-hard combinatorial optimization problem. A hierarchical optimization algorithm with a dual-cache synced tuning mechanism (HOA-DSTM) is proposed to solve the DFJSP in this article. The HOA-DSTM consists of two distinct stages: the evolutionary stage and the optimization stage. In the evolutionary stage, an elite retention strategy is designed in the crossover process to preserve the knowledge of high-quality individuals during each iteration. A dual-reinforcement learning (dual-RL) mechanism based on a conversion factor is employed to adjust the crossover probability ( ${P}_{c}$ ) and mutation probability ( ${P}_{m}$ ) to increase the optimization efficiency. The optimization stage includes a local search with seven operators and a DSTM for the optimum elite in the population. The DSTM leverages the coupling characteristic of the DFJSP encoding scheme to adjust the operation sequence (OS) and factory assignment (FA) in the current optimal individual. The experimental results on benchmark datasets demonstrate that the HOA-DSTM outperforms state-of-the-art algorithms in solving the DFJSP.