关键词: 流程工业, 分布估计算法, 异构并行机, 全局搜索, 局部开发

Abstract: Aiming at a kind of complex heterogeneous parallel machine scheduling problem that was Heterogeneous Parallel Machine Scheduling Problem with Job Processing Constraints and Sequence-dependent Setup Times(HPMSP_JPCSST)existed in process industry,an Estimation of Distribution Algorithm(EDA)was proposed to minimize the makespan criterion.By making reference to Double King Division Spring Multiplication(DKDSM)method,a new update mode of probability model was presented,and the sampling probability model was constructed by crossing and fusing the sub-probability model.Thus the evolutionary direction of population was guided to improve the global search depth of the algorithm.An Insert-based mutation operation was designed to maintain the population diversity level,and the global search wide of the algorithm was improved.To enhance the local exploitation ability,the local search based on insertion method and first move strategy was introduced.Computational experiments and comparisons demonstrated the effectiveness and the robustness of proposed algorithm.