The problem of pick sequencing in the rotary rack S/R system (PPS-RRS) is investigated with the objective of minimizing the execution time. The rotary rack S/R system consists of one S/R machine and multiple levels of...The problem of pick sequencing in the rotary rack S/R system (PPS-RRS) is investigated with the objective of minimizing the execution time. The rotary rack S/R system consists of one S/R machine and multiple levels of carousels that can rotate independently in bi-directions. The routing policy, namely the decision on the storage or retrieval sequence, dominates the efficiency and the throughput for such S/R systems, due to the complicated relationship between all levels of carousels and the S/R machine. For the purpose of optimizing the PPS-RRS, a computational model is developed in terms of execution time for picking multiple items in one trip. Characteristics of the PPS-RRS are analyzed and a local search heuristic based on a newly proposed neighborhood is presented. Integrated with the proposed local search procedure a new hybrid genetic algorithm is developed. Experimental results demonstrate the structure characteristics of good sequence and the efficiency and effectiveness of the proposed sequencing algorithms.展开更多
智能制造系统采用了物联网等大量先进信息技术,使得车间积累了大量的实时生产数据。同时,复杂制造系统在运行过程中容易出现一系列干扰事件,这对车间实时响应能力提出了更高的要求。因此,在工业大数据支撑的制造环境下,针对考虑序列相...智能制造系统采用了物联网等大量先进信息技术,使得车间积累了大量的实时生产数据。同时,复杂制造系统在运行过程中容易出现一系列干扰事件,这对车间实时响应能力提出了更高的要求。因此,在工业大数据支撑的制造环境下,针对考虑序列相关设置时间和阻塞的混合流水车间调度问题(Hybrid flow shop scheduling problem with sequence-dependent setup times and blocking,HFSP-SDST-B),提出一种基于深度强化学习的实时调度方法,从而实现制造资源的合理分配和完工时间最小化。作为一个序列决策问题,HFSP-SDST-B可以被建模为一个马尔科夫决策过程。在每个调度点,智能体根据当前的生产状态选择相应的调度规则,从而进行合理的工件排序和机器分配。为了实现生产数据驱动的实时调度方法,依次设计考虑阻塞因素的调度点、通用生产状态特征、基于遗传规划的启发式规则和奖励函数。然后提出一种基于近端策略优化算法的训练方法,从而让智能体构建状态与规则之间的有效映射。最后试验结果表明,与现有的动态调度方法相比,该方法具有优越性和通用性,并且通过学习能够有效处理随机扰动时间和新订单插入的未知情况。展开更多
基金This work was supported by the National Natural Science Foundation of China (60104009)the Natural Science Foundation of Shandong Province, China (Z2000G01).
文摘The problem of pick sequencing in the rotary rack S/R system (PPS-RRS) is investigated with the objective of minimizing the execution time. The rotary rack S/R system consists of one S/R machine and multiple levels of carousels that can rotate independently in bi-directions. The routing policy, namely the decision on the storage or retrieval sequence, dominates the efficiency and the throughput for such S/R systems, due to the complicated relationship between all levels of carousels and the S/R machine. For the purpose of optimizing the PPS-RRS, a computational model is developed in terms of execution time for picking multiple items in one trip. Characteristics of the PPS-RRS are analyzed and a local search heuristic based on a newly proposed neighborhood is presented. Integrated with the proposed local search procedure a new hybrid genetic algorithm is developed. Experimental results demonstrate the structure characteristics of good sequence and the efficiency and effectiveness of the proposed sequencing algorithms.
文摘智能制造系统采用了物联网等大量先进信息技术,使得车间积累了大量的实时生产数据。同时,复杂制造系统在运行过程中容易出现一系列干扰事件,这对车间实时响应能力提出了更高的要求。因此,在工业大数据支撑的制造环境下,针对考虑序列相关设置时间和阻塞的混合流水车间调度问题(Hybrid flow shop scheduling problem with sequence-dependent setup times and blocking,HFSP-SDST-B),提出一种基于深度强化学习的实时调度方法,从而实现制造资源的合理分配和完工时间最小化。作为一个序列决策问题,HFSP-SDST-B可以被建模为一个马尔科夫决策过程。在每个调度点,智能体根据当前的生产状态选择相应的调度规则,从而进行合理的工件排序和机器分配。为了实现生产数据驱动的实时调度方法,依次设计考虑阻塞因素的调度点、通用生产状态特征、基于遗传规划的启发式规则和奖励函数。然后提出一种基于近端策略优化算法的训练方法,从而让智能体构建状态与规则之间的有效映射。最后试验结果表明,与现有的动态调度方法相比,该方法具有优越性和通用性,并且通过学习能够有效处理随机扰动时间和新订单插入的未知情况。
