Abstract: To make easy and efficient production planning for mixedmode process and improve performance of the planning, a parallel backward inferring algorithm based on flow networks was proposed. According to the logical relationships between processing orders and units, as well as working procedures, a dynamic mixedmode production flow networks was established. Based on the networks, a universal task allocation algorithm with parallel backward inferring was introduced to conduct production planning. This algorithm was used to determine start time of predecessor process for all tasks by parallel backward inferring in reverse direction of processing routes from the end working procedure. And then, each task was allocated to the predecessor processing unit. Case of the production planning of steel making and continuous casting showed that this algorithm could establish a feasible production plan with no operation conflict quickly and efficiently.

Key words: mixed-mode production processes, production planning, flow networks, parallel backward inferring algorithm

摘要： 针对混合流程作业计划编制困难、效率低、生成的计划可执行性差等问题，提出了基于生产流程网络图的时间并行倒推算法。该算法按照生产工序和工位间的逻辑关系，以及产品的生产工艺路径，构建混合流程生产作业动态网络图，用逆工艺路径的时间并行倒推算法来编制作业计划，从流程最后一道工序开始，依据网络图中生产工序和工位关系，所有任务逆工艺路径方向进行时间并行倒推，确定其在当前作业工序的紧前工序的开始作业时间，再将任务分配给紧前工序的适宜工位加工。通过炼钢-连铸作业计划编制案例表明，该算法能快速有效地编制出任务间没有作业时间冲突且可执行的生产作业计划。

关键词: 混合流程, 生产计划, 流程网络图, 时间并行倒推算法