基于扩展状态任务网的制造供应链计划

Manufacturing Supply Chain Planning Based on Extended State Task Network

制造供应链计划是制造供应链管理的关键问题,它不仅需要分配生产任务和控制库存,还需要解决不同工厂(企业)间的运输配套问题.为统一描述具有复杂产品生产过程(包括装配型、分解型和多输入多输出型等)的生产任务、存储任务和不同模式(包括单种物料独立运输模式和多种物料组合运输模式)的运输任务,提出了扩展状态任务网(extended state task network,简称ESTN).扩展状态任务网用比例转化任务统一描述生产任务、存储任务和单种物料独立运输任务,用虚比例转化任务和组合移动任务共同描述多种物料组合运输任务.应用扩展状态任务网,meta启发式方法在求解制造供应链问题时更容易编码和操作.为求解基于ESTN的制造供应链计划模型,提出了具有多样性检测的参考解集更新策略与分散性解变异策略的路径重连算法.路径重连算法维护一个由高质量解(精英解)组成的参考解集,将一个向导精英解的属性逐步引入一个起始精英解而形成的中间解序列(路径),并用此中间解序列更新参考解集以获得进化.计算实例表明,该路径重连算法比标准遗传算法、标准Tabu搜索算法以及普通路径重连算法能够获得更好的解,证明了多样性检测对参考解集更新的关键作用以及分散性解变异策略在提高解的质量上的能力.

Manufacturing supply chain planning is a key factor of manufacturing supply chain management. Assignment of the production tasks, inventory control and transportation between the plants or the enterprises needs to be considered in manufacturing supply chain planning. Extended state task network （ESTN） is proposed in order to abstractly describe the production tasks with complex product production processes （assembly process, distilling process, process with many kinds of input material and many kinds of output material）, the store tasks, and the transportation tasks with different modes （one mode is that transfering only one kind of material and another several kinds of material）. In the extended state task network, proportion transform task is used to abstractly describe the production task, the store task and the transportation task that transfers only one kind of material. The combination of virtual proportion transform task and the combination move task are applied to describe the transportation task that transfers several kinds of material. It is easier to encode and operate the solution of the manufacturing supply chain planning if metaheuristic methods are used to solve the problem based on ESTN than on other models. A path relinking algorithm is developed to solve manufacturing supply chain planning model based on ESTN. The algorithm maintains a reference set of solution with good quality in evolutionary process. A list of the new solution （path） is created by inserting properties of a guide solution into an initiate solution. The list is used to update reference set in order to evolute the solution. The solution update method of the reference set with diversification detection and decentralization mutation strategy is proposed in the path relinking algorithm. The results of computations demonstrate that the path relinking algorithm can obtain better solutions than the standard genetic algorithm, the standard Tabu search procedure and the general path relinking algorithm, and prove that the solution update method of the reference set with diversification detection is very important to path relinking. Moreover, the decentralization mutation strategy can enhance the capacity of searching a better solution.