克服库存不精确的鲁棒集成补货、生产控制及分销策略

Inventory inaccuracy does exist in main stages of supply chain management, e.g., raw materials replenishment, production control and distribution scheduling, which leads to significant negative impact on its performance. Therefore, in this project, we will investigate the robust integrated replenishment, production control and distribution scheduling problem to hedge against inventory inaccuracy. The objective of this problem is to maximize the probability that the operation cost of a supply chain is less than an expected value in the situation of inventory inaccuracy. This probability is defined as robustness measurement. To solve this problem, we propose the robust integrated replenishment, production control and distribution policy by integrating state prediction method, robust optimization theory, robust control theory, and backstepping method to overcome the difficulties of ordering time-delay, and the coupling between replenishment, production and distribution. We also prove that the robustness of the proposed policy is better than that of traditional replenishment, production control and distribution policies, analyze the sensitivity of the robustness with respect to inventory errors, and find the quantitative condition of adopting the proposed robust policy. Base on that, we combine the proposed robust policy and the radio frequency identification (RFID) technology, which has been applied extensively in supply chain management, and solve the integrated optimization problem of robust policy adoption and RFID devices deployment so that the cost of RFID system is minimized while robustness of a supply chain system with inventory inaccuracy can be ensured. Finally, an experimental system of robust integrated replenishment, production control and distribution scheduling will be developed.

库存不精确性在供应链管理的主要环节(如原材料补货、生产控制与分销调度)中客观存在，并给其性能带来很大负面影响。因此，本项目研究克服库存不精确性的鲁棒集成补货计划、生产控制及分销调度问题，其目标是最大化库存不精确条件下供应链运行成本低于预期值的概率，并定义此概率为鲁棒性度量。为解决此问题，把状态预测方法、鲁棒优化理论、鲁棒控制理论与逆推方法相结合，克服订货时延和补货、生产、分销三者之间的耦合等困难，提出鲁棒集成补货、生产控制和分销策略；从理论上证明这一策略的鲁棒性优于传统的补货、生产控制和分销策略；对这一策略的鲁棒性关于库存误差的灵敏度进行分析，并提出其定量化适用条件。在此基础上，把这一鲁棒策略与供应链管理中广泛应用的射频识别技术结合起来，解决鲁棒策略应用与射频识别设备部署综合优化问题，在保证鲁棒性指标的基础上，最小化射频识别系统实施成本；并开发鲁棒集成补货、生产控制与分销调度实验系统。

库存记录误差普遍存在，且会对生产/库存系统产生严重不利影响；本课题研究了克服库存不精确的鲁棒集成补货、生产控制及分销策略。. 在鲁棒补货策略方面，针对由不可见库存损失导致的库存不精确，提出一种递归算法来估算真实库存水平的概率分布，并提出了一种鲁棒再订货时间/订货量策略。.在鲁棒生产控制策略方面，首先分别针对单阶段和多阶段单品种生产/库存系统，考虑补货提前期，提出一种库存精确时的近似最优补货及生产控制策略；在此基础上，考虑库存不精确性，提出一种在系统状态空间区域边界上光滑变化的鲁棒补货及生产控制策略，有效抑制了库存不精确性的负面影响。进一步针对多阶段、多品种、考虑物料运输时延的生产库存系统，应用动态规划获得库存记录精确时的近似最优补货及生产控制策略，然后针对库存不精确的情形，提出了一种基于库存水平条件数学期望的鲁棒生产/库存控制策略。. 在鲁棒分销策略方面，首先研究含补货提前期的库存路径规划问题：针对一个供应商两个零售商的简单情形，证明了最优库存路径策略具有切换曲线结构特征；推广到多零售商情形，提出了一种近似策略并将其与遗传算法结合来优化库存路径。进一步针对库存不精确、考虑补货提前期的多零售商分销调度问题，提出了当前及未来库存概率分布的估算方法；在此基础上，针对候选配送路径，提出了鲁棒补货时间、补货量及补货周期优化算法；最后，设计了含成本估计规则的遗传算法来优化配送路径。. 基于上述结果，提出了库存不精确时鲁棒集成补货、生产控制及分销的逆向递推策略，并开发了实验系统。进一步研究了鲁棒生产控制策略与射频识别系统的综合应用方法，建立了库存不精确时多阶段生产系统的马尔可夫决策过程模型，通过策略迭代获得了阈值型鲁棒生产控制策略，建立了射频识别设备部署优化的整数规划模型并提出了解此模型的启发式算法。. 本课题成果对库存、生产、分销控制理论有新的贡献，对克服库存数据误差带来的不利影响有重要应用价值。