Optimizing Safety Stock Placement in General Network Supply Chains
Optimizing Safety Stock Placement in General Network Supply Chains
复制标题
优化通用网络供应链中的安全库存配置
DOI：
--
复制
发表时间：
2004
期刊：
影响因子：
--
通讯作者：
Ekaterina Lesnaia
中科院分区：
文献类型：
--
作者：
Ekaterina Lesnaia
对于面临不确定需求并需要为客户提供高水平服务的制造企业来说，在供应链的每个阶段保持的安全库存量是一个重要问题。库存量应该小，以尽量减少持有和储存成本，同时保持及时为客户服务的能力，并满足大部分（如果不是全部）需求。本文利用确定性服务时间模型的框架对这一问题进行了分析，并提出了一般网络供应链中的安全库存配置算法。我们首先证明了一般问题是NP难的。接下来，我们开发了几个条件，表征一般网络问题的最优解。我们发现，我们可以确定所有可能的候选人的最佳服务时间的一个阶段，通过构建路径从阶段到其他阶段的供应链。我们使用这个结构，即这些路径，作为一般网络问题的分支和定界算法的基础。为了生成下界，我们创建并解决了一般网络问题的生成树松弛。我们提供了一个多项式算法来解决这些生成树问题。我们进行了一组计算实验，以评估性能的一般网络算法，并确定如何设置各种参数的算法。除了一般网络的情况下，我们考虑两层网络的问题。我们开发了一个专门的分支和定界算法，这些问题和计算表明，它是更有效的比一般的网络算法适用于两层网络。导师：Stephen C.管理科学与工程系统Abraham J. Siegel Professor of Management Science & Engineering Systems
The amount of safety stock to hold at each stage in a supply chain is an important problem for a manufacturing company that faces uncertain demand and needs to provide a high level of service to its customers. The amount of stock held should be small to minimize holding and storage costs while retaining the ability to serve customers on time and satisfy most, if not all, of the demand. This thesis analyzes this problem by utilizing the framework of deterministic service time models and provides an algorithm for safety stock placement in general-network supply chains. We first show that the general problem is NP-hard. Next, we develop several conditions that characterize an optimal solution of the general-network problem. We find that we can identify all possible candidates for the optimal service times for a stage by constructing paths from the stage to each other stage in the supply chain. We use this construct, namely these paths, as the basis for a branch and bound algorithm for the generalnetwork problem. To generate the lower bounds, we create and solve a spanning-tree relaxation of the general-network problem. We provide a polynomial algorithm to solve these spanning tree problems. We perform a set of computational experiments to assess the performance of the general-network algorithm and to determine how to set various parameters for the algorithm. In addition to the general network case, we consider two-layer network problems. We develop a specialized branch and bound algorithm for these problems and show computationally that it is more efficient than the general-network algorithm applied to the two-layer networks. Thesis Supervisor: Stephen C. Graves Title: Abraham J. Siegel Professor of Management Science & Engineering Systems