Coordination of supply chain inventory systems with private information

摘要

This dissertation considers the problems of coordinating different supply chaininventory systems with private information under deterministic settings. These systemsstudied are characterized by the following properties: (a) each facility in the system hasself decision-making authority, (b) cost parameters of each facility are regarded asprivate information that no other facilities in the system have access to, and (c) partialinformation is shared among the facilities. Because of the above properties, the existingapproaches for systems with global information may not be applicable. Thus, newapproaches for coordinating supply chain inventory systems with private information areneeded.This dissertation first studies two two-echelon distribution inventory systems.Heuristics for finding the replenishment policy of each facility are developed underglobal information environment. In turn, the heuristics are modified to solve theproblems with private information. An important characteristic of the heuristicsdeveloped for the private information environment is that they provide the samesolutions as their global information counterpart.Then, more complex multi-echelon serial and assembly supply chain inventorysystems with private information are studied. The solution approach decomposes theproblem into separate subproblems such that the private information is divided asrequired. Global optimality is sought with an iterative procedure in which thesubproblems negotiate the material flows between facilities. At the core of the solution procedure is a node-model that represents a facility and its corresponding privateinformation. Using the node-model as a building block, other supply chains can beformed by linking the node-models according to the product and information flows. Bycomputational experiments, the effect of the private information on the performance ofthe supply chain is tested by comparing the proposed approach against existingheuristics that utilize global information. Experimental results show that the proposedapproach provides comparable results as those of the existing heuristics with globalinformation.