Research On Coordination Optimization Of Decentralized Supply Chain System

Decentralized supply chain management has become the key character of the supply chain management. This dissertation first illustrates that collaborative planning is the most suitable coordination mechanism of decentralized supply chain. It then studies the optimization of collaborative planning of decentralized supply chain, with single-attribute objective and multi-attribute objective respectively. The main contributions of the dissertation consist of the following four parts.(1) After researching the features of decentralized supply chain management decision making, this dissertation suggests that collaborative planning is the most suitable coordination mechanism of decentralized supply chain. Especially when it comes with multi-attribute objective, collaborative planning needs to be adapted on time and the model needs to be rerun to get an optimum solution. Different techniques working under such circumstance in supply chain management are shown in the dissertation.(2) For the single-attribute objective decentralized supply chain system, a synchronous coordination model is constructed. To solve this model, an optimal method, combining Lagrangian relaxation algorithm with genetic algorithm, is presented. This dissertation modifies a fuzzy subgradient algorithm, which is superior to subgradient algorithm, and uses the modified algorithm, whose convergence is proved, to update Lagrangian multiplier to increase the efficiency of entire coordination strategy. Based on the internal punishment, a coordination strategy is proposed and the coordination optimization is improved.(3) For collaborative planning of decentralized supply chain system without a coordinator and with its members all having single-attribute objective, an asynchronous coordination model is presented. The surrogate subgradient algorithm is used to update the Lagrangian multipliers. Combining the genetic algorithm with the interactive heuristics algorithm, the dissertation develops an asynchronous coordination strategy based on the Lagrangian relaxation algorithm. (4) For multi-attribute objective collaborative planning of the decentralized supply chain, coordination optimization problems under two circumstances: one has a mediator and the other allows local information private, are researched. Coordination models are constructed for these two circumstances and decomposition methods are used to solve the models with integer variables to reach a global Pareto optimal solution. When there exists a mediator, the solution approach is based on Primal Decomposition methods. In this approach, members of the supply chain have to disclose their local information. On the other hand, if they want to keep local information private, Dual Decomposition methods should be adopted. Therefore, the coordination optimization strategy also changes. Instead of pushing restricting constraints upwards from local sub-problems to a global master problem, Dual Decomposition methods push linking constraints down to local problems. In order to make the coordination optimization more efficient, Primal-Dual Decomposition is suggested to improve the optimizing approach.