Research On Reliability Management Model Of Supply Chain Based On Multi-Agent System

The competence is not among businesses but among supply chains in 21th century. To gain cost advantage and market share, many companies implement various initiatives such as source goods globally and leanagile supply chain management. These initiatives are effective in a stable environment, but they could make a supply chain more vulnerable to various types of disruptions. The failure which occurs at any node or link of supply network affects rapidly all companies in the web, and then, influences the normal operations of the whole network.Based on application of operational research, intelligent algorithm and optimization theory, this dissertation researches supply chain reliability management, in light of systems engineering, reliability engineering, multi-agent system(MAS) and distribution chain integrative operation management. The concrete research is as follows:(1)Study on reliability design of supply chain. A reliable supply chain network design model based on ontology and multi-agent technology is proposed, which integrated a variety of calculating methods. Service publication and discovery of the member enterprises are realized by the service ontology, the order task is dynamically decomposed by the workflow ontology. The prediction of supply chain target reliability is carried out with BP neural network, the bidding enterprise's reliability is evaluated by using fuzzy logic method, the multi-objective genetic algorithm is applied to solve this multi-objective optimization problem to obtain the successful tender, the reliable supply chain is designed optimally.(2)Study on reliability early warning of supply chain. The function and operating process of supply chain reliability early warning system is analyzed. Based on the induction and analysis of various early warning methods, an infrastructure of supply chain reliability early warning system is built from two different views. A MAS-based reliability comprehensive early warning model in supply chain is put forward. The node enterprise agents employ the genetic projection pursuit method to predict the reliability of each enterprise. The subsystem agents get the weights of the enterprises' reliability by the eigenvalue weighted method, and then the reliability evaluation result of each subchain and each echelon are acquired by the weight. At the top layer, the system agent is responsible for predicting the whole reliability of supply chain.(3)Study on reliability improvement of supply chain. The production dependencies and the degree of production dependence in supply chain are defined, and the algorithm of operation reliability of supply chain based on the network analytical method is put forward. The echelon needed for redundant design is found through analysis of the component importance. If there are several tender companies, fuzzy reasoning is used to obtain the risk probability from the risk index and its trend. The optimal company is chosen in combination with risk comprehensive response capacity of these companies. In order to realize reliability improvement of supply chain, a hybrid agent is designed which merged the reactive agent with the deliberative agent(4)Study on reliability regulation of member enterprises. The model of member enterprises reliability regulation based on MAS is constructed. In order to inspirit the member enterprises to improve the operating efficiency and service reliability, an incentive regulation model and approach based on yardstick competition for the member enterprise is presented. By combing price regulation and service reliability control, the model uses price-cap regulation and considers the service reliability and product eligibility rate for the quality control. The DEA approach is applied to establish the regulation model.