Research On Reliability Probability Risk Assessment Method Based On Bayesian Network

With the rapid development of science and technology, the structures of all kinds of systems are becoming more and more complicated, the capacity parameters of system are gradually improved, the environmental conditions also become more and more tough, so the safety and reliability problems in the operation process are becoming more and more attention by people. Based on the diesel engine as the research object in the paper, the method of improving its reliability was found through the risk assessment research of reliability probability, so as to control the failure risk probability of system and reduce the related maintenance costs of system. From the angle of the required specification size of the total joint table with storing network nodes and computational efficiency, various reliability modeling methods are compared and analyzed based on the subsystem reliability in the diesel engine, and the risk assessment of reliability probability for the diesel engine system was proceded using the optimal Bayesian network modeling method. The main research contents in the paper were as follows.(1) On the basis of understanding and reviewing the Bayesian network theory and traditional reliability methods, five reliability methods based on the Bayesian network(Naive^Explicit Connectivity> Explicit Disconnectivity^MLS and MCS)were analyzed and were established the corresponding models based on the supply system reliability in the diesel engine, and the Bayesian network modeling methods with the minimum required specification of the total joint table and the highest computational efficiency was found out through the comparison of the required specification size of the total joint table with storing network nodes and computational efficiency for various methods, and the Bayesian network was optimized by two heuristics to get the optimal Bayesian network model finally.(2) On the basis of the established reliability analysis model based on the Bayesian network, the system reliability was analyzed. For the starting difficult fault model of diesel engine, the optimization Bayesian network model was established. Then the system reliability was analyzed through the model, including the consequences probability, fault diagnosis, causal reasoning and important degree analysis (RRW、FV、RAW and BM). The failure probability of system, weak links of system and weight of components in the system were obtained, so as to provide the necessary references for improving system reliability, carrying out diagnosis and forecasting through the reliability analysis.(3) According to the characteristics of multi-state system, the reliability modeling and analysis methods of multi-state system based on Bayesian network were provided. First the state functions of several typical multi-state systems were established, second the Bayesian networks of the multi-state systems reliability analysis were established according to the characteristics of state functions, third the multi-state system reliability were analyzed through the model. Last the effectiveness and advancement of the proposed method were verified through the fuel fault example in the diesel engine.