The Application Of Bayesian Networks In Reliability Analysis

Bayesian Networks (BN), which was made up of Directed Acyclic Graph(DAG) and Conditional Probability Table(CPT) , is also called as Belief Networks. DAG is composed with nodes and its pathway. Node is the symbol of variable. Pathway between nodes represents correlation of the nodes (from father nodes point at son nodes). It related the nodes with conditioned probability in CPT. And the nodes without father nodes have prior probability. The inaugurator is Pearl that researched it upon probability theory and graph theory. Bayesian networks has excellent performances in expressing and reasoning with uncertainty knowledge. Such as in artificial intelligence, data mine, it is a powerful tool and has abroad applications.Reliability theory is a synthetical subject about the life character of production. And its important academic tool is probability theory. The practical analyses of the reliability always depend on the graphs. Thus, this paper introduced BN to the analyses of the reliability, and compared it with the fault trees and structural graphs of reliability analysis. For the superiority of the BN's reasoning, it can omit the traditional inconsistent process of the minimum path sets. At the same time, by adjust the probability table; it exceeded the traditional fault trees analysis by getting over the localization of the fault tree analysis, such as independence, binary state of the components. Using the CPT, BN can calculate interesting probability of any node or their sets and provide the diagnoses and maintenance ofthe reliability system. Furthermore, using the generalization of the BN--intervalBayesian networks ,made the reliability analysis of the components whose faulty rate just in a set feasible. It gets over the localization of single point value in traditional analysis, widen the field of reliability analysis.