Application of bayes network on estimation of machine fault probability
JI Zhong-hua · He'nan kexue · 2004
On the base of bayes network, a method of fault diagnosis for machine is presented in the paper. The purpose of using the network is to deduce the probability of a given fault from the condition data of a running machine as well to offer the evidences for further diagnosis. Bayes network is a directed acyclic graph with a series of conditional probabilities. A two-layered network model is constructed in the paper, and the upper layer represents a fault set and the lower layer represents a symptom set. First, the prior probabilities of faults and the conditional probabilities and leak probabilities are given. Second, the posterior probabilities are calculated by the process of reasoning and computation on the bayes network. By comparing all of posterior probabilities a maximum posterior probability is obtained and the trend of the machine faults is completed. The validity and accurateness of the method is checked by an example.