The Research Of Motor Fault Detection Based On Bayesian Network

Bayesian network is developed to deal with the problems of uncertainty in artificial intelligence research area. It is a tool to apply the theory of probability and statistics to the uncertainty reasoning and data analysis in complex field. In recent years, the theory of Bayesian network is developed rapidly, and its technology has been successfully applied to the fields of finance, speech recognition, robotics, and so on. In the application of Bayesian networks, Bayesian network structure construction is a core issue. It still lack of a mature Bayesian network structure learning algorithm, so it’s of great significance to solve Bayesian network structure optimization problem by novel and efficient algorithm.We have been a lot of research on BN structure learning in this paper. Applies Cuckoo Search algorithm to Bayesian network structure learning. At the same time, the advanced search strategy is used to improve structure learning algorithm. This algorithm is also applied to the fault diagnosis of induction motors.First, the paper expounds the research background and development status of Bayesian networks. Then introduces the basic theory of Bayesian network.Second, on the issue of the construction of the Bayesian network, the newly proposed cuckoo search algorithm is applied to Bayesian network structure learning, Cuckoo Search algorithm is a heuristic search algorithm which simulates species of parasitic cuckoo brood to find the optimal solution. A novel Bayesian network structure stochastic evolution method is given based on mutual information theory and binary Levy flights. And, a competition mechanism is introduced in cuckoo search algorithm to improve search ability.Then, with detailed introduction of the Hilbert Huang transform, this paper introduces the method of extracting the marginal spectrum characteristics of the stator current signal of the induction motor. The method of quantification of marginal spectrum is described in detail, and the relevant feature vectors are obtained.Finally, the paper presents fault diagnosis model which combines the induction motor stator current Hilbert marginal spectral characteristics and Bayesian network classifier constructed by cuckoo search algorithm. Experimental results show that the fault diagnosis model is very effective, high accuracy rate of diagnosis.