The Study On Fault Diagnosis Method Based On Wavelet And RBF Neural Network

In recent years,with the development of computer technology and pattern recognition technology,fault diagnosis has developed gradually,especially the fault diagnosis based on the knowledge.With the development of neural network and wavelet analysis,they are applied to various fields gradually.This paper combines wavelet analysis and neural network with fault diagnosis based on the good pattern recognition performance of neural network and the data processing performance of wavelet decomposition.In this paper,aiming at the large volume of data and redundant data,wavelet decomposition is used to deal with the signal processing,including the noise reduction of data and the extraction of feature vectors.After wavelet decomposition,decreasing the dimension of the fault signal,filtering out the redundant signal components,and highlighting the fault components,which does not destroy the information contained in the signal therefore,the accuracy and speed of fault diagnosis are greatly improved in practical application.In this paper,a radial basis function(RBF)neural network based on particle swarm optimization(PSO)is proposed.Because of the good pattern recognition performance of RBF neural network,and the rapid diagnostic ability compared with other neural networks,RBF neural network is adopted as the diagnostic technique in this paper.In addition,aiming at the center determining of the basis function in implicit layer,the PSO algorithm is adopted to optimize the determination of base function center,so that the optimal center of the basis function in implicit layer is obtained and the accuracy of RBF neural network is greatly improved.The diagnostic performance of the fault diagnosis put forward in this paper is proved by diesel engine.Design the overall structure of the on-line monitoring and fault diagnosis of diesel engine,including the acquisition module,monitoring module,transmission module and diagnosis module,forming a local area network(LAN).Acquisition module has been designed based on wavelet transform,and at the same time,diagnosis module has been designed based on RBF neural network optimize by PSO algorithm.It is proved thatthe fault diagnosis method designed in this paper shows good performance in speed and precision of diagnosis.