Research On Wavelet Transform And Neural Network Based Rotating Machinery Fault Diagnosis

Rotating machine is the key equipment in areas such as aviation, chemical and power, so thefault diagnosis research has a great practical significance. With the continuous evolution of vibrationdetection, signal processing and other related technology, it has became an important researchdirection of fault diagnosis based on the vibration signal detection, processing and analysis.Meanwhile, the research of intelligent fault identification and diagnosis technology based on neuralnetwork, has opens up a new way to research and apply of fault diagnosis.This paper describes the wavelet transform, neural network and other related contant. On theone hand, it describes the continuous wavelet transform, discrete wavelet transform, orthogonalwavelet packet transform and analysis the problem of edge effects; on the other hand, it describes thebasic theory of the BP neural network and a BP algorithm based on wavelet neural network, analysisthe learning algorithm and problems of the standard BP neural network, by the simulation examples,the performance of BP neural network and BP algorithm based on wavelet neural network iscompared.Meanwhile, in order to extract the characteristics of rotating machinery fault, the theory and thespecific implementation process of extract fault characteristic value based on continuous waveletmodulus maximal method and optimal orthogonal wavelet packet method have researched.Finally, through multi-rotor test stand simulates common rotating machiney fault, usingcontinuous wavelet modulus maximal method and orthogonal wavelet packet method to extract rotorsystem common characteristics of fault, then put it into BP neural network for fault diagnosis, theresults indicates that the above method can diagnosis fault of rotor system effectively.