Research Of Fault Diagnosis In Traction Converter

In this dissertation significances of fault diagnosis are clarified in the beginning. Then the state of studying up till now is summarized. The weakness for Electric Locomotives Traction Converter by using principles of fault diagnosis is analyzed. New intelligent fault diagnosis methods employing wavelet transform and neural network for Traction Converter are presented.The structure and principle of the CRH5 Converter is introduced in this dissertation. Based on analysis of the Converter's working mechanism, a new method of making the output circuit of the Converter as fault character parameters is presented. We have got different waveforms of output circuit while there are fault components in the Converter. By making simulation of it by MATLAB/SIMULINK. The feature extraction of the waveform is used through wavelet transform. Character vectors are constructed by obtaining the power spectrum of decomposed coefficients.In this paper, a new efficient intelligent fault detection method based on the wavelet transform and the artificial neural networks was presented, the theoretical background of wavelet transform was given, and a much better BP algorithm based on the traditional BP algorithm was introduced. Accordingly, a new wavelet neural networks fault diagnosis systems was developed. The proposed intelligent fault diagnosis system was simulated. The results showed that the developed system was effective. By using this method, performances of fault diagnosis are improved. And avoiding experts spend more time on analyzing frequency data. This intelligent fault diagnosis is endowed with promising future.