Research On Fault Analysis And Diagnosis Method Of CRH2Traction Converter

With the development of technology of multilevel inverter, three-level inverter has been widely applied in many occasions, such as high voltage, large power fields. The increasing number of the power device improves the failure rate, which makes the fault problem more and more serious. Besides, the study of its fault diagnosis has become one of the hot issues of the power electronics. CRH2EMU adopted the three-level inverter, its performance is one of the most important indexes for evaluating the safety and efficient operation, and the relevant fault diagnosis is necessary important.The paper starts from the topological structure and the working principle, establishing the fault model of the traction inverter of CHR2and making the fault analysis and classification. On the basis of describing the structure, working principle of neural network, the wavelet analysis theory and the application of the fault diagnosis, the paper puts forward the theory of combining the wavelet analysis and neural network theory, and then applies it to the fault diagnosis of traction converter of CRH2.In the view of the excellent time-frequency analysis capability of the theory of wavelet, this paper uses the feature vectors, which are extracted by the wavelet theory, as the input vector of neural network for training. In order to get rid of the shortcomings that neural network training is easy to fall into local minimum, the paper uses genetic algorithm to optimize the structure、weight and the threshold of the neural network, and then the better neural network is got. Finally, the CRH2EMU traction inverter fault detection platform is designed. The platform uses the graphical programming language Lab VIEW to compile computer program, called MATLAB procedures, and realize the fault diagnosis, verify the feasibility of fault diagnosis using genetic algorithm to optimize neural network.