Research On Fault Diagnosis Of Emu’s AC Drive Main Circuit

With the increasing speed and density of high-speed rail,fault diagnosis system of high-speed EMU’s main circuit becomes more and more important.Firstly, this paper introduces the urgency and the significance of fault diagnosis of high-speed EMU,and illustrates some common methods which have been applied to the fault diagnosis system of high-speed EMU’s main circuit.This paper analyzes some deficiencies of this research field in China and key technologies for this field.Then it puts forward the research approach to fault diagnosis system based on artificial intelligent.The object of this research is CRH2’s AC drive main circuit.This research analyzes the principle and the structure of the main circuit of CRH2’ AC drive and summarizes the reasons of CRH2’ fault on the basis of trouble records.Secondly,three-level inverter simulation mode is built by Matlab /simulink.This pager simulates and analyzes the faults of the power switch device,and samples the signal in fault situations.Then it extracts fault signal amplitude and analyzes the signal in a method which uses wavelet package analysis to select and extract the feature.Finally, this paper creats and trains a back propagation nerual network to diagnose the fault,and further optimizes the parameters of nerual network by PSO and QPSO.The result of this paper proves that:the method of BP artificial neural network based on L-M algorithm can converge fast and reach a great accuracy.The artificial neural network optimized by PSO converges more fast and the accuracy is great.The artificial neural network optimized by QPSO not only converges fast but also can prevent training from getting into local extreme point.