Based On Empirical Mode Decomposition And Genetic Neural Network Of Railway Vehicles Bearing Fault Diagnosis Research

As one of the most important units in the rail vehicle, the rolling bearing has magnificent meaning to the reliable running of the railway vehicles. So it is important for accurate and efficient fault diagnosis of the rolling bearing in rail vehicles and which issue needed to be addressed. This dissertation generalizes the previous achievements, and the railway vehicle rolling bearing fault diagnosis method which combines EMD(Experience model decomposition) and the RBF neural network optimized by genetic algorithm is proposed.First, the paper introduced the mechanism of railway vehicles rolling bearing failure, failure modes and causes and the vibration model. Common methods of characteristic information extraction for rolling bearing fault and their respective characteristics are discussed. Simulation experiments using wavelet packet analysis and EMD were done, the result of which proved that the wavelet packet analysis and EMD can be effective for bearing fault characteristic information extraction.Then, BP and RBF neural network as two typical neural networks are selected for the trouble-mode recognitionare, and the four fault diagnosis models of the wavelet packet-BP, the wavelet packet-RBF, EMD-BP and EMD-RBF are established. The simulation results with Benchmark data have proved that EMD has an advantage over the wavelet packet on fault characteristic information extraction and RBF has better results than BP on trouble-mode recognitionare.Afterward, the RBF neural network parameters were optimized using genetic algorithm for further enhancing performance. The wavelet packet-GA-RBF and EMD-GA-RBF models are established. The simulation results with Benchmark data have proved that GA-RBF has been greatly improved performance compared to the RBF neural network.Finally, based on above research and analysis, the wavelet packet-GA-RBF bearing fault diagnosis model and EMD-GA-RBF bearing fault diagnosis model were established respectively, and simulation experiment with the measured rail vehicles bearing fault data were done. The results prove that the model based on EMD combined with the genetic algorithm optimized RBF neural network can be used to track vehicles bearing fault diagnosis.