Approaches To Bearing Fault Diagnosis Of Go Line In Subway Trains

The rolling bearing system plays an important role to the safety of metro train.When the defect of the rolling bearings occurs,the vibration signal exhibits unusual behavior with non-linear and non-Gauss properties.Higher Order Statiscs(HOS)is powerful tool to deal with Non-Linear and Non-Gaussian signals.In these years,the Non-Linear?Non-Stationary and Non-Gaussian signal processing tecchniques have been paid more attention to in machine fault diagnosis in order to satisfy the demand of precise diagnosis.First,it presented Artificial neural network(ANN)and the HOS theory,and studied the bisspectrurn analysis method,algorihm,characteristics and neural network learning and neural networkdiagnosis,then put them into the fault diagnosis of rolling bearing.Here,the bi-spectrum analysis is applied to extract fault information from the vibration signals of outer ring,inner ring,rolling body and normal condition.Based on bi-spectrum of typical fault feature vectors and six commonly used non-dimensional indexes: kurtosis,kurtosis,skewness,margin,waveform and pulse,the positions of fault rolling bearing can be identified by Back-Propagation(BP)neural network configuration.There different situation ware studied:(1)different loadings.(2)different serious of the same fault under the same loading.The experimental data shows that this novel method,combined BP neural network configuration with multi spectral features,implies high accuracy in fault classification and fault identification,the bispectrum analysis method is feasible and effective,thus,provides a new conception for the intelligent fault diagnosis of rolling bearing.