Research On The Fault Diagnosis And Remaining Life Prediction Of Metro Train Running Gear’s Rolling Bearing

With the rapid development of metro in China, overload has become the norm in metro train operations, long-term overload running is bound to break the safety operation life of metro trains. Running gear’s rolling element bearings are key elements in metro trains, and are one of the multiple fault components, it directly affects the safety of metro train. Therefore, research on the fault diagnosis technology of running gear’s rolling bearings is of great practical significance to the safety and reliability for metro trains. In this thesis, rolling bearing operation experiments are designed to approach various operating conditions and multi speed running. And vibration acceleration signal is collected with a signal collecting device, the reasonable fault feature is extracted, two new bearing intelligent fault diagnosis algorithms is proposed, namely SVM-AdaBoost and BP-AdaBoost algorithm. The result shows that these two algorithms are significant in application, and can meet the needs of the rolling bearing fault diagnosis. Meanwhile, according to bearing experimental data in the whole life cycle, bearing performance degradation degree is perceived, the remaining life prediction method of Grey Model based on EMD is studied, which is of great importance in preventing the accident, maintenance plan and the repairing process.The main research of this paper is as follows:(1) In order to collect the vibration signal of bearings, experiments of fault diagnosis for metro train’s rolling bearing is designed. To simulate the running condition of the metro train, the experimental scheme of vibration signal collection for rolling bearing is designed under various working conditions.And the vibration signal are taken as the basic state parameter of fault identification.(2) Fault feature is extracted based on vibration signal of rolling bearing. Time domain analysis of vibration signal analysis, spectrum analysis and wavelet packet decomposition are included, the fusion of time domain and the wavelet packet energy feature parameter are used as the data input of the next step intelligent fault diagnosis algorithm.(3) Rolling bearing health status is identified by SVM-AdaBoost and BP-AdaBoost intelligent AdaBoost algorithm, which are separately based on the support vector machine (SVM) and BP neural network as the sub classifier. Two new multi classification algorithms are built to solve the problem of multiple fault classification problems. Compared with the single SVM classifier and BP classifier network fault diagnosis results, the results show that the proposed integrated learning algorithm of SVM-AdaBoost and BP-AdaBoost has great superiority in the diagnosis accuracy.(4) The prediction method of bearing remaining life is researched. According to bearing signal data in the whole life cycle, the traditional time-domain analysis and fatigue life prediction method of Grey Model (GM) prediction bearing based on EMD is studied and compared. The mean value of the IMF’s envelope amplitude is used as the regression feature, which reflects the bearing health condition. The result shows that residual life prediction method of GM based on EMD can accurately predict the remaining life of the rolling bearing.(5) On the basis of the research method and actual survey, metro train bearing fault diagnosis and life prediction software system is designed, thereby realizing in-transit fault diagnosis and remaining life prediction, achieving the purpose of train health status real-time monitoring, guaranteeing the safety and reliable operation of metro train.