Research On Fault Diagnosis And Health Prognostic Of High-Speed Train Bogie Based On Data-driven Technology

With the rapid development of rail transit,the high-speed train gradually become the main body of urban transportation in the "made in China of the 2025 plan" as the driving the development of green,the important role of stimulating the regional economic,with the rapid development of it,its design,manufacture,use,especially the protection and security cost is more and more higher,so how to improve the efficiency of train system safety,operation and maintenance,to reduce the use and care,with minimal maintenance in the as and autonomous maintenance is the real-time security has become a relevant enterprises and the attention of the researchers.The traditional state monitoring and early fault warning has been unable to meet demand,so this paper by high-speed train bogie system as the research object,carry out failure prediction based on the high-speed train bogie system and health management research,the real-time running state monitoring system,through the failure analysis,trend analysis and combined with the operation of the system historical data,prediction of key system components of real-time health state,carry out new maintenance model research based on the Condition-based maintenance and the predicting maintenance,the thesis is based on data driven high-speed train bogie forecast about the key technology of fault diagnosis and health,and the paper main research content includes:(1)Based on the theories and standards related to Prognostics and health management,the technical framework of high-speed train PHM system based on the key technologies of Prognostics and health management and the PHM system architecture suitable for high-speed trains are proposed.(2)The high-speed bogie system is based on the function and structure,analysis of the main components of the high-speed train bogie system failure mode and failure mechanism of the construction of high-speed train bogie and the vehicle dynamics model,through the use of the SIMPACK software simulation high-speed train different operation condition and failure mode,including the research object model parameter setting,orbital excitation spectrum,different working condition limit,realize data acquisition under different working conditions and related data pretreatment,etc.(3)To study the high-speed train bogie PHM system of real-time data acquisition and feature extraction,in different conditions,different sensor data acquisition channels get the vibration acceleration signal,the time domain characteristics,based on ICEEMDAND adaptive decomposition of a variety of information entropy and the way of a coarse graining of scattered multiscale entropy parallel feature fusion,realizes the multi-channel multiple domain fusion feature extraction,the complete characterization of the bogie system under different working condition of health.(4)Fault diagnosis of high-speed train bogie based on multi-domain feature fusion and SVM optimization is studied.Single channel multi-domain feature fusion and multi-channel multi-domain feature fusion,minimum redundancy maximum correlation(mRMR)feature selection method was used for feature dimensionality reduction,and the proposed fault diagnosis method was verified to be effective in real-time fault type judgment through comparative analysis of support vector machine(SVM)fault diagnosis with different optimization algorithms.(5)The health condition prediction of high-speed train bogie based on LSTM circulating neural network is studied.Build can reflect the train monitoring the state of the health factor curve,root mean square value and kurtosis as HI training LSTM forecast model,with real-time data update forecast model,predict the train of real-time health state,RMSE and root mean square error and the mean absolute error evaluation prediction algorithm performance,verify the validity of the algorithm and the robustness.