Feature Analysis Of Monitoring Data Of High-Speed Train Bogie Based On Time-Frequency Atom Approach

With the rapid development of high-speed railway, the traffic mileage continues to increase, China has entered the High-speed railway age. However, the decline of high-speed train’s safty status in the process of long-term service will threaten the running safety and quality of the train. Currently, mass monitoring data, which contains a large number of train running status information, has been obtained through sensors installed to track and monitor the key part of the train. Critical to the train’s running status recognition and safety assessment is extracting the useful information from the monitoring data. Considering the monitoring data is nonlinear and non-stationary, this thesis analyzes the high-speed train’s monitoring data with time-frequency atom approach which has a good effect on non-stationary signal analysis. This thesis mainly researches on the four statuses including normal status, air spring without gas, anti-hunting damper demoliton and lateral damper demolition, the main work and results are as follows:1. Analyze the functions of high-speed train bogie’s key components of air sping, lateral damper and anti-hunting damper, then analyze the characteristics of monitoring data perliminarily both in the time and frequency domain to diagnose the differences of different statuses. Make the monitoring data pre-processing plan aiming at eliminating the abnormal value and getting the useful information concenteating in the range of 0-20Hz.2. Further study of the process of time-frequency atom approach and propose the energy attenuation rate feature. Carry out the experiment of extracting feature of the entire demolition of fault monitoring data in time and frequency domain separately. The support vector machine is adopted to achieve the status recognition of the seven kinds of statuses which contain normal status of train, air spring without gas, lateral damper demolition, anti-hunting demolition, anti-hunting demolition and lateral damper demolition, anti-hunting demolition and air spring without gas, lateral damper demolition and air spring without gas, then analyze the experiment result based on the background of the monitoring data.3. In order to get close to the actual condition, this thesis studies the single damper fault status. Firstly, combine the fault types by the installation sites of anti-hunting dampers. Based on the analysis of the recognition result of support vector machine, propose the multi-channel-combined recognizition scheme to judge the fault location. Take the bogie for the target to recognize by analyzing the distribution of the feature vector of monitoring data of single lateral damper faults in the feature space. Finally, judge the the fault bogie.This work was supported by National Natural Science key Foundation of China. (No.61134002)...