The Characteristic Analysis Of High Speed Train Monitoring Data

Along with our country’s railway leap-forward development, High-Speed Train has become an important part of transport system, The High-Speed Train normal operation has an great significance in the economic development and personal security. However, because of the improvement of the train speed, the vibration has become more severe, the wheel force and sinusoidal influence has become the most important hidden trouble, we collect lots of data from the train sensors, how to analysis from vibration out of the train running status and situation is the most important to safe operation, most scholars used frequency domain method to analysis, but the premises are assumed to non-stationary data, whether it smoothly or not has not been shown, therefor, how to start from the characters of data to analysis is the key to the research of the page, this paper analysis large number of monitoring data to find out the answer, Among four condition such as Original car, Spring airless, No transverse shock absorber, No resist sinusoidal shock absorber to analysis to find out the conclusion.This page is based on research projects of monitoring data of High-Speed train service under the safety state evaluation of the key issues, Analysis and research parts of stationarity, correlation and fusion feature, the main contents are as follows:(1)Stationarity test is the precondition of random signal to analysis, At first. This paper select several kinds of representative method for stationarity signal test to comprehensive narrative, including autocorrelation image observation method, reverse column test, rounds test and ADF test, Compare the method and look for the pros and cons, select several with progressive relationship method to analysis High-Speed Train monitoring data. The sensors need to be test are chosen from original car and three fault conditions of nine categories of acceleration and displacement, for no transverse shock absorber condition speed under 200km/h, therefor, analysis 40km/h to 200km/h six speed statistical is feasible, in short sampling distance different kinds of conditions showed non-stationarity, but long sampling distance often show the characteristics of stability, so in the processing of stability analysis in monitoring data should select the appropriate method.(2)Correlation analysis is used to analyze the degree between interval variables, this page analyze the sensor of same condition with Pearson correlation, threshold Cr=0.8 is cut-off point, screening the redundancy of high correlation sensor, analyze the relationship between high correlation data and train operation of monitoring data, then statistic the three parameters in sensor with low correlation in the table, to find a kind of sensor with the most representative.(3)Data fusion is used to simplify large number of monitoring data between original car and fault condition, this page take the same working condition of the same kind of direction(X,Y,Z) on sensor fusion analysis and compare different kinds of working conditions at the same time, the experimental analysis indicates that the features of no transverse shock absorber condition is obviously in direction Y, the sensor in direction Z is not obvious in fault condition, choose the sensor of fault condition with representative feature.