Research On Statistic Distribution Modeling And Prediction Method Of Time Between Failures For On-Board Equipment Of CTCS

The proposal of "eight vertical and eight horizontal" high-speed railway network makes our country enter the era of rapid development of high-speed railway construction.High efficiency,safety and stability make high-speed trains gradually become the first choice for people to travel.How to make high-speed train run more safely and efficiently that gradually become the emphasis of railway research.At present,the research of fault prognostics in railway field mainly focuses on ground equipment,but high-speed train is the main part of passenger transport,once the fault will lead to inestimable losses.Time between failures can reflect the inherent evolution law of system or equipment failures,therefore,this paper takes the micro time between failures of on-board equipment as the main research object,then conducts statistical distribution modeling and reliability analysis on it,and establishes combined prediction model to predict it.The main research content of this paper is as follows:(1)By consulting a large number of literatures,this paper summarizes the research status of fault prognostics technology and time between failures in detail,and introduces the structure and function of on-board equipment.Based on the shift record sheet in quality analysis area of bullet train equipment and the historical record sheet of replacement of ATP on-board equipment accessories,the on-board equipment faults are divided into six categories and 16 fault types.The paper’s data source is that the fault information recorded in the ATPCU-LOG file of a type on-board equipment,then preprocesses the fault data,finally,uses the word segmentation technology and Apriori association rule data mining algorithm to obtain the fault rule library,laying a foundation for the follow-up study.(2)According to fault rule library,the paper extracts the micro time between failures sample sequence of on-board equipment from the historical fault data.Firstly,the paper carries on the statistical distribution analysis for sample sequence,and then assumes that the sample sequence obeys the two-parameter Weibull distribution or exponential distribution,and the model parameters are solved by the least square method and the maximum likelihood method respectively.Finally,the optimal statistical distribution model is obtained based on the grey correlation analysis method,so as to establishes the basic reliability model.Then the basic reliability analysis of on-board equipment is conducted according to the optimal statistical distribution model,mean micro time between failures,the micro failure rate function and basic reliability function of on-board equipment are obtained.(3)In order to solve the nonlinear problem of sample sequence,combined prediction model of the time between failures based on temporal decomposition is established.Firstly,the STL algorithm is used to decompose sample sequence into seasonal component,trend component and remainder component.Then the decomposed data are predicted by the echo state network,BP neural network and support vector machine prediction model respectively.Finally,the final prediction results are obtained by adding the predicted results of four kinds of combined models,and compared with three kinds of single prediction models.The paper found that the predictive effect of combined prediction model is better than single prediction model.The optimal combined model is ESN+SVM+SVM,that the prediction accuracy can reach 96.49%.To sum up in conclusion,this paper takes the system-level and component level micro time between failures of on-board equipment as the main research object,then establishes the optimal statistical distribution model for it,and the reliability analysis of on-board equipment is conducted.Then the paper establishes the combined prediction model based on temporal decomposition to forecast micro time between failures,simulation results show that predictive effect is better than single prediction model.The prediction results make the train driver and field maintenance personnel can predict failure condition ahead and make the corresponding countermeasures.This paper contains 60 figures,12 tables and 70 references.