ANFIS-based Calculating Model Research Of Rail Break Risk Event Probability

Rail break will cause the railway line failure,causing train delays,and may cause train derailment when serious,which will drive safety accidents.Therefore,it is the key to ensure the safety of railway lines to master the rule of rail break and to control the risk of rail break.According to the production process of dangerous and harmful factors in the national standard classification,causal factors of rail break risk events are divided into people,equipment,management and environment four categories in this paper.Based on the experience of railway management in China,these four classes of causal factors are further subdivided,and the key causal factors of rail break risk are determined.In order to realize the accurate positioning of the rail break risk on the railway line,and to accurately quantify the causal factors of rail break,this paper divides the continuous railway line into several grids with 200m as the basic unit.The different grids are regarded as different individual to be analyzed individually.Causal factors data of each grid are collected,to compute rail breaking probability of each grid,accurately identifying the space-time location of rail break risk,and spatial position accuracy of probability calculation of rail break risk events being controlled in hundred-meter level.This paper applies the Adaptive Neural-Fuzzy Inference System(ANFIS)in the calculation of rail break risk event possibility.With causal factors as input variables,the possibility of rail break risk event as the output variable,the ANFIS-based calculating model of rail break risk event possibility is built,to calculate the daily probability of rail break risk event occurrence of each grid according to the states of various causal factors.The possibility of the rail break risk event occurrence in each grid changes when the causal factors change,realizing the dynamic management of the rail break risk.In order to verify the validity of the ANFIS-based calculating model of rail break risk event probability,this paper selected 100 grids in the upward section from Zhulu to Shachengdong of Daqin line in Taiyuan Railway Bureau as research objects,the mileages of which are ranged from K190+000 to K210+000.The author collected the data of causal factors and rail break events from December 1st 2014 to March 5th 2015 with a grid as the basic unit.Based on the collected data,the proposed model was used to calculate the probability of the rail break risk event occurrence in each grid,and the validity of the model was verified by comparing the calculated results with the collected data of rail defects and rail break accidents.