Anomaly Identification In Temperature Data Of High-speed Train

High-speed railway plays an important role in transportation,environment and economy because of its large transportation capacity,fast speed,convenience and energy saving.With the progress of the times,the high-speed rail plays a pivotal role in solving the problem of inter-city traffic and increasing passenger demand.It has become the preferred mode of transportation,bringing huge benefits to the country.Therefore,people put forward higher requirements in the safety and stability of high-speed trains.Today,sensor networks have been established on high-speed trains.The sensors are distributed over key location,they can collect information,such as temperature,speed and stress,monitoring comprehensively the train’s working status during the running.The abnormal state of the high-speed train can be expressed by some physical quantities,such as the data of temperature and vibration.In this paper,anomaly identification is used in the temperature data which collected by the sensor,so as to grasp the working status of the components in the running part of high-speed train,such as gear,motors.It also can find the hidden dangers of train failures early,and prevent the occurrence of sudden accidents and stopovers.Based on the characteristics of temperature data collected by train sensors,this paper discusses the anomaly identification in three aspects,containing sequence anomaly,pattern anomaly and point anomaly.Then the paper expounds their definition,put forward detection methods,and verifies the reasonableness of methods.In the research of the point anomaly identification,this paper trains and learns the temperature data of high-speed train,finally,the 53 H and LOF are selected to solve this problem,which by starting from the global and local sequence.In the research of pattern anomaly identification,the sequence segmentation and sequence similarity measure of the subsequence are important steps.A new linear segment algorithm based on important points is proposed in this paper,by combining the top-down algorithm and special point algorithm.And Dynamic Time Warping distance is selected to measure the similarity of subsequences,because of inconsistent length in the sequences after splitting.Combining with abnormal index,a complete recognition system of pattern anomaly is constructed in this paper.In the research of sequences anomaly recognition,this paper solves two problems,which are sequence similarity measure and sequence abnormality discrimination.This paper introduces the similarity measure methods based distance,and defines the abnormality index for abnormality discrimination analysis.Besides,the paper defines the potential correlation matrix,which is a completely new way,and gives the abnormality discrimination threshold interval.