Fault Diagnosis Of Track Circuit Based On Unsupervised Learning

As one of the key equipment in railway signal system,the working state of track circuit directly affects the safety and efficiency of railway operation.Because of the characteristics of many kinds of equipment,complex structure and long working-time,the equipment failure of track circuit is frequent.At present,some intelligent monitoring systems have realized the collection of the track circuit electrical characteristics and set the threshold value according to the track circuit adjustment table,when the collection curve exceeds the threshold value,the corresponding fault alarm will be given.However,in some abnormal cases,the track circuit will fluctuate below the threshold value.At present,there is no good method to diagnose these faults.In this paper,the main rail out voltage data is taken as the research object,and the unsupervised learning method is used to establish the fault diagnosis model.Firstly,the monitoring data of the track circuit is analyzed,and the main rail out voltage is selected as the research object,and given the reason.According to the characteristic of the main rail exit voltage curve,the typical faults are analyzed by using Rheinda criterion and other methods.Then,the main rail voltage is extracted by statistical method,and the dimensionality is reduced by PCA and correlation analysis.Secondly,the related methods of unsupervised learning are analyzed,and k-means algorithm,CLARA algorithm,PAM algorithm and DBSCAN algorithm are used to establish clustering models,and model evaluation and comparison are performed.Then use weighted voting to do integrated clustering.Compared with the previous models,the analysis shows that the effect of integrated clustering is better.Finally,the ZPW-2000A Track Circuit Fault diagnosis system is developed according to the designed clustering integration model.The user login module,data interface module,Data pre-processing module,Cluster Integration Model Information Module,Fault Diagnosis Module,History Query Module,Fault Storage and Statistics module are designed,and carry out system flow analysis.