| Railway signal system plays an important role in the safe operation of rail transit.At present,ZPW-2000R type jointless track circuit widely laid in China’s railway section,is an important part of railway signal system.Track circuit works outdoors all year round,and the environment is complex,so it is a fault prone equipment.However,the existing fault diagnosis mainly depends on the regular maintenance of the field staff and the analysis of the data monitored by the microcomputer,and the judgment is made through the circuit principle.How to prevent the fault as soon as possible,find the fault in time,locate the fault accurately and repair the fault faster is an urgent problem to be solved.Therefore,this paper studies the fault diagnosis method of ZPW-2000R track circuit based on neural network.(1)Through in-depth analysis of the track circuit principle,common fault types are obtained,orthogonal experiments are designed,and the monitoring data are obtained through the laboratory track circuit simulation track bed experiment.The data are sorted and analyzed,and different network structures are divided according to different situations.(2)In view of the situation that there are many monitoring data and the network model is complex when there are outdoor monitoring equipment,a fault diagnosis model based on multiple parallel BP neural networks is proposed.The fault types are divided into three categories first,and then 29 fault modes are distinguished.The simulation results show that the fault diagnosis accuracy is high.In view of the imperfect monitoring data without outdoor monitoring equipment,a fault diagnosis model based on RBF neural network is proposed.Through simulation experiments,it is proved that the diagnosis accuracy of sending channel fault and track fault is high,while the diagnosis accuracy of receiving channel fault is low,which can provide reference for field staff.(3)In view of the situation that there are many monitoring variables and complex fault types with outdoor monitoring equipment,a fault diagnosis model based on convolution neural network is proposed.The model can directly and quickly determine the probability of 29 kinds of faults,which can provide guidance for field maintenance personnel and judge the fault types more quickly.From the diagnosis results,it can be seen that BP neural network,RBF neural network and convolution neural network are all suitable for ZPW-2000R track circuit fault diagnosis,BP neural network is more suitable for dealing with more comprehensive monitoring variables,RBF neural network also has good effect for data missing,convolution neural network is suitable for multi classification problems,and can quickly judge 29 kinds of faults type.On the one hand,combined with the monitoring data collected by the computer,this paper uses neural network algorithm to realize the intelligent fault diagnosis of track circuit,which replaces the original fault diagnosis method relying on the experience of maintenance personnel and circuit principle,and is more efficient and accurate.On the other hand,this paper comprehensively summarizes 29 kinds of fault types of track circuit,and the fault diagnosis accuracy is high.Finally,the research contents of this paper are summarized,and the future development of track circuit fault diagnosis is prospected. |
