Track Circuit Fault Diagnosis Based On Multiple Models And Evidence Theory

With the rapid development of railway,the status of railway transportation is increasing day by day.People also pay close attention to the safety and efficiency of railway operation.As an important link to ensure the normal operation of railway,the working state of track circuit is related to the operation and safety of the whole railway.The complex working conditions and various running conditions of track circuit determine that the occurrence of track circuit fault is inevitable.At present,the track circuit fault diagnosis is still mainly based on manual inspection and troubleshooting,and there are problems such as low troubleshooting rate and low efficiency,which also aggravate the work intensity of staff and increase unnecessary company expenses.How to combine the emerging technology to make the correlation diagnosis efficient and intelligent is a big difficulty for researchers to solve.Aiming at the fault diagnosis problem of ZPW-2000 R uninsulated track circuit,this paper fully combines the previous research experience and the specific situation of track circuit fault,proposes a fault diagnosis method based on multi-model and evidence theory.The research content of this paper includes the following parts:(1)The structural composition and working principle of ZPW-2000 R track circuit are introduced.On the basis of fully understanding the structure and principle of track circuit components,the overall operating principle of track circuit is analyzed and summarized according to the composition sequence of track circuit.According to the actual working condition of track circuit,the typical fault types and related electrical volume are summarized,and the four-terminal network model of track circuit is established.Fully combined with the data given by the manufacturer and the data obtained by modeling,the two verify each other and complement each other.(2)The artificial intelligence diagnosis model was established,and the BP diagnosis model,GA-BP diagnosis model,SVM diagnosis model and GSA-SVM diagnosis model were obtained by combining the intelligent algorithm with the track circuit fault diagnosis.All the four diagnostic models have indispensable diagnostic advantages,so they can be used as BPA(Basic Probability Assignment)sources for evidence theory.In general,the construction of the model is simplified and the use of the model is more comprehensive.(3)Establish a fault diagnosis model based on multiple models and evidence theory.The subjective weight of evidence was determined by the overall diagnostic accuracy of the model.The uncertainty and reliability of BPA constructed by the model output were used to determine the objective weight of the evidence body.The two were combined to reconstruct BPA,and the reconstructed BPA was fused.According to the rules of evidence theory,the diagnosis was made.(4)Compared with the single diagnostic model,the diagnostic accuracy of each fault category and the overall fault is improved.In this paper,the fusion of multiple diagnostic models is effective and feasible,and a new idea is provided for the utilization of multiple models.