Research On Compensation Capacitor Fault Diagnosis Of Track Circuit Based On Neural Network

Track circuit is an important basic equipment of train control system,its working condition is directly related to the efficiency and safety of train operation.Ensuring the healthy operation of track circuit is the key link to the vigorous development of railway enterprise.However,the fault diagnosis technology of track circuit can't meet the needs of the rapid railway development in our country,how to diagnose the track circuit fault timely and accurately is an issue that needs to be addressed urgently.Therefore,it has important practical significance and applicable value to study the fault diagnosis technology of track circuit.In this paper,the ZPW-2000A track circuit is used as the study object and the emphasis is on the study of compensation capacitor fault.The main work of this paper is summarized as follows:First of all,based on the basic structure of the ZPW-2000A track circuit,the model of an unoccupied track circuit and the short-circuit model of an occupied track circuit are established based on transmission-line theory.The accuracy of the models is verified by comparing the simulation data from the built models with the test data from experimental platform.The causes of errors are analyzed at the same time.Secondly,according to the shape and characteristics of short-circuit curve,a method based on improved included angle chain is presented to describe the short-circuit curve of track circuit.On this basis,through calculating the difference between the normal curve and the curves of different compensation capacitor fault models,the fault location and fault type can be realized.Then,the study discusses the influences of single and two compensation capacitor faults on the data of an occupied&unoccupied track circuit under different railway ballast resistances.The characteristic parameters of an occupied&unoccupied track circuit are determined and extracted by correlation analysis.Finally,the study builds a fault diagnosis model for the occupied and unoccupied track circuit.The compensation capacitor faults of the occupied and unoccupied track circuit can be diagnosed effectively base on PNN,PSO-SVM and BP algorithm.The diagnostic effect of the three models can be get from the prediction accuracy and prediction time.It is concluded that PNN algorithm has great advantages over prediction accuracy and prediction time for the fault diagnosis,which shows that this method has practical value and significance.