Research On Intelligent Fault Diagnosis Method Of Metro Turnout Equipment Based On Action Current Data

Due to the long-term outdoor operation of metro turnout equipment and complex working conditions,the fault frequency is high.However,the fault diagnosis of the existing turnout equipment is still dominated by human,which has the defects of low efficiency and poor reliability,and has brought adverse effects on the train operation safety.Intelligent fault diagnosis method based on data can significantly improve the efficiency and accuracy of fault diagnosis.It has attracted much attention in the field of fault diagnosis of turnout equipment,but there are still some deficiencies in the current research.First,the existing intelligent fault diagnosis methods based on time-series data lack of time-domain characteristic indexes that can reflect the energy barycenter characteristics of fault data;Second,when using the traditional convolutional neural network(CNN)algorithm for fault diagnosis,the time-series feature loss caused by the dimension transformation of fault data is not considered.Based on the investigation of relevant literature at home and abroad,aiming at the problem that the time-domain characteristic index in the existing fault diagnosis methods can not reflect the energy barycenter distribution of data,this thesis constructs three indexes to describe the energy distribution characteristics of action current data of turnout equipment,and proposes an intelligent fault diagnosis method for turnout equipment based on energy distribution characteristics and support vector machine(SVM).In addition,in order to further improve the efficiency of fault feature extraction of action current data of turnout equipment and solve the problem of time-series feature loss caused by the dimension transformation of fault data,a feature extraction model based on one-dimensional convolutional neural network(1DCNN)is designed,and an intelligent fault diagnosis method for turnout equipment based on one-dimensional convolutional neural network and support vector machine(1DCNN-SVM)is proposed.The main research work of this thesis is as follows:(1)The working principle of metro turnout equipment is explored,and the simulation model of action circuit for ZD6 switch is established.The control process of control circuit of the switch is mainly combed,and the common faults of the turnout equipment are analyzed according to the action current data of the control circuit.Under the background of insufficient existing fault data,the simulation model of the action circuit in the control circuit is established to simulate the operation under normal conditions and six types of fault conditions.At the same time,the simulated current data is compared with the real current data,which proves the effectiveness of the model,and provides additional data for the subsequent verification of the effect of fault diagnosis methods and comparative experiments.(2)An intelligent fault diagnosis method for turnout equipment based on energy distribution characteristics of action current data and SVM is proposed.Aiming at the problem that the existing time-domain characteristic indexes can’t reflect the energy barycenter distribution of fault data,three indexes describing the energy distribution characteristics of action current data of turnout equipment are constructed.The data after extracting the energy distribution characteristic indexes through windowing is inputted into the SVM model for fault identification,and it is compared with the other two commonly used methods to verify the effectiveness and superiority of this method in turnout equipment fault diagnosis.(3)An intelligent fault diagnosis method for turnout equipment based on1DCNN-SVM is proposed.In order to further improve the efficiency of fault feature extraction of action current data of turnout equipment and solve the problem of time-series feature loss caused by the dimension transformation of fault data,a feature extraction model based on 1DCNN is designed.The action current data is inputted into the 1DCNN model for preliminary training,and then its full connection layer data is extracted as a feature set and inputted into the SVM model for fault identification,so that the intelligent fault diagnosis method of turnout equipment based on 1DCNN-SVM is formed.The effectiveness and superiority of this method in fault diagnosis of turnout equipment are verified by comparing this method with other two commonly used methods.