Research On Fault Diagnosis Method Of High-Speed Railway Turnout Based On Convolutional Neural Network

As one of the important ground signal equipment in the high-speed railway operation system,the main function of turnout is to change the direction of the train in order to realize the turning and cross-line operation of the train.However,it has the characteristics of large number,short life and frequent faults.In order to ensure its safe and reliable operation,it is of great significance to study intelligent fault diagnosis methods to promote the rapid development and safe operation of high-speed railway.At present,the methods used for turnout fault diagnosis often face some practical problems,such as the dependence of feature extraction on experts’experience and the small amount of valid label sample data.In this thesis,an intelligent fault diagnosis method for high-speed railway turnout based on convolutional neural network is explored,and the relevant diagnosis algorithm is realized by using the data of turnout action current curves to drive convolutional neural network model training.Firstly,the working principle of high-speed railway turnout and the structure of S700K AC electric switch machine are analyzed.By establishing simulation model based on turnout control circuit,the shape of action current curves are studied and their fault mechanisms are explored.Secondly,the real and effective turnout data are used to train the neural network model,and the advantage of convolutional neural network in the deep learning model for automatic feature extraction is given full play so that the intelligent fault diagnosis technology of high-speed railway turnout with self-reliance and high efficiency is explored and studied.The specific work is as follows:(1)Active current curves analysis and control circuit simulations modeling.On the basis of introducing the working principle of high-speed railway turnout and S700K electric switch machine,in this thesis,the normal operating current curve of turnout and several common fault operating current curves are summarized,and the turnout control circuit is simulated through the software of Matlab.By simulating its normal state and several fault curve shapes,the operating mechanism is explored.(2)A turnout fault diagnosis method based on LeNet-5 is proposed.Taking the typical LeNet-5 model as an example,the basic principles of neural network and convolutional neural network are summarized in this thesis.Then,actual turnout data are input into convolutional neural network for training by transforming one-dimensional data into two-dimensional gray images,compared the classification accuracy of the test sets when the number of neurons in the whole connected layer are different.By comparing with other traditional machine learning classification algorithms,it is shown that the proposed fault diagnosis method has good superiority.(3)A hybrid turnout fault diagnosis method based on DCNN-SVM is proposed.In view of the fact that the actual turnout fault data are less,and the advantage of convolutional neural network in automatic data features extraction.Then referring to the design structure of convolutional layer in VGG-Net model,two or three convolutional layers of smaller convol utional filter banks are used instead of one convolutional layer of larger convolutional filter banks to extract data features in this thesis,and the support vector machine classifier also has good classification performance when dealing with small amount of data.According to these,a deep convolutional neural network model in line with the actual situation is proposed.Finally,compared and analyzed the experimental results by using different kernel functions under different numbers of fully connected layer neurons.The fault diagnosis method based on convolutional neural network can effectively extract the features of turnout action current curves automatically,and further improve the accuracy of training model and eliminate the influence of manual feature extraction.The proposed method can effectively adapt to the complex working environment of high-speed railway turnout and improve the efficiency of turnout fault diagnosis.It has important application value in guaranteeing the safe and reliable operation of high-speed railway.