Research On Railway Switch Fault Diagnosis System Based On Ensemble Learning

High-speed railway plays an important role in the railway transportation system of China.Railway Switch serves as a vital part of track structure for high-speed railway and its condition play an important role in train running safety and quality.The traditional fault diagnosis technology of railway switch is time-consuming and laborious.The accuracy is not high.It is still an important and urgent problem for railway staff to improve the accuracy and efficiency of fault diagnosis.Therefore,the research on the fault diagnosis of railway switch device has a great significance.According to the research topic“ Railway Electrical Service Fault Diagnosis Based on Big Data Processing”in the institute of applied mathematics of Hebei academy of sciences,this paper takes the current data of railway switch monitoring as the research object and puts forward a fault diagnosis model of railway switch based on ensemble learning.In order to improve the accuracy and efficiency,the ensemble learning’s stacking algorithm is adopted to combine the traditional classifier with the deep learning.The main research contents include:(1)Analysis and preprocessing of railway switch current data: based on the analysis of the Railway Switch current data,the current waveform of different railway switch represents different railway switch fault types.Aiming at the problem of inconsistent data dimensions,Hermite interpolation algorithm is adopted to unify the dimensions of data in the process of data preprocessing.(2)Research on railway switch fault diagnosis based on traditional classifier:Support Vector Machine,BP neural network and Decision Tree are used to preprocess the data of railway switch fault diagnosis.By optimizing the parameters,the accuracy is 90.16%,85.53% and 83.39% respectively.(3)Research on railway switch fault diagnosis based on ensemble learning: in order to improve the fault diagnosis rate of the algorithm,stacking algorithm isproposed.The algorithm is divided into two layers.The first layer adopts SVM,BP neural network and decision tree for fault diagnosis.In the second layer,a meta classifier is used.The deep belief neural network and the one-dimensional convolution neural network are used as the meta learners respectively to constructed two kinds of railway switch fault diagnosis models based on ensemble learning.By comparing the two models,the accuracy of the latter is higher and its time consumption is shorter.(4)Railway switch fault diagnosis system of high-speed railway: railway switch fault diagnosis system of high-speed railway is designed and developed by Pycharm,Visual Studio,SQL Server database and other tools.Its aim is to help the railway staff to judge whether the railway switch is in fault and identify the type of fault in time.