Fault Diagnosis And Prediction Of S700K Switch Machine

The switch machine is one of the most important equipment in the railway system.It has the characteristics of frequent movements and complex working conditions.If there is a failure,it will have a serious impact on railway operations.However,the maintenance methods currently used are mostly based on experience and knowledge.This method judges the type of fault based on the experience of experts or staff,and then repairs it,which has the problem of relying on labor and low efficiency.In order to solve the above problems and realize the prediction of failures so as to reduce maintenance costs and ensure the safe and efficient operation of the railway,this thesis studies the fault diagnosis and fault prediction of the switch machine.(1)Analysis of working principle of the switch machine and data preprocessing.Study the reason for the formation of the normal operating power curve of the switch machine,and analyze the power curves of 8 common faults.Since it is hard to obtain the fault data of the switch machine,using the synthetic minority oversampling technique(SMOTE)to solve the problem of imbalance between normal data and faulty data.According to the characteristics of the power curve,divide the curve into stages from the time domain and frequency domain respectively.Calculate the features at each stage,and perform dimensionality reduction.(2)Fault diagnosis.Use the method based on multi-layer perceptron(MLP)and the method based on support vector machine(SVM)to perform the fault diagnosis research respectively.The results show that the method based on support vector machine has the advantages of being more accurate and efficient.The 10-fold cross-validation results show that the accuracy has been improved and it is practical.Aiming at the problem of unknown faults,this thesis conducts the research based on support vector data description(SVDD)to classify known faults and unknown faults at the same time.The result shows that the accuracy of this method is relatively high,and there is no misreporting of unknown faults as normal or known faults,and it is practical.(3)Fault prognosis.Use the model-based method to obtain the degradation signal curve of the switch machine,and the fault prediction research is carried out based on this.Carry out fault prediction research based on support vector machine(SVM)and multilayer perceptron(MLP)respectively,and compare them.The results show that the method based on the multi-layer perceptron(MLP)is more useful which can effectively predict the development trend of the degraded signal of the switch machine.Figure 63,Table 16,Reference 56.