Research On Fault Diagnosis Of S700K Switch Machine Based On Improved KPCA Algorithm

As one of the three major parts of railway outdoor signal equipment,the switch machine is put into use in a large amount and is an important part of railway turnout system.At present,S700K switch machine is mostly used on high-speed railway,but its structure is complex and the working environment is changeable,and the fault frequency has been rising,among which electrical faults account for a large proportion.The condition monitoring of the switch machine on the railway site mainly depends on human observation.According to the previous work experience,the fault is judged by comparing the changes of the power curve.The fault identification time is long,and it is easy to misjudge and miss judgment.The traditional regular maintenance mode is also easy to cause equipment under maintenance or over maintenance,and the work efficiency is low.This requires the railway site to adopt a more efficient and safe intelligent diagnosis method to ensure the normal operation of S700K switch machine,so as to effectively ensure traffic safety.Therefore,the research on intelligent diagnosis method based on power curve is of great significance to help railway field maintenance.The main work of this thesis is as follows:(1)Taking S700K switch machine used on high-speed railway as the research object,the changes of power curve under its normal operation and six common fault conditions are analyzed,and its fault phenomena and causes are summarized.Ten time domain eigenvalues are extracted from their power curves to form the common fault feature set of S700K switch machine.(2)The KPCA data dimensionality reduction method is used to sort and reduce the dimensions of the selected 10 features to verify the effectiveness of KPCA algorithm in dimensionality reduction in 3D projection space.The KPCA algorithm is improved,and the optimal feature subset which is most suitable for the fault diagnosis model is selected to reduce the feature dimension and diagnosis time while ensuring high fault diagnosis accuracy.(3)The SVM method is used to construct the fault diagnosis classifier,and the SVM classifier is used as the base learner to diagnose the fault of the power curve characteristic data of S700K switch machine.AdaBoost algorithm is introduced to iteratively integrate the diagnosis results of SVM classifier,design new voting rules,and combine all weak classifiers into the final strong classifier.The experimental results show that the diagnosis accuracy of the method proposed in this paper can reach 100%,and the diagnosis time and accuracy can meet the requirements of railway field.At the same time,four intelligent fault diagnosis methods commonly used in previous research are used to verify and compare with this method,which proves the effectiveness and superiority of the method proposed in this paper.