Research On Fault Diagnosis Method Of High-Speed Railway Turnouts

As the rapid development of railway transportation in China, transport efficiency has been increased greatly. At the same time, it brings heavy maintenance pressure to one of the important signal equipment-turnout, especially in high-speed railway and passenger dedicated.However, the periodic repair method for maintenance of turnout is generally adopted by major domestic railway companies, and fault diagnosis for turnout mainly relies on maintenance experience. So the operation conditions of turnout cannot be acquired in time. Maintenance staff whose experience is not rich cannot accurately determine the cause of the fault turnout and ant maintenance it as quickly as possible. This will lead to the condition that failure of the turnout could not get timely or proper maintenance, and influence the efficiency of rail transport, it may cause heavy loss of life and property.In order to solve the problems above and achieve the condition based maintenance, it is very necessary to design the fault identification and diagnosis method for turnout in high speed railway, so the fault information of turnout and possible causes can be provide to the maintainer for troubleshooting, get rid of periodic maintenance mode, and enhance the maintenance efficiency.In this paper, the principles and failure modes of high-speed rail turnouts are described, because it is difficult to establish precise mathematical model of turnout, but turnout various modes of conversion power and current curves have different characteristics, so the machine learning methods-Support Vector Machine (SVM) is used to constructed empirical model of the above curve data for fault diagnosis, and how to use SVM construct diagnostic methods is focused on for optimal diagnostic performance. The main work is as follow:1) Present research review on fault diagnosis for turnout and fault diagnosis requirements analysis:with the analysis on current situation and fault diagnosis problems of turnout, the requirements of its fault diagnosis requirements are proposed. Some common characteristics of fault diagnosis method are analyzed, the core algorithm Fault diagnosis method is selected combined with the above requirements.2) Summary and analysis of failure modes:failure sample data is obtained from railway field through survey, common failure modes are classified by exchange with the maintenance experts, and the reasons for corresponding failure patterns are analyzed combined with the actual working principle and structure of turnout.3) Study on fault diagnosis Method:sample data is classified with summary of the failure mode before, the SVM is used as the core combined with a variety of feature extraction algorithm and optimization algorithm to construct diagnostic methods, the classified samples are tested and the performance of various methods is compared, the optimization algorithm based on grid search parameters is used for SVM fault diagnose methods.4) Fault diagnosis software design:the software functional requirements are determined by exchanging with railway field maintenance experts, and software modules and system function are designed according to the requirements.5) Diagnostic software field test:the performance of the software fault diagnose is tested by a simulated outdoor environment, the software is also installed on the computer based monitor for the furfure test and improvement of performance.The turnout fault diagnosis software is designed and its corresponding diagnostic methods are studied. Diagnose software in the testing phase get a satisfactory results, it prove the feasibility of this diagnose method.