Research On Fault Diagnosis Method Of Multifunctional Vehicle Bus

As the medium of train information transmission,the train communication network has the functions of controlling train operation,detecting and diagnosing vehicle failures.Its normal operation is an important guarantee for the safe operation of trains.Therefore,in order to ensure the reliability of the Multifunctional Vehicle Bus(MVB)operation,speed up the troubleshooting of the MVB,technologies such as health management and network fault diagnosis have become research hot spots in the field of train communication network.The research on MVB fault diagnosis technology is still in its infancy.The on-site network troubleshooting mainly relies on the work experience of network engineers.It is gradually eliminated through manual repairs such as replacing equipment and cables.The work efficiency is low and it is easy to cause excessive maintenance.Therefore,in the context of the development of artificial intelligence diagnosis technology,in order to shorten the troubleshooting time of MVB and improve the security and reliability of MVB operation,this paper deeply analyzes the working principle of MVB and common network faults,introduces Support Vector Machine(SVM)fault diagnosis method diagnosing MVB faults,and conducts in-depth research on its modeling method.The main research results of this article are as follows:(1)MVB working principle is deeply studied.Based on the analysis of the physical waveform and link information of the MVB under different network failures,a feature extraction method combining the time-domain characteristics of the network waveform and the link data parameters is proposed;(2)The experimental platform is built to simulate six network conditions,such as bus open circuit and intermittent open circuit.The fault data collection of MVB is completed.The fault data are preliminarily processed separately by Min-Max standardization,Z-score standardization and decimal calibration standardization methods,and then PCA is used to reduce dimension of the obtained data Finally,the processing effect is compared through the experiment;(3)The MVB fault diagnosis model is established based on SVM.Three common algorithms,grid algorithm,genetic algorithm and firefly algorithm,are used to optimize the key parameters of SVM i.e.,penalty factor C and Gauss kernel function coefficient g.For this phenomenon that the optimal parameters of SVM are basically concentrated in the same area,a FA-Grid two-step optimization algorithm model which aims to have fast location before refined search is proposed and experimentally verified;(4)In view of the complex calculation problems in the modeling process,considering the specific needs of on-site fault maintenance personnel,a visualized MVB fault diagnosis system software is designed and developed based on the MATLAB GUI platform,which provides a convenient fault diagnosis tool for on-site fault maintenance personnel.Finally,the research contents of the full text are summarized,and the future research work and direction are prospected.