Fault Detection In Traction System Of High-speed Trains Based On Canonical Correlation Analysis

At present,the high-speed train has become the most important means of transportation between cities.The traction system is the core system of the high-speed train.It is a multifunctional,highly integrated,and complex system integrating mechanical,electrical and communication subsystems.It not only provides sufficient power for trains,but also guarantees the reliability of train operation.However,with the increase of high-speed train mileage,the wear and aging of traction system components are becoming more and more prominent,which is very easy to cause various faults.If the faults are not handled in time,they will gradually aggravate and bring serious hidden danger to train operation.Therefore,the research on fault detection of the high-speed train traction system is of great significance.In this paper,the high-speed train traction system is the research object,based on the datadriven method to establish the system model,and an effective traction system fault detection strategy is designed.The main research contents of this topic are as follows:(1)In this paper,the operation mechanism and basic structure of the high-speed train traction system are introduced.At the same time,the working mechanism of signal acquisition in the traction system is expounded.On this basis,the typical faults of the traction system are studied,the causes and manifestations of different faults are discussed,especially the sensor faults are described,and the models of different sensor faults are analyzed.(2)Aiming at the non-Gaussian signal problem of high-speed train traction system,a fault detection method of high-speed train traction system based on K-L divergence(KLD)and canonical correlation analysis(CCA)is proposed.Not only the single variable and multivariable K-L divergence evaluation functions are given through strict mathematical theoretical analysis,but also two nonlinear coordinate transformations are introduced to realize the conversion from non-Gaussian signal to Gaussian signal.The signal processing strategy effectively avoids the estimation of the non-Gaussian probability density function,and improves the processing accuracy and computational efficiency of the algorithm.At the same time,the feasibility of practical application in traction systems is greatly improved.(3)Aiming at the problem that it is difficult to detect the micro fault of high-speed train traction system,the incipient fault detection method of high-speed train traction system based on Hellinger distance(HD)and canonical correlation analysis is proposed.Considering the traction system under actual working conditions,the incipient faults of the traction system are analyzed qualitatively,and the noise information in the fusion signal is accurately modeled.Then,Hellinger distance is introduced from various conditions and statistics are constructed to realize the incipient fault detection of traction system.Finally,the scheme is verified in the data of high-speed train traction system.The algorithm has high sensitivity to incipient faults and can meet the needs of real-time fault detection of traction system with high-frequency sampling.