Research On Fault Diagnosis Method Of High Speed Train Traction System Based On Independent Component Analysis

The traction system of a high-speed train is a complex system consisting of several components working together to provide sufficient power for the high-speed train.The stable operation of the traction system is an important basis for ensuring the safety and reliability of high-speed trains.The failure of one or more of the components will cause the train to stop,derail or even cause casualties.Therefore,it is important to study the fault diagnosis method of the traction system.In this thesis,fault diagnosis work is carried out for highspeed train traction systems based on the independent component analysis method.The main research contents of this thesis are as follows:(1)Firstly,the research status of three commonly used traction system fault diagnosis methods is introduced,and the data-driven fault diagnosis method,which is the most effective and most widely applicable,is elaborated.Secondly,the basic structure and operation mechanism of the traction system are introduced in order to better analyse the operation data of the traction system,and several typical faults of the traction system are analysed.(2)A traction system fault detection method based on Dynamic inner Independent Component Analysis(Di ICA)is proposed,which takes into account the non-Gaussian and dynamic nature of traction system data and improves fault detection.The feasibility is first verified theoretically by combining the independent component analysis with the dynamic inner algorithm through a rigorous mathematical derivation.Dynamic and static features in the data are taken into account while ensuring maximum non-Gaussianity of the extracted latent variables.Secondly,the Hotelling statistic is used to monitor the dynamic and static spaces separately during the fault detection phase.When a fault occurs,the statistic exceeds a threshold value and an alarm is issued.Finally,experiments on a high-speed train traction system simulation platform verify the feasibility and effectiveness of the method.(3)A traction system fault diagnosis method based on Di ICA and Hidden Markov Model(HMM)is proposed to apply HMM to the fault diagnosis field based on the research in(2)above.In the offline stage,the dynamic latent variables in the traction system are firstly extracted using the Di ICA algorithm and the prediction errors of the dynamic latent variables are obtained.Secondly,the Baum-Welch algorithm is used to build an HMM model of the prediction error of the dynamic latent variables,and the data of the traction system under various operating conditions are formed into an HMM fault library.In the online phase,the unknown fault data is input into the HMM fault library after extracting the prediction errors of the dynamic latent variables.A forward-backward algorithm is used to output the most likely fault type of the unknown data,thus completing the fault diagnosis.Finally,experiments on a high-speed train traction system simulation platform validate the soundness and efficiency of the method,and the fault diagnosis accuracy of this method can reach98.18%.