Research On Fault Detection Method Of High-speed Train Traction System Based On Canonical Correlation Analysis And Neural Network

High-speed trains are one of the fastest and most convenient modes of transportation,and the traction system is the core power device of high-speed trains.Due to the increasingly harsh operating environment and gradually increasing operating time of highspeed trains,different types of faults may occur in the traction system,and the location and components of the faults may also be different,such as motor,inverter,control unit,and sensor positions.These faults can affect the safety of train travel or reduce the effectiveness of the traction system,resulting in speed reduction or forcing the train to stop running.In order to avoid various types of faults in high-speed trains,improve operational efficiency,and ensure passenger safety,this paper conducts an in-depth study on the traction system of high-speed trains,taking traction system faults as the research object,and using data-driven methods combined with traditional machine learning algorithms and artificial intelligence to detect and diagnose faults.The main research contents of this paper are as follows:(1)For early faults in the traction system of high-speed trains,this paper explores the problem that traditional neural networks cannot simultaneously extract data spatiotemporal features from the perspective of improving detection accuracy and shortening detection time,and proposes a fault diagnosis method based on correlation analysis and improved two-stream neural network.The algorithm mainly uses long short-term memory and convolutional neural networks to extract data features.The typical correlation analysis algorithm is used as the activation function to train the two-stream neural network deeply.A related experiment was designed for comparative analysis,and the results showed that the proposed algorithm significantly improved detection accuracy and data smoothness.(2)Regarding the problem of fault detection and diagnosis in high-speed train traction systems,this paper proposes a fault diagnosis method based on graph neural networks and two-stream neural networks by extracting dataset feature values and using prior knowledge to predict data in advance and form a correlation graph of different features within the dataset.The algorithm utilizes the correlation graph and normalization method for data preprocessing.In the model training phase,a correlation graph is constructed based on the diagnostic results,while weight coefficients are set to ensure the contribution of the training set to the network promotion.During offline training,process quantities and expert knowledge are used to determine whether the test sample has a fault.Simulation results have shown that the proposed algorithm in this paper can accurately locate sensor faults and has a high fault detection rate.Based on the traditional canonical correlation analysis algorithm,this paper proposes a solution for fault detection and diagnosis in high-speed train traction systems by combining correlation analysis methods,graph convolution methods,and neural networks.A comparative analysis with similar existing algorithms shows that the proposed algorithm in this paper has a higher fault detection rate and has certain application value in the fault diagnosis and detection of nonlinear dynamic systems.