Bearing Fault Diagnosis Method Based on Graph Fourier Transform and C4.5 Decision Tree

Abstract

Bearing is a key component of rail vehicles. Its operational status greatly affects the safety of passengers and cargo on the train. Therefore, it is especially important to find a fault diagnosis method suitable for train bearings. In order to adapt to the railway application background, the graph Fourier transform (GFT) is introduced into its fault diagnosis. As the foundation of graph signal processing (GSP), GFT is the expansion of graph signal in terms of the eigenfunctions of graph Laplacian matrix. The vibration signal is converted into a path graph signal. Using GFT to extract the graph spectrum domain feature as fault feature set combines with the C4.5 classification algorithm to identify the fault of the rolling bearing. Taking into account the practicality of the application, train axle rolling bearings vibration signal with real faults have been collected from on-site trains to verify its validity. By comparing with the time domain feature and the frequency domain feature classification result, it reflects the practicability of the method under the complex operation conditions of the train.