A Study of Fault Diagnosis Method for the Train Axle Box Based on EMD and PSO-LSSVM

The vibration signals of train axle box bearings are nonlinear and non-stationary, and the previous time-frequency analysis methods couldn't accurately extract the instantaneous characteristics of the signal that include short-time Fourier transform, Wigner-Ville distribution, wavelet transform. In this paper, empirical mode decomposition (EMD), a relatively new time-frequency analysis method, was used to decompose the signals into a series of single-frequency intrinsic mode functions (IMFs) in order to extract feature parameters. In fault pattern recognition, particle swarm optimization least square support vector machine (PSO-LSSVM) algorithm was applied for identifying the fault type. Results show that the proposed method has a prediction accuracy of 92.5% in identifying four kinds of working conditions which include normal bearing, outer ring fault, inner ring fault and rolling element fault, higher than SVM, PSO-SVM and LS-SVM.