EEMD based incipient fault diagnosis for sensors faults in high-speed train traction systems

In this paper, the incipient fault diagnosis problem is studied for traction motor sensor fault. The data used for the fault diagnosis is from sensors of the closed-loop traction motor system, in which the deviations between normal and bias faulty data are of 1 %~5 % and between normal and gain faulty data are of 1%~10%. Considering the non-stationary of the data, the Ensemble Empirical Mode Decomposition method is used to transform the original data to intrinsic mode functions (IMFs). Then, the fault feature vector is extracted from IMFs. To deal with the fault diagnosis problem, the kernel support vector machine is utilized to be trained by the feature vector, and achieve the classification of the feature vector. The simulation results show that incipient faults can be diagnosed completely by the Kernel SVM based on EEMD method.