Improved deep PCA and Kullback–Leibler divergence based incipient fault detection and isolation of high-speed railway traction devices

Abstract

Enhancing the reliability of high-speed railway traction system is critical important to the safety of entire trains. Data-driven based FDD (Fault Detection and Diagnosis) schemes focused on electric locomotive traction system have received more and more attention. An improved Deep PCA (Principal Components Analysis) algorithm is presented in this paper, based on which, a KLD (Kullback–Leibler Divergence) based incipient FDD scheme is proposed at the same time. The main contributions are summarized as: (i) The improved Deep PCA algorithm by using the covariance matrix of the dataset (not the original dataset as references) can highlight more useful incipient fault information with much faster data decomposition; (ii) The diagnosis scheme based on KLD can improve the accuracy of non-Gaussian process; (iii) The study proposed in this paper can realize the update of fault database and the diagnosis of unknown type of incipient faults. The experiment results performed on TDCS-FIB (Traction Drive Control System-Fault Injection Benchmark) platform demonstrate the superiority of the proposed data-driven based incipient FDD scheme in comparison with the original Deep PCA algorithm.