Extracting Fault Features of High-Speed Train Bogies Using Copula Function

To monitor the working condition of key components of high-speed train bogies in real time, an approach using Copula function to extract features is proposed. Vibration signals of a certain high-speed train are obtained under four typical working conditions, including normal condition, yaw damper fault, air spring fault, and lateral damper fault. The vibration signal is decomposed by ensemble empirical mode decomposition first. Then, the joint probability distribution of intrinsic mode functions is computed by Gaussian Copula function. The Kullback-Leibler distance of two marginal distribution functions, and the mean and variance of the joint probability density function are extracted as the features. The support vector machine is used to classify the working conditions. The experimental result shows that the average recognition rate is above 95% at the speed of 200 km/h, which verifies the effectiveness of the proposed feature extraction method.