Dynamic Data Analysis of High-Speed Train Based on MEMD and Compressive Sensing

Aiming at the characteristics of multi-channel vibration signal of high-speed train, this paper proposes a data analysis method based on Multivariate Empirical Mode Decomposition (MEMD) and compressive sensing. The method decomposes vibration signal using MEMD to obtain a series of Multivariate Intrinsic Mode Functions (MIMF), and extracts information entropy of these MIMF components, constituting the original high-dimensional feature set. Then compressive sensing algorithm is adopted to compress the high-dimensional feature to eliminate redundant and undesirable feature, obtaining the low-dimensional optimal feature. The experimental results of classification and identification of conditions via Support Vector Machine (SVM) show the good classification performance is 100 % of proposed method in each channel, verifying the effectiveness of the proposed method.