Blind source separation of single-channel train signal based on EEMD and ICA
GU Qian-we · Jisuanji yingyong yanjiu · 2014
Blind source separation is an effective method for multiple fault diagnosis.This paper proposed a new blind source separation algorithm based on ensemble empirical mode decomposition(EEMD) for fault diagnosis of train signal.Nonlinear mixed signal filtered wave was decomposed into intrinsic mode function(IMF) containing different source signal characteristics,they became the new multidimensional signals.The application of principal component analysis(PCA) could accurately estimate the number of source signals to solve the underdetermined problem of single-channel blind signal separation.At last fast independent component analysis algorithm(FastICA) realized the blind separation of signals.The experimental signal used the simulation signal and train mixed fault signal.Experimental results show that this algorithm can effectively analyze the characteristics of train single fault and has important practical value.