Abstract:The Independent component analysis (ICA) algorithm was proposed based on Ensemble empirical mode decomposition (EEMD), which aims to solve the bearingscrew composite fault signal separation in single channel. First of all, through the EEMD method, the composite signal was decomposed in different channels to get a series of component of the IMF. Then the kurtosis value of the fund and phase relationship value was calculated to select some numerical larger IMF component. A set of observations was newly formed with the original signals, as input of the ICA to get a series of IC component. Lastly, selecting for impact ingredients of larger IC components, envelope analysis was carried out to diagnose the fault type identification. Through the experiment, separate and identify two types of fault successfully to prove the effectiveness of the proposed method.