An ICA-EMD Feature Extraction Method and Its Application to Vibration Signals of Hydroelectric Generating Units

In order to understand the operating status and the fault development trend of hydroelectric generating units,extracting the failure feature or incipient symptom from vibration signals is foundational.However,vibration signals of hydroelectric generating units presents non-linear,aliasing and non-stable character,its feature extraction is a challenging problem in this field.Thus,a new feature extraction method(ICA-EMD) of vibration signals for hydroelectric generator units is proposed by combining independent component analysis(ICA) with empirical mode decomposition(EMD).Firstly,the signals form multi vibration sensors were separated into statistics independent component by ICA.Then,each of statistics independent components was analyzed by using the method of autocorrelation analysis to eliminate the interference of the background noise.Finally,all of the statistics independent components were decomposed by using EMD,and the fault feature signals were obtained by accumulating and reconstructing all the co-channel components.The results of simulations experiment and an application example show that this method can extract efficiently incipient symptom,weak signals,transient signals and other fault feature signals.Compared with other methods,this method is an effective way to practical projects.