Application of ensemble empirical mode decomposition in failure analysis of rotating machinery

For suppressing the phenomenon of mode mixing in empirical mode decomposition (EMD) and increasing the analysis accuracy, an improved algorithm named ensemble empirical mode decomposition (EEMD) was presented. A moderate Gauss white noise generated randomly was added to the original signal, which changed the local time span of the signal and rendered the analysis scales of EMD in a trial. By sufficient trials, considering as extracting the nature of the signal from different aspects, an ensemble mean of certain intrinsic mode function (IMF) decomposed by the EMD method was output as the final result of the new algorithm. The IMF eliminated bad effects of artificial noise, and indicated clearly the intrinsic processes of the signal with full of real meanings. EEMD method was validated by both simulation experiment and real rub-impact case, and then was compared with basic EMD algorithm and high-frequency-harmonic method. The results showed that EEMD was more precise but a little time-consuming, EEMD has good prospects of application in failure analysis of rotating machinery.