Clustered ensemble empirical mode decomposition (EMD) and its application to windturbine noise and co-seismic landslide-induced ground motion

Clustered ensemble empirical mode decomposition (EEMD) is proposed to resolve the multimode problem of EEMD. EEMD is a noise-assisted data analysis method which is used to decompose signals into a collection of intrinsic mode functions (IMFs). The multi-mode problem of the method is referred to that signals with a similar time scale are decomposed into different IMF components and form an over-complete set of IMFs and the occurrence of this problem may lead to mis-interpretation of the signals. The solution to this problem is to recombine the multi-mode IMF components into a proper single IMF. Instead of the previous heuristic manual approach, we incorporate a statistical clustering analysis to assist the diagnosis of multi-mode IMFs and to guide the recombination of the multi-modes based on the classified clusters. As a result, signals are reorganized into a condensed set of clustered intrinsic mode functions (CIMFs). The method is first examined using an artificially synthesized signal and is shown that the multi-mode problem can be largely eliminated in a statistically reliable manner. Then the method is applied to two sets of practical signals: wind turbine noise and co-seismic landslide-induced ground motion. For the former, three CIMFs are found to be associated with noise generation mechanisms. For the latter, the first non-Gaussian CIMF is closely related to the landslide fracture induced ground motion.