IMPROVED EMPIRICAL MODE DECOMPOSITION AND APPLICATION IN FAULT DIAGNOSIS

摘要

The mechanical vibration signal can be decomposed by the empirical mode decomposition (EMD), that any complicated one-dimensional signals can be decomposed to a finite and often small number of 'intrinsic mode functions' (IMF). It is very useful for fault diagnosis or others. However, EMD and its Hilbert spectrum are distorted by the end effect, i.e., both beginning and ending point of data series are not local maximum or minimum, the decomposition quality would be polluted further alone with the decomposition. An improved empirical mode decomposition method is derived in the paper. The time series model is used to forecast both ends of data till it is the local maxima or minima. Meanwhile, a decomposition precision criterion of EMD is defined by the residual square sum. The simulation shows that the improved EMD is better than the general one by the criterion. At last, as an example, the improved EMD is applied to fault diagnosis, and it is very well to identify the gear fault.