Abstract: The nonlinear and non-stationary data set can be decomposed into a finite and often small number of ‘intrinsic mode functions’ by empirical mode decomposition(EMD). The intrinsic mode functions usually are the fault signal in fault diagnosis problem. The most serious problem of EMD method is the end effects due to the spline fitting at the data ends, I.e. the envelope curve fitted may have wide swings at the data ends if the ends are not the extremum. The decomposition quality would be polluted further alone with the decomposition. An improved empirical mode decomposition based on the time series analysis is derived. The envelope curve can be well fitted with the predicted extremum at the data ends from the time series model. It’s useful to eliminate the end effects. Compared with the wavelet analysis, EMD method is adaptive. An example from the gearbox signal is given to demonstrate the power of the proposed method.

Key words: Empirical mode decomposition, Forecast, Time series analysis