Abstract:

Empirical mode decomposition (EMD) is a useful method in nonlinear and non-stationary signal analysis. We propose an improvement to the EMD method to reduce the end effects based on linear prediction. Data are extended by linear prediction until a local extreme appears at each end. The same procedure is applied to each intrinsic mode function. Experiments show that the proposed method works well in restraining the end effects of EMD compared to other methods. Because striation distance in fatigue fracture image is quasi-periodical, striation distance in the horizontal and vertical directions are calculated based on improved EMD algorithm, and the real striation distances obtained through triangle conversion. The striation distance measurement of actual images shows that application of EMD to striation distances measurement is feasible.

Key words: empirical mode decomposition (EMD), linear prediction, fatigue fracture image, striation distance