Improvement of the EMD method and applications in defect diagnosis of ball bearings

Based on the analysis of the empirical mode decomposition (EMD) method in this paper, an improved arithmetic for local mean is presented in detail. In the new arithmetic, the local mean is obtained by connecting all window means of the signals between two successive extrema with the cubic spline interpolation. The new arithmetic is more time saving than that used in the conventional empirical mode decomposition method. The improvement of the new arithmetic is validated by decomposing simulating signals. Moreover, the improved empirical mode decomposition method is successfully applied in decomposing vibration signals of ball bearings and provides a new method to diagnose defects of ball bearings. In contrast to the traditional envelope analysis, the empirical mode decomposition method could decompose the signal into different frequency bands adaptively, and the central frequency and bandwidth of band-pass filter, which was used to extract the resonance vibration, should no longer be decided. The results show that the EMD method is powerful for defect diagnosis of ball bearings.