MORPHOLOGICAL UNDECIMATED WAVELET AND ITS APPLICATION TO FEATURE EXTRACTION OF IMPULSIVE SIGNAL

Abstract

To avoid the decrease in signal length in morphological wavelet decomposition, a novel morphological undecimated wavelet (MUDW) based on multi-scale Top-Hat transformation is proposed. Morphological opening operator and multi-scale Top-Hat transformation are adopted as a signal analysis operator and a detail analysis operator under a MUDW framework to ensure an invariant signal length and sufficient signal information in morphological wavelet decomposition. The proposed MUDW is also examined by vibration signals of a rotor with impulsive feature. The results confirm that the proposed MUDW is feasible and effective in fault diagnosis.