Abstract: The mathematical morphological analysis is a nonlinear method of digital signal processing. To extract fault feature of rotating machinery, a novel method proposed based on mathematical morphology. The morphological filter that constructed by mathematical morphological transform is able to identify the shape feature of the signa1. A combined morphology filter used in fault signal decomposition of rotating machinery, the original vibration signal is decomposed into fundamental component, fault component and noise component via different length of the morphology structure elements, the signal that containing fault information produced by rotor fault is extracted independently. The signal length and information are not lost during the decomposition procedure. A practical rubbing signal decomposed and the fault component extracted by the method, then Hilbert-Huang transform (HHT) adopted in the further comparing analysis, the results show that the fault feature is successfully extracted and non-stationary characteristics of fault described accurately, the effectiveness improved. It is also fast and simple in computational realization.

Key words: Fault diagnosis, Feature extraction, Mathematical morphology, Rotating machinery, Signal decomposition