A Novel Method for Periodical Impulses Detection and Its Applications in Rubbing Fault Diagnosis

Periodical impulses detection is the key issue for vibration-based condition monitoring and fault diagnosis of rotating machinery. Inspired from spectral distribution of faulty vibration signal, a novel harmonic-decomposition based periodical impulses detection method is proposed in this paper and applied in rub-impact fault diagnosis for rotor system. Firstly, the resonance band where the impulses implied in sensed vibration signal locate, is identified by the fast kurtogram algorithm. It is demonstrated that the fault-stimulated periodical impulses can be expanded as a series of harmonics equipped with a uniform frequency interval in spectral distribution. An adaptive harmonic-decomposition method is thus developed to decompose the harmonics within the identified band with periodical impulses reconstructed. The rubbing fault diagnosis for rotor system of rotating machinery is finally achieved according to the envelope analysis for reconstructed periodical impulses. Compared with detecting results obtained by band-pass filtering based approach, the proposed method is able to detect the periodical impulses with higher reconstructed signal-to-noise (SNR) and more distinct fault features extracted, which is very helpful to reach a reliable diagnosis conclusion.