Application of wavelet packets algorithm to diesel engines’ vibroacoustic signature extraction

Abstract

Multisource unstable impulsive excitations, time-varying transmission path, concentrated mode, dispersion and reverberation that are important characteristics of reciprocating machines such as diesel engines result in wide-band non-stationary vibroacoustic responses which influence the effective extraction of vibroacoustic signatures and become a key factor limiting diesel engines’ vibration diagnosis. In this paper, a serial theoretical deduction of the unstable dynamic properties of diesel engines is made; the mechanism of non-stationary vibroacoustic responses is elucidated. On that basis, the wavelet packet technique is introduced into application research on diesel engines’ vibroacoustic signature extraction. The reason for the existence of frequency aliasing in the Paley series from wavelet packets’ decomposition is analysed, and the wavelet packet frequency-shifting algorithm is given. The given algorithm can reduce the noise significantly, can heighten signal-to-noise ratio and can effectively extract the desired parameter. Experiments on 190 serial diesel engines confirm its significant validity in vibroacoustic signature extraction and reciprocating machines’ vibration diagnosis.