Fault detection of gearing system based on multiwavelet packets

This paper introduces a new method to detect the fault of gearboxes. A novel multiwavelet packets system with reduced-size coefficients is introduced. The wavelet function spaces at higher scales are broadband, so that the frequency content of the narrow-band phenomena hidden in the signal that occur at relatively high frequencies, cannot be diagnosed precisely, that is unsuitable for extracting mechanical fault information. This deficiency of the scalar wavelet systems can be overcome by dividing the frequency bands of different levels. Using the multiwavelet packets transform, one attains this achievement automatically and more precisely. To lessen the cost of computation, a new multiwavelet packets coefficients reduction algorithm is proposed. As an example, non-stationary vibration signature of a gearing system is analyzed. It verify the effectiveness of the proposed method.