Gearbox Fault Diagnosis Using Multiscale Sparse Spectrum

Due to its special modulation mechanism with multiple units (eg. shafts, gears, etc.) under various conditions, the related fault information of gear fault would distribute in a broad frequency band. In this manner, it is not easy about accurately detecting the early-stage gear fault by detecting the fault frequency in a limited frequency band. In this paper, a new spectral analysis, called multiscale sparse spectrum (MSS), is proposed to achieve fault frequency detection in a sound way. The overall frequency information about the raw signal is firstly sensed by a series of frequency-window function, which can be reached by short-frequency Fourier transform. Then, according to orthogonal matching pursuit, harmonic atoms are further employed to sparsely mine the modulation components from these multiscale pseudo mono-components. Finally, MSS is proposed to synthesize the existing harmonic-related components. Furthermore, a synthesized sparse spectrum (SSS) is obtained by searching the frequency-frequency ridge from MSS. Compared with EMD and fast-kurtogram analysis, the results show the effectiveness of the proposed method of gear fault detection.