Application of improved reweighted singular value decomposition for gearbox fault diagnosis based on built-in encoder information

Due to the harsh operating condition and persistent heavy load, planetary gearboxes as the key transmission parts are prone to damage. With lower cost and better accessibility, the built-in encoder signal has been considered an alternative tool for health state monitoring of gearboxes in recent researches. However, how to extract the feature signature without phase shift or waveform distortion has become a major challenge for current methods. Motivated by this, an improved reweighted singular value decomposition (IRSVD) method is designed to deal with the aforementioned problem in this paper. Firstly, through mathematical derivation combined with numerical analysis, the characteristics of the encoder signal from faulty gear are studied. To accurately evaluate this kind of fault feature, a new index, namely the normalized proportion of harmonics (NPH), is introduced. Without any prior information, it can be used for the optimal selection of the signal component (SC) when applying the proposed IRSVD method. Benefiting from the merit of NPH, IRSVD can choose the optimal SC as the denoised signal, which greatly simplifies the application of traditional singular value decomposition (SVD). Finally, the effectiveness of IRSVD is verified by case studies with both simulated data and real experimental data from a faulty planetary gearbox, containing the challenging corrosion fault and two-worn-teeth fault.