Maximum Synchronous Average Margin Deconvolution for Bearing Incipient Fault Diagnosis

Effectively capturing impulsive features holds great significance for incipient fault diagnosis of rolling element bearings. Whereas, transient fault impulses often get obscured by irrelevant components and background noise, restricting the efficacy of existing blind deconvolution methodologies for bearing incipient fault diagnosis. To tackle this issue, this study tailors a new metric termed the synchronous average margin (SAM). The blind deconvolution process is formulated as an optimization problem, where maximizing the SAM serves as the objective function, referred to as maximum synchronous average margin deconvolution (MSAMD). To eliminate dependence on prior period information, the envelope harmonic product spectrum is adopted for the period estimation in each iteration. A two-step denoising operation, combining sparsification with time-domain synchronous averaging, is then presented to suppress noise and irrelevant harmonic interference. The synchronous average signal is subsequently computed to determine the SAM indicator. MSAMD iteratively updates the inverse filter towards maximizing the SAM, ultimately yielding a filtered signal with the highest SAM. A simulation case and two experimental cases are schemed to validate the effectiveness and competitiveness of the proposed method. Comparative results with several mainstream blind deconvolution methods demonstrate that MSAMD exhibits superior adaptability and robustness in suppressing strong noise and harmonic interference, making it a promising tool for early-stage fault diagnosis of rolling element bearings.