A local transient feature extraction method via periodic low rank dynamic mode decomposition for bearing incipient fault diagnosis

摘要

The characteristic components of faulty rolling bearing are submerged by noise and interference components. The challenge of bearing incipient fault diagnosis is to accurately estimate transient impulses from noisy signals. In this paper, a periodic low rank dynamic mode decomposition (PLRDMD) algorithm is proposed to accurately extract transient features while preserving amplitude. First, a periodic window selection strategy is proposed to preserve the amplitude of transient impulses. By setting the sampling window as one period, localized de-noising can eliminate the reduction of transient impulses especially. Then, a local low rank approximation framework is designed to extract the incipient transient impulses of faulty rolling bearing. Therefore, the proposed PLRDMD algorithm can accurately extract the transient features from incipient fault signal of faulty rolling bearing. Lastly, numerical simulations and experimental researches illustrate the validity and superiority of the proposed PLRDMD algorithm over the standard DMD.

关键词