Oscillation search robust dynamic mode decomposition method and its application in rolling bearing fault diagnosis

Mechanical equipment has complex structures and operating environments, where bearing fault signals are frequently affected by harmonic interference from internal components and strong external noise. Therefore, it is a struggle to extract periodic impulse transients of fault signals from strong noise and interference. To address this challenge, this paper proposes an oscillation search robust dynamic mode decomposition (OSRDMD) method. Firstly, to enhance the reconstruction accuracy of the dynamic matrix A, a novel scheme is proposed to select valuable left singular subspaces to reduce large energy interference and noise. Then, to further reduce the interference from non-periodic components, the dominant periodic oscillatory modes of DMD are selected to reconstruct the fault signals. Simulation results demonstrate that OSRDMD consistently and reliably extracts fault frequencies even in low signal-to-noise ratio (SNR) environments ranging from -10 dB to -15 dB. Furthermore, in real dataset analysis, the proposed method exhibits superior fault diagnosis accuracy compared to existing decomposition techniques.