Step robust mode decomposition method and its application in rolling bearing fault diagnosis

Rolling bearings subjected to external impacts or uneven loads often exhibit non-periodic transient step disturbances, which distort fault features and degrade diagnostic accuracy. This paper proposes a step robust mode decomposition (SRMD) method to address these challenges. SRMD employs a physically constrained collaborative optimization mechanism, using derivative penalty terms and sparse modeling to force step components into piecewise constant forms, effectively separating disturbances while preserving fault feature integrity. Additionally, an adaptive iterative mechanism eliminates the need for a preset mode count, enabling the autonomous determination of modal components and enhancing stability. To ensure precise identification, a fused evaluation index is defined for the efficient extraction and enhancement of critical fault modes. Experimental results demonstrate that SRMD accurately separates step disturbances and enhances fault features across various noise environments, exhibiting superior robustness.