Period oriented weighting sparse denoising for compound faults of rolling bearing

Compound faults (CFs) present substantial challenges in the fault diagnosis of rolling bearings. This study introduces a novel methodology, termed Period Oriented Weighting Sparse Denoising (POWSD), designed specifically to diagnose and differentiate CFs in rolling bearings. Initially, the type of faults is determined based on the outcomes of sparse denoising. Subsequently, POWSD employs the fault periods identified for each type, calculating a new metric, FPER, which leverages the multiscale features and periodic relationships within sparse coefficients. These relationships are weighted differently at various levels, aiding in the targeted denoising of sparse coefficients during the POWSD process. Through the implementation of POWSD, repetitive impulses associated with CFs are effectively extracted and isolated. Both simulation experiments and practical tests conducted on a specialized bench confirm the efficacy of POWSD, showcasing its superior performance in denoising and separating CFs in comparison to existing methods.