A New Swarm Decomposition Technique for Bearing Fault Detection and Diagnosis

Although swarm decomposition (SD) can provide flexibility in extracting oscillatory components, it has limitations in parameter sensitivity and the absence of a systematic mode selection, which may obscure weak fault features and lead to incomplete decomposition and high computational costs. To address these limitations, a new recursive swarm decomposition (RSD) technique is presented in this work for more efficient bearing fault detection and fault type diagnosis. The proposed RSD technique employs a two-layer decomposition framework with an impulse-modulation energy (IE) score mode evaluation for component selection. The selected sub-IMFs are then applied in harmonic suppression to attenuate rotational interference and enhance impulsive features. Its effectiveness is verified by simulation and experimental tests. Test results show that the proposed RSD technique can enhance representative features for bearing fault detection.