Defect length estimation of rolling element bearings via adaptive separation of dual-impulse

Dual-impulse responses are characteristic phenomenon of rolling element bearings with a defect, which can be utilized to quantify the defect length of bearings. Previous researches have demonstrated that separated extraction is more accurate than simultaneous extraction for the dual-impulse responses. However, accurately extracting dual-impulse responses from vibration signals is challenging due to interference, and separated extraction approaches for these responses in bearing vibration signals are seldom reported. To address this issue, a defect length estimation method based on vibration signals of bearings is proposed. Firstly, the proposed weighted EEMD (ensemble empirical mode decomposition) enhances the stronger impulse response within the dual-impulse. Then, both the raw signal and the signal enhanced by the weighted EEMD are decomposed using Fast ICA (Independent Component Analysis) to yield two separate signals with an impulse response. Finally, the defect length is quantitatively estimated based on the two separate signals. Meanwhile, the proposed approach is validated through simulation and experiment. The results indicate that the proposed approach can estimate the defect length, and the stronger impulse response within the dual-impulse can be enhanced by weighted EEMD.