Bearing multi-fault diagnosis based on improved template denoising source separation

Processing noise improperly or the number of sensors less than the number of faults may lead to failure of multi-fault diagnosis based on source separation. Improved template denoising source separation was proposed and used to diagnose multi-fault of bearings. Shift independent component analysis was carried out first and the independent components obtained were used as templates to the template denoising source separation algorithm. The proposed method can separate source signals through underdetermined source under reverberant environments and diagnose multi-fault of bearings. The effectiveness was verified by both simulated and experimental analysis.