Motor-current-based electromagnetic interference de-noising method for rolling element bearing diagnosis using acoustic emission sensors

The high sensitivity of AE sensors enables engineers to detect tiny fault signals of a bearing in advance of failure. However, this process is also easily corrupted by noise, due to the sensitivity of the sensors. Among possible noise sources, electromagnetic interference (EMI) generated by variable frequency drives (VFD) is one of the most difficult noises to address because of its highly nonstationary characteristics. This disturbs the envelope spectrum, which is the conventional method of bearing diagnosis. Thus, in this paper, a method is proposed to adaptively remove EMI from the AE signal for more accurate bearing diagnosis. The proposed method eliminates EMI peaks in the enveloped frequency spectrum, using a motor current signal. To this end, the proposed method employs empirical mode decomposition (EMD) and probabilistic filtering techniques. The proposed method is verified by examining bearing testbed data and effectively eliminates the unwanted peaks of the EMI for AE data.