Wavelet Decomposition and Energy Distribution as Ball-Bearing Diagnostics Tools in Direct-Drive Motors

In the last decade different methodologies have been proposed for the detection of incipient bearing failures. Vibration measurements in both time and frequency domains have been used for the detection of localized defects. In this research, particular attention is given to dif-ferent wear evolution. In case of generalized roughness (due to contaminations, lack or loss of lubrication, corrosion, humidity, etc.), classical spectral analysis of the vibration signal does not exhibit characteristic frequencies but only unpredictable broadband changes. Furthermore, when direct-drive motors are employed – e.g. in the packaging machines – to provide complex laws of motion, interactions between different parts of the bearings are not periodic, so the detection of failures cannot be performed in the frequency domain. A wavelet based decomposition of the stator current is here proposed for the detection of generalized roughness of bearings in direct-drive motors. Stator current energy, computed in each node of the decomposition, is used as fault indicator and the frequency band, which is most sensitive to the degradation process, is identified. Experimental results are presented for different mo-tors with different working hours, operating in an industrial environment. In particular analy-sis of energy distribution among bandwidth in the wavelet decomposition is done and a com-parison between the energy level in healthy and faulty cases completes the paper. The results are normalized with respect to a healthy motor.