Experimental study of induction motor misalignment and its online detection through data fusion

Most of the induction motor (IM) fault detection schemes are based on one sensor with one detection logic which are generally incapable of bringing out any consistent feature related to rotor misalignment. Moreover, these logics do not consider simultaneously the asymmetric load condition with variable speed operation. In this study, a data fusion-based misalignment related fault identification algorithm is presented, which isolates fault features from similar features generated because of other operating conditions. In the proposed scheme, the feature vector is constructed by using signatures created from frequency-domain characteristics obtained from stator vibration and line current measurements. Thereafter, the feature fusion technology, by means of the weighted linear combination concept, is adopted to take advantage of the best features from both sensors and to discern the pattern of misalignment with other signatures. The technique is validated experimentally on a 5.5 hp IM and the results are presented.