Half-broken rotor bar detection on IM by using sparse representation under different load conditions

Currently, the Induction Motor is widely used in industry, due to its easy installation and operation. Induction motors require a more reliable monitoring due to constant operation increases the possibility of faults, for example, a broken rotor bar fault. Early stage, broken bar is not easy to detect, and its evolves is slow and quiet. In the most of cases, it is detected when the fault is critical and other faults have appeared. Many techniques have been proposed in the literature, but majority of these performs analysis in frequency domain, applying additional transformation or preprocessing methods. In this paper, a novel methodology to detect a half-broken bar fault is proposed, making use of the vibration signal from induction motor under two fault conditions: healthy and half-broken bar; and three load conditions: unloaded, half-loaded and three-fourths loaded. The detection is possible due to the sparse representation of the raw signal which is obtained and then evaluated by minimal decomposition error criterion. In this way, preprocessing methods are not needed, and the fault is detected early and directly. These tests were developed in Matlab software, with vibration signals from induction motors in steady state.