Off-the-grid compressive sensing for broken-rotor-bar fault detection in squirrel-cage induction motors

In this paper, we propose an off-the-grid compressive sensing based method to detect broken-bar fault in squirrel-cage induction motors. To validate our method, we first build a dynamic model of squirrel-cage induction motor using multi-loop equivalent circuit to simulate motor current under fault conditions. We then develop an off-the-grid compressive sensing algorithm to extract the fault characteristic frequency from the simulated motor current by solving an atomic norm minimization problem. Comparing to other fault detection methods via motor current signature analysis, our method yields high resolution in extracting low-magnitude fault characteristic frequency with only 0.7 second measurements. Simulation results validate our proposed method.