Robust Motor Current Signature Analysis (MCSA)-based Fault Detection under Varying Operating Conditions

Abstract

Motor current signature analysis (MCSA) has been widely used in motor fault detection including bearing fault, broken-bar, and eccentricity, etc. When a motor’s fault is in its early stage or a faulty motor is operating in varying load conditions, fault signature may be submerged in the background noise and interference, making fault detection a very challenging problem. In this paper, we address the problem of extracting small fault signature of frequency components under a varying load condition and a noisy background. To this end, we segment the time-domain stator current into overlapped sequences, and treat each sequence as an independent measurement of an imaginary sensor. A minimum variance beam-forming method is then employed to generate the current frequency spectrum with robust performance under varying-load operations. Our method is validated with experimental data collected on a motor with a minor eccentricity fault operating in varying conditions.