Low Complexity Motor Current Signature Analysis Using Sub-Nyquist Strategy With Reduced Data Length

Abstract

Motor current signature analysis requires high sampling rate and a large number of data samples to detect faults at light loading conditions. However, it would be interesting to investigate the alternative methods to bring down these requirements using the principle of sparse signal processing. In this paper, a low-complexity fault detection algorithm based on the sub-Nyquist sampling of the analytic current signal has been proposed and compared to other similar algorithms. The acquired current data is first converted into an analytic signal using a low-cost time-shift-based method. The processing of the analytic current signal has been performed at a rate 20 times lower than the Nyquist criterion using multicoset sampling technique. The computational complexity has been reduced further by partially reconstructing the spectrum only for the regions of interest. The regions of interest were found out from the estimated supply frequency and the armature current magnitude. The supply frequency has been estimated from the transformed analytic current signal. The algorithm has been tested for broken rotor bar (BRB), eccentricity, and broken bearing faults, using the recorded data from a 22-kW induction motor drive at various loading conditions in the laboratory. In addition, the BRB fault has also been tested for motors with 1.5- and 3.7-kW ratings. A receiver operating characteristic curve has been generated to test the performance of the algorithm. The proposed algorithm uses only a single current sensor to acquire one of the armature phase currents.