Induction Motor Current Signature Analysis for Speed Detection Using Spectral Estimation Techniques

Abstract

In this research work, real-time performance of various spectral estimation techniques have been investigated to detect speed of an induction motor from rotor slot harmonics using motor current signature analysis. It has been demonstrated that novel methods Matrix-Pencil and Principle Component Auto-Regressive (PCAR) can be applied in realtime for speed detection. Steady-state and dynamic performance of Matrix-Pencil, PCAR, FFT, and Covariance methods were compared. The experimental results obtained indicate that, using PCAR and Matrix-Pencil methods, it is possible to achieve 0.5 Hz resolution, accuracy ±1 to ±2 rpm and settling-time of about 0.25 to 0.3 seconds for fewer data sets. The results showed that PCAR or Matrix-Pencil method leads to speed detection inaccuracies, predominant at light loads, after certain small data samples number. The trade-offs between samples number, accuracy and order number, for both PCAR and Matrix-Pencil methods, are more predominant for smaller data-sets at light loads.