Histogram of Oriented Gradients for Rotor Speed Estimation in Three-Phase Induction Motors

This article presents a novel approach for sensorless speed measurement of induction motors (IMs) using a time-domain stator current analysis. More particularly, we propose the use of histogram of oriented gradients (HOGs), usually applied to the computer vision field and image processing applications, for estimating the rotor speed in squirrel cage IMs. The features extracted from HOG were used as inputs for an artificial neural network predictor (ANNP) to estimate the final rotation of the electrical machine for some load and operational scenarios. This research addressed the evaluation of the best HOG and signal processing parameters, such as the HOG bin angle, filtering cutting-frequency, and sampled time window to optimize the ANNP performance. Some experimental tests from a 7.5 kW squirrel cage IM were carried out to validate the approach to motor fed by a sinusoidal power supply and by an inverter. The method was also validated for the IM subjected to broken rotor bars and running at very low slip conditions. The results showed a good approximation between the speed measurements using a tachometer and the estimated ones.

Automated summary

This article proposes a novel approach for sensorless speed measurement of induction motors using time-domain stator current analysis. The approach utilizes histogram of oriented gradients (HOG) and artificial neural network predictor (ANNP) to estimate the rotor speed in squirrel cage IMs. Experimental tests on a 7.5 kW motor validated the effectiveness of the method, even under broken rotor bars and low slip conditions.