An Online Artificial Neural Network Speed Estimator for Sensorless Speed Control of Separately Excited DC Motor

This paper presents a speed estimator for sensorless speed control of separately excited DC motor by using artificial neural network. To estimate the speed, a coefficient, which concerns with the speed in the armature current estimation equation, is adjusted accordingly to an artificial neural network learning principle until the difference between the actual and the estimated armature current is minimal. Widrow-Hoff learning rule is adopted to perform this task. The consequence of using this algorithm leads to the ability of online speed estimation and simple artificial neural network structure. The simulation and experimental results have verified the effectiveness of the introduced method.