Design of artificial neural network (ANN) based rotor speed estimator for DC drive

Abstract

This paper describes the design of ANN based rotor speed estimator for separately excited DC motor using MATLAB Toolbox. A comparative analysis of the DC motor drive's behavior with and without ANN based was performed. It is shown that rotor speed feedback by suitably trained ANN enables very good quality of the drive performance over a wide range operating conditions for both open and close loop systems. For the purpose of the training, the Levenberg-Marquardt back-propagation algorithm was used. A standard three layer feed-forward neural network with tan-sigmoid (tansig) activation functions in hidden layer and purelin at the output layer was applied in this design. The result shows that by using only one hidden layer, minimum error can be obtained and the performance of the estimator is excellent. The proposed solution seems to be attractive to the conventional speed estimator, resulting a mechanically simpler motor and consequently increasing the degree of reliability for the whole drive systems.