Novel adaptive artificial neural networks based speed regulators for PMDC motor drives

This paper presents a new adaptive controller for PMDC motor drive using a flexible artificial neural network (ANN) structure. When the ANN regulator is used to control the motor drive directly, the only output available is the speed/position error of the controlled motor. The actual output error of the ANN network is usually unknown. This paper presents a simple and robust training algorithm, which enable the ANN based speed controller to be trained online using the speed and current motor outputs errors. The ANN based speed regulator employs an effective online tuning mechanism based on the actual drive output error; it only requires the knowledge of the speed reference trajectory, past speed error and current deviation variables