Robustness, simplification, and time consumption reduction of online-trained ANN-controller fed IM-servo drives

Abstract

A proposed ANN-based tracking controller for induction motor (IM) servo-drive is introduced in this paper. In this technique, the IM is used as a high-performance drive (HPD) in servomechanism applications as robot systems to track certain position/speed trajectories in all time. The robustness of the controller was tested against parameter variations (especially, inertia of the mechanical system) and sudden repetitive changes in reference speed/position trajectory. Good results were obtained that insure the robustness of the controller versus traditional PI-controllers. Also, this paper studies parameters affecting the ANN online training. The authors proposed a simple ANN architecture taking advantage of the main features of the ANN as linearity and adaptive learning rate. The proposed controller contributes towards time-consumption reduction of the ANN online training. The proposed controller was implemented using Texas Instruments TMS320C31 floating point DSP.