Adaptive Neural Controller for Speed Control of PMSM with Torque Ripples

This paper presents a novel approach to backpropagation adaptation of neural controller for motion control application. Authors propose modification to classical error backpropagation algorithm along with general tuning methods. Multistep process of adaptive neural controller development is presented, including: simulation results, hardware-in-the-loop experiment and final experimental verification on direct drive with permanent magnet synchronous motor. Main research goal was to develop real-time training method for neural-network-based controller that will remain stable for long periods and is easy to deploy on various hardware platforms.