RBF neural networks based quasi sliding mode controller and robust speed estimation for PM Synchronous Motors

This paper presents a neural networks based discrete time variable structure control and a robust speed estimator designed for a Permanent Magnet Synchronous Motor (PMSM). Radial basis function neural networks are used to learn about uncertainties affecting the system. A cascade control scheme is proposed which provides accurate speed tracking performance. In this control scheme the speed estimator is a robust digital differentiator that provides the first derivative of the encoder position measurement. The analysis of the control stability is given and the ultimate boundedness of the speed tracking error is proved. The controller performance has been evaluated by simulation using the model of a commercial PMSM drive. Simulations show that the proposed solution produces good speed trajectory tracking performance.