Predictive Control of Traveling Wave Ultrasonic Motors using neural network

Traveling Wave Ultrasonic Motors (TWUSMs) possess extreme nonlinear properties such as saturation reverse effect and dead-zone, which are reliant on the driving conditions. These characteristics make modeling and control of TWUSMs highly challenging. Thus, deriving a simple and precise mathematical model suitable for controlling USMs has been a major problem for researchers. In this paper, a multi-layer perception neural network (MLPNN) based on the Hammerstein structure of TWUSMs is utilized to annul the nonlinear subsystem of TWUSM. Subsequently, a Generalized Predictive Controller (GPC), along with the inverse model characterized by MLPNN, is utilized to control the angular position of a TWUSM. The inverse model is able to cover all the variations in initial conditions, load torque, and the driving frequency. Simulation results based on the proposed scheme are presented which validate the scheme's performance.