Generalized Regression Neural-network-based Modeling Approach for Traveling-wave Ultrasonic Motors

Abstract As the dynamic characteristics of the traveling-wave ultrasonic motor are highly non-linear and time varying, an analysis model is difficult to obtain. It is difficult to design a suitable controller to achieve high-precision position control using conventional control techniques. A new identification approach is proposed and implemented, and it aims to provide a practical model for controller design and simulation. As a result, a generalized regression neural-network-based model is developed to identify the relation between input excited frequency and phase difference of two-phase AC voltages and the output driving torque generated by the traveling-wave ultrasonic motor. One major advantage is that transfer function identification is no longer required. The other advantage is that it allows for the application of traditional controller design and standard software simulation.