Identification and speed control of ultrasonic motors based on neural networks

An ultrasonic motor (USM) is a newly developed motor that has many excellent performances, useful features and extensive applications. The operational characteristics of the USM are affected by many factors. Strongly nonlinear characteristics could be caused by the increase of temperature, the changes of load, driving frequency and voltage and many other factors. Therefore, it is difficult to perform effective control on USMs using traditional control methods based on mathematical models of systems. Recently, artificial intelligent methods based on neural networks have become the main approaches to perform USM control. However, the existing neural-network-based methods for USM control have some shortcomings, such as complex network structures, slower convergent speeds and lower convergent precision, as well as no theoretical guarantee on the convergence of control. Furthermore, it is difficult to obtain accurate control input for the USM by using a speed controller with a single control variable. In this paper, a bimodal controller is designed where both the driving frequency and amplitude of the applied voltage are used as control inputs. A novel input–output recurrent neural network (IORNN) identifier is constructed to dynamically identify the input–output relation of the ultrasonic motors. To guarantee convergence and for faster learning, the adaptive learning rates are derived using discrete-type Lyapunov stability analysis. Numerical results show that the proposed IORNN identifier can approximate the nonlinear input–output mapping of ultrasonic motors quite well. Compared with the existing method, the control precision can be increased by about three times and the convergence time can be decreased by about two times when the proposed method is employed. Good effectiveness of the proposed control scheme is also obtained for various reference speeds.