Identification and control of rotary traveling-wave type ultrasonic motor using neural networks

Neural networks (NNs) with varied learning rates are proposed to identify and control a nonlinear time-varying plant. First, the network structure and the online learning algorithm of an NN are described. To guarantee the convergence of error states, analytical methods based on a discrete-type Lyapunov function are proposed to determine the varied learning rates of a three-layer NN with one hidden layer. A rotary traveling-wave type ultrasonic motor (USM), which is driven by a newly designed high-frequency two-phase voltage source inverter using double inductances double capacitances resonant technique, is studied as an example of nonlinear time-varying plant to demonstrate the effectiveness of the proposed control system. Then, a robust control system is designed using two NNs to control the rotor position of the USM. In the proposed control system, the Jacobian of the USM drive system is identified by a neural-network identifier to provide the sensitivity information to a neural-network controller.