RECURRENT WAVELET NEURAL NETWORKS BASED ADAPTIVE CONTROL FOR SERVO DRIVE SYSTEM
OF INDUCTION MOTOR

Wu Qinghui  Shao Cheng

(Institute of Advanced Control Technology,Dalian University of Technology, Dalian, 116024)

Abstract: An adaptive control system using recurrent wavelet neural networks (RWNN) is presented, which is based on vector control for an induction servo drive motor characterized by non-linear, multivariable, strong coupling, slow time-varying properties, etc, and many uncertainties such as mechanical parametric variation and external disturbance. The new method is to overcome the limitation of conventional position and velocity PID that depends on the accurate model and cannot guarantee satisfactory control performance and better dynamic characteristics. A back propagation algorithm based on grads descent is developed to train the RWNN on line using delta adaptation law. Simulation results demonstrate the effectiveness of the proposed method.

Key words: Servo induction motor  Recurrent wavelet neural networks  Adaptive control  Robust control

CLC No: TP13

Received 20040721, received in revised form 20020040925