Speed control of elliptec motor using adaptive neural-fuzzy controller with on-line learning simulated under MATLAB/SIMULINK

Ultrasonic motors (USM's) possess extremely nonlinear characteristics which depends on environmental conditions such as temperature, load torque, driving frequency, voltage, etc. These characteristics vary with fluctuation of driving parameters and environmental conditions. The piezoelectric USM Elliptec motor is a standing wave type that feeds from single power source which was modeled already by GMDH neural network approach and the aim of this paper is to speed control of model by controlling frequency and voltage feeding USM. Sugeno-type Fuzzy Inference System which consequent parameters adapted by online Extended Kalman Filter and premise parameters by online back propagation algorithm, was employed to control driving voltage whereas a fuzzy inference system applied to finding appropriate frequency to reduce great speed error. The ANFIS controller uses the error signal and as first input and error variation as second input and outputs voltage in order to fine tuning of speed error. The model implemented under MATLAB/SIMULINK which is a suitable environment to develop real time applications in an easy way.