Research On Magnetic Bearings Control Based On Fuzzy Neural Network Control Algorithms

Active Magnetic Bearings(AMB) are a new type of mechatronic system.It supports the rotor untouched by the force of active electromagnetic and magnetizer and the position of rotor is controlled by the control system.The control system is to suspend the rotor steady and freely through adjusting the magnetic field according to the suspension status of rotor.The performance of controller affects the dynamic characteristic and the control precision of AMB.AMB need that the control system have small overshoot,fast response and strong disturbance rejection ability.The AMB using PID controller have big overshoot,weak disturbance rejection ability and difficult adjustment of parameters,so it is not satisfied.The simplied mathematical model is set up through analyzing one-freedom AMB.First the AMB system is simulated in Matlab using fuzzy interface system(FIS),then a fuzzy controller is designed.Through analysis it is concluded that the two key factors affecting the steady error of fuzzy controller are 'ke' parameter and quantization strategy of 'zero' in the discrete domain of error.Simulation indicates that choosing a proper 'ke' parameter and a proper quantization strategy of 'zero' in the discrete domain of error can smaller the steady error obviously.The control strategy of Nine-point controller is analized,two kinds of Nine-point controllers with two kinds of different control rules are used to control of AMB.Aiming at a big overshoot,a forty-nine-point seven state controller is proposed.Simulation shows that the AMB system using a forty-nine-point seven state controller have smaller overshoot.A Neuro-Fuzzy controle scheme is presented,in which fuzzy control and neural network are combined.on the base of it an adaptive Neuro-Fuzzy controller is designed,in which the weights of neural network are modifying on line according to back propagation algotithms. Simulation shows that the adaptive Neuro-Fuzzy controller is better adaptive when the parameters of AMB model are changed.