Recurrent Fuzzy Neural Network Variable Structure Position Controller Based On Vector Control Of PMLSM

Position control of permanent magnet linear synchronous motor (PMLSM) is realized by MATLAB/SIMULINK software in this thesis. Based on the mathematics model of PMLSM which is analyzed in chapter 2, the simulation model of vector control system with SVPWM is built up, and the simulation results show that it has fast dynamic response and easy to realize, but has the disadvantages of big ripples of both thrust and current, and less robustness when external load disturbances exist.In order to reduce the ripples, and improve robust performance, variable structure control which belongs to modern control strategy is introduced. A variable structure adaptive (VSA) position controller is adopted, where a simple adaptive algorithm is utilized to estimate the uncertainty bounds. The simulation results demonstrate that its dynamic response and position track performance are greatly improved, and they have much smaller thrust and current ripples. With the external disturbance, position error and speed perturbation are apparently reduced, and the system is robust to parameter variations. However, the convergence rate of adaptive mechanism is quite slow, it can not deal with the external disturbance effectively.To compensate the disadvantages, the combination of variable structure and intelligent control is expected. By means of the nonlinear adaptive and learning ability of fuzzy neural network (FNN), a recurrent fuzzy neural network (RFNN) variable structure position controller is investigated, in which the RFNN is utilized to estimate the real-time lumped uncertainty. Comparing with the VSA controller, RFNN has nearly zero thrust and current ripples, less position error and no speed perturbation, the rate of estimation convergence is further improved, and the chattering is reduced. It is proved that RFNN has high dynamic characteristics and is robust to parameter variations and external disturbance.