Model Identification Based On NN And Variable Structure Control Research Of A Magnetic Levitation System

This paper studies on the system identification method based on neural network and control arithmetic of a set of NC magnetic levitation system. First it briefly describes the general composition and working principle of MLS, then the basic principle and arithmetic of neural network used for system identification Meanwhile applies the arithmetic to this system, so as to achieve a neural network model which could reflect the characteristic of the system best.Due to the speediness, robustness and simply-realization of the sliding mode variable structure control, a variable structure controller aiming at MLS is designed in the paper, meanwhile a simulation model is established by MATLAB/SIMULINK. The paper applies the sliding control law for neural network model achieved by system identification and gains a good effect. However, chattering exists inevitably on the switch surface of sliding mode variable structure controller. The frequent change of systematic parameters in short period will damage hardware of the system. Thus we should try to reduce chattering. The paper combines neural network with side mode control, optimizes the variable structure control of slide mode through neural network, studies on the equivalent side mode control arithmetic based on RBF neural network and side mode control arithmetic based on RBF network compensation Besides, a novel two-degree-of-freedom control method is designed based on neural network and sliding mode variable structure control, that is, first to design the input controller through side mode control, then the output feedback controller by neural network. This method takes full advantage of the self-study and management of neural network, in order to eliminate the influence brought by chattering during side mode control. Finally, control rule is added to neural network mode to simulate through MATLAB.The result shows sliding mode variable structure control has a good control effect on this system and proves web-based optimized side mode control ratio could combine both advantages, further improving control performance and enhance system robustness to suspend the system stably.