Abstract: Aiming at first kind of nonlinear system, a sliding mode variable structure control method based on system identification by using of neural network was proposed. The controlled system object was dynamically identified by two neural networks. Combining with variable structure control, the limit of variation range for controlled object in the conventional variable structure control was overcome. The simulation results show that the control system has good stability and convergence, meanwhile, system chattering is reduced and good robustness is achieved.

Key words: neural network, dynamic approach, sliding mode variable structure control, robustness