The Study Of Nonlinear System Neural Network Sliding Mode Variable Structure Control

At present, nonlinear system identification and control has become the forefront and hot problem domestically and abroad; sliding mode variable structure control has opened a new way in dealing with complex control of nonlinear systems, thus it is important in theory and application significance to research sliding mode variable structure control.In this paper, adaptive control, sliding mode variable structure control and neural network are combined, a superior neural network adaptive variable structure control is introduced. At the same time, the convergence of the algorithm training the neural network is strengthened.A nonlinear system sliding mode variable structure control based on system identification is given ;plants are identified dynamically through two neural networks, system's stability, convergence and quickness being combined; the needing the plant's variable scope of parameters as SMVSC before is avoided, and chattering becomes smaller, so the robustness is strengthened.In order to raise the training speed and the convergence of learning algorithm of the neural network, a new neural network adaptive variable structure control based on variable structure two-order learning algorithm is developed. A new two-order Adaline neural network is put forward first,which can train quickly,and approach to nonlinear systems well.Identifying systems by using it can realize the weights of the neural network are adjusted on line and it has the real-time property.The algorithm has two-order convergence speed. The learning error converging to a narrow-band sliding mode area has been proved.In order to lessen the time of reaching sliding mode surface, and to play the merits of neural network and SMVSC, a SMVSC system which is combined with neural network is studied, and gives the design method of neural network sliding mode surface, realizing the omnidistance sliding mode control over the movement of the system. Then it gives a way which can online adjust controller's parameters based on neural network, so as to lessen the chattering. The simulation results prove the effectiveness of the developed scheme in this paper.