Fuzzy Modeling Of Nonlinear System And Its Application In The Inverted Pendulum System

According to the conventional control theory, it is necessary to establish a model of the controlled plant. In general cases, modeling with high accuracy is very difficult for some complicated nonlinear system with serious coupling. Even if the model of system can be given in some cases, its actual use is usually restrained because of the difficulty of solving an effective and easy realized control strategy. Therefore, how to establish an effective model is a main problem in design of control system. The method of fuzzy modeling can be taken as a good candidate of this problemIn this thesis, the inverted-pendulum system is firstly analyzed by using T-S fuzzy model due to its property of arbitrary approximation for the non-linearity. Based on the parameters identification, simulations of system are made by using fuzzy logic toolbox of Matlab. The results of comparison with LQ approach show the performance improvement in the aspects of overshoot and fast response.The fuzzy hyperbolic model, a new approach for modeling of nonlinear system, is subsequently discussed for its validity. Based on selecting the initial weights of neural network according to the priority of each input, the parameters can be identified by using BP neural network, which is good for reducing the randomness of initial weights and computing time.By combining self-study ability of neural network with knowledge-expression of fuzzy set theory, fuzzy neural network with self-correction of membership function is lastly constructed to make the approximation of somenonlinear functions.