Fuzzy adaptive control based on RBFN

Aiming at practical plants with strong nonlinear characteristics or changing operating points, the paper develops a fuzzy adaptive control strategy based on RBFNs, continuing the work in Chen Xiaohong, Gao Feng, and Qian Jixin (1996, 1997). Theoretically, neural networks can approximate any given sample sets accurately, but an industrial plant is often highly non-linear or with changing operating points. It is very difficult to approximate such a plant by only one neural network. On the other hand, the multi-model method brings about the oscillation problem. The control strategy proposed in this paper possesses not only the performance of high accuracy like that of the multi-model method, but also eliminates the disadvantages of the method. Simulation results from pH control in a CSTR demonstrate the above properties