A fully dynamical fuzzy neural network

A fuzzy neural network with dynamic weights is proposed. The network topological architecture and the supervised learning algorithm are given. This novel network has some distinct features and considerable advantages: (1) each basic dynamic weight of the network is a dynamic subsystem; (2) the input space is partitioned into a number of fuzzy boxes; (3) no a priori knowledge including the order of controlled systems is required; (1) it does not require the structuring of feedforward or inverse models of plants through neural networks; (5) the network is essentially a nonlinear controller with learning abilities; and (6) the initial basic dynamic weights can be widely interpretation. The proposed adaptive and learning control system is applied to the control of the pH neutralization process. The simulation investigations show that the dynamic learning control system has better dynamic quality, stronger robustness, and more adaptation and intelligence compared to the present conventional control techniques using an explicit and quantitative mathematical model.