Neuro-fuzzy modeling of nonlinear systems for control purposes
T. Culliere, André Titli, J.M. Corrieu · 2002
In this paper we propose a neuro-fuzzy model for the identification of the nonlinear dynamic systems. The new model is composed of two stages. The first stage consists of a cut-out of the input space in areas. This global treatment is done by the fuzzy module. The second stage consists of a local identification of the system by several simplified neural networks. This article describes the first stage with an independent simple fuzzy model and the second stage with a neural one. Then it presents the complete model and shows the modifications of the backpropagation algorithm for the multiple neural network's learning. Simulations on examples and in particular on invert pendulum showing the neuro-fuzzy's ability.>