Recurrent neuro-fuzzy modeling of a biotechnological process
Gregorio Ismael Sainz-Palmero, M.J. Fuente, Pastora Vega · 2001
This paper deals with the develop of a new recurrent neuro-fuzzy model for complex systems from input-output data. In this paper a recurrent fuzzy neural network, called RFasArt (Recurrent FasArt), has been applied to model a complex biotechnological process: a wastewater treatment plant. This network is based on the Adaptative Resonance Theory (ART) but introducing formalisms from the fuzzy set theory and taking into account the contextual information in its processing stage. In order to compare the results obtained with this fuzzy neural network, a classical recurrent neural network has been implemented and used to model the same process. The results shows the better behaviour of the RFasArt system. At the same time with this architecture it is possible to obtain a knowledge base of fuzzy rules where is stored all the knowledge of the plant learned by the fuzzy neural network.