Neuro-fuzzy control using self-organizing neural nets

Abstract

This paper discusses a new approach to design a fuzzy logic control system, based on the self-organizing map (SOM) neural network. SOM is used to generate multivariate fuzzy state space from system's input-output data through unsupervised training. The trained SOM is then used as a part of an inference mechanism for a fuzzy logic controller. The proposed method is compared with other fuzzy neural network approaches. Sample data from a chemical plant is used to demonstrate the technique.<>