Abstract
This paper presents application of a class of hybrid neuro-fuzzy network to the solution of a particular complex problem. The primary objectives are both to investigate the capability of adaptive neuro-fuzzy networks and to justify their application to predict the v-i characteristics of nonlinear, multi-variable, complex systems such as electric arc furnaces. The novelty of this work is proposing a feedforward neuro-fuzzy structure suitable for long-term prediction. Successful implementations of feedforward neuro-fuzzy predictors are described and their performances are illustrated using the results obtained from adaptive neuro-fuzzy networks and recorded data.