Modeling and control of an electric arc furnace using a feedforward artificial neural network

Previous studies have shown that the electric arc furnace is chaotic in nature and hence standard control techniques are not effective. However, human (heuristic) control is used every day on electric arc furnaces. A furnace operator assesses the performance of the furnace and makes judgments based on past experience and intuition. In order to improve the effectiveness of this control, a qualitative understanding of the operating conditions of the furnace is required. Artificial neural networks are capable of learning the system dynamics of the electric arc furnace. This article describes a feedforward neural network trained to model arc furnace electrical wave forms taken from an experimental arc furnace. The output of this model is then used in estimating the future state of the furnace for control purposes.