The intelligent arc furnace controller: a neural network electrode position optimization system for the electric arc furnace

The amount of power delivered to the scrap metal in an electric arc furnace is controlled by positioning three large electrodes. The authors present quantitative performance results of the intelligent arc furnace regulator and analyze qualitatively how a neural network improved furnace operation. A neural network control system can learn to predict the relationships between the electrode position and stability of furnace operation. Real-time adaptation to changes in furnace conditions enables the system to increase productivity while reducing furnace and electrode wear, lowering power consumption, and achieving an unprecedentedly smooth melt. The intelligent arc furnace controller has been able to create substantial cost savings, and has proven to be remarkably stable during five months of full-time operation