Comparison Of RPROP And Improved BP Neural Networks Performance In Electric Arc Furnaces

Electric arc furnaces are preferred for the production of liquid steel because of the flexibility of the processes in the production of liquid steel, investment and operating costs are lower than other production equipment. The principle of operation of electric arc furnaces with alternating current is based on melting the electric current through scrap by controlling the carbon electrodes. Nonlinear, dynamic, have more than one parameter, have a complex characteristic system. The current-voltage fluctuations and variable parameters of the electric arc furnace, as well as the operation mode and the intuitive operation of the operators, further complicate. Control of similar complex systems is more difficult than linear systems. It is therefore important to control the motion of the electrodes that perform efficiency and power transfer in electric arc furnaces. In this study, an artificial neural network with a multilayered network structure and a resilient back propagation algorithm (RPROP), as well as improved back propagation algorithm (BP) performances, which have proven effective in previous studies, were compared. Both methods have been found to be successful in terms of unstable systems, but it has been determined that resilient back propagation algorithm is learned with faster and lower error in the experiments with the same data set. The study was performed using Knime open source data analysis tool and artificial neural networks and the results were evaluated