Energy efficiency of an Electric Arc Furnace with SVM-RFE

Melting scrap for the manufacturing of special steels in Electric Arc Furnace (EAF), requires high energy consumption at a considerable cost. Identifying influential variables in the consumption of energy in this process is a priority. In this paper we propose using a data mining algorithm, specifically implementing the Support Vector Machine - Recursive Feature Elimination (SVM-RFE) algorithm, to select and rank the most influential variables to achieve efficient energy consumption in the melting process in an EAF. The SVM-RFE algorithm obtains coefficients and generates a ranking list which identifies highly discriminating variables or of greater influence on energy consumption in the melting process in an EAF. The variables identified by the data mining algorithm were monitored and adjusted in the control system of the EAF, obtaining as a result, reduction in energy consumption in steel casting tests, regarding the historical records in 15.81 kilowatts hour per charged metal tons (kWh/TCM) per steel casting; i.e., 3.55 ±0.5% saved and a shorter time 1 ± 0.3 minutes per steel casting.