Self-organizing maps as a tool to assess possible substitution of fused by sintered MgO aggregates in MgO–C refractories
Sebastian Sado, Ilona Jastrzębska, W Zelik, Jacek Szczerba
Abstract
This work aimed to investigate if it is possible to substitute fused MgO aggregate with the sintered MgO in MgO–C refractories which could save significant amounts of energy. The Self-Organizing Maps (SOM) algorithm, belonging to the Machine Learning techniques , was used to classify 20 variants of MgO–C refractories, based on 10 different compositions with various fused-to-sintered magnesia ratios and shaped under 120 and 180 MPa, into clusters of comparable corrosion resistance . The industrially-produced MgO–C refractories were subjected to corrosion tests in an induction furnace at 1650 °C and 1720 °C using metallurgical slag with a CaO/SiO 2 ratio of 0.8. Results of wear rates and physical properties were analyzed using the SOM algorithm, which allowed us to prove that high-purity fused MgO in MgO–C composition can be substituted with lower-purity sintered MgO aggregate with no loss of corrosion resistance . These findings can greatly contribute to more sustainable production of MgO–C refractories.