Analysis of predictors for modification of alumina inclusions in medium carbon steel

Currently in steelmaking, “inclusions engineering” has occupied a relevant space in terms of improvements in processes and products. This is because these compounds, depending on the chemical nature, morphology, physical state, size and distribution, can harm both the processes and the mechanical properties of the final product. Thus, the present work aimed to evaluate the modification of alumina inclusions by Ca addition in a medium carbon steel, by modeling the relationship between possible predictor variables in order to establish the fraction of liquid inclusions at the end of the secondary steel refining process. To achieve this goal, it was performed a statistical significance analysis by multiple linear regression. Feature engineering and feature selection methods were used, such as Pearson correlation filtering, “best subsets” method and statistical metrics like p-value, standard error, residuals plot, Durbin–Watson test and adjusted R2. Then, a statistically significant model was finally reached. Through the statistical and metallurgical discussion, the model indicated that the oxidation of steel, the titanium and sulfur content in steel are the main obstacles to the modification of non-metallic inclusions. In addition, computational thermodynamics proved to be an important ally in making decisions about the process.