ABSTRACT The slag foaming practice enables high energy efficiency in the electric arc furnace, and its optimization is based on empirical knowledge and process modeling. Thus, it is essential to predict the physical properties of slags, as this phenomenon depends on their density, surface tension, and viscosity. Although thermodynamic calculations provide insights into the chemical composition of slags in equilibrium with liquid steel and refractories, and commercial software packages (such as FactSage) include modules to estimate slag viscosity, there are limitations regarding the chemical species considered and the properties that can be forecasted. In this study, predictive models of surface tension, density, and viscosity are evaluated for their ability to estimate experimental properties of metallurgical slags. These models, developed for glass compositions using machine‐learning algorithms, were subsequently validated on slags in this work. On the basis of their performance and limitations, a framework coupling FactSage and GlassNet was proposed and applied to investigate potential slag conditioners for controlling foamy slag. The tested candidates were silica sand, sodalite, and steel scale, and their effects on the characteristic times of foam collapse (or aging) were predicted in the results.
Falsetti et al. (Tue,) studied this question.