Thermodynamic based numerical modelling of inclusion removal in a bottom stirring ladle furnace

<p dir="ltr">Steel strength enhancement remains a critical challenge in steelmaking, with non-metallic inclusions significantly impairing mechanical properties. Although bubble blowing in ladle furnaces effectively reduces inclusions, challenges persist in optimizing bubble size, predicting removal efficiency, and characterizing bubbles and inclusions in high-temperature melts. Additionally, while alloying and hot rolling principles are well-established, accurately predicting mechanical properties for diverse steel grades remains complex due to high-dimensional variable interactions.</p><p dir="ltr">This study addresses these challenges by combining numerical modelling and machine learning. A CFD-based model simulates bubble generation from nozzles of varying sizes and geometries under different gas flow rates, with predicted bubble sizes validated experimentally and integrated into a bubble-inclusion attachment model. To improve removal efficiency prediction, interfacial energy changes during bubble-inclusion collisions were analysed, with the energy gap serving as a criterion for attachment feasibility. A comprehensive model incorporating flow and temperature fields was developed to quantify inclusion removal.</p><p dir="ltr">For mechanical property prediction, a machine learning approach with statistical feature engineering was proposed, reducing input variables from 46 to 13 while maintaining high accuracy (R > 0.94) on an industrial dataset of 12,000 samples. The model’s reliability was confirmed by aligning feature importance rankings with empirical data. This work demonstrates the potential of integrated CFD and machine learning methods to advance steel quality control and strength prediction.</p>