CFD-Derived Regression Model to Predict Surface Velocity for a Continuous-Casting Round Billet Mold

Abstract

Highlights A 3D electromagnetic–fluid dynamic model was established to simulate mold surface flow in continuous casting. Predictive formulas for maximum surface velocity and high-velocity range were derived under varying process parameters. A CFD-derived regression model predicts key surface-flow characteristics within milliseconds with satisfying accuracy. Abstract In the continuous casting process, the flow velocity of molten steel at the surface of the mold directly influences the melting of the flux, slag entrainment behavior, and the uniformity of heat transfer. Fast and accurate prediction of the magnitude and distribution of surface-flow velocity is crucial for defect suppression and ensuring continuous casting quality. In this study, a three-dimensional electromagnetic–fluid dynamic coupled model was established and validated to simulate the molten steel flow behavior under various casting speeds, Electromagnetic Stirring (EMS) currents, EMS frequencies, and Submerged Entry Nozzle (SEN) immersion depth. Subsequently, the maximum surface velocity in the mold was fitted to derive a predictive formula under different conditions; the median range and width of the surface high-velocity range were predicted in a similar way. As a result, a regression model derived from Computational Fluid Dynamics (CFD) was built to predict the maximum mold surface velocity and the high-velocity range within milliseconds. The reliability of the regression model was verified by newly generated simulation results, with the relative error between the predicted and simulated maximum surface velocities kept within 2%. It was shown that the maximum surface velocity increases with casting speed and EMS current but decreases with EMS frequency. The median range of the high-velocity range decreases with increasing casting speed and current. The immersion depth of SEN had a minimal impact on the maximum surface velocity in the mold and the high-velocity range. This regression model is beneficial for rapid mold design and for predicting potential slag entrainment, breakout risks, and billet defects, for the purpose of intelligent continuous casting.