Rapid velocity magnitude flow field prediction in electromagnetically stirred continuous casting via a proper orthogonal decomposition-based reduced-order model

A reduced-order model (ROM) based on proper orthogonal decomposition (POD) is proposed, integrating POD methodology with regression techniques to predict velocity magnitude fiow field morphology in continuous casting electromagnetic stirring processes under varying operational parameters. Computational fiuid dynamics (CFD) simulations were performed to calculate fiow fields at various casting speeds and applied currents, resulting in a comprehensive sample database. The velocity matrix was decomposed via POD, and regression models were subsequently trained to correlate operational parameters with mode coefficients. Validation against CFD simulations proved the ROM’s effectiveness in predicting velocity magnitude fiow fields under electromagnetic stirring, with maximum relative errors of 6.5% in the mold region and 8.4% in the turbulent stirring zone. Notably, the POD-based ROM achieved a computational efficiency three orders of magnitude higher than conventional CFD simulations (prediction time of about 1/1000 of CFD).