Abstract: The flow and heat transfer state of molten steel within the slab continuous casting mold is a critical factor determining the quality of the final slab.Utilizing artificial intelligence technology to achieve real-time,precise prediction and intelligent control of this complex multiphysics field is of great significance for improving the quality of high-end steel and promoting the intelligent transformation of the steel industry.To this end,this study first established a mechanistic model of molten steel flow,heat transfer,and solidification under electromagnetic stirring（EMS）in a slab continuous casting mold.Furthermore,a set of flow field evaluation criteria for the mold was proposed—namely,a steel-slag interface slag entrapment-freezing index,a shell uniformity index,and an inclusion removal index—with the aim of optimizing the EMS process.Secondly,based on the dataset of 3 Dflow and temperature fields generated by the aforementioned model,a largescale multiphysics prediction model for the mold was developed using a deep neural network（DNN）architecture,enabling rapid prediction of the multiphysics field within the mold.The results show that,compared to traditional numerical simulation results,the prediction errors of the large model for the multiphysics fields,including the flow and temperature fields within the mold,are all within 10%.Meanwhile,the model′s computational speed was significantly increased,with the average computation time to obtain the multiphysics field within the mold being drastically reduced from the original 24 hours to 2 seconds.This research provides key technical support for achieving online optimization and closed-loop control of the EMS process and for the construction of a " Digital Twin" system.