Graphics‐Processing‐Unit‐Accelerated 3D Online Heat Transfer and Solidification Model for Continuously Cast Slab

Abstract

A graphics‐processing‐unit (GPU)‐accelerated 3D heat‐transfer and solidification model is proposed to predict the temperature field and solidification field for continuously cast slab. The fully 3D model is developed based on the finite difference method and implemented on the compute unified device architecture (CUDA)/C++ platform. The model is verified by solving the 1D Stefan problem and validated by surface temperature measurements. The model is proven to be 18.91 times faster for the coarsest mesh of 35 million nodes and 22.92 times faster for the finest mesh of 50 million nodes than a central‐processing‐unit (CPU)‐based parallel version with 28 threads. The corresponding relative calculational times are 0.19 and 0.32, which indicate that the model has great computation efficiency. The model is proven to be robust enough for the dynamic continuous casting (CC) process via a test of dynamic casting speed. Further analysis of the nonuniform heat transfer and solidification shows that the model is able to predict the nonuniform heat transfer efficiently. In the studied cases, a higher casting speed will lead to a remarkable enhancement of the unevenness of heat transfer and solidification in general.