GPU加速的定向凝固枝晶生长的元胞自动机模型

GPU-accelerated Cellular Automaton Model for Dendritic Growth during Directional Solidification

基于元胞自动机模型,采用GPU并行计算技术,在单个GPU上实现了凝固过程枝晶生长的元胞自动机计算模型模拟。以Al-Cu二元合金为例,对单个等轴晶生长过程和柱状晶生长过程进行了数值模拟。随着过冷度增加,等轴枝晶逐渐发展出二次枝晶臂和三次枝晶臂;在定向凝固枝晶竞争生长中,择优取向枝晶的二次枝晶臂阻挡了非择优取向枝晶,进而淘汰了非择优取向枝晶。与基于CPU的并行模型进行计算速度比较,验证了基于GPU的计算模型的加速效果。结果表明,在测试范围内,基于GPU计算模型的加速比最大可达10.90。

Based on the cellular automaton model and GPU parallel computing technology, the cellular automaton computing model for dendritic growth during the solidification process was realized on a single GPU card. Taking Al-Cu binary alloy as object, the single equiaxed crystal growth process and columnar crystal growth process were numerically simulated. With the increase of undercooling degree, secondary and tertiary dendrite arms are gradually developed in equiaxed dendrite, where the secondary dendrite arms of non-preferentially oriented dendrites are blocked by the preferentially oriented dendrites during directionally solidified dendrite competition growth, which is then eliminated. Comparison of the calculation speed with the CPU-based parallel model verifies the acceleration effect of the GPU-based calculation model, indicating that the speedup ratio can reach up to 10.90.