基于GPU并行计算的风沙流SPH数值分析

在研究风沙流动方面,光滑粒子流体动力学方法(SPH)的无网格性有着独有的优势。利用SPH方法研究风沙流动时,需要将整个计算区域离散成数量庞大的单个粒子,因此计算规模大、计算效率低。为提高SPH方法的计算效率,采用支持并行计算的CUDA平台,利用GPU大规模并行计算技术,实现SPH方法数值模拟时的加速运算。以二维气沙两相耦合模型作为数值算例,利用GPU并行计算详细分析颗粒群的运动规律。比较在不同粒子数下CPU与GPU的计算效率以及GPU线程数对计算效率的影响。对所得结果进行统计分析后,得到了单颗沙粒的典型抛物线形和变异的跃移轨迹。模拟结果证明:SPH-GPU并行计算技术能够应用在风沙流结构的数值模拟研究中。

SPH Numerical Analysis of Wind and Sand Flow Based on GPU Parallel Computing

Smooth particle hydrodynamics(SPH)has a unique advantage in the study of wind and sand due to its meshlessness.When the SPH method is used to study the wind and sand flow,it is necessary to discretize the entire calculation area into a large number of single particles,so the calculation scale is large and the calculation efficiency is low.In order to improve the computational efficiency of SPH method,CUDA platform which supports parallel computing and GPU massively parallel computing technology were adopted to realize the accelerated operation of SPH method in numerical simulation.Taking the two-dimensional air-sand two-phase coupling model as a numerical example,the particle swarm movement law was analyzed in detail by using GPU parallel computing.The computational efficiency of CPU and GPU under different particle numbers and the influence of the number of GPU threads on the computational efficiency were compared.After statistical analysis of the results,the typical parabolic shape and the variant saltation trajectories were obtained.The simulation results show that the SPH-GPU parallel computing technology can be applied to the numerical simulation research of the wind and sand structure.