SPH-GPU并行计算在风沙流中的应用

为了实现小尺度范围风沙运动的真实感模拟，采用基于拉格朗日力学无网格形式的光滑粒子流体动力学（smooth particle hydrodynamics，SPH）方法解决了基于欧拉网格法因网格大变形或者变形边界等引起的各种问题，并克服了不能用固定欧拉网格追踪任意单颗粒子运动轨迹的困难，因此该方法在研究风沙运动方面有着独特的优势。然而，随着风沙流动中SPH粒子数目的增加，该方法计算效率低，计算规模大的缺陷在风沙模拟过程中尤为明显。为了提高其计算效率，在CUDA软硬件平台上，建立SPH-GPU并行加速的二维气沙两相耦合模型，对串行的热点程序进行分析，找出最耗时且适合并行的热点程序；其次对GPU并行计算模型进行验证，宏观上得到了沙粒群运动的时空变化规律，微观上得到了典型沙粒的跃移轨迹和变异的尖角轨迹；最后对比了三种不同粒子数下CPU与GPU的计算效率。模拟结果证明SPH-GPU并行计算方法能够进一步应用在风沙流的数值模拟研究中。

Application of SPH-GPU Parallel Computing in Wind and Sand Flow

In order to realize the realistic simulation of the wind and sand movement in a small scale, the smooth particle hydrodynamics smooth particle hydrodynamics（SPH） method based on the Lagrangian mechanics meshless form is used to solve the grid based on the Euler grid method various problems which caused by large deformation or deformation boundary and overcome the difficulty of tracking the trajectory of any single particle with a fixed Euler grid. Therefore, this method has a unique advantage in the study of sand movement. However, with the increase of the number of SPH particles in the flow of wind and sand, the calculation efficiency of this method is low while the defects of large calculation scale are especially obvious in the process of wind and sand simulation. In order to improve its computational efficiency, software and hardware platform, a two-dimensional gas-sand two-phase coupled model of SPH-GPU parallel acceleration is established on the CUDA. Firstly, the serial hotspot program is analyzed to find the most time-consuming and suitable parallel hotspot program. Secondly, the GPU parallel computing model is verified. The temporal and spatial variation laws of the movement of the sand particles are obtained macroscopically while the jumping trajectories and the sharp trajectories of typical sand particles are obtained microscopically. Finally, the computational efficiency of CPU and GPU under three different particle numbers is compared. The simulation results prove that the SPH-GPU parallel computing method can be further applied to the numerical simulation study of wind and sand flow.