GPU加速的SPH方法在溢洪道水流模拟中的应用

    Application of GPU-based accelerated SPH method in flow simulation of spillway

    • 摘要: 基于拉格朗日描述的光滑粒子动力学方法(SPH)擅长于处理自由面剧烈变化的水流现象,十分适合水利工程中泄洪等问题的数值模拟。然而,SPH方法通常采用均匀分布的粒子对流体计算域进行空间离散,对于工程问题而言需要的粒子数量较多、计算量大。为了突破SPH方法在实际大规模计算中的适用范围,采用C++和CUDA混合编程的技术,借助GPU实现了对SPH方法的并行加速。通过WES三圆弧段组成的光滑溢洪道过流问题,验证了GPU加速的SPH方法的计算精度和可靠性,计算效率相对原始的SPH仿真过程提高了61.8倍。最后,将GPU加速的SPH方法应用于水利工程的溢洪道泄流问题,分别模拟了光滑溢洪道和台阶式溢洪道流动特性,通过自由面的演化过程及泄流沿程截面上的速度分布状态,对比分析了台阶对泄流现象的影响。

       

      Abstract: Smoothed particle hydrodynamics (SPH), a smooth particle dynamics method based on Lagrangian description is good at numerical simulation of flood discharge with drastic changed free surface. However, the SPH method usually uses uniformly distributed particles to discretize the computational domain of the fluid in space. For engineering problems, the particles numbers and calculation complexity are large. In order to break through the application scope of SPH method in practical large-scale computing issues, the parallel acceleration of SPH method was realized by the mixed programming technology of C++ and CUDA and the help of GPU hardware equipment. Through calculating the flow passing issue of smooth spillway composed of three arc segments of WES, the accuracy and reliability of GPU-based SPH method in this paper are verified. The computational efficiency is 61.8 times higher than that of the original SPH simulation process. Finally, the GPU-based accelerated SPH method was applied to the discharge simulation of hydraulic engineering, the flow characteristics of smooth spillway and stepped spillway were simulated respectively. By observing the variation process of the free surface and the velocity distribution along the discharge section, the effect of the step on the discharge phenomenon is analyzed.

       

    /

    返回文章
    返回