一种基于DirectCompute加速的实时流体仿真框架  

Real Time Fluid Simulation Framework Based on DirectCompute

作　　者：刘璐[1] 刘箴[1] 何高奇[2] 陈田[3] 刘婷婷[1] 刘翠娟[1] 

机构地区：[1]宁波大学信息科学与工程学院,宁波315211 [2]华东理工大学计算机科学与工程系,上海200237 [3]上海电机学院,上海200000

出　　处：《系统仿真学报》2016年第10期2467-2475,共9页Journal of System Simulation

基　　金：国家自然科学基金(61373068);宁波市科技计划项目(2015A610128;2015C50053;2015D10011;2016D10016);高等学校博士学科点专项科研基金(20133305110004)

摘　　要：针对流体仿真中流体粒子规模增大导致的仿真效率较低和较复杂固液边界处理耗时较长的问题,提出了一种基于DirectCompute加速的实时流体仿真框架,仿真中采用一种基于网格的局部碰撞检测方法来加快搜索固流边界,使用DirectCompute技术将流体计算、碰撞检测和流体表面重构部分放入GPU(Graphics Processing Unit)中并行计算来加速SPH(Smoothed Particle Hydrodynamics)流体仿真,并利用其特性优化了SPH流体仿真的存储结构,节省了存储空间。实验结果证明,该框架有效提高了带有较复杂静态场景的流体仿真速度,并能展现出流体水花细节。A real time fluid simulation framework based on DirectCompute acceleration was proposed to solve problems of low simulation efficiency that caused by the increasing fluid particle size and long processing time that caused by the complex solid-liquid boundary. In the simulation, a grid local collision detection method was adopted to speed up the search for the solid-liquid boundary. Fluid calculation, collision detection and surface reconstruction were computed parallel in the GPU with DirectCompute technology to accelerate the SPH fluid simulation. It could optimize the storage structure of fluid simulation, and save the storage space. Experimental results show that the framework can effectively improve the fluid simulation speed with more complex static scene, and can show the details of the fluid spray.

关 键 词：流体仿真 光滑流体动力学方法 碰撞检测 GPU加速 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]