基于 PCISPH 的流体粒子飞溅改进方法-《计算机技术与发展》

文章信息/Info

Title:: Improved Spattering of Fluid Particles Based on Predictive Correction of Incompressible SPH

作者:: 钮倩倩¹ ; 林绿开¹ ; 李毅¹; 2*; 1. 温州大学计算机与人工智能学院,浙江温州 325000;
2. 浙江大学计算机科学与技术学院,浙江杭州 310058

Author(s):: NIU Qian-qian1 ; LIN Lyu-kai1 ; LI Yi1; 2*; 1. School of Computer and Artificial Intelligence,Wenzhou University,Wenzhou 325000,China;
2. School of Computer Science and Technology,Zhejiang University,Hangzhou 310058,China

Keywords:: smoothed particle hydrodynamics; fluid simulation; particle splash improvement; prediction correction incompressible SPH; Poisson equation; Taichi programming

摘要:: 流体飞溅是自然界中最常见的流体现象。流体模拟是计算机图形学的一个重要研究分支。流体模拟已经广泛应用于电影、游戏和其他工业数值领域。由于飞溅场景中流体的密度和压力的速率变化非常大,因此模拟对离散化解的精度要求很高。对于流体粒子飞溅模拟数值不稳定,且缺乏真实效果,针对经典的预测校正不可压缩 SPH(PCISPH) ,该文提出了基于粒子表面流体飞溅改进的泊松压力解的方法,通过替换压力源项提高了模拟精度。实验结果表明,该方法中流体的压力分布较之经典方法更接近现实的流体粒子飞溅模拟效果,并且改进后的方法经过剧烈运动后帧率仍符合实时性仿真效果的要求,保证了良好的压力稳定性和真实感。该文通过 Anaconda3 集成 Python3 和 Taichi 环境实现流体仿真实验。

Abstract:: Fluid splashing is the most common fluid phenomenon in nature. Fluid modeling is an important branch of computer graphics.Fluid simulation has been widely used in film,games,
and other industrial numerical fields. Because the density and pressure of the fluidin the splash scene vary largely,the simulation requires high accuracy of discrete solution. For the unstable and lack of real effect of fluidparticle splash simulation,we propose an improved Poisson pressure solution method based on particle surface fluid splash for the classical　predictive correction of incompressible SPH ( PCISPH) ,which improves the simulation accuracy by replacing the pressure source term.The experimental results show that the pressure distribution of the fluid in this method is closer to the realistic fluid particle splashsimulation effect compared with the classical method,and the frame rate of the improved method still meets the requirements of real-timesimulation effect after intense movement,which ensures good pressure stability and realism. We implement fluid simulation experimentsusing Anaconda3 to integrate Python3 and Taichi environments.