Improvement And Verification On Anisotropic Kernels Reconstruct SPH Fluid Surface

With the continuous improvement of the proformance of the computer hardware,computer virtual reality technology has been widely developed. As an important part ofcomputer simulation, fluid simulation is indispensable to show the realistic world. How tomake your computer simulate a more realistic fluid fastly has important practicalsignificance and application value.Firstly, using the widely used Smoothed Particle Hydrodynamics (SPH) method tosimulate the movement of fluid, Marching Cubes method to generate the threedimensional isosurface of fluid particle and shading language Cg to draw the fluid surface,we build the basic model of fluid simulation.Secondly, we analyze the randomness of the distribution of the fluid particle. In orderto make the fluid surface more smooth and the fluid simulation more anthenticity, thecalculation method of the position of the fluid particles is improved. We propose theclassification calculation method of the average position of the fluid particles: compute theaverage position for the particles outside the fluid, and do not compute the averageposition for the isolated particles and the particles inside the fluid, so that the amount ofcomputation can be reduced and the rendering speed can be improved.Thirdly, we carefully analyze the advantage and disadvantage of using sphericalmetaball or ellipsoidal metaball to build the surface of the fluid. Because sphericalmetaball make the fluid surface more smooth and vivid, while ellipsoidal metaball build asimple fluid surface and achieve a fast speed, we present a improved ellipsoidal metaballto reconstruct the fluid surface: using spherical metaball for the isolated paticles and theparticles inside the fluid, while ellipsoidal metaball for particles outside the fluid. At thesame time, because the traditional anisotropic kernels compute average position for eachparticle in the fluid, we simplified the anisotropic kernels to only need to do averagecalculation for the particles outside the fluid. At the same time, because the size ofmetaball is different after using the conventional ellipsoidal metaball to determine theanisotropy of every metaball, the method of using ellipsoidal metaball to build the fluidsurface is optimized, so that, the size of every metaball is similar and the reconstruction accuracy of the surface can be improved.Fourthly, carefully analyzed the traditional Marching Cubes grid search method usedin the process of the fluid simulation, in view of the traditional Marching Cubes gridsearch method retrieval each cube one by one in each frame, we propose a marked pointsearch method: just search cubes around the marked points in each frame to complete thework of searching for the cutted grid in each frame, thereby improving the computingspeed.Finally, by using VC++, OpenGL on Windows platform, we design a realistic fluidsimulation system to validate the effectiveness of the methods we proposed.