Research On Phsically-Based Fluid-Solid Coupling Animation

With the rapid development of computer hardware, virtual reality and computer graphics, computer animations have penetrated into every aspect of daily life and work. Computer animations were created to explain, analyze, or visualize information for applications including scientific research, architecture, engineering, systems simulations, education, and documentary projects.With increasing demands for realism, the realistic image alone can not meet the demand. The demands of interaction effect and motion detail are growing strongly. This has brought great challenges to computer animation, especially the fluid-solid coupling animation. The traditional animation technology that draws fluid frame by frame, becomes no longer executable. Although the blocking animation can simulate the effect of the fluid to some extent, but it can be difficult to achieve the fluid-solid coupling animation. So the physically-based animation is becoming a widely popular method for the fluid-solid coupling animation.This thesis has fluid-solid coupling animation as research object, implemented a fluid-solid coupling animation system and improved the efficiency and fidelity of the traditional method. The main research includes four aspects: Firstly, we study the pyhsically-based fluid simulation; Secondly, research on the surface reconstruction of the particle-based fluids; Thirdly, research on the reconstruction thechnology for rigid body and the modeling for deformable bodies; Fourthly, we studied of the coupling fluid to solid bodies, and the parallelizing algorithm based on GPU. In this thesis, the fluid is imcompressible, like water. Solid is include rigid and non-rigid.The main contributions and innovations of own algorithm are as follows:1) In this thesis, the SPH (smooth Particle Hydrodynamics) technique is used to simulate fluid. An adaptively Stepped SPH framework was proposed to solve the constant time step problem in basic SPH. In the framework, each particle has its own time step. Compared with the basic SPH and Globally adaptive time stepping method, the proposed framework obtained an expected speedup.2) Since the SPH method uses the particles to simulate fluids, it is difficult to track the surface and reconstruct a smooth surface. To handle this, a surface reconstruction method based on anisotropic kernels was presented in this thesis. The proposed method simplified the basic anisotropic kernels, and classified the particles into inner particles and around surface particles based on the eigenvalue. The surface is reconstructed only by around surface particles. Experiments show that, the proposed method can create smooth surface, maintain the features, and improve the efficiency.3) In order to handle the solid surfaces, this thesis applied the free-form deformation technique, which commonly used in the computer animation, to surface reconstruction in a novel way. The proposed method solved the modeling problem that create animation model from slices data, which enriched the application areas for this thesis. For the reconstruction from slices, the vertices correspondences are very important. The reconstruction based on free-form deformation can build the correspondences very well, and the surface is approximate to the input slices, so it’s accuracy. The deformable solid can be created by add the contraints between the vertices.4) It presents fluid-Solid coupling animation system’s key thechnologies. In addition to fluid simulation and surface reconstruction, it includes following details:On the solid surface resampling, this thesis presented a resampling method based on the SPH. The proposed resampling method can obtain a points set with blue noise and very efficiently; Introduced the simulation of coupling fuild to the solid body, for calculating the forces between fluid and solid; Using GPU to parallize the system, sliding vectors was used to handle the multi-sort particles.