입자-격자 혼합형 유체 시뮬레이션에서의 시각적 향상 기법 : Visual Enhancement Methods in Hybrid Particle-Grid Fluid Simulation

다국어 초록 (Multilingual Abstract)

The realistic simulation of fluids largely depends on a temporally coherent surface tracking method that can deal effectively with transitions between different types of flows.We model these transitions by constructing a very smooth fluid surface and a much rougher, splashy surface separately, and then blending them together in proportions that depend on the flow speed. This allows creative control of the behavior of the fluids as well as the visual results of the simulation. We overcome the well known difficulty of obtaining smooth surfaces from Lagrangian particles by allowing them to carry normal vectors as well as signed distances from the level set surface and by introducing a new surface construction algorithm inspired by the moving least-squares (MLS) method. We also implemented an adaptive form of the Fluid-Implicit-Particle (FLIP) method that only places particles near visually interesting regions, which improves performance. Additionally, we introduce a novel subgrid solver based on the material point method (MPM) to increase the amount of detail produced by the FLIP method. We present several examples that show visually convincing water flows.

목차 (Table of Contents)

• 1 Introduction 7
• 1.1 Motivation 7
• 1.2 Problem Definition and Contributions 10
• 2 Background and Related Work 13
• 2.1 Particle-Based Fluid Simulation 13
• 2.2 Grid-Based Fluid Simulation 15
• 2.3 Particle In Cell 16
• 2.4 Material Point Method 18
• 2.5 Level-Set Method 19
• 2.6 Surface Reconstruction 20
• 3 Algorithm 24
• 3.1 Incompressible Material Point Method 24
• 3.1.1 MPM Explicit step 25
• 3.1.2 Incompressible Fluid 27
• 3.1.3 Hybrid FLIP-MPM Solver 28
• 3.2 Responsive Surface Tracking 31
• 3.2.1 Smooth Surface Construction 31
• 3.2.2 Velocity-Based Blending 33
• 3.2.3 Surface Reconstruction with Anisotropic Kernel 34
• 3.3 Adaptive Particle Placement 37
• 4 Results and Discussion 41
• 4.1 Visual Enhancement of FLIP with MPM Sub-grid Solver 41
• 4.2 Performance of the Adaptive FLIP 45
• 4.3 Application of the Responsive Surface Tracking 46
• 5 Concluding Remarks 50
• 5.1 Conclusion 50
• 5.2 Future Work 51