Solid-Fluid Interaction Simulation System Based on SPH Unified Particle Framework
Solid-fluid Interaction Simulation is the important research point in the field of fluid simulation. Existing researches mainly focus on the phenomena such as motion, deformation, infiltration, erosion, etc., and less attention is paid to the solid fracture caused by fluid impact. The simulation solutions used by the related researches have weakly coupling with the SPH solid-fluid interaction framework because their solid fracture methods need the geometric form for representing and processing the solid model. For the same reason, their methods have difficulty to parallelize and achieve the balance between Realism and real-time performance. To deal with these problems, a full process parallelization simulation framework is proposed for parallelize each key stage of the simulation system. The framework is based on the SPH unified particle framework and has the advantage of high parallelism of the SPH method because of the particle form of solid used to describe and solve solid field quantities. For solving the solid fracture, a physics-geometry hybrid method with the high parallelism implementation is proposed. This method is also based on particle form to solve the solid field quantities so that have strong coupling with SPH framework. In term of neighborhood particle search, a spatial three-level index sorting based on the unified grid method is proposed to optimize neighborhood search processes including solid particles. Because the spatial characteristics of the solid fracture method is considered, the two methods can be closely combined with each other. In the stage of fluid rendering, in order to avoid the communication between GPU and CPU caused by surface extraction, the Screen Space Fluid Rendering method is adopted. In SSFR, the particle data is directly input into the rendering pipeline on GPU for rendering without transfer to CPU for extracting fluid mesh. For verifying the stability of the parallelized framework, performance and Realism of the simulation system, some scenes with different complexity are constructed for simulating. The results show that the framework can achieve well real-time performance in the scene with 650k particles and 180 solid chunks, and simulation is stable with the effects conformed to the phenomena of real-world physics.