Dynamically evolving fluid simulation

Abstract

The Eulerian framework is widely used to simulate large scale fluid with surface and motion detail. Grid-based simulation is the most common scheme, but it requires a large computational cost and memory usage. To reduce memory usage, our method allocates fluid and nearby cells only. The allocated region is dynamically evolved during simulation. Speed up is another main focus of the paper. Each simulation step is bounded by the magnitude of the maximum velocity. A high velocity requires a large number of small time steps. Instead of relying on implicit scheme, which inevitably sacrifices simulation quality, propose a new method to keep time step large by reducing maximum velocity via moving frames. The moving frame approach allows the reduction in maximum velocity. Our method reduces memory usage by dynamic allocation of cells on grid, and provides speed-up strategy for numerical simulation by applying galilean invariance. We show the validity of our method with a set of liquid simulation.