Multilinear Fluid Simulation with Model Reduction for Pressure Projection

Grid-based simulation usually requires high resolution grid and heavy computation time to produce high quality simulation results. We propose a novel example-based method that reduces the simulation time. The main strategy for the time reduction is to avoid accurate but expensive calculation for mass conservation. Instead, we show that a linear combination of precomputed examples can approximate the intended computation. A tensor-based model reduction scheme is applied to handle the linear combination efficiently. The experiment results tested in 2D verify that our method produces high quality simulation results efficiently even on a high resolution grid.