Cell-Constrained Particles for Incompressible Fluids

Abstract

Incompressibility is a fundamental condition in most fluid models. Accumulation of simulation errors violates it and causes volume loss. Past work suggested correction methods to battle it. These methods, however, are imperfect and in some cases inadequate. We present a method for fluid simulation that strictly enforces incompressibility based on a grid-related definition of discrete incompressibility. We formulate a linear programming (LP) problem that bounds the number of particles that end up in each grid cell. A variant of the band method is offered for acceleration, which requires special constraints to ensure volume preservation. Further acceleration is achieved by simplifying the problem and adding a special band correction step that is formulated as a minimum-cost flow problem (MCFP). We also address coupling with solids in our framework and demonstrate advantages over prior work.