Evolving Sub-Grid Turbulence for Smoke Animation

Abstract

We introduce a simple turbulence model for smoke animation, qualitatively capturing the transport, diffusion, and spectral cascade of turbulent energy unresolved on a typical simulation grid. We track the mean kinetic energy per octave of turbulence in each grid cell, and a novel 'net rotation' variable for modeling the self-advection of turbulent eddies. These additions to a standard fluid solver drive a procedural post-process, layering plausible dynamically evolving turbulent details on top of the large-scale simulated motion. Finally, to make the most of the simulation grid before jumping to procedural sub-grid models, we propose a new multistep predictor to alleviate the nonphysical dissipation of angular momentum in standard graphics fluid solvers.