(605c) Physics-Based Modeling of Free Surface Dynamics Using Graphics Processing Units (GPUs)

Free-surface and two-phase systems are abundant in pharmaceutical equipment and processes, including vortex formation in agitated systems, nozzle sprays, tank filling, sloshing, spray coating atomization. Processes involving free surfaces are almost always time-varying, and involve transport processes that simultaneously span multiple length and time scales. This disparity in length and time scales makes first-principles modeling of free-surface and two-phase systems difficult.

In this work, we show how the Navier-Stokes equations with free surface boundary conditions can be solved efficiently on graphical processing units. We begin by introducing the concepts governing GPU-performance, as applied lattice-Boltzmann and free surface modeling. We then show how, given identical CFD algorithms, a single scientific GPU can execute simulations 100x faster than a single CPU. We then validate this approach by comparing predictions from simulations involving agitated tanks, shaker flasks, and sloshing tanks.