Parallelizing Solid Particles in Lattice-Boltzmann Fluid Dynamics

Computational fluid dynamics (CFD) is a widely used numerical method for simulating complex systems, with applications ranging from aerodynamic science to hydrology. The latticeBoltzmann method is a relatively recent technique that has been shown to be as accurate as traditional CFD methods, but with the ability to integrate arbitrarily complex geometries at a reduced computational cost. Although fluid flow in the lattice-Boltzmann method is easily parallelized, this is not true of the specific problem of modeling solid particles suspended in a fluid. The data structures representing solid particles and the interactions required to communicate information about them constitute time-consuming and serial portions of the simulation. In this paper, we describe a parallel implementation of the lattice Boltzmann method that incorporates new data structures and a modification to the basic algorithm. The goal of these modifications is to enable strictly local, hence parallel, computations on solid particles and reduced communication overhead.