Detailed analysis of the lattice Boltzmann method on unstructured grids

TL;DR

This paper analyzes two implementations of the lattice Boltzmann method on unstructured grids, deriving their accuracy, viscosity relations, and proposing improvements, with validation on benchmarks and porous rock samples.

Contribution

It provides a detailed theoretical and practical analysis of lattice Boltzmann methods on unstructured grids, including derivations, accuracy proofs, and boundary condition improvements.

Findings

Both methods are first order accurate in time and second order in space.

Derived relations between kinetic viscosity and time step for both methods.

Validated implementations on benchmark geometries and porous rock samples.

Abstract

The lattice Boltzmann method has become a standard for efficiently solving problems in fluid dynamics. While unstructured grids allow for a more efficient geometrical representation of complex boundaries, the lattice Boltzmann methods is often implemented using regular grids. Here we analyze two implementations of the lattice Boltzmann method on unstructured grids, the standard forward Euler method and the operator splitting method. We derive the evolution of the macroscopic variables by means of the Chapman-Enskog expansion, and we prove that it yields the Navier-Stokes equation and is first order accurate in terms of the temporal discretization and second order in terms of the spatial discretization. Relations between the kinetic viscosity and the integration time step are derived for both the Euler method and the operator splitting method. Finally we suggest an improved version of the bounce-back boundary condition. We test our implementations in both standard benchmark geometries and in the pore network of a real sample of a porous rock.