Soutenance de thèse Daniel Walter Lagrava Sandoval

M. Daniel Walter Lagrava Sandoval soutiendra, en vue de l'obtention du grade de docteur ès sciences, mention informatique, sa thèse intitulée

Revisiting grid refinement algorithms for the lattice Boltzmann method

Abstract :

Fluid motion is an intrinsically multi-scale phenomenon. For instance, in the turbulent case, the range of excited scales of motion spans over many orders of magnitude. This becomes of great concern when trying to simulate fluids numerically. One solution to this problem is the usage of non-uniform grids or grid refinement. The main idea is to keep the computational time reasonable, while simulating all or most of the scales present in the problem. In particular, in this thesis, we are interested in grid refinement for the lattice Boltzmann method.

Although the topic has been already studied by the community, we are convinced that there is room for improvements, both in the theoretical and practical explanations provided so far.

We present thoroughly all the theory of the grid refinement for the lattice Boltzmann. As an interpolation operation is necessary in the selected approach, the minimal order of interpolation is explained. Further, we introduce a new filtering operation that is mandatory to study turbulent flows. Our generic implementation, which was developed on top of the Palabos open-source library, is explained in detail. Our approaches are tested on several two and three-dimensional benchmarks. We have specially chosen challenging cases, where, among others, large gradients cross the grid-refinement frontiers. Finally, we end with two other subjects related to grid refinement, namely a grid refinement algorithm that does not use an overlapping zone between the grids and a criterion to choose zones where local refinement is necessary.

Date: Mardi 18 décembre 2012 à 14h00

Lieu: Battelle bât. A - Auditoire rez-de-chaussée

13 décembre 2012

À la Une