Séminaire Calcul Scientifique et Modélisation

This work is devoted to the numerical simulation of low Mach number flows, modeled by the compressible Euler system. Commonly used solvers for discretizing this model are Godunov-type schemes. These schemes exhibit poor performance at low Mach number in terms of efficiency and accuracy.

Indeed, regarding the accuracy problem observed with quadrangular grids, it arises from the fact that the discrete solution fails to converge to the incompressible solution as the Mach number tends to zero. To overcome this accuracy problem, many fixes have been developed and consist in modifying the numerical diffusion of the original scheme. These corrections improve the accuracy of compressible schemes as the Mach number goes to zero. Unfortunately they introduce other problems, such as the appearance of numerical oscillations (checkerboard modes on a Cartesian grid) in the numerical solution, or the damping of acoustic waves as the Mach number goes to zero. Efficiency is also compromised as these schemes are stable under a more restrictive CFL condition compared to the original scheme.

In this talk, we propose to study the phenomenon of oscillations that plagues some of the fixes proposed in the literature. We focus on Roe-type fixes, in particular those that reduce the numerical diffusion on the jump of the normal velocity. The asymptotic analysis of these schemes leads to a discretization of a wave system in which the pressure gradient is centered. To better understand the phenomenon, we focus on the linear wave system. We then show that this fix is not TVD, unlike the Godunov scheme, which explains the appearance of numerical oscillations in the unsteady solution. Next, we study the long-time behavior of the numerical solution. It turns out that spurious stationary oscillations appear on the velocity field, preventing mesh convergence of the numerical solution. Moreover, the dimension of the space of spurious elements is huge, making it difficult to consider any form of filtering for these modes. We conclude this work by noting that the modification of the numerical diffusion makes it difficult to develop numerical methods that are both stable and accurate for low Mach number flows.

Je montrerai tout d'abord comment à partir de considérations énergétiques et du principe de moindre action on peut utiliser une fonction indicatrice de phase (fonction couleur) pour modéliser l'effet des forces capillaires sur la dynamique d'un film mince. Je montrerai ensuite comment construire un solveur HLLC pour discrétiser le système d'edp obtenu dans le cas où on peut négliger les effets liés à la pression capillaire (donc liés à la courbure de l'interface film-air). Je terminerai par l'état d'avancement de nos travaux dans le cas où toutes les forces capillaires sont prises en compte dans le modèle. Des résultats numériques permettront d'illustrer la présentation et de mettre en évidence ce qui marche mais aussi ce qui ne marche pas encore ... Il s'agit d'un travail commun réalisé avec B. Delacroix, G. Blanchard, M. Bouyges et C. Laurent.

On the mesoscopic level, motion of individual particles can be modeled by a kinetic transport equation for the population density f(t,x,v) as a function of time t, space x and velocity v \in V. A relaxation term on the right hand side accounts for scattering due to self-induced velocity changes and typically involves a parameter K(x,v,v') encoding the probability of changing from velocity v' to v at location x:

\partial_t f(t,x,v) + v \cdot abla f(t,x,v) = \int K(x,v,v') f(t,x,v') - K(x,v',v)f(t,x,v) dv'

This hyperbolic model is widely used to model bacterial motion, called chemotaxis.

We study the inverse parameter reconstruction problem whose aim is to recover the scattering parameter $K$ and that has to be solved when fitting the model to a real situation. We restrict ourselves to macroscopic, i.e. velocity averaged data $\rho = \int f dv$ as a basis of our reconstruction. This introduces additional difficulties, which can be overcome by the use of short time interior domain data. In this way, we can establish theoretical existence and uniqueness of the reconstruction, study its macroscopic limiting behavior and numerically conduct the inversion under suitable data generating experimental designs.

This work based on a collaboration with Kathrin Hellmuth (Würzburg, Germany), Qin Li (Madison, Wisc., USA) and Min Tang (Shanghai, China).