Examples of research activities in physical modeling with Trio_U
 Large-eddy simulation (LES) in single-phase flows: issues and research topics

The large-eddy simulation approach (LES or SGE in French) proposes an intermediate way between averaged approaches RANS and direct simulation DNS.

Unlike the RANS approach, LES gives access to fluctuating quantities without requiring the resolution of all spatio-temporal scales, as for DNS; the effect of small non-resolved scaled (“filtered” scales) on the resolved ones have to be modeled.

Since modeling is based on the assumption of isotropy for the smallest resolved scales (Smagorinsky model, WALE, structure-function ...), this should require the use of very refined meshes.

In industrial applications and despite the use of multi-million element meshes, this last criterion cannot be fulfiled near walls where flows are locally strongly anisotropic.

As a consequence, at the walls, where turbulence is produced and small coherent structures called streaks are generated (see figure), the mesh-cells are much larger than these structures.

The “universal” character of LES models being inappropriate to correctly resolve the flow in the near-wall region in such conditions, a specific wall-modeling has to be employed to reproduce the local effect of the wall-turbulence on the rest of the flow: this specific modelling is called wall functions.

Wall functions in single-phase LES

 Research in numerical simulation of flows with interfaces

At the smallest scale, a two-phase flow is made of phases separated by moving interfaces. The direct numerical simulation of such flows consists in simulating all the scales in time and space of the flow. In particular, all the inclusions (bubbles or droplets) of the flow are accounted for individually. In contrast to the numerical simulation of single-phase flows, the main difficulty deals with the numerical tracking of the moving interfaces. Several numerical methods are dedicated to such a numerical tracking. Among them, two are more particularly studied and developed within the Trio_U project: the diffuse-interface method and the « Front-Tracking » method. Whatever the method, the goal is to have at our disposal a method that is accurate and efficient in the sense that it must allow to simulate complex flows with several inclusions.

This constraint is related to our goal to get information that is of industrial interest. Indeed, the development of direct numerical simulation methods is part of a long-term multi-scale modeling approach of two-phase flows. The only two-phase flow models that can be used to study industrial applications are averaged models (in space, in time or statistically). Like any other averaged model, they must be closed. The determination of these closures is very important, in particular in complex physical situations such as boiling at high heat flux. The direct numerical simulation is a tool that is particularly attractive in this regard because it allows to have access to all the local information necessary to the validation or development of closure relations of the averaged models.

In this perspective, the main research topics consist in developing efficient numerical methods and, thanks to them, studying and modeling, on the one hand, the interactions between interfaces and turbulence and, on the other hand, flows with liquid-vapor phase-change.

Simulation of flows with interfaces

TRIO U - Last update : 15/09/2009 18:35:15 - Legal Notice
© CEA 2003-2017 All Rights Reserved