Simulation of SCAlar by Large Eddy Simulation: development and application – SCALES

Methodological development for turbulent mixing is the main goal of this project. Two main ways are investigated. First, to be able to perform accurate simulations for real flow configurations, new sub grid-scale models are proposed based on the optimal estimator theory. Moreover, original numerical technics are developed to allow us to perform direct numerical simulations at very high resolution.

For the two followed ways, first results are very promising. We first demonstrated the ability to use «optimal estimator« as subgrid-scale models for large-eddy simulation. Moreover, the proposed «Eulerian/Lagrangian« bybrid numerical approach allowed us to perform simulations for mixing configurations never performed before, with more than 28 billions of grid points.

The next steps will be now to use optimal estimator to improve subgrid-scale models by taking into account the stability properties of the models. Moreover, the hybrid approach will be developed in a more generalist code to be able to perform simulations on a wider range of flow configurations.

Peer review journal : 1
Book chapter : 2
Conference : 3

In many industrial applications related to fluid dynamics, the chief point is the prediction the scalar field dynamics in a turbulent flow e.g. the scalar field is used to describe the temperature in thermohydraulic problems and it can also describe the concentration of chemical species with applications in combustion or air quality. Morever, it can be used as the level-set function to capture the interface in multiphase flows. The goal of the proposed research is to study the scalar dynamics in turbulent flows in order to develop accurate modeling procedures to predict the scalar behavior. The project is split into four tasks. Each task focuses on a specific objective, which is to develop accurate methodologies for numerical simulations of scalar transport on a broad range of industrial and geophysical applications.

To this purpose we plan to develop a new subgrid-scale model to perform accurate large-eddy simulations (LES) with scalar transport. The large-eddy simulation approach allows to simulate industrial flow configurations with moderate computational cost. This approach proposes to compute explicitly the largest scales of motion only and to model the small scales behavior with a subgrid-scale model. In this project, we aim to develop accurate subgrid-scale for scalars taking into account the molecular properties of the scalar. Indeed, only few models take into account the influence of these properties. Moreover, the model dependency with these properties needs to be elucidated. The model development will use a systematic procedure of model development. This new procedure will be determined during the project. It will be based on the optimal estimation theory, which allows to define, a priori, an optimal model. Optimization and error reductions techniques will be thus used to reduce the error model toward the optimal error. This new procedure will be useful to extend the application domain of LES by defining new subgrid-scale quantities. In this project, a systematic procedure of model development will be also used to model the subgrid-scale scalar variance and the subgrid-scale scalar dissipation. Both quantities are needed for Large-Eddy Simulations of turbulent combustion.

As another approach, we will combine grid-based solver (Eulerian approach) for the velocity field with particle methods (Lagrangian) to compute the scalar field. This technique will allow obtaining a finer discretisation of the scalar field than the velocity field. Whereas the velocity field computation will be done by LES. This will allow an explicit computation of the entire scalar scales for a broad range of applications. Finally, we plan to develop subgrid-scale models to be applied in the context of particle methods. The goal will to simulate turbulent scalar transport without diffusivity in highly turbulent flows. In this case, the number of particles needed is theoretically infinite and thus a subgrid-scale model has ti be developed. This is the case when the scalar is a level-set function in multiphase flow simulations.