A framework for assessing the key statistical parameters of complex flows in choosing appropriate turbulence prediction methods on a quantitative basis is developed. These parameters characterise flow/modelling matching conditions quantified in this work. Matching conditions are important in classifying complex turbulent flows in order to frame best practice for model predictions to inform computational aerodynamics design optimisation in the context of virtual test beds. In the incompressible low Reynolds number shear flows considered here, the boundaries of the 'conforming domain' within which turbulence models are valid need to be defined, based on basic mechanisms of turbulence, and the statistical parameters. This has led to a new guideline ‘localness map’ for standard model applications. Since the choice of turbulence model depends on the complexity of the flows considered, it is useful if systematic sets of the parameters indicate the type of flow. They are that of residence time, the degree of spatial non-locality, the straining, and the non-Gaussianity, each of which is appropriately normalised. It can be demonstrated that the quantified map, in particular that of localness for the shear flows, provides a firm foundation for evaluating a wider range of Underlying Flow Regimes, including locating the Underlying Flow Regimes on the generalised localness modeling map as a framework for best practice guidelines. This work produces 7 sets of quantitative localness-structural parameters, which are used as baseline sets for grouping the Underlying Flow Regimes, and hence it opens the possibility of having complete modelling maps for Application Challenges to assess the need for zonal modelling.

No description available.

L’étude multi-échelle d’un lit fluidisé gaz-particule est réalisée afin de comprendre les raisons de la mauvaise prédiction de la combustion dans ce type de réacteur. Dans un premier temps, des simulations numériques directes à l’échelle de quelques particules montrent le couplage entre la fraction volumique de particules et la fraction massique des espèces dans l’écoulement. Ensuite, une analyse des équations flitrées en LES montre qu’un terme de corrélation fluide-particule apparaît lorsque l’on explicite le taux de réaction gaz-particules. On comprend alors que si ce couplage n’est pas pris en compte correctement dans les simulations aux grandes échelles, le résultat ne peut pas être bon. Des simulations numériques à l’échelle du réacteur sont alors réalisées avec différents maillages pour tenter de mettre en évidence les effets de sous-maille liés à ce couplage à l’échelle des petites structures solides dans l’écoulement. / The multiscale study of a reactive gas-solid fluidized bed is performed to understand the reason of the bad prediction of the combustion in such a reactor. First, Direct Numerical Simulations at the particle array lenght scale show the dependency of the species mass fraction released in the gas phase on the solid volume fraction. Then, the analysis of the filtered continuity equations for the eulerian granular model highlights that a fluid-particle coupling term appears when expliciting the heterogeneous reaction rate. Therefore, we understand the need to take this coupling into account in large scales simulations to obtain good results. Computations at the lab-scale reactor are eventually performed using different grid refinements in order to try to highlight the subgrid terms due to this coupling at the small solid structure scale.