Laboratoire d'accueil : Laboratoire de l'incendie et des explosions (LIE)
Date de début de thèse : octobre 2015
Nom du doctorant : Adithya Ramanathan Krishnan
Descriptif du sujet
The study of turbulent flows evolving in a stable stratified place interests a large number of safety issues (both natural and industrial). In the framework of safety studies conducted by IRSN, an important issue concerns the risk of accumulation of hydrogen in the upper part of the nuclear containment mainly from the oxidation of the zircaloy alloy under the effect of buoyancy. The risk of explosion of this gaseous mixture can be reduced if the concentration levels of hydrogen are lowered under the effect of turbulent mixing. In this context, IRSN participates in the validation of CFD codes for the prediction of erosion of stratification by a turbulent jet. The results show that standard RANS first order turbulence models are insufficient in predicting the proper turbulent mixing of the stratification. Use of LES for such large scale experiments appears prohibitive. Hence, a hybrid RANS / LES approach is carried on.
In the present study a hybrid approach based on an explicit algebraic Reynolds stress model (EARSM) will be studied and implemented wherein an algebraic relation is provided for the Reynolds stress tensor by replacing the Boussinesq eddy viscosity hypothesis. This model integrates the robustness of a two-equation eddy-viscosity model and accounts for the turbulence anisotropy effects as well. When compared to the original RSM approach, the complications in anisotropy are limited by having an algebraic relationship. To take this forward, EARSM will be implemented in the P2REMICS code in a pure RANS context and validated on well-known academic cases. Following which the hybrid approach will be introduced through the Partially Integrated Transport Model (PITM) framework.
Amount of resolved turbulence in the transition and LES region is negligible. Hence simultaneously, a study on the forcing term within the momentum equation is done to increase the amount of resolved turbulence. As a first step, validation of the pre-implemented isotropic Random Fourier Model (RFM) with the available DNS data is done for the spatially developed channel case. Further, introducing anisotropy in the synthetic fluctuations generalizes this method. After which, it will be implemented to analyze the much relevant case in a hybrid context.