In this thesis, we focus on modeling mass and energy transfer associated with wall condensation in a turbulent flow of a vapor–noncondensable gas mixture. The flow is two-phase and turbulent, where forced, mixed, and natural convection modes may occur. The framework of this work relies on the RANS approach applied to the compressible Navier–Stokes equations, in which wall condensation is described using semi-analytical wall functions developed in a previous doctoral study \cite{iziquel2023}. These functions account for the different convection modes as well as suction and species interdiffusion effects, but neglect the presence of a liquid film.
In the literature, the influence of film formation and flow on mass and heat transfer is often neglected, since it is generally assumed that, in the presence of noncondensable gases, the resistance of the gaseous layer to vapor diffusion is much greater than the thermal resistance of the liquid film.
The objective of this thesis is to improve the prediction of heat and mass transfer by investigating, beyond the thermal resistance of the condensate, the dynamic effect of the liquid and its interaction with the gaseous diffusion layer during wall condensation. The study will first consider laminar film flow, and then attempt to extend the analysis to the turbulent regime.
In the gas phase, the wall-function model developed in \cite{iziquel2023} for a binary mixture of vapor and a single noncondensable gas will be extended to mixtures of vapor and $n>1$ noncondensable gases (N2, H2, …), in order to address hydrogen risk issues.
The validation of the implemented models will be carried out using results from separate-effect (SET) and coupled-effect (CET) experiments available in the literature (Huhtiniemi \cite{huhti89}, COPAIN, ISP47-MISTRA, ISP47-TOSQAN, RIVA). Comparisons at the CFD scale, using wall functions for condensation neglecting the film, will be performed on benchmark cases from the literature and condensation experiments (COPAIN) to assess the impact of this assumption as well as the improvement provided by the new model in terms of accuracy and computational cost.