Simulation aux grandes échelles des transferts thermiques dans les échangeurs de chaleur en fabrication additive

In aeronautical and aerospace engineering, heat exchangers have a fundamental role. Recent progress in additive manufacturing is a great opportunity for innovation aiming at compact heat exchangers to reach new performances. Pressure loss and heat transfer performances are the two main characteristics to be optimized for heat exchanger efficiency. However, the significant roughness introduced by additive manufacturing strongly impacts these performances. In addition, a gap still exists between the computationally optimized structures and the industrial reality due to the lack of fundamental knowledge of the new additive manufacturing structures. Consequently, current wall models used in steady and unsteady three-dimensional Navier-Stokes simulations do not take into account typical roughness induced by additive manufacturing. The H2020 CleanSky2 STREAM project was dedicated to the performance improvement of the new generation of heat exchangers, taking advantage of additive manufacturing, topological optimization, and high-fidelity simulations. Within this project, this PhD's aim was to generate a large high-fidelity database of roughness-resolved Large-Eddy Simulations (RRLES) with the YALES2 CFD platform. This database was for the development and assessment of novel wall models in collaboration with the LEGI laboratory. To reach this aim, three steps were necessary beforehand. The first achievement done during this PhD was the development of a rough surfaces generator. This tool is able to generate multiple rough surfaces mimicking those encountered in additive manufacturing. Among available geometries with this generator, there are parallel planes, square and cylindrical channels, and more industrial configurations like plates with tube fins. Test cases have highlighted good performances for accuracy and computational time. Obtaining body-fitted meshes was the second step. A roughness-resolved mesh generator, developed at the CORIA laboratory, enables to provide automatically a large number of body-fitted unstructured meshes with control of the cell size distribution of the final mesh. The numerical setup and the methodology for RRLES have been set and validated. For this purpose, a recycling method has been developed. The latter's principle is to interpolate velocity at a distant plane in the computational domain and to impose this velocity at the inlet. This method allows to perform periodic channel flow. In addition, a fully automated workflow with integrated post-processing has been built and has enabled to run numerous simulations on a remote supercomputer. Among the database's cases, three configurations have been particularly studied in order to highlight the impact of additive manufacturing printing direction on the flow topology, velocity and temperature profiles. RRLES simulations have also been conducted for square and cylindrical channels, and for industrial-like applications like plates with tube fins. Finally, the database has been analyzed and some rough wall laws have been derived. The work done during this PhD has led to a better understanding of the impact of roughness induced by additive manufacturing. This led to propose modeling strategies for the industrial partners of the STREAM project and paves the way for new efficient heat exchangers.