Master thesis and internship[BR]- Master's thesis : Wall functions on CFD RANS calculations in high CPU resource configurations applied to boosters[BR]- Integration internship

Abstract

The application of wall laws in constrained environments such as low-pressure compressors remains a long-standing challenge. This work investigates the applicability of the standard wall law on both a stator and a rotor, using the k-ϵ turbulence model. The study begins with a 2.75D analysis on a midspan section of the stator, only for a reference inlet angle aimed at rapidly identifying representative y+ distributions. The stator case was first tested at a Reynolds number of Re = 3 × 10^5, but this value proved insufficient to ensure a fully turbulent boundary layer from the inlet. A higher Reynolds number, Re = 1.2 × 10^6, was therefore adopted. For the rotor, simulations were performed at Re = 3 × 10^5. Preliminary results revealed a general loss of accuracy as the first-cell height increased, though performance improved again when approaching the end of the logarithmic layer of the inner boundary layer. From this parametric analysis, two representative y+ distributions were selected for the subsequent three-dimensional study. In 3D, three inlet angles are considered for the stator : the reference angle previously mentionned and a deviations of plus and minus 5°. For the rotor, only a reference incidence angle is considered. In the case of the stator, wall laws encounter the greatest difficulties when dealing with the positive deviation with respect to the reference inlet angle. Time savings of around 40% for the stator and more than 50% for the rotor were achieved, although the associated loss of accuracy often offsets these benefits. Overall, The results remain relatively unconvincing, as significant discrepancies were observed; however, an ability to mention of wall laws to capture shocks, even if they mispredict their intensity and location. In addition, the mesh used inherently limited the applicability of wall laws. With the Autogrid5 topology, where cell size can only increase away from the wall, the refinement strategies recommended in the litterature could not be followed. As a result, careful calibration of the first-cell size and aspect ratio was required, underlining that the choice of mesh strongly conditions the accuracy of wall-law approaches. Finally, a key subtlety is that the notation (average of) y+ merely refers to a mean value along the chord, whereas it is the full distribution range of y+ that determines validity. Therefore, relying on the representative mean of y+ alone can be misleading, and attention should be given to the underlying y+ distributions when assessing wall-law performance.