Validation of RANS Turbulence Models for Subsonic Turbulent Boundary Layer Flows Under Families of Pressure Gradients

Kurzfassung

The present thesis aims to provide a validation database of existing RANS models for high Reynolds number flows with history effects due to streamwise changing mild pressure gradients, and to assess the predictive accuracy and its uncertainty of each RANS model. For this purpose, two recent wind tunnel experiments with mild pressure gradients were selected, and the two-dimensional computational setups for RANS simulations were defined. The first test case is the Virginia Tech wind tunnel experiment conducted in the framework of the North Atlantic Treaty Organization (NATO) Science & Technology Organization (STO) Air Vehicles Technology (AVT) 349 project (Fritsch et al. (2022)). The second test case considered is the Laboratoire de Mécanique de Lille (LML) wind tunnel experiment by Cuvier et al. (2017). Both flow cases encompass streamwise changing mild pressure gradients, which result in the non-equilibrium effects and history effects. For both test cases, the influence of changes in the computational setup and their sensitivities were investigated. The RANS simulations were carried out using three turbulence models, the SA-neg, the Menter SST, and the SSG/LRR-omega models and the results were compared with the experimental data. The experimental and RANS results are in good agreement in regions where the flow is in near equilibrium. Some appreciable discrepancies are observed in the region where the flow is at APG and in non-equilibrium. However, the deviations are relatively small due to mild pressure gradients, which is a further reason to demand highly accurate measurement data. The largest deviations of the RANS predictions from the experimental data are found in the regions of the largest non-equilibrium, i.e., in the regions of the largest streamwise changes in the local pressure gradient. Also, the greater discrepancy is observed in the LML test case than the Virginia Tech test case, which is most likely due to the larger pressure gradient coefficient. These observations might suggest that the existing RANS models, that are calibrated for equilibrium boundary layer flows, fail to capture all details of non-equilibrium effects. Comparisons of different RANS models reveal a fairly less model dependency on both test cases than might be expected. This could be explained by the fact that the SSG/LRR-omega model uses the same length-scale equation as the Menter SST model, and all models are mainly calibrated for equilibrium turbulent boundary layer flows. Also, comparisons of different numerical set-ups demonstrate the difficulty of reproducing a 3D experimental set-up in 2D computations.