Executive Summary : | The affordability of modern supercomputers has significantly impacted the design and analysis of industrial components, with numerical simulations replacing costly wind tunnel or experimental testing procedures. Computational fluid dynamics simulations are now essential for conceptual and preliminary design and validation of certification in various sectors. Two-equation steady RANS models remain the high-fidelity method for replicating turbulent flows in industrial settings, but their shortcomings are well-known, particularly in turbulent separated flows. Large-eddy simulations (LES) are considered the future of reliably modeling engineering applications, as they resolve energy-containing large scale eddies and include the effect of small scales using a heuristic model. Advancements in CFD algorithms, particularly GPUs, are necessary to make LES a practical tool. The recently proposed single time-stepping artificial compressibility methods can be used as a building block for efficient algorithms. This project aims to develop an open-source, efficient, and high-accurate LES solver to simulate high Reynolds number turbulent flows. The solver will use upwind and central compact schemes to discretize convective and diffusive terms on a staggered grid, with low-storage Runge-Kutta schemes applied for time integration. Residual stresses in LES will be modelled using the dynamic version of Vreman's model with global coefficients. The solver will be thoroughly validated for attached and separated boundary layer flows. |