High performance computing methodologies for solving complex turbulent flows

点击翻译按钮获取中文摘要

Computational fluid dynamics (CFD) methods are commonly used as tools for solving industrial and environmental engineering problems. As most of these problems are turbulent, the realistic simulation of the turbulence effect is essential. The Reynolds-Averaged Navier-Stokes (RANS)-based approach has become the workhorse for calculating turbulent flows, especially in hydraulics and aerodynamics. However, RANS models have been found to be not sufficiently accurate for complex flows with large separations and where unsteady effects are of interest. The increase in computer power and the development of more accurate and efficient numerical methods have led to the development of large eddy simulation (LES) methodology. LES has been found superior to RANS, but is too expensive to be deployed in real-scale engineering problems. A more affordable approach combines RANS and LES’ different strengths; one can use LES only where needed because the flow is too complex, while in other parts of the domain, such as near walls, the flow is computed with the less expensive RANS model. This has become known as the hybrid RANS-LES approach, which is an appealing methodology for modern CFD practice. This research program is focused on the development of a high performance flow solver based on a hybrid LES-RANS approach and on the investigation of modern numerical methods to allow high accuracy and computational performance. More specifically, we will investigate overset-grid and higher-order discretizations methods to achieve the high accuracy required by turbulent flows. The applications are geared towards solving the complex flow problems encountered in hydraulics and in aerodynamics engineering, such as free surface flows over the spillways or the penstocks of dams, and compressible flows around aircraft structures.