Physics-inspired Coarsening for Turbulent Flow Simulations

点击翻译按钮获取中文摘要

Fluid turbulence constitutes a key physical phenomenon with vital implications in a wide range of engineering and scientific fields. However, turbulent flows are notoriously costly to simulate and predict due to their essentially nonlinear, chaotic behavior involving a wide range of coupled length and time scales. An increasingly attractive method is to compute only the dynamics of large-scale motions with models to represent the net effect of unresolved features, an approach known as Large-Eddy Simulation (LES). The underlying theory of LES based on spatial filtering, however, suffers from several shortcomings, including the lack of an accurate, consistent treatment of nonuniform grid resolution, boundaries, and interfaces. This proposal introduces the idea of physics-inspired coarsening as a potentially transformative concept for addressing these challenges. The main objective of the project is to develop the physics-inspired framework for turbulent flow simulations and to test it on representative flows sharing the essential characteristics with many turbulent flows of interest. As such, the proposed work aims to develop game-changing simulation technologies relevant to applications including oceanography, atmospheric science, renewable energy, aerospace, propulsion, astrophysics, and even human health. The project will include the development of a new undergraduate course in numerical methods and a fully integrated undergraduate research project to serve as on-ramps to computational science and engineering research at UC Irvine, a Hispanic Serving Institution.The essential hypothesis of the physics-inspired approach is that low-cost digital representations of turbulent flows can be generated in a particularly advantageous way by leveraging physical processes that limit dynamical complexity in nature. Of foremost importance for turbulent flows, this means mimicking the physics of viscosity to develop a coarsening procedure. The physics-inspired and spatial filtering approaches are mathematically equivalent in the very simplest of cases, where spatial filtering has proven most effective. For more complex, realistic cases such as flow simulations with non-uniform grid resolution or turbulent flow modeling near a solid surface, the physics-inspired method provides a novel modeling framework that avoids the shortcomings associated with spatial filtering. The project will perform Direct Numerical Simulations together with LES to rigorously test new models in a range of flow conditions including isotropic turbulence, wall-free shear flows, and wall-bounded flows. Evaluations will include both direct testing of model accuracy prior to implementation as well as integrated testing of model performance within LES.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.