Physics-Inspired Coarsening for Large-Eddy Simulation of Turbulent Flows

Large-eddy simulation (LES) is widely used for computing high Reynolds number turbulent flows. LES is typically framed in terms of solving the spatially-filtered Navier-Stokes equation, but the use of spatial filtering for LES is not without its shortcomings. Issues include: (i) how to define filtering for wall-bounded flows, (ii) commutation errors for non-uniform filters, and (iii) extensibility to flows with additional complexity, such as multiphase flows. While it is sometimes possible to ignore these issues in practice, the emergence of data-driven modeling techniques, with the necessity of training data, amplifies the need for a robust, comprehensive coarsening framework for LES that improves upon spatial filtering. In this presentation, I will review turbulence theory relevant to LES, most notably, the kinetic energy cascade. Physics-inspired coarsening will be introduced as a new perspective that leads to novel physical insight into energy cascade dynamics. Furthermore, it provides a generalizable procedure for LES theory that can address the aforementioned weaknesses of spatial filtering. Some preliminary modeling results, including a generalizable dynamic procedure that does not require a test filter, will be shown to demonstrate the feasibility of this new approach to LES.