Skip to main content

Grid-Adaptation for Large Eddy Simulations: A Sequence of Ideas over 10+ Years

Speaker: Johan Larsson, University of Maryland, College Park

Presentation

Seminar Slide Deck (PDF-4.8MB)

Abstract

While grid-adaptation has reached a certain level of maturity in several areas of CFD, the application to turbulence-resolving simulations (most notably, large eddy simulations, or LES) is still in its infancy. The chaotic and broadband nature of the dynamics in an LES leads to 2 major challenges. First, the grid affects both the numerical and modeling error in LES, compared to only the numerical error in non-broadband problems. As a result, the estimation of the local error production cannot be purely mathematical but must also require physics-informed reasoning and assumptions. Secondly, the chaotic nature causes adjoints to diverge exponentially, which makes direct application of the adjoint-weighted residual method cumbersome.

The talk will discuss some progress made on this problem over the last 5-10 years in the PI's group, along with outstanding issues and ongoing/future work.

Biography

Photo of Johan Larsson
Johan Larsson is a Professor at the University of Maryland where he works on multiple problems in the field of computational turbulence including wall-modeling for large eddy simulation, grid-adaptation for turbulence-resolving simulations, high-speed turbulent flows, and uncertainty estimation for turbulence problems. He earned his undergraduate degree in Sweden and worked at Volvo Car Corporation before moving to Canada and earning his Ph.D. at the University of Waterloo in 2006. He then worked at the Center for Turbulence Research as a postdoctoral fellow and Research Associate for 6 years before joining the University of Maryland in 2012.