Efficient Solution Strategies for Multiscale Flow Problems - Combining Adaptive High Order Discretizations, Models and Data

Explore advanced computational fluid dynamics techniques in this seminar presentation from the FEM@LLNL series, where Andrea Beck from the University of Stuttgart discusses efficient solution strategies for multiscale flow problems. Discover how to combine adaptive high-order discretizations, models, and data to tackle complex multiphysics phenomena that span wide ranges of temporal and spatial scales. Learn about highly accurate and locally adaptive numerical schemes based on discontinuous Galerkin spectral element methods (DGSEM) for solving compressible Navier-Stokes equations on pre-exascale computing systems, with specific applications in aerospace engineering. Examine the challenges of simulating turbulent flows and flows with phase interfaces, where different physical processes dominate at opposing ends of the scale range. Understand how scale-bridging models supplement upscaled governing equations to enable practical engineering simulations that would otherwise be computationally prohibitive. Investigate the role of machine learning methods in developing numerically and mathematically consistent scale-bridging models, including successful examples of CFD/ML hybrid applications. Gain insights into adaptive discretization strategies that tailor approximation schemes to locally dominating physical effects, essential for efficient simulation of multiscale phenomena on modern high-performance computing architectures.