Data-Driven Discontinuous Galerkin FEM via Reduced Order Modeling and Domain Decomposition

Explore a cutting-edge seminar on data-driven Discontinuous Galerkin Finite Element Method (DG FEM) presented by Kevin Chung at Lawrence Livermore National Laboratory. Dive into the innovative approach of combining reduced order modeling and domain decomposition to address the challenges of scaling scientific technologies from laboratory to industry applications. Learn about the proposed scalable, component reduced order model (CROM) method that utilizes Discontinuous Galerkin Domain Decomposition (DG-DD) to break down complex physics equations into manageable components. Discover how proper orthogonal decomposition (POD) is employed to identify critical physics modes and create efficient reduced order models (ROMs). Understand the potential of this method in solving equations up to 40 times faster with minimal error, even at scales 1000 times larger than unit components. Gain insights into the application of these techniques to Poisson and Stokes flow equations, and their potential extension to more complex physics like the Navier-Stokes equation. Delve into the future implications of this research for transitioning laboratory-scale technologies to practical industrial use in this 54-minute presentation from the FEM@LLNL Seminar Series.

FEM@LLNL | Data-Driven DG FEM Via Reduced Order Modeling and Domain Decomposition