Past Event: CSEM Student Forum

A Dynamic Flux Surrogate Approach for Interface Couplings of Full-Order Models and Data-Driven Surrogate Models

Edward Huynh, CSEM Student

Abstract

Partitioned methods for coupled problems rely on data transfers between subdomains to synchronize the subdomain equations and enable their independent solution. These methods enable code reuse, increase concurrency, and provide a convenient framework for plug-and-play multiphysics simulations. However, accuracy and stability of partitioned methods depends crucially on the type of information exchanged between the subproblems. Data-driven system identification methods allow the construction of accurate, computationally efficient and minimally intrusive data transfer surrogates. This approach leaves the main computational burden to an offline phase and consequently the application of this approach as the sole additional cost during the online simulation phase. In this presentation, we formulate and couple together fully data-driven (DMD-DMD) and hybrid (FOM-DMD) solvers for a model advection-diffusion transmission problem by using Dynamic Mode Decomposition (DMD) to learn the dynamics of both the subdomain states and the interface flux from data. Our results demonstrate that these couplings can be accomplished accurately at a cheaper computational cost.

Biography

Edward (also goes by Ed) was born in Orange County but grew up in the jewel city of the desert known as Las Vegas. He previously received his BS in Math and BA in English Literature from the University of Nevada, Las Vegas. After that, he obtained an MS in Applied Mathematics from the University of Arizona. In Fall 2024, he joined the Oden Institute as a PhD student in CSEM. His current research interests are in optimal transport, partial differential equations, and optimization. In his free time, Ed likes to play table tennis, play video games, and play music on his guitar. He also enjoys good food and drinks and spends time either cooking or finding restaurants and bars.