Interface learning of multiphysics and multiscale systems
This work addresses the problem of efficiently simulating complex systems with multiple scales and physics for researchers and engineers in computational science, though it appears incremental as it builds on existing domain decomposition and physics-informed learning concepts.
The paper tackles the challenge of modeling multiphysics and multiscale systems by introducing an interface learning paradigm with data-driven closure using memory embedding to provide physically correct boundary conditions at interfaces, demonstrating its promise on canonical illustrative problems. It also highlights potential benefits for high-performance computing environments to reduce communication costs in heterogeneous platforms.
Complex natural or engineered systems comprise multiple characteristic scales, multiple spatiotemporal domains, and even multiple physical closure laws. To address such challenges, we introduce an interface learning paradigm and put forth a data-driven closure approach based on memory embedding to provide physically correct boundary conditions at the interface. To enable the interface learning for hyperbolic systems by considering the domain of influence and wave structures into account, we put forth the concept of upwind learning towards a physics-informed domain decomposition. The promise of the proposed approach is shown for a set of canonical illustrative problems. We highlight that high-performance computing environments can benefit from this methodology to reduce communication costs among processing units in emerging machine learning ready heterogeneous platforms toward exascale era.