Skip to main content

High-Order Adaptive Multistep Code Coupling

Speaker: Laurent François, ONERA

Presentation

Seminar Slide Deck (PDF-2.9MB)

Abstract

Multiphysics simulations have become a key ingredient in the study and design of complex systems. For such simulations, it is usually desirable to rely on code heritage and design a coupling strategy that allows the reuse of existing model-specific solvers with minimal modifications. The main issue that has been previously tackled is that of spatial interpolation between different meshes, which enables the exchange of coupling data (e.g., heat flux, source terms...) between the solvers. However, most coupled frameworks rely on low-order explicit coupling schemes in time, where the coupling variables are frozen during each coupling step, and the coupling time step is constant or adjusted heuristically, without any guarantee in terms of accuracy and stability.

This talk will present a novel temporal coupling strategy, the multistep coupling scheme, which aims at providing high-order accuracy in time for the coupling terms, as well as a dynamic error-based adaptation of the coupling time step. The theoretical foundation of the scheme will be presented. Numerical experiments on a conjugate heat transfer problem will demonstrate the adequacy of the scheme, and an application to the study of the heatshield response during a probe entry into the atmosphere of Uranus will be presented. Finally, the ongoing implementation of the strategy within the open-source CWIPI HPC coupling library developed by ONERA will be discussed, with the goal of providing a complete space-time code coupling library.

Biography

Photo of Laurent François, ONERA
Laurent François graduated in aerospace engineering at ISAE Supaéro (France) and obtained his Ph.D. in applied mathematics at ONERA and École Polytechnique on the topic of modeling and numerical methods for the simulation of solid rocket motor ignition. He now is a research scientist at ONERA, with a main focus on solver coupling and numerical methods for multiscale PDEs involving applications in high-speed reactive flows and solid propellant combustion.