Flexible and Efficient Data Mapping for Simulation of Coupled Problems

• Applicant: Professor Dr. Benjamin Uekermann
• Subject Area: Computer Architecture, Embedded and Massively Parallel Systems
• Term: since 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 515015468

Multi-physics and multi-code simulations are an increasingly important topic of research today. New application insights can often no longer be extracted from single simulation models, but need the coupling and flexible interplay of more and more simulation models, which become more and more disparate. To deal with the increasing complexity, generic coupling software, which allow application scientists to easily couple existing simulation codes with each other, can be of great help.In the proposed project, we build on the open-source coupling software preCICE. preCICE is a community project with hundreds of users spread over many application fields: aerodynamics, astronautics, automotive manufacturing, wind energy, biomechanics, biomimetics, marine engineering, nuclear fusion, reactor safety, geophysical systems, and more. Originally, preCICE was developed for low-order, mesh-based, surface-coupled problems, such as fluid-structure interaction. In recent years, more and more users, however, have started to use preCICE for non-standard coupled problems, such as mesh-particle coupling, volume coupling, or coupling with higher-order discretizations. To this end, workaround implementations are used, which can work for specific setups, but have many drawbacks: they are inflexible, hard to maintain, and often computationally inefficient. The limiting factor for all these cases is the way that preCICE currently handles coupling meshes and mapping of coupling data between these meshes. The current concept is too rigid.The overarching goal of the proposed project is to develop and implement new concepts for more flexible solutions allowing solver-based data-mapping, large-scale volume data mapping, and dynamic changes of coupling meshes. To evaluate the increase in flexibility and efficiency, we consider three test setups. The first setup, fluid-structure interaction using the highly-optimized ExaDG code (building on existing deal.II-preCICE integration), involves solver-based, higher-order data mapping. The second setup studies volume data mapping problems with partition-of-unity radial-basis-function interpolation in an artificial coupling environment, but with real geometries from the applications areas mentioned above. The third setup is a highly-dynamic volume coupling between the particle codes XDEM and MercuryDPM and the CFD codes OpenFOAM and SU2.

• DFG Programme: Research Grants
• International Connection: Luxembourg, Netherlands
• Cooperation Partners: Dr. Xavier Besseron; Professor Dr.-Ing. Bernhard Peters; Professor Dr. Anthony Thornton; Professor Dr. Thomas Weinhart