Strategic Partnership in Computational Science for Advanced Simulation and Modelling of Engineering Systems (ASiMoV)

ASiMoV poses a computational challenge that requires close collaboration between domain experts from industry and academic experts in supercomputing, computational modelling and numerical methods. ASiMoV is jointly led by EPCC and Rolls-Royce, and includes the Universities of Bristol, Cambridge, Oxford and Warwick.

In order to achieves it goals, ASiMoV requires breakthroughs across the simulation domain and uniquely combines fundamental engineering and computational science research to address a challenge that is well beyond the capabilities of today’s state-of-the-art. ASiMoV simulations couple computational fluid dynamics, combustion and structural models, thus creating highly complex multi-scale and multi-physics Digital Twins of aero-engines with trillions of cells. A result of this complexity is the need for extreme-scale computing, and thus for highly-scalable software.

The ultimate, long-term goal of the research undertaken as part of ASiMoV is to enable the “virtual certification” of aero-engines, which would reduce both the costs and the timescales of the certification process. However, although the necessary simulation capability is essential, so is the evidential basis for trusting the simulation. In order to form part of a trusted certification process, simulation results must be accurate, verified and fully validated against experimental results. The project will demonstrate the need for techniques that can exploit future computing platforms and the unprecedented amounts of data they consume and produce, robustly, securely and affordably.

EPCC’s role in the project is focused on extreme application scaling and performance, code coupling and code correctness:

• Performance analysis and optimisation of the CFD and structural mechanics applications Hydra and FEniCS-X, in particular on ARCHER2;
• Development of a new scalable and extensible combustion suite, called ASiMoV-CCS;
• Coupling of Hydra CFD and FEniCS-X structural mechanics simulations;
• Development of the C++ library Morpheus to transparently switch sparse matrix storage formats;
• Investigation of I/O performance in large-scale engineering applications;
• Development of the OFT overflow checking tool to detect scale-dependent overflow bugs.

Image courtesy Rolls-Royce.