Data-driven multi-physics and multi-scale modelling

Creating a “digital” product

Research objective

The ability to perform high-fidelity simulations of realistic engineering problems is an essential component of creating a “digital” product. A fundamental prerequisite is enabling a variety of software packages, from atomistic applications to engineering-scale systems, to communicate with each other efficiently. Modelling for many engineering systems will therefore require the coupling of solvers from a range of different physics domains. For example, fluid flow may interact with structures such as a sailing vessel, which is affected by both water and air, or an oil reservoir model may include multi-phase fluid flow interacting with the chemistry of rock. Even for a single domain, multiple interacting models with different length or time scales and different approximations may be required to capture all essential phenomena.

The overall objective of our research is to develop multi-physics and multi-scale modelling methods applicable to target industries which make effective use of IBM’s Data Centric Systems architectures to model large, complex systems in a reasonable time. A companion objective is co-design of future HPC architectures to better match these real-world applications.

Publications

[1] D.E. Keyes, et al.,
Multiphysics simulations: Challenges and opportunities,“
International Journal of High Performance Computing Applications 27(1), 4-83, 2013.

[2] D. Groen, S. Zasada, P. Coveney,
Survey of Multiscale and Multiphysics Applications and Communities,“
Computing in Science & Engineering 16, 2012).

[3] T.-R. Teschner, L. Könözsy, K.W. Jenkins,
Progress in particle-based multiscale and hybrid methods for flow,”
Microfluid Nanofluid 20, 2016.

Ask the experts

Robert Sawko

Robert Sawko

Małgorzata Zimoń

Małgorzata Zimoń

Chris Thompson

Chris Thompson

John Magerlein

John Magerlein
IBM T.J. Watson Research Center