The scientific areas covered by this proposal merge multiphysics and multiscale modelling together with High Performance Computing (HPC). The areas covered by multiphysics modelling will start with Conjugate Heat Transfer (CHT), where traditional Computational Fluid Dynamics (CFD) is coupled to solid heat diffusion. Fluid Structure Interaction (FSI) will also be dealt with to show how to couple fluid and structural mechanics to study solid structure deformations. The multiscale aspect will be tackled by looking at flows under non-equilibrum (micro-scale) where extra modelling, rather than the traditional Navier-Stokes-Fourier equations, is required to account for the rarefied nature of the gas.

Turbulence modelling will be also covered, by looking at advanced modelling through Large-Eddy Simulation (LES) and different types of Reynolds Averaged Navier Stokes (RANS) models. LES is a technique which relies on filtering the Navier-Stokes equations, that was first introduced by Smagorinsky in 1963 [1]. It has become much more popular recently because of the computing power increase, even if still not routinely used by industry, for instance. Vast computing power is required because modelling is only carried out at the sub-grid scale level. This aspect of turbulence modelling will be coupled with heat diffusion in a solid for a CHT case of a heated cross flow inside a rod bundle, which corresponds to a typical application for heat exchangers in thermal power generation.  

Advanced RANS modelling, which is able to capture turbulent flow anisotropy, and is based on second order moment closure will be used to study the problem of a cylinder oscillating inside a rod bundle. The methodology will use the Reynolds Stress Model of Speziale, Sarkar and Gatski (SSG) [2] within a FSI coupling framework based on a Quasi-Newton method [3].

Rarefied gas flows and their requirements in terms of modelling will be introduced. The Regularised 26 Moment Method [4], which has proven to be able to properly model the flow when the traditional Navier-Stokes-Fourier equations fail (Knudsen number between 0.1 and 1, e.g. corresponding to early/mid-transition regime) will be used.

The School will insist on both interlinked aspects between advanced multiphysics and multiscale modelling, and HPC, showing that advanced modelling for some cutting edge applications related to power generation cannot be carried out without HPC capabilities. The path for developing/optimising a massively parallel multiphysics and multiscale+HPC software will also be introduced [5].

REFERENCES

[1] J. Smagorinsky, Mon. Wea. Rev., 91, 99-164 (1963).
[2] C. Speziale, S. Sarkar, T. Gatski, J. Fluid Mech., 227, 245-272 (1991).
[3] K. Scheufele. “Robust Quasi-Newton methods for partitioned fluid-structure simulations”. Master’s thesis. 2015.
[4] X. Gu, D. Emerson, J. Fluid Mech., 636, 177-216 (2009).
[5] Y. Fournier, J. Bonelle, C. Moulinec, Z. Shang, A. Sunderland, J. Uribe, Computers & Fluids, 45, 103-108 (2011).