Extreme-scale high-fidelity turbulence simulations of convection and boundary layers using accelerators

Turbulence is one fundamental problem of classical physics that has eluded a satisfactory description by engineering models.

The main reason is the high-dimensional chaotic nature of turbulence. Practical predictions of turbulent flows thus rely extensively on novel theories, experiments and numerical simulations.

Using modern simulation approaches, two fundamental questions will be answered in this project: In Rayleigh-Bénard convection, the question of the transition to the ultimate regime is addressed: Do the thin boundary layers become turbulent, and thus significantly increase the turbulent heat transfer at Rayleigh numbers as high as Ra~10^15 and beyond? Which effect do wing-tip vortices, formed at the wing end with a turbulent boundary layer, have onto the development of the boundary layer, the separation, and the changes of lift and drag of the whole lifting surface?

A better understanding will allow for improved wing-tip design leading to reduction in both the lift-induced drag and parasitic drag, as well as designs that are more resilient to flow separation. In both cases, detailed direct numerical simulations, resolving all turbulence details, are needed to predict all relevant interactions.

Numerically, we rely on a novel spectral-element code that delivers both accuracy and high speed on modern CPUs and GPUs.