Simulation of Rayleigh Benard convection at Ra = 1e13 (very high Rayleigh number) скачать в хорошем качестве

Simulation of Rayleigh Benard convection at Ra = 1e13 (very high Rayleigh number)

This video shows simulation results of compressible natural convection. The Rayleigh number is very high (Ra = 1e13) and the temperature difference is large (T_hot/T_cold = 4). The parameters for this simulation are: Ra = 1e13 T_hot/T_cold = 4 L/H = 16/9 dy_Wall = 3.3e-5 H The simulation implements an explicit Gas-Kinetic Scheme (GKS) for low Mach number flows. Gas-Kinetic Schemes are Finite-Volume methods, that derive fluxes from kinetic theory. In difference to the Lattice Boltzman method (the most famous kinetic method) GKS is directly applicable to compressible flows. The simulation is performed on a rectangular mesh with cells streched towards the top and bottom boundaries. The resolution of this simulation is 8192x8192 cells, hence the domain is covered by approximately 67 million cells. The high resolution and the cell stretching allows a good resolution of the thin thermal boundary layer and the two-dimensional turbulent structures. The left and right sides of the domain are connected (periodic boundary conditions). Top and bottom walls implement no-slip for the velocity field and Dirichlet boundary conditions for the temperature. The simulation is performed on a single (yet massively parallel) Nvidia Tesla P100 GPGPU which allows highly performant simulations of explicit schemes provided that the data structures are adapted to the architecture. In physical dimensions with the properties of air this simulation corresponds to temperatures T = 25°C at the top wall and 920°C at the bottom wall. The distance of the walls corresponds to 5.7 meters. Hence, these conditions are comparable to the thermal dynamics of a small scale building fire (if reactions and radiation effects are neglected). The numerical solver has been developed by Stephan Lenz at the Institute for Computational Modeling in Civil Engineering of TU Braunschweig. Valuable discussions and support by Prof. Manfred Krafczyk, Prof. Martin Geier, Martin Schönherr and Konstantin Kutscher are gratefully acknowledged. https://www.tu-braunschweig.de/irmb