Petascale Simulation of Turbulent Two-Phase Flows
In the transportation sector, most propulsion devices, due to weight and volume restrictions, rely on liquid fuels that have high energy density. Ever more stringent regulations on noise production and pollutant formation complicate the manufacturers' task to meet these constraints in a way that is cost effective in both design and operation stages. This means relying on more aggressive designs, where the underlying physics often challenge our understanding. The multi-scale, multi-phase, and multi-physics character of these flows, difficult to reproduce experimentally, has led computational fluid dynamics to play an increasingly important role in understanding the physics governing the transfers and transport between and within different media. The phenomena at play span a wide range of physics, such as turbulence, capillarity, cavitation, and combustion, often encountered in other industrial applications as well as in nature.
An abundance of these flows involves two or more media. Along the interface that separates them, transfers take place in the form of exchanges of mass, momentum, and heat. The local rates at which these transfers occur depends on the transport phenomena occurring in the vicinity of the interface. The transfers in turn affect the dynamics of the flow and other coupled phenomena (combustion, turbulence, . . . ). The non-linear character of the equations governing the evolution of these systems renders their prediction very challenging. The formulation of well-posed models is therefore critical. In order to develop and validate such models, Direct numerical simulation capabilities, dedicated analysis tools, and novel modeling strategies are essential.
This project will focus on three different aspects. First, the scale of Blue Waters will enable us to apply recently developed numerical techniques to the largest turbulent two-phase flow simulations ever performed. Second, the computational performance will be tested and optimized by adapting the algorithms and the structure of an in-house high-performance scientific software to exploit the unique capabilities of the heterogeneous XK nodes. Finally, the carefully conducted simulations will provide valuable information for the a priori analysis of the next generation of lower-order models for turbulent two-phase simulations.