Massively Parallel Computing in Computational Fluid Dynamics
Massively Parallel Computing in Computational Fluid Dynamics: Insight Into the Physics of Turbulence of Real and Canonical Flows
Computational Fluid Dynamics (CFD) researchers have been among the earliest proponents, and the largest users (both by way of allocations as well as the numbers of users) of high-performance computing (HPC) to simulate complex flow phenomena (which are almost always turbulent). Whilst much of the HPC resources have, traditionally speaking, been devoted to understanding the physics of turbulence in "canonical" turbulent flows, such as shear flows, channel flows, and boundary layer flows, this trend is beginning to shift towards simulating complex flows in complex geometries that consist of O(B) grid points. Needless to say, the main scientific challenge in fluid dynamics remains that of gaining better insight into the physics of turbulence and its role in the transfer of momentum, heat, and mass in engineering applications which include the aerodynamics of high lift devices, chemically reacting flows in combustion systems, such as combustors of jet engines, and the aeroacoustics of low and high speed flows.
The availability of HPC platforms, such as Intrepid, the IBM Blue Gene/P at the Argonne Leadership Computing Facility (ALCF), has led to significant advances in direct numerical simulation (DNS) of turbulence, and has made possible the development and application of large-eddy simulation (LES) for "real" engineering flow scenarios. Further, the data from DNS, obtained from simulations on Intrepid, has begun to provide valuable insight into the physics canonical turbulent flows and promises to lead to the development of better subgrid models.
In this talk, I will present the results of turbulence simulations of both canonical flows and "real" flows in complex geomeries, and discuss the challenges of a new computational initiative with GE Global Research - that of simulating coupled combustor-turbine simulations on Mira, the IBM Blue Gene/Q at ALCF.