Wall-modeled Large-Eddy Simulation (LES) of the HIFiRE scramjet

Описание: Visualization extracted from wall-modeled large eddy simulations (WMLES) of the HIFiRE scramjet, targeting the dual-mode operating regime at a flight Mach number of 6.5. Flow features are explored through instantaneous flow fields from particular snapshots, as well as their evolution in time.

The simulations solve the spatially-filtered Navier-Stokes equations in a finite volume formulation on unstructured hexahedral meshes, integrating subgrid-scale models to account for effect of the unresolved turbulent scales, an equilibrium wall model to lower the otherwise computationally impractical near-wall resolution requirements, and a flamelet-based combustion model to account for the complex chemical reactions associated with the JP-7 surrogate hydrocarbon fuel.

A solution-adaptive methodology is implemented to account for the different dominant physics that develop locally in each region of the flow: an essentially non-oscillatory (ENO) second-order shock capturing scheme is applied near shock waves, identified by a shock sensor, whereas a non-dissipative centered numerical scheme is applied elsewhere, to accurately represent the turbulent flow inside the scramjet.

The massively parallel CharLESx compressible flow solver developed at the Center for Turbulence Research, Stanford University/NASA is used to perform these simulations on meshes of increasing resolution, up to 700 million control volumes. The simulations are run on the Mira IBM Blue Gene/Q cluster at the Argonne Leadership Computing Facility (ALCF). Data exploration and visualizations were done in part using the Tukey AMD/NVIDIA cluster at ALCF. The authors also acknowledge the following award
for providing postprocessing resources that have contributed to the research results re-
ported in this visualization: MRI-R2: Acquisition of a Hybrid CPU/GPU and Visualization
Cluster for Multidisciplinary Studies in Transport Physics with Uncertainty Quantifica-
tion (http://www.nsf.gov/awardsearch/showAw.... This
award is funded under the American Recovery and Reinvestment Act of 2009 (Public
Law 111-5). Computing time during the early-science testing period on the Sequoia and Vulcan IBM Blue Gene/Q clusters at Lawrence Livermore National Laboratory through the National Nuclear Security Administration (NNSA) within the Predictive Science Academic Alliance Program at Stanford University is also acknowledged, as it allowed us to scale these simulations to 1.57 million MPI tasks, using 786 K hyperthreaded cores on 48 K nodes.

Financial support from the U.S. Department of Energy, Air Force and NASA is gratefully acknowledged. The authors are grateful to Parviz Moin, David Davidson, Mark Gruber, Gianluca Iaccarino, Frank Ham, Mohammad Shoeybi, Paul Covington, Javier Urzay, Lee Shunn, Amirreza Saghafian, Matthias Ihme and Mirko Gamba for useful discussions and interactions in the course of this research.

Research carried out by: Iván Bermejo-Moreno, Julien Bodart, Johan Larsson and Ronan Vicquelin.

Visualizations created by Iván Bermejo-Moreno, using, among other tools, ParaView, Blender, Python, the VTK library, GIMP and ffmpeg on Linux.

Stanford, California - USA - November, 2013