In the area of Computational Fluid Dynamics (CFD), the Computational Science Division uses a number of numerical methods, including mesh-based schemes (e.g., finite volume, discontinuous Galerkin, spectral element, and pseudo-spectral), to develop flow simulations. The division also develops pseudo-particle-based flow solvers that employ the Lattice Boltzmann and Direct Simulation Monte-Carlo techniques. These codes/solvers are used to perform large-scale computations using scalable algorithms and answer elusive questions in flow physics. Projects include:
Collaboratory for the Computational Study of Canonical Wall Bounded Flows: Together with researchers from industry and from US/European laboratories and academia, Argonne’s Computational Science Division is simulating canonical, wall-bounded flows at high Reynolds numbers to provide reference datasets to compare against other solutions and evaluating the effectiveness of hybrid RANS/LES methods and wall-modeled large eddy simulations (WMLES) in capturing the details of turbulent flow physics. To date, the project, focused on separated flows, has simulated highly resolved flow fields for the periodic hill, converging-diverging channel, and wall-mounted cube. The flow fields are currently being used to devise and validate new RANS models and evaluate their potential for use in hybrid RANS/LES formulations in the spectral element code Nek5000.
Hypersonic Flows: For compressible supersonic flows, the Mach number (in addition to the Reynolds number) is a scaling parameter. For these flows, the no-slip boundary condition creates a subsonic region near the wall; the location of the sonic line moves closer to the wall at higher (hypersonic) Mach numbers. Also, a significant temperature gradient develops across the boundary layer at supersonic speeds because of the increased viscous dissipation near the wall. In fact, the static-temperature variations can be very large, even in an adiabatic flow, resulting in a low-density, high-viscosity region near the wall. This phenomenon, in turn, leads to a thicker boundary layer than at an equivalent free-stream Reynolds number in subsonic flow. The Computational Science Division is developing highly scalable, higher-order, finite-difference, and discontinuous Galerkin codes to simulate flows over canonical shapes for free-stream Mach numbers M∞<= 5.0. Our aim is to simulate flows under conditions that are relevant to validate the results of our computations against the HiFiRe and BOLT experiments.
Particle-Based Methods: We are currently exploring particle-based methods, such as the Lattice Boltzmann Method (LBM), to simulate incompressible and weakly compressible turbulent flows. Interest in LBM stems from its superior scaling on heterogeneous computing platforms (i.e., CPU+GPU platforms). Its applications include flows over complex terrain and wall-bounded flows. The division is also engaged in pseudo-particle simulations of hypersonic flow over blunt bodies in the continuum-transition régime.