The large-scale, computational simulation of turbulent fluid-dynamical phenomena will continue to have an enormous impact on many diverse areas of science and engineering ranging from climate modeling of planet earth, to environmental fluid dynamics and to industrial and engineering applications at large Reynolds numbers. The ideal is direct-numerical simulation (DNS) in which all relevant physical processes are properly represented and all length scales are resolved within numerical simulation. At the extreme Reynolds numbers required for many engineering applications, however, full DNS is unlikely to be practicable within the foreseeable future. The standard engineering prediction tool has been Reynolds-averaged modeling (RANS). Whilst RANS will remain useful for many applications, there exists a growing need for a more detailed but computationally tractable numerical simulation capability in engineering development work. Examples include internal pipe flows and external streamlined and bluff-body flows where physically realistic modeling of turbulent boundary-layer dynamics including transition, curvature, separation and Reynolds-number effects is required for accurate prediction.
Large-eddy simulation (LES), where the large scales of turbulent motion are resolved on the computational grid while the effects of small, unresolved eddies are modeled, is intermediate between RANS and DNS. LES has been very successful for free-shear and mixing turbulence in a wide variety of settings. In unbounded flows, the large eddies carry most of the turbulent kinetic energy, dominate momentum transport and set the length and time scales that condition the small-scale turbulence dynamics. This picture is reversed near a smooth or rough wall, where the most energetically productive eddies are necessarily part of the small-scale motion. Hence despite decades of effort, the accurate numerical prediction of wall-bounded turbulent flows remains a challenging area for computational fluid dynamics. The broad objective of the present research is to construct a robust LES capability for wall-bounded flows at Reynolds-numbers typical of practical engineering interest.
The project will aim to develop a subgrid-scale methodology for LES of wall-bounded turbulence with emphasis on application to spatially evolving smooth or rough-wall turbulent boundary-layer flows in the presence of favorable and adverse pressure gradients, wall curvature, laminar-turbulent transition and flow separation at essentially arbitrarily large Reynolds numbers. The novel element is a subgrid-scale wall model, based on a wall-normal integration of the stream-wise momentum equation, which enables dynamical calculation of the wall shear stress without requiring near-wall scale resolution, but which incorporates local surface roughness, wall-normal momentum transport and pressure-gradient effects. The LES modeling resulting from this work will be available for incorporation into general computational fluid-dynamics codes. It is expected that this will provide a significant advance in our capability for the numerical simulation of complex turbulent flows at very large Reynolds numbers.
The research will form an important part of the education and training of individual graduate students. Additionally the work will support dissemination of the concepts and applications of modern computational engineering technology through participation in K-12 outreach programs.
