There is an urgent need for better methods of predicting the thermal features of turbulent flows that are technologically important. Although progress is needed at all levels of turbulence modeling, the recent advances in computer technology and the outlook for continued advances in the future suggest that it may be timely to give increasing attention to research on direct and large eddy simulation of turbulent flow to move these capabilities toward larger and more relevant problems. Problems to be given particular emphasis in this study include separated flows with heat transfer. The objectives of the proposed research are to: (1) develop a better understanding of complex turbulent flows (particularly separated flows) with heat transfer with a view toward identifying improved turbulence modeling for the Reynolds-averaged Navier-Stokes equations, (2) evaluate and compare the characteristics of several subgrid-scale models for the large eddy simulation of turbulent separated flows with heat transfer including a dynamic model, and (3) work toward a more efficient computational strategy for spatially non-periodic flows to allow more relevant problems to be addressed with current computational resources. The proposed research will attempt to make contributions both to the expansion of basic knowledge and to the advancement of the scientific predictive capability that is needed to maintain a competitive edge in design procedures for products such as gas turbine engine components, electronic devices (such as digital computers) and in manufacturing processes. The research will employ direct and large eddy numerical simulations based on a preconditioned finite-volume discretization technique to solve the compressible Navier-Stokes equations for turbulent flows with heat transfer. The direct numerical simulations will provide basic knowledge of some separated flows with heat transfer, particularly the rearward facing step flow, which have been very difficult to model through the Reynolds-avera ged Navier-Stokes equations. This information will guide the development of improved closure models for the Reynolds-averaged Navier-Stokes equations and aid in the evaluation of subgrid-scale models for the large eddy simulations. Particular attention will be given to establishing the merits and limitations of a dynamic subgrid-scale model suitable for flows with heat transfer. The results should provide fundamental new knowledge about flows with heat transfer and result in improved means of predicting such flows and will enhance national competitiveness.
