APPROACHES TO TURBULENT FLOW COMPUTATION - APPLICATIONS TO FLOW OVER OBSTACLES

Abstract Reliable methods of predicting turbulent flows over three-dimensional obstacles are crucial in wind engineering and other fields. Accurate numerical methods, including those for constructing computational grids, are a pacing item. Methods of treating the effects of turbulence are equally important but more difficult to develop. In this paper, issues connected with numerical methods are discussed, but attention is centered on the treatment of turbulence. For complex flows, two methods may have the potential for accurate prediction, namely methods based on the Reynolds-averaged Navier-Stokes (RANS) equations and large-eddy simulation (LES). RANS models have the benefit of longer history and lower cost, but existing models are not sufficiently accurate for three-dimensional or separating and reattaching flows. Furthermore, it is unclear whether accurate models for these flow phenomena will be forthcoming. Zonal modeling may simplify the task of model construction but still requires models that can deal with three-dimensionality and separation. LES should have less difficulty with these issues. However, costs are much higher and experience is limited, so it is not yet known whether LES will fulfil its promise; to deal with obstacle flows, new models for wall layers will be required, especially for those near separation. A research program that could lead to reliable methods in the next five to ten years is outlined. It combines the use of LES and experimentation for producing accurate data and RANS models for routine calculations.