Comparison between CFD and Wind Tunnel experiment for slender bodies of aspect-ration O(1) in the prescence of cross-wind

Kurzfassung

It is of interest to assess the influence of aerodynamic loading on modern high-speed train systems characterized by low structural weights. These vehicles should be expected to operate safely under a diverse range of operating conditions. The impact of operational conditions on vehicle stability requires a comprehensive understanding of the aerodynamics of a train in the presence of cross wind. Traditionally, measurements have been made using in-service vehicles or via wind tunnel programs which are costly and cover only a fraction of the possible conditions of interest. The concept of the numerical laboratory, whereby Computational Fluid Dynamics (CFD) methods provide high fidelity predictions for arbitrary combinations of operating condition and geometry, is still some way from realization. This pre-sentation discusses work undertaken by the German Aerospace Center (DLR) in the analysis of numerical tools against experiment for prediction of a model train configuration over a range of cross-wind conditions. Measurements reported in the literature [1] [2] reflect the dynamic nature of the flow surrounding a train under cross-flow conditions. These measurements suggest that vortex-dominated flow, boundary layer separation and unsteady flow phenomena play a critical role in determining aero-dynamic vehicle loadings under cross-flow. Numerical prediction of these flows is demanding in terms of both computational and physical modelling requirements. The Large Eddy Simulation (LES) technique has been shown [3] to offer improvement over RANS at significantly greater computational costs. Hybrid RANS/LES methods, such as Detached Eddy Simulation (DES) [4] are significantly cheaper than LES. Hybrid methods, and in particular DES, suffer however from limitations when flow instabilities originate in attached boundary layers and are not externally forced - the formulation of an adequate description of the region bounding the RANS/LES interface remains an open question. Adequate resolution of free shear layers remains another unresolved issue. For the complex dynamical systems that are formed by flow/vehicle interactions computational methods available today cannot, individually, recover all relevant flow physics so that questions of the nature of "What can we resolve?", "What should we resolve?", and "Where/what is the cost break-even point?" are relevant. The purpose of this work is to assist in under-standing the flow processes involved so that these questions can be answered.