High-speed trains push air to the front, sides and over the top to form a train slipstream. The extension of the slipstream to the side, top and wake flow depends on train speed, train shape, ambient conditions and the environment in which the train operates. In this paper, the slipstream and wake flow of a 1/20th scale model of a simplified five-coach ICE2-shape train running in two different environments; in open air and when passing a platform, were obtained using large-eddy simulation (LES). The flow Reynolds number was taken to be 300,000; based on the speed and height of the train. The effect of the platform height on the train slipstream was investigated by performing simulations on a platform of different heights: 20, 60, 90 cm. To investigate the effect of mesh resolution on the results, two different computations were performed for the case of the flow around the train running in the open air using a different number of mesh nodes; a fine mesh consisting of 18,000,000 nodes and a coarse mesh consisting of 12,000,000 nodes. The results of the coarse mesh simulation were deemed to be comparable to those from the fine mesh simulation. The LES results were also compared with full-scale data and a good agreement obtained. A number of different flow regions were observed in the train slipstream: upstream region, nose region, boundary layer region, inter-carriage gap region, tail region and wake region. Localized velocity peaks were obtained near the nose of the train and in the near wake region. Coherent structures were formed at the nose, roof and inter-carriage gaps of the train. These structures spread in the slipstream and extend a long distance behind the train in the far wake flow. The maximum slipstream turbulent intensity was found in the near wake flow. The results showed that there is a significant effect of the platform height on the slipstream velocity and nose and tail pressure pulses. However, there is only a minor effect of the platform height on the static pressure along the body of the train compared with that on the nose and tail pressure pulses. In general, the slipstream velocity in the lower region of a train running in the open air was found to be larger than that around a train passing a platform. This has been related to the effect of the underbody complexities of the train.

Currently, there are three different methodologies for evaluating the aerodynamics of trains; full-scale measurements, physical modeling using wind-tunnel, and moving train rigs and numerical modeling using computational fluid dynamics (CFD). Moreover, different approaches and turbulence modeling are normally used within the CFD framework. The work in this paper investigates the consistency of two of these methodologies; the wind-tunnel and the CFD by comparing the measured surface pressure with the computed CFD values. The CFD is based on Reynolds-Averaged Navier–Stokes (RANS) turbulence models (five models were used; the Spalart–Allmaras (S–A), k-ε, k-ε re-normalization group (RNG), realizable k-ε, and shear stress transport (SST) k-ω) and two detached eddy simulation (DES) approaches; the standard DES and delayed detached eddy simulation (DDES). This work was carried out as part of a larger project to determine whether the current methods of CFD, model scale and full-scale testing provide consistent results and are able to achieve agreement with each other when used in the measurement of train aerodynamic phenomena. Similar to the wind-tunnel, the CFD approaches were applied to external aerodynamic flow around a 1/25th scale class 43 high-speed tunnel (HST) model. Comparison between the CFD results and wind-tunnel data were conducted using coefficients for surface pressure, measured at the wind-tunnel by pressure taps fitted over the surface of the train in loops. Four different meshes where tested with both the RANS SST k-ω and DDES approaches to form a mesh sensitivity study. The four meshes featured 18, 24, 34, and 52 × 10⁶ cells. A mesh of 34 × 10⁶ cells was found to provide the best balance between accuracy and computational cost. Comparison of the results showed that the DES based approaches; in particular, the DDES approach was best able to replicate the wind-tunnel results within the margin of uncertainty.