A comparison of methods to simulate the aerodynamic flow beneath a high speed train

The introduction of dedicated high speed railway lines around the world has led to issues associated with running trains at very high speeds. Aerodynamic effects increase proportionally with air speed squared, consequently at higher speeds aerodynamic effects will be significantly greater than for trains travelling at lower speeds. On ballasted trackbeds the phenomenon in which ballast particles become airborne during the passage of a high speed train has led to the need for understanding of the processes involved in train and track interaction (both aerodynamical and geotechnical). The difficulty of making full-scale aerodynamic measurements beneath a high speed train has created a requirement to be able to accurately simulate these complex aerodynamic flows at model-scale. In this study results from moving-model tests and numerical simulations were analysed to determine the performance of each method for simulating the aerodynamic flow underneath a high-speed train. Validation was provided for both cases by juxtaposing results against those from full-scale measurements. The moving-model tests and numerical simulations were performed at 1/25$^{th}$ scale. Horizontal velocities from the moving-model tests and computational fluid dynamics (CFD) simulations were mostly comparable except those obtained close to the ballast. In this region the multi-hole aerodynamic probes were unable to accurately measure velocities. The numerical simulations were able to resolve the flow to much smaller turbulent scales than could be measured in the experiments, and showed an overshoot in peak velocity magnitudes. Pressure and velocity magnitudes were found to be greater in the numerical simulations than the experimental tests. This is thought to be due to the influence of ballast stones in the experimental studies allowing flow to diffuse through them; whereas, in the CFD simulations the flow stagnated on a smooth non-porous surface. Additional validation of standard deviations and turbulence intensities found good agreement between the experimental data but an overshoot in the numerical simulations. Both moving model and CFD techniques were shown to be able to replicate the flow development beneath a high-speed train. These techniques could therefore be used as a method to model underbody flow with a view to train homologation.