CFD Study of the Flow around a High-Speed Train

Abstract: This document is a report summering the master thesis work dealing with the Computational Fluid Dynamic (CFD) study of the flow around a high-speed train. The model is a scaled 1:50 generic train with two cars, one inter-car gap and simplified bogies. A platform is set on the side of the train since one of the aim of the study is to look at the consequences of the phenomena in the wake on people or objects standing on the platform. The slipstream is one of this phenomena, it is due to the fact that the viscous air is dragged when the train is passing. If too strong, it can move or destabilize people or objects on the platform. In addition of the slipstream study, a velocity profile study, a drag and lift coefficients analyze as well as a Q-factor study and a frequency study have been realized. Some results of these different studies are compared with the ones obtained on the same model with a Delayed Detached Eddy Simulation (DDES). Since the flow is turbulent, for those different studies, the flow has been simulated with a Reynolds Averaged Navier-Stokes equation model (RANS) which is the k-ω SST model for the turbulence. The study of the slipstream allowed to calculate the Technical Specification for Interoperability (TSI) which must not be higher that the European Union requirement set at 15.5 m/s, the result obtained is 8.1 m/s which is then lower than the limit. The velocity profile shows similarities with the DDES results even though it is less detailed. The same conclusion is done for the Q-plot where is clearly visible the two counter-rotating vortices in the wake. Finally, a Fast Fourier Transform algorithm has been applied to instantaneous velocity results in the wake of the train in order to get the frequency of the aerodynamic phenomena in that wake. The main frequency is 25 Hz and corresponds to a Strouhal number of 0.1, quite closed to the results obtained with DDES which is 0.085. The results of the RANS and DDES are reasonably similar and by regarding at the large difference between the cell numbers (respectively 8 500 000 and 20 000 000) it can be conclude that in some ways the RANS model can be preferred at the DDES to save time for the computation but it does not contain the small scales resolved by the DDES.