Simulation of high Reynolds number flow over a backward facing step using SPH

Except as may be otherwise specifically indicated or acknowledged herein, the data and information contained herein is the property of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) Australia. The content of this document is only to be used for the purpose for which it is specifically supplied and must not be used, reproduced or disclosed to third parties, in whole or in part, without the prior written permission of CSIRO. Summary The Lagrangian Smoothed Particle Hydrodynamics method has recently been used to simulate fluid flows. Often, turbulence effects are neglected, and most flow simulations are performed in the low Reynolds number laminar regime. Therefore, in this project, a quasi-compressible Smoothed Particle Hydrodynamics (SPH) code is used to simulate a high Reynolds number flow past a backward facing step. A preliminary attempt to quantitatively evaluate the ability of SPH to predict high Reynolds number flows is made here. For this purpose, the SPH code belonging to the Computational Fluid Dynamics Group of CSIRO Mathematical & Information Sciences is used to perform all baseline flow simulations. A sub-particle scale model, similar in concept to a sub-grid scale model in the grid-based Large Eddy Simulations (LES) approach, is incorporated into the SPH code to predict the effects of turbulence. Simulations are performed at three different SPH resolutions and compared with experimentally and numerically established results. Results of the SPH simulations showed several flow characteristics that have previously been reported by other researchers, mainly: formation and detachment of secondary recirculation regions, and fluctuations of the reattachment length. The flow evolution as predicted by the SPH code compares qualitatively with similar work by other researchers. Numerical results predicted by the SPH code, mainly downstream mean velocity profiles and the mean reattachment length, do not compare quantitatively with experimental results. It is concluded that this is primarily due to the limitations associated with a two-dimensional LES computation. Future work in this project must then be prioritised in the regard of performing fully three-dimensional simulations.