A code framework based on the Smoothed Particle Hydrodynamics (SPH) method has been used to investigate the liquid disintegration processes of an air-assisted atomizer. As the flow physics includes spatial and temporal scales which cover at least 4 orders of magnitude, the use of HPC resources is indispensable. The application of the SPH method is rather new to computational fluid dynamics (CFD). We therefore compare our in-house code to established CFD tools in order to assess the computational performance as well as the quality the physical results. It can be shown, that SPH is able to outperform commonly used grid based methods concerning the scalability behavior as well as the absolute computing speed. The three dimensional test case to be presented consists of 1.2 billion particles. The simulation has been run on the ForHLR I cluster, where 2560 cores have been used for 60 days. The simulation is the most detailed numerical investigation of a prefilmer based atomizer and one of the largest SPH multi-phase flow simulations ever. It did capture the experimentally observed bag breakup regime with good agreement of the spatial liquid disintegration and the breakup time scales.
Smoothed Particle Hydrodynamics for Numerical Predictions of Primary Atomization
High Performance Computing in Science and Engineering ´16: Transactions of the High Performance Computing Center, Stuttgart (HLRS) 2016