EFFICIENT COMPUTATION FOR GENERALISED FREE-SURFACE MULTI-PHASE SMOOTHED PARTICLE HYDRODYNAMICS (SPH) USING GRAPHICS PROCESSING UNITS (GPUs)

点击翻译按钮获取中文摘要

Great progress has been made in the last 20 years in computational fluid dynamics (CFD) with major developments in commercial general-purpose codes and more specialised shallow flow and wave propagation modelling. However, complex, generally highly nonlinear and distorted, surface motion, involving for example wave breaking, bore propagation, aeration and debris transport, has remained largely intractable. In this respect the novel method of smoothed particle hydrodynamics (SPH) holds great promise. SPH is a meshless method and possesses some unique advantages over conventional grid-based approaches; no explicit treatment of the free surface and no computational grid mean that sophisticated meshing is not needed for complex geometries. Recent theoretical advances have been proven in 2-D but the major practical limitation of the method lies in the very large number of particles required for converged solutions. To address this, High Performance Computing (HPC) has been applied in 3-D as well as 2-D in several academic and industrial research centres (EDF, VA TECH Hydro, ESI BV). There is a most attractive alternative however, originating from the computer games industry which has given rise to Graphics Processing Units (GPUs) of parallel architecture which may be exploited for SPH. The aim of this proposal is to exploit GPUs for 3-D free-surface problems, including violent and multi-phase effects, spanning a range of industries such as for offshore or marine engineers or naval architects, civil engineers undertaking flood risk management and other areas with complex free-surface, generally multi-phase flows such as in the nuclear industry.