Incompressible SPH (ISPH) with fast Poisson solver on a GPU
Journal: Computer Physics Communications
Publication Date: 06 February, 2018
Department of: Mechanical, Aerospace and Civil Engineering
A new engineering simulation tool for design against highly violent and destructive flows
This research develops a new methodology for predicting accurately the violent impacts from ocean and coastal waves and is the first to provide a computer simulation tool for fully 3-D calculations which are accessible to all engineers/users by avoiding the use of massive supercomputers. The accurate prediction of pressures and subsequent forces is vital for engineering design and resilience of offshore and coastal structures.
Flows of such violent nature are incredibly complex, involving extreme deformations of the fluid surface and large amounts of fragmentation, which traditional simulation approaches are unable to model easily, therefore an alternative technique is required. The “incompressible smoothed particle hydrodynamics” (ISPH) method is such a technique for simulating these flows whilst providing a highly accurate pressure field. However, the method is very time consuming because a matrix with billions of entries needs to be solved efficiently. Researchers at the University of Manchester have addressed this expense with a novel approach of accelerating ISPH with the scientific computing power of computer graphics cards (known as GPUs) in the open-source SPH code DualSPHysics. For the first time, the new simulation tool, Incompressible-DualSPHysics, can simulate real large-scale engineering applications in hours, whereas previously they would take weeks or months.
- The new GPU-accelerated incompressible smoothed particle hydrodynamics (ISPH) code, Incompressible-DualSPHysics, is a the first of its kind providing an engineering tool for predicting highly accurate pressures in violent hydrodynamic fluid-structure impact flows.
- GPUs are a cost-effective and energy efficient alternative to traditional supercomputers, which can fit into an ordinary desktop PC or laptop. Incompressible-DualSPHysics is therefore a highly accessible research and engineering tool.