We have run a demonstration simulation with our GPU capable CFD code DrNUM. Goal of this demonstration was to show how DrNUM could be used to efficiently simulate environmental flows. 

The DrNUM CFD software was designed from scratch with massively parallel hardware in mind (e.g. GPUs). As a result it is able to perform a complex LES (large eddy simulation) on comparatively cost efficient hardware. We believe this is ideally suited for environmental fluid mechanics. It could, for example, be used to improve wind park layouts or compute pollutant transport.

Below is a small introductory video which shows an example computation for a region close to our office. It uses real terrain data which is freely available from an old radar mission done by the Space Shuttle. The terrain data has a resolution of 90 metres and the CFD mesh uses a cell size of 12 metres which results in a total mesh size of about 80 million cells. The whole simulation, as it can be seen in the video, run for approximately 20 hours on a single PC.

The DrNUM technology can bring high end CFD simulations to normal engineering projects, because it removes the need for expensive high performance hardware.

The satellite imagery we have used for the video has been retrieved from Google Earth and the map is from the OpenStreetMap project.

​​​We are looking for partners to bring this development forward!

​If you are interested in high end CFD for environmental flows (e.g. wind farm design or pollutant transport), please do not hesitate to contact us!

​References​

In 2007 we have started the development of an open-source meshing software called enGrid. At the time there was no other open-source tool available which was capable of creating anisotropic grids (e.g. for boundary layers). We first concentrated on the implementation of a prismatic boundary layer algorithm. Surface meshes could, at the time, only be imported from third party tools. One tool we regularly used for surface meshing at the beginning was Gmsh. The volume meshing outside the prismatic layers was done by the open-source software Netgen, which could easily be integrated into the enGrid GUI.
Later on we implemented a surface mesher based on a discrete surface representation (e.g. STL files). Initially the we used geometric tools which had been implemented in-house and after a few years moved to the open-source library CGAL. At about the same time we changed the tetrahedral mesher from Netgen to TetGen, because TetGen had changed the licence to an open-source licence.
Other important features, which got implemented throughout the years, are the support for polyhedral grids, export functionality for different solvers (e.g. OpenFOAM®, SU2, Dolfyn) and the possibility to create case templates for OpenFOAM®.