Last month's SU2 Conference brought together the community of researchers and practitioners who are focused on the SU2 open-source CFD code. This flow solver originated at Stanford (SU2 = Stanford University Unstructured) and is now managed by the SU2 Foundation. The foundation's mission is to "support global software development and education in the engineering sciences" and it's led by a multi-organizational group of CFD experts. Pointwise has been a proud supporter and sponsor of SU2 for many years and is a regular participant in their workshops and for the past two years their annual conference.
Cadence's Travis Carrigan presented "Automation, Adaptation, and Advanced CFD Meshing in Pointwise" at this year's event to bring the group up to speed on the software's latest capabilities.
Automation is at the top of our feature development plan and we've given this feature suite the name Flashpoint. The premise behind Flashpoint is to minimize the amount of user action required to obtain a suitable mesh: not just any mesh but one that's ready for CFD. The key to obtaining a suitable mesh is to go beyond just blind automation and add best practices honed over decades of experience. Shown below is what's possible with Flashpoint surface meshing today. Given a suitable geometry model, the UI on the left illustrates the minimum set of parameters required to generate a CFD-ready surface mesh including information where anisotropy should be used to resolve features such as wing leading and trialing edges. The meshes show below are the result of changing a single parameter to obtain a more highly refined mesh.
Another recent addition to Pointwise's meshing capabilities is the use of voxels on a mesh's interior. Our T-Rex (anisotropic tetrahedral extrusion) technique generates semi-structured layers of boundary layer-resolving cells near no-slip surfaces and transitions to isotropic cells away from those surfaces. One technique for filling the volume region is a simple isotropic tet mesh. A recently added option is the use of voxels, axis-aligned hexahedra with multiple levels of refinement. The advantages of voxels over tetrahedra are well-aligned, well-shaped cells, and more efficient volume filling.
These two cuts (gray = symmetry plane, blue = tail section fuselage station) through a volume mesh around an Extra 300 aircraft show the use of voxels away from the aircraft surfaces. Multiple levels of refinement are stitched together to be point-to-point continuous.
Another powerful capability in Pointwise is mesh adaptation using a highly efficient point-cloud technique for driving the adaptation process and coupling Pointwise to the CFD solver. Rather than have the flow solver produce an adaptive metric at each mesh point, the solver only need deliver a sparse set of points at which a cell edge-length metric is specified. That point cloud is used to drive the adaptation process in Pointwise (using its source clustering technique) and the updated mesh is sent back to the flow solver to continue the solution. Mesh adaptation relieves the burden of having to generate an optimal mesh in advance and instead begins with a suitable mesh that's adapted to an optimal form by the specific needs of the flow solver.
This cut through a volume mesh around an automobile illustrates a mesh that has been adapted to the flowfield.
A recording of the full presentation of Automation, Adaptation, and Advanced CFD Meshing in Pointwise is available on the conference website and is well worth 15 minutes of your time. It looks like most of the other presentations, slides and videos are also available for your viewing pleasure.
If I've piqued your curiosity about SU2 and its capabilities, you can download it from GitHub here.