Spencer Owen spent the summer on our Applied Research team working on new technology that’s still a version or two away from release. He made two new features quite a bit more robust through his testing and feedback.
Working in our Applied Research team gave Spencer access to two technologies coming in future releases of the software. The first was high order mesh generation and curving. The second was something we haven’t talked about publicly too much: an effort to automate (in the literal sense of the word) meshing given only a geometry model.
High-Order Mesh Generation
Coming in Pointwise V18.2 is the first production release of our degree elevation and mesh curving technology for the purpose of generating non-linear meshes up to polynomial degree four.
The technology to be released in V18.2 is the baseline version we’ve been developing ourselves for the past several years and for which we have been awarded a NASA SBIR Phase II contract to take the technology to a higher level (pardon the pun). We wrote about this award in the blog post High-Order Mesh Curving Research Selected by NASA.
Spencer tested this new capability on a suite of cases that pushed the mesh curving technique to its limits. The results of his work included several recommendations and best practices for how to control the process. (Curving of the high order mesh so that it conforms to the curved geometry model and blends smoothly onto the interior of the volume is actually the hardest part of generating a high order mesh.)
Pointwise has been working for several months under subcontract to MIT on automatic generation of body-conforming, hybrid meshes that are suitable for viscous CFD simulation. The overall goal is to require very little user input (default meshing parameters) and require no prior knowledge of the geometry model.
An active area of this research involves proximity-based mesh clustering to ensure a smooth variation of cell size in the presence of several components of the geometry model. The example shown below for a wing, pylon, and nacelle illustrates the use of a clustering source for this purpose.
Another aspect of this work involved properly identifying where to use anisotropic cells for the purpose of geometric and viscous resolution via our T-Rex meshing technique. Spencer was able to achieve some good results for a variety of geometries including the space launch vehicle shown below.
It’s important to keep in mind that the three meshes shown above were all generated using the exact same code with only different default meshing attributes for cell size, etc. The technique is proving to be geometry agnostic.
And the Ten Gallon Hat
Spencer enjoyed his time in Fort Worth so much he bought himself a ten gallon cowboy hat and used that as a test case for much of his research work this summer. We hope he wears it a lot on campus at the University of North Carolina at Charlotte where he’s pursuing a PhD in mechanical engineering. You can read more about Spencer in his This is How I Mesh profile.