What Our Interns Did This Summer on Cadence's Pointwise Meshing Team

There's just one problem with summer interns. As soon as you get to know them, they have to return to school. This loss is not diminished by the fact that ours (and probably yours too) had to work remotely this summer. (Which is still better than last summer when we rescinded our two internship offers due to the pandemic, one of the lowlights of my career.) Anyway, all three of our interns have come and gone and are back on campus working toward their degrees. And now is the time when I share what they accomplished during the summer.

(This post talks about their work. If you're interested in Intern Reading Club, that's another blog post.)

Logan Tan

Logan just completed his undergraduate degree in mechanical engineering at Washington University in St. Louis before coming to us this summer. He'll be continuing his education there in the fall as part of a 5-year master's degree program. He worked with our team of applications engineers on applied meshing.

As is the case for most technical interns, Logan began by taking the online Pointwise Meshing Foundations course which is available to all our customers and friends and provides comprehensive instruction in all aspects of mesh generation using Pointwise. Here's what a NASA engineer had to say about it.

"The Pointwise Mesh Generation Foundations course was very detailed and informative. Even as an experienced Pointwise user, I was able to learn useful tips for improving workflow as well as discover existing features that will be helpful for future meshing projects. If you want to learn how to use Pointwise and/or learn how to better utilize the software to become more efficient at mesh generation, I would highly recommend the Pointwise Mesh Generation Foundations course."

- CFD Engineer, NASA Langley Research Center

Logan didn't stop at just learning the software by taking the course and running the tutorials. He also improved both by identifying places where updates are needed and by making suggestions on features that would improve usability. So when you take the course and run the tutorials, you can thank Logan for the improved experience.

Logan used his new Pointwise knowledge to test the Flashpoint automatic surface mesher (released in V18.4 last October) on a plug nozzle geometry we had been generating for a CFD workshop. Given a geometry model, Flashpoint takes a handful of mesh attributes and applies all of the software's techniques and our best practices to automatically generate a surface mesh with curvature and anisotropic refinement.

Pointwise Flashpoint automatic surface mesher was used to generate this mesh for a benchmark nozzle geometry.

More QA testing was done by Logan for a new feature coming in a future release, Thin Surface Interpolation (TSI). This technique is for quad-dominant meshes on thin surfaces (hence the name) like wing trailing edges. Typically you would generate a structured quad grid on the trailing edge and diagonalize it into triangles. However, some of these thin regions don't lend themselves to a strict IxJ structured grid. TSI provides more flexibility by adding just a bit of unstructured-ness to the structured grid.

And because I challenge the interns each summer to demonstrate something totally new (to me) in Pointwise, Logan wrote a Python script that takes a photograph of an object and uses Pointwise's Glyph scripting language with Python to make a surface within the borders of the photographed object.

Mayank Sharma

Mayank is a Ph.D. student in the Gas Turbine Simulation Laboratory at the University of Cincinnati. His research involves turbomachinery optimization using a gradient-based method for CFD solutions on a structured grid. He spent the summer investigating new formulations of the control functions for use in Pointwise's structured grid solver (elliptic PDE based) that might provide improved behavior in high-curvature regions while also minimizing any user interaction. It's important to note here that Pointwise's elliptic PDE method splits the control functions into those on the interior of the grid (for smoothness and clustering) and those on the boundaries (for orthogonality).

Mayank's exploration began with the Villamizar-Acosta method which tends to produce grids that have a uniform distribution of cell area. This quality of the method was confirmed in his implementation with the exception of convex and concave corners. Of course, combining V-A with one of Pointwise's existing boundary control functions helped in corners while also providing an even smoother control of the interior mesh.

Next the Spekreijse method was implemented. Unlike most structured grid methods which transform computational space directly to physical space, this method introduces an intermediate parameter space such that one algebraic transformation maps computational to parameter space and then a second transformation (this one elliptic in nature) goes from parameter to physical space. When compared to V-A, the S method better controls clustering in addition to smoothness but it still suffers similar problems in corners.

This structured H-grid for the top half of a turbine blade domain was generated using Spekreijse's method.

This side-by-side comparisons illustrates the difference between solving the elliptic PDEs using Pointwise's default Thomas-Middlecoff (interior) and Steger-Sorenson (boundary) control functions (left) and replacing Thomas-Middlecoff with Spekreijse (right). The latter does seem to improve upon one of Steger-Sorenson's drawbacks which is how rapidly the mesh clustering grows out of concave regions (see the dip on the left of each grid).

Overall, the conclusion was that either of these new control function formulations was better than using nothing (aka Laplace) for the interior control functions but that the boundary control functions were still required to maintain clustering at the boundaries. Research will continue.

Sarah Hope Swaim

Sarah Hope is a 4th year senior at Arizona State where she's majoring in Computer Science with a minor in Innovation in Society. You may be wondering, like I was, exactly what one learns when studying innovation in society. As it turns out, Arizona State is the home of the School for the Future of Innovation in Society that has the stated goal of “linking innovation to public value.” Or as Sarah Hope told us during her presentation, “You can play an important part in designing a better future for all of us when you understand the roles and impact science and technology have in society.”

One of Sarah Hope’s first accomplishments was to update the Glyph scripts on our GitHub page. This update was originally driven by the need to rebrand from Pointwise to Cadence because of the acquisition. In the process of making those changes, Sarah Hope also tested all the scripts and converted three of them from Tcl to Python. (If you haven't yet visited our GitHub page you probably should. We share all sorts of handy Glyph scripts as well as the source code for several CAE plugins. It's a great set of learning tools and launching points for your own work)

Some of the scripts on the Glyph Script Exchange are now available in both Tcl, Glyph's native language, (left) and Python (right). They have identical look, feel, and functionality.

The Plugin SDK was the next task Sarah Hope tackled by adding both CAE export and mesh import for the MFEM format (to be released in Pointwise soon). She also contributed to building and testing of an update to the Kestrel exporter.

When asked whether her internship experience met her expectations, Sarah Hope said the following.

"Throughout the summer, my knowledge of software development, CFD mesh generation, and so much more grew immensely. Working alongside such welcoming and passionate people provided an amazing experience that I will never forget."

And that's what I call a pretty good way to end the summer.