Stanford ADL SU2 News
The Aerospace Design Lab at Stanford University released SU2 Version 2, their open source, C++ based CFD code. It is reported that SU2 has been downloaded over 2,700 times in the year since its first release.
Notable new features in Version 2 include (but aren’t limited to):
- Better parallelization and multigrid
- SST turbulence model
- Reacting flow models
- Improved grid adaption
The SU2 team presented a paper about their code at the recent AIAA Aerospace Sciences Meeting. Their paper titled Stanford University Unstructured (SU2): An Open-Source Integrated Computational Environment for Multi-Physics Simulation and Design is available for download from their website.
And to wrap-up SU2 news, CFD Online opened a new forum dedicated to SU2.
Choices, Choices, Choices – How to Mesh for Analysis
Desktop Engineering published the article Meshing Your Design for Analysis: Which Path to Take? in which they discuss with industry representatives the tradeoffs of three different approaches:
- Use CAD-embedded meshing and analysis.
- Use meshing tools native to your analysis package.
- Use specialized meshing tools.
Pointwise’s Rick Matus contributed to this article and spoke about all approaches but, obviously, represented use of specialized meshing tools.
If I was to cite one benefit for each approach they would be:
- CAD-embedded – avoid geometry transfer/translation
- native to analysis – guaranteed mesh/analysis compatibility
- specialized – broad functionality portable to many analysis tools [OK, that's 2. Sue me.]
Without meaning to contradict anything said in the article, it’s clear that the demarcations between the three approaches aren’t necessarily all that distinct. Native and specialized meshers have strong CAD import capabilities. CAD-embedded meshers aren’t always as simplistic as folks think. And specialized meshers have customization and automation features that allow non-expert use.
What’s also unsaid is that the mesh must provide a solid foundation for your analysis – the needs of the software itself and your needs for simulation results. If your solver needs hexes and the mesher doesn’t make them it doesn’t matter what other attributes it has. If you need resolution of viscous boundary layers to compute separation or heat transfer and your mesher only makes uniformly-spaced meshes you can’t use it.
[Comment: At the top of the article there's an image of what looks like an overset mesh and flow solution for a weapon dropping off the pylon of an aircraft. However, the image is so terribly distorted that all the grid lines have dropped out making the image a waste of pixels. You'd think that the magazine would know how to avoid this and/or have fixed it by now. For example, the image appears to be in JPEG format which, because of the compression, is an absolute no-no for engineering graphics with lots of fine lines. I'm also willing to bet they scaled the image up which is a losing proposition, literally and figuratively.]
Meshing for FEA
Desktop Engineering also ran another meshing article, this time focused on FEA. Meshing for FEA is based on material from NAFEMS’ Intro to FEA course and includes several interesting factoids.
- In some cases meshing can account for 80% of project time. [I found this surprising for FEA where I thought that meshing was much more mature and automated.]
- The practical performance limit for FEA meshes is 5 million elements.
- Hexes and quads are preferred over tets and tris for accuracy reasons. If you must use a tet for reasons related to meshing time, never use a linear tet – use one with 10 nodes.
- Badly shaped elements effect the accuracy of the FEA results.
- And of course, CAD geometry can have an adverse effect on meshing.
Best of all, the author advises that you limit the amount of time you spent on meshing because meshing is such a “therapeutic pastime” it’s easy to get carried away. I like where his head’s at!
News in Brief
- CFD Online opened a new general forum for visualization.
- The OpenFOAM Foundation has opened a repository for unsupported contributions to the code.
- Siemens PLM Software released Femap Version 11 including the ability to create geometry from legacy meshes, better XY plotting, and improved NASTRAN compatibility.
- Beta CAE released versions 14.0.0 of ANSA and μETA for pre- and post-processing, respectively.
- 30 billion cells in 120 seconds? That’s what FieldView was able to do for a researcher at the University of Tokyo who’s computing 10-20 high fidelity CFD runs per night.
- Can Carnegie Mellon’s Capability Maturity Model for software development be molded into a Simulation Maturity Model? [This is an old blog post rediscovered in Altair's year-end summary.]
- Can a wind turbine operate efficiently when installed in the middle of Case Western Reserve University’s urban campus? (Full paper available at the link.)
Trelis for Meshing
Computational Simulation Software (aka csimsoft) announced the release of Trelis 14.0 – “powered by CUBIT” – for mesh generation. According to their website, after more than two decades of developing CUBIT in conjunction with Sandia National Labs and distributing CUBIT commercially, Trelis is a new product with CUBIT at its core with csimsoft’s upgradges.
[It's not clear to me at this time how CUBIT, Trelis, and the announced (but as of yet unseen by me) open source version of CUBIT are related. Is this the beginning of a divergence? Or will all 3 move forward in lock-step?]
Drop a red hot ball of nickel into water? Let me sing you the song of my people.
When a red-hot ball of nickel is dropped into water it’s surrounded by a layer of bubbles and makes a pretty interesting sound. Click the image for the video.