By John Chawner
Over 2,000 engineers, scientists, managers and fans of all things aeronautical spent a late June week in Dallas for the AIAA Aviation and Aeronautics Forum and Exposition. This event featured 16 individual conferences including Aeroacoustics, Applied Aerodynamics, CFD, and Fluid Dynamics.
Because Aviation was held just down the road from our Fort Worth office, in neighboring Dallas, Pointwise took the opportunity to send virtually all of our technical staff to experience the event firsthand, whether that meant mingling with customers and partners in the exhibit hall or engaging in a bit of professional development by attending technical sessions. Three papers were authored or co-authored by current or former Pointwise staff and five of us are also involved with technical committees. All this activity made for a very full week.
One of the CFD (computational fluid dynamics) topics I heard most about was high order methods and meshes. Tools and techniques for elevating the polynomial degree of linear meshes was exemplified by meshCurve (Reference 1), software from the University of Kansas that is said to be planned for a free release soon.
Pointwise's work to date on high order meshing has been done via the Glyph scripting language and was presented at Turbo Expo 2014 for an application involving structural mechanics (Reference 2). We are now in the process of transitioning that technology from script to Pointwise itself.
Anisotropic mesh generation was also a topic of considerable interest at Aviation, not only in the context of a priori generation but also from the standpoint of adaption. As someone pointed out, anisotropic adapted meshes would fail any a priori mesh metric calculation but still provide the basis for accurate CFD results.
Pointwise's Dr. John Steinbrenner presented nearly a decade's worth of work on the T-Rex meshing technique in a 30 minute talk (Reference 3). At its core, T-Rex (anisotropic tetrahedral extrusion) is an advancing layer mesher that creates only tetrahedra. In order to reduce cell count and provide cells that are more amenable to accurate computation of the boundary layer regions, strategies have evolved over the years for post-processing the near-wall anisotropic tetrahedra into layers of prism and hexes. Highlights of the cell agglomeration strategies are listed below in the chronological order of their implementation.
The details of Dr. Steinbrenner's paper delve into the complications introduced into cell agglomeration by symmetry plane matching, baffle meshes that may extend into the cell combination layer, and the insertion of multiple extrusion trajectories from sharp, convex corners.
Taking a break from CFD and mesh generation, I attended the panel discussion on “Whatever Happened to the Four-Year Airplane?” during which panelists (Bell Helicopter, Lockheed Martin Skunk Works, Boeing, Johns Hopkins APL, and Gulfstream) delved into the issues that lengthen design cycles, perhaps unnecessarily.
One light-hearted moment from the discussion came when one of the panelists, speaking on the use of formal processes as checks and balances against risk, said that engineers tend to turn into lawyers when they start parsing the language in written requirements.
As noted above, a panelist pointed out that the sparsity of aircraft design programs means there are few opportunities to transfer wisdom from experienced engineers to their more youthful counterparts. In a relatively minor coincidence, the editor's letter in the July/August 2015 issue of Computing in Science and Engineering (Reference 4) addresses this from the computational standpoint. Dr. Douglass Post (head of the U.S. DoD HPCMP CREATE program) cites CFD simulation as an efficient and effective tool to use in a simulated design process by which mentors can transfer their expertise to younger engineers.
It is always a pleasure to hear Prof. Philip Roe speak. The last time I had the pleasure was 2013's symposium celebrating the careers of Roe, Prof. Bram van Leer, and Prof. Antony Jameson (Reference 5). This time Prof. Roe was giving the Fluid Dynamics Award Lecture as this year's recipient (“for seminal contributions to the design of numerical algorithms for simulations of compressible flows”) and chose as his topic “Searching for Truthiness in CFD.”
Roe admittedly borrowed the word “truthiness” from TV's Stephen Colbert who applies it to politicians. In the context of CFD, Roe defines truthiness as “true enough to win an argument” but otherwise “not true enough.” Extending his wordplay a bit further, Roe introduced the term “physicisity” which is the quality or appearance of being based on physical principles.
Later during the week I had the honor and pleasure of being one of the four judges for the first ever Flow Visualization Showcase sponsored by the Fluid Dynamics Technical Committee. Held outside the exposition hall, four video stations were arranged for 30 minute use by the 18 participants during which time they could show off their visualization and answer questions. Three awards were given:
Congratulations to everyone who participated in this inaugural event and let's hope this showcase becomes an annual event.
I was heartened to hear no fewer than three NASA officials cite CFD or the CFD Vision 2030 Study (Reference 6).
What is the take-away from this? The Vision 2030 Study is more than a paper tiger. It is a significant and actionable vision and a guide for moving forward.
Prior to leaving Pointwise to pursue a physicist's dream job, Dr. John Rhoads co-authored a paper with researchers from NASA Langley on aeroacoustic simulations using FUN3D (Reference 7). This paper was a result of Pointwise's collaboration with participants in AIAA's series of BANC (Benchmark Problems for Airframe Noise Computations) workshops (Reference 8).
The results of the computations turned out quite well, showing better comparison with experimental data relative to a baseline mesh. See Figure 4 for a representative CFD solution. The difference is likely due to Pointwise's ability to control clustering in regions of interest. The authors graciously wrote “Based on our current experience, Pointwise offers a more up-to-date and efficient alternative to VGRID for generating grids that are suitable for aeroacoustic computations of such complex configurations.” This is high praise indeed because of the high regard in which VGRID and its author are held. (There is a reason why the Meshing, Visualization, and Computational Environments (MVCE) technical committee's annual best paper award is named in honor of Dr. Pirzadeh.)
In the wake of the publication of the CFD Vision 2030 Study, the MVCE technical committee (of which I am a member) took ownership of all meshing and geometry related problems cited by the study. At AIAA SciTech 2015, MVCE hosted a panel discussion on these geometry and grid issues. MVCE continued its focus on the 2030 Vision with a session on Visualization for CFD 2030.
In addition to a panel discussion on visualization the session featured an invited paper on geometry issues and a summary paper about SciTech 2015's grid generation panel.
Dr. Nigel Taylor (MBDA), shown in Figure 5, was invited to present on the topic of Industrial Perspectives on Geometry Handling for Aerodynamics (Reference 7). The bottom line? Airfoil geometry mathematics is still an active area of research. Knowing that, how can we expect to accurately model an entire aircraft? (It is not actually that dire.)
As Taylor put it, “there's a lot of geometry about.” And a lot of it has to be meshed, where preparation of the geometry, or more specifically the outer mold line (OML), requires a substantial human effort. Why must this be so?
Besides providing a great overview and foundation for further discussions about geometry issues for CFD, Taylor's paper does present a path for moving forward.
What is the true bottom line? “It is often necessary to resolve [in the mesh] regions of the flow adjacent to the OML in significantly finer detail [by orders of magnitude] than the magnitudes usually associated with manufacturing or assembly tolerances.”
Also, Taylor asks that we stop using the phrases “dirty CAD” and “sloppy CAD” because they imply intent, either malicious or ignorant. What is really manifest in MCAD models to which those labels are applied is lack of suitability for a particular purpose where that purpose is CFD.
Prof. John Dannehoffer (Syracuse University) presented the summary paper from SciTech 2015's panel discussion on The Path to and State of Geometry and Meshing in 2030. I was a panelist and co-author of the paper and previously wrote extensively about the panel discussion (Reference 8). This summary paper is a good way to get the panel discussion “on the record.”
But the interesting news announced by Dannenhoffer was the beginning of planning for the first Geometry and Meshing Workshop (in the great tradition of other AIAA CFD-related workshops) to address current status of these technologies and progress toward addressing the CFD Vision 2030 goals. Certainly, more details will follow over time as planning matures.
Finally, a very enlightening panel discussion on CFD visualization consumed the remainder of the morning. The panel included Tecplot's Scott Imlay, Intelligent Light's Earl Duque, CEI's Sean Ahern, and Guoning Chen from the University of Houston (Figure 6).
Imlay had the laugh of the day with his observation that the number of people announcing the death of Moore's law is doubling every 18 months. But his observation also had a salient point: we really need a Moore's law for I/O bandwidth because CPU gains are outpacing our ability to read and write data, which is a lot of what traditional post-processing is about.
After about 1.5 hours of discussion, the cloud and mobile devices were mentioned. In general, all the panelists felt that the cloud as an on-demand resource that scales to your level of need and enables collaboration is already here to a certain degree and will likely be part of standard CFD visualization usage in 2030.
The CFD Vision 2030 Study's principle author, Jeff Slotnick, attended the panel and admitted that the knowledge extraction portion of the study is sparse on the details due to the relative lack of expertise of the authors in that area. He is hoping that this panel discussion and the Visualization Subcommittee of MVCE will flesh-out the details of visualization in 2030 and help map the plan for achieving those goals.
AIAA Aviation and AIAA SciTech are two of my favorite conferences, and I recommend them to all of you who have not yet had the opportunity to attend.
If you would like more information on any of the Pointwise-related papers described above you can email me directly at email@example.com. And if you would like to generate meshes for aeroacoustics, FUN3D or any other application, let's get you a demo license of Pointwise.