optimizing fin shapes

[Buckeye]

If someone wants me to use my fairly beastly rig to do some intense CFD work, just say the word .

Ooh. What kind of hardware and software do you run?

 

[dhbarr]

Ooh. What kind of hardware and software do you run?

i7-5930, 64GB RAM, raid0 512G SSD, Titan GPU.

No NVME or spinning rust, but could do if I had a project that needed it. No ECC as this is quite obviously a gaming rig which happens to be excellent at crunching.

Software is the issue, only really have access to OpenFoam or similar. So I'm running nothing extremely fast

 

[blackbrandt]

I got an i7-6700HQ, 16GB DDR4, and nVidia GTX 960M.

 

[raptor22]

I would think that CFD is good enough these days to run comparison studies at much lower costs. The students can crank the numbers 24/7.

In our department, CFD is a "graduate students only" level class. The faculty here is of the opinion that, unless you are qualified to write your own solvers, you don't understand the inner workings well enough to know whether or not you are getting valid results. We do have some fantastic professors who are willing to help students build CFD models, but that type of involvement is reserved for special cases where the professor takes an interest in a project that needs it.

We have access to FLUENT with machines built to run it, and building a FLUENT model for "Spiked spherical blunt bodies at M=2.0" was considered. However, the time required to build an accurate converging model was more than the time required to fabricate and run our windtunnel models, so we chose to use the windtunnel to fit our time constraints.

Unless the culture at other universities is vastly different, most undergrad aerospace students aren't spending their days crunching numbers in FLUENT.

Honestly, though, what would be the point of running standard fin planforms with standard airfoil shapes? Testing is expensive, either in time, money, or both, and this is a well understood subject. The industry has huge amounts of data compiled on the drag and lift of supersonic lifting surfaces, and it makes a lot more sense to pull it from an accepted source after someone else does the legwork if at all possible.

IMO, its only worth it to run the tunnel or CFD in this instance if you have some novel promising shape in need of investigation.

 

[Buckeye]

Hey raptor22, thanks for the boots on the ground perspective. Yep, even when I was in school back in the Stone Age, CFD was rarely taught to undergrads. The CFD snobbery from the profs is not doing the students any favors. In my business, we rely on well-established commercial software, and I recruit CFD engineers with a good understanding in applications, not so much in the inner workings of solvers. Since most undergrads never get a wiff of a CFD model in school, they get left out of our hiring efforts. Shame. Virtual engineering is only going to keep growing.

Interesting experience with Fluent vs. your wind tunnel.

At your level and at the industry level, all this airfoil info is easily found and understood. Most hobbyists don't know where to look.

 

[dhbarr]

At your level and at the industry level, all this airfoil info is easily found and understood. Most hobbyists don't know where to look.

And / or don't have access to AIAA archives, which is where I keep coming up short.

 

[Winston]

And / or don't have access to AIAA archives, which is where I keep coming up short.

Ditto.

 

[raptor22]

Hey raptor22, thanks for the boots on the ground perspective. Yep, even when I was in school back in the Stone Age, CFD was rarely taught to undergrads. The CFD snobbery from the profs is not doing the students any favors. In my business, we rely on well-established commercial software, and I recruit CFD engineers with a good understanding in applications, not so much in the inner workings of solvers. Since most undergrads never get a wiff of a CFD model in school, they get left out of our hiring efforts. Shame. Virtual engineering is only going to keep growing.

Interesting experience with Fluent vs. your wind tunnel.

At your level and at the industry level, all this airfoil info is easily found and understood. Most hobbyists don't know where to look.

That is very interesting, thank you for sharing your experience from industry. Myself and other students are definitely very interested in learning how to use CFD as an effective tool.

For those looking for the data: it seems to me that instead of purchasing access to AIAA reports, the most effective use of resources is probably a book compiling data.

The books I mentioned are great resources, and "Fundamentals" lists tons of sources for a more detailed look. Other good authors for compiled data on flying vehicles include Raymer in "Aircraft Design A conceptual approach" and Jan Roskam. They present the data in nicely aggregated graphs with explanations on their use.

These books also discuss how to find optimum configurations through parametric studies, which is a nice addition to the analysis toolbox for a hobbyist.

 

[bobkrech]

Today, all serious number crunching is performed on the world's largest supercomputers using GPUs (Graphics Processing Units) which are massively parallel floating point processors. https://www.top500.org/lists/2016/11/

Nvidia is the major player in GPUs. https://www.nvidia.com/page/home.html If you have a gaming computer, you are likey to have a TFLOP Nvidia GPU inside your computer. https://www.geforce.com/hardware https://www.nvidia.com/object/quadro.html https://www.nvidia.com/object/tesla-supercomputing-solutions.html

The Orion study was probably done about 10 years ago. NASA was trying to develop their own TPS for Orion instead of using the tried and true AVCOAT heat shield developed by AVCO Wilmington in the mid-60s for Apollo. After years of finding nothing better, that's indeed what they are using for Orion. It appears that no one had bothered to read the thousands of NASA documents developed during and after the Apollo program to document the physics, calculations, and engineering done to successfully go to the moon and safely return to earth.

The engineering calculations and case modeling conducted performed for the Apollo heat shield development could be performed an a PC based GPU system in less than a second......

 

[boatgeek]

10 days on how many CPUs? That doesn't sound too bad for such a detailed and lengthy simulation. The fin designs in question here would be much less time consuming. I do automotive (subsonic) aerodynamics with CFD (often more difficult than aerospace applications because of ground effects, turbulence, and bluff shapes), and we get one to two day turnaround on a couple hundred cores per job.

If a university can afford to build and maintain supersonic wind tunnel labs and fab shops, they can easily purchase (or rent on the cloud) a few thousand cores in a High Performance Computing system and still come out ahead.

I would guess (I'm not involved at all) that the 10 days comes from simulating short time steps over the entire descent from not-quite-orbital speed at the Von Karman line to the surface. With all of the changes in density and speed in between, that gets to be a pretty impressive number of simulations. You can do simulations of the scale we're talking about here in relatively short timeframes on small numbers of cores. On my current HPR project (https://www.rocketryforum.com/showt...don-A-square-rocket-that-doesn-t-quite-fit-in), a coworker ran a few CFD simulations on off time, and we were getting convergence in 20-30 minutes on 12 cores. He also had a really cool system set up that tested whether the Mach number changed significantly across cells and then remeshed those areas with a tighter mesh. Getting to finding a 0.1% to 1% difference in fin shape will take a very fine mesh and a lot of work, though.

Incidentally, we use CFD for ship resistance at the office, so you get the added fun of the interface between air and water. My coworker said the rocket was a lot easier than trying to capture a free surface.

 

[Buckeye]

My coworker said the rocket was a lot easier than trying to capture a free surface.

Absolutely!

 

[Igotnothing]

Meanwhile, us dummies will just model a rocket in Open Rocket, pick a motor that approximates what we might run in said rocket, and then play with the fin shapes taking note on the impact to top speed and altitude. And then we might spend some time pondering whether we want to buy the necessary wonder materials needed to keep the flutter fiend far away.

And then we just decide to fly it slower.

 

[BEC]

The books I mentioned are great resources, and "Fundamentals" lists tons of sources for a more detailed look. Other good authors for compiled data on flying vehicles include Raymer in "Aircraft Design A conceptual approach" and Jan Roskam. They present the data in nicely aggregated graphs with explanations on their use.

These books also discuss how to find optimum configurations through parametric studies, which is a nice addition to the analysis toolbox for a hobbyist.

Interesting - I'll have to spend a little quality time with Roskam and Raymer and see what might be useful in their books. After Christmas.....

 

[OverTheTop]

There are a couple Apogee newsletters on this topic, written in simple language (sometimes too simple). I remember one in particular that said fins with square trailing edges were low drag.

I just commented on this in another thread somewhere. IIRC the flat edge is about 1% chord for optimum effect. I can't remember what the advantages were or under what conditions. If you have a look in the book "Fundamentals of Aerodynamics" by Anderson, I think it was covered in there. I could also have been in his book "Introduction to Flight".

Search for a free pdf of that first one for a great read. Highly recommended.

 

[Grog6]

Thanks for the tip; that looks like a great read.