Gapped fins....

[Dan Griffing]

I don't know how fluid dynamics is solved numerically but I assumed the problem is somehow discretized and equations are applied to each point based on the geometry. It doesn't seem that it would require solutions of large systems of simultaneous equations or nonlinear iteration but I'm not an aerospace engineer.

Your assumptions are wrong.

Numerical simulations require large 3D or 4D grids that are iteratively solved with each of the millions of grid elements variable’s being recalculated with the values of its neighbors from the previous iteration.

This requires very fast, high precision floating point operations, and millions of storage locations for each grid element, and for the last few calculation iterations for each element. That’s how complex sets of fluid dynamics differential equations are numerically calculated.

I’m a retired engineer who has worked with these kinds of problems. Although it might not seem that way to you, this does require a substantial amount of computational resources.

There may be simpler approximations that can work for a very limited subset of CFD, but that takes someone who fully understands CFD to extract and derive this subset, and you don’t run into this kind of person very often.

 

[bjphoenix]

Your assumptions are wrong.

Numerical simulations require large 3D or 4D grids that are iteratively solved with each of the millions of grid elements variable’s being recalculated with the values of its neighbors from the previous iteration.

This requires very fast, high precision floating point operations, and millions of storage locations for each grid element, and for the last few calculation iterations for each element. That’s how complex sets of fluid dynamics differential equations are numerically calculated.

I’m a retired engineer who has worked with these kinds of problems. Although it might not seem that way to you, this does require a substantial amount of computational resources.

There may be simpler approximations that can work for a very limited subset of CFD, but that takes someone who fully understands CFD to extract and derive this subset, and you don’t run into this kind of person very often.

No, I understand. I didn't know if it required iteration. That obviously requires a lot of calculations and the size of the model would be an important factor, i.e. the number of points to be calculated. It seems to me that a pair of split fins may require a lot fewer points than an entire curved wing for instance, or turbine blade. When I started writing FEM type programs we were solving hundreds of linear equations, certainly less than a thousand. Programs we use these days would be solving many thousands of equations, but generally linear behavior and not nonlinear so they don't require iteration. Not that difficult to do on a PC but iteration would obviously require a longer solution.
What prompted this- it's not that a phone couldn't do it, it's that you couldn't effectively define the model or view the results with the limited I/O of a phone. You could crank it up on your phone and let it run for awhile, but anyone who needs to solve this type of problem will have a fairly powerful PC available to them. At my office everybody has transitioned to powerful laptop computers so they can carry them home to work from home or they can carry them to meetings, I think the gaming PC that I bought 2 years ago is quite a bit more powerful that the laptops we issue at work and many times more powerful than the desktop computer that I use at home.

 

[jderimig]

I have run into this before years ago when I was doing gap finned rockets but i don't remember the issue. As an experiment is to model your gap fin but make the gap very small, almost zero. If there is a big shift in Cp then then the calculation is wrong. I think that is what I found.

 

[jqavins]

I have run into this before years ago when I was doing gap finned rockets but i don't remember the issue. As an experiment is to model your gap fin but make the gap very small, almost zero. If there is a big shift in Cp then then the calculation is wrong. I think that is what I found.

And that's been done here, reported upthread. The calculation is confirmed to be wrong.

As for the computing power needed, first let me say that I regret bringing up phones. I was simply illustrating the huge advance in computing power vs. size in the last 30 years, not suggesting a phone as a platform for these computations.

As for stress analysis vs. CFD, to summarize, they're both FEA, where one has more variables and more equations, some of which are non-linear, than the other, so it's a big bunch harder. And the shapes for rockets are rather complex, and the analysis is actually for the space outside the solid object extending to a not so simple extent, so the mesh will require a whole lot of nodes in order to even be pretty good, let alone great. And that's just the aspects I can identify. So yeah, a very big bunch harder.

All that said, I'm amazed (and a little bit skeptical) to hear that the theory and approach, including automatic meshing that's reasonably good (for some value of reasonable) are not established art, requiring only a coding exercise. To do a very high quality job would, undoubtedly, require expertise in the meshing process and a real crapload of nodes. But as for the computing power, every hard core gamer today has a box under his/her desk that makes 1993's supercomputer look stupid.

 

[jderimig]

A couple thoughts.
1. Listen to @Dan Griffing. DIY CFD will likely produce a result that is less accurate than just ignoring the fin gap and assuming it isnt there.
2. The turbulence in the fin gap is probably less that all the other real surface turbulent sources you are ignoring in the model.
3. DIY emprical studies are superior to DIY theory. Make a model, fill in the gap with flush fitting insert. Fly it a few times. Take the insert out, fly it again a few times. Compare the data, make you own conclusion.

 

[bjphoenix]

I would guess that a reasonably good wind tunnel would be a good way to investigate this.

 

[BABAR]

I would guess that a reasonably good wind tunnel would be a good way to investigate this.

Good thought. I wonder, though, even WITH a wind tunnel, there are nearly an infinite number of variations of gap sizes and fin lengths and widths and fin bevels, the complexity of the data points may be challenging, as now you have Empiric data that you are trying to find a formula to fit.

then again, once you get beyond Bernoulli's principle i am pretty much lost…..

 

[jqavins]

Good thought. I wonder, though, even WITH a wind tunnel, there are nearly an infinite number of variations of gap sizes and fin lengths and widths and fin bevels, the complexity of the data points may be challenging, as now you have Empiric data that you are trying to find a formula to fit.

then again, once you get beyond Bernoulli's principle i am pretty much lost…..

How much closer to the very complicated truth are would you like to get? If one were to:

• Take one or two or five sample fin shapes;
• Test them as one piece with the edges rounded;
• Cut them and round the new cut edges;
• Then separate them in steps based on some chosen measure - let's say the combined root cord - and test at - let's say - 1%, 2%, 5% , 10%, and then in 10% steps until the changes become insignificant;

then fit a correction curve to the data, I bet it would be pretty good, for some reasonable, useful value of "pretty good".

Yes, that's potentially a lot of data to generate, a lot of wind tunnel runs. If it takes five fin shapes, and 10% steps to 100% before the changes fade away, it would be 65 runs. It's a big task, but I wouldn't call it Herculean. One could reasonably stop there. But...

That's only testing shapes where the leading edge of the trailing fin and the trailing edge of the leading fin fit together, i.e. they are parallel, straight, and have the same length. The whole thing could be repeated with other gap shapes. Maybe five sets, including the one above ( draw that as | | ) plus \|, /|, and \/, and /\. One could do // and \\, but I expect parallel is parallel so those two are not worth doing. That brings the total to 325 runs, so I guess "herculean" could apply.

AND even that is assuming the pairs all have matching spans, i.e. it doesn't account for \/.

So, how much closer to the very complicated truth are would you like to get?

On the other hand, this

Alternatively, knock together a quick and easy 3/4FNC with split fins, and do multiple swing tests with varying nose weights to determine the true CP. Or, for that matter, someone here might have access to an actual wind tunnel.

Then, for OR, use whichever method is demonstrated to give the better answer AND add a warning.

would take somewhere from zero to one wind tunnel runs and no math. A single series of wind tunnel runs would be a good idea so that one can determine when the warning is needed and when it isn't.

 

[BABAR]

Not sure if wind tunnel tests are similar to other experiments, but for many empirical data sets, you need to do MULTIPLE runs at each setting and take an average rather than just 1 at 1/8”, 1 at 1/4”…… You may need to do 5 or 10.

also need to consider the surface prep. Open Rocket allows for this, I imagine however if would be even more complex with gap fins. Not sure I exactly know what ”boundary layer” and “laminar flow” really are anymore, it’s been 40 years since I took aerodynamic engineering and those neurons are at best rusty if not completely rewritten, but I am guessing they would have a dramatic effect on flow in and around gaps of various sizes.

okay, REAL question. Aside from aesthetic concerns (and I agree gaps look cool), and potential based on lots of discussions of how to achieve “whistling rockets”. (Multiple people have done it, but I have yet to see a published design that EVERYONE says consistently works), WHY does anyone care?

I guess a PRACTICAL experiment would be, for the same total surface area, if you throw in a gap of any size, but keep, say, the “center of lateral area” of the fin as a whole on the same place, does it make the rocket LESS stable. my MindSim says ”no”, I think the extra drag would likely make it as stable if not more stable, but I have discovered lots of things are counterintuitive. I’d also guess that the additional imparted drag would make it less efficient from a performance standpoint, so only reason I can see to do it is for cosmetic reasons or to get a whistle (and please if anyone has a design for low power motors that consistently whistles, publish it! I have been chasing that Unicorn for a while.)

 

[jqavins]

I'm pretty sure that the look and the change of maybe getting a whistle are the only reasons in our level of stuff. In "real" missiles they are also used for steering, where one part pivots and the other is fixed.

 

[SolarYellow]

In addition to planform variations, one would have to consider the thickness of the fins, the profiling shape of the fins (square, rounded, airfoiled, details of airfoiling, etc.), angle of attack of the rocket body, angle of the fins (3 fin vs. 4 fin and relationship to angular orientation of departure from flight aligned with the rocket centerline), and probably a bunch of other stuff I can't spout off the top of my head.

For our hobby world, I'm down with getting close in design and swing testing the final result. Be conservative if you don't want to rebuild. If you're going for maximum performance and can't afford to be conservative, you're probably going with 3FNC or 4FNC in any case.

Consider building a quick and dirty mockup with smaller, standard body tube for cheap before you commit big composite parts and tens or more man-hours to a build.

 

[Jeff Lassahn]

okay, REAL question. Aside from aesthetic concerns (and I agree gaps look cool), and potential based on lots of discussions of how to achieve “whistling rockets”. (Multiple people have done it, but I have yet to see a published design that EVERYONE says consistently works), WHY does anyone care?

Situations that look a lot like gapped fins occur in multistage rockets. So figuring out how much the sustainer fins derate the performance of the booster fins is a nice capability for OR to have.

 

[BABAR]

Situations that look a lot like gapped fins occur in multistage rockets. So figuring out how much the sustainer fins derate the performance of the booster fins is a nice capability for OR to have.

Somewhere I read that for same number fins on booster and sustainer (so say three fins each) that for maximum STABILITY you wanted the fins completely disconjugated (so from front would look like 6 fins). This is highest drag. For best PERFORMANCE you put them IN-line (so looks like three fins).

Of course, all the best performance in the world doesn’t help if rocket is unstable

 

[jqavins]

Situations that look a lot like gapped fins occur in multistage rockets. So figuring out how much the sustainer fins derate the performance of the booster fins is a nice capability for OR to have.

Oh, now that's a good point.

In addition to planform variations, one would have to consider the thickness of the fins, the profiling shape of the fins (square, rounded, airfoiled, details of airfoiling, etc.), angle of attack of the rocket body, angle of the fins (3 fin vs. 4 fin and relationship to angular orientation of departure from flight aligned with the rocket centerline), and probably a bunch of other stuff I can't spout off the top of my head.

All of those things have measurable effects even without split fins, and OR and RS either don't count them or count them very little of it to begin with. (As far as I know, changing the airfoil shape is accounted for in drag, but not in CP.) That means that all that is in the noise as far as improving OR's and RS's results. Any effect of those factors on performance is most certainly in the noise compared to motor variations. This conversation started with weird behavior of the software with regard to the CP.

 

[Alan15578]

Situations that look a lot like gapped fins occur in multistage rockets. So figuring out how much the sustainer fins derate the performance of the booster fins is a nice capability for OR to have.

You could calculate or estimate the downwash angle induced by the sustainer fins at the location of the booster fins, and use that to decrease the Cn(alpha) of the booster fins However, you are probably already covered for that level of detail by your simple Barroman CP static margin.