Counter-intuitive AeroFinSim Lite result

[Adrian A]

I decided to check out FinSim before committing to the fins of my new build.

Looking into it, I'm finding a very counter-intuitve result that makes me very skeptical about its answers.

My current design looks like this:



and I get results for divergence of Mach 2.51 and flutter at Mach 2.31 using the NACA TN-4197 method.

But if I drastically shorten the root chord, which should reduce the torsional stiffness, the critical velocities went up 3.28 and 5.53 for the divergence Mach and the the flutter threshold. Really? The fin dimensions below are is less flutter prone than the fins above?




Playing with this some more, increasing the root chord length consistently decreases the critical velocity. Increasing the tip chord consistently increases the critical velocity. Both of those have to be incorrect, unless I'm completely missing something. Is it possible he got those two geometry values swapped?

Looking at the additional results, the bending natural frequency is constant regardless of tip chord and root chord, which is clearly wrong since it should be a function of both. If the tip chord is larger than the root chord, the natural frequency would be much lower than if the fin is the other way around. Also, the torsional natural frequency goes down with increasing tip chord (correct) and also goes down with increasing root chord (incorrect).

Am I missing something or are the stiffness calculations needed to find the critical speeds in FinSim just fundamentally broken?

 

[jderimig]

If it were me i would ask the author of the software to explain because it doesn't make sense to me.

 

[user 30299]

Adrian,

When using FinSim Lite be sure to check the analysis method under the CN-alpha tab.

When you first open the program it's usually set on the Classical 2-D Lift Slope method.
The results you're getting suggests it's the 2-D Slope method.

You should select NACA TN 4197 Method under CN-alpha.

 

[Adrian A]

Adrian,

When using FinSim Lite be sure to check the analysis method under the CN-alpha tab.

When you first open the program it's usually set on the Classical 2-D Lift Slope method.
The results you're getting suggests it's the 2-D Slope method.

You should select NACA TN 4197 Method under CN-alpha.

Changing the CN-alpha changes the lift model, and the resulting critical velocities, but it doesn't do anything for the stiffness shown in the "additional results" which is where I think there are major problems.

 

[jderimig]

I wonder if the algorithm doesn't model the fin as a beam but just models the "drum skin" natural frequencies? Need clarity from the author.

 

[Adrian A]

I looked up an equation for torsional stiffness, and for a thin beam with chord length a and thickness b, the twist angle is

torque * length / (modulus * 1/3 * chord length * thickness^3)

where "length" would be the height of the fin.

http://www.eng.cu.edu.eg/users/amamansour/TorsionME.pdf
The torsional stiffness would then be modulus * chord length * thickness^3 / (3 * height). This would be directly applicable for a rectangular fin. For a normal trapezoidal fin, I'd need to dredge out the calculus part of my brain to come up with the equation, but basic reasoning would say that since the lift is generated along the whole fin height, the base of the fin will see the torque of the whole fin, and the parts farther from the base will see less torque, and so there would be a stiffness benefit for having the fin at the root be longer and/or thicker than it is at the tip.

 

[Adrian A]

NACA Tech note 4197 is relevant
https://ntrs.nasa.gov/citations/19930085030
A key formula from page 12 of that says that

"Coleman (ref. 11) has given the frequency of a tapered beam of constant thickness ratio in terms of the frequency of an untapered, uniform beam having the same root chord as

tapered natural frequency ~= rectangular natural frequency * (1 + 1.87 * (1 - (tip chord) / (root chord)) ^ 1.6.

So for a triangular fin with 0 tip chord, the natural frequency of the fin would be 2.87^1.6 or 5.4 times the natural frequency of a fin with rectangular profile. If I plug in a rectangular fin, where the tip chord and the root cord are the same, then the equation cancels out to give a value of 1, as expected.

So for my fin with 0.8" tip chord and 4.7" root chord, the torsional natural frequency should be (1+ 1.87 * .829)^1.6 = 4.47 times the natural frequency if the tip chord were 4.7" long. If the fin were rectangular with 4.7" chord length, 0.04" thickness and 2" span height, the torsional natural frequency is 511 Hz. If I decrease the tip span to 0.8" long, it should go up to 2287 Hz. FinSim shows it going to 871 Hz instead.

I will probably dig deeper into that NACA 4197 later. It appears to have all the relevant information.

 

[mikec]

Purely on the basis of "looks good, flies good" it seems to me like you have too much sweep. I understand the desire to move the CP back as far as you can, but I'd lengthen the root aft for better support.

I've always wished there was a robust method to assess flutter. And good luck understanding the real properties of a composite layup. We still have challenges there in big-money pro aerospace.

 

[kjhambrick]

I have to ask Adrian ...

Are you studying fin flutter as a preventive measure to optomize the structural integrity of a rocket on the drawing board ?

Or have you seen inertial data suggesting that fin flutter might be stealing altitude from your flights ?

If so, are you ready to share what you've seen ?

Thanks

-- kjh

 

[Adrian A]

I have to ask Adrian ...

Are you studying fin flutter as a preventive measure to optomize the structural integrity of a rocket on the drawing board ?

Or have you seen inertial data suggesting that fin flutter might be stealing altitude from your flights ?

If so, are you ready to share what you've seen ?

Thanks

-- kjh

I'm about to cut fins for my StratoSpear sustainer and trying to plan for whether I will need to do a tip-to-tip layup to stiffen them. I'd like to get my sustainer working this week and fly it next weekend on a single-stage Mach 2.8 flight if the launch isn't cancelled again.

 

[robopup]

Here's a table I just put together for my project, looking at a couple methods for simulating flutter speeds vs demonstrated flight speeds on a few of my rockets. In general, every fin flutter method I've tried is conservative to the point it is useless.

The fins on all of these had a similar profile to yours, just a bit less sweep (but trailing edge of tip was still aft of trailing edge of root).

	Fin Count	Fin Laminate	Fin Total Thickness (in)	E1/E2 (psi)	G12 (psi)	Altitude at max speed (MSL ft)	Static pressure at max speed (psi)	Dynamic pressure at max speed (psi)	Howard Flutter Prediction (ft/s)	Finsim (Pine's Method Prediction) (ft/s)	Demonstrated Max Flight Speed (ft/s)	Simulated Max Flight Speed (ft/s)
38mm J570	3	Uni T700 Quasi-iso	0.045	7.40E+06	2.80E+06	5900	11.8	50	1337	1276	2755	N/A
54mm K250	3	Uni T700 Quasi-iso	0.045	7.40E+06	2.80E+06	14600	8.4	27	1115	1085	2400	N/A
54mm L800	3	G10	0.1	2.80E+06	4.20E+05	7500	11.1	43	1094	1800	2610	N/A
75mm M1100*	3	G10	0.125	2.80E+06	4.20E+05	13400 (sim)	8.8	54	1129	1718	3300 (sim)	N/A

*75mm M1100 was never recovered (only had directional radio tracker), but I'm confident it did not shred - nominally flight time to apogee and total time of flight. All other flights use data recovered from altimeters.

 

[jderimig]

Here's a table I just put together for my project, looking at a couple methods for simulating flutter speeds vs demonstrated flight speeds on a few of my rockets. In general, every fin flutter method I've tried is conservative to the point it is useless.

The fins on all of these had a similar profile to yours, just a bit less sweep (but trailing edge of tip was still aft of trailing edge of root).

	Fin Count	Fin Laminate	Fin Total Thickness (in)	E1/E2 (psi)	G12 (psi)	Altitude at max speed (MSL ft)	Static pressure at max speed (psi)	Dynamic pressure at max speed (psi)	Howard Flutter Prediction (ft/s)	Finsim (Pine's Method Prediction) (ft/s)	Demonstrated Max Flight Speed (ft/s)	Simulated Max Flight Speed (ft/s)
38mm J570	3	Uni T700 Quasi-iso	0.045	7.40E+06	2.80E+06	5900	11.8	50	1337	1276	2755	N/A
54mm K250	3	Uni T700 Quasi-iso	0.045	7.40E+06	2.80E+06	14600	8.4	27	1115	1085	2400	N/A
54mm L800	3	G10	0.1	2.80E+06	4.20E+05	7500	11.1	43	1094	1800	2610	N/A
75mm M1100*	3	G10	0.125	2.80E+06	4.20E+05	13400 (sim)	8.8	54	1129	1718	3300 (sim)	N/A

*75mm M1100 was never recovered (only had directional radio tracker), but I'm confident it did not shred - nominally flight time to apogee and total time of flight. All other flights use data recovered from altimeters.

How do you know you didn't have flutter?

 

[robopup]

How do you know you didn't have flutter?

You're right, I don't. But I think a lot of the time, both on the forums and in the tools themselves, anything past the flutter speed is branded as "thar be dragons." For example, Finsim:


I do know I didn't shred at 2x and sometimes 3x the predicted flutter speed on the selection of flights in my table, so for structural design of fins I struggle to see the value in paying attention to these predicted flutter speeds. Particularly when the tools also seem questionable to do relative comparisons (this fin has a flutter speed 10% higher than that fin) as Adrian pointed out.

 

[TonyL]

Something to keep in mind is that those equations are only really valid for rectangular, flat plate fins, and any variation from that leads to some degree of model inaccuracy [but better than nothing]. If you want to exercise the tool, the NACA format data is the more correct way to go about it. That however requires some FEA results, and the elastic axis never seems to occur like it is shown in the diagram, so one gets to think about that too...
I have not used the Lite version, so the NACA format may or may not be implemented.

br/

Tony

 

[Maxwelljets]

I decided to check out FinSim before committing to the fins of my new build.

Looking into it, I'm finding a very counter-intuitve result that makes me very skeptical about its answers.

My current design looks like this:

View attachment 594782

and I get results for divergence of Mach 2.51 and flutter at Mach 2.31 using the NACA TN-4197 method.

But if I drastically shorten the root chord, which should reduce the torsional stiffness, the critical velocities went up 3.28 and 5.53 for the divergence Mach and the the flutter threshold. Really? The fin dimensions below are is less flutter prone than the fins above?


View attachment 594784

Playing with this some more, increasing the root chord length consistently decreases the critical velocity. Increasing the tip chord consistently increases the critical velocity. Both of those have to be incorrect, unless I'm completely missing something. Is it possible he got those two geometry values swapped?

Looking at the additional results, the bending natural frequency is constant regardless of tip chord and root chord, which is clearly wrong since it should be a function of both. If the tip chord is larger than the root chord, the natural frequency would be much lower than if the fin is the other way around. Also, the torsional natural frequency goes down with increasing tip chord (correct) and also goes down with increasing root chord (incorrect).

Am I missing something or are the stiffness calculations needed to find the critical speeds in FinSim just fundamentally broken?

Were you changing the fin CG location? That's quite important to get it to spit out the correct answers. Here's a fin CG calculator I wrote to aid with using finsim.

 

[Buckeye]

I do know I didn't shred at 2x and sometimes 3x the predicted flutter speed on the selection of flights in my table, so for structural design of fins I struggle to see the value in paying attention to these predicted flutter speeds. Particularly when the tools also seem questionable to do relative comparisons (this fin has a flutter speed 10% higher than that fin) as Adrian pointed out.

Totally agree. If the tool can't do reasonable A-B comparisons, then the absolute numbers it spits out are rather meaningless.

I assume you are using an accelerometer altimeter to get max speed and not a baro altimeter? The speed from baro will be wildly inaccurate. For sake of completeness, it would be nice to see the simulated speeds (preferably RASAero II) in your table, just to verify that the measured speeds are in the right ballpark.

This has been an enlightening thread. If the observations hold true, then add FinSim Lite to the heap of over-building myths that flourish on the forum.

 

[jderimig]

Love to model using FinSim and fly a rocket with strain gages on the fins and compare.

 

[user 30299]

Here's link to a thesis that does a decent deep dive into fin flutter.

https://ntrl.ntis.gov/NTRL/dashboard/searchResults/titleDetail/ADA502110.xhtml

"Aeroelastic Optimization of Sounding Rocket Fins"​

Worth reading ( or skimming ) before pushing FinSim to the side of the road.

As for fins not shredding at 2x or 3x the predicted speed, we don't know if the altimeter's max speed
was momentary or sustained. We also don't know how long a fin can sustain a flutter before it degrades
the fin's layup or it's attachment to the body tube.

We have all seen some exciting shreds, but how many where due to poor construction, or severely under-designed
fins in the first place. Some fins have an immediate shred, and some get a bit of time/altitude before shredding,
and why is that.

"Postmortems" are always helpful but not always done. The above thesis offers a pretty good postmortem.

 

[jderimig]

^ flutter != mechanical failure. The fin can flutter but stress < strength can be true for the entire flight.

 

[user 30299]

^ flutter != mechanical failure. The fin can flutter but stress < strength can be true for the entire flight.

What's I like are those flight videos that have a clear picture of the fins as they begin to flutter.

 

[robopup]

Totally agree. If the tool can't do reasonable A-B comparisons, then the absolute numbers it spits out are rather meaningless.

I assume you are using an accelerometer altimeter to get max speed and not a baro altimeter? The speed from baro will be wildly inaccurate. For sake of completeness, it would be nice to see the simulated speeds (preferably RASAero II) in your table, just to verify that the measured speeds are in the right ballpark.

This has been an enlightening thread. If the observations hold true, then add FinSim Lite to the heap of over-building myths that flourish on the forum.

Well aware of effects of baro. These max speeds are pulled from kalman filtered baro and accelerometer data. I can also pull max accelerometer based but it's about the same. Here is an updated table with Rasaero sims (although a couple of these flights were 2 years ago now, I believe I had final numbers entered properly for sims but you know how it goes...).

Flight	Fin Count	Fin Laminate	Fin Total Thickness (in)	E1/E2 (psi)	G12 (psi)	Altitude at max speed (MSL ft)	Static pressure at max speed (psi)	Dynamic pressure at max speed (psi)	Howard Flutter Prediction (ft/s)	Finsim (Pine's Method Prediction) (ft/s)	Demonstrated Max Flight Speed (ft/s)	Simulated Max Flight Speed (ft/s)
38mm J570	3	Uni T700 Quasi-iso	0.045	7.40E+06	2.80E+06	5900	11.8	50	1337	1276	2755	2425
54mm K250	3	Uni T700 Quasi-iso	0.045	7.40E+06	2.80E+06	14600	8.4	27	1115	1085	2400	2275
54mm L800	3	G10	0.1	2.80E+06	4.20E+05	7500	11.1	43	1094	1800	2610	3000

Here's link to a thesis that does a decent deep dive into fin flutter.

https://ntrl.ntis.gov/NTRL/dashboard/searchResults/titleDetail/ADA502110.xhtml

"Aeroelastic Optimization of Sounding Rocket Fins"​

Worth reading ( or skimming ) before pushing FinSim to the side of the road.

As for fins not shredding at 2x or 3x the predicted speed, we don't know if the altimeter's max speed
was momentary or sustained. We also don't know how long a fin can sustain a flutter before it degrades
the fin's layup or it's attachment to the body tube.

We have all seen some exciting shreds, but how many where due to poor construction, or severely under-designed
fins in the first place. Some fins have an immediate shred, and some get a bit of time/altitude before shredding,
and why is that.

"Postmortems" are always helpful but not always done. The above thesis offers a pretty good postmortem.

Thanks for the link, looks like a nice summary (although just looking at their fin shape screams flutter to me). All the flights in my table were over the predicted flutter speed for at least 5 seconds, and other than the 38mm, at least 10 seconds. Agreed, I think a lot of shreds that get chalked up to fin flutter were simply poor constructions - a little crooked, poor quality laminate on tip to tip or fin stock etc.

^ flutter != mechanical failure. The fin can flutter but stress < strength can be true for the entire flight.

Exactly.

I don't know enough about flutter or these analytical equations to know where exactly we're outside their assumptions, but I'd be willing to bet that for our purposes there are several. I also think these analytical equations are just too simplistic to capture what we actually care about - structural failure. One off the top of my head which bothers me is that you simply input a "shear modulus". Most of our fins are composite laminates, and using the actual stiffness matrix rather than a homogenized laminate G totally changes the answer on these frequency type problems.

This is a hard enough problem to properly simulate that, at least for me, it's more instructive to do flight tests instead.

 

[kjhambrick]

^ flutter != mechanical failure. The fin can flutter but stress < strength can be true for the entire flight.

Yes, but since there is only one source for the energy that goes into fin bending, will flutter not 'steal' altitude, even if the fins don't fail ?

-- kjh

 

[robopup]

Yes, but since there is only one source for the energy that goes into fin bending, will flutter not 'steal' altitude, even if the fins don't fail ?

-- kjh

Good point, and probably if the fin is actually fluttering. But enter the fin thickness required to satisfy these fin flutter programs and watch what your altitude does. I think like most rocket design, it's all about tradeoffs.

 

[kjhambrick]

Silliness: One thing I've wondered on a rocket this size is if someone with perfect pitch or better yet, a frequency analyser, coldn't measure the resonant frequency of a fin by plucking a fin mounted on the rocket or on a solid base ?

 

[watheyak]

Silliness: One thing I've wondered on a rocket this size is if someone with perfect pitch or better yet, a frequency analyser, coldn't measure the resonant frequency of a fin by plucking a fin mounted on the rocket or on a solid base ?

I tune mine all to B sharp.

 

[Adrian A]

Well aware of effects of baro. These max speeds are pulled from kalman filtered baro and accelerometer data. I can also pull max accelerometer based but it's about the same. Here is an updated table with Rasaero sims (although a couple of these flights were 2 years ago now, I believe I had final numbers entered properly for sims but you know how it goes...).

Flight	Fin Count	Fin Laminate	Fin Total Thickness (in)	E1/E2 (psi)	G12 (psi)	Altitude at max speed (MSL ft)	Static pressure at max speed (psi)	Dynamic pressure at max speed (psi)	Howard Flutter Prediction (ft/s)	Finsim (Pine's Method Prediction) (ft/s)	Demonstrated Max Flight Speed (ft/s)	Simulated Max Flight Speed (ft/s)
38mm J570	3	Uni T700 Quasi-iso	0.045	7.40E+06	2.80E+06	5900	11.8	50	1337	1276	2755	2425
54mm K250	3	Uni T700 Quasi-iso	0.045	7.40E+06	2.80E+06	14600	8.4	27	1115	1085	2400	2275
54mm L800	3	G10	0.1	2.80E+06	4.20E+05	7500	11.1	43	1094	1800	2610	3000

This is great, thanks.

I think I have seen flutter in the data from one of my 29mm flights with thin fins and Mach 1.1 or so. It showed up in the gyro and accelerometer data as a high-frequency oscillation. I'll see if I can find it.

Edit: Here it is:

https://www.rocketryforum.com/threa...hread-inertial-navigation.172320/post-2347707

 

[jderimig]

Silliness: One thing I've wondered on a rocket this size is if someone with perfect pitch or better yet, a frequency analyser, coldn't measure the resonant frequency of a fin by plucking a fin mounted on the rocket or on a solid base ?

We used to do this for material char with a strain gage. Not only can you get resonant freq but damping coefficient also.

 

[kjhambrick]

We used to do this for material char with a strain gage. Not only can you get resonant freq but damping coefficient also.

Yes !

It's been years but I was a cable monkey for a an old timer mechanic-turned technician who ran a vibration analyser on gas turbines in the late '70s and early 80's ...

He was a real artist with 'perfect pitch' -- he could ID problems by the signature of the output on the scope and sometimes he could tell more than he could see on the scope by listening to various spots on the machine thru the handle of a screwdriver.

-- kjh

 

[OverTheTop]

Yes, but since there is only one source for the energy that goes into fin bending, will flutter not 'steal' altitude, even if the fins don't fail ?

-- kjh

Maybe not as much as you might think. It will cause more drag with the fin not being at zero AoA, but the resonance itself might need little input. That's how resonance works and why massive failures start with small excitations.
https://en.wikipedia.org/wiki/Resonance

Silliness: One thing I've wondered on a rocket this size is if someone with perfect pitch or better yet, a frequency analyser, coldn't measure the resonant frequency of a fin by plucking a fin mounted on the rocket or on a solid base ?

Absolutely. We have used this to characterise voice-coils in interferometers in FTIR spectrometers, and then tune a notch filter in the control loop.

I'll also add that driving through the resonance frequency is a way of dealing with it. If you are not there long enough the resonance levels can stay below destructive.

 

[kjhambrick]

Maybe not as much as you might think. It will cause more drag with the fin not being at zero AoA, but the resonance itself might need little input. That's how resonance works and why massive failures start with small excitations.

Dooh !

If course, that's what resonance is all about, isn't it ?

Absolutely. We have used this to characterise voice-coils in interferometers in FTIR spectrometers, and then tune a notch filter in the control loop.

OT Alert !

Sigh ... nostalgia again !

I miss the old Baird Atomic Differential IR that we ran at a Lube Oil Testing Lab where I worked as an intern in the early 70's.

We eventually got an IBM FTIR in the early 80's which did all the tedius stuff for us but it was very satisfying to be able to filter out the known components in a sample, a little at a time 'by hand' by identifying them in 'the book' ( dang, I've forgotten the name of that 6-inch thick book of IR Scans )

Dr Jordan, my boss, taught me a ton and he was fun to work with on that spectrometer !

-- kjh