Flap force simulation

Hi all,

I'm working on simulating the hydrofoil flap forces for a Moth and am struggling to get believable results, especially when modeling high flap deflections.
Here's what I've tried so far:

1. Basic Servo Equations: I initially started with some simple servo sizing equations. This approach gave highly inconsistent torque estimates, with values ranging anywhere from 20-80 Nm. I suspect this is because these models are too simple and don't account for the hydrofoil's geometry or, more importantly, stall.
   
   
2. XFLR5: My next step was to use XFLR5. Unfortunately, it fails to converge for any flap deflection greater than about 10 degrees.
   
   
3. AVL (Athena Vortex Lattice): I then moved to AVL, hoping to get more stable data. While AVL does solve, I'm skeptical of the results. It's producing $C_L$ values as high as 2.3, which seems unrealistic. I'm not convinced it's accurately capturing stall behaviour (as expected from a vortex lattice method).
   

This brings me to my main question:

Does anyone have empirical data for Moth foils (or similar high-performance foils) that I could use to validate my AVL results?

Alternatively, am I using the wrong tools for this? I'd appreciate any advice on a better approach to model these high-deflection cases where stall is clearly a significant factor.

Thanks in advance.

 

Have you seen this moth tank testing paper ?
https://www.boatdesign.net/attachments/csyspaperfeb09-beaver-paper-on-moth-pdf.136024/

openFoam is slow and hard to validate too, but instructive to play with. Do you have a precise 3d geometry to share (Solid without naked edges in Rhino or a high-res .stl, perhaps at the AoA and flap angle you expect it to stall) ? Perhaps you have access to a university cluster or a serious workstation ? I could give it a quick try in my openFoam 13 workflow tuned for much larger wings, and pass it on to you from there if the basics check out. You'd only have to swap geometries and adapt speeds for each run.

 

My CAD model is still in its infancy.
The plan is to machine the center part out of 7075 aluminum or 316 stainless steel,
and then have the carbon fiber wing sections just plug straight into it.
I know that joint where the wings meet the main vertical part is the most critical spot.
It’s going to be under a ton of stress, and it will need a lot of FEA just to have any hope of surviving.

I was wondering about simscale to model this interface.

I spotted OpenFOAM, but I basically wimped out when rI saw it ran on Ubuntu. Trying to wrestle with Docker, Ubuntu, and programming all at once is just way out of my comfort zone.

I'll re read the beaver paper, I hadn't spotted the raw data at the end..

 

If the 6 pieces in your CAD fit well enough to merge (boolean union) into a single solid, there should be no problem throwing that at openFoam. And the FEA could probably wait till you are happy with the hydrodynamics

Indeed, due to abyssal documentation, slow iteration due to long calculation times, lack of really usable workflows for appended fast boats, and the total lack of coordination if not outright chapel wars in the openFoam community, it does take months to figure out from scratch. Or perhaps weeks with a polished-up generalist interface like Simscape, at the end of which you've mostly learned their ergonomy and start to get locked into their monetization scheme.

That's why i was proposing to take over my raw but case-similar scripts. Then it would just be a matter of doing a standard openFoam 13 install on Ubuntu (= pasting 2-3 lines in a terminal), copying over an .stl in the reference case, (perhaps changing speed, water properties inc. turbulence level in a couple text files), launching my script, waiting a couple days, then watching it on paraview using my templates.

But since i can't block mine for weeks, you'd need access to a modern workstation or HPC node. That would mean Xeon or Epyc cpu's of the last couple years with 8-12 channels of actually populated ECC ram each. (CFD is memory-bandwith constrained, and it needs the ECC ram error correction because with the enormous quantities of data handled and a real need for coherence, any little bit getting lost will crash the calculation. So forget about running that stuff on gamer hardware. I've lost thousands and months trying ). Think 6-10k€ of hardware, or twice as much if you just call IBM or HP, and 0.5-1 kW continuous power usage + fan noise at the levels of an industrial vacuum cleaner under load.

Cheers and all the best for the project !

 

I'm not sure why you would need to block your CFD machine for weeks for a 'simple' foil? Is you PC from the 90's?

If the sectional shape does not really change you could even simplify the simulation to 2D to retrieve an estimation of all the necessary coefficients.

Would be lucky to get a a drawing or 3D model to help out with the calculation.

Interesting paper though!

Edit: For the case you want to invest in CFD hardware, you also don't need a machine worth thousands of thousands dollars. I'm happy with my machine 20 cores 128GB RAM for ~400€.

 

Just for turnaround time, and to limit the attention spent babysitting the computation to an amount where you feel like working on a wing as opposed to optimising a CFD process.

The development is going to need quite some runs for various AoA's, flap angles, versions, speeds, and if you want to account for the free surface, it'll take a few seconds of simulated time for the wave system to build every time. Also if you don't want to invest massive learning time to become able to simplify while retaining confidence, you'll lean towards the brute force side of resolutions and timesteps/courant numbers.

Sure, 2D or very coarse 3D can be run on a laptop, but i wouldn't quite trust that for figuring out where a stall starts etc. Is it really worth it compared to the panel methods then ?

Or perhaps you really have a much better starting point than i extrapolated from the planing boat W3 tutorial
Maybe one day one of us is going to get a half-decent foiling boat tutorial into the distribution?

 

The challange is the inteception between foils, the bulb in between.
If the paper is correct and it accounts for 50% of the drag its well worh the optomisation.

Be great to try your scripts.. I'm going to need to build a full model anyway.

It's a CAD station, It's okay.. I9-14900K, 64gb memory, Shame it won't use the GPU as that's a Quadro RTX 5000

The model starts as one lump then gets chopped up. I'll need to be able to change the flap angle anyway, then remerge into a solid.

 

When you go for CFD, use 2D simulations and you can run multiple simulations with different speeds, angles of attack, changed flaps, flapangles, with/without free surface, etc. in a fraction of the time of that of a 3D simulation of the whole foil. Sure, then you're gonna miss the strut(s) and foil tips. However there's no need to run multiple 3D calculations to get an general indication of the results. Lateron 3D calculation for one or a few setups can be performed to get a correction factor between 2D and 3D.

Keep it simple

 

The openfoam installation comes together with several tutorial for foils, and, the internet is flooded by tutorials about foils. Simply adapt a tutorial case, change the medium to water and eventually apply multiphase flow to include free surface effect. I insist and only can recommend my suggestion of going 2D instead of 3D. 3D would not only take much longer calculation time but also make it harder and things would get more complicated for you.

 

What software would you recommend XFLR5 / xfoil won't solve above about 10 degrees flap angle (Main foil at 4 degrees) ?

AVL always solves , but I'm getting some crazily high CL values and massive hinge torque values.. you won't get a CL value of 2.3 from a simple single element foil.. (exluding your multi element Formula 1 monster foils)

 

Well, the RAM is enough for the case but th I9 has 2 memory lanes instead of the 24 of my twin-Epyc server. That means you'll run about 12 times slower, like a week 24/7 where i'd expect overnight runs for a 3d case like that. It might indeed be more realistic to start with 2d sections like MC_Rash suggested then.

(i can run a test in my 3d workflow nevertheless if you send me geometry. I'd need to tweak it slightly and see how it behaves before passing it on, else it's gonna be hours of explanations and debugging of your first runs instead of 10mn of tuning.)

 

Give me a week to update the CAD model, Do you want a challanging Flap angle, something like 20 degrees?

 

It might be faster to build and test it.. fasten it to sheet of plywood behind a boat, with some strain guages.. possibly 3d print some different bulb shapes..

I'm use to FEA where i can do 5 different layups and geometry tweeks in a morning..

 

Cool! 20° flap sounds challenging, do known moths go that far ? Perhaps send me a second version with moderate or neutral flap angle too then.

 

Just sent you a link to the step files.