Ultra heavy tailsitter key parameters?

ArduPlane
1

Hello everyone, I was wanting to design a very heavy dual motor tailsitter in excess of (100-500 lbs || 45 - 227 kg). It would use internal combustion engines (ICE).

Can anyone please comment a parameter or two that I should focus on that relates to the excessive weight? Or perhaps just any parameter you think i should focus on…

Thanks!

using: pixhawk 2.4.8

2

copying my reply from the other topic, thanks for starting this as a new topic.

That will be quite a machine! Got any pictures?
Note that none of us have experience with tailsitters of that size, and this would be the first ICE tailsitter with ArduPilot that I am aware of.
If you have funding to pay for consulting then I’d suggest Leonard Hall or Paul Riseborough. I can give contact details if needed.
Otherwise here is what I suggest:

  • build a good RF9 model. The model will almost certainly not allow for getting a good tune with your real vehicle, but it will likely get you in the ballpark and should allow you to find any major issues with this scale of vehicle. Concentrate on the mass distribution in the model and the motor response curves and timings
  • ensure you have a really good isolation system, and test it with a tethered test stand, ensuring the EKF stays completely happen with motors running
  • build a really good hanging tether for testing. Constrain it to one axis at a time initially. Have a single simple motor kill switch on RC transmitter and make sure that tether can handle large forces
  • take it slowly and carefully, documenting your experiments as you go. Expect it to take dozens (possibly hundreds) of short flights in the tether to get a good tune.
  • do not try a fwd transition until the QSTABILIZE flight is really solid, including rapid non-oscillating recovery from large disturbances

and most importantly, feed us with lots of videos :slight_smile: Maybe start your own blog post with updates after each test?
Cheers, Tridge

3

Hi tridge, I’m resposting here in case you wanted to respond here instead of in the dual motor tailsitter topic.

Tridge, thanks for your response, I know all developers are busy people.

I don’t have any pictures, nothing is built yet (except for a really ugly bent up plank that I’ve been using to learn ardupilot).

I may take up Mr. Hall and Riseborough’s help at some point, I would like their contact information if you would.

I’m not familiar with any flight simulation software, would RF9 be able to estimate rotation rates based on the size of the elevons? Or perhaps it doesn’t allow that kind of customization. Very good tutorial you made here btw: https://ardupilot.org/dev/docs/sitl-with-realflight.html

I will indeed focus on minimizing vibrations with isolation mechanisms.

Good advice on the tether, I have a switch that disarms the craft, that will always kill the motors no?

I will share what I can.

I have a thing about fwd transitions and FBWA, I don’t trust these modes mainly because I don’t have yaw stabilization (everytime I switch to FBWA my plank yaws out of control, I won’t be using fins so I need yaw stabilization). I’m actually able to fly the plank at high aoa’s in Qstabilize (imitating fwd flight) by increasing the Q_angle_max parameter. There’s significant oscillation at around 90 deg though, I need to try at 80 and see what happens. It’s also difficult to turn unless I ease up on the angle to around 70 deg. It’s to be expected, but I kind of want to be able to have it as my back up mode if Qacro doesn’t work.

Tridge, really important question for you- when in Qstabilize does the angle calculating algorithm use the GPS at all to help correct its orientation (or estimate its attitude)? I am aware that it does use an airspeed sensor to scale gains as discussed here https://ardupilot.org/plane/docs/qacro-mode.html

I think QACRO provides rate based yaw stabilization, so I think I’ll mess around with that mode and see what happens. Why doesn’t FBWA support differential thrust yaw stabilization anyways?

Any insight would be much appreciated, thanks

Shef

4

you’d need to enable the yaw damper and differential thrust for yaw control. This would be a good thing to get right in the simulator.

approximately, yes

that depends on the ignition setup. See the wiki page on ICE setups.

yes, GPS velocity is important got attitude estimation

it does, but you need to enable it. Planks without fins are not common, and may need some more work as well.

5

Hi Tridge, thanks for your response. I now can observe yaw stabilizing outputs to motor left and motor right in the servo output. Now I just need to tune. It’s snowing outside, so guess I’ll have to wait…lol.

Here are the settings I changed (JeanPhilippeHell recommended these specific ones)

KFF-RDDRMIX: 0 WAS .2
YAW2SRV-SLIP: 0 WAS 0
YAW2SRV-DAMP: 0.08 WAS 0 (although a higher value may be necessary)
RUDD-DT-GAIN: 100 WAS 10

Wow, I wouldn’t have thought a mere flight simulator could estimate rotation rates based off elevon size, that means it incorporates computational fluid dynamics (CFD), or at least mimics it. Amazing, I’m much more interested in RF9 now.

I’ll be sure to see the ignition setup in the ICE wiki, need to have a reliable true motor kill switch.

Hm, ok, I suppose GPS is a must in almost every scenario.

I’ll update here on how FBWA performs, even if it’s a little redundant since Jean has already done it. Maybe I’ll find out something new.

Thanks again Tridge, have a great day.

6

I’m sure it’s not CFD. We don’t really know what the algorithm it uses is, but it is useful for getting ballpark numbers. It won’t get the final tune, but should allow you to experiment with options and help us develop any new features needed.

7

Ah, yes, that makes more sense.

Shef

8

Hey Tridge I was wondering, if you only implement the “physics” for a model in RF9 and forget the rest of the model (say you just leave the cat tailsitter visual) would you see an accurate portrayal of how the real thing would fly?

When I say “the physics” I refer to RF9’s wing design as well as motor/engine/prop selection and all of the parameters and variations that go along with it.

I’ve been learning a lot about SITL and RF9 over the past week since you recommended it, it is a powerful tool! Thank’s for taking the time to tell me about it.

Shef

9

the graphics model also deals with ground interactions. You should at least scale the existing model (there are menu items for scaling graphics and physics)

10

That is a good point, although its good to know that it will fly the same without doing the heavy visual lifting of creating the model. Thank you.

Shef