Beagle Pup FDM

[sanhozay]

As suggested, a topic about my problems with JSBSim in my Beagle Pup model. I see these some, but not all of the time, and iHAM-COL encountered same during the recent USA tour. It may be caused or exacerbated by wind direction and strength.

I haven't done much work on the FDM. I chose JSBSim, not because I have a raft of wind-tunnel data, but because I have a single airframe with three engines and propellers. Yasim would require three different solutions, whereas JSBSim should allow me to create a single aerodynamic, flight control and ground reaction model with different engines.

Pup 100: 100hp Rolls-Royce Continental 0-200, 24 imp gal tanks
Pup 150: 150hp Lycoming 0-320-A2B, 24 imp gal tanks
Pup 160: 160hp Lycoming 0-320-D2C, 34 imp gal tanks

The FDM is generated using Aeromatic++ (the command line tool) using parameters measured from the model. I've added structure contact points and moved the gear to the correct positions. In the absence of a CoG range, CoG is set to the geometric centre of the fuel tanks. I've adjusted the control surface ranges to match specifications and reduced the pitch effect due to elevator to calm down twitchy pitching during turns.

Engines are very, very roughly tuned. I've plugged in numbers from specifications and adjusted the BSFC to get into the ballpark of cruise speed at 3000ft. There's more work to do there -- I know the RPM is coming out too high, for example.

The nagging problem I have is that the aircraft often gets kicked to one side after spawning (with engines not running). Just as the splash screen fades, there is a transient force that can skid the wheels sideways. Sometimes it's a matter of a foot or so, other times, it ends up halfway to the runway edge and there may also be a turning force that can leave the aircraft facing sideways. This was sometimes causing very extreme ground reactions, where the aircraft would tip over and ping hundreds of feet into the air. I think I have reduced this by adding square rebound damping on the gear, which also reduces the tendency to trip over apparently invisible obstacles on taxiways. I have a basic crash detection on the structure contacts which reports the contact and pauses the sim but this just gets annoying when it's a sim-generated crash rather than pilot error.

I think it's worse when it's windy. Probably also worse when that wind is anywhere but from the nose.

[bomber]

Can you get it to reliably reproduce the kick ?

Have you considered replacing the prop with one that produces no thrust ?

[sanhozay]

I can't reproduce it reliably, but once I get it, starting at the same position usually makes it happen. I got a slight kick at EGHI with a 9kt crosswind and no movement at all at EGNM with variable 3kt.

Zeroing out the C_THRUST and C_POWER tables on the propeller didn't change the behaviour, and I confirmed that this had taken effect by starting the engine, applying full throttle and releasing the parking brake with no movement.

In an attempt to try different crosswinds, I tried using the --metar option but even with a strong crosswind it doesn't kick. The wind speed and direction seems to be interpolated slowly when passing a metar string like this and I'm starting to think that there is some sudden transition when using real weather fetch that's nothing to do with the FDM, or perhaps only highlighted by aircraft with low inertia.

So I tried the Dragonfly, which is JSBSim and much lighter than the Beagle Pup, and it does similar. The default Cessna does the same, but only nods its nose slightly. Yasim aircraft (I tried the Navion and 14bis), don't seem to move at all.

That's about all I have time to do today, and I'm away until Friday. When I get back, I think I'll set up some logging of the wind speed at startup and try and tally that with the forces on the model.

[bomber]

Jsbsim does a trim routine.... you can normally see the results of it in the console window.... is it trimming before spawn


Also a simple question.... how do I get hold of it so as I can look at the engines and prop configuration for when you get back on Friday.

Also any data folder for the engines and prop configuration you're using to base the xml files on would help me having to search the net and taking a punt.

Cheers

[sanhozay]

Jsbsim does a trim routine.... you can normally see the results of it in the console window.... is it trimming before spawn

This is what I got with a big kick to the left (I've stripped out the scenery load messages):

Code: Select all

Starting and initializing JSBsim
T,p,rho: 518.67, 2116.22, 0.0023769
Start common FDM init
...initializing position...
FGJSBsim::set_Longitude: -0.0236113
FGJSBsim::set_Latitude: 0.889374
 cur alt (ft) =  0
Checking for lon = -1.35283deg, lat = 50.9574deg, alt = -9999ft
FGJSBsim::set_Altitude: 39.6472
...initializing ground elevation to 39.5472ft...
...initializing sea-level radius...
 lat = 50.9574 alt = 39.6472
...initializing Euler angles...
FGJSBsim::set_Euler_Angles: 0, 0.0074002, 3.47914
...initializing velocities...
FGJSBsim::set_V_calibrated_kts: 0
End common FDM init
Ready to trim, terrain elevation is: 39.5414
  Trim complete
  Initialized JSBSim with:
  Vn,Ve,Vd= 0, -0, 0 ft/s
  Bank Angle: -0.0891999 deg
  Pitch Angle: 0.210416 deg
  True Heading: 199.34 deg
  Latitude: 50.769 deg
  Longitude: -1.35283 deg
  Altitude: 43.1159 feet
  loaded initial conditions
  set dt
Finished initializing JSBSim
FGControls::get_gear_down()= 1
0: GEAR_CONTACT: 0 seconds: NOSE 1
1: GEAR_CONTACT: 0 seconds: LEFT_MAIN 1
2: GEAR_CONTACT: 0 seconds: RIGHT_MAIN 1
3: GEAR_CONTACT: 0 seconds: TAIL_GUARD 1
4: GEAR_CONTACT: 0 seconds: PROPELLER_BOTTOM 1
Splash screen progress
5: GEAR_CONTACT: 0.00833333 seconds: TAIL_GUARD 0
6: GEAR_CONTACT: 0.00833333 seconds: PROPELLER_BOTTOM 0
7: GEAR_CONTACT: 3.63333 seconds: RIGHT_MAIN 0
8: GEAR_CONTACT: 4.01667 seconds: RIGHT_MAIN 1
9: GEAR_CONTACT: 4.725 seconds: RIGHT_MAIN 0
10: GEAR_CONTACT: 4.84167 seconds: RIGHT_MAIN 1
{copilot} Welcome aboard! Need help? Use 'Help -> Tutorials'.

Hmm. Odd that it shows contact of TAIL_GUARD and PROPELLER_BOTTOM. These are the lowest structure contact points (apart from the gear) but these shouldn't be hitting the ground unless it's starting with the wheels below ground.

Tail guard is at -16.38", propeller bottom is -35.32" and the gear are at -51.8" (nose) and -51.26" (main). My reference point is the tip of the propeller cone.

Also a simple question.... how do I get hold of it so as I can look at the engines and prop configuration for when you get back on Friday.

That would be very helpful, thanks.

https://github.com/FGMEMBERS/Beagle-Pup

This is missing my last commit for the square rebound damping on the gear. I'll ask IH-COL to refresh it from FGADDON.

Also any data folder for the engines and prop configuration you're using to base the xml files on would help me having to search the net and taking a punt.

This has the type certificates and the operator's manual for each engine and propeller.

https://dl.dropboxusercontent.com/u/283 ... anuals.zip

General specifications:

http://www.pilotfriend.com/aircraft%20p ... %20100.htm
http://www.pilotfriend.com/aircraft%20p ... e/A500.htm

Many thanks.

[jwocky]

Okay, here are a few things to check out, if I understand the problem correctly:

If the plane is just blown by the wind sideways, it's maybe the weight. The Pup is a very light plane and if the wind is strong enough ... in that case, you maybe start it up with full tanks and pilot weight in it and parking brake on.

If the plane flips over sideways, it is definitively a force outside of the x-axis. The usual suspects are tanks, pax-weights and so on out of the middle line, or, usually a bit harder to spot different damper settings on the gears laft and right. I had a plane once, that had 230000lbs/sec damper on one side and accidentally only 23000 on the other. When the tanks were full, it hung on the side with the weaker damper and flipped over during the trim.

and the same can happen around the y-axis if the CoG is too far behind the main gear. Nose goes up, tail strikes the ground, since the tail is defined as collision point, like another giant spring, the plane jumps up and you get blown away because you had no ground contact at all for seconds and you see only the end result when the splash screen fades.

Just to check it out, give yourself with manual weather a good crosswind or go to a place with adequate nasty wind and comment the tail strike struct out and see whether it still happens. Iwill take time to tinker on planes next week, I can try the Pup then and have a look if that is good enough.

[IAHM-COL]

This is missing my last commit for the square rebound damping on the gear. I'll ask IH-COL to refresh it from FGADDON.

Updated now at 9955c2

[jwocky]

Okay, I was curious, got me the Pup 160 and just browsed over the files (I just read, no changes), and took her in the air and some things caught my attention

1.) The plane is, as it starts, of course slightly asymmetric because it starts with a pilot of 170 lbs. With a spring coefficient of 3000 lbs/sec, you have about 1 inch more compression on the left than on the right. Which would be normally uncritically, but look at the z-coordinates (-51.26 inches, -48.56 inches uncompressed)

2.) The z-coordinates are between -51 inches (gear) and +52 (rudder top). 103 inches or more than 3 meters? Is she that high?

3.) She is a little nose light, that leads to a nodding motion and if there is enough wind, she slightly hits with the tail loses direction control for a second, bumps back on the nose wheel and back. And with every repeat, cross wind gets her a little blown away. Funny effect, you can only see on really light planes.

So, two suggestions:

1.) put CoG and pressure point a little further back to get more weight on the nose wheel

2.) For an empty weight of 1090 lbs, 2100 lbs/sec on the nose wheel and 3000 lbs/sec on the main gears appear a little bit high. Especially the nose wheel appears quite jumpy given, most of the weight sits on the main gear. Same with the rebound.

The being blown over is pretty clearly a side effect from loosing her ground contacts while the splash screen still covers her. So you see only the end result, when she is already blown over. But you can do the following: Go to KXTA, one of the green runways (they are so nice bumpy). Give yourself some good crosswind and just taxi slowly around. You can see how she bumps up with the nose gear at every little pothole and then gets a side push the moment, the nose gear is up.

Now, off to the engines ***spreads a cloud of spare parts and some oil traces in his wake ***

[jwocky]

I had no time for a closer look, but it seems, the p-factor is high, also resulting in a lot of torque and I am not sure, do they really run the propellers up to 2700rpm or only the engine? There is really no gear in between?

[bomber]

In need of incidence data at root and tip if anyone has it ?

Simon.