Ground trim from a python script?

[clolsonus]

Hi, I am working on a little side project that uses the python interface to JSBSim. I would like to do script-based initialization instead of defining initial conditions in an xml file (for future flexibility.) However, when I try to do a ground trim I end up with all my gear points (including all the hard contact points) making contact and the plane immediately bounces in the air.

Question: is there any examples or tutorials of how to do a purely python-script-based ground trim? I haven't been able to find anything that doesn't just load initial conditions out of an xml file.

Otherwise I just realized I don't have all my most recent changes committed, but later this evening (or tomorrow) I could post a link to my python code if anyone was willing to try it out and help me.

I tried to work on this a few months ago and finally gave up, but was able to at least get things started mostly reliably by using trimType=6 (no trim at all.) But I'd really like to get this and in-air trims working from python. At the end of the day I want to be able to specify and airport or runway or 5 mile final or any place for startup similar to how it is done in FlightGear.

Can any one offer any help or insights or point me to a good working example?

Thanks in advance!

Curt. (yes that Curt ... I have not messed with this stuff for at this level for many years so I have forgotten everything I once knew and have become dumb again, but I have a uav project I've been still working on and I'd like to beef up my simulation portions. ) :-)

[bcoconni]

Ground trim can be done with Python by setting the initial conditions properties then by setting the property simulation/do_simple_trim to 2.
In the example below, the pitch angle theta-deg is set to 0 then the ground trim is triggered.

jsbsim/tests/CheckTrim.py

Lines 61 to 63 in 70505e6

	fdm['ic/theta-deg'] = 0.0
	# If the trim fails, it will raise an exception
	fdm['simulation/do_simple_trim'] = 2 # Ground trim

Note that an exception jsbsim.TrimFailureError will be raised if the trim fails.

[clolsonus]

Hi Bertrand, this seems to work exactly as described for the two simple models I tested, thank you very much!!!

I feel 1% smarter now than before, but still 99% dumb. I tried to extend your example to do an in-air trim and my trim fails.

The example script you pointed to appears to also include an in-air trim, but that loads the initial conditions from an xml file for the in-air trim test. I assume it should be possible to do a simple python script-based in-air trim by setting the right properties? Is there a similar python example somewhere? Or maybe my trim fail output tells you something.

I am trying to run the FlightGear JSBSim Rascal 110 model standalone. As far as I know the Rascal 110 model should be able to easily cruise at 30 kts @ 7500' MSL. I set these ic properties:

     # St George, UT
     self.fdm["ic/lat-geod-deg"] = 37.03
     self.fdm["ic/long-gc-deg"] = -113.52
     self.fdm["ic/terrain-elevation-ft"] = 2500
     self.fdm["ic/h-agl-ft"] = 5000
     self.fdm["ic/vc-kts"] = 30

     self.fdm["propulsion/set-running"] = -1

And here is the output:

 Trim failed
 Trim Results: 
       Angle of Attack:   2.54  wdot:  1.43e+01 Tolerance: 1e-03  Failed
              Throttle:   0.50  udot: -2.18e+00 Tolerance: 1e-03  Failed
            Pitch Trim:  -0.33  qdot: -1.03e+00 Tolerance: 1e-04  Failed
            Roll Angle:  -0.01  vdot:  4.98e-03 Tolerance: 1e-03  Failed
              Ailerons:   0.00  pdot:  3.78e-06 Tolerance: 1e-04  Passed
                Rudder:  -0.51  rdot:  1.08e-01 Tolerance: 1e-04  Failed

AOA seems reasonable for these conditions, throttle of exactly 50% seems suspicious. The other values seem ok. Thanks in advance for any little nudges in the right direction!

Edit: I did get a separate model (from work that I cannot share right now) to trim in the air ok ... so there maybe is something with the rascal 110 model? Is it possible to get help debugging why the rascal model fails the in-air (longitudinal) trim?

Thanks again!

[bcoconni]

The trim solver sometimes struggles to converge (see the issue #360 for example). There are typically 2 reasons why the convergence would fail:

1. The engines are not running (despite calling self.fdm["propulsion/set-running"] = -1). Check that the engines are indeed running (RPM and thrust should be non-zero).
2. The initial conditions are too far from (or not close enough to) the solution. For that you might need to tweak the IC's to help convergence. A good start could be to set the pitch to 1deg (self.fdm['ic/theta-deg'] = 1.0) and see if that helps.

[clolsonus]

Thanks Bertrand,

The Rascal 110 has an <electric_engine> for power. I don't know if I did the right thing for printing the values you suggest, but right before the trim I printed a set of values and right after I printed the same, and indeed propeller_rpm, power, and thrust are all zero after the trim:

Before trim:

propeller_rpm 0.0
power_hp 2.9599973179562826
power_W 2.9599973179562826
blade_angle 30.0
advance_ratio 0.0
thrust_lb 21691533.212531038
thrust_N 21691533.212531038

After failed trim:

propeller_rpm 0.0
power_hp 0.0
power_W 0.0
blade_angle 30.0
advance_ratio 30.108009732337035
thrust_lb -0.0
thrust_N -0.0

Is there something different that needs to be done for trimming with an <electric_engine> ? Ground trims appear to work and I immediate have power without needing to do anything additional.

Thanks in advance,

Curt.

[bcoconni]

The thrust values before trimming are suspicious: 21691533.212531038 lbs is way too high. Also the property propeller_rpm is 0.0 so in all likeliness your engine is not producing any thrust since the propeller is not spinning. Are you sure your propeller is linked to your engine ?

In addition you get the same values for different units:

• power_hp and power_W should be different.
• thrust_lband thrust_N should also be different.

[clolsonus]

Hi Bertrand, thanks once more for your patience in walking me through this learning (debugging?) process.

If it helps, the version of the model I am attempting to trim can be found at the following link It should be a straight copy of the jsbsim portions of the FlightGear Rascal110 model: https://github.com/NorthStarUAS/Simulator/tree/main/nstSimulator/data/models_jsbsim

I agree those pre-trim values are clearly weird and wrong, but is that something I am responsible for setting? (and if so, what is the recommended way to set them? ... I am trying to do this all dynamically from a python script and not use xml init files.) My naive expectation is that JSBsim would initialize all it's important values to 0 or whatever values are necessary for the code to function correctly?

The aero/engine/prop model comes right out of FlightGear and sadly I am not yet familiar enough with JSBSim to determine if everything is setup correctly? The top level xml file appears to have the following basic structure:

<fdm_config>
    <propulsion>
        <engine>
            <thruster>

I apologize again in advance for not jumping into the discussion here knowing enough to provide all the important details up front, so please feel free to continue to ask for what you need me to do to figure this out.

Thanks again!

Curt.

[bcoconni]

I have run a simulation with the following parameters:

self.fdm.set_dt(0.001)  # 1000 Hz time step
self.fdm["ic/vc-kts"] = 70.
self.fdm["ic/h-sl-ft"] = 5000
self.fdm["fcs/throttle-cmd-norm"] = 1.0  # Full power

self.fdm.run_ic()
for _ in range(2000):
  self.fdm.run()

The time step is reduced to 0.001s (1000Hz) because this is a small aircraft. The initial velocity has been increased to 70kts which is, as shown below, unsufficient for a level flight.

The simulation gives the following results:

There is this weird behavior with the angle of attack alpha-deg:

• A ~6Hz oscillation during the first 0.25s i.e. 1.5 period
• Then alpha stabilizes to -2 deg

When the simulation continues, one can see that the aircraft just dives. The velocity stabilizes to ~83kts, the pitch to -20deg and no phugoid is taking place.

Given these results, I am wondering if trimming this aircraft at 30kts is possible at all ?