Cessna 172 Pitches Up and Stalls With Elevator Neutral

[tejaDhulipala]

Hi all,

I am trying to use an autopilot for the Cessna 172x as part of a larger project. When testing the simulation, I set the elevator to neutral and noticed that the plane pitches up significantly until it stalls. This doesn't seem realistic, as Cessnas should have posititve static stability. I've included a 3D matplotlib graph of the flightpath. Throttle is set to 0.8, micture is set to 0.8, ailerons are controlled by a heading PID, and fcs/elevator-cmd is set to 0.

It appears to try to climb 400 ft or so, then it stalls. Is this how it's supposed to behave, or is there something wrong with how I am using it?

[seanmcleod70]

You say you set fcs/elevator-cmd to 0, but are you sure that the autopilot isn't changing it based on the autopilot mode, it's set point(s) etc.?

Let's see some plots of the controls over time, airspeed, attitude etc.

[tejaDhulipala]

Thanks for the response. I'm pretty sure the autopilot is not changing the elevator. In fact, the only reason that I'm using the autopilot mode it to hold the heading constant. Otherwise, with both elevators and ailerons neutral, the plane will start rolling as well. I just wanted to isolate the issue with it pitching up. Here are my plots:

Another problem I noticed is that when I turn on the heading autopilot, the aircraft is able to align to the heading well. However, the property "fcs/left-aileron-pos-norm" is constant at zero despite this.

Here is my code for reference:

aircraft = jsbgym.c172x

sim = SimulationInterface(render_mode="graph") # render mode can be None, flightgear, human, graph, human_fg, or graph_fg
[print(k, v) for k, v in sim.initialize().items()]


num_steps = 0
max_steps = 50
while num_steps < max_steps:
    actions = {
        prp.throttle_cmd: 0.8,
        prp.mixture_cmd: 0.8,
        prp.elevator_cmd: 0,
        prp.ap_hold_heading: 1,
        prp.ap_heading_setpoint: 0
    }
    obs = sim.step(actions)
    num_steps += 1
sim.close()

SimulationInterface is a wrapper for jsbsim.FGFDMExec(), and prp are wrappers for properties.
prp.throttle_cmd = fcs/throttle-cmd-norm
prp.micture_cmd = fcs/mixture-cmd-norm
prp.elevator_cmd = fcs/elevator-cmd-norm
prp.ap_hold_heading = ap/heading_hold
prp.ap_heading_setpoint = ap/heading_setpoint

[seanmcleod70]

That's a pretty aggressive pitch up. There are really only 2 moments that produce a pitch up, from the elevator and from the engine thrust if the propellor is mounted below the cg of the aircraft.

I'd still suspect the elevator at this point. Plot the elevator position, which is what is used in calculating the pitch moment, let's see what that shows.

fcs/elevator-pos-rad

This particular model has 3 inputs that get summed to produce the elevator deflection. You've plotted fcs/elevator-cmd-norm above, but what about the other 2?

jsbsim/aircraft/c172x/c172x.xml

Lines 340 to 362 in c8af924

	<channel name="Pitch">
	<summer name="fcs/pitch-trim-sum">
	<input>ap/elevator_cmd</input>
	<input>fcs/elevator-cmd-norm</input>
	<input>fcs/pitch-trim-cmd-norm</input>
	<clipto>
	<min>-1</min>
	<max> 1</max>
	</clipto>
	</summer>
	<!--
	Map the -1 to +1 input into the range specified below, then
	multiply by 0.01745 to convert to radians
	-->
	<aerosurface_scale name="fcs/elevator-control">
	<input>fcs/pitch-trim-sum</input>
	<range>
	<min>-28</min>
	<max> 23</max>
	</range>
	<gain>0.01745</gain>
	</aerosurface_scale>

[tejaDhulipala]

I plotted all the relevant command variables, and all of them are at 0, except elevator position in radians, which is 0.

These are the precise elevator properties I am graphing, for clarity:

self.sim["fcs/elevator-cmd-norm"], self.sim["fcs/elevator-pos-rad"], 
        self.sim["ap/elevator_cmd"], self.sim["fcs/pitch-trim-cmd-norm"]

[tejaDhulipala]

I also tried to debug by checking how different planes react (without the autopilot heading hold of course). I started them at the same speed, throttle (0.8), and elevator cmd (set to 0). All other conditions I left as default, so I don't know what they are.

• The Cessna 172P climbs about 500 ft and stalls. It also turns 60 degrees to the left.
• The Boeing 747 pitches down to about 20 degrees and descends (probably because of the low starting speed)
• The piper cherokee 28 does a pitch up maneuver, but pitches down before stalling and it isn't as extreme. Seems somwhat realistic.

[seanmcleod70]

Plot all the moments in the PITCH axis, i.e. aero/coefficient/Cmde and all the others. The overall pitch moment will be the sum of these plus any moment from the engine.

Just to double-check, have you modified the cg?

Print out the cg values.

jsbsim/src/models/FGMassBalance.cpp

Lines 405 to 410 in c8af924

	PropertyManager->Tie("inertia/cg-x-in", this,1,
	(PMF)&FGMassBalance::GetXYZcg);
	PropertyManager->Tie("inertia/cg-y-in", this,2,
	(PMF)&FGMassBalance::GetXYZcg);
	PropertyManager->Tie("inertia/cg-z-in", this,3,
	(PMF)&FGMassBalance::GetXYZcg);

[tejaDhulipala]

I have not modified the CG. I'm going to print those out right now

[tejaDhulipala]

CG X in: 45.48951612903226
CG Y in: 4.22741935483871
CG Z in: 35.42701612903226

[seanmcleod70]

Any reason you're not trimming the aircraft? Unless the aircraft is trimmed or there is an autopilot active for all axes you can't be surprised that starting an aircraft at some airspeed and altitude with the controls all centered is going to give you stable flight.

[tejaDhulipala]

I don't know how to trim the aircraft, so maybe that coul help. The context of this post is that I was trying to tune an autopilot using a PID for altitude control with the elevator, but no matter what I did the plane would pitch up and stall. The inbuilt altitude hold autopilot in c172xap doesn't seem to work either.

[tejaDhulipala]

Here are my graphs with the inbuilt autopilot active:

These were my inputs:

actions = {
        prp.throttle_cmd: 0.8,
        prp.mixture_cmd: 0.8,
        prp.elevator_cmd: 0,
        prp.ap_hold_heading: 1,
        prp.ap_heading_setpoint: 0,
        prp.ap_hold_altitude: 1,
        prp.ap_altitude_setpoint: 5000
    }

[seanmcleod70]

Take a look at the JSBSim python examples - https://github.com/JSBSim-Team/jsbsim/tree/master/examples/python for examples of how to trim the aircraft.

Having said that, even if you have a successful trim solution, if you then engage an autopilot that is badly tuned the aircraft could spiral out of control.

[tejaDhulipala]

Yeah I'll check out auto trimming the aircraft. Also, after modifying the altitude hold pid it is a lot better.

My bigger concern when I say the aircraft pitched up with no elevator deflection was that there was something wrong behind the scenes in the properties I wasn't noticing. So just to reiterate, everything with jsbsim (and its properties) are fine, and that agressive pitch up is accurate?

[tejaDhulipala]

Yeah, I think everything looks right. I played around with the PID values and got a decent altitude hold. I also think starting the plane at 120 knots (even though that is its cruise speed) may have caused a lot of lift and a strong upward moment. Thank you so much for your help!

[seanmcleod70]

You generally want to start the aircraft in a trimmed state, even if you're immediately going to enable an autopilot.

[tejaDhulipala]

That makes sense!

[agodemar]

@tejaDhulipala Trimming an aircraft model is a fundamental subject in Flight Dynamics. To learn more, you can watch Ben Dickinson's YouTube video on trimming a 6-DoF model.