Trim Failed - no clue about

[gasilva]

HI,

Full Trim
  Sorry, qdot doesn't appear to be trimmable

  Trim failed
  Trim Results: 
       Angle of Attack:  10.16  wdot: -2.52e+00 Tolerance: 1e-03  Failed
              Throttle:   0.81  udot: -2.59e+00 Tolerance: 1e-03  Failed
            Pitch Trim:   0.00  qdot: -1.46e+00 Tolerance: 1e-04  Failed
            Roll Angle:   0.00  vdot: -1.04e-17 Tolerance: 1e-03  Passed
              Ailerons:   0.00  pdot:  6.28e-10 Tolerance: 1e-04  Passed
                Rudder:   0.00  rdot: -8.18e-10 Tolerance: 1e-04  Passed
Traceback (most recent call last):
  File "/Users/gasilva/Documents/GitHub/jsbsimNew/examples/python/RunScriptLongFlight.py", line 59, in <module>
    while fdm.run():
  File "jsbsim.pyx", line 1010, in jsbsim.FGFDMExec.run
RuntimeError: Trim Failed

I am implementing a Global5000 aircraft with DATCOM aerodynamics. Everything is running OK. But when I run trim-cruise_ap.xml I got the error above.

How can I debug it and know that is happening? Is there any clue to solve this problem?

Regards,

[seanmcleod]

What are the conditions you're trying to trim at, i.e. airspeed, altitude etc.?

For example maybe the airspeed you're using is too low, such that given the the alpha range the trim routines are using it isn't possible to generate enough lift for the given weight etc. Or maybe it can, but there is too much drag at that alpha such that even 100% throttle isn't enough.

[gasilva]

Basically it was a wrong forces balance model from DATCOM Pro. They implemented a totally different force balance model and a table for drag instead of using Drag-basic and Drag-induced. I migrated all DATCOM data to Aeromatic format and it worked. Thanks. PS: I can put the wrong file to your comments if you want.

[seanmcleod]

There is no reason that a table for drag shouldn't work. JSBSim doesn't care how the forces and moments are defined.

So I'd be interested to see what the original DATCOM data looked like.

[agodemar]

@gasilva Interesting work!
Your model of Global5000 and your tests on trimming, are they going to be available or explained somewhere in our Wiki?

[gasilva]

Maybe yes. But I need to know how to setup a blog like @seanmcleod did.

[seanmcleod]

Take a look at - https://pages.github.com/ or as I mentioned previously you can also just add jupyter notebooks to a repo and reference them.

[agodemar]

@gasilva @seanmcleod @bcoconni
I think it will be the case to host a JSBSim Blog. The supporting technology is GitHub Pages, much as we do for our JSBSim Online Reference Manual.
This time I'll try out Docusaurus, based on Node.js and Yarn.
The sources of the blog will be public, contributions by users will be welcomed. All contributions will be subject to the usual PR workflow.
What do you think?
(I'll move this post to a separate discussion)

[gasilva]

global5000_aero.xml
file created by DATCOM Pro

global5000_aero.xml.zip

Aircraft file that used global5000_aero.xml

global5000DAT.xml.zip

[gasilva]

Global5000 file with DATCOM information in the format of Aeromatic.
This actually worked.

[gasilva]

global5000AeroDAT.xml.zip

[gasilva]

The script is not trimming is this one.

trim-cruise_ap.xml.zip

[gasilva]

you will need this also to run it
airborne.xml.zip

[gasilva]

Consumption with a Mach step in autopilot

Consumption with a Mach ramp in autopilot

[seanmcleod]

@gasilva have you compared the drag data from DATCOM versus from Aeromatic? For example plot Cd vs AoA for both.

[gasilva]

[gasilva]

DATCOM table was generated by the Software
JSBSim is the Aeromatic
DATCOM Analytic is the $C_D=C_{D0}+k \cdot C_L^2$

[gasilva]

Just another fitting fro DATCOM data

[gasilva]

the Drag Analytic done with fitting of DATCOM table

    <function name="aero/force/Drag_basic">
       <description>Drag at zero lift</description>
       <product>
          <property>aero/qbar-psf</property>
          <property>metrics/Sw-sqft</property>
          <table>
            <independentVar lookup="row">aero/alpha-rad</independentVar>
            <tableData>
             -1.57    1.500
             -0.26    0.013474
              0.00    0.011474
              0.26    0.013474
              1.57    1.500
            </tableData>
          </table>
       </product>
    </function>

    <function name="aero/force/Drag_induced">
       <description>Induced drag</description>
         <product>
           <property>aero/qbar-psf</property>
           <property>metrics/Sw-sqft</property>
           <property>aero/cl-squared</property>
           <value>0.0583</value>
         </product>
    </function>

[gasilva]

Just another fitting fro DATCOM data

The grey line from DATCOM transformed to Drag basic + induced works! The problem is the table.

[seanmcleod]

Out of interest what did you use to plot the graphs? Is it Matplotlib with a particular style?

So comparing the DATCOM (orange) with the Aeromatic (blue) I can see from about +9 AoA and larger you could run into a case where the trim works using the Aeromatic drag data but fails with the DATCOM data because there isn't enough thrust available.

At what sorts of speed/AoA did the trim start failing for the DATCOM data?

[gasilva]

Sorry, it is a plain Excel plots. I did not have time to make txt files and code. For sure, Python is better but...

The problem with DATCOM are not the Drag values. I tried to implement a Drag table as it is in DATCOM aero.xml but it did not worked. When I used Drag Basic and Induced Drag it worked.

The CD_0 and k worked well. But the table not.

 <axis name="DRAG">

    <function name="aero/force/Drag_basic">
       <description>Drag at zero lift</description>
       <product>
          <property>aero/qbar-psf</property>
          <property>metrics/Sw-sqft</property>
          <table>
            <independentVar lookup="row">aero/alpha-rad</independentVar>
            <tableData>
             -1.57    1.500
             -0.26    0.013474
              0.00    0.011474
              0.26    0.013474
              1.57    1.500
            </tableData>
          </table>
       </product>
    </function>

    <function name="aero/force/Drag_induced">
       <description>Induced drag</description>
         <product>
           <property>aero/qbar-psf</property>
           <property>metrics/Sw-sqft</property>
           <property>aero/cl-squared</property>
           <value>0.0583</value>
         </product>
    </function>

the table is like that

	<function name="aero/coefficient/CD_wbh">
		<description>
			Drag coefficient due to wing, body, and horizontal tail
			Main contributor to Phugoid damping: greater Cd, better damping
		</description>
		<table>
			<independentVar lookup="row"> aero/alpha-deg </independentVar>
			<tableData>
				     -8.0000         .0350
				     -6.0000         .0255
				     -4.0000         .0183
				     -2.0000         .0138
				       .0000         .0121
				      2.0000         .0139
				      4.0000         .0186
				      6.0000         .0265
				      8.0000         .0379
				     10.0000         .0526
				     12.0000         .0706
				     14.0000         .0922
				     16.0000         .1171
				     18.0000         .1453
			</tableData>
		</table>
	</function>

and the force balance is like that

	<axis name="DRAG" frame="STABILITY">

		<function name="aero/force/drag-lbs">
			<description>
				Drag force = ( CD_wbh + dCD_Min_Flap_1 + dCD_Flap + CD_Gear + dCD_Elev + CD_induced)
				             * q_bar * wing area
			</description>
			<product>
				<sum>
					<property> aero/coefficient/CD_wbh </property>
					<property> aero/coefficient/dCD_Min_Flap_1 </property>
					<property> aero/coefficient/dCD_Flap </property>
					<property> aero/coefficient/CD_Gear </property>
					<property> aero/coefficient/dCD_Elev </property>
				</sum>
				<property> aero/qbar-psf </property>
				<property> metrics/Sw-sqft </property>
			</product>
		</function>

	</axis>

[gasilva]

I am running the simulation in Mach=0.43 to 0.5. The alpha is from 11 to 7 more or less.

[seanmcleod]

I don't see anything obviously wrong with the drag table.

When you say it doesn't work, does this mean you can't get it to trim at all across the M0.43 to M0.5 range, or just a portion at say the lower end of the Mach range?

[gasilva]

the trim does not work in 0.4 to 0.52 range with the table and force balance. Em seg., 2 de ago. de 2021 às 13:15, Sean McLeod ***@***.***> escreveu:

…

I don't see anything obviously wrong with the drag table. When you say it doesn't work, does this mean you can't get it to trim at all across the M0.43 to M0.5 range, or just a portion at say the lower end of the Mach range? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#490 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADUT7AS67Q4SIHRPFDMKXLT23ADXANCNFSM5BIVMTJQ> .

[gasilva]

maybe it worked in a previous version of JSBSim...

[gasilva]

it does not work even with force balance of JSBSim. If I define a force instead of coefficient.

[seanmcleod]

I'll try and take a look by running it and see if I can spot any issues.

In the meantime I'd also suggest that you contact the developer of DATCOM Pro to let him know.

[gasilva]

OK!

[gasilva]

Just did that and referred to this discussion.

[gasilva]

He know something like this happened in the past. He will inform soon.

[gasilva]

The error changed now when I changed the script.

Full Trim

Trim successful
Trim Results: 
     Angle of Attack:  12.01  wdot: -6.60e-04 Tolerance: 1e-03  Passed
            Throttle:   0.77  udot: -8.20e-04 Tolerance: 1e-03  Passed
          Pitch Trim:  -0.47  qdot:  2.80e-05 Tolerance: 1e-04  Passed
          Roll Angle:   0.00  vdot:  2.24e-17 Tolerance: 1e-03  Passed
            Ailerons:   0.00  pdot: -1.73e-15 Tolerance: 1e-04  Passed
              Rudder:   0.00  rdot:  8.55e-19 Tolerance: 1e-04  Passed

Repeating Notify (Event 2) executed at time: 0.008333
  position/h-agl-ft = 30000.000000
  ap/mach_setpoint = 0.430000
  fcs/throttle-pos-norm[1] = 0.767165
  fcs/throttle-cmd-norm[1] = 0.768530
  ap/throttle-cmd = -0.001562
  velocities/mach = 0.430000
  propulsion/engine[0]/n1 = 83.706485
  propulsion/engine[1]/n1 = 83.706485
  propulsion/engine[0]/thrust-lbs = 3711.430011
  propulsion/engine[1]/thrust-lbs = 3711.430011
  velocities/vc-kts = 157.641416
  aero/alpha-deg = 12.006342
  simulation/frame = 1.000000


Hold heading and altitude (Event 1) executed at time: 10.016663
  Cal. airspeed (kts):   = 157.636832
  Altitude (AGL, ft):    = 30000.373858
  Weight (lbs):          = 80101.397946


Repeating Notify (Event 2) executed at time: 1800.007613
  position/h-agl-ft = nan
  ap/mach_setpoint = 0.500000
  fcs/throttle-pos-norm[1] = nan
  fcs/throttle-cmd-norm[1] = 0.768530
  ap/throttle-cmd = nan
  velocities/mach = nan
  propulsion/engine[0]/n1 = 30.000000
  propulsion/engine[1]/n1 = 30.000000
  propulsion/engine[0]/thrust-lbs = nan
  propulsion/engine[1]/thrust-lbs = nan
  velocities/vc-kts = 0.000000
  aero/alpha-deg = 0.000000
  simulation/frame = 216001.000000

[seanmcleod]

So we can close the discussion?

[gasilva]

Actually not. The drag table does not work. Only if transformed to Drag Basic (CD0) and Drag Induced (k). I tried to debug but no success.

[seanmcleod]

I'm confused, yesterday I received a Github notification for this discussion in which you mentioned - "no issue with DATCOM JSBSim file".

When I copied one over another...I do not know what was different. Simply as that.
But no issue with DATCOM JSBSim file. It was the trim-cruise_ap.xml (trim cruise with autopilot).

And another one.

I do not know why it started working. Just copy the script that worked on top of the other script. But now I do not have any bug tracking.

[gasilva]

Yes false alarm! I was happy at that time. But I discovered that I was running the other model by accident.

[seanmcleod]

Have you had any further feedback from the DATCOM Pro developer?

[gasilva]

Not yet. I gave the false alarm to him also. Enviado do meu iPhone

…

Em 10 de ago. de 2021, à(s) 08:32, Sean McLeod ***@***.***> escreveu:  Have you had any further feedback from the DATCOM Pro developer? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.