Could you please help analyse this Pixhawk crash with log?

Posted by Tom Moore on May 28, 2016 at 2:09am

Hi all,

I'm a relatively new Pixhawk pilot and have been flying this flight controller for about a year and a half. Last night (flying at night to try and record bats as part of a project), I set up a test auto route for a new area and flipped the plane to auto as usual on the transmitter.

Unfortunately, it then took a dive that I couldn't correct in time despite an altitude for the waypoint of 40m (relative).

If you could help look through the logs (attached) then this would be really helpful. The plane is fixable but it will take quite some time!

Thanks for your help and taking the time to look at these. I'm not really sure what to be looking for so any help would really be appreciated. From looking at the path in Google Earth it looks like the auto path thought it was above the ground but the attitude of the plane shows it was nose down the whole time.

All the best, Tom

2016-05-27 21-35-50.log

2016-05-27 21-35-50.kmz

You need to be a member of diydrones to add comments!

Join diydrones

Email me when people reply –
Follow

Replies

  • Tom Moore June 1, 2016 at 12:53pm

    Hi,

    Following on from the replies below I have gone through the autotune procedure and this produced the attached log.

    Using the link to the AUTOTUNE page that Tridge supplied, I've also graphed ATRP.Demanded vs ATRP.Achieved and this can be seen attached as well. It doesn't look exactly as the AUTOTUNE page describes.

    Can anyone who knows about these things tell me if the AUTOTUNE has worked? Please note that I was using a slightly different plane to the previous one that crashed (now it's this one http://projecterebus.weebly.com/blog/bat-uav-mk-ii-part-ii and previously it was a normal Bix3 from Hobby King).

    Thanks, Tom

    ATRP demanded vs achieved.jpg

    Bat UAV Mk II - Part II
    After an initial successful flight to prove the wings would stay on and the plane doesn't generally disintegrate in flight we've been able to fly the…
  • Tom Moore May 28, 2016 at 5:44am

    Hi Tridge and Hein,

    Thanks for taking the time to look through the files. I'll definitely be going through the autotune process and will re-enable the arming checks (not sure why I disabled these!). Is there anything else you would recommend before attempting another auto mission?

    One thing I wanted to query is that I've flown quite a few auto missions before and never experienced this behaviour. It seems that the plane wanted to reduce it's altitude when in auto. When I look at the flight in Google Earth it seems that the log shows that the plane was flying level then a decrease in altitude then an increase but what the plane was actually doing during this period was a fairly set slope descent towards the ground. I've included the jpeg to show what I mean. Can I ask where the log takes the altitude from for this as it seems to be different from what I actually experienced?

    Just one more point, the increasing error in altitude - is this something I need to be concerned about? Was it a factor in the crash?

    Thanks for the help, Tom

    ps repairs are underway! Will hopefully have something flyable by the bank holiday :)

    Crash flight.jpg

    https://storage.ning.com/topology/rest/1.0/file/get/3702248222?profile=original
    • 100KM
      Hein du Plessis > Tom Moore May 28, 2016 at 11:24am

      Hi Tom

      See my attachment - the altitude error is the difference between the required altitude and the actual altitude. The fact that this keeps growing means APM knows it needs to go higher, but it can't. Another indicator is the throttle that goes to max - APM tries to increase airspeed when it can't achieve pitch.

      Check under GPS - RelAlt for altitude that's read by APM - that shows APM knew it was losing altitude. Again check my image above - not sure why your KML shows otherwise.

      Cheers,

      Hein

    • Tom Moore > Hein du Plessis May 29, 2016 at 4:00am

      Thanks, Hein. That explains things very well.

      I'm still a little unclear on why the path in Google Earth shows the plane's course and altitude as different from what it was actually doing? 

      Thanks, Tom

  • Developer
    Andrew Tridgell May 28, 2016 at 2:46am

    Hi Tom,

    The problem is that your pitch gains have never been tuned. When you are flying in manual I can see you putting in quite large stick inputs in order to make the aircraft climb. For example, elevator input went to 1200.

    The autopilot needs to know how much elevator input it needs to use to produce a desired pitch rate. That is the PTCH2SRV_P parameter. The flight log shows that is the default value of 0.4. That produced only a very small elevator input. It may be enough to keep the aircraft level, but it isn't enough to make it pull out of a dive.

    You also had a very low integrator gain (PTCH2SRV_I) which is used to compensate for too small values of P.

    What you need to do is run an AUTOTUNE. See:

    http://ardupilot.org/plane/docs/automatic-tuning-with-autotune.html

    I would also strongly recommend you re-enable arming checks (the ARMING_REQUIRE and ARMING_CHECK parameters). In this case the lack of arming checks didn't cause the crash, but you will eventually get a crash if you continue to have all pre-flight checks disabled.

    Cheers, Tridge

    Automatic Tuning with AUTOTUNE — Plane documentation
    • 100KM
      Hein du Plessis > Andrew Tridgell May 28, 2016 at 3:16am

      Hi Tom

      I agree with Tridge, your PTCH2SRV_P is 0.4 which is very low - my planes vary from 0.9 to 1.6. You said "flipped the plane to auto as usual" so I assume you've had successful auto missions before, but perhaps they didn't have to cope with recovering from a steep descent.

      I can see on the logs that altitude error goes up continuously while throttle goes to max (ch3), but elevator stays very low (ch2) so APM simply did not have enough elevator authority.

      Also pre-flight checks is a good idea :)

      Sorry for your loss, hope you have it repaired soon - great place to fly btw!

      2016-05-28_12h15_26.jpg

      https://storage.ning.com/topology/rest/1.0/file/get/3702634462?profile=original
This reply was deleted.