I am trying to analyse the crash shown in this video:
This was a test flight which lasted around 5 minutes, which includes autonomous take-off, 6 waypoints and autonomous landing. Everything went fine until the landing part. Previously we tried a similar scenario more than 5 times and they all worked flawlessly so it’s very puzzling to see it crash like this.
I tried to follow the steps here : copter.ardupilot.com/wiki/common … sing-logs/ but I still can’t get my head around it and pin-point for sure the reason for the crash.
I generated the following plots in an attempt to understand the crash (this log contains two flights, the crash is in the second flight)
More higher res images with the log file can be found here: mega.co.nz/#F!I8RywRrC!9PC-S2kopzQTDbDP8zaYUw
What I noticed was the following:
[li] The Yaw keeps on circling, basically it goes from 0 to 360 and back again, very weird ![/li]
[li] The Roll and the desired roll change just before the crash, indicating a possible mechanical failure ?[/li]
[li] The number of GPS during the auto landing was 7-8 [/li]
[li] Acceleration in the Z and Y directions were high too during the crash [/li][/ul]
It’s just hard to exactly figure out the source of the crash. One thing for sure is that the heading of the UAV keeps on rotating for no obvious reason, the magnetometer seems to be fine.
We have a competition this weekend and we would really like everything to be in control and understandable, we don’t want these sudden crashes to happen.
My take on this
your clock wise motors are working quite a bit harder than your clock wise motors. Looks like you have/had a twisted arm creating a torque. This is consistent for the 3 flights in the file.
By the third flight you don’t have a lot of headroom on the clock wise esc outputs, I guess combination of your torque I mention and battery voltage dropping with use. Then shortly after you lose alt channel 4 looks pegged to me. Not a good situation to be in.
Plot RCOUT 1 to 8.
The 4 harder working outputs rise quite quickly prior to the alt drop, quicker than volt sag I think. Looking at des yaw and yaw it looks like its trying to turn so increases the already hard working motors, …but you don’t have enough power left in reserve to make the yaw and the copter starts to struggle trying too maintain attitude, sacrificing altitude.
So my conclusion at the moment is that twisted arm resulted in requiring too much effort by one set of motors, not leaving you enough in reserve. Its one of the first things I look at when looking at my logs, especially for a new build. It wastes power, makes the yaw controller think harder and as seen here can cripple the overall performance of your copter.
Might be hard to tell if a twist was there if it got a bash on the landing. I don’t think it came loose in the air though with how consistent the motor speeds are prior to the crash for the 3 hops.
I try to align the motors as best I can prior to first flight too…eyeballing the ends of the props aligning with the adjacent motor tips, gets it pretty close. It doesn’t take very much misalignment for it to be seen in the logs. Unless you look specifically for misalignment you probably won’t have noticed.
Hope you get it back together…I do like the octaquad config, my DIYer is an octaquad.
Wow, great analysis RabbitStu, so many things to learn from experienced people like yourself.
I think your analysis are inline with what we thought, the way these Steadidrones motors are mounted is problematic. They can easily get misaligned during transporation, and it’s hard to get a precise allignment.
The puzzling thing is that we always had these circling behaviour prior to the crash in multiple test sites and multiple occasions, even when the autonomous take off and landing was working perfectly. We initially thought that the 0 degree yaw in the waypoint command was relative to the second waypoint, and to be honest with you, it’s very hard to see the correct orientation from far away anyways. From what I read now, the 0 degree yaw means that it should change the heading to point to the next waypoint, a behaviour we never managed to observe ! Any ideas on that ?
Thanks a lot for your feedback
Yeah those mounts aren’t too different to what I use. Infinite adjustment but obvouosly can be problematic if knocked.
Your yaw could be a symptom of the twisted arm too! The yaw controller might not be coping. I will try and have another look at your logs when I’ve had my tea.
sent from my phone so apologies for any typos
OK another quick look focusing on yaw, I didn’t look too hard at that before.
In auto mode its harder to see whats going on as des yaw does’t act in the same way as in the other modes. In other modes if the yaw controller can’t maintain yaw, the des yaw is pulled around to follow yaw, this makes it easy to see that des yaw is chasing yaw.
But, it looks like your copter is yawing clockwise when viewed from above, is that correct? Your clock wise motors (also viewed from above) are trying a lot harder so the brain is trying to turn the copter anti clockwise when viewed from above, or in other words its trying to stop the clockwise spin but failing. In this case if the yaw controller tried any harder I don’t think it would have enough thrust to fly or maintain attitude.
I think if you get it flying again and with straight untwisted arms it will fly a lot better without the yaw issue.
First flights should not be an auto mode see if yaw is ok, get some logs and check the RC outs again, and adjust any twist as necessary.