Help Deciphering A Crash - GPX File

I noticed that you took off in position hold. I do not think that this was causal in the mishap. However I do not think it is a good practice to take off with any of the automation on AH, PH or what ever. You do not know what the aircraft will do it could be having a satellite glitch and takeoff and fly straight into something or yourself. You should get the A/C airborne in a safe altitude and stable before turning on PH. I had a moment had I not done this I would have lost an A/C. I turned on PH at 20 feet and it started to move away rapidly. Had I not immediately turned PH off I would have lost the A/C.

It looks like your aircraft did not have the sufficient power to fly level at the speed you accelerated to, this coupled with a low power state your aircraft ran out of power to fly at 35 ft/S (24MPH) and stay level. It looks like right before the crash the aircraft attempted to power up to climb but then your battery went well below limits and was not able to push the current required to stay level at that pitch angle. I think if you had leveled of a bit you may have been able to save it. Any time the automation is not doing what you want you should turn it off and stabilize. I do not know what controller you are using but I have the DX8 getting battery info from the TM1000 telemetry module, This tells you if you are pushing your battery beyond limits and alerts you to discontinue what you are doing. I have had several instances where my controller barks at me when the battery is going beyond limits for what I am demanding in addition to simply letting you know that your battery is low.

 

Colin

Not sure if this is a problem you are experiencing now or not but to eliminate any future problems or possible doubts of what can be future problems I would twist all your wires. Personally I think your Lipo wire is too long, but if you twisted it, it will become much shorter. I would also twist the lipo battery cables as well. Every cable that you have soldered to the flight control board or the PDB should be twisted. The motor and LED wires that run through the booms should also be twisted. What I do is twist the cables and then heat shrink them in place so that the wires don't untwist. For the motor wires I braid them. You can see the picture of my lipos wires on the battery as an example but it applies to just about all your wiring.

 

This is how an NTSB aircraft safety investigation report would have sounded like.

This may sound harsh not my intention. This is based solely on my analysis of your flight data

Initial Synopsis

At 10:50:54 Mishap Aircraft (MA) took off with 15.8 V on the battery and using 650W of power. Mishap Pilot (MP) had Position Hold (PH) engaged prior and during the takeoff. The initial take off was uneventful and the aircraft climbed to an altitude of 38 feet. The nominal power during takeoff was at 650 watts. At 10:51:06 the aircraft was level at 38 feet. MP engaged Altitude Hold (AH). At 10:51:25 MP disengages position hold. At 10:51:26 MA pitch is lowered to 10 degrees of pitch momentarily and accelerates to a ground speed of 5.5 ft/s(GS) MA is maintaining altitude. The MA was flying on a 150 heading. at 10:51:41 MA was accelerating to 10 ft/s(GS) with a 4 degree nose down pitch. At 10:51:47 the MA was flying at 21 Ft/sec and slowly descending. MA was at 35 feet using 850W of power. At 10:51:54 MA aircraft descended to 30 feet and was traveling at 33 Ft/S(GS). At this time a low battery indication is enunciated. The Power spiked to 1009.58W, the battery was registering 14.3V at 70.6 Amps altitude hold is still engaged at this time. At 10:51:55 the ground speed was 35 Ft/S at 28 feet AGL 10 feet below target altitude. The battery is well below a safe state registering at 13.5V and pushing 72.7 A a total of 981.45W. After this the MA was no longer recording the data to the MA impacting the ground.

Conclusion

The MA took off with a battery that was 1V below a full charge. MA aircraft reported no known maintenance anomalies in the data, during the initial Takeoff and flight. Maintenance issues do not seem to be causal in this mishap. MA took off with PH engaged but is not determined to be causal. The MA flew to an altitude of 38 feet. MP disengages PH accelerates the MA to a speed that the MA was unable to provide enough thrust to maintain altitude. MA was in a continuous descent until ground impact. Prior to ground impact MA's autopilot attempted to increase thrust until battery went well below safe battery limits. MA still had AH engaged through ground impact. MP may have failed to recognize the battery state either because the data was not available through a battery monitoring system or the system did not alarm MP. The MA with altitude hold engaged was not able to keep up with descending trend. Two possible cause that lead to the mishap. MP failed to recognize that the aircraft was descending through the target altitude in time to react. MP may not have been aware of the battery state. It is possible if the MP pilot recognized the descending trend or the low battery warning, the MP may have discontinued all maneuvering. MP could have disengaged altitude hold and stabilized the MA. Had MP been aware of the descending trend then MP could possibly have discontinued the forward acceleration to allow for the MA to regain the intended target altitude or level off. Their may have been sufficient time to safely land without incident.

I know this sounds terse, not my intention. Being in the flying business for 15 years and consulting drone crashes at the Air Force Safety Center, this is a small sample of what one of these reports sound like. Just for comfort to not throw spears I almost lost my CS8 doing something similar I was testing how aggressive I could descend while moving forward. I ended up at full throttle an the AC touched the ground momentarily and went airborne again I was able to stabilize it. Had I hesitated on the throttle I would have been in the same situation. Mine was not exactly the same I had full battery, but it does give me an appreciation that these things can get away from you pretty quickly if you are not judicious in your flying.

 

Andy bounce I like it

 

I would have to agree with your "baseless speculation" I would guess that he did not get a battery warning because the downlink to the DX8 was getting interference. I had the same issue with a TM1000 on my small Hexa. I had my TM1000 mounted on top of an AR8000 it was near my Arducopter 2.5 I was getting intermittent downlinks. When the downlink goes out you no longer get the alarm. As soon as I repositioned it away from the other electronics it seemed to be more stable

 

Do have a Grob 109?

In the predator if it bounces once we have to initiate an immediate go around or else it risks shearing the nose wheel.

 

I ran a couple of simple regresion models to your data and indeed I find a good statistical correlation to motor 3 current and pitch angle. The more forward you move your number 3 works harder. It seems that number 5 had some significance as well in some of my models. I would check number 5 it is possible that motor may be fighting 3. I would definitely have speed displayed in addition to battery info, if you don't. I like having bat info up as a redundant backup in case of the TM1000 failing. In fact I reference the OSD more for my bat info. I prefer to be heading back to a landing prior to the DX8 barking at me. I like to set what we call in Mil parlance a "bingo" fuel state if you will. I gauge how much power it takes me to get to a certain point then I subtract that amount from the minimum MAH's left. Once I reach that point I bring the copter back. If I do it just right I start getting the DX8 warning just as I am in the descent to landing. I tend to find the Miliamp counter useful. Although if you have to keep in the back of your head previous power used if you shutdown and relaunch the battery counter re-starts from 0.

Here is how I have my OSD configured. It is busy but I feel it has all of the essential information. I use the Commercial Aviation display standard of groundspeed left, alt right HDG on top. I placed the annunciator panel below that displays all of the mode and some warning information.

 

I did my glider training in Hobbs New Mexico back in 93 flying crappy Sweitzer 2-33's.

Llano I think is where they do training for the civillian contractors that fly the MQ-1's for the Army. For the Air Force I went through normal Undergraduate Pilot Training in Vance AFB,(Enid) OK( the armpit of america) I flew KC-135 Stratotanker's most my active duty career. I went through MQ-1 Training at Creech Afb, a small base 30 miles north of Las Vegas in the middle of no where. That is where the schoolhouse used to be until it moved in 2011 to Holloman in Alamagordo, NM. I was based out of Creech the rest of my stint doing combat ops until separating and going reserve last year.

 

Andy I ran a couple models in JMP. It definitely is not a pretty model for sure. I did not consider non linear effects. I regressed motor current for each motor against Nick Angle, Roll Angle, Elevation, Ground Speed, Variometer and course. I figured since there is intuition that the a/c fighting the course, I figured I would throw that in there. The first model was very dirty lots of outliers and had some issues with variance. I cleaned up a little, not great, but good enough for government work. I removed from the model all data points where the nick and roll angles were both zero, figuring that my have some effect in the correlation between those variables, in not adding to effect.

I was able to come up with this model for the Motor 3 Current

Parameter Estimates

Term Estimate Std Error t Ratio Prob>|t|
Intercept113.84593 7.363634 15.4 <.0001
NickAngle1.0027161 0.390663 2.57 0.0148*
RollAngle0.9698023 0.502042 1.93 0.0618
Variometer 0.3827967 0.214633 1.78 0.0834
Course -0.039359 0.017856 2.20 0.0344*
GroundSpeed 1.9823365 0.130977 15.14 <.0001*
Elevation(ft) -0.300519 0.192212 -1.56 0.1272

This shows the parameters Nick Angle, Course and Ground speed to be statistically significant in the model, the others not so much. The way I handled course is not ideal considering, I probably should have looked the absolute value at an overall change in course. I Just used the raw data. Course test well but does not do much to the model in terms of the parameter estimate, so we pretty much can ignore it. But the model does seem support the data in some respect. Nominally the amp draw on that motor is 11.3A. It is showing some correlation that if we are moving at 35ft/s that we'll increase the motor amps another 7A. If our pitch angle is 10 degrees we add another amp to the motor putting it around 19A where it was before the data shut off. So it seems to do a decent job in predicting amp draw for nick angle and ground speed. Overall the data seemed to test well for normality and I did not get as much of an issue variance, the scatter plot looked okay.

 

Thanks

Luckily I have a copy on a school licence through AFIT, I am finishing up a grad certificate in Test and Evaluation. It was something that I started while in the Air Force I am wrapping up my last class as we speak.

You are right about course, I should have chosen heading instead, I guess fixed wing logic still in my head. I found that to be odd as well that nick would be significant. I used elevation of the GPS(EDIT) because I felt that the Baro might not have been as stable and could contribute to outside factors such as pressure differential. I figured that the GPS elevation would be a little more stable in the data. As a you have said in one of your own posts in the my "Base Less Guess," could be that since that motor is in the rear and is compensating for one of the others. He is flying a CS6 so his rear center boom is 4 and 3 is the rear right adjacent boom. I was staring at my small DJI Hex to get an insight on what motors I am looking at. It is possible that the culprit motor is 6. 6 Is the front left boom. If that one is not level, it could be that 6 is causing the yaw effect and a differential thrust component, possibly forcing 3 to work much harder to compensate for the yawing and thrust moment on 6, while the copter is in an aggressive forward acceleration. If you look at the raw data prior to it going offline, You can see 6 is barely contributing anything, and 1 ramps up to try to compensate. It looks like near the end the rear engines were working more aggressively than the front motors. Its as if the logic of the copter was working harder to maintain a level platform over gaining or maintaining altitude. With all those factors combined it could be possible that it was pushed to its max envelope, with stuff out of balance caused that 3rd motor to work extra hard.

An interesting model might be to regress number 3 against the other motors and see which motor is contributing an effect on 3. I did it and it looks like number 4 is the culprit.

I did as such below and it looks like Motor 4, not 6 is being the bastard!

 

I regressed engine 3 to all of the other motors and here is some insight.



Parameter Estimates

TermEstimate Std Error t Ratio Prob>|t|
Intercept 31.971632. 0.80938 1.54. 0.1334
MC1 0.0830837 0.160587 0.52. 0.6081
MC2 1.1079499 0.503323 2.20. 0.0344*
MC4 1.4082313 0.269697 5.22. <.0001*
MC5 0.190791 0.195378. - 0.98 0.3355
MC6. - 0.485089 0.219955. - 2.21 0.0341*

This is a pretty good model. It predicts the data to +-6 Amps on the earlier data but is almost spot on in the lattar 30 data points. In fact my predicted value for the last data point was 19.83A which is damn close to the actual value of 19.1. All of the clock wise motors 2,4 and 6 test well for significance. Probability for motor 4 for statistical error value is fractional, meaning for is really significant in the overall effect that motor4 has on 3.


All of his counter-clockwise motors affect 3 which makes sense. But look at number4, the center rear motor! Number 4 seems to be really affecting 3 the most. So what that indicates to me is that 4 is providing the yawing moment that 3 is fighting. As 4 is spinning it is providing something that is making 3 counteract. If 3 is trying to balance itself laterally with 4 it is possible that 3 is trying to compensate the counter yawing tendencies or some induced roll of 4 (the center boom) to command a stable pitch attitude. 3 will ramp up to stabilize any moment that is causing the copter yaw as well as bank on that side which is probably why we do not see 5 ramping up as much. 5 is probably ramping enough to stabilize left banking tendencies leaving 3 to do the job of compensating off balance effects of yaw being induced by the rear motor

What the data is telling us is that as motor 4 ramps up that 3's current increases. This validates my model above with pitch angle. As the copter changes pitch, motor 4 is the motor that is mostly ramping up to pitch nose down as we would expect, but is causing 3's current to increase. Long story short, motor 3 Gets something that it does not like from 4 so 3 ramps up to counter that.