WHAP! How long does it take to crash an airplane?

[Tim Wescott]

It just occurred to me that I could write a crash sensor into a TUT by looking for a sudden rotation that's way faster than "real", and for the (unreleased) big TUT, a sudden acceleration.

To work reasonably well, there'd have to be a rule that basically says "if the TUT rotates more than X overall, over the space of T milliseconds, then there's a crash and shut off the motor".  The fun part is finding the time span and rotation amount that's inclusive enough to encompass most crashes, yet won't make the motor stop if the plane gets slapped by a wind gust in the overhead 8's.

For the "how long does it take" part I'm thinking of just measuring how long it takes me to say "WHAP!" -- that should be pretty close.

Doing it with an accelerometer is a bit easier, at least for sudden crashes -- I've done post-hole crashes that don't involve much rotation, but a crash can be summed up as "sudden unauthorized change in direction", so that's pretty easy to detect.

Has anyone delved into this?  Got any numbers?

I wish there was a control line crash compilation video out there -- I could look at videos and get an idea of how much rotation I should look for.

[Tim Wescott]

I could probably detect a broken prop blade, too, but that's more problematical, because it depends on the airframe.  Even given that it's electric flight, the amount of vibration induced in normal flight on a plane with the wrong structural resonances may be worse than the amount of vibration induced on a really rigid airframe and a prop with an inch taken off one tip.

Folks may be a bit irritated with a timer that cuts out on the first corner.

[Mark wood]

I could probably detect a broken prop blade, too, but that's more problematical, because it depends on the airframe.  Even given that it's electric flight, the amount of vibration induced in normal flight on a plane with the wrong structural resonances may be worse than the amount of vibration induced on a really rigid airframe and a prop with an inch taken off one tip.

Folks may be a bit irritated with a timer that cuts out on the first corner.

The loss of a propeller should be detectable via frequency for certain. The acceleration is actually huge and cyclic. It would be very easy to calculate since the acceleration of a blade of a complete propeller is canceled by the acceleration of the blade opposite to it. A blade release event would then allow the remaining blade to pull really hard on the shaft resulting the the fuse having the sit shook out of it. Its easy a=v^2/r integrated over the span. This would turn up in the all of accelerometers wit ha frequency of the motor. All you need is to watch for a frequency dependent blossom.

Loss of load, bird shots and blade release testing is some of the most entertaining thigs I've done.

[Tim Wescott]

The loss of a propeller should be detectable via frequency for certain.

Hmmm - yes, but without knowing the motor RPM you'd have to do some fairly fancy signal processing to identify a strong repetitive signal.

If one could do it reliably just by seeing lateral acceleration over some threshold, I think I'd call that good.  Given how much vibration you see in acceleration on a slimer between engine running and engine off, I suspect that would be enough.

[Mark wood]

Hmmm - yes, but without knowing the motor RPM you'd have to do some fairly fancy signal processing to identify a strong repetitive signal.

If one could do it reliably just by seeing lateral acceleration over some threshold, I think I'd call that good.  Given how much vibration you see in acceleration on a slimer between engine running and engine off, I suspect that would be enough.

I think such a shut down would be good and if implemented you'll save at least one airplane at some point. I also don't think it has to be even that sophisticated. Whenever we would do one of these tests all of the accelerometers blossomed off the stirp chart. It's like a very violent earthquake, everything shakes like crazy. Just watch for an all channel spike in excess of 15 - 20 G something higher than normal for flight. There'll be peaks well in to the tens possibly hundreds of Gs if you're processing fast enough. Filter the detection for 4 or 5 processor cycles so noise doesn't trigger it from a stray power spike or something like that. Maybe omit the longitudinal axis which is likely to be lower or lower the threshold.  Or you could assume a frequency threshold using 5k RPM,  80 hz. Only the motor will likely ever have any tones in that range or above, maybe a pushrod but it won't have detectable energy. Hook one up to a monitor and thump it with your finger and watch how it responds, you'll see an all channel blossom. Think about it, I calculated the tension on the blades to be around 900 pounds to size the pivot pin on my hub for the 10" hinged propeller. What kind of acceleration do you think will result from that if it let go? Something 900/4 = 225 ish G's. More for a 12-13 inch prop. There's a reason airplanes self destruct when a prop blade comes off.

Mounting one of the IMUs on a slimer might provide some interesting data actually. It certainly would help characterize the frequency content and get an idea of minimum threshold. Take it for a ride and see what happens.

Most ESCs have an overcurrent shut down which does a good job of preventing damage such as what you are concerned about during a crash. The fires usually are  result of the battery being compromised. I'm not sure trying to protect for that kind of crash would be necessary albeit an interesting exercise trying to figure out how you would.   

[Mark wood]

Watch this video. This is one of the the things I used to do before the assimilation into the electronic engine controls side of the company.



Here's another piece of background:



Sorry but yet another video of interest. The engine control system has to be able to manage all of the elements within this discussion and more. A turbine engine is a very interesting machine chalk full of all kinds of aerodynamic pieces. Some engine controls I've worked on even managed the structural properties of the engine and include active clearance control where we manage the cooling of specific parts of the engine to keep it running optimally.

[John McFayden]

Tim,

I once had the up-line break while flying at normal level flight. As I recall, I didn't yell WHAP but another 4 letter word often yelled at intense moments. Sharpest outside corner that plane ever made and I am fairly certain I did not get the entire word out before the ground caused the engine to go silent. Go with a 3 letter word!

John McFayden

[Ken Culbertson]

Tim,

I once had the up-line break while flying at normal level flight. As I recall, I didn't yell WHAP but another 4 letter word often yelled at intense moments. Sharpest outside corner that plane ever made and I am fairly certain I did not get the entire word out before the ground caused the engine to go silent. Go with a 3 letter word!

John McFayden

Problem with electric is that everyone hears it!

Ken

[CircuitFlyer]

My experience using a mems accelerometer:  They are inherently a rather noise devise and some sort of filtering is required to get usable data. For stunt purposes you are likely only interested in anything under 10hz. A low pass filter can remove any higher frequency vibration and noise. A “WHAP!” Can occur in a very short period, in other words, a short but high frequency pulse. A 200hz sample rate and a threshold of about 8g could still miss a “rapid acceleration incident”. It could be expensive to test your design under real world conditions.  It’s likely best to contract that part out and leave it to someone else to collect the data for you.

[Tim Wescott]

My experience using a mems accelerometer:  They are inherently a rather noise devise and some sort of filtering is required to get usable data. For stunt purposes you are likely only interested in anything under 10hz. A low pass filter can remove any higher frequency vibration and noise. A “WHAP!” Can occur in a very short period, in other words, a short but high frequency pulse. A 200hz sample rate and a threshold of about 8g could still miss a “rapid acceleration incident”. It could be expensive to test your design under real world conditions.  It’s likely best to contract that part out and leave it to someone else to collect the data for you.

My wife has expressed an interest in learning to fly, so that test effort will be made, whether the planes are instrumented or not.