What the hell??? Violent random motion

[epicycloid]

Last night I was running a program from Aspire. Pretty straightforward. After it had been running for a little while, made a few passes, and I thought it looked safe, I hit F5 on the keyboard to halt the machine so I could slide my dust shoe under the spindle and let it continue (I was following the bit with a hand vacuum until then, as I often do).

My Halt routine just raises up the spindle 1" and turns it off. Not enough room to slide the dust shoe under, so I jogged the Z axis up a ways and turned around to grab the dust shoe.

When I turned back around the spindle (not running) was feeding down into the wood, and jammed into it about 1/2" below the surface. Then the router violently started moving in some random path, well outside the programmed space, and knocked all the clamps and other stuff around on the table.

I hit F5, nothing, I hit the e-stop on the front of my machine. Whew! That stopped it. Now it's in some weird unknown feed hold mode, with the bit embedded into the wood and parts strewn all over the spoil board. I'm in a semi-panic, so no, I didn't really pay attention to much on the screen. I just wanted to jog Z up, free the impaled wood, and collect my wits.

I hit Init, thinking that would let me jog, but the violent movement started again. I hit the e-stop and switched off power to the machine. I'm still in a semi-panic.

KMotionCNC now has a hundred dialogs telling me it can't see the KFLOP, and of course I can't dismiss them and quit / exit, because the next one pops right up. Ok, Ctrl-Alt-Del, Task Manager, end task, power the machine back up, Init, jog Z up, wiggle off the destroyed workpiece, shut everything down and go in the house. Take a deep breath. That was scary.

A friend with a similar machine said he's had a similar thing happen twice, no idea why. He's had his machine a lot longer than I have. Mine's closed-loop, his is open-loop, other slight differences, but same custom builder.

I'm not sure there is any way to reproduce this, but I can try in a few days after I'm past my current deadline.

FWIW, I'm running 4.34H. My Init, Home, Safe Z on Halt and Touch Off are attached, just in case there's something really odd I'm doing.

My question is, is this something others have seen?

Is it some known issue when the KFLOP gets in some odd state?

--Jon

[TomKerekes]

Hi Jon,

Yikes! Sorry to hear this.

I'm not aware of any such issue.

I see you have a Touch Off program that would likely drive Z down forever if inadvertently activated. Is it possibly assigned to a Hot Key right next to F5?

It might be that jamming the Z down forced and held the XY servos off position long enough for them to eventually result in a violent over correction.

I see the Following Error Limits are maxed out at 1 billion counts. You might consider setting them to a small value. In the event of hitting a hard object or being forced off position the axis should then disable and stop and cause less damage. You would need to determine the max following error you ever have under normal conditions and then set the limit slightly larger.

Please let us know if you are able to reproduce the issue.

[Moray]

Only bizarre things I've had involved Mach 3, and it randomly running the turret into the spindle on my lathe. Since moving to KMotionCNC, the only crashes have been self inflicted.
Tom, should the MaxI be as high (10'000)?

[epicycloid]

Thanks Tom. I'll dig into this a bit more next week. I'll let you know if it's reproducible.

I ran a few things today with no issues. Which of course means last night's issue is probably a hard to find gremlin.

--Jon

[TomKerekes]

Jon,

As Moray points out MaxI of 10000 doesn’t really make sense with max Output of 250. No point of allowing the Integrator to ramp to 10000 where the Output will only ramp to 250. This would require a long time to Ramp back down to recover after a persistent error.

[epicycloid]

Thanks Tom. I was just using numbers from Carl, that were probably inherited (passed down?) from other prior machine configurations. So I haven't actually done anything to asses what better values might be.

That said, I think our (two identical) machines are the first ones he's built using these new-ish Leadshine closed-loop steppers and drivers. I have no idea where his values came from, but he did say in a conversation that the MaxOutput should be at (no more than) 250, but I don't remember the reason for that.

I'll dive back into this when I get back home from the symposium I'm attending, and can dig into the code and documentation a bit more.

--Jon