Catastrophic spagetti nest  

Re: Catastrophic spagetti nest

Hi Janis

I think most of us have had that...

I think the best advice would be to either reduce speed or set mode to high power.

Are the default acceleration settings in the start GCode?

M201 X1000 Y1000 E600 ; set default acceleration

Peter

Re: Catastrophic spagetti nest

Peter

Many thanks, I will check! 🙂 Glad it is not just me, but common thing and printer is OK.

But I'm also very much interested in understanding mechanics of this -
- is this loss of coordinates? (I had it once, it produced a shift, but not this spagetti cloud)
- multiple hits agains the mesh and constant shif in corrdinates?
- anything else besides that?
- any attempts to detect this situation using computer vision and machine learning?

So, extruder head is moving, hitting and obstacle, stalled, then again, and again ... right?

Janis

Re: Catastrophic spagetti nest

Janis

I think in this case, the top of the part was broken off as well and there was nothing underneath to receive the filament.

This is not a "common occurrence" but it is one thing that can happen due to user error... Every time it has happened to me has ultimately been my fault.

One thing to remember - the processor is only a 16 MHz single-core. 3D printing is pushing the processor quite hard, so to build in any sort of intelligence is rather wishful thinking. Imagine trying to 3D print with one of the first 80286 PCs from 30+ years ago and then try to add AI...

Peter

Re: Catastrophic spagetti nest

Ah! Now it's cllear and logical - part of the model was knocked off.
Also, your remark about processor power vs printing speed is very usefull to me.

Thank you, Peter! Avesome help 🙂

Re: Catastrophic spagetti nest

hmmm, while the printer and it's on-board controller may not have much processing power to handle such an endeavor (machine vision), it would be very cool if it would be possible to integrate a host PC (which could be much more powerful) running something like Repetier Host and a camera. If you could pull the gcode from repetier and use the layers as your expected scenario and then compare to an actual view from the camera, you might be able to detect variations. Perhaps if you see excessive material outside of the target build envelope, the printer would auto-kill the print and send you a text with picture saying it failed...would be nice.

So essentially you might not need any AI or machine learning at all, just some basic edge detection integrated with the slicing software.

This is very pie-in-the-sky, but its fun to dream 😀

Re: Catastrophic spagetti nest

Kevin

People like Janis and I do not print via a PC, but use OctoPrint running on a Raspberry Pi.

Now whilst the Pi is maybe 50 times more powerful than the printer processor, it is still nothing like sufficient to optically recognise a relatively very fast moving printer in order to determine if there was an issue.

Edge detection works very well on stationary objects, but fast, random movement would make things extremely difficult. Ant that is assuming you could see the whole of the printed model.

Peter