You open the enclosure to find a tangle of extruded strands, a nozzle drawing shapes in mid-air, and nothing recognisable as the part.
This page cannot fix the spaghetti
The strands are not a defect with a cause of their own. At some point the nozzle stopped having anything to print onto — the part came loose, a layer moved out from under it, or a support tower fell — and the machine, which has no way to know that, carried on faithfully extruding into empty space for however many hours were left.
So the useful investigation is not "why did it produce spaghetti". It is "what was underneath the nozzle when it stopped being there".
What it cost, so the investigation is worth doing properly
Take a 180 g part that failed at hour nine on a machine drawing 120 W, printing filament bought at $24/kg, with electricity at 17.5 cents per kWh and machine time costed at fifty cents an hour. The filament is $4.32, the electricity is $0.19 for 1.08 kWh, and the machine time is $4.50 — $9.01 gone, and nine hours of a machine that could have been printing something else.
The failure rate is deliberately set to zero in the cost calculator, because $9.01 here is one job's raw loss rather than a surcharge spread across good parts. The point of the number is that it is large enough to justify twenty minutes of diagnosis rather than an immediate reprint with the same settings.
The wreckage still holds the evidence
Look at these in order, before clearing the plate:
- Is the part still attached anywhere? A part that let go entirely leaves a clean outline of first-layer plastic or a brim still bonded to the plate. That is detachment mid-print.
- Is the loose part curled at the corners? Then it lifted before it let go, and the real fault is warping — the nozzle probably knocked the raised corner and pushed the whole thing free.
- Is the part intact but sitting displaced on the plate? It moved as a unit, which means it was pushed rather than peeled. Look at nozzle hitting the print.
- Is there a collapsed tower of support material in the mess? Supports collapsing removes the floor under an overhang, and the print above it has nowhere to go.
- Is the lower part of the print offset from the upper? Then it did not detach at all — a layer shift moved the toolhead off the part.
At what height the strands begin also tells you when it happened, which pairs with anything else that occurred at that time — a draught from a door, someone opening a window, a spool that ran out.
Reducing the damage rather than the rate
Failures cannot be driven to zero, but their cost can be cut sharply.
Watch the first layer finish. Most detachments are decided in the first few minutes, and the four minutes it costs to watch it complete are the cheapest insurance available on a long print.
Start long prints when you will see them again soon. A failure at minute forty of a twenty-hour job costs almost nothing if it is caught; the same failure discovered next morning costs the whole spool section and the whole night.
Treat camera-based failure detection as a partial net. It reliably catches an established tangle. It does not catch the layer shift that caused it, and it will not catch a part detaching under a solid top surface.
Split very tall parts. Height is what makes both detachment and collision likely, and two joined halves fail half as expensively.
Clearing the nest without creating a second fault
Strands wrap around the hotend and weld into a solid collar, and the temptation is to pull it off cold with pliers. That collar is gripping the silicone sock, the thermistor and the heater cartridge leads, and a hard pull takes them with it. A thermistor tugged half out of the block is how a tidy-up becomes a thermal runaway halt — or, if it stays put but reads the air instead of the metal, something considerably worse.
Heat the block to the material's printing temperature, take the toolhead somewhere you can reach without leaning over the bed, and peel the softened plastic away by hand with a glove on. Then check the sock is seated and the wires are where they were before starting the next print.
The materials most involved
Everything that contracts hard: ABS, ASA, the polyamides and polycarbonate. On those, spaghetti is usually the last act of an adhesion failure that started as a lifted corner, which is why an enclosure and a hot plate prevent more of these than any firmware feature does. On PLA it is much rarer, and when it happens the cause is more often a collision or a support failure than a bond that gave up.