The print ran beautifully all evening. At some point it came off the plate, and the nozzle spent the rest of the night pushing it around or extruding into the space where it used to be.
A good first layer is not evidence
This is the sentence worth internalising, because it is the reason people keep being surprised by this failure.
The strength of the bond to the plate is essentially set once the first layer is down. It does not increase afterwards. What increases is the force trying to peel the part off, which grows with every layer added on top. A print that survived two hours did not have adequate adhesion; it had adhesion adequate for two hours.
The warping page covers where that force comes from in detail. What matters here is that it is cumulative and one-directional, so this failure has a characteristic timing rather than a characteristic appearance.
What actually changes between layer one and the failure
Five things shift during a long print, and each of them is a candidate.
The part gets taller. A tall part concentrates the same peel force at the perimeter through more leverage, which is why a slender column detaches when a flat plate of the same volume does not.
The bed cooled. Profiles that drop the bed temperature after the first few layers are common, and on a high-contraction material that is exactly the wrong economy — a base allowed to go fully rigid can transmit tension that a base held above its softening point simply cannot.
The room changed. Central heating going off at eleven at night is a real cause of failures at midnight. So is a window opened at bedtime, or a machine sitting near a door that gets used.
A draught started. Cooling one side of a part faster than the other makes it contract unevenly, which puts a twisting load on the bond rather than a symmetrical one.
The geometry narrowed. Many models have a large footprint that reduces as they rise. Once the cross-section above is heavier than the base can hold, the failure is arithmetic rather than bad luck.
What to change, and what to observe
- Keep the bed at one temperature for the whole print. If your profile drops it after layer three, remove that. This is the most common single fix for a mid-print release on ABS, ASA and polyamides.
- Enclose the machine. Not for the temperature at the plate, but for the air the part stands in, and to remove the draught entirely.
- Add a brim of at least 8 mm with the separation gap set to zero. If a brim is already there and lifting with the part, read brim not preventing warping — that is a different failure of the same fix.
- Split tall parts and join them. A part half the height experiences much less peel at the base, and two shorter prints also fail half as expensively.
- Print in the middle of the plate. The outer edge of a large bed runs cooler than the centre on almost every machine, sometimes by enough to matter on a marginal material.
Success criterion: run the same file again with one change and note the height the failure reaches. Failing later is real progress on this fault, and it is the only signal you get short of a completed print.
When it lets go at exactly the same height every time
That points away from adhesion and toward something positional in the machine — see prints failing at the same height. An adhesion failure has a fuzzy, load-dependent timing; a mechanical fault repeats precisely.
What people try that does not work
Increasing bed temperature past the material's range. Beyond a point the base of the part is soft rather than better bonded, and you have traded one failure for elephant's foot and a part that distorts where it meets the plate.
Layering on more adhesive. A thick glue film shears within itself. If adhesive is genuinely needed, it wants to be a thin even coat on a warm plate — see adhesive not improving adhesion.
Assuming the plate is clean because it looks clean. Skin oil is invisible. A plate that has been handled since its last wash is a plate with a reduced bond, and that shows up as a mid-print release long before it shows up as a first layer that never sticks.
The materials that let go
ABS, ASA, polycarbonate, the polyamides and polypropylene — everything with high contraction. On those, an enclosure and a hot plate held constant are worth more than every other measure combined. PLA detaching mid-print is unusual enough that it is nearly always a contaminated plate or a cold draught rather than a material property.