Corners curl up off the plate. Long flat parts come out banana-shaped even when they never fully detached. A part that looked fine at layer twenty has visibly lifted by layer two hundred.
What is actually happening
Every layer is laid down molten and immediately begins to contract as it cools. The layers below it have already finished contracting and are dimensionally fixed, so each new layer pulls on a base that will not move with it. The result is a stack of layers in tension, anchored only at the plate.
That tension has to go somewhere. The least restrained place is a corner of the first layer, where the bonded area is smallest relative to the leverage applied to it, so that is where the part lifts first. It is also why a long flat part warps more than a tall narrow one of the same volume — the leverage grows with the distance from the centre.
This is why the material matters so much. PLA contracts 0.2 to 0.5% as it cools; ABS and ASA contract 0.4 to 0.8%; unfilled nylon contracts up to 2%. Roughly speaking, four times the contraction is four times the load working to lift a corner away from the plate, applied continuously for the whole print.
Ranked causes
- A high-shrinkage material printed without an enclosure. If the material is ABS, ASA, polycarbonate, unfilled nylon or polypropylene, this is the cause until proven otherwise.
- Moving air across the part. An open window, a fan, an air-conditioning vent, or the printer's own part-cooling fan running at full speed on a material that does not want it.
- Bed temperature below the material's glass transition. A base held above its softening point cannot build tension; a base allowed to go rigid can.
- A large flat footprint with sharp corners. Geometry alone can turn a marginal case into a failure.
- High infill. More material inside means more contracting material pulling inward.
Fixes, in the order worth trying them
Take the plate up to the hottest setting the material allows. Free, immediate, and it addresses cause three directly. Keep it there for the whole print rather than dropping it after the first layer — a bed that cools mid-print is a common cause of a part that lets go at hour two.
Enclose the printer. Even a cardboard box over an open machine makes a measurable difference, because most of the problem is convective cooling rather than radiation. Close the window in the same room while you are at it.
Turn the part-cooling fan down or off for ABS, ASA, polycarbonate and nylon. This feels wrong if your instincts were formed on PLA, where cooling is nearly always good. On high-shrinkage materials, cooling is the enemy.
Add a brim, and add mouse ears at sharp corners. Ten millimetres of brim, with the separation gap set to zero so it is genuinely bonded. A narrow brim with a gap is decoration.
Round the base corners in CAD if the design allows it. A radius spreads the peeling force that a sharp corner concentrates, and it is often the difference between a part that lifts and one that does not.
Reduce infill. Counter-intuitive, but less material inside means less internal tension. Increase wall count instead if the part needs to be strong.
What does not work
Increasing bed adhesion indefinitely. You can reach a point where the bond is stronger than the plate's own surface, and then the plate loses. Adhesion is one side of the competition; on a high-shrinkage material the winning move is to reduce the force on the other side.
Switching to a "low-warp" grade of the same polymer without changing anything else. Fibre-filled ABS and nylon genuinely do warp much less — the fibre restrains contraction — but they still want the enclosure, and buying one without the other usually disappoints.
Telling warping apart from things that look like it
A part that is dimensionally wrong but flat is shrinkage, not warping, and it is compensated in CAD or in the slicer rather than fought on the plate. A part that split horizontally partway up is delamination, which shares the same contraction cause but is fixed with more heat and less cooling rather than with better adhesion. A part with a bulging base that otherwise sits flat is elephant's foot, which is a Z-offset problem in the opposite direction.
Getting this distinction right saves a lot of wasted effort, because the three failures have overlapping fixes and non-overlapping causes.
When it is the machine's fault
If a low-shrinkage material warps on a machine that used to print it fine, suspect the bed rather than the profile. A bed mesh is a snapshot of the plate's shape at the moment it was probed, and plates take a permanent set over months of heat cycling. Re-probe at printing temperature — a heated bed is a different shape from a cold one, and a mesh measured cold and used hot is a mesh measured on a plate that no longer exists.