3D Printing Warping: Why 3D Printed Parts Warp, and How to Stop It
A part that comes off the plate already bent was not damaged by the machine. It was held flat while its stress built up, and it bent the moment the constraint was removed. That distinction matters, because it changes what you do about it: 3D printing warping is not a machine fault, it is an energy problem, and the order you attack it in decides how much of it you ever have to argue with a supplier about.
Nobody can give you a warping figure that will hold for your part. Not a percentage, not a maximum deflection in millimetres, not a guarantee. Warping is set by your geometry, your orientation, your machine and your material together, so any number quoted without those four is decoration.
What can be given is an order. There is a cheapest move, then a cheaper-than-the-machine one, then two that cost an extra cycle. Work down that list and most warping stops being a conversation at all.
What actually causes a part to warp?
Every layer is melted, or bonded, while the layer beneath it is still hot. The top of the part is therefore hot and wants to shrink. The bottom is already cold, and it is bolted to a steel plate that is not going anywhere. The top contracts, the plate refuses, and the difference between those two facts has to go somewhere. It is stored in the part as residual stress.
While the part is on the plate, the plate holds it flat. Nothing moves, and the part looks perfect — which is exactly why this defect arrives as a surprise. The moment the part is cut off, the constraint is gone and the stored stress resolves itself. The part does not have to be damaged for it to bend. It only has to be free.
Two properties of your geometry decide how much bending you get, and both are structural rather than thermal. The first is how much of the part is free to move: a large flat unsupported plate has far more opportunity to distort than a ribbed or dished one of the same mass. The second is how stiff the section is along the gradient: a thin, wide section bends easily, and no process setting changes that.
Swipe the diagram sideways to read it →
Is warping a machine fault, and should you change machines?
Usually no. If a part warps on one machine and not another, the machine is not necessarily at fault — the two machines simply have different thermal strategies, and you have not yet compensated for the difference. Changing machine is the most expensive way to attack a problem that usually starts with geometry.
There are machine-level parameters that genuinely affect this: bed temperature, scan strategy and how the thermal field is shaped around the part. These are set per machine and per material, and they are not something you control from a drawing. Which is why the honest answer to a warping complaint is not "buy a better machine" but "tell us what the part is, and we will tell you which of these six moves applies to it".
Roughly speaking, 3D printing warping divides cleanly into the part you designed and the process you chose. The part you designed accounts for most of it, and you have more control over that half than you think — which is the argument for working through the list below before you start comparing machines or suppliers.
One genuine machine-side exception is worth naming: an uneven or inconsistent thermal history between parts on the same plate, or a plate that is not flat or not properly prepared. If some parts on a plate warp and others do not, that is a setup question, not a design question.
Why does SLM warp more than SLS or SLA?
The material matters more than the process name. Metal has no stage where you can relieve stress by letting the part cool slowly under constraint, so in selective laser melting (SLM) the stress has to be managed deliberately. Our SLM page lists the two measures that matter: planning orientation to keep witness marks off functional faces, and a controlled stress relief cycle that relaxes the residual stresses built up during melting, which is what stops thin or large flat sections from distorting.
Polymer behaves differently, and better, for a reason that is easy to state. In selective laser sintering (SLS) the whole part sits in a powder bed at a temperature well below the melting point of the material, and the bed itself supports the part. No supports are required at all, which is the single biggest structural difference: a supported part is a part whose shape is being held by something else, and removing that something is a bigger change than removing a support.
That does not make SLS immune. A large flat SLS plate can still curl at the edges, because the edges are the part that cools first and is least supported by loose powder. It is a smaller effect, and it responds to geometry in the same way.
SLA is a third mechanism again, and it is worth separating because it is often misdiagnosed. Resin does not melt and does not shrink like a melt pool, so the warping people report in SLA is usually cupping from support removal and peel forces, plus the slight shrinkage our FAQ notes as the resin cures — not a stored thermal gradient. If your resin part is bowed, adding stress relief will not help you, because there is no residual thermal stress to relieve. Fixing it means changing how it is supported and how it comes off the plate.
The practical summary is in the tolerance figures we publish for each process: typical tolerance is ±0.2 mm for SLM, ±0.3 mm for SLS and ±0.1 mm for SLA. Those are dimensional figures, not warping allowances, and a part that bows by more than its tolerance has a real defect. Which is the argument for treating warping as something to specify rather than something to expect.
What order should you attack warping in?
Cheapest first, and this ordering is the actual content of the answer.
- Change the geometry. Ribs, fillets, dished or dished-out surfaces, and symmetry. Stiffness comes from section shape, not from adding material everywhere. A ribbed plate of the same weight as a flat plate will behave completely differently, and this costs nothing on the drawing.
- Change the orientation. Put the direction of greatest stiffness along the thermal gradient, and orient the part so that its stiffest axis runs with the plate. This is a decision made once, costs nothing, and frequently removes the problem entirely. It is also the one most often skipped.
- Anchor the part properly. More deliberate support contact area means more constraint, held on purpose rather than by accident. This is where supports stop being an accessory and become part of the distortion strategy.
- Ask what the thermal strategy is. Bed temperature and scan strategy are machine parameters. You are not going to set them, but you are entitled to know whether your part's geometry is compatible with the machine the supplier actually runs.
- Stress relieve before release, not after. A controlled thermal cycle while the part is still held flat relaxes the stress without giving it a reason to move. This is the single most commonly missed step, and it is the reason parts arrive bent.
- Fixture the part during stress relief. An unrestrained part in a furnace simply bends in the furnace instead of on the plate. If the part goes into a stress relief cycle loose, you have changed when it warps, not whether.
The first three are drawing and planning decisions that cost nothing, and they remove most of the problem. The last two are process decisions with real cost in cycle time. How the process actually runs is worth reading before you ask for any of them, because the order the steps happen in is the whole point.
This is also why 3D printing warping is so often discussed as if it were a material property. It is not a property of the material at all — the same titanium and the same stainless steel behave completely differently on the same machine depending on how the part was drawn and which way up it was built. Fix it at the drawing and it stops being a material conversation.
What can you control, and what should you ask a supplier?
Four of the six moves above are decisions you can make before the part is ever quoted. Geometry, orientation, support strategy and whether stress relief is in the route: those four are yours, and specifying them is free. The other two are the supplier's process, and the only thing you can do is ask.
| What to ask | Why it matters | If the answer is vague |
|---|---|---|
| What orientation will you build this in, and why? | Orientation is the cheapest distortion control there is. | The default is whatever nests best, not what distorts least. |
| Is stress relief in the route, and when? | Before release it helps. After release it is too late. | They will not know the difference between the two. |
| How is the part held during stress relief? | An unrestrained part bends in the furnace. | No fixturing is being considered. |
| What tolerance applies to flatness, not just size? | Size tolerance says nothing about bow. Flatness is a separate callout. | Bow is being treated as out of scope. |
| What is the largest flat section you will build this way? | Every process has a sensible ceiling for unsupported span. | The part is being attempted past the point where it is predictable. |
That last row is the one worth pressing on. We would rather tell you that a wall or a span is too thin or too wide before you order it than print it and hope — a part that is geometrically marginal behaves differently on every machine, which makes it unquotable rather than merely imperfect. That is the same position we take on wall thickness, and it is also what the SLS versus SLA versus SLM comparison is really about: which process you choose determines which of the six moves above you are actually fighting.
Where this leaves you
The useful summary is short. 3D printing warping is residual stress finding a new equilibrium, not damage, and the part looked fine right up until the moment nothing was holding it flat. Geometry and orientation are free and solve most of it. Support and anchoring solve some of the rest. Stress relief solves the rest only if it happens while the part is still constrained, which is the step that gets skipped.
None of this requires a better machine. It requires the geometry to be right, the orientation to be chosen rather than inherited, and the release and stress relief steps to be in the right order. If you send us a part that has warped before, tell us which of those six was missing — that is usually a faster conversation than sending the file again and hoping.