You cut the model in two because it was taller than the machine, printed both pieces overnight, and this morning you set them face to face and there is a step you can catch a fingernail on. Split 3D prints never line up on the first attempt because three unrelated errors all land on the same seam, each one small enough to ignore on its own, and the fix for one of them does nothing for the other two. Sanding harder is what most people try first, and it is the right answer to only one of the three.
This post contains affiliate links. If you purchase through these links, 3D Prototyping Hub may earn a small commission at no extra cost to you.
This post also contains Amazon affiliate links. As an Amazon Associate we earn from qualifying purchases.
Read the Seam Before You Sand It
Where the error sits tells you which of the three you have. Dry fit the halves on a flat surface, press them together, and look at the joint from the side.
| What you see |
What it means |
Where to go |
| Faces touch, but one half overhangs the other all the way round |
The two pieces are different sizes |
Cause 2 |
| Faces meet in the middle and gape at the edges, or the pair rocks |
The mating faces are not flat |
Cause 1 |
| Faces sit flush, but features are rotated or offset across the seam |
The alignment features are not locating anything |
Cause 3 |
| Fit was good dry and bad after glue |
Adhesive film thickness and clamping |
Dry fit is the only fit |
| A step on one edge only, same end of both parts |
First-layer squish or a non-flat bed on that corner |
Cause 1 |
Do all of this dry. A joint that is wrong dry gets worse with glue in it, and glue removes your ability to check anything afterwards.
Cause 1: The Mating Face Is Not Flat
The face you cut the model on is almost always the face that ends up on the build plate, and the build plate is where the two most common flatness errors come from.
Elephant foot spreads the first few layers outward into a lip around the whole perimeter. Two of those lips meet before the faces do, so the halves rock on a ridge and gape in the middle — why 3D prints have elephant foot covers the first-layer squish and the slicer compensation that removes it at the source. Put a 0.4 to 0.6mm chamfer on the mating edge in CAD as well, which both hides any lip that remains and gives the glue somewhere to go.
Warp lifts corners away from the plate during the print, and a face that lifted 0.3mm is a face that will never sit flat afterwards. Large flat-bottomed parts are exactly the geometry that warps most, so the split that was supposed to make the print easier can make this failure more likely; why 3D prints warp and how to fix it is the one to read if the gap is at the edges rather than in the middle.
Plate texture transfers. A textured PEI sheet puts its pattern into both mating faces, and two textured faces meet on their high points with a film of air between them.
The repair is a flat reference, not a sanding block. Lay a sheet of 120 to 400 grit abrasive on a piece of float glass or a glass build plate, hold the part square, and move it in a figure of eight. Stop when the scratch pattern is even across the whole face — an uneven pattern is a map of where the face was high, and it is the only flatness gauge you need.
Cause 2: The Two Halves Are Not the Same Size
Every thermoplastic shrinks as it cools, and it does not shrink equally in every direction. PLA and PETG are usually quoted in the region of 0.2 to 0.4%, ABS and ASA closer to 0.5 to 0.8%, and the figure along Z is not the figure in X and Y because the layers are constrained by the plate while they cool and the part is not.
That matters at a seam because of arithmetic. Across a 200mm face, a 0.4% difference between the two halves is 0.8mm of step — visible from across the room, impossible to sand out, and entirely invisible when each piece is measured on its own and found to be within a few tenths of nominal.
Four rules remove most of it:
- Same orientation. If one half prints flat and its mate prints standing up, they shrink along different axes and will not match. Orient both the same way, even when that costs you supports.
- Same material and same spool. Two spools of the same brand and colour can carry different pigment loading, and pigment changes shrinkage. Why a new spool of filament prints worse covers how much a batch change actually moves.
- Same session, same conditions. A part printed in a cold garage in the evening is not the same size as one printed in the afternoon.
- Measure against the model, not against each other. A pair of digital calipers and the dimension from CAD will tell you which half moved, which is what decides whether you compensate or reprint. Why 3D prints are the wrong size works through the per-axis compensation.
The effect scales with the part, which is why this cause barely exists on a 60mm model and dominates on a 400mm one — why 3D printed parts fail when you scale them up is the longer version of that argument.
For ABS, ASA and nylon the controlling variable is ambient temperature during the print, because a part cooling in moving room air shrinks differently from one cooling in a warm chamber. That is a hardware answer rather than a settings answer: QIDI's own store lists enclosed machines with heated chambers, and why your printer can't print ABS or nylon explains what the chamber is actually doing.
Cause 3: Your Alignment Pins Are Fighting You
Printed pins and sockets are the standard way to register two halves, and they are also the standard way to guarantee they never sit true.
A hole printed on an FDM machine comes out undersize, because the extrusion is laid on a polygonal approximation of a circle and then pulls inward as it cools. A pin printed on the same machine comes out oversize, for the mirror-image reason. Both errors stack in the direction that makes the fit tight, so a pin modelled at nominal diameter either will not go in, or goes in with enough force to split the socket. The usual reaction is to open the clearance until it drops in easily — at which point the pins are no longer locating anything and the halves can sit a few tenths of a millimetre off in any direction.
What works:
- 0.15 to 0.25mm diametral clearance on a 0.4mm nozzle, confirmed on a test coupon rather than assumed. Print one small plate with four sockets at different clearances before committing it to a 12-hour part.
- Two features, not four. Two pins and a flat face fully locate a joint. Four pins over-constrain it, and when the four holes disagree by a tenth each, the part finds a compromise position that satisfies none of them.
- One round pin and one slot is better still. The round pin fixes position, the slot fixes rotation while letting the pair tolerate a little shrinkage difference along the seam. This is the standard answer in tooling, and it costs nothing to model.
- Chamfer the socket mouth. The lead-in does the finding, so the fit does not have to be loose to be assemblable.
When the alignment genuinely has to be right, stop printing the pin. Model the hole around 0.3mm undersize, open it with a brad-point drill bit run slowly by hand, and fit a steel dowel pin. The pin is round, straight and constant in diameter, which is three things no printed pin is, and the hole is now a drilled hole rather than a printed one. Go slowly: fast drilling melts the bore and the plastic closes behind the bit.
Where You Split It Decides How Much It Shows
The split line is a design decision and it is usually made by the slicer's default cutting plane, which has no opinion about your part.
- Cut on an existing edge or feature line. A seam in a corner radius or along a panel line is nearly invisible. The same seam across a flat face is the first thing anyone sees.
- Leave both faces printable flat. A cut that gives one half a flat bottom and the other a 40-degree overhang has traded a seam problem for a support problem.
- Use a lap or step joint rather than a butt joint. A stepped seam doubles the glue area, self-aligns along one axis, and stops the joint being a single plane of weakness.
- Keep the seam out of bending. A glued joint is weaker than the plastic around it, so put it where the part is loaded in compression. Why glued 3D prints come apart goes through adhesive choice against the load the joint actually sees.
Dry Fit Is the Only Fit That Counts
Adhesive has thickness, and on a large seam that thickness is part of your tolerance budget. A thin cyanoacrylate contributes a few hundredths of a millimetre; a gap-filling epoxy can contribute half a millimetre if you let it, and it will set with whatever misalignment you clamped in.
Pick the adhesive by working time rather than by strength. Gel cyanoacrylate grabs in seconds, which is fine on a 50mm seam that already fits dry and useless on a 300mm one — you cannot position two large halves in the time it gives you, and once it has grabbed crooked the joint is finished. A five-minute two-part epoxy gives you a window to slide the parts true, check the seam from both sides, and clamp.
Clamp against a flat reference rather than against the parts. Set the assembly on glass or on the build plate while it cures, so gravity holds the two faces coplanar instead of holding them wherever your hands left them. Masking tape pulled across the seam in several places is a better clamp than a bar clamp on a printed part, because it applies even pressure without crushing the walls.
If the joint has any chance of needing to come apart — a prototype, an electronics enclosure, anything you will iterate — do not glue it at all. Heat-set threaded inserts and a pair of bolts pull the faces together with real force, survive being undone, and let you re-seat a seam you got slightly wrong instead of living with it.
Fixing a Seam You Already Glued
A hairline joint is a finishing problem, and finishing is good at hairlines. Two or three light passes of filler primer, sanded back between coats, will fill a seam and the surrounding layer lines in the same operation so the repair stops reading as a line. Why 3D prints still show layer lines after sanding covers the sequence that actually gets a surface flat, and it applies unchanged here.
What filler cannot fix is a step. If one half stands proud of the other by half a millimetre, sanding the step down removes material from one side of the joint and leaves a visible dished band where the profile no longer matches. That is a reprint, and knowing it early saves an evening.
Or Stop Splitting It
Every section above is a way to manage a joint. The joint is the problem, and the only fix that removes it is a build volume the part fits in.
That is worth pricing honestly against the time. A larger machine prints the part whole, which eliminates the seam, the pins, the glue, the clamping and the finishing in one move — Anycubic's US store and ELEGOO's US store both list large-format FDM machines alongside their standard sizes, and our large-format printer guide compares what the extra volume costs in print time and footprint.
For a one-off, buying a machine for a single part is the wrong answer and a bureau is the right one. A shop with a large-format machine prints in one piece at sizes a desktop printer has to cut up, and large-format 3D printing services covers how to specify that job. When you want quotes, the providers in our directory are listed by location and process, and sending the whole model is usually cheaper than the third attempt at the seam.
Hero photo by Efe Yağız Soysal via Unsplash, unsplash-free licence.
