Almost every printer has a box somewhere in the corner of the workbench filled with parts nobody talks about: the torso printed in two halves, the enclosure panel that was 15 mm too long for the build plate, the prototype that had to be split before it could be sliced. Joining printed parts back together is one of the most common things anyone with a printer actually does — and one of the least documented.
The instinct is to treat it as a glue question: which adhesive is strongest? That framing is where most failures begin. A glued joint is not primarily limited by the adhesive. It is limited by geometry — specifically, how much load-bearing area the joint offers and how that area is oriented relative to the layer lines the printer left behind. A mediocre adhesive on a well-designed joint will outperform the best adhesive on a butt seam, every time.
This guide covers the design side of joining: what makes printed surfaces bond differently from injection-moulded ones, how to choose an adhesive by material, and how to shape a joint so the load goes through the material instead of through the glue line.
Why Printed Surfaces Bond Differently
A printed face is not a machined face. Three properties of FDM and resin parts change the rules of adhesion.
- Layer lines are both an asset and a liability. Slightly rough, they give an adhesive far more surface area to key into than a smooth face would — which is genuinely helpful for shear. But the same lines mean the bonded interface is a series of ridges, and any peel or cleavage load finds the valley between ridges first. Printed joints fail from the edges inward, not uniformly.
- The material is often low-surface-energy and chemically inert. PLA, PETG and polypropylene-adjacent materials do not want to be wet by a polar adhesive. PETG in particular is famously slippery in this sense — it resists adhesion to itself even when hot.
- Parts are not flat. Even on a well-tuned machine, a printed face has measurable waviness, plus elephant's foot at the bottom edge and slight lifting at the corners. Two "flat" faces pressed together usually touch at a few high spots, leaving most of the intended bond area as a gap for the adhesive to bridge rather than fill.
The practical consequence is that the strongest joints do not rely on glue alone. They are designed so that the adhesive's job is to hold the parts in place while a mechanical feature carries the load. Geometry and adhesive are not alternatives; they are a division of labour.
Surface Preparation Beats Adhesive Choice
Given identical parts and identical adhesive, preparation can change joint strength by a factor of two or more. It is also free. The order matters.
| Step | What it does | Watch out for |
|---|---|---|
| 1. Remove supports and clean | Support scars and dust are the most common cause of a clean-looking joint that will not hold | Do not sand away the layer texture entirely — you are removing usable keying area |
| 2. Scuff the bond face | 120 to 240 grit. Creates mechanical keying and removes mould-release-like surface films | Wipe away all sanding dust; dust is a bond breaker |
| 3. Degrease | Isopropyl alcohol, 90% or higher. Removes skin oils, which are the number one contaminant | Let it flash off fully. Solvent trapped at the interface weakens the joint |
| 4. Improve surface energy (for hard-to-bond plastics) | A brief pass with a flame or an atmospheric plasma wand oxidises the surface so adhesive can wet it | Keep the flame moving; you want a sheen change, not a melted corner. Never on thin walls |
| 5. Bond within the open time window | Surface energy decays within hours. Prepare and bond in the same session | Do not prepare parts tonight and glue them tomorrow |
Step 4 is the one most people skip and the one that unlocks materials that "cannot be glued". A five-second pass of a butane torch across a PETG or nylon face — enough to see it change from matte to slightly glossy — measurably improves wetting. It is the same principle the industry uses at scale with plasma treatment, and it costs nothing.
The same logic drives the finishing and post-processing discipline: the surface you present to the adhesive is a decision you make deliberately, not something you inherit from the printer.
Choosing an Adhesive by Material
There is no universal best adhesive. There is only the best adhesive for a given plastic pair. The table below reflects how each family actually behaves.
| Material | Primary choice | Alternative | Notes |
|---|---|---|---|
| PLA / PLA+ | Cyanoacrylate (CA) gel | Two-part epoxy, solvent welding with dichloromethane | Bonds readily. Brittle joints — use epoxy where shock is expected. PLA+ behaves the same |
| PETG | Two-part epoxy or polyurethane | CA with a primer, or flame treatment first | Notoriously hard to bond. Requires prep and a flexible adhesive that tolerates movement |
| ABS / ASA | Solvent weld with acetone or an ABS slurry | Two-part epoxy | Solvent welding produces a joint as strong as the base material. See ASA for outdoor parts |
| Nylon (PA, PA-CF) | Two-part epoxy after flame or plasma treatment | CA with primer | Moisture-sensitive: dry the part before bonding or the joint will fail at the interface |
| TPU / flexible | Flexible polyurethane or contact adhesive | Shoe-goo type flexible cement | Rigid adhesives crack as the part flexes. Match the adhesive's flexibility to the part's |
| Photopolymer resin | CA gel or two-part epoxy | Resin-and-light: re-cure a brushed resin layer with a UV torch | Re-curing with the original resin gives an optically seamless joint on clear parts |
| Carbon- or glass-filled | Two-part epoxy with a thickened filler | Scuff aggressively first; fibres resist keying | Fibres also make the surface abrasive to your tools |
Two rows in that table deserve emphasis. ABS and ASA are the only common filaments that can be solvent welded — dissolved at the joint until the two parts become one continuous material, which is why enclosure and automotive parts are usually joined this way. And resin parts have an option no filament has: brush on unpolymerised resin, press the parts together, and cure the seam under UV. On clear resin this produces a joint you cannot find by eye.
Six Joints That Carry Load
Adhesive strength is expressed as force per unit area. Every structural joint design is an attempt to convert the load into the mode the adhesive handles best — shear — and away from the mode it handles worst: peel.
| Joint | How it works | Best for | Design note |
|---|---|---|---|
| Plain butt | Two flat faces, adhesive only | Cosmetic seams, non-structural | Weakest option. Never use where the joint carries bending |
| Rabbet / lap | Overlapping step gives shear area and self-alignment | Panels, boxes, enclosures | The single highest-value upgrade over a butt joint, and easy to add in CAD |
| Dowel pins | Short printed or steel pins locate the parts and take shear | Long seams, alignment-critical joints | 0.2 mm clearance; consider a tolerance test coupon first |
| Tongue and groove / key | Mechanical interlock along the seam | Parts loaded in torsion | Add 0.25 to 0.4 mm clearance, or print the key separately and sand it to fit |
| Dovetail / sliding | Interference geometry resists pull-apart without adhesive | Assemblies meant to come apart | Print clearance generous; dovetails are sensitive to elephant's foot |
| Threaded insert + bolt | Fastener clamps the joint | Anything that will be serviced | Strongest and most reversible. Heat-set brass inserts beat printed threads for repeated cycles |
The design rule that follows from this table is simple: give the joint a step before you give it a pin, and give it a pin before you give it more glue. A rabbet joint on a two-half print costs an extra five minutes in CAD and roughly doubles the bond area, while also making the parts self-align so they cannot slide out of position during clamping.
Controlling the Glue Line
Beginners assume a thinner glue line is always stronger. That is true for rigid structural adhesives, but only down to a point, and the failure mode of a too-thin line is different from the failure mode of a too-thick one.
- Too thick (over 0.3 mm): the adhesive becomes the load path rather than the parts. Most adhesives are weaker in bulk than the plastic around them, so a thick line creates a soft layer in the middle of a stiff assembly.
- Too thin (under 0.03 mm): the adhesive is starved. There is not enough material to fill the surface waviness, and the bond depends on the few high spots that happen to touch — which is exactly the "I clamped it hard and it still popped off" failure people blame on the glue.
- Working range: roughly 0.05 to 0.15 mm. That is thin enough for stiffness and thick enough to bridge the waviness of a printed face.
The practical consequence: stop clamping so hard. Clamping pressure should be enough to hold alignment and squeeze out excess, not enough to starve the joint. For CA, a light finger press for 30 seconds is usually correct; for epoxy, finger-tight with tape is plenty. The exception is solvent welding, where a firm clamp is right because you are not relying on the glue line as a layer at all.
Accounting for the Glue Line in CAD
This is the detail that separates a joined part that looks printed-on-purpose from one that looks repaired. When you split a model into two halves, the surfaces that meet are not the surfaces you designed — they are the surfaces plus a glue line plus whatever surface imperfections the printer added.
If the two halves are simply cut at the same plane, they will either interfere (printing two halves that are each exactly at nominal and then trying to clamp them leaves an invisible-but-real gap at the ends) or sit visibly apart. The fix is a deliberate allowance, in the same spirit as any other dimensional tolerance decision.
- Deduct a glue gap from each face: subtract 0.05 mm from each side for CA, 0.1 mm for epoxy.
- Add vertical alignment features rather than relying on hands: dowel pins, tongue and groove, or a stepped lap. Alignment under clamping pressure is impossible to hold by feel.
- Chamfer the bond face's outer edge by 0.3 to 0.5 mm. It sounds trivial. It creates a reservoir that holds the squeeze-out instead of letting it run down a visible face, and it removes the sharp rim that tends to chip.
- Test the fit dry first. Print a short section of the joint only — 20 mm of the 300 mm seam — before committing to the full print. Test coupons are covered in depth in the fit and tolerance workflow.
- Keep the joint off the visible surface where possible. If the seam can hide on an inside corner or be covered by a later feature, it should. The sanding, priming and painting sequence can hide a lot, but hiding a bad seam is harder than avoiding one.
When Not to Glue
Not every large part wants to be joined. It is worth knowing the cases where the honest answer is "do not do this".
- The load is bending across the seam. A glued butt joint in bending is a peel joint with extra steps. Either convert it to a rabbet with pins, or reorient so the load is in shear.
- The part will be hot in service. Most adhesives soften far below the temperatures where engineering plastics still work. A glued joint in an engine bay or near a heat source is creep waiting to happen.
- The part needs to be watertight under pressure. A glue line is a seam, and seams leak. If you need a pressure-tight join, print it in one piece or use a mechanical seal — see the discussion of capillary leaks in printed parts.
- The joint has to survive repeated cycles of disassembly. Use threaded inserts and bolts instead. Adhesive joints and serviceable joints are different design problems.
- You are joining two different materials with wildly different expansion rates. The differential will eventually crack the bond. Design in a compliant element — a flexible adhesive, a grommet, or a gap.
When a join is genuinely required, the goal is to make it an assembly rather than a repair. That means designing for it from the start, which in turn means the part design decisions come before the slice, not after.
A Working Sequence for Large Parts
Put together, the workflow for a part that does not fit the build plate looks like this.
- Plan the split before slicing. Decide the cut plane so it falls on the least visible, least loaded location — and so each piece still prints without either half needing extreme support.
- Add the joint features in CAD. Rabbet or tongue and groove, plus dowel holes at 40 to 60 mm spacing, plus a 0.3 mm outer chamfer on the bond faces.
- Print a short test of the joint region only. Check fit and clearance before committing 12 hours of machine time.
- Adjust the CAD from the test. This is the step that reliably saves the project.
- Print both pieces. Keep the orientation consistent between them so the layer lines at the seam run the same direction; orientation sets the strength of the seam as much as the adhesive does.
- Dry-fit and mark. Tape the parts together, confirm the alignment, and mark any spot where the faces interfere.
- Prepare both faces. Scuff, clean, degrease, and for hard-to-bond plastics, flame treat — then go straight into bonding.
- Bond with a measured glue line and light clamp. Alignment features hold position; pressure only needs to close the gap.
- Let it cure for the full rated time. Most adhesives reach handling strength long before they reach structural strength. A joint loaded at one hour is not the joint you tested.
- Dress the seam. Once cured, sand flush and finish if the joint is visible.
For big multi-part builds it is also worth reconsidering the nozzle before committing to a joint. Larger nozzles and high-flow hotends reduce the number of pieces needed in the first place — see the discussion of nozzle sizing and the H2 high-flow nozzle for parts that would otherwise have to be split. Where the split is unavoidable, keeping the filament dry across a long print matters more than usual: a moisture-weakened surface bonds poorly no matter what adhesive you choose.
What Holds and What Comes Apart
The single idea worth carrying away is that gluing printed parts is a design activity, not a consumable choice. The adhesive is the last 10 percent. The other 90 percent is the joint's area, the direction of the load relative to the layer lines, and the surface treatment you applied twenty minutes earlier.
Get those right and a two-part epoxy or a cheap CA gel will hold a functional part together for years. Get them wrong and no adhesive on the market will save the seam. The printer gives you a surface; the joint gives you the strength. For parts that must survive outdoors after bonding, a fibre-reinforced ASA-CF shell pairs with the same joint rules and simply adds UV and heat resistance to the seam.