Designing for heat-set inserts
A printed thread survives a handful of assemblies and then starts to round off: the plastic ridges crumble layer by layer, and the screw spins without biting. If one of your parts is going to open and close dozens of times — a PCB cover, a motor mount, a battery compartment — the printed thread is the link that fails first. The fix isn't to print a better thread: it's to not print one at all. You heat-set a threaded brass bushing into the plastic and let the metal be the thread. The rest of this article is about the one piece you actually design: the boss that receives it.
The brass melts the plastic, and the plastic grips as it cools
A heat-set insert is a brass bushing with the metric thread already machined inside and a knurled outer surface: diagonal or crosshatched grooves. You seat it with the tip of a soldering iron, and the temperature that matters is the tip's, not the plastic's print temperature. For PLA a tip at 200–230 °C is enough; for PETG, a little more, around 240–250 °C. A soldering iron at that temperature dumps far less heat into the mass of the brass than an extruder does, so contact time matters as much as the temperature itself: you want to soften the wall of the hole without scorching or boiling the plastic.
The hot brass softens the wall of the hole; you push gently and the molten plastic flows into the grooves of the knurl. As it cools, that plastic solidifies around every groove and locks the insert in the two directions that matter: it won't pull out under the screw (axial pull-out) and it won't spin when you tighten it (torque resistance).
The advantage is physical. A printed thread puts the whole load on a few plastic ridges built up out of layers, weakest right at the inter-layer plane — the part's flimsiest seam. An insert puts the load-bearing thread in brass and spreads the pull-out force along a whole knurled cylinder that grips the plastic over its entire surface, not across a few fragile teeth. You can drive in a steel screw and torque it down fully without stripping the thread. For M2 through M5 — exactly the sizes where printed threads strip most easily — it's the most reliable threaded joint you can design.
Size the hole a little smaller than the insert
The pilot hole isn't drilled to the insert's outer diameter — it's a bit narrower. Match it to the outside of the knurl and the insert drops in loose: the molten plastic never fills the grooves, and you end up with a bushing that wobbles and pulls out at the first tug. Make it too narrow and the brass has to displace so much material that the wall splits and cracks. The balance you want is a knurl that bites enough plastic to fill its grooves without overloading the wall.
The number that actually governs is the knurl's minor diameter, not its outer one: the pilot hole sits close to that minor diameter, so the crests of the knurl drive into solid plastic. That's why the pilot ends up well below the insert's outer diameter, not just a few tenths under. A hole sized close to the outer diameter gives you exactly the loose bushing you were just warned to avoid.
| Thread | Insert OD (approx) | Pilot hole Ø |
|---|---|---|
| M3 | ~4.6 mm | ~4.0 mm |
| M4 | ~5.6 mm | ~5.3 mm |
| M5 | ~6.4 mm | ~6.4 mm |
Treat the table as a starting point, not gospel: the real outer diameter and the recommended hole vary from one maker to another, and the datasheet for the insert you actually bought overrides any generic table. Print a test boss, melt an insert in, and confirm it grips without splitting the wall before you commit to a whole panel of bosses.
Make the boss wall thick, or it splits
The boss is the plastic cylinder you model around the hole, and its wall thickness is what separates an insert that holds forever from one that cracks the part as it goes in. The most common failure by far is a thin wall splitting — either as the insert goes in, when the hot brass expands the plastic and pushes it outward, or later, when you tighten the screw and the metal thread pulls radially on the wall.
The reason is hoop stress, the same stress that shows up in any press fit: the insert pushes the wall outward and the wall resists by working in tension all the way around. A thin wall has no cross-section to spread that tension over, so it gives way at its weakest point, which on a part printed upright is the perimeter seam — it unzips top to bottom along a layer line. A comfortable rule of thumb at these sizes is boss outer diameter ≈ 2× the insert outer diameter, that is, a wall thickness of at least one insert radius on each side. Plenty of manufacturer sheets get by with less — on the order of a wall thickness equal to the insert OD — but doubling it leaves margin for your printer's variation without your having to think about it.
And that thickness has to be real wall, not infill. Sparse infill barely resists hoop tension: it's a mesh full of gaps, not a continuous ring. What resists is the perimeter, the bead that goes all the way around the hole without a break. Design the boss so the wall is almost all perimeter — with a 0.4 mm nozzle, three or four perimeters are ~1.2–1.8 mm of continuous wall wrapping the hole — and add perimeters in that zone rather than infill, which counts for next to nothing here. Walls, perimeters and infill covers this in detail.
Chamfer the mouth and tune the depth
Add a small lead-in chamfer at the top of the hole, a 45° cone roughly as deep as the wall is thick. It does three things: it guides the insert as it enters, it catches the first displaced plastic instead of letting it overflow the edge, and it leaves a clean, flush face once the insert finishes seating. Bear in mind that a chamfer printed upright comes out stepped layer by layer; a smooth cone modeled in from the start guides cleanly.
Make the hole a little deeper than the insert is long, so it can seat flush or just below the surface. That small pocket under the insert isn't cosmetic: in a blind hole the molten plastic and the air the brass pushes ahead of it have to go somewhere; with nowhere to escape, they push back up and the insert sits proud or is forced out as it cools. A hole that's too short leaves the insert sticking up and fouling whatever bolts on top; too deep is no problem — the insert just ends up slightly recessed.
Too much heat is its own failure
Cracking isn't the only failure mode, and heat fails in two directions. Overdo it — tip too hot or insertion too slow — and you melt too much: the insert sinks in crooked, molten plastic overflows and bubbles at the edge, and the boss ends up domed — a bulged top face instead of a flat one. Come up short — tip too cool or insertion too fast — and the plastic has no time to flow into the grooves: the insert sits proud, the knurl is half-filled, and the bushing is loose, just like a hole that's too wide.
Between the two extremes there's a comfortable window. Start at the cool end of the temperature range: if the insert won't go in smoothly, raise it a few degrees before you push harder. When you see plastic overflowing or bubbles at the mouth, you've gone too hot or too long; when the insert sits proud and spins with the screw, you've come up short.
Melt the insert perpendicular to a flat face
Always orient the insert hole perpendicular to a flat face of the part, never against a side wall. The reason is structural and comes from how the layers stack. When the insert goes in perpendicular to a flat face printed against the bed, its radial push spreads the hoop stress within the layer planes, which is the favorable orientation: the horizontal plane carries the tangential tension without pulling apart.
If instead you melt the insert against a side wall — its axis parallel to the layers — the brass's radial push pulls in exactly the direction that separates one layer from the next. That's the weak direction of any FDM part, where the material retains only 40–80 % of its strength, and the insert loads it in direct tension: the wall delaminates, it splits between layers like the pages of a book, and the insert loosens or pulls out. If the geometry forces an insert on a side, reorient the part on the bed so that face prints flat, even if you have to add some support elsewhere. Why orientation decides which way the weakness points is covered in Walls, perimeters and infill and in Layer adhesion and anisotropy.
With the boss sized right, heat-set inserts handle any threaded joint meant to be opened and closed many times. For large threads — M6 and up, wheels, caps — it often pays to print the thread directly — see Modeling threads for the details. And when you'd rather the screw enter from behind the part than thread into it, the answer is a captive nut: that's covered in Captive nuts and clearance holes.