The threads that holdare not printed.

Where a printed thread gives up

A printed fixture almost always fails at an interface: a bonded joint, an unsupported wall, or a thread. The first two get caught in design review, because you can see them in the model. However, printed threads pass review, pass the first bench check, and then strip in production, on the shift when an operator torques a mount plate down one more time than the plastic could take. The fix is not more plastic, but rather metal, placed in the model before the file is ever sliced.

The mechanism explains where the metal has to go. A printed thread is a helical feature built from extruded perimeters, so every crest and root crosses the layer boundaries that carry the weakest bond in the part. Torque shears those small plastic features, while pull-out separates the layers along the Z axis, the weak direction in any fused filament fabrication part for three reasons at once: polymer bonding between layers is incomplete, layer lines act as crack initiators, and sparse infill leaves no solid material behind the boss. Fiber placement carries real load when it is aimed correctly, as we lay out in what actually makes a composite part strong, but no layup rescues a thread form cut into plastic.

What a heat-set insert actually does

A heat-set insert is a threaded brass component melted into the plastic, and the operative word is melted. As the iron brings the surrounding material above its melt point, polymer flows into the knurling on the insert body and freezes there, producing a mechanical interlock over the insert’s full length and diameter. The load path through the joint changes completely: the bolt loads brass, the brass loads the bulk of the part through that interlock, and no plastic thread form is asked to do both jobs at once.

This is why inserts resist pull-out and torque far better than printed threads, and why they keep resisting through repeated assembly and disassembly. In production tooling that distinction decides the service life, because a fixture gets opened, shimmed, and re-clamped for years, and every one of those cycles takes a little more material out of a printed thread, which makes it a consumable in that environment. An insert is hardware.

The cavity is a datasheet dimension

The design work happens in CAD, not at the bench, and it starts with the manufacturer’s datasheet. Model the cavity using the exact dimensions published for the insert you will actually install: the maximum hole diameter and the insert length, taken from that part number rather than a similar one in the same drawer. Those two numbers are the specification, and no care at the bench substitutes for getting them right. Everything after them is technique.

Both directions of error leave a signature that shows up at first assembly. A cavity that is too large leaves too little material to melt and flow into the knurling, so the interlock is thin and the insert may spin under torque. A cavity that is too small displaces more material than the boss can absorb, which can bulge the wall and leave the insert proud of a face meant to be flat. Both are free to fix in the model. Neither is free to fix in a part that has already come off the printer.

Install from the back side of the load

One orientation rule does more for pull-out resistance than anything else, and it costs nothing at the model stage: design so the insert is installed from the back side of the load direction. Arranged that way, a pull-out load presses the insert deeper into the part and loads the surrounding material in compression, rather than drawing the brass back out. The bolt still enters from the working face, and the assembly sequence does not change. What changes is which face the brass goes into, and what material stands between it and the pull.

In practice, the load case has to be settled before the boss is placed. Where the pull comes from, which face the fastener enters, and how the fixture mounts together decide the orientation of every insert, which is why we ask for the load case along with the model. Reversing an insert after the fact is not an adjustment that anyone makes at the bench. It is a reprint.

An insert installed from the wrong face still torques to spec on the bench. It still pulls out on the line.

The orientation rule costs nothing in CAD

A temperature rule, and three ways to ruin it

The installation itself is simple enough that the real risk is complacency rather than difficulty. Set the soldering iron 50 to 100 degrees C hotter than the material’s nozzle printing temperature, which puts an Onyx part at roughly 390 degrees C, and keep the tip aligned and perpendicular to the insert axis. Let the heat do the work, and stop when the top of the insert is flush. A cheap $8 to $10 iron with a standard conical tip is sufficient: this is a temperature problem, not a tooling problem.

Bolts, captive nuts, and machine keys

An insert solves a thread. It does not solve a part being pulled apart along its layers, and that is where two more pieces of metal earn their place. Bolts that run completely through a part into captive nuts add the tensile strength of the bolt to resist Z-axis separation forces, and they compress the part together so the layer interfaces stay closed. The clamp load does what the layer bond cannot.

Metal machine keys address the opposite problem, which is shear arriving in one place rather than tension pulling the layers apart. A key running through the part distributes that localized shear across the entirety of the Z axis, and it presents a strong, wear-resistant flat surface where a printed face would wear. Both decisions come out of the same exercise: visualize how the part would fail or deform under its expected loads, then place the bolts, keys, and splits where that mental model says it wants to come apart.

Design the joint the hardware runs through

Hardware and splitting are the same conversation, because the reason to split a part into co-parts is to print each section in its strongest orientation. Once a part is in two pieces, the joint is a designed feature rather than a byproduct of the split. Add chamfers to the mating edges (8 degrees is a workable starting point) so the co-parts guide themselves together instead of binding on a corner.

How you plan to join the co-parts decides the fit you should model in the first place. An arbor press and a tight press fit need no extra room; liquid adhesives such as cyanoacrylate, epoxies, and silicone do, because the bond line occupies space the model has to give it. Use the Move Face tool in CAD to add that clearance between mating faces before export, rather than sanding it in at the bench.

Metal where the metal belongs

Overall, the three techniques share one logic: put metal where the load concentrates, and let the composite do what it is genuinely good at, which is holding accurate geometry at low weight. Printed threads are not a material failure. They are a design decision that can be made differently, for a few dollars of brass, at the stage where changing it is still free.

The parts worth reviewing first are the ones that come apart on purpose: quick-change mount plates, end-of-arm tooling, and any fixture an operator handles every shift. As printed tooling takes on more of the work that used to belong to machined aluminum, the fixtures that hold will be the ones where the metal was designed in from the datasheet, not added at the bench after something stripped.