The enclosure came apart beautifully the first time. The second time one screw felt a little looser. By the tenth opening the hole had widened into a smooth funnel, the thread had nothing left to grip, and the whole box was being held together by friction and hope. Nothing was overloaded, nothing was dropped, and the screw was the same screw. The plastic simply gave way, slowly, exactly where a small hard steel thread was pressing into it.
Anyone who designs multi-part printed assemblies runs into this, and the usual advice is unsatisfying: use heat-set inserts. That advice is correct and incomplete. An insert is not a drop-in fix for a badly designed boss. Put a 4 mm brass insert into a 5 mm wall of PLA and it will split on installation or pull out under load, and you will blame the insert instead of the geometry that surrounded it.
This article is about the joint, not the hardware. We will look at why a thread in plastic is a fundamentally weaker thing than a thread in metal, compare the four ways people actually put a reusable thread into a printed part, and then work through the dimensions that decide whether the joint survives fifty openings or five.
Why a thread in plastic is always a compromise
A machine screw is a wedge wrapped around a cylinder. Tightening it converts torque into axial tension, and the flanks of the thread push outward against whatever they are engaged with. In metal, that outward pressure is resisted by a material that yields only at hundreds of megapascals. In a thermoplastic it is resisted by a material that yields at thirty to sixty, and that is also viscoelastic, meaning it keeps moving under a constant load long after the load was applied.
Two consequences follow, and both are mechanical rather than chemical. First, the engaged thread area is tiny. A short M3 engagement may have less than 30 mm² of load-bearing flank, and the stress on that area is what decides whether the thread strips. Second, the plastic creeps. Every time you tighten a screw you apply a preload, and every preload relaxes over days, which is why a joint that felt solid in October is loose in January. The relaxation mechanisms are the same ones described in creep, fatigue and preload, and the practical fix is the same: keep the preload low, keep the joint short, and give the plastic somewhere to be compressed rather than sheared.
This is why the honest summary of every method below is the same. You are not trying to make plastic behave like metal. You are trying to arrange the joint so that the plastic is loaded in compression over a large area instead of in shear over a small one.
The four ways to put a reusable thread into a printed part
Every method in common use is one of four approaches, and they differ enormously in effort, strength and how many times the joint can be opened.
| Method | What holds the thread | Openings before it degrades | Best use |
|---|---|---|---|
| Drive a screw straight into printed plastic | One or two moulded-in thread forms cut once | 1 to 3 | One-time assembly, disposable fixtures, prototypes |
| Thread-forming screw (PT / Plastite type) | Purpose-shaped thread that cuts and then locks | 5 to 10 | Production enclosures with consistent boss geometry |
| Heat-set brass insert | Metal thread, plastic melted around it and re-solidified | 25 to 50+ | Anything the user will open, service or upgrade |
| Printed thread | Plastic thread printed as geometry | Depends entirely on size and clearance | Large threads (M12 and up), caps, adjusting mechanisms |
The first row is where most people start and where most disappointment comes from. Driving a standard machine screw into a printed hole works because the screw cuts its own path, but the material it cuts is not removed, it is displaced. That displaced plastic sits under tension around the hole, and the next time you drive the same screw it cuts a slightly different path through already-damaged material. Two or three cycles and the hole is a smooth bore.
The second row is the underrated option. Thread-forming screws for plastics have a coarser pitch, a wider flank angle and a sharp cutting edge, and they are designed to be driven into a moulded boss once and stay. They are cheaper than inserts, need no heat, and are perfectly adequate for a housing that gets opened twice in its life. They are not adequate for a battery door.
Heat-set inserts: what actually happens when you press one in
A heat-set insert is a brass or stainless bush with a knurled or helically grooved outside and a machine thread inside. You press it into an undersized hole with a soldering iron, a purpose-made tip, or an ultrasonic setter. The insert conducts heat into the surrounding plastic, the plastic softens, the insert sinks to depth, and when the heat stops the plastic re-solidifies around the knurls and grooves.
The result is not a friction fit. It is a mechanical interlock, and the strength comes from two separate things: the plastic that has flowed into the grooves resisting pull-out, and the plastic that surrounds the insert resisting the boss splitting outward. That second one is why boss wall thickness matters more than insert depth once you are past a certain point, and it is the dimension people get wrong most often.
The boss dimensions that decide everything
These numbers are for standard brass inserts with an outer diameter of 3.5 to 5 mm, which covers M3 and M4. Scale the wall relative to the insert diameter rather than memorising millimetres.
| Dimension | Rule of thumb | What happens if you break it |
|---|---|---|
| Hole diameter | Insert outer diameter minus 0.15 to 0.25 mm | Too small: boss splits during installation. Too large: no melt contact, insert pulls out cold |
| Hole depth | Insert length plus 1 to 2 mm of clearance | Bottomed out insert cannot sink to depth; trapped air pushes molten plastic back up the hole |
| Boss wall thickness | 1.0 to 1.5 times insert diameter, minimum 2 mm | Thin wall splits along a layer line under installation pressure |
| Engagement under the insert | At least 1 insert diameter of solid material below | Insert pushes through the floor of a thin shell |
| Boss height above the surface | Insert length plus 0.5 mm, or flush with a countersink | Insert sits proud and the mating part does not seat |
Notice that the wall thickness rule and the engagement rule both add material in directions that a shelled part does not naturally have. A 1.6 mm wall with a boss stuck on it will split; the boss needs to be a solid column with its own walls, and it needs to be connected to the shell with fillets rather than meeting it at a right angle. This is the same set of rules that governs where plastic should sit for stiffness, described in load paths rather than shapes, applied at a much smaller scale.
Installation technique is half the result
The temperature question is the one that ruins the most parts. Too cold and you force the insert in mechanically, which splits the boss or shears the plastic instead of melting it. Too hot and the plastic becomes so fluid that it flows away from the knurls entirely, leaving a loose insert in an oversized hole. The right temperature is the lowest one at which the insert sinks with light pressure, typically 20 to 40 °C above the material's glass transition or, for semi-crystalline materials, just above melting.
Three practical rules follow. Set the iron temperature, do not guess it by feel; a variable-temperature iron with a flat insert tip is the whole tool kit. Press straight and let the insert find its own depth, because side load at an angle opens one side of the boss first. And give the part time to cool before you drive a screw into it, because the insert is only locked once the plastic around it is solid again.
One more thing that is rarely mentioned: an insert installed into a dry part behaves differently from one installed into a damp part. Moisture in the filament becomes steam at insert temperatures, and steam under an insert produces the same voids and reduced bond that wet filament produces everywhere else. If the part has been sitting in a humid workshop, dry it first, exactly as described in moisture by material.
Printed threads: when they work and when they are a trap
Printed threads get a bad reputation, and it is mostly deserved at small sizes. The reason is arithmetic. A printed M6 thread has a thread height of about 0.5 mm, which at a 0.2 mm layer height is two and a half layers. The thread does not have a smooth flank, it has a staircase, and the load is carried on the corners of that staircase. Add a clearance that has to absorb print tolerance, and the engaged area drops further.
At larger sizes the arithmetic changes. An M16 thread has a thread height of about 1 mm, five layers, and a much longer helix, so the load is spread over far more material. Printed threads also become genuinely useful wherever the thread is doing a job other than clamping: a screw-adjust mechanism, a cap, a levelling foot, a large jar lid. Those applications load the thread in slow compression and rarely need to hold a preload for years.
If you print a thread, print it in the orientation that puts the load along the layers rather than across them, use a trapezoidal or buttress profile instead of a sharp V, and give it more clearance than you think. The same orientation logic that governs pivots and pins applies here, and it is set out in pins, pivots and bushings.
Resin threads are a different animal
Stereolithography prints threads far better than FDM, because there are no layers to staircase and the resolution in the build direction is tens of microns rather than tenths of a millimetre. Fine M3 threads are genuinely practical in resin. The limitation moves from geometry to material: standard resins are brittle, so a thin-walled resin boss will crack rather than yield. A tough engineering resin is the right choice for anything that gets tightened, such as an ABS-like resin, and it is worth reading the notes on what attacks a printed part before putting a resin thread near a solvent.
Pull-out, strip-out and torque: the three failure modes
A threaded joint in plastic fails in one of three ways, and they need different fixes. Confusing them is why people add material where it does not help.
Pull-out is the fastener being extracted axially, thread and all. It happens when the engaged material shears in a cylinder around the thread. The fix is more engagement depth, up to about two times the fastener diameter, and more material around that cylinder. Beyond two diameters the extra depth buys very little, because the load is no longer distributed evenly down the thread.
Strip-out is the thread itself being torn out while the fastener stays in place. It happens when the flanks are overloaded, which is a function of the material's shear strength and the engaged area. The fix is a larger diameter or a metal insert, not more depth.
Boss splitting is the wall around the hole failing outward. It is the most common failure with inserts and the least discussed. It happens because the insert is an interference fit in a hoop that has a weak seam, and FDM parts have a perfect seam exactly there: the vertical layer boundaries run radially outward from the hole. This is why a boss printed with its axis vertical is far more split-prone than one printed lying down, and why the same boss in resin, with no layers at all, behaves completely differently.
The practical torque limit for a plastic thread is low enough to surprise people. An M3 machine screw in a heat-set insert can typically take around 0.5 to 0.8 N·m before the plastic begins to yield, against 2 to 3 N·m for the same screw in an aluminium thread. If you are designing something that a user will tighten by hand with a long handle, you are designing for a metal thread or a much larger fastener. The relationship between hand force and torque is the subject of grip, knurl and handwheel, and it explains why a small knob with a long lever can destroy a joint that felt fine with fingers.
Which material for the boss
The material choice matters less than the geometry, but it is not irrelevant. What you are looking for is a combination of shear strength, creep resistance and toughness, and these do not always come together.
| Material | Insert behaviour | Watch out for |
|---|---|---|
| PLA | Sets beautifully, holds well cold | Creeps and softens above about 50 °C; a warm car undoes the preload |
| PETG | Forgiving to install, good toughness | Stringy and gummy when hot; insert can shift while cooling |
| ABS / ASA | Excellent insert results, holds preload | Needs an enclosure; warping moves the boss |
| PETG-CF / ABS-CF | Stiffer boss, less creep | Abrasive; use a hardened nozzle such as this tungsten one |
| PA6-CF20 | Best all-round for load and heat | Must be bone dry; hygroscopic and dimensionally fussy |
| PET-CF17 / PPS-CF10 | Holds torque at high temperature | Expensive; high print temperature; needs a hot, rigid machine |
| TPU | Not a thread material | Use a metal insert or a through-bolt with a washer instead |
The pattern is that glassy amorphous materials set a clean insert but do not hold preload when warm, while semi-crystalline and fibre-filled materials hold preload but are fussier to install. If the joint will live at room temperature and be opened occasionally, PLA is genuinely fine. If it will see heat, sun or a long service life, the filled engineering materials pay for themselves.
Designing the joint, not just the hole
Everything above concerns a single fastener. A joint is a system, and the system-level decisions decide whether the fastener ever sees the load you designed it for.
Use the screw to clamp, not to locate
The single most effective change you can make to a multi-part printed assembly is to stop using screws as alignment features. Give the mating parts a shoulder, a tongue, a chamfer or a pair of pins that establish position, and let the screws do nothing but pull the faces together. A screw in a clearance hole with 0.4 mm of slop cannot hold two parts in alignment, and the assembly will shift every time it is opened. This is the same conclusion reached from a different direction in tolerance stack-up.
Spread the clamp load over a flange
A screw pulling directly into a boss in a thin shell puts a point load on a flexible plate, and the plate bends away from the screw instead of compressing. The result is a joint that feels tight and is not. A local flange around the boss, two to three wall thicknesses wide, gives the clamp load somewhere to go. Where the parts need to seal as well as clamp, the flange has to be wide enough for a gasket groove, which is covered in sealing printed parts.
Prefer compression over bending
If a screw has to hold two parts at an angle, or pull against a long flexible arm, the plastic under the head sees a peeling load rather than a compressive one, and peeling is what opens layer lines. A small compressive pad, a spherical washer face or simply a thicker local wall converts the load back into compression. Printed washers or captive spacers are a cheap way to get there, and the same principle behind wall count and shell placement applies at this scale.
Plan for the number of openings
Decide early how many times the joint will be opened, and let that number choose the method. A housing opened once for assembly can use a self-tapping screw and never think about it again. A battery door opened weekly needs an insert, and possibly a captive screw so the fastener does not get lost. A lid opened several times a day needs something that does not depend on thread friction at all, which is where clips, quarter-turn features and the compliant elements in compliant mechanisms earn their place.
When to stop threading plastic entirely
There is a category of joint where no amount of boss design produces something reliable, and recognising it early saves a lot of iteration.
- The joint carries vibration. Threaded plastic under vibration loses preload fast. Use a metal insert plus a thread-locking element, or move to a through-bolt with a nut.
- The part is small. Below about M2.5 there is not enough material for a reliable boss. Use a metal insert, a clip, or design the parts to snap together.
- The service temperature approaches the material's heat deflection point. Preload relaxes and the joint self-loosens. Go to a high-temperature engineering material or change the joint type.
- The assembly needs to come apart blind, by hand, without tools. Threads are the wrong answer. Use a clip, a cam or a quarter-turn.
- The joint has to carry a sealing load and be opened repeatedly. The combination is hard on plastic threads; a metal insert with a captured O-ring is usually the only stable answer.
For joints that have to transmit motion rather than just hold two parts together, threads are often the wrong mechanism altogether, and the alternatives in gears you can actually print or the sliding surfaces in bearings you can print are better starting points.
A practical workflow
If you are building a multi-part assembly this week, this is the order that avoids most of the rework.
- Decide the opening count before drawing anything, and pick the method from the table above.
- Pick the insert by outer diameter, not by thread size. The outer diameter drives the boss, and the boss drives the layout.
- Size the boss from the insert diameter using the table: wall 1.0 to 1.5 times the diameter, floor engagement at least one diameter.
- Print a test boss, ideally four of them in the same orientation as the real part. Install an insert in each and try to split the first three deliberately.
- Measure the result rather than eyeballing it, using the approach in inspecting printed parts. Insert depth and hole diameter are the two numbers that predict success.
- Test the torque limit on a duplicate joint. Tighten until something yields, and then design your assembly instructions to use half that value.
- Add the alignment feature and the flange, then re-check that the mating faces actually touch.
- Dry the filament before the real print, in a dryer cabinet or a drying box. Wet filament and inserts do not get along.
- Record the settings that worked: insert size, hole size, iron temperature, material and orientation. The next part will use them.
None of this is difficult, and all of it is easy to skip. The difference between a printed assembly that opens fifty times and one that strips on the third attempt is rarely the insert you bought. It is whether the boss around it was sized for a metal thread in plastic rather than a metal thread in metal.