Every functional 3D printed part eventually fails. The interesting question is how. Break a tensile bar in a testing jig and the answer is boring: the layers let go. But watch a real assembly for a year — a hinge, a clamp, a bolted bracket, a snap-fit lid — and a completely different set of failure modes shows up. The part does not break. It slowly changes shape, or it develops a hairline crack that grows one print at a time, or the screw that held it tight last spring is now loose in its hole.
These are the failures that surprise people, because they happen far below the load the part was designed for. This guide is about the three mechanisms behind them — creep, fatigue and preload loss — and the design habits that make an interface survive time, not just a single pull test.
Why Long-Term Strength Is a Different Property
When people talk about a filament being strong, they usually mean one of two short-term numbers: tensile strength at break, or layer adhesion in a single direction. Both are measured in seconds to minutes. Neither tells you what happens under a load that never goes away.
FDM parts sit at a particular disadvantage here, and the reason is structural rather than chemical. A test bar is printed as a continuous extrusion, so its strength is dominated by how well adjacent roads fused. A finished assembly bolts, presses, or clips that same material against another material — metal screws, aluminium inserts, or a mating printed face. That interface concentrates stress in a way the test bar never experiences.
Three things then decide whether the interface holds:
- Creep — constant load, gradually accumulating deformation. The material flows away from where you put it.
- Fatigue — repeated load, accumulating micro-damage. The part cracks without ever being overloaded once.
- Preload loss — the clamping force that makes a joint work quietly decays, and a loose joint fails at a fraction of the load a tight one handles.
All three get worse with temperature, and all three get dramatically worse when the stress is applied across layer lines rather than along them.
Creep: The Load That Never Goes Away
Creep is the tendency of a polymer to deform continuously under a load that is held constant. It is not a defect. Every thermoplastic does it, and at room temperature it is slow enough that most hobby parts never reveal it. Raise the temperature 20 °C, or double the sustained stress, and the rate can jump by an order of magnitude.
The practical shape of creep is a curve with three regions. First comes a short burst of rapid extension as the material rearranges. Then a long plateau where it creeps steadily but slowly. Then, past a threshold and given enough time, the third stage arrives — accelerating flow that ends in rupture. The trap is that in a real part, stage one and stage two look identical for months. Nothing appears to be wrong.
Where creep shows up in printed assemblies
The classic cases are all cases of sustained stress in a warm place:
- A clamp or a spring clip that is held open. The deflection is constant, so the force relaxes rather than the shape changing — the clamp grips less every week.
- A shelf bracket or a wall hook with a static load. Over a summer the arm sags a few degrees, which is invisible until a long straight edge is laid against it.
- A threaded boss under a tightened screw. The material under the head flows, the screw loses tension, and the joint that felt solid in March rattles in November.
- A spacer or a shim in a stack of bolted parts. It thins, and every fastener in the stack loosens with it.
Note that these are all static loads. Creep is not about impact or vibration. It is about a part being asked to hold a position for a very long time.
The material ranking
Creep resistance tracks the glass transition temperature of the polymer, which is why the engineering grades pull so far ahead of the commodity ones. Amorphous polymers such as ABS and ASA hold their shape better than PLA because PLA’s transition sits close to typical indoor temperatures, so a PLA part on a sunlit shelf is already in the regime where it flows. Semi-crystalline materials behave differently: they resist creep well below their melting point because the crystalline regions act as physical crosslinks, but they soften quickly as they approach it.
| Material | Relative creep resistance at room temp | Behaviour under a warm sustained load | Reasonable use |
|---|---|---|---|
| PLA | Low | Sags noticeably above ~50 °C; can flow under load at indoor summer temperatures on a sunny surface | Display pieces, short-lived jigs, prototypes |
| PETG | Moderate | Better than PLA, still softens and deforms well below its print temperature | Containers, brackets with light static loads |
| ABS / ASA | Good | Holds dimension reliably in ambient and mild engine-bay-adjacent conditions | Structural brackets, outdoor hardware, enclosures |
| Carbon-fibre filled ABS, ASA or PETG | Very good | Fibres add stiffness, so the same load produces less initial strain and less creep to accumulate | Loaded brackets, mounts, machine parts |
| TPU | Depends on hardness | Elastic but not immune — a permanently compressed TPU gasket takes a set and stops sealing | Dampers, feet, seals with a defined service life |
The carbon-fibre point deserves emphasis. Filled grades do not just add stiffness; by stiffening the part they reduce the strain the polymer has to accommodate, and creep deformation scales with that strain. A carbon fibre ABS filament bracket that deflects half as much under load will also creep at roughly half the rate.
Design rules that cut creep
- Reduce sustained stress, not just peak stress. A rib that halves the bending moment halves the creep rate. This is the single highest-leverage change.
- Keep the part cool. Anything above about 40 °C needs a harder look at material choice. Moving a bracket out of direct sun is a real engineering fix.
- Never rely on a printed spring to hold force indefinitely. If a spring is needed, use a flexible filament with a defined hardness and design the deflection so that a 20 % loss of force still does the job.
- Support the load along the layers where you can. Reorienting a bracket so the stress runs in-plane removes an entire failure path.
Fatigue: Why a Part Cracked Without Being Overloaded
Fatigue is different from creep in one crucial way: the load is repeated. Each cycle does a tiny, mostly invisible amount of damage near a stress concentration, and over thousands of cycles those tiny damages link up into a crack that grows until the part separates. The stress needed to cause fatigue failure can be a small fraction of the stress needed to break the part once.
FDM parts are unusually prone to fatigue because of where the stress concentrations sit. Every layer boundary, every seam where a perimeter starts and stops, and every internal void left by an infill pattern is a place where stress cannot flow smoothly. Under a static load these features are mostly harmless. Under a cyclic load they are focus points, and the crack usually starts at whichever one sits closest to the surface.
The unhelpful role of infill
Low infill is a fatigue problem, not a strength problem. A 15 % grid inside a part looks solid from the outside, but the walls carry almost all the load, and the infill joints are repeated stress risers sitting in the load path. A fatigue crack that begins in the wall has an easy route inward along the infill boundaries. For anything that will flex repeatedly, higher wall count and a denser or continuous infill are worth more than extra external thickness.
What counts as cyclic in real use
Designers tend to picture fatigue as a machine-vibration problem. In practice, the most damaging cycles are ordinary ones:
| Application | Cycle source | Typical cycle count | What fails first |
|---|---|---|---|
| Living hinge | Opening and closing | 10,000–100,000+ | Crack at the hinge root, propagating across layers |
| Snap-fit latch | Assembly and disassembly | 10–1,000 | Root of the cantilever, or the layer line at its base |
| Robot arm or car mount | Acceleration and vibration | Millions (low amplitude) | Screw holes elongating, then wall cracking |
| Bicycle or scooter part | Road vibration and load reversals | Millions | Any 90° internal corner |
| Tool handle or grip | Manual use | Thousands | Bond between the grip and the core |
The living hinge deserves special mention because it is the one case where fatigue is not merely tolerated but expected. A printed hinge needs a material with real ductility, a thin section, the bend axis perpendicular to the extrusion direction so the fibres are not being pulled apart, and a single continuous wall rather than a stack of layers crossing the flex line. It is also a component with a finite service life, and treating it as permanent is the mistake.
Design rules that cut fatigue
- Radius every internal corner. A sharp internal corner is a stress concentration of roughly three times the nominal stress. A fillet costs nothing and removes most of it.
- More walls before more infill. Walls are continuous load paths; infill boundaries are interruptions.
- Print so the cyclic stress is in-plane. Bending a part so layers are being peeled apart is the fastest route to a fatigue crack.
- Use a tougher material for load reversals. ABS, ASA and carbon fibre reinforced PETG tolerate cyclic strain far better than a brittle high-stiffness grade. Stiffness and fatigue life are not the same property, and optimising only for stiffness can shorten life.
Preload Loss: The Quiet Failure of Every Bolted Joint
Bolted and screwed joints are the most common mechanical interface in printed designs, and they fail in a way that has nothing to do with the screw. A fastener works because it is stretched. The stretch is the clamp load that presses the two parts together and lets friction carry the load. If the printed material under the screw head or around the insert creeps even slightly, the stretch is lost, the clamp load drops, and the joint starts moving. Once it moves, the load shifts onto the screw in shear and onto the hole walls, and the joint fails quickly.
Heat-set inserts versus printed threads versus self-tapping screws
There is a real hierarchy here, and it is worth knowing what each option actually buys you.
- Heat-set threaded inserts are the strongest choice for anything that will be assembled more than once. A metal insert spread into a properly sized printed boss carries load far above what the surrounding plastic would allow, and it protects the threads from wear. The trade is that the boss needs adequate wall thickness and the insertion temperature has to match the material.
- Printed threads work well at large diameters, in coarse pitches, and for parts that are assembled once or twice. At small diameters they strip easily, and creep under load is a genuine concern. A stiff filled ABS or ASA-CF filament is a better host for a small thread than an unfilled grade.
- Self-tapping screws into a plain hole are fast and appropriate for enclosures that come apart rarely. They cut their own thread, so they lose grip after a handful of cycles.
For a full treatment of thread geometry, clearance and pitch selection, the earlier guide on designing functional threads and inserts covers the dimensioning side. What that article does not cover is what happens after six months of service, which is where preload enters.
Practical rules for joints that stay tight
- Give the boss thickness. The material around a thread or insert carries the load, so thin walls are the usual cause of stripped and loosened joints. Aim for a wall thick enough that the thread is fully engaged and surrounded.
- Use a washer or a flanged screw head. Spreading the clamp force over more area lowers the local stress and slows the creep that causes the loss.
- Take the load off the threads. Design a shoulder or a step so shear is carried by the printed geometry rather than the screw, and the fastener only has to hold the parts together.
- Re-torque once, deliberately. A joint that has been in service for a few weeks should be checked once. If it has lost tension, the boss design needs attention — not just another turn of the driver.
Where the joint also has to stay airtight, alignment and clamp uniformity matter as much as torque. The guide on stopping capillary leaks in printed parts covers the sealing side of the same problem.
Feed the Problem at the Source: Moisture and Temperature
Creep, fatigue and preload loss are all accelerated by two things a printer operator actually controls: how dry the filament is and how hot the part runs.
Moisture is a genuine contributor to long-term deformation, not just a cosmetic defect. Water absorbed into a filament flashes to steam in the hot end, and the resulting micro-voids are permanent features in the finished part. Each void is a stress riser, and a wall full of them fatigues sooner than a solid one. Drying before printing is therefore part of the durability strategy, not merely a fix for surface finish. A heated drying cabinet that also stores filament at low humidity removes the guesswork, which matters more for nylon and TPU than for PLA.
Temperature is the second lever. Because creep rate is strongly temperature dependent, a part that operates 30 °C cooler can have a dramatically longer service life at the same load. Three practical moves help: choose a material whose transition temperature is well above the service temperature rather than marginally above it, keep parts out of direct sun and away from warm enclosures, and print high-glass-transition grades such as ASA filament or flame retardant ABS where elevated temperatures are unavoidable.
Extrusion stability contributes too. Wear on a brass nozzle changes the extrusion width over its life, and a part printed slightly under-extruded has proportionally weaker walls everywhere. A tungsten carbide nozzle holds its bore far longer when running filled or abrasive materials, and stable extrusion means the wall thickness you designed is the wall thickness you get — on the first part and on the thousandth. Consumables are only part of the picture, but the foundation below everything else is a stable first layer, which is what a textured PEI build plate is for.
A Service-Life Checklist for Loaded Printed Parts
Run through this before committing a printed part to a permanent application.
- Identify the load type. Static, cyclic, or one-time assembly? Each points to a different failure mechanism and a different fix.
- Estimate the sustained stress and the service temperature. If the temperature is more than about 40 °C, assume creep is in play and derate accordingly.
- Check the orientation against the load. Stress should run in-plane, not peel layers apart.
- Radius every internal corner and taper every load transition. Stress concentrations drive fatigue.
- Budget more walls, not more infill, for anything that flexes repeatedly.
- Design joints for loss of preload. Use inserts where the part will be reopened, thicken the bosses, and let geometry carry shear rather than the threads.
- Dry the filament and verify extrusion. Voids and under-extrusion shorten life silently.
- Test one part to failure before printing twenty. Cycle it, load it, leave it in service for a week. A printed prototype costs a few grams of filament; a field failure costs the customer.
Common Misunderstandings
- “It is strong enough, I tested it.” A static pull test measures short-term strength. It says almost nothing about behaviour after a month under load.
- “Stiffer is always better.” Stiffness helps creep but can hurt fatigue life, because a brittle high-modulus grade reaches its strain limit sooner under cyclic bending.
- “The screw came loose, so the screw is bad.” Almost always the printed material crept, not the fastener. Replacing the screw does not fix the design.
- “PLA is fine, it is indoors.” Indoor is not cool. A dark PLA part in sunlight can reach temperatures where creep matters.
- “More infill density solves everything.” Density helps short-term strength far more than it helps fatigue, because the wall-to-infill boundary remains regardless of how densely the interior is filled.
The Takeaway
Short-term strength and long-term reliability are separate design problems, and the second one is decided by details that never appear in a filament spec sheet: whether the load is sustained or repeated, where the stress concentrations sit, how much material surrounds a screw, and how hot and how wet the part gets in service. A part that is 20 % less stiff but properly oriented, filleted, thickened at the bosses and printed from dry filament will still be working long after a stiffer, sharper, under-dried one has cracked or sagged.
Design for the load you can see and the time you cannot. That is the whole difference between a print that works and a part that lasts. InnoStation 3D is an authorized reseller of ANTINSKY filaments and printer parts, so every material referenced here is a current production product with published processing parameters — if you are matching a material to a long-term application, start from the datasheet, then verify with one instrumented test part before committing the design.