Most failed 3D printed parts do not snap in half. They lose a dimension. A rail gets a millimetre of slop, a hose neck starts weeping, a gasket stops sealing, a bolt that was tight in March is loose by October. In every case the geometry was fine on day one. What changed was that material quietly left the part, or quietly moved out of the way.
That is a load path problem, not a modelling problem. Understanding where force actually travels through a printed part tells you which surfaces to protect, which cross-sections to thicken, and which features to stop printing altogether and buy instead.
Where the Load Actually Goes
Engineers draw load paths as lines of force flowing from where the load enters a part to where it leaves. In a printed part those lines cross two very different kinds of material: the solid perimeter around the outside, and the sparse lattice inside. They behave nothing alike under sustained load.
Perimeters carry load along their length almost as well as the bulk polymer. Infill carries load through a series of crossings, and each crossing is a place where the extrusion was laid down in a different direction. Under a steady pull, the crossings are where the part starts to give. This is why adding infill often makes a part worse, not better — extra lattice adds weight and print time but very little tensile capability compared with another perimeter.
The practical consequence is a routing rule: make the solid shell continuous from the load's origin to its exit, and treat infill as a shape holder rather than a load carrier. If a bracket's load enters at a bolt hole and exits at a wall mount, the perimeter should trace that route without interruption. A hole drilled through the middle of the path does not just remove material, it forces the load to divert around the hole edge, where it concentrates.
The Four Load Path Failures Nobody Models
Four mechanisms account for nearly every "it worked in CAD" part that fails in service. Each has a different signature on the failed part, which makes diagnosis possible after the fact.
| Mechanism | Signature on the failed part | What it is really doing | Fix that works |
|---|---|---|---|
| Sliding abrasion | Polished, dished or blackened contact surface; fine dust nearby | Both surfaces are the same polymer; the softer one transfers and tears | Change material pair, add a dissimilar wear surface, increase contact area |
| Notch stress | Crack starting exactly at an inside corner or at a layer line beside a hole | A sharp geometric change triples local stress; the layer line is the weakest plane | Generous radius, blend the transition, orient so layers cross the notch |
| Fastener relaxation | Bolt still present but loose; witness marks rotated; joint has play | The polymer creeps under the bolt's preload, so clamp force decays | Metal insert, larger washer area, or spring element to keep force constant |
| Compression set | Seal or pad has taken a new, flattened shape and no longer springs back | The material was compressed past its elastic limit and stayed deformed | Choose a material that recovers, reduce compression, constrain the squeeze |
All four share a root cause: the part was designed for a shape, not for a history of forces. A printed part in service spends months or years under loads that never appear in a CAD screenshot.
Sliding Contacts: Why Identical Materials Weld and Wear
A printed slide, rail, guide or shoe runs two surfaces of the same polymer against each other. That is the worst pairing available. Identical polymers have identical surface energy, so the two surfaces form tiny local bonds wherever they touch, and sliding tears those bonds out of whichever side is softer at that instant. The result is a dished, polished, blackened track and a pile of fine powder that nobody notices until the play becomes visible.
Worse, the heat goes into the contact rather than out of it. Polymer is a poor conductor, so friction heat stays at the surface, softens it, and raises the real contact area — which raises friction further. That runaway is how a tight-but-movable slide becomes a seized one.
The design answer is to stop running the same material on itself. Four options, in ascending order of effort:
- Pair with a dissimilar surface. A printed slider running on aluminium extrusion, a glass-filled nylon, or a PTFE-lined strip costs one part and changes the tribology completely. There is no self-bonding because the surfaces have different chemistry.
- Embed the wear surface. Print a pocket for a strip of acetal (POM), UHMW tape or a piece of the same plastic used for cutting boards. These materials are chosen by industry precisely because they slide well and recover.
- Run on a rolling element instead of a sliding one. If the motion is short and repeatable, a printed pivot with a steel pin, described in this guide to pivots, shafts and bushings that stay tight for years, removes the sliding interface entirely.
- Buy the bearing. For continuous rotation under load, a printed sleeve will always lose to a purchased bushing. The engineering question is not whether you can print it, but whether the service life justifies the experiment.
One more rule that saves a lot of rebuilds: never design a sliding pair to run dry under high pressure and high speed at the same time. Pressure and velocity multiply. A high-pressure, slow-moving slide is fine. A low-pressure, fast-moving one is fine. Both at once is where printed parts die, and the failure is thermal, not mechanical — the surface reaches its softening point long before any stress limit is approached. This is the same failure family covered in the deeper look at printed bearings and sliding surfaces that weld themselves shut.
Fasteners Come Loose Because the Plastic Moves
A bolt does not loosen because it unscrews itself. It loosens because the material under its head changes thickness. Polymer exhibits creep; under a sustained clamp load it continues to deform, slowly, for weeks. The bolt head follows that deformation inward, the preload decays, and the joint loses its grip. Vibration then finishes the job.
Two design moves fix this, and they are complementary. First, spread the clamp load over an area large enough that the local pressure stays inside the material's long-term capability — a printed washer or a boss top of at least three bolt diameters across. Second, put a metal insert in the hole so that the thread itself is not polymer, as detailed in the guide to threads, engagement depth and inserts that survive real loads.
Where neither is possible, add compliance. A joint that includes a spring washer, a printed wave washer or an elastomer pad keeps the clamp force roughly constant even as the polymer underneath creeps. The joint then works with the material's behaviour instead of betting against it. The same creeping mechanism is what limits functional parts months after printing, and it is worth reading before designing anything that must stay tight.
Stress Concentrations Are Where the Load Path Turns
A load path cannot turn a corner without concentrating. Printed parts turn corners badly because the layer interfaces cross the path at right angles, and the interface is the weakest plane in the whole part. A sharp inside corner therefore produces a crack that starts at the corner and runs along a layer line, exactly where the part is least able to resist.
The standard fixes apply, but they need to be applied at the right scale. A 0.5 mm fillet on a 0.2 mm layer height is only two and a half layers of material; it is a rounding, not a stress relief. For functional parts, use a radius at least a quarter of the local thickness, and blend the transition with a taper rather than a radius where the geometry allows. The ribs, gussets, fillets and bosses article covers how these features should be shaped so the part breaks where you intend it to.
Orientation is the second half of the same problem. A part whose load path runs across layers will fail at the interfaces. A part whose load path runs along the layers will fail much later. That is the central argument of designing parts before they are sliced, and it explains why the same model printed on two machines can behave like two different materials.
Pads, Feet and Hose Necks: The Three Places Parts Quietly Give Up
Certain geometries appear in almost every printed device and fail in almost the same way every time. They deserve names.
Pads and feet that take a set
A printed foot or bumper is asked to do two opposite jobs: hold a shape under load and spring back when the load leaves. Most rigid filaments do the first well and the second badly. PLA and PETG under a sustained load will compress and stay compressed — compression set — so a device that sat level in January wobbles in June. Flexible materials recover far better, and the article on compliant mechanisms explains how to make an elastic element that does not rely on the bulk material's memory. Where the load is heavy and the life is long, print a hard shell and buy the elastomer.
Hose necks and cable glands that weep
The neck of a printed hose fitting is the worst possible geometry for a layered part. The load comes from the hose pulling sideways, the fluid applies internal pressure, and the whole assembly sits at the end of a lever arm. A printed neck should be thick, should be reinforced with perimeters running axially, and should clamp the hose with a barb rather than a smooth wall. A printed Barb is not a compression fitting and should not be expected to seal without a clamp. Where the connection carries fluid continuously, the sealing principles in gaskets, O-rings and groove geometry matter more than the fitting's shape.
Cable runs are easier but have their own trap: a printed hole that a cable passes through will chafe the insulation if the edge is sharp, and the layer lines make an excellent saw. A generous radius, a grommet or a smooth-drawn loop around the cable removes the risk. This is a case where the geometry costs nothing and the omission is expensive.
Corner bosses that pull away
A boss at the corner of a printed box sits in a region where the shrinking part pulls away from the bed and where the wall's own contraction competes with it. The boss is the stiffest local feature, so it takes the load and often the crack. Keeping bosses away from corners, tying them to walls with ribs, and letting mounts survive the screw hole are the standard defences. The warping guide covers the print-side half of the same problem.
Thermal Load: The Path Nobody Draws
Load paths are usually drawn from mechanical forces only. Heat adds a second path that behaves identically but arrives by a different route. A part that carries warm fluid, sits near a heated bed, or holds a motor expands and contracts with every cycle, and a restrained part converts that expansion into stress.
Where two printed parts with different thermal expansion are bolted together, the joint will work slightly on every cycle. Over thousands of cycles, that working loosens the fasteners and frets the contact faces. This is the mechanism behind failed printed gears in warm enclosures and the reason motor mounts fail at the screw hole rather than in the body. The practical defence is to allow one degree of freedom. A slotted hole, an oversize clearance, or a compliant element absorbs the expansion instead of fighting it.
When the Load Path Leaves the Polymer Entirely
There is a point where no amount of design intelligence saves a printed part, and recognising it early is a skill worth developing. The following are reasons to change the approach rather than the geometry:
- The contact must hold a dimensional relationship over thousands of cycles. Printed polymer will wear; metal will not in the same way.
- The load is continuous and the material's long-term creep strength is the limiting factor, not its short-term strength. Test it, or move to a different material.
- The part must seal against fluid without a clamp or gasket. Layers are not a sealing surface on their own.
- The temperature in service approaches the material's glass transition. Stiffness collapses there, and no wall thickness restores it.
None of these is a failure of printing. They are a description of where printing stops being the right tool. The honest answer, in each case, is a hybrid: the printed part provides the shape and the purchased part provides the precision, the seal or the wear surface. A design that combines a printed housing with a metal insert, an acetal slide or a silicone seal usually costs less than a printed part engineered to defeat physics.
Material Choice Driven by Load Path, Not by Hobby Convention
Most material discussions start with "which filament is strongest" and end with a chart. Load paths suggest a different question: which property matters at the point where the load actually has to travel? The table below is organised that way.
| What the load path demands | Sensible material | Why | What will disappoint |
|---|---|---|---|
| Stiffness with light weight (arms, spars, plates) | PLA, PETG with thick skins | High modulus per unit cost; perimeters do most of the work | Thin sheets that flex; infill used as a substitute for walls |
| Impact and rough handling | ABS or ASA with a slightly warm enclosure | Higher ductility; fails by bending rather than cracking | Brittle parts printed cold and fast; sharp internal corners |
| Wear and sliding contact | Acetal or UHMW insert in a printed body | Low friction, good recovery, dissimilar to the counterface | Printing both halves in the same filament |
| Sustained compression (pads, seals, gaskets) | TPU 85A to 95A, or silicone | Recovers elastically; resists taking a set | Rigid filament used as a spring |
| Repeated flexing (clips, straps, hinges) | TPU, PEBA, or a properly designed compliant feature | Elastic strain to failure is large | PLA hinges, which crack at the first few cycles |
| Clamped joints that stay tight | Any polymer with metal inserts and a wide washer | No creep of the thread form; clamp load spreads | Threaded polymer directly under a small bolt head |
Two of these rows point at the same underlying discipline, which is inter-layer bonding and orientation. In a printed part the material's properties are direction-dependent, so the load path has to be chosen with that asymmetry in mind. A part that is strong along the layers and weak across them is not a compromise; it is a design statement, and it should be deliberate.
Printing the Load Path Properly
Once the geometry is right, the print settings determine whether the path is actually continuous. The variables that matter most are not the ones usually discussed.
- Perimeters run the load. Add walls, not infill. Two additional perimeters are worth more than fifteen percent of extra infill for almost every load-bearing part.
- Layer lines are the weak plane. A part in tension along the build direction is a stack of bonded plates. Rotate it on the plate so the load runs in the plane.
- Temperatures matter more than speed. A slightly hotter nozzle and reduced part cooling improve inter-layer bonding, which is exactly the property a load path depends on. The material-specific temperature guidance in effect and texture filaments applies equally to plain materials.
- Do not add supports where the load path runs. Support removal leaves a rougher, thinner surface with a stress riser at the break line. Redirect the orientation instead. The support guide covers when supports are genuinely needed.
- Dry the filament. Moisture turns into steam in the hot end and leaves voids at the interfaces, which is invisible until the part is in service. Moisture by material explains why some polymers suffer more than others.
The nozzle matters as a second-order effect: a worn or oversized nozzle changes extrusion width and therefore the effective wall thickness, which is the dimension the load path depends on. That is why an abrasive material and a soft brass nozzle are a poor combination, and why a hardened tungsten nozzle pays for itself on carbon-filled filaments.
A Check List for Load-Bearing Parts
- Trace the load path from where the force enters to where it leaves and mark it on the drawing.
- Confirm the solid perimeter is continuous along that entire route.
- Identify every turn, step and hole on the path and give it a radius of at least a quarter of the local thickness.
- Check that the load does not run across layer lines. If it does, re-orient or redesign.
- Replace every sustained sliding contact on the same polymer with a dissimilar surface.
- Replace every thread and every small bolt head with an insert and a wide washer.
- Give thermal expansion one free direction, or expect the joint to work loose.
- Reduce compression on pads and seals until the material still recovers after a week under load.
- Add perimeters rather than infill where stiffness is short.
- Decide, explicitly, which parts of the assembly should be purchased rather than printed. Write it down.
None of these steps requires a CAD system beyond the one you already have. They require knowing where the force goes, which is the part of engineering that a printer cannot do for you.
If you are building functional assemblies and want your material and hardware choices to match the load path rather than a hobby chart, browse the ABS, PETG and TPU ranges at InnoStation 3D, and pair them with a hardened tungsten nozzle and a filament dryer so the interfaces hold the strength the design assumed.