Why 3D Printed Parts Crack Under Vibration

The bracket held a 5kg weight without complaining. You bolted it to the machine, ran it for two hours, and found it in two pieces with a crack that runs dead straight across one layer line. Nothing about the static load changed. 3D printed parts crack under vibration because cyclic loading is a different failure mechanism from the one you tested for: it does not need to beat the part's strength, it only needs to repeat a much smaller stress enough times to start a crack at the weakest plane available. In a printed part, the weakest plane is already there, several hundred times over, in the form of layer boundaries. If the part is load-critical and the duty cycle is long, the provider directory lists shops that will make it by a process without that weak plane.

This post contains affiliate links. If you purchase through these links, 3D Prototyping Hub may earn a small commission at no extra cost to you.

This post also contains Amazon affiliate links. As an Amazon Associate we earn from qualifying purchases.

Find Out Which Failure You Actually Had

Five distinct things get called "it cracked from the vibration", and they want different fixes. What you find when you take the assembly apart tells you which one happened.

The first two are the common pair, and they often cause each other: the joint slackens, the part starts moving a fraction of a millimetre per cycle, and that movement becomes the load that cracks it.

Fatigue Is Not the Property You Tested

Hanging a weight on a bracket tells you it can carry that weight once. Vibration asks a different question — can it carry a fraction of that weight a few hundred thousand times an hour — and the answer is usually much less encouraging.

Thermoplastics have no dependable endurance limit. Below some stress, steel can cycle essentially forever; polymers mostly cannot, and every cycle at any meaningful stress accumulates a little damage. That damage concentrates wherever the local stress is highest and the material is weakest, and in an FDM part those two things coincide at a layer boundary beside a sharp corner. The crack starts invisibly, grows a few micrometres per thousand cycles, and the part looks perfect right up to the moment the remaining section cannot carry the load. That is why the failure feels sudden when the process was anything but. The underlying weakness is the same one behind parts that snap along layer lines — vibration is simply applying that load over and over.

There is a second, less obvious mechanism. Polymers are viscoelastic, so some of the energy in every flex cycle comes out as heat rather than going back into the spring. At high frequency and reasonable amplitude a part can warm itself measurably, and since PLA begins softening around 60°C, a self-heating PLA bracket gets weaker exactly while it is being worked hardest. If a failed part feels warm when you touch it, that is what you are feeling — and it is the same softening trap as prints that deform in a hot car, self-inflicted.

Orientation and Walls Decide This; Infill Barely Does

Most people respond to a fatigue failure by raising infill. It is the least effective change available.

A bracket under vibration is loaded in bending, and in bending the stress lives at the outer surfaces of the section while the middle carries almost nothing. The outer surfaces of a printed part are its perimeter walls. Going from 25% to 60% infill adds print time and mass in the region doing the least work, and the extra mass drops the part's natural frequency, which can move it closer to whatever is shaking it rather than further away.

Three changes do the work instead:

• More walls. Four to six perimeters on a loaded part puts solid, continuous extrusion where the stress actually is. In most slicers this is one number.
• Orientation. Print so the cyclic stress runs along the layers rather than across them. A bracket printed flat on its back resists bending far better than the same bracket printed standing up, where every bending cycle tries to peel one layer off the next.
• Section, not density. A rib, a gusset or a taller section raises stiffness much faster than infill does, because bending stiffness grows with the cube of depth.

If the geometry forces the weak orientation on you — a tall arm that has to be printed vertically — that is a strong signal to change process rather than settings.

Every Sharp Internal Corner Is a Crack Waiting to Start

Fatigue cracks are extraordinarily consistent about where they begin: the single highest stress concentration in the part, which is almost always an internal corner with no radius.

Printed parts are more notch-sensitive than the same polymer moulded, because the notch and the layer boundaries compound. A 2-3mm fillet in every internal corner is free in CAD, costs nothing in print time, and routinely multiplies fatigue life — it is the highest-value change on this page. Related rules follow from the same mechanic: do not put text, logos or a QR code on a face that flexes, since each recessed letter is a notch; do not run a screw hole out to within a wall thickness of an edge; chamfer the base so the flare of elephant foot is not sitting at the root of a cantilever.

One more, specific to FDM: tell your slicer to randomise or align the Z seam away from the loaded corner. The seam is a vertical line of start-stop points where each layer's wall is weakest, and a seam stacked up the inside of a fillet is a crack already drawn on the part.

The Bolted Joint Is Usually the Real Failure

More printed parts are killed by their fasteners than by their geometry, because a bolted joint in plastic loses the thing it depends on: preload.

Tighten a screw against printed plastic and the material under the head immediately begins to creep. Clamp load falls over hours and days, the joint reaches the point where friction no longer holds it, and the surfaces start to move microscopically against each other. From then on the bracket is being hammered rather than held, holes go oval, and the fatigue crack follows. The slow half of this mechanism is the same creep behind printed brackets that sag.

Fix it at the joint, in this order:

1. Take the clamp load out of the plastic. A metal compression spacer through the hole, or a brass insert long enough to bottom out, lets the screw clamp metal to metal. The plastic then locates the part instead of carrying preload, and creep stops mattering.
2. Spread what load is left. A wide washer under a cap head puts the force over several perimeter loops rather than pressing a cone into sparse infill. A cap screw and nut assortment with washers covers most of what a bench needs.
3. Use inserts rather than tapped plastic. Heat-set brass inserts melt into an oversized hole, bond as the plastic cools around them, and give a steel thread that survives being undone. Install them square and at a controlled temperature with an iron with conical insert tips; a crooked insert is a stress riser rather than a fix. Why cut threads fail so fast is covered in why printed threads strip.
4. Stop the nut rotating mechanically. Nylon-insert lock nuts hold where split washers do not — split washers have been shown repeatedly to do almost nothing against transverse vibration. Where a nut cannot fit, thread-forming screws designed for plastics displace material rather than cutting it and hold better than a machine screw in a drilled hole.

Thread locker is the last resort, not the first, and it needs a chemistry check: anaerobic lockers and thin cyanoacrylates can cause stress cracking in ABS and polycarbonate, which turns the fix into the next failure.

Isolate It Instead of Stiffening It

If the part only tears itself apart at one particular speed, you have a resonance problem, and no material upgrade will solve it — a resonant part sees several times the amplitude the input suggests, so you are chasing a moving target.

Two honest options. Shift the frequency: shorten the unsupported span, add a rib, bolt the free end to something, or remove mass from the end of a cantilever. A stiffer, lighter part has a higher natural frequency, which is why a carbon-fibre nylon version of the same geometry sometimes stops failing for reasons that have nothing to do with strength. Or kill the input: a viscoelastic pad such as Sorbothane, or a rubber grommet mount, converts vibration into heat before it reaches the plastic. Isolation is usually cheaper and faster than a third print, and it protects whatever is bolted to the part as well.

Material Decides How Many Cycles You Get

Same geometry, same bolts, very different service life.

For most people the honest sequence is PLA → PETG → ASA if heat and sun are involved, or straight to nylon if the part genuinely has to live with vibration for hundreds of hours. The material comparison in full sits in how to choose a 3D printing material, and our tier-by-tier notes on the tough grades are in the nylon filament guide and the ASA guide.

Two practical conditions attach to that advice. Nylon must be dry — it takes on moisture from room air within hours, and wet nylon prints with foamed, poorly fused layer welds, which is precisely the defect fatigue exploits, so a dryer box is part of the purchase rather than an accessory to it. And ASA, nylon and polycarbonate all want a chamber: if you are buying a machine for this class of work, QIDI's enclosed high-temperature machines are built around that requirement, and both Anycubic's US store and ELEGOO's store carry engineering filaments alongside their machines if you only need the spool.

Change These, In This Order

1. Fillet every internal corner, 2-3mm, and move the Z seam off the loaded ones.
2. Reprint in the orientation that puts the layer planes along the load, not across it.
3. Raise wall count to four to six. Leave infill where it is.
4. Put a metal spacer or a brass insert in the clamp path so the screw is not squeezing plastic.
5. Add wide washers and nylon-insert lock nuts. Torque to feel, not to the limit.
6. If it still fails at one specific speed, isolate the mount or change the part's stiffness.
7. Only then change material — and if you change to nylon, dry it first.

Work down that list and stop when it survives. Most parts stop failing at step 4.

When FDM Is the Wrong Process for the Job

Some parts should not be printed on a filament machine at all, and recognising that early is cheaper than four iterations.

The clearest case is a long duty cycle on a part whose failure costs more than the part: a drone arm, an engine-bay bracket, a pump mount. FDM's weak plane is a property of the process rather than a tuning problem, and walls, fillets and nylon reduce it without removing it. SLS nylon behaves far better because sintered parts are much closer to isotropic in every direction, which is the whole argument in SLS printing services. A machined or cast part removes the question entirely — the trade-offs are laid out in CNC machining vs 3D printing.

If the application is one of the obvious vibration-heavy ones, we have specific pages for what to ask a shop: drone and UAV parts and motorcycle and powersports parts. For everything else, our provider directory lists shops by location and process, and most will quote from an uploaded file inside a day.

A printed part can live a long time under vibration. It just has to be designed as if it will be shaken, rather than designed as if it will be weighed.

Hero photo by Locanam 3D Printing via Unsplash.