Every failed 3D printed part tells you where it broke. Far fewer people notice that the break location was drawn months earlier, in the CAD file, at a corner that looked perfectly innocent on screen. Brackets snap at the screw hole. Housings split where the wall meets the floor. Standoffs shear off flush with the base. Gears lose teeth at the root rather than at the tip. In every one of those cases the part was strong enough in bulk — the filament was fine, the walls were thick, the infill was sensible. What failed was the local geometry: a transition so abrupt that the load had nowhere to go except a single line.
This article is about that level of the problem. Not which material to buy, and not how parts connect to each other, but how a single printed component is shaped in cross-section so that the load spreads out instead of concentrating. Ribs, gussets, fillets, chamfers and bosses are the five tools that do that work. They are also the five features most consistently missing from the models people download and print unchanged.
Stress Concentration: The One Idea That Explains Every Cracked Corner
Push a part in tension, and the stress is spread evenly across its cross-section — until the section changes shape. At any step, notch or sharp inner corner, the stress field has to squeeze through a smaller area and bend around a corner, and the peak stress at that point can be several times the average. The ratio between peak and average is the stress concentration factor, and it depends almost entirely on one geometric ratio: the fillet radius divided by the section thickness. Small radius, thick section, large factor.
Here is what makes this brutal for printed parts. Aluminium brackets are usually designed with generous radii because a machinist can cut them and the material is isotropic. Printed parts are not isotropic, and the failure mode at a sharp corner interacts with layer direction: a stress concentration that would only be a nuisance in aluminium becomes the initiation site for a delamination crack in a part whose layers run perpendicular to the load. The corner does not just concentrate stress; it concentrates it exactly where the material is weakest.
Fillets: How Big, and Where They Actually Matter
A fillet is an inside radius where two surfaces meet. Its job is to give the stress field a smooth path to follow instead of a corner to pile up against. The relationship is not linear: going from a sharp corner to a small radius buys you most of the available relief, and going from small to generous buys progressively less.
| Fillet radius (relative to the member thickness) | Approximate relief vs. a sharp corner | Practical use |
|---|---|---|
| No fillet (sharp internal corner) | Baseline — stress concentration factor commonly 2 to 4 | Never acceptable on a loaded corner |
| 0.25 x thickness | Roughly half the available relief | Absolute minimum if space is tight |
| 0.5 x thickness | Most of the available relief | The sensible default for printed brackets |
| 1.0 x thickness or more | Diminishing returns; approaches the smooth-body value | Worth it when the outer envelope allows |
The practical rule for printed parts: fillet every inside corner that carries load to at least half the thickness of the thinner member, and never leave an inside corner sharp. On a 4 mm-thick bracket arm, that is a 2 mm radius. On a 6 mm rib meeting a 3 mm wall, it is 1.5 mm.
There is a second, printer-specific reason to fillet inside corners that has nothing to do with stress. A sharp internal corner is where the toolpath has to reverse direction in a tight space, and it is the classic site of under-extrusion, bulging and layer stacking artefacts. A radius gives the slicer room to lay material down properly. So the same feature that fixes the mechanics also fixes the finish.
Where Not to Fillet
Fillets on outer edges of a bearing surface, a sealing face, a mating face or a flat that must sit flush are a mistake — they remove the contact area you need. And a fillet cannot rescue a corner that is loaded in pure peel. If the load pulls two layers apart at the joint, the geometry inside the layer plane is irrelevant; the fix is orientation, not radius. This is the boundary condition most people miss: fillets manage in-plane stress, they do nothing for inter-layer peel. The inter-layer bonding and orientation fundamentals are the other half of the picture.
Ribs and Gussets: Adding Stiffness Without Adding Solid Plastic
Stiffness in bending scales with the cube of depth and only linearly with width. That single fact is why a thin, deep rib does far more work than a thick, shallow one, and why a hollow ribbed section beats a solid section of the same mass by a wide margin. A gusset is simply a rib placed diagonally across an inside corner — a bracket's most common weak point.
| Reinforcement | What it resists | Printability | Design note |
|---|---|---|---|
| Vertical rib (in the Z direction) | Bending of a wall or panel | Excellent — no support, walls run along the load | Best printed reinforcement by a wide margin |
| Horizontal rib (flat in XY) | Bending of a horizontal span | Good if it lies flat on the bed | Keep the rib's own layers perpendicular to the panel's layers |
| Diagonal gusset in a corner | Rotation of one member relative to another | Good on a vertical corner; needs support if it bridges in air | The single most effective fix for a cracking bracket |
| Rib on a curved surface | Bending plus torsion | Moderate — the rib root is a curve and needs care | Fillet the entire root, not just the ends |
Rib thickness is where most designs go wrong. The instinct is to make ribs thick because thick feels strong. In practice the useful range is 0.5 to 0.8 times the wall thickness, and the reason is print physics rather than mechanics: a rib that is thicker than the wall it joins creates a thick-to-thin transition that shrinks at a different rate, pulls the wall in, and leaves a visible witness line on the opposite face. It also produces the classic sink mark. Thinner than the wall is the correct direction, and it costs you almost nothing in stiffness because the rib's contribution is dominated by its depth.
Spacing matters as much as thickness. Space ribs roughly 10 to 15 times the wall thickness apart, or you end up with a nearly solid block whose mass gain buys very little extra section depth. If your ribs are creeping closer together than that, reduce the count and increase the depth instead.
The Rib Root Is a New Corner You Just Created
Adding a rib to a panel creates two fresh inside corners where the rib meets the surface. If you leave those sharp, you have moved the failure site rather than removing it — and you have done it in a location with less material around it. Every rib root needs a fillet of at least a third of the rib thickness, ideally more. This is the most common self-inflicted wound in functional part design.
Chamfers: Where a Rounded Corner Is the Wrong Answer
A 45-degree chamfer and a fillet are often treated as interchangeable cosmetic options. They are not. A chamfer removes material in a straight cut and creates a hard edge where it lands; a fillet adds material in a curve and blends. Both relieve stress at the transition, but the chamfer does it with less volume and less risk of the outer surface bulging.
| Situation | Use | Why |
|---|---|---|
| Inside corner under load | Fillet | A chamfer at an inside corner still leaves a stress-raiser at each end of the cut |
| Outside edge that a person or cable will touch | Chamfer, 0.5 to 1 mm | Cheap to print, no bulge, removes the sharp edge and reduces elephant-foot interference |
| Bottom edge sitting on the build plate | Chamfer 0.4 to 0.8 mm | Directly counteracts first-layer squish and elephant foot without a separate slicer setting |
| Lead-in to a hole or bore | Chamfer | Guides the fastener, removes the burr where the perimeter meets the opening |
| Transition between two very different thicknesses | Forty-five degree taper, several millimeters long | Spreads a step change over distance, which no single fillet can do |
The taper case is worth dwelling on. When a 6 mm boss meets a 2 mm wall, a fillet at the base helps but the section still changes sharply. A 45-degree tapered skirt, 3 to 5 mm tall, distributes that change over a real distance and is far more effective. Think of it as a fillet drawn at the scale of the whole transition rather than the scale of the corner.
Layer Direction Changes What Every One of These Rules Means
None of the geometry above exists in a vacuum. A printed part is a stack of ribbons weakly bonded to each other, and reinforcements only work if the load path stays inside a ribbon rather than crossing between them.
| Feature orientation | Load behaviour | Practical consequence |
|---|---|---|
| Wall or rib printed vertically, load in-plane | Load runs along the extruded strands | Best case — this is what a gusset should look like |
| Wall or rib whose layers cross the panel's layers | Fine for stiffness, poor for peel | Acceptable for stiffness-only ribs, not for load-bearing joints |
| Fillet plane parallel to the bed | The radius is printed as a clean curve | Best finish and best mechanical result |
| Fillet plane perpendicular to the bed | The radius becomes a stairstep of layer ends | Still relieves stress, but the surface is rough and the outer edge is fragile |
The takeaway: a reinforcement is only as good as the orientation of the layers inside it. A beautifully filleted gusset on a part printed in the wrong orientation will still delaminate, because the crack travels between layers, not through a corner. If you cannot reorient the part, you sometimes have to reorient the feature — or accept that a printed bracket needs tolerance and fit compensation elsewhere rather than a geometry fix.
Bosses and Standoffs: The Features That Fail Most Often
A boss is the cylindrical pad that a screw, insert or pin goes into. It is the highest-stress feature on most functional parts because it concentrates force from a small fastener head into a large flat panel, and because people commonly print it as a thin-walled tube standing on a flat floor with a sharp corner at the base.
The rules that matter, in order of impact:
- Wall thickness. A boss needs at least three perimeter lines of solid wall. On a 0.4 mm nozzle that means 1.2 mm minimum, and 2 mm is safer. Infill inside a boss contributes almost nothing — the load travels through the wall.
- Outer diameter. Two times the fastener diameter is the floor; 2.5 to 3 times is comfortable. A boss narrower than that has too little wall to carry the load.
- A fillet or taper at the base. This is the single most important detail. The boss-to-floor corner carries the bending moment the fastener applies, and it is where the crack starts. A fillet of a third of the wall thickness at minimum, or a tapered skirt, is not cosmetic.
- Full connection to adjacent walls. A boss floating in the middle of a panel is a lever waiting to snap. If a wall runs nearby, connect the boss to it with a short rib.
- Enough material under the fastener head. You need at least one full perimeter ring of solid plastic around the hole — a wall that stops right at the hole edge leaves a split-prone collar.
A common instinct is to make the boss short and the panel thick to compensate. This usually fails. The panel gives little local support, and the short thick boss transfers load straight into the panel's weakest layer direction. Local geometry beats global thickness nearly every time.
Translating All of This Into Slicer Settings
| Design feature | What the slicer must do | Typical value |
|---|---|---|
| Ribs and gussets | Print as solid walls, not infill | Rib thickness a multiple of the extrusion width; 4+ perimeters locally |
| Filleted corner | Continue the perimeter smoothly around the radius | Fine layer height near the radius; avoid gap-fill-only areas |
| Boss with a hole | Add perimeters around the hole rather than relying on infill | 4 to 6 perimeters within the boss |
| Overhanging gusset underside | May need support or a chamfered underside | Chamfer 45 degrees below 45 degrees overhang angle |
| Elephant foot on base chamfer | Slight negative z-compensation plus the physical chamfer | 0.1 to 0.2 mm compensation |
One setting deserves a specific warning: do not let the slicer fill a structural rib with infill. Most slicers treat a thin vertical feature as a candidate for sparse infill if it falls below a wall-count threshold. A rib that ends up 30 % infill is a rib in name only. Check the preview in cross-section before committing to a long print.
The extrusion side matters too. A radius is a place where the extruder is constantly varying its pressure, and a part printed with poor wall order and flow control will show gaps at exactly the fillets you added for strength. This is where good filament matters more than exotic filament: consistent diameter means consistent pressure means the radius actually comes out solid. A spool like PETG that holds its tolerance well is worth more here than a high-modulus blend that prints unevenly.
Material Choice, Through the Lens of Local Geometry
Different filaments respond differently to the same stress concentration, because the failure mode at a sharp corner depends on whether the material yields, crazes or delaminates.
- ABS and ASA tolerate stress concentration comparatively well — they yield locally and distribute the load, which is exactly what you want at a boss base. Their weakness is inter-layer adhesion, so the fillet must be oriented so the crack cannot travel between layers.
- PETG sits in a useful middle: better layer adhesion than ABS, more ductility than PLA. It is the safest default for printed brackets and reinforcement geometry that has to survive handling.
- PLA is stiff and strong in bulk but has very little capacity to redistribute local stress — a sharp corner that ABS would merely yield at, PLA will crack at. If you print PLA, the fillets are not optional.
- Fibre-filled materials such as PETG-CF raise stiffness and print well at radii, but they are less forgiving in peel. Use them when the load path stays in-plane, and fillet generously.
- Resin parts behave differently again: they have no layer lines to delaminate, but they are typically more brittle in thin features. The reinforcement logic is the same, but the impact-resistant resin family is the better place to start when a boss has to take a screw load.
The pattern is consistent: the more brittle the material, the more the corner geometry carries the design. A part that is fine in ASA with a sharp corner is a failure waiting to happen in the same shape printed in a stiff, low-ductility blend.
A Deliberate Limitation: What Geometry Cannot Fix
Not every failure is a corner. It is worth being honest about the situations where no amount of filleting will help, because chasing geometry in these cases wastes time and filament:
- Peel loading across layers. If the load pulls layers apart, the fix is orientation or a different joint, not a radius.
- Creep under sustained load. A fillet reduces peak stress but does not change the fact that a plastic part under continuous load will slowly deform. That is a different failure mechanism with a different fix.
- Warping driven by thermal shrinkage. Not a local geometry problem. Fillets can slightly reduce the restraint at a corner, but the real levers are chamber temperature and part cooling.
- Gross under-designed section. If the whole member is too thin, adding ribs and fillets just moves the failure. Section size comes first.
The useful mental model: geometry changes where and how abruptly a part fails. It does not change whether the part is fundamentally big enough, and it does not change the physics of sustained load or thermal shrinkage.
The Pre-Print Checklist
- Every loaded inside corner has a fillet of at least half the thinner member's thickness.
- No rib is thicker than the wall it joins; ribs sit at 0.5 to 0.8 times wall thickness.
- Every rib root has its own fillet — adding a rib creates two new corners.
- Ribs are spaced 10 to 15 wall-thicknesses apart, not packed together.
- Every corner that carries a bending moment has either a gusset or a taper, not just a radius.
- Every boss has at least three walls of thickness, an outer diameter of 2x the fastener, and a filleted base.
- Bosses are tied into nearby walls with a short rib instead of floating free.
- Outer touch edges have a 0.5 to 1 mm chamfer; the base edge has 0.4 to 0.8 mm.
- The part is oriented so that load paths run along layers, and the fillet planes are parallel to the bed where possible.
- Structural ribs are verified in the slicer preview as solid walls, not as sparse infill.
None of these are exotic. They are the same principles a mechanical engineer has applied to cast and machined parts for a century, adjusted for one inconvenient fact: a printed part is a stack of weakly bonded ribbons, and geometry is the only tool that lets you route load through the ribbons rather than across them. Get the corners right and the material almost stops mattering. Get them wrong and no filament on the shelf will save the part.