There is a reflex, when a printed part flexes too much or breaks, to raise the infill. It is the wrong lever most of the time. The material that resists bending is the material furthest from the middle of the part, and in a printed object that is the walls.
This is not a printing insight, it is a structural one, and it is the same reason a scaffolding pole is a tube rather than a rod.
Four runs of the same bracket
The subject is a bracket 80 mm tall with 95 cm³ inside it, sliced at 0.2 mm through a 0.4 mm nozzle at 80 mm/s. Two variables move and nothing else does: how many perimeters, and how dense the fill.
| Shell | Infill | Time | Extruded volume |
|---|---|---|---|
| 2 walls | 20% | 5.429 h | 106,036 mm³ |
| 4 walls | 20% | 6.22 h | 121,811 mm³ |
| 6 walls | 20% | 6.991 h | 137,178 mm³ |
| 2 walls | 50% | 11.259 h | 222,273 mm³ |
The last row is the one that settles the argument. Going from two walls to six costs about an hour and a half; going from 20% to 50% infill on two walls costs nearly six — so if what you want is strength per hour, shell is the cheaper purchase. The print time estimator will run the same pair on your own part.
Read the first and third rows against the fourth. Tripling the shell adds under an hour and a half and under a third more plastic. Raising infill from 20 to 50% on the original thin shell adds nearly six hours and more than doubles the plastic. And of the two parts, the six-walled one resists bending better, because its extra material is out at the surface where bending stress is highest rather than distributed through the middle where it is lowest.
That is the whole argument, and it holds for almost every functional part you will print.
Why the position of the material matters so much
Bending stress in a beam is zero at the neutral axis and greatest at the surfaces. Material sitting near the middle of a part contributes almost nothing to its bending stiffness, and infill is by definition material near the middle.
The relationship is steep. A section's resistance to bending grows with roughly the cube of its depth, which means the outermost tenth of a part's thickness does a wildly disproportionate share of the work. Adding a wall moves material to precisely the place where it counts; adding infill moves it to the place where it counts least.
Infill's actual job is different, and worth stating so it is not maligned:
- It supports the top surface. Without enough infill the top layers sag between the gaps and you get pillowing.
- It stops the walls buckling inward under compression, which is why a thin-walled hollow part crushes rather than bends.
- It carries shear between the top and bottom skins of a flat plate.
- It resists point loads where something bears directly on a face.
Those are real jobs. Global bending stiffness is not on the list.
The numbers to reach for
For a functional part in a rigid material, a sensible default is a shell of three to five walls with infill somewhere between 10 and 25%, and then the following adjustments:
- It bends and you do not want it to — add walls, two at a time, before you touch infill.
- The top surface looks poor — raise infill, or add top layers, or both. This is an infill problem.
- Something screws into it — add walls locally around the boss, or add a modifier region. Infill does not help a thread.
- It is crushed rather than bent — raise infill; this is the compression case where it earns its keep.
- It has to be light — reduce infill hard and keep the shell. A part at very low infill with a good shell is stiffer and lighter than the reverse.
The case for solid, and why it is rarer than people think
Full-density infill has a legitimate use: small parts where the shell is most of the volume anyway, parts that will be tapped or machined, and parts where impermeability matters — a printed container that must not weep needs a solid wall section rather than a solid interior, but the two are usually specified together.
Outside those, 100% infill is a common and expensive mistake. It roughly doubles or triples the print time, it uses the plastic in the least effective place, and on materials that shrink it actively increases warping and internal stress, because there is more contracting mass fighting the plate. A solid ABS block is much harder to print successfully than a shelled one.
Wall count interacts with the nozzle, and people forget
Walls are counted, not measured. Three walls through a 0.4 mm nozzle is a thinner shell than three walls through a 0.6 mm one, so a profile carried across nozzle sizes silently changes the part. If you keep profiles per nozzle — and you should — check the wall count in each rather than copying it.
The related trap is designing a wall thickness that is not a whole number of extrusion widths. A 1.5 mm rib on a 0.45 mm line width is three and a bit lines wide, and the slicer will either leave a gap up the middle or overlap to fill it. Designing ribs and walls in multiples of your line width is a five-minute habit that removes a whole class of gaps between infill and walls.
What to do with an existing part that failed
Do not reprint it stronger everywhere. Look at where it broke.
If it snapped along a layer line, the fix is orientation and interlayer bonding, not walls — layer height and strength covers that case. If it bent and stayed bent, it was under-walled or the material creeps. If it crushed, raise infill. If it split at a screw, the boss was too thin.
One reprint aimed at the real cause beats three reprints with every setting turned up, and it costs a great deal less in hours.