Shell Count, Wall Thickness and Infill: Why Adding Plastic Often Makes a 3D Printed Part Weaker

Ask ten people how to make a 3D printed part stronger and eight of them will say the same thing: increase the infill. It is the most visible number in the slicer, it feels like the obvious strength dial, and it is very often the wrong one. A part printed at 100 % infill with two perimeters can be measurably weaker in bending than the same part printed at 25 % infill with six perimeters, using less filament and less time. That result surprises people the first time they see it, and it stops being surprising once you understand where load actually travels inside a printed part.

This article is about that division of labour. We covered infill pattern and density selection separately, and the internal structure trade-offs elsewhere, so the question here is narrower and more useful: given a fixed material and a fixed shape, how should the plastic be distributed between the shell and the inside — and what happens physically when you get it wrong?

The Skin Carries the Load, the Core Mostly Fills Space

Bend a beam and the stress is not uniform through its thickness. It is highest at the outer surfaces and drops to zero at the neutral axis in the middle. In a printed part, the outer surfaces are exactly where the perimeters are printed, and the neutral axis is exactly where the sparse infill sits. The material that carries the load is already in the right place; the material in the middle is doing almost nothing for bending stiffness, which is why hollowing out the core costs so little.

Quantitatively, bending stiffness scales with the second moment of area, which grows with the cube of depth and only linearly with width. Throwing additional material into the core increases the cross-section by a small amount while doing nothing to move material away from the neutral axis. Adding perimeters does the opposite: it thickens the load-carrying shell exactly where stress is highest.

Where plastic goes	Effect on bending strength	Effect on mass	Verdict
Extra perimeter lines	Strong positive — thickens the high-stress shell	Moderate	First thing to change
More infill density	Mild positive at best, often negligible	Large	Last resort
More top/bottom solid layers	Positive for local stiffness and surface, little for global bending	Moderate	Situational
Thicker design wall in CAD	Strong positive, larger than anything the slicer can do	Large	The real lever
Taller part (increased depth)	Very strong — cube law, no extra material needed	Zero	Free if geometry allows

The table is worth reading twice, because the second row is the one people refuse to believe. Infill density is a weak lever specifically for bending, which is what most functional printed parts experience.

What a Perimeter Actually Contributes, and Why It Is Not Linear

A perimeter line is a continuous strand of filament extruded along a path lying in the XY plane, bonded to the line beside it and to the layers below. Unlike the infill lattice, which is a collection of crossings and gaps, the shell is a set of concentric closed loops. Under load, those loops act like a laminated skin: continuous in the direction of the load, with no internal voids to concentrate stress.

The consequence is that shell contribution is roughly proportional to the shell's cross-sectional area in the load path — which means it scales with wall thickness, and wall thickness scales with the number of perimeter lines. Going from two perimeters to four on a 0.4 mm nozzle takes the wall from 0.8 mm to 1.6 mm, which is a doubling of the load-carrying shell. Going from four to six adds another 50 %. There is no magic threshold, but there is a very practical one: below three perimeters, the shell is frequently interrupted by the infill boundary, and the load path looks more like a chain of links than a continuous skin.

Perimeter count (0.4 mm nozzle)	Approximate wall	Where it is appropriate
1 to 2	0.4 to 0.8 mm	Display models, vase-mode style single-wall surfaces, nothing load-bearing
3	1.2 mm	Absolute floor for a part that holds a screw or takes handling load
4 to 5	1.6 to 2.0 mm	The sensible default for brackets, housings and anything functional
6 to 8	2.4 to 3.2 mm	High-load pockets, bosses, screw holes, loaded lugs — local reinforcement
More than 8	3.2 mm and up	Diminishing returns unless the CAD wall really is that thick

Note the asymmetry in that table: the useful number is not one value for the whole part. Structural regions and cosmetic regions want different wall counts, and most slicers now let you set a per-object or per-modifier value. A single print with four perimeters globally and eight around the mounting holes is a better part than eight everywhere, at lower cost.

The 1.5 mm Trap: When a Wall Is Neither Thin Nor Thick

There is a specific wall dimension that causes more disappointment than any other, and it sits just above one and a half millimetres on a 0.4 mm nozzle. It is the thickness at which a slicer feels entitled to place sparse infill between two thin shells — and because the gap between the shells is only a few tenths of a millimetre, the result is a wall built out of two skins with a void between them.

That construction is weak in a way that is invisible on a printed part. The two shells are bonded only through whatever sparse material caught in the gap, and there is a genuine internal free surface. Under bending, the two skins act as separate thin members rather than one thick one, and the strength is closer to the thin member than to the thick one. Worse, the void is a stress raiser in exactly the location where the design intended solid material.

Designed wall thickness (0.4 mm nozzle)	What the slicer typically does	Fix
Below 1.2 mm	Solid, 2 to 3 lines, no interior	Fine — but too thin for load
1.2 to 1.6 mm	Solid lines only, occasionally with a thin gap	Usually safe; verify in preview
1.7 to 2.4 mm	Two shells plus a sparse or gap-filled interior	Thin the wall to 1.6 mm, or thicken it to 2.5 mm and force solid infill locally
2.5 mm and above	Two shells plus meaningful infill	Acceptable if the interior is solid or high density

The clean solution is to stop designing walls that sit in the trap. Either keep a functional wall at or below the thin solid threshold, in which case the slicer produces solid plastic and the part is as strong as the drawing says, or commit to a genuinely thick wall and set a high local infill so the interior is real. The worst outcome is the accidental middle, which happens by default.

Why Adding Infill Cannot Fix a Bad Wall

Suppose you keep the 2 mm wall and raise infill to 50 %. What have you actually built? Inside each wall there is now a denser web, still sitting between two shells, still connected to them only where extrusion lines happen to touch. The part gets heavier and stiffer, but the load path is still the two thin shells, and the failure mode — the shell separating from the core — is unchanged. You have spent filament without moving the structure into a better configuration.

This is the mechanism behind a result that seems paradoxical: adding plastic can make a part fail sooner. A denser core has a different thermal mass and shrinks differently from the surrounding shell, which increases residual stress at the shell-core boundary. It also converts a part that would yield and deform into one that transfers load abruptly into a stiff interior and cracks. And because a printed part is a stack of weakly bonded layers, that boundary is where the separation will start.

• Where it does help. Compression along the Z axis — infill resists crushing load reasonably well, because the load path runs down the extrusion strands rather than between layers.
• Where it barely helps. Bending and torsion, which is most real duty cycles.
• Where it hurts. Thin-walled parts where the denser core raises local stiffness enough to change the failure from yielding to fracturing, and any part where the added mass matters for the mounting hardware.

Layer Direction Decides Which Wall Rule Applies

A shell printed as a vertical wall has its extrusion strands running horizontally, in the plane of the wall. That is the good orientation: the load travels along the strands. The same shell printed so the wall's thickness is vertical has strands running across it, and then the wall strength is governed by inter-layer bonding rather than by wall thickness — which is much lower, often by half.

This is why the same wall count can produce a strong part and a weak part. Four perimeters on a vertical wall is four continuous strands resisting the load. Four perimeters on a horizontal plate is four bonded interfaces resisting delamination. The number is identical; the mechanics are not.

Wall orientation	Strands relative to the load	Effective strength	Practical rule
Vertical wall, load in the wall plane	Parallel	High — governed by strand tensile strength	Wall count is the real lever
Vertical wall, load pushing the wall sideways	Perpendicular in bending	Moderate — bending stiffness plus shell	Add ribs rather than walls
Horizontal plate, load pulling it apart vertically	Across the bond	Low — governed by layer adhesion	Reorient, or add mechanical interlocking
Wall whose thickness is vertical	Across the bond, in thin sections	Low	Reorient, or over-thicken the wall

The corollary: no slicing decision survives a bad orientation. If the load must cross layers, wall count and infill density become secondary, and the honest fix is a geometry change — the subject of the section geometry discussion on fillets, ribs and bosses.

Where the Solid Regions Should Actually Live

The productive way to think about a printed part is not "how much infill" but "which regions must be solid". Load concentrates at features, not in the middle of panels, so the solid plastic belongs exactly where fasteners, pivots and joints are.

• Around every screw hole. At least one full closed ring of solid material, four to six perimeters, extending roughly 1.5 times the hole diameter from the edge. Threads and inserts need material around them that is not a sparse lattice.
• At bosses and standoffs. Three perimeter lines minimum all the way up, and a solid base. Infill inside a boss contributes almost nothing because the load path is circumferential.
• In the first and last few millimetres of any loaded member. Buttons the sparse interior in, and prevents the shell from peeling away at the ends.
• Under any bearing or sliding surface. A solid region resists local crushing; a sparse lattice deflects under a point load and wears unevenly.
• Where the part attaches to another part. Bolt flanges, hinge leaves and press-fit bores should be locally solid even if the rest of the part is not.

Everything between those regions can be much lighter than instinct suggests. Panels, covers, standoff shafts and cosmetic volumes are almost never the failure site, and printing them hollow saves both filament and print time for no cost in strength. That is the real trade worth making, and it is the opposite direction from turning the global infill dial up.

Choosing the Numbers for a Real Part

The decision sequence that works reliably, in order:

1. Identify the load path. Trace where force enters and leaves the part. Only those routes need thickness.
2. Set the load-bearing walls from the strand direction. If the strands run along the load, four perimeters is a good default; if the load crosses layers, fix the orientation before touching anything else.
3. Set structural regions to solid, non-structural regions to 15 to 25 %. Use per-object or modifier settings rather than a global value.
4. Check the wall thickness trap. Anywhere the CAD wall sits between 1.7 and 2.4 mm on a 0.4 mm nozzle, either thin it or thicken it — do not leave it ambiguous.
5. Add top and bottom solid layers generously. Four to six is reasonable; they stiffen the shell where loads enter and exits.
6. Only then consider raising infill, and only in regions where the load is compressive.
7. Verify in the slicer preview, in cross-section, that the regions you designed to be solid actually are.

One filament-side note matters here. A part with thin walls and small extrusion widths puts more demand on extrusion consistency than a part with thick solid regions, because pressure variation shows up directly as gaps between adjacent strands. Running a spool whose diameter holds its tolerance makes the difference between a wall that is four bonded lines and a wall that is four lines and three pinholes. This is also the setting where a flow and wall-order calibration pays for itself immediately.

Material Choice Changes the Wall Budget

The correct perimeter count is not material-independent, because ductile materials tolerate a small amount of internal imperfection while brittle ones do not.

• PETG is a forgiving default: good layer adhesion and enough ductility to redistribute load around a small void. It will let you get away with three perimeters where a brittle material will not.
• ABS and ASA are reasonably tough in bulk but weak across layers, so wall count helps only when the strands align with the load. They reward solid local regions around fasteners.
• PLA is stiff and strong in the strand direction but has very little capacity to deform, so internal voids become crack initiation sites rather than harmless imperfections. If you print structural parts in PLA, treat the wall thickness trap as a hard constraint, not a preference.
• Fibre-filled blends such as PETG-CF and PA6-CF are stiffer and print cleanly at consistent widths, which makes thin walls more predictable, but they are less forgiving in peel. Keep the load path in-plane and run a hardened or tungsten nozzle so the wall width does not drift as the abrasive filament wears the orifice.
• Flexible filament follows the same principle with a twist: the shell dominates almost completely, because a sparse flexible core deforms with little resistance. Bending stiffness in TPU is essentially a wall-count story.
• Resin parts have no layer direction to fight, so wall thickness translates more directly into strength. Thin resin sections are brittle, and the tougher impact-resistant resin grades are the right starting point for anything with a wall under two millimetres.

In all cases, dryness is not optional. Wet filament produces micro-voids along the extrusion, which is precisely the failure mode a thick, well-bonded shell is supposed to prevent — a filament dryer addresses it at the source rather than trying to compensate in the slicer.

What Wall Count Cannot Fix

Being precise about the limits keeps this from becoming another "just add perimeters" article:

• Sustained load and creep. More shell does not stop a plastic part from deforming slowly under continuous load. That is a time-dependent failure with its own remedies.
• Load crossing layers. Orientation, not wall count. If the strands run perpendicular to the load, extra perimeters only add more weakly bonded interfaces.
• A section that is simply too thin. Below a certain absolute dimension there is no wall count that produces useful strength; the design needs more section, not more lines.
• Dimensional accuracy requirements. Thick walls and dense infill increase shrinkage and warp, and can make tolerances harder to hold, not easier.
• Fatigue at a stress raiser. A sharp corner will still initiate a crack under cyclic load regardless of how many perimeters surround it.

The Pre-Print Checklist

1. Load paths are identified, and only those regions are set to solid or high wall count.
2. Every load-bearing wall has at least three perimeters, and preferably four to five.
3. Fastener holes, bosses and bearing surfaces have solid 360-degree material around them, not sparse infill.
4. No designed wall sits in the 1.7 to 2.4 mm ambiguous band on a 0.4 mm nozzle.
5. Wall strands run along the load wherever the geometry allows; where they cannot, orientation is reconsidered rather than patched.
6. Infill is set per region, with non-structural volumes at 15 to 25 % rather than a global value.
7. Top and bottom solid layers are four or more to button the sparse core.
8. Infill density was not used as the primary strength adjustment.
9. The slicer preview has been inspected in cross-section at a thin wall and at a boss.
10. Filament is dry, extrusion width is consistent, and the nozzle is suited to the material's abrasiveness.

None of this requires a new printer or a more exotic filament. It requires accepting that a printed part is a laminated shell around a mostly empty volume, and that the shell is what carries the load. Once that picture is in place, the slicer stops being a dial to turn up and starts being a way of placing material exactly where the stress is — which is also, conveniently, the cheaper way to print.