Lattices, channels, and parts you cannot mold easily

01

Start from the honest advantage list

Layer-based manufacturing changes one economic fact: shape complexity no longer requires tool access. A cutting tool must reach every surface it makes; a mold must release its part. Deposition and fusion processes carry no such obligation, so re-entrant forms, enclosed voids, and graded structures become buildable. ISO/ASTM 52910 catalogs these design opportunities alongside their constraints, treating them as decisions with consequences rather than free upgrades ¹. That framing matters, because every freedom on the list creates a downstream obligation somewhere else in the process chain.

Research literature on design rules groups the freedoms into four broad kinds of complexity: shape, hierarchical, material, and functional ². What none of them change is physics. Minimum printable features still exist, surfaces still carry process texture, properties still vary with direction and parameters, and anything hidden inside a part must eventually be cleaned, verified, or trusted blind. The advantage list is honest only when it is read next to the obligation list.

• Shape complexity: re-entrant, hollow, and freeform geometry
• Hierarchical complexity: features spanning coarse to fine scales
• Material complexity: graded or multi-material construction
• Functional complexity: mechanisms and consolidated assemblies
• Every freedom adds a cleaning, verification, or qualification duty

02

Lattices: mass removed, questions added

A lattice replaces solid volume with a repeating network of struts or surfaces. Done for a reason, it removes mass from low-stress regions, tunes stiffness, absorbs energy progressively, or multiplies surface area for heat or fluid exchange. In powder-bed processes, lattice shells also print without supports because surrounding powder carries them, which is why nested lattice forms are a signature of that process family. The unit cell is a design variable: its type, size, and strut thickness set the behavior, and the transition regions where lattice meets solid skin usually govern failure.

The honest boundary is minimum feature size. Design-rule research exists precisely because processes impose numeric limits on thin walls, small holes, and slender features, and those limits move with process, material, machine, and orientation ². A strut near the capability floor prints as a chain of near-defects; a lattice of such struts is structurally worse than the solid it replaced and much harder to inspect. Where margins are generous and mass is cheap, a lattice is often decoration. Decoration is allowed. It should simply be labeled as such.

03

Channels and conformal features

Internal channels are the second headline freedom: passages that follow function rather than the straight lines a drill demands. Cooling passages can track a surface at constant offset, manifolds can merge smoothly, and flow paths can avoid dead corners. The constraint arrives immediately: whatever filled the channel during building must leave it afterward. Powder needs escape openings and a pouring route, resin needs drainage and wash access, and support structures inside a closed passage may be impossible to remove at all, a scenario design guidance explicitly warns against ¹.

Channel design is therefore exit design. Cross-sections that support themselves, such as teardrop or diamond profiles in fusion processes, avoid internal supports entirely. Diameter budgets grow with length because process texture, particle adhesion, and access limits accumulate; a long narrow passage that cannot be flushed, gauged, or imaged is a promise nobody can check. A lattice or channel that cannot be cleaned is a filter you did not mean to make. Decide the cleaning method and the verification method when the passage is drawn, not after it is trapped.

• Design escape openings and drainage before styling the channel
• Prefer self-supporting cross-sections in fusion processes
• Grow diameter allowances as passage length increases
• Plan flushing, gauging, or imaging access from the start
• Never place removable supports where tools cannot reach

04

Consolidation: fewer parts, heavier bookkeeping

Part consolidation merges an assembly into one built piece. The wins are real and sometimes decisive: fasteners disappear, joints and leak paths vanish, tolerance stacks collapse into single features, and assembly labor drops toward zero. Design-for-additive literature treats consolidation as a core strategy with its own published guidelines, not an afterthought ². Brackets plus spacers plus clips becoming one bracket is the everyday version; fluid manifolds losing every gasketed interface is the industrial one.

The costs are administrative as much as mechanical. A consolidated part carries every function of its ancestors, so one local defect condemns the whole piece, and repair becomes replacement. One build orientation must now serve all merged sub-shapes, which forces compromises a multi-part assembly never faced. Inspection scope expands to every merged feature, including internal ones that used to be open faces on separate parts. And when a single interface needs a tolerance or finish the process cannot deliver, machining allowance or a deliberate split must return. Consolidate where interfaces are the enemy; stay modular where service, tolerance, or inspection is.

05

Internal geometry must answer to inspection

These processes can build what no one can see, and acceptance does not run on faith. NIST's part-qualification work exists because complex internal features, rough as-built surfaces, and process variability make additive parts genuinely hard to verify with conventional dimensional and nondestructive methods ³. A conformal channel or embedded lattice moves inspection from calipers to flow tests, witness coupons, sectioned samples, or computed tomography, each with its own cost, resolution floor, and honest limits.

The discipline is to name the acceptance method while the feature is still a sketch. If a channel's function is flow, a flow or pressure test may accept it cheaply. If a lattice's function is stiffness, a proof load on the part or a sectioned sibling may serve. If nothing can verify the feature at a cost the part justifies, the design claim must shrink to what can be verified, or the geometry should be redesigned toward inspectability. Unverifiable performance is not performance; it is a story attached to a part.

06

Decide with capability data, not catalog promises

Generic minimums from marketing pages are not your machine's minimums. ISO/ASTM 52902 defines test artifact geometries, including fine features, holes, and slots, specifically so a system's geometric capability can be measured rather than assumed ⁴. Building and measuring such artifacts in the actual material and parameter set tells you the smallest strut that prints cleanly, the narrowest channel that clears, and how capability drifts across the build area. That data converts lattice and channel design from hope into engineering.

Then decide the boring way. State the requirement each geometric feature serves; strike any candidate that violates escape, access, or acceptance; prefer the simplest geometry that meets the requirement, because simple geometry cleans faster, inspects cheaper, and fails more legibly. Record the measured capability, the chosen geometry, and the reason, and re-measure after meaningful changes to machine, material, or parameters. The parts this notebook celebrates are not the most intricate ones; they are the ones whose intricacy can be justified in one sentence and verified by one named method.

• Measure minimum struts, walls, holes, and channels on your own system
• Check capability across the build area, not one lucky corner
• Let the smallest reliable feature govern the design
• Re-verify after material, machine, or parameter changes
• Record requirement, geometry, and acceptance method together