Most parts that fail in 3D printing were not badly printed. They were badly drawn for printing. A model that is perfectly valid for machining or injection moulding can be unprintable, fragile, or dimensionally wrong once it comes off the build plate — not because the printer failed, but because the geometry asked for something the process cannot do.
This is a practical guide to the design decisions that determine whether a part prints well, aimed at engineers and designers sending a model out for the first time.
Wall thickness is the first thing to get right
Every process has a minimum feature size below which geometry either disappears or becomes too fragile to survive handling. On extrusion-based processes that floor is set by the nozzle: a wall thinner than a small multiple of the extrusion width cannot be built reliably, and a single-track wall has very little strength across the layer direction.
Two practical consequences. First, thin ribs and fins you would happily machine may need thickening for printing. Second — and more often missed — scaling a model down scales its wall thickness down too. A design that prints cleanly at full size can fall below the process floor at 60%, and nothing in the model warns you.
Keep walls reasonably consistent as well. Large jumps in thickness cool at different rates and are a common source of warping and internal stress, particularly on larger polymer parts.
Orientation decides strength, finish and cost
A printed part is not isotropic. Layer-based processes build up in one direction, and the bond between layers is generally weaker than the material within a layer. That means the same geometry has different strength depending on how it is oriented on the build plate.
If a part has a known load direction — a bracket that takes a bending load, a clip that flexes — say so. A note on the drawing or in the order that identifies the critical load path lets the part be oriented so that load is not trying to peel layers apart.
Orientation also drives surface finish and support. Downward-facing surfaces and steep overhangs need support material, and the faces where support was attached will be rougher after removal. Where one face has to be cosmetic or a sealing surface, flag it, because that constraint changes the setup.
Overhangs, bridges and the shapes that print themselves
Unsupported overhangs beyond roughly forty-five degrees from vertical begin to need support on most extrusion processes. You can often design the need away:
- Chamfer the underside of a protruding feature rather than leaving it square
- Replace a horizontal round hole with a teardrop or diamond profile so its top self-supports
- Break a tall thin part into sections that print flat and assemble, rather than printing it upright with a scaffold
- Add a small fillet where a vertical wall meets the plate to reduce the chance of a corner lifting
Every support you design out is cost and post-processing removed, and one less surface damaged on removal.
Tolerances and fits are not machining tolerances
Additive processes are dimensionally capable but not to machining precision, and the achievable tolerance varies by process, material, part size and orientation. Holes in particular tend to finish undersized, because the perimeter is laid down with a finite track width and the material shrinks as it cools.
Where a fit genuinely matters — a bearing seat, a dowel, a threaded insert — the reliable approach is to print undersized and finish by machining or reaming, or to design in a clearance generous enough that process variation does not consume it. Expecting a press fit straight off the printer is the single most common cause of disappointment on a first part.
For functional assemblies, moulded-in threads are usually better replaced with heat-set inserts or a tapped boss designed with enough wall around it to take the load.
Sending the right file
File format is where a lot of avoidable information gets thrown away:
- STL — a triangulated mesh and still the common currency. It carries geometry only: no units, no colour, no tolerances. Export at a fine enough resolution that curved surfaces are not faceted, but not so fine that the file becomes unwieldy.
- STEP — true solid geometry rather than a mesh. Preferred where the part may need modifying, repairing or machining after printing, because the real surfaces survive.
- 3MF — a modern container that can carry units, colour and other metadata alongside the mesh, which removes a whole class of scaling confusion.
Whatever the format, send a drawing or a note with anything that the geometry cannot express: critical dimensions, which faces are cosmetic, the load direction, and the intended material. Terminology for the processes themselves is standardised internationally, and research on additive process capability is published by bodies including the National Institute of Standards and Technology; Australian adoptions of the relevant standards are available through Standards Australia.
A short pre-flight checklist
- Are all walls above the process minimum — at the size you are actually printing?
- Is the load direction known and stated?
- Have obvious overhangs been chamfered or reoriented away?
- Do any fits need post-machining, and is that allowed for in the model?
- Is the file in a format that carries what the printer needs to know?
If you would like a model checked against these before it goes to print, send it through our 3D printing service or talk to us about the part. Where a physical part already exists and needs turning into a model first, that is a job for 3D scanning.