Design rules for metal 3D printing: the table you should have on hand

Almost everything that goes wrong with a metal printed part was decided in the CAD software, weeks before the machine was even turned on. And almost always for the same reason: a part was designed with a milling machine in mind and then sent to be printed.

This is the reference you should keep open while drawing. None of these numbers are set in stone: they depend on the alloy, the machine, the layer thickness, and the parameters, and there are shop floors that exceed them daily because they have optimized their processes to do so. They are the framework within which a part comes out perfectly the first time without any further discussion.

The table

All of this applies to the part in the position in which it will be manufactured, not in the position shown on the drawing. This is the most common misconception: a hole is "vertical" or "horizontal" depending on how it is positioned on the build plate, not how it is shown on the drawing.

Thicknesses: why there is a minimum

A wall cannot be thinner than the width of the laser cut, and below approximately twice that width, the wall lacks the rigidity to support itself during build. The screed passes over each layer, pushing against it, and a wall that is too thin will buckle, drag powder along, or even bring the entire structure down.

There is another, less obvious effect: a thin wall heats up and cools down faster than the material next to it, and this difference creates stress. That's why thin walls warp the most, even if they emerge intact.

The rule of thumb: if a wall is going to bear a load, don't go below one millimeter; if it is purely geometric —a lid, a fairing, a lattice— you can go lower, but accept that it will arrive deformed.

Holes: the vertical one is easy, the horizontal one is not

vertical hole—its axis parallel to the growth direction—comes out easily and without supports, because each layer rests on the one below it. Below approximately 1.5 mm, powder begins to become trapped and sintered in the wall, and below half a millimeter, it simply closes.

A horizontal hole is another story: the top part is an overhang that grows towards the center with nothing underneath. Up to about 6 or 8 millimeters in diameter, the material holds; above that, the upper crown droops and the hole is no longer round.

The solution isn't to add supports inside—they can't be removed—but to change the shape of the hole: give it a teardrop, rhombus, or pointed arch cross-section, so that no area is flat and horizontal. If the hole has to be round, it's already dimensioned, printed smaller, and then drilled.

Internal channels: this is where the parts get lost

This is the rule that is most costly to ignore, and it comes in two parts.

First: the powder has to be able to escape. A channel less than three millimeters wide is very difficult to empty, and the powder that remains inside compacts over time and with vibration. In a cooling circuit, that means a channel that doesn't cool; in an aerospace component, unaccounted-for mass that no one knows is there.

Second: if it cannot be emptied, it cannot be inspected. And an internal cavity that cannot be viewed or measured is a cavity that cannot be certified.

From this come three specific rules: drainage holes at the low points of any circuit, designed from the beginning and not added later; layouts without dead spots, where powder can fall by gravity towards an outlet; and each channel, accessible from the outside in at least two places, so that it is possible to blow from side to side.

It's the same thing that's explained in conformal cooling, and it's the difference between a mold that works for ten years and one that has to be thrown away.

Overhangs and angles

The 45-degree rule is a decent starting point: above that angle to the horizontal, the face stands on its own. Below that, it either needs to be supported or redesigned.

I'm not elaborating on it here because it has its own article, and it's one of the biggest money savers: supports in metal 3D printing, with the ten design modifications that reduce supports and why supports are the most underestimated cost item.

Radii, corners and section changes

No sharp interior edges. A right-angled inside corner concentrates stress during build just as it does in service: that's where cracks appear. A radius of one or two millimeters costs nothing and eliminates a whole problem.

Sudden changes in cross-section are the second leading cause of warping. A part that transitions from a solid block to a thin wall in a single layer has two different cooling rates working against each other within that layer. This is resolved with gradual transitions, chamfers, or ribs—anything that distributes the change across several layers.

Threads, pins and small features

Threads: Below M6, printing the thread is a waste of money. It comes out with an approximate profile, lacking the strength of a machined thread, and you still have to tap it. The sensible thing to do is print the hole with the drill bit size and then tap it. From M8 upwards, a printed thread can be acceptable for non-critical functions.

Pins and thin protrusions: More fragile than they appear, not because of the finished part but because of the stress they endure during build and when supports are removed. Those less than half a millimeter in diameter are likely to be lost.

Text: Intaglio printing is easier to read than relief printing, withstands shot blasting better, and doesn't round off. And place it on a vertical or top surface: it won't be legible on a bottom surface.

Size, height and aspect ratio

Everything that fits within the building volume can be accommodated, but there are two practical limits before that:

Height is time. The cost of an LPBF part depends much more on its height than its volume, because you pay for layers. Laying a part flat can cut the time in half—with the caveat that changing the orientation changes the properties, which is what anisotropy and orientation is all about.

Slenderness vibrates. A part much taller than it is wide receives the force of the squeegee in each layer with an increasingly longer lever arm. Above an eight-to-one ratio, it should be well anchored or split.

The three rules that are worth more than the whole table

If you only take away three things from this article, let them be these.

1. Design for orientation, not for the drawing. Decide how the part will be manufactured before you finish drawing it, and draw it in that orientation. Almost all the rules above depend on it.

2. Mark from the beginning which faces will be machined. And give them extra dimensions. It's much cheaper to leave two-tenths of a millimeter extra on three faces than to discover at the end that the part is at the rough dimensions but with the wrong finish. This is what the related article is about: tolerances and finish.

3. If the part has an interior, resolve the powder exit before the geometry. This is the only rule in this list that, if ignored, cannot be fixed later.

How we verified it

In SamyStudio, thickness, overhang, and accessibility checks are performed on the already oriented part, before generating supports, which is when it's still inexpensive to make changes. And at the open days, the exercise is exactly this: bring in a part, place it on the build plate, and see which rules it breaks.

Which is, by far, the fastest way to learn them.

—

Why some parts shouldn't be printed even if they meet all these rules, in five parts you shouldn't print. What dimensions can be ordered in rough form, in tolerances and finish. And why orientation is part of the specification, in anisotropy and orientation.

And when it's time to export, the settings that do change the part are in prepare the file: STL, 3MF and the mesh.