How are you handling GD&T on additively manufactured parts?

Anders NairPhysical Metallurgy Engineering · 14 yrs

We kept printed datums on a mid-volume titanium bracket programme, and the thing that made it work was not a tolerancing trick — it was fixing the build orientation in the part definition rather than in a drawing note. Once orientation is a controlled characteristic, an as-built upskin face becomes repeatable enough to serve as a secondary or tertiary datum, because every part sees the same thermal history at that face. The primary datum still gets machined, because the primary carries the inspection repeatability for everything downstream and you do not want that riding on surface texture. But two machined datums out of three becoming one out of three took a real cost out of the part.

The inspection side is where the scheme earns or loses you money. Touch probing an as-built additive surface gives you scatter that tracks the surface texture, not the form, so we moved to structured-light scanning with a best-fit datum established from a defined surface region rather than three discrete points. ASME Y14.5 lets you do this through a datum target area, and calling out an area target on the upskin face rather than point targets is what makes the measurement reproducible between the supplier's CMM and ours. Point targets on a printed face will give you two different answers in two different labs and a supplier argument that nobody wins.

Ananya WatanabeMetallurgical Failure Analysis · 25 yrs

Verified expertise in Metallurgical Failure Analysis

Your read on Y14.5 is right but the gap is narrower than it looks — profile tolerance is the correct tool and it was not written with an assumption about how the surface got there. The place additive genuinely breaks the standard is surface texture, because Y14.5 and the texture standards behind it assume a directional lay from a tool path. A printed surface has no lay in that sense, and an Ra callout on a downskin face is close to meaningless as an acceptance criterion. We replaced texture callouts on non-critical printed faces with an areal parameter and a reference sample, and referenced the sample in the drawing rather than trying to force a profile-based number to carry the meaning.

On profile versus size-and-position, one caution: profile controls form, orientation and location together, so if you apply a single profile tolerance to a printed feature you are also implicitly tightening its form to that value. On additive parts the form error is often the dominant term, which means a profile tolerance sized for the location requirement will fail on form even when the part is perfectly usable. Composite profile solves this properly — a looser upper segment for location, a tighter lower segment for form within the feature. It is the same mechanism you would use on a sheet-metal weldment and it maps onto additive cleanly.

Ananya OkaforRotating Equipment Engineering · 15 yrs

Slightly different angle: we stopped trying to make one drawing serve both the build and the inspection. The build gets a model with orientation, support strategy and machining allowance baked in, and the inspection drawing dimensions the finished part only. Trying to tolerance the as-printed condition and the finished condition on one sheet is where most of the confusion in our shop came from, because the same nominal surface meant two different things depending on which operation you were reading for.

That split also fixes the cost conversation. Once the machining allowance is a build parameter rather than a tolerance consequence, you can ask the direct question — what does this datum cost per part in secondary operations, and what does it buy in inspection repeatability — and answer it with numbers. On our bracket, two datums were worth the machining and the third was not, and we would never have been able to demonstrate that while the allowance was implicit in a tolerance stack.