Printed spare parts are not legally the same part
The spare parts argument for additive manufacturing is the most persuasive one in the field and the most frequently oversold. Yes, a scanned impeller can be printed in days rather than bought in months. No, the result is not the part that was in the drawing, and in a regulated or warranted installation that difference is the whole subject rather than a detail to be sorted out later.
Start with what is genuinely true. For obsolete equipment, for one-off geometry, for anything where the original supplier has left the market, printing a replacement is often the only route that exists. Shops doing this well have saved installations that would otherwise have been scrapped. The case does not need inflating.
The inflation happens one step later, when the demonstration part becomes a plan to print spares generally. At that point the question stops being whether the geometry can be reproduced and becomes whether the organisation can state, on paper, what the replacement is made of, by what process, to what tolerance, verified how, and against whose authority. Those five statements are what a purchased spare part silently carries. A printed one carries none of them unless somebody builds the record.
Terminology is worth getting right before the argument starts, which is what ISO/ASTM 52900 is for. It fixes the vocabulary of additive manufacturing, and precise vocabulary is what keeps a discussion about process qualification from sliding into a discussion about printers. The process is the thing being qualified. The printer is one input to it, alongside the powder, the parameter set, the build layout, the thermal treatment and the inspection method. Change any of them and the qualification question reopens, which is unwelcome news for anybody who imagined a part file as the deliverable.
The deliverable is a package. It contains the geometry, but also the material specification and its evidence, the parameter set with the machine it was developed on, the post-processing route, the inspection plan with acceptance criteria, and a statement of who is responsible for the part's fitness. In sectors with formal airworthiness, pressure or medical regimes, that last statement has a legal addressee and an approval process behind it. In unregulated industrial maintenance it still exists, informally, as the question of who gets asked why the pump failed.
Warranty is where this bites first and most quietly. Fitting a part the original equipment manufacturer did not supply, into an installation still under warranty, is a contractual act. Sometimes it is permitted, sometimes it voids cover on the subsystem, occasionally it voids more than that. The clause exists in the supply agreement and can be read in an afternoon. It is read, in practice, after the failure.
Reverse engineering adds a second layer that has nothing to do with manufacturing. A scanned part reproduces the geometry of an object, not the intent of a drawing, and the drawing may be somebody's intellectual property. Scanning a component to restore an out-of-production machine and scanning a component that is currently on sale are different acts with different legal exposure, and the fact that a scanner cannot tell them apart is not a defence.
There is also a measurement problem that the enthusiasm tends to skip. A purchased part comes with an implied statement that it was made the way parts of that kind are made, with the internal structure that implies. A printed part's internal condition is a function of the build, and the inspection that would confirm it — porosity, internal geometry, residual stress — is not something a calliper reaches. Deciding what level of inspection the application actually warrants is engineering judgement, and recording that judgement is the part of the file that auditors read.
A workable position
The shops that make this work do not treat printed spares as a general capability. They maintain a short list of parts for which the package exists, built deliberately, part by part, usually starting with items that are non-critical, geometrically simple and expensive to source. The list grows by addition rather than by policy. Everything not on the list is still bought.
That is a less exciting proposition than a printer that replaces the stores, and it is the one that survives contact with an insurer. The technology is ready for far more than most organisations have the documentation to support, and the gap between those two facts is where the disappointments live.