Industries / Industrial equipment and services / 3D printing
Qualifying parts
Qualifying parts is the work of proving that a printed part is good enough for its job: writing the standards it must meet, measuring and inspecting it, and getting a regulator or customer to accept the evidence.
7
Milestones achieved, from 2011 to 2025.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
2011
Two standards bodies join up
ASTM International formed its additive manufacturing committee, F42, in 2009. ISO formed its own committee, ISO/TC 261, in 2011; it grew out of guidelines on "rapid technologies" written by the German engineers' association VDI, and the German standards body DIN runs its secretariat. In September 2011 ISO and ASTM signed a cooperation agreement to write joint standards instead of two competing sets. The first joint terminology standard, ISO/ASTM 52900, was published on 15 December 2015 and replaced by a 2021 edition, so buyers, makers and regulators now use the same defined words.
LessonTwo bodies chose to share one set of documents rather than split the market between rival rules.
committee.iso.orgcommittee.iso.orgcommittee.iso.orgcommittee.iso.org
Credit
ASTM Committee F42; ISO/TC 261; DIN
2015
Printed parts cleared to fly
In February 2015 the US Federal Aviation Administration (FAA) certified GE's T25 housing, a printed cobalt-chrome case for a compressor inlet temperature sensor. GE called it the "first 3D-printed part certified by the U.S. Federal Aviation Administration (FAA) to fly inside GE commercial jet engines" and planned to fit it to more than 400 GE90-94B engines on Boeing 777s. In February 2017 Norway's Norsk Titanium and Boeing finished a test program to FAA criteria for structural titanium parts on the Boeing 787, built up from melted wire by Norsk's Rapid Plasma Deposition process. Norsk's chief executive said the 787 would become "the first commercial airplane to fly with FAA-approved, 3D printed titanium parts in structural applications."
Credit
GE Aviation; Norsk Titanium and Boeing; FAA
2017
Regulators publish printing guidance
The European Union Aviation Safety Agency (EASA) first issued its Certification Memorandum CM-S-008, "Additive Manufacturing", on 4 April 2017. Issue 04, dated 3 September 2025, sorts printed parts by how much harm their failure could cause and makes the certification effort proportionate to that. The US Food and Drug Administration issued "Technical Considerations for Additive Manufactured Medical Devices" on 5 December 2017, after a draft in May 2016. It tells makers to control where parts sit in the build volume, to validate the process, and to print test coupons (small sample pieces built alongside the part) in the build's worst-case locations.
Credit
EASA; US Food and Drug Administration
2017
NASA rules for flight parts
NASA's Marshall Space Flight Center released MSFC-STD-3716 on 18 October 2017, a framework for metal parts made by laser powder bed fusion for spaceflight. NASA-STD-6030, approved on 21 April 2021, widened this to all printed hardware for crewed spaceflight, with 115 numbered requirements. Each machine must hold a qualified material process record, and parts are sorted into classes A, B and C by risk. The standard says monitoring during the build must be proven like any nondestructive evaluation (NDE, inspecting a part without damaging it, for example by X-ray CT scanning), and that "even a qualified in-situ process monitoring method cannot be considered a complete replacement for NDE."
LessonIn NASA's rules, watching the build adds to inspecting the finished part; it does not replace it.
Credit
NASA Marshall Space Flight Center; NASA Office of the Chief Engineer
2018
Public benchmark data arrives
The US National Institute of Standards and Technology (NIST) completed the first round of its Additive Manufacturing Benchmark Test Series (AM Bench) in 2018, the second in 2022 after a one-year COVID-19 delay, and the third in 2025. Each round publishes tightly controlled measurements of real builds and sets blind challenge problems, where modelers predict results before the data are released. In 2018 none of the 10 groups that predicted the size of laser melt pools came close to the measured values; NIST says those problems were largely solved by 2022. The 2025 round drew 85 challenge submissions, and NIST counted about 1,670 unique-user downloads a month of its datasets between May 2025 and May 2026.
LessonBlind tests showed how far simulations were from real builds, and then measured how fast they improved.
Credit
NIST, with Lyle Levine among the organizers
2023
Build images matched to CT
Oak Ridge National Laboratory's Peregrine software uses AI to flag flaws in each layer of a powder bed printer while the part is being built. Since its disclosure in 2019 it has gained five non-exclusive research licensees, Blue Origin, Cummins, GE Additive, Raytheon Technologies and Ultra Safe Nuclear Corporation, and its licensing work won a 2022 Federal Laboratory Consortium award. ORNL has published open Peregrine datasets since 2021. In September and October 2023 it released builds whose layer images are co-registered (aligned in the same coordinates) with X-ray CT scans of the finished parts: an Arcam Q10 electron-beam build in nickel alloy IN738 (about 28 GB) and a Concept Laser M2 laser build in 316L stainless steel with tensile tests (about 59 GB). An April 2025 release added fatigue tests on 60 of 64 printed cylinders (about 128 GB).
federallabs.orgdoi.ccs.ornl.govdoi.ccs.ornl.govdoi.ccs.ornl.govrdworldonline.com
Credit
Luke Scime, Vincent Paquit and colleagues, ORNL Manufacturing Demonstration Facility
2025
Plastic elbow fails in flight
On 18 March 2025 a kit-built Cozy Mk IV, G-BYLZ, lost engine power on final approach to Gloucestershire Airport in England, landed short and hit the landing-guidance antenna; the pilot had minor injuries and the aircraft was destroyed. The UK Air Accidents Investigation Branch (AAIB) found that a 3D-printed plastic air induction elbow had softened in engine heat and collapsed, choking the engine's air. The owner had bought it at a US airshow and was told it was carbon-fiber ABS with a glass transition temperature (the point where a plastic turns from hard to soft) of 105°C, above the 84°C of the epoxy laminate in the plans. AAIB tests of two samples measured 52.8°C and 54.0°C. The elbow was missing from the parts list sent to the Light Aircraft Association, so it was never assessed, and the association plans an alert on printed parts for all its inspectors.
LessonA seller's datasheet number is not proof; the measured softening point was about half the claimed one.
Credit
UK Air Accidents Investigation Branch (report published 4 December 2025); Light Aircraft Association
Watch for
Qualifying parts by simulation
A steering group of aviation makers, government agencies and universities, Computational Materials for Qualification and Certification (CM4QC), aims to lower the cost of qualifying printed aircraft parts by using computer models of materials as evidence. NIST says the group's strategy document was published in early 2026 by NASA and the FAA. On 13 January 2026 the group and NIST began a joint test for AM Bench 2028: a 7-vane jet engine stator segment from Pratt & Whitney, about 16.5 cm by 9 cm by 7.5 cm, printed in nickel alloy 718 by laser powder bed fusion, with build monitoring, inspection and mechanical tests. Boeing, GE Aerospace, Lockheed Martin, NASA, the Air Force Research Laboratory and ORNL attended the kickoff.
Achieved whenNIST publishes the stator measurements, blind model predictions match them within NIST's stated uncertainty, and the FAA or EASA accepts model results as part of an actual part certification.
Claimed by
CM4QC Steering Group and NIST, January 2026
Printed-metal values in MMPDS
MMPDS is a handbook of allowables (statistically based minimum strengths that designers may rely on) for metals; Battelle compiles its meetings for the FAA's William J. Hughes Technical Center. Its new Volume II, first published in 2024, sets out how data on printed metals must be gathered and analyzed, but the 2024, 2025 and 2026 editions all state that it "does not currently have any material properties." At the March 2026 MMPDS meeting, test plans for printed Ti-6Al-4V titanium, 15-5PH steel and nickel alloy ABD900 were still in progress. One powder batch for the titanium plan broke the doping limits in its material specification and new specimens are being built, and Boeing was waiting for Air Force Research Laboratory approval to release raw data to Battelle.
Achieved whenan MMPDS Volume II edition lists property values for at least one printed alloy.
Claimed by
Boeing, the Air Force Research Laboratory, America Makes, Wichita State's National Institute for Aviation Research (NIAR) and Battelle, March 2026
Open
01
Score build monitors against CT
An open scoring kit and public results table for in-process flaw detectors. It takes a detector's layer-by-layer flaw calls on ORNL's public Peregrine builds and checks them against the pores found in each build's co-registered X-ray CT scan. For each detector and each machine it reports a probability-of-detection curve (the chance of catching a pore of a given size, the usual way inspection methods are rated) and the number of false calls per cubic centimeter.
The briefCredit · unclaimed02
A machine-readable NASA printing checklist
A free, openly licensed, machine-readable version of NASA-STD-6030's 115 numbered requirements (AMR-1 to AMR-115), one record per requirement with the standard's revision stored alongside, plus a checker that follows fixed rules (not an AI model's judgment) and that reads a part's build record (machine, powder batch, settings, witness specimens, CT results, process record) and marks each requirement as shown, missing or not applicable. Its output is a filled draft of the standard's Appendix A compliance matrix with every "shown" item pointing at the file that proves it.
The briefCredit · unclaimed
Open 01 · the brief
Score build monitors against CT
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open scoring kit and public results table for in-process flaw detectors. It takes a detector's layer-by-layer flaw calls on ORNL's public Peregrine builds and checks them against the pores found in each build's co-registered X-ray CT scan. For each detector and each machine it reports a probability-of-detection curve (the chance of catching a pore of a given size, the usual way inspection methods are rated) and the number of false calls per cubic centimeter.
Why it is still open
NASA-STD-6030 requires that a build monitor used to accept parts first show "a proven level of reliability for detection", yet no public, scored comparison of detectors against CT was found in searches of ORNL's dataset portal, NIST AM Bench, GitHub, Hugging Face, Kaggle, Zenodo and arXiv in November 2026. ORNL has published three builds with layer images aligned to CT (an Arcam Q10 IN738 build, a Concept Laser M2 316L build, and a 316H build with fatigue tests); each ships ORNL's own neural-network flaw calls but no scoring kit. ORNL's own January 2024 report plotted its detection hit rate against CT flaw size for four builds on one EOS M290 machine and says Peregrine has internal tools for comparing models between builds and machines; those tools are not public. Makebench, a public benchmark from Flanders Make, ranks models on another task, working out laser settings, and its promised CT and porosity benchmark had not appeared. The closest GitHub code is NIST's exp_and_sim_pod, which computes probability of detection for CT scanning of printed flaws, not for build monitoring.
doi.ccs.ornl.govdoi.ccs.ornl.govdoi.ccs.ornl.govinfo.ornl.govarxiv.orggithub.comstandards.nasa.gov
Built with what exists
The three ORNL Peregrine datasets (downloaded through Globus); Python with NumPy, SciPy and scikit-image to match CT pores to build layers; PyTorch for a baseline detector; NIST's exp_and_sim_pod code for fitting detection curves. No printer or laser is needed.
First step
Download the 28 GB Arcam Q10 IN738 release this month and count how many of the pores in its CT scan overlap a flaw call that ORNL shipped with the data.
Done looks like
A public repository with a written rule for matching a flaw call to a CT pore (distance tolerance, size bins, and which pores near the surface are left out), that scores at least three detectors (ORNL's shipped flaw calls, a simple brightness-threshold baseline, and one trained model) on all three builds, with a detection curve and a false-call rate for each pair, and scripts that let anyone reproduce every number.
Open 02 · the brief
A machine-readable NASA printing checklist
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A free, openly licensed, machine-readable version of NASA-STD-6030's 115 numbered requirements (AMR-1 to AMR-115), one record per requirement with the standard's revision stored alongside, plus a checker that follows fixed rules (not an AI model's judgment) and that reads a part's build record (machine, powder batch, settings, witness specimens, CT results, process record) and marks each requirement as shown, missing or not applicable. Its output is a filled draft of the standard's Appendix A compliance matrix with every "shown" item pointing at the file that proves it.
Why it is still open
The standard's Appendix A matrix is a table in a PDF that people fill in by hand ("Enter Yes or No"), and the 2021 baseline is still the only version on NASA's page, although its five-year review fell due in April 2026. A GitHub search for "NASA-STD-6030" found no repositories, and web searches naming Materialise, Authentise, Oqton, 1factory and Additive Assurance found nothing tied to this standard. The nearest product is StandardFlow, launched on 27 October 2026, an AI platform that checks printing build reports and certificates against technical standards; its launch report does not name NASA-STD-6030 or give a price or license.
Built with what exists
The public NASA-STD-6030 PDF (approved for public release); pdftotext; an AI model such as Claude Fable 5.2 (see the Note "Three working devices") to draft the requirement list, with a person checking each item against the PDF; JSON Schema and Python for the checker; the public ORNL Peregrine 316L dataset, which includes build settings, CT, tensile tests and micrographs, as a sample build record.
First step
Extract the text of Appendix A this month and confirm that the requirement numbers run from AMR-1 to AMR-115 with no gaps.
Done looks like
All 115 requirements captured with their section numbers and checked by a second person with none missing, and a checker run on the Peregrine 316L build that outputs a compliance matrix in which every requirement marked "shown" cites a file in the dataset.