Automotive staged-pilot and crash-validation records are not substantiating data for the aerospace certification of a generative-design part. US approvals attach to a design or a modification, through type certificates and supplemental type certificates. A Designated Engineering Representative (DER) approves the substantiating data for a specific part, and the authority keeps the certification basis and the design approval itself. An airworthiness finding rests on substantiating data for the specific design.

That way the package a DER reviews is assembled from records instead of reconstructed afterwards. It is not certification or regulatory sign-off, which stay with the applicant, the DER and the authority.

Certification attaches to the part, not to the generative-design tool

EASA’s Certification Memorandum CM-S-008 (Issue 4, September 2025) states that every aviation product, part and appliance must meet the certification specifications of its type certification basis, whatever material, process or fabrication method produced its engineering properties. In the US, approval comes through a type certificate, an amended type certificate or a supplemental type certificate, each tied to a particular design or modification.

The method still has to earn trust. In an interview study of 20 aerospace additive manufacturing practitioners published in Design Science, one respondent stressed that the topology optimisation process itself has to be qualified, not only the material and the print process. A vendor saying its software is “aerospace certified” therefore answers the wrong question. The useful questions are which part, on which programme, under which authority and against which basis.

What a DER signs, and what stays with the FAA

A DER approval covers substantiating or descriptive data; it is not a design approval under 14 CFR part 21. Structural DERs may approve engineering reports, drawings, material and process specifications and other structural data, within the limits of their appointment. The FAA does not delegate setting the certification basis, special conditions, exemptions, equivalent-level-of-safety findings, or issuing type certificates, STCs or PMAs.

Two limits matter most for generative design. A DER is not authorised to interpret regulations and must consult the FAA project office when interpretation is in question, including for new or unconventional materials and processes. An applicant proposing new or unproven technology must also coordinate with that office to find out whether delegating the compliance finding is appropriate at all. The DER records findings on FAA Form 8110-3, which is the DER’s only means of approving technical data. The form must not be signed until all substantiating data is available, understood and listed on it.

EASA also ties requirements to each part’s type certification basis. It also says no standardised approach exists yet, so agreement with the competent authority is needed until one does. Even no-criticality “D” parts handled as minor changes come with an expectation that the design organisation tells EASA it intends to use additive manufacturing and supplies an impact assessment.

What’s different about validating a generative-design part for aerospace versus automotive?

The difference lies in what the evidence has to prove, and who has to accept it. Under 14 CFR 25.307, strength and deformation compliance must be shown for each critical loading condition. Analysis may be used only where experience shows it to be reliable for that kind of structure. Otherwise substantiating tests are required. FAA AC 25.307-1 says compliance can be shown by analysis supported by previous test evidence, by analysis supported by new test evidence, or by test alone. It also says finite element methods on complex structure count as reliable only when validated by full-scale tests.

The advisory circular defines new structure as structure with significantly different details, geometry, load paths or materials from previously tested designs. For new structure, substantiation up to limit and ultimate loads normally requires testing of sub-components, full-scale components or assembled components. The AIA’s 2025 guidance on certifying additive components describes the same climb as a test pyramid. It runs from coupon and design-detail characterisation through single parts and sub-assemblies, then components, up to full air-vehicle validation.

Compare the automotive precedent. In 2018, GM and Autodesk described using generative design on a seat bracket in an initial proof-of-concept project, describing design exploration and a compressed development process. Nothing in it is a substantiation record an airworthiness reviewer could accept. The staged sequence of simulation, shadow mode, limited pilot and staged rollout is better suited to autonomous machines such as vehicles on public roads.

Section 25.571 requires evaluating all structure that could contribute to catastrophic failure for fatigue, corrosion and accidental damage, and setting inspections to prevent that failure through the aircraft’s operational life. For small aeroplanes, FAA AC 23-13A warns that a design change that looks minor from a static-strength standpoint can have a major effect on fatigue. It also notes that a one percent increase in stress level can shorten expected fatigue life by more than five percent. Static and stiffness load cases alone are therefore the wrong objective set for an aerospace part. A vendor blog on generative design validation makes the matching point from automotive and robotics: the failure mode is the load case nobody modelled, such as vibration fatigue on a bracket that passed its static load.

Why a staged pilot or crash test does not count as airworthiness evidence

A staged pilot builds up operating experience across a fleet or a site. AC 25.571-1D allows service experience to inform fatigue evaluations, and the applicant identifies the intended approach in a compliance plan.

Before witnessing a test, a DER must check that FAA conformity inspections are done and that the test article conforms. When DERs approve test data, they state that the test followed the FAA-approved test plan. If a single-load-path structure is substantiated by test alone, the test loads must be raised to account for material variability. Fatigue life carries its own margin: AC 23-13A defines safe-life as the mean test life divided by a scatter factor. If test results do not match the analysis, the reasons must be investigated whether or not the article failed below ultimate load, and that can mean redesign and more testing.

Autodesk Research reports that the generatively designed Airbus bionic partition was undergoing 16G crash testing as part of its certification for the A320 fleet.

Checklist: can an existing validation record be reused?

Before assuming any earlier record carries over, check it against what an aerospace finding needs:

• Certification basis named: the record shows compliance against a specific basis and amendment level, which an applicant has to identify.
• Every critical load condition covered: static strength for each critical loading condition, plus fatigue, corrosion and accidental-damage evaluation.
• Conformed test articles under an approved plan: with conformity inspection before witnessing.
• Statistically based design values: derived from specimens made with production processes.
• Test and analysis correlated: with discrepancies investigated.
• Configuration traceable: each component traced to its drawings, CAD revisions and build records.

A record that misses any one of these is development evidence, not substantiation.

Material allowables: A-basis, B-basis and where S-basis falls short

The FAA’s Part 23 guidance is explicit on design values. For single-load-path structure, 14 CFR 23.613 requires design values at 99 percent probability and 95 percent confidence (A-basis). For multiple-load-path structure, it requires 90 percent probability and 95 percent confidence (B-basis). The same text says S-basis values have unknown statistical assurance and do not meet the probability requirements of 23.613 at amendment 23-45. It adds that no statistically determined alternative to A-basis or B-basis is allowed. For projects at amendment 23-64 and later, applicants are told to build FAA-accepted means of compliance into their certification planning instead of seeking an equivalent-level-of-safety finding. On the transport side, AC 25.613-1 calls for statistically determined design values based on enough testing, using specimens made with production processes and tested to accepted standards such as ASTM, EN or ISO.

For Class C and D parts, S-basis data from MMPDS or CMH-17 may support proportionate means of compliance, with additional factors needed because coupon data does not represent a complex configuration. The texts cover different scopes: the FAA text addresses 23.613 design values, while the EASA allowance is limited to Class C and D parts.

The AIA guidance says the applicability of bulk material allowables must be demonstrated for each individual component. It says allowables have defined limits, such as machine build parameters and surface condition. Design values must account for geometric feature, location in the part and post-processing. The guidance also notes that extracting coupons from complex or thin-walled geometries can be difficult, and may require purpose-built coupons and test methods. Thin, organic members are exactly what topology optimisation produces, so the allowables question for a generative part usually turns into a coupon-design question early on. The Design Science study adds that a lack of AM material data for simulation tools makes simulation-based design support too vague to predict robustness reliably.

Criticality decides how far substantiation has to go

Issue 4 of CM-S-008 makes certification effort proportionate to novelty, criticality and complexity. Its policy scope is limited to new parts or repair by replacement of no- and low-criticality parts. Under the AIA guidance, a part’s criticality class comes out of a failure modes, effects and criticality analysis covering fatigue, crack growth and residual strength. Under the FAA’s Part 23 safety emphasis items, proposals to use additive manufacturing for principal structural elements, primary structure or structure whose failure may prevent continued safe flight and landing must be coordinated with the Policy and Standards Division to determine the level of FAA involvement. EASA expects certifying higher-criticality AM parts (Classification A and B) to be a particular challenge in the near future.

So the practical entry point is secondary structure, a view one generative-design vendor also gives: brackets, mounts, fittings and interior parts face lower certification barriers than spars or frames.

Has any generative-design tool produced a part that passed DER review or Part 25/Part 23 structural substantiation?

The Airbus bionic partition is a generatively designed part that has entered certification testing. Airbus described it in 2016 as a generative-design component 45 percent (30 kg) lighter than current designs. Autodesk Research describes it as still undergoing 16G crash testing as part of certification, without stating that certification was completed.

Printed parts, on the other hand, do have certification precedents. GE reported that the T25 sensor housing was the first 3D-printed part certified by the FAA to fly inside its commercial jet engines. Liebherr-Aerospace reports that its A350 nose landing gear sensor bracket was the first Airbus system part qualified for titanium additive layer manufacturing. A Lufthansa Technik A-Link for the IAE V2500 was reported as the first EASA certificate for an additively manufactured load-bearing spare part. Norsk Titanium and Boeing completed a testing programme with FAA certification deliverables for structural parts on the 787. None of these announcements presents its geometry as generatively designed. They are manufacturing precedents, not precedents for the design method.

Missing public evidence does not prove the work cannot be done. GAO reported FAA staff noting that, under the performance-based Part 23 rules, applicants’ own means of compliance are proprietary and not made public. If a tool vendor claims a certified part, the reasonable ask is the part number, the programme, the authority and the test reports behind it. If the vendor has none, they should say so plainly instead of implying the software is certified.

How is an optimised geometry traced back to certified static and fatigue test reports?

The sources describe the links such a trace needs: configuration-controlled records, conformed testing, and a Form 8110-3 signed only once all substantiating data is listed. In order, the links are:

1. Analysis inputs under version control. NASA’s guide to certification by analysis lists documentation of the analysis configuration, model inputs and version control as the first element for flight modelling analysis.
2. Geometry clean-up recorded as revisions. Raw topology output can carry imperfections from mesh discretisation, and some organisations simplify it into “two-and-a-half D” shapes that look machinable.
3. Digital control plan. The AIA guidance requires CAD file names and revisions, slicing and build parameter files, and software and firmware revision levels to be tracked. The build definition and machine parameters must be traceably linked to each part.
4. Fixed process. Key process variables are fixed once qualified, and changing one triggers requalification. Software or firmware updates that could affect them trigger machine requalification.
5. Records package. Each AM component must trace to drawings, CAD revisions, PCD revisions, feedstock lot, build reports, inspection records and machine qualification records.
6. Conformed test and correlation. The test article is conformity-inspected before a witnessed test, and the review covers loading representativeness, instrumentation and comparison of results with analytical predictions.
7. Form 8110-3. Signed only once all the substantiating data is available and listed.

One respondent in the Design Science study called fragmented workflows the “Achilles’ heel” of generative design, because every data transition “costs traceability and data quality”. The solver version, mesh settings, and the hardware and libraries an optimisation ran on belong in the same configuration record as the CAD revision. A rerun that cannot reproduce the geometry breaks link one, and every link after it inherits the break.

What should a certification-aware study cover before a generative-design tool is trusted for aerospace parts?

A certification-aware study fixes the certification basis and means of compliance first, and only then tests whether the tool’s outputs can be substantiated under them. In design verification and validation terms, as the Design Science study puts it (citing America Makes, 2023), verification confirms that the specified requirements are met and validation confirms fitness for the intended use. Both need objective evidence. The study should cover:

• Means of compliance, settled early. One interview respondent said the means of compliance “should be pretty much locked in” when judging whether a part suits AM. Under the performance-based Part 23 rules, FAA staff must review and accept each proposed means of compliance, and GAO’s report GAO-21-85 found staff asking for more guidance on how to do so.
• Criticality class and the load cases that follow from it. Fatigue spectra and damage-tolerance cases belong in the optimisation as constraints, not as checks run after the geometry exists.
• Qualification of the optimisation method itself. It has to be qualified alongside the material and the process.
• Allowables coverage. Confirm that the part’s features fall inside the limits of the available design values.
• Inspectability. Designing for inspectability is described as a critical area, and building in inspection access costs strength and mass.
• Analysis credibility. NASA’s guide names establishing confidence, through verification, validation and uncertainty quantification, as one of the top impediments to certification by analysis. It notes that individual applications of certification by analysis have been accepted case by case.
• The test-method gap. The FAA’s 2022 REDAC briefing says AM certification can use a building-block approach, but standards for those test methods do not yet exist.

The study produces artefacts: a basis and compliance plan, an optimisation set-up whose constraints match it, a reproducible trace, and a test plan sized to the part’s criticality. TechnoLynx treats those artefacts as an engineering deliverable that a reviewer can read and sign against, not as a policy document. A study can also conclude that a given tool is fit only for low-criticality brackets on today’s evidence. That is a useful answer, because it tells a programme where generative design can save weight now and where it still needs a qualification campaign nobody has published.

FAQ

What is Part 23 in aviation?

Part 23 is the set of FAA airworthiness rules for small aeroplanes: those weighing 19,000 pounds or less and carrying 19 passengers or less. In December 2016 the FAA replaced its prescriptive design requirements with performance-based regulations.

What is the main purpose of generative design in aerospace engineering?

In the Airbus partition the purpose was cutting weight while staying strong, because less weight means less fuel burned. Airbus describes generative design as generating thousands of alternatives that meet set goals and constraints, with additive manufacturing needed to build them.

Who is responsible for issuing airworthiness directives?

In the US, the FAA issues airworthiness directives. AC 25.571-1D notes that maintenance actions specified by the type certificate holder after certification may be made mandatory by FAA airworthiness directives.