Contacts
Get in touch
Close

Contacts

Romania, Cluj-Napoca – 400157
Str. Ploiești Nr. 9

Romania, Vama – 447350
Str. Fermei Nr. 14

+40 774 463 429

info@inteliscience.net

Automated First Article Inspection (AS9102) for Siemens NX

Automated First Article Inspection (AS9102)

Your PMI already is the characteristic list. Stop re-typing it into Form 3.

Every aerospace supplier knows the ritual. A first article is due — new part, design change, process change, or simply two years since the last production run — and someone sits down with the drawing to do what AS9102 demands: number every single characteristic with a balloon, transcribe every requirement into the Form 3 characteristic list, and account for all of it, measurement by measurement. A part with 120 characteristics means 120 balloons placed by hand and 120 rows typed by hand. Hours to days per report — and at the next revision, the ritual starts over.

The absurd part: for anyone working model-based in Siemens NX, those characteristics already exist as structured data. Semantic PMI carries the dimension, the tolerance, the datum reference, the associated geometry — everything Form 3 asks for. Transcribing it manually is copying data from one structured system into another, through a human.

We automate exactly that copy: characteristics extracted from NX PMI (or from 2D drawings where that’s what exists), balloons placed automatically, Forms 1–3 pre-filled, and at the next revision a delta report that tells you which seven of the 120 characteristics actually changed. Your quality team measures and releases — the paperwork assembles itself.

// First article inspection, in numbers

0+
characteristics on a typical machined part — each ballooned and transcribed by hand today
4–8 h
of pure paperwork per FAI report — before a single measurement is taken
0%
characteristic accountability required by AS9102 — a missed balloon is a finding
minutes
for the revision delta once extraction is automated — re-check what changed, not everything

The most regulated copy-paste job in manufacturing

First article inspection exists for a good reason: before a part enters serial production, someone must prove that the process produces exactly what the design defines — every characteristic, accounted for. Nobody argues with the purpose. The problem is the mechanics that grew around it.

Ballooning is manual: a quality engineer or designer numbers each characteristic on the drawing, one by one, and any characteristic missed is an audit finding waiting to happen. Transcription is manual: each requirement, nominal, tolerance and inspection method is typed into the Form 3 characteristic list. And revisions are brutal: when the drawing changes, most suppliers re-balloon and re-transcribe the whole part, because working out precisely which characteristics changed is itself a manual comparison job.

Multiply by reality — AS9102 triggers on new parts, design changes, process or supplier changes, tooling changes, and production lapses over two years — and FAI paperwork quietly becomes a permanent tax on your quality and engineering capacity. It is skilled-person time spent on data transport between two structured systems, with a human keyboard as the interface.

In — NX model with semantic PMI, auto-ballooned

Out — AS9102 Form 3, pre-filled

#CharacteristicRequirementSource
1Hole Ø 12 H7, 4ר 12 +0.018/0PMI
2Position 4× holes to A|B|C⌖ Ø 0.1 ⓂPMI
3Base length150 ±0.1PMI
4Flatness, mounting face⏥ 0.05PMI
…118 further characteristicsextracted & numberedauto

Result column stays empty on purpose — measuring and releasing remains your quality team's job.

From model to report: how the automation works

The pipeline follows the way FAI actually flows through a supplier — and it works from both worlds: semantic PMI in NX where you are already model-based, and AI extraction from 2D drawings where that is what the program delivers.

01 — EXTRACT

Characteristics from PMI or drawing

  • Semantic PMI read via NXOpen
  • 2D drawings via AI extraction
  • Dimensions, GD&T, notes, specs
  • Nothing paraphrased — source linked
02 — BALLOON

Numbered automatically, missed nothing

  • Balloons placed per your convention
  • Ballooned views/PDF generated
  • Completeness check: 100% accountability
  • Manual overrides always possible
03 — PREFILL

Forms 1–3, assembled

  • Form 3 characteristic list generated
  • Forms 1–2 from Teamcenter/ERP data
  • Export to Net-Inspect, 1factory, Excel
  • Your template, your numbering
04 — DELTA

Revisions compared, not redone

  • Old vs. new revision diffed
  • Changed characteristics flagged
  • Partial FAI scoped in minutes
  • Audit-ready change record

When does FAI automation pay off?

The math is simple: effort scales with FAI volume and characteristic count. You are a candidate if most of these apply:

  • You produce dozens to hundreds of AS9102 reports per year — across new parts, design changes and process changes.
  • Ballooning and Form-3 transcription cost 4–8 hours per part, done by people you need elsewhere.
  • Revision changes trigger full re-work because nobody can quickly say what changed.
  • You already author semantic PMI in NX — or want to (the MBD page is the natural companion offer).
  • Your customers mandate Net-Inspect, 1factory or their own FAI portals — and you re-type data into them today.
  • FAI backlogs delay production starts — the paperwork, not the parts, sets the schedule.

What the automation is built on

Native NX and Teamcenter interfaces on the input side — your customers' FAI systems on the output side.

NXOpen PMI API Semantic PMI AI drawing extraction (2D) Auto-ballooning AS9102 Forms 1–3 Revision delta compare Teamcenter item & rev data Net-Inspect / 1factory export Excel / PDF / your QMS template

Before and after: the same first article, twice

A machined part, 120 characteristics, design change at revision D — partial FAI required. Here is the paperwork on each side of the change:

Before — balloon & type

  • 120 balloons placed by hand — 2–3 h
  • 120 rows transcribed into Form 3 — 2–4 h
  • Manual rev C vs. rev D comparison — or full re-work to be safe
  • Transcription slips: a missed balloon becomes an audit finding
  • Production start waits for paperwork, not parts

After — extract & verify

  • Characteristics extracted from PMI, balloons placed — minutes
  • Form 3 pre-filled, exported to your FAI system — minutes
  • Rev delta: 7 changed characteristics flagged — partial FAI scoped instantly
  • Completeness checked by machine — 100% accountability, every time
  • Quality engineers measure and judge — instead of typing

Why inteliScience

The FAI tools your customers mandate — Net-Inspect, 1factory and their kin — are good at what they do: managing reports, approvals and supplier workflows. What they don’t do is reach into your CAD system. The gap sits between NX and those tools: someone has to get 120 characteristics out of the model or drawing and into the report, correctly, every time. That gap is where your hours disappear — and it is exactly where we work.

We build on the NXOpen PMI API and Teamcenter interfaces daily; characteristic extraction, auto-ballooning and form generation are engineering on supported extension points, not magic. Where the program still runs on 2D drawings, the same AI extraction that powers our drawing-mining offers reads the characteristics from the sheet. And the output lands wherever your process needs it — customer FAI portals, Excel templates, your QMS.

The honest boundary, as always: the automation prepares, your people decide. Measuring, judging conformance and releasing the report stays with your quality team — they just stop being typists first. And if you are heading towards MBD anyway, the two offers compound: PMI authored by rule today is a Form 3 that writes itself tomorrow.

Fixed price — no day rates, no surprises

FAI Automation Pilot

€30,000 fixed
  • Characteristic extraction for one part family — from NX PMI and/or 2D drawings
  • Auto-ballooning per your convention + Form 1–3 generation in your template
  • Revision-delta compare, demonstrated on real change cases
  • Export to your FAI system (Net-Inspect, 1factory, Excel/QMS)
  • Measured on 10 real parts: time and completeness, your numbers — plus rollout blueprint
Start the pilot conversation

8–10 weeks · built on your NX/Teamcenter installation · tooling and results are yours, whatever you decide afterwards.

For comparison

The cost of hand-made FAI

€110,000+ per year
  • 200 reports × 6 h ballooning & transcription — 1,200 engineer hours
  • Production starts delayed by paperwork backlogs
  • And the un-priced risk: a missed characteristic found by the customer's auditor

Frequently asked questions

We still work with 2D drawings on most programs. Does this only help MBD users?
No — the pipeline has two front ends. Where semantic PMI exists in NX, extraction is deterministic and essentially perfect. Where the program delivers 2D drawings, AI extraction reads the characteristics from the sheet with confidence scoring and a review step for anything doubtful. Most suppliers run both in parallel for years; the automation covers the mix — and every program you migrate to MBD makes its FAIs cheaper as a side effect.
Who is responsible for the report — the software or us?
You — unchanged, and by design. The automation prepares: it extracts, balloons, pre-fills and diffs. Your quality engineer verifies the characteristic list against the source (one click per entry), performs or assigns the measurements, judges conformance and signs. What changes is where their hours go: into judgement instead of transcription. The completeness check actually strengthens the accountability you sign for — a machine does not overlook balloon 87.
Our customers mandate Net-Inspect / 1factory. Does this replace them?
No — it feeds them. Those platforms manage reports, approvals and supplier workflows on the customer side; your customers chose them, and that is not a fight worth having. The gap they leave open is on your side: getting characteristics out of NX and into the report. Our automation fills exactly that gap and exports in the format the mandated system expects — so the tool your customer sees stays the same, and the hours behind it collapse.
How reliable is the extraction from 2D drawings?
Reliable enough to be useful, honest enough to be safe: every extracted characteristic carries a confidence score, low-confidence items go to a review queue, and the completeness logic flags regions of the drawing with no extracted characteristics so nothing silently disappears. In the pilot we measure precision and completeness on ten of your real parts — you see the numbers per drawing type before rollout, not after.
How does the revision delta work?
Both revisions are extracted into the same structured characteristic model, then compared field by field: changed nominals and tolerances, added and removed characteristics, renumbered balloons mapped to their predecessors. The output is the scope of your partial FAI — plus an audit-ready record of what changed between revisions, which is a document your customers ask for anyway and nobody enjoys producing by hand.
Can this connect to our CMM / measurement systems too?
The characteristic list is the natural bridge: it exports with stable IDs that measurement planning and CMM software can consume, so results flow back against the right balloon numbers. Full closed-loop integration (characteristics → measurement program → results → Form 3) depends on your metrology stack and is scoped as a rollout step, not squeezed into the pilot — the pilot proves the extraction-to-form chain first, because that is where the daily hours are.

Your quality engineers didn't study engineering to place balloons. Give them their hours back.

Send us one part — model or drawing — and we'll show you the extracted characteristic list and the pre-filled Form 3, generated from your own data.

Talk to an FAI Automation Engineer
Home » Services » inteliServices » Automated First Article Inspection (AS9102) for Siemens NX

get in touchWe are always ready to help you and answer your questions

Always here to support you – reach out to us with your questions, and we’ll guide you every step of the way.

Call Center

+40 774 463 429

Our Location

+40 774 463 429

Romania

Email

info@inteliscience.net

Social network

info@inteliscience.net

Get in Touch

Define your goals and identify areas where AI can add value to your business