Composite wing control with AI — an internal defect only shows up in ultrasonics. iLEAN combines vision, UT and the autoclave cycle.

The problem

In composites, an internal defect costs a whole part — and a whole part costs a fortune.

The composite process lead knows that a wing cured in the autoclave is a point of no return: if porosity, delamination or a disbond with the core turns up after the cycle, the part is expensive scrap — there is no recovery. And they also know that the truth about how it got there is scattered:

1. The layup — fiber orientation ply by ply, no wrinkles and no FOD between plies, correct impregnation. Today this is verified by the operator's eye and a checklist; some shops run ATL/AFP with data that is never integrated.
2. The autoclave cycle — temperature, pressure, vacuum, ramp. It lives in the autoclave controller; nobody later cross-references it with what the operator saw during layup.
3. NDT — ultrasonic C-scan, sometimes thermography. In the lab's vertical software, in yet another system.
4. The veteran's knowledge — the "I went over that area myself" that no system captures.

The classic setup works 99% of the time. And even so, a single wing with porosity found after the autoclave is a scrapped part whose material, labor and autoclave slot are gone for good. The question is not whether it will happen; it is how much it costs every time it does.

How it fits the IRIS system

iLEAN does not add a fifth system — it seals the cracks between the ones you already have.

The problem with composite wing control is not a lack of information: it is information split into islands that, at the moment of decision (next ply, autoclave entry, NDT release), never quite reaches the same file. iLEAN acts as the putty that fills those gaps, without asking you to change your ATL/AFP, your autoclave or your NDT equipment.

Edge sees each ply before the next one goes down. Connect captures the autoclave cycle and the NDT. The agent cross-references against the recipe and, if something does not add up, holds the part. The person signs — never the other way round.

The three iLEAN pieces applied to composite wing control:

• Edge — a terminal with machine vision (CNN) over the layup station. It detects wrinkles, off-axis fiber orientation, FOD between plies, poor impregnation and vacuum bag failures. Actuator at 45 ms when the part has to be held before the next ply. It works with no network.
• Connect — captures the full autoclave cycle record (temperature, pressure, vacuum, ramp), captures the ultrasonic scanner readings (whether by direct integration, serial reading or PDF digitization), captures the supplier's prepreg batch certificate, and captures the operator's voice during layup. Information arriving by email from a supplier (a change in the resin batch) enters at second zero.
• Agent — cross-references the wing recipe, the ply-by-ply layup (as seen by Edge), the autoclave cycle (Connect), the ultrasonic NDT and the material certificates. If any of them deviates, it holds the part before the next step and alerts the person responsible. The person signs; the wing does not move on by itself.

See the full IRIS architecture →

Before and after

Manual control with a checklist vs. cross-referenced control with iLEAN

Aspect	Manual control + checklist	With iLEAN Edge + Connect + Agent
Fiber orientation ply by ply	Operator's eye, checklist	Edge vision after every ply, image archived
FOD between plies	Visual walkthrough	Edge detection before the bag is closed
Autoclave cycle	Trace inside the autoclave system	Cross-referenced with the recipe and the layup in the file
Ultrasonic NDT	PDF from the lab software	C-scan indexed to the serial number + cross-referenced with the cycle
Operation with no network	n/a	Edge keeps running on the cabinet's own power
Dossier for the auditor / OEM	Rebuilt by hand, weeks	Live dossier per serial number, automatic

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• A CFRP/GFRP layup line for a full wing or a subassembly (spar, rib), a consolidated autoclave and in-house ultrasonic NDT.
• Edge pilot at the layup station + Connect over the autoclave cycle and the ultrasonic scanner + a dossier agent per serial number. First value expected within a few weeks.
• Indicative payback between 4 and 9 months, depending on how often you have scrapped parts for defects found after the autoclave in recent years and on the cost of a single part (material + hours + autoclave slot).
• Reduction of scrap caused by defects found too late: conservative estimate ≥ 30%. The hard lever is a single wing part recovered before the autoclave: that alone pays for the pilot.

And the quality data point for the committee

“What if the AI gets it wrong and lets a defective layup through?” — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely compares an image against a learned pattern and recontextualizes an autoclave data point into a file, the best models brought error below 1.5% ^[1]. And even so, the critical part is never signed off alone: iLEAN holds and the person signs. The three safety rings exist precisely for this.

^[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.

Cross-links: CFRP/GFRP composites with AI vision · ultrasonic NDT in aerospace · aluminum fuselage control