Carbon fiber prepreg traceability with AI — the clock starts the moment it leaves the freezer.

The problem

Two clocks that almost nobody adds up roll by roll.

Carbon fiber prepreg is not an inert material you store and then use: it is a living resin with two clocks that the aerospace auditor demands for each individual roll.

1. Shelf life — the months the roll can survive at −18 °C in the freezer. It is usually on the supplier's certificate and in the ERP, if somebody entered it.
2. Out-of-freezer life — the cumulative hours at room temperature, adding up every time the roll comes out of the cold to be cut, laid up, or simply left on the table. It usually lives in a spreadsheet kept by the materials manager, if somebody updates it.
3. Actual room temperature — the declared out-life assumes 23 °C. If the room is at 26 °C because the air conditioning is down, the clock runs faster. It is usually nowhere anyone can act on.

The quality manager knows this, but cannot recalculate the balance of every roll each time a cabinet is opened. The laminator, gloves on and in a hurry, glances at a yellow label, thaws and cuts. The classic system works 99% of the time. That 1% is the expired roll that reaches the layup tool — and the AS9100 audit NCR does not forgive.

How it fits the IRIS system

iLEAN does not add another spreadsheet — it seals the cracks between the freezer, the table and the ERP.

The prepreg problem is not a lack of information: it is information living on separate islands that, at the critical moment (the cabinet door opening, the second thaw, the shift change with a half-used roll), does not reach the decision-maker in time. iLEAN acts as the putty that fills those gaps, without asking you to change your ERP, your composites MES or your layup room.

The terminal reads the roll label as it leaves the cabinet. The agent adds up the clock against the actual room temperature. The per-roll file builds itself and is handed to the auditor signed.

The three iLEAN pieces applied to carbon fiber prepreg traceability:

• Edge — a computer-vision terminal (CNN) at the freezer cabinet door and at the cutting table. It reads the roll label (Hexcel, Toray, Solvay, Cytec), records the in/out event and, if you want, triggers a stack light when a roll has been out too long. It works with no network. If the plant loses WiFi, Edge keeps reading and recording.
• Connect — captures the layup recipe, the supplier's delivery note, the manufacturer's certificate and the reading from the ambient temperature sensor. And it also captures what comes in from outside (an email from the supplier with a change to the batch out-life, a WhatsApp message from the R&D manager with an authorized exception for a specific roll) at second zero.
• Agents — add up the clock roll by roll, cross-reference it with the actual temperature, maintain the balance and assemble the per-roll file in auditor format (certificate + temperature curve + out-of-freezer hours + signature of the person who released it). If a roll goes past the limit, the agent holds the material and alerts the quality manager through whatever channel they use. The person validates and signs; the roll is never released on its own.

See the full IRIS architecture →

Before and after

Manual prepreg traceability vs. a live file with iLEAN Tracer

Aspect	Spreadsheet + yellow label on the roll	With iLEAN Edge + Connect + Agents
Out-of-freezer clock per roll	Added up by hand; depends on the laminator writing it down	Added up automatically from the label reading on the way out
Actual room temperature	Assumed (a theoretical 23 °C)	Integrated sensor; the clock runs accordingly
Real-time balance	Calculated when somebody asks	Visible to operator and quality on screen and earpiece
Expired roll	Caught when unwrapping it or at the audit	Held before it leaves the cabinet
Operation without a network	n/a	Edge keeps reading with its own light
Per-roll file for AS9100	Rebuilt by hand, weeks of work	Per-roll dossier, automatic, ready for the auditor

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.

• Aerospace composites plant with a layup room, 2-4 prepreg freezer cabinets and ATL/manual cutting tables. Volume of rolls opened per month in the tens to the low hundreds.
• iLEAN Tracer pilot on one cabinet and one table: camera for label reading + ambient temperature sensor + integration with the composites ERP/MES. First value expected within a few weeks (out-life balance per roll visible on screen and traced into the file).
• Indicative payback between 4 and 9 months, depending on the number of rolls scrapped per year because the clock was uncertain, the average cost of a structural roll, and the weight of historical AS9100 audit NCRs.
• The hard levers: rescued rolls that the system certifies as still within out-life (instead of scrapping on "I don't know", scrapping on "it exceeds"), a reduction of ≥ 30% in the time spent preparing the audit file, and one NCR avoided in primary structure — which on its own pays for the pilot.

And the quality manager's reasonable doubt

“What if the AI makes up an exit time that never happened?” — hallucination is a problem of free generation, not of anchored tasks. Reading the label and computing the clock are tasks anchored to the source: read a code, add up minutes, compare against a threshold. In that family of tasks the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN holds the roll and the person signs. The three safety rings are there for exactly this — ring 1 (the OT network, the machines) accepts nothing that has not been validated and signed by a person.

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