Coil-to-stamped-part traceability — without the link, you pay for the claim against the steel supplier yourself.

The problem

The claim arrives three months later — and the link broke at the uncoiler.

The quality manager at an automotive stamper always tells it the same way: the OEM or the Tier-1 customer files a claim over a steel defect (delamination appearing after drawing, an inclusion visible when painting, a mechanical property out of spec in a later test). The question is always the same — "which coil did this part come from?" — and in most plants the answer follows this painful route:

1. Look for the part's delivery note to find the production batch.
2. Look for that day's press log and work out which coil was on the line.
3. Cross-reference by hand with the steel mill's delivery note to pull the heat certificate.
4. If the coil marking was lost or somebody linked it wrongly, there is no way to close the loop.

When the loop does not close, there is no claim against the steel supplier — and the cost stays at home. Without a live coil-to-part link, any defect of origin is paid for by you. And the link does not break at the press: it breaks in the first meters of uncoiling, when nobody made sure to link the physical coil, the certificate and the work order.

How it fits the IRIS system

iLEAN does not add a new traceability system — it seals the link where it breaks.

The problem is not a lack of information: the mill delivers its certificate, your MES has the work order, the press logs what it did. The link between the three breaks at the physical moment — the coil entering the uncoiler — because nobody ever had time to make the link by hand every single time. iLEAN acts as the putty that fills that exact crack.

Vision identifies the physical coil as it comes in. Edge verifies that it matches what was planned. Tracer keeps the coil → part link from that second on, batch by batch, for the whole life of the part.

The three iLEAN pieces applied to coil-to-part traceability:

• Vision — a terminal with a camera and a convolutional neural network in the uncoiling area. It reads the coil's face marking (QR, label, mark painted on the edge) in milliseconds and links it to the heat certificate. If the marking was lost, Vision raises a flag before start-up and requires a link signed by the operator — the coil is never processed on its own without traceability closed.
• Edge — cross-references the ID of the identified coil with the work order in the ERP/MES. If it does not match (the wrong coil on the uncoiler), it holds the start-up and triggers the actuator (stack light, controlled stop). It works with no network: if the plant loses WiFi, Edge keeps verifying and holding, because what is critical cannot depend on connectivity.
• Tracer — keeps the link from the coil to every part that leaves the press, batch by batch, for the whole life of the product. When the claim arrives three months later, nothing has to be rebuilt by hand: the quality manager asks Tracer for the list of coils used in the customer's batch, the heat certificate for each one, and sends the claim to the steel mill with the data attached.

See the full IRIS architecture →

Before and after

Traceability rebuilt by hand vs. a live link with iLEAN

Aspect	Classic traceability (delivery note + log)	With iLEAN Tracer + Vision + Edge
Coil identification on entry	Operator reads it and writes it down by hand	Vision reads the marking and links the certificate
Coil ↔ work order match	Assumed to be correct	Edge verifies and holds if it does not add up
Coil → outgoing part link	Rebuilt by hand when a claim comes in	Live, batch by batch, for the life of the product
Response to an OEM claim	Days or weeks, sometimes unsuccessful	Hours, heat certificate included
Claim against the steel mill	Hard to prove — you swallow it	Documented coil by coil — it gets paid
Operation with no network	n/a	Edge keeps identifying on its own panel light

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 stamper with 3-6 presses, buying from 3-5 steel mills, one or two steel-defect claims a year that today never get closed.
• Vision + Edge pilot in the uncoiling area of one main line, integrated with the ERP/MES for the work order. First value expected within a few weeks: the first month of operation produces closed traceability from day one, without rebuilding the past.
• Indicative payback between 4 and 9 months, with a hard measurable lever: a single claim collected from the steel mill instead of swallowed pays for the pilot. Real claims in automotive are not small.
• Expected reduction of over 30% in response time to an OEM claim, automatic dossier included.
• A recurring benefit that does not enter the ROI but carries weight: the change in the relationship with the steel mill — knowing there is traceability coil by coil changes the commercial terms and the discipline of whoever delivers to you.

And the plant manager's reasonable doubt

"What if the camera misreads the marking and the part goes out with no traceability?" — Vision never decides alone. If the reading is not clear, it holds the start-up and requires a link signed by the operator from the app. Hallucination is a problem of free generation, not of anchored tasks: in tasks where the AI simply reads a marking and compares it with a certificate, the best models brought error below 1.5%^[2]. And even so, what is critical is never decided alone — Edge holds, the person signs. The three safety rings exist precisely for this.

^[1] Reference automotive quality standard in the order of 25 PPM (parts per million). Symestic.

^[2] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.