Four rings between the coil and the pallet

The problem

The capital pain: a nonconforming wheel leaving the plant.

A single nonconforming wheel at the customer triggers a cost chain that starts in the tens of thousands of euros and has no defined ceiling. And unlike other sectors, here the part is critical to vehicle safety: there is no concession available and no negotiable deviation. Its internal twin is unplanned downtime, in a plant that must deliver millions of wheels a year against a sequenced schedule with no inventory buffer. Both problems share the same root: information that exists in one system but is not cross-checked against another in time. The plant already knows this and already tries. What it lacks is not will or procedure: it is cross-checked evidence in real time between systems that today do not talk to each other.

• A single nonconforming wheel at the customer triggers a cost chain that starts in the tens of thousands of euros and has no defined ceiling.
• The wheel is a safety part: no concession is possible, no deviation is negotiable. At worst, the scale is a safety campaign across thousands of vehicles.
• Its internal twin is your own unplanned downtime, in a plant that must deliver millions of wheels a year against a sequenced schedule with no stock buffer.
• Both problems share the same root: information that exists in one system but is not crossed with another's in time.
• The plant already knows this and already tries. What it lacks is neither will nor procedure: it is real-time cross-checked evidence between systems that do not talk to each other today.

How it fits the IRIS system

Four rings that cross-check — if they do not agree, the line stops.

No single ring solves the problem: the strength is in the cross-check. The wheel passing the camera has to agree with what the order says, with what the marker engraved and with what the pallet dossier declares. Four independent sources contradicting each other is an alarm; a single source is an assumption.

• Ring 1 · The order is the truth (business and manufacturing system). The active order defines part number, coil heat, die set, process parameters and destination customer.
• Ring 2 · The marker engraves what is published (case 8). It engraves exactly what the system publishes and returns confirmation of what was engraved. Zero intermediate keying, no possible divergence between the order and the wheel's physical identity.
• Ring 3 · The camera cross-checks and JIDOKA stops (cases 9 and 10). The camera reads the weld, the finish and the actual marking on the wheel going past and cross-checks them against the active order. On any discrepancy, JIDOKA AI stops the line on the spot. And the changeover could not even have started without visual validation of the tooling.
• Ring 4 · The pallet ships with its dossier (case 11). Every pallet ships with an evidence pack cross-referencing the previous three rings plus the photo of the marking, the welding parameters read from the panel and the signature of whoever validated the first piece.

Underneath runs the predictive layer (cases 2 and 4): the welder curve and the spoken reports from the maintenance round feed the reliability model, so downtime is anticipated before something breaks and parameter drift is corrected before the first bad part is produced. And SMED AI speeds up changeovers so that running four active rings is never paid for in productivity — the condition that makes the system acceptable on the floor.

See the full IRIS architecture →

Before and after

The four rings, and what each one closes

Ring	What it contributes	What failure it closes
1 · The order is the truth	Part number, heat, tooling, parameters and destination customer	Ambiguity about what should have been produced
2 · The marker engraves what is published	Engraved with no keystroke in between, confirmation returned	Divergence between marking and order
3 · The camera cross-checks and JIDOKA stops	Bead, finish and actual marking against the live order	A nonconforming part that keeps moving
4 · The pallet ships with its dossier	Evidence from the three rings plus marking photo and parameters	Traceability that has to be rebuilt afterwards

Impact estimate

Estimated impact — to validate with your own numbers.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• Estimated payback 6-12 months against the first customer line stoppage avoided.
• Protection against a safety campaign: a single impact potentially running into several hundred thousand euros, with no defined ceiling.
• From detection at the customer to detection on the line, and from reactive unplanned downtime to anticipated downtime.
• Protection of the OEM relationship and of the award of future programs, on a multi-year horizon.

estimated payback 6-12 months against the first customer line stoppage avoided. Add the protection against a safety campaign — a single impact potentially running into several hundred thousand euros with no defined ceiling — and the protection of the customer relationship and of future programme nominations, which is multi-year in value. *Estimate to validate with the quality manager and plant management.*

And the fair question from the production manager

«Do we have to deploy all twelve cases to get this?» — no, and framing it that way would be the way never to start. The rings are built in layers: each case pays for itself from the first month and, as they add up, they close the cross-check. A plant that starts with coil receiving and the welder panel already has half of ring 1 and half of ring 3 running without committing to a two-year project.

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