Pattern plate changeover validated with evidence, not a blind signature

The problem

The most fragile moment of the day starts on a blind signature.

A badly started changeover is paid in two currencies. The cheap one: the first molds to scrap and a long hour of lost cadence at the heart of the plant. The expensive one: a run with the wrong core or residual sand that passes visual inspection and ends up as a defective brake part at the customer. On top of that, the number of changeovers keeps growing. Part mix increases with electrification — differential cases, control arms and calipers for hybrid and electric vehicles run alongside traditional production — so the pattern change goes from being an event to being a daily routine. And there is no data: changeover duration is "whatever it takes", different on every shift, with nobody knowing where the time goes.

• The line stops, several pairs of hands work together, and start-up depends on somebody signing that pattern plate, cores and chamber are ready — with no objective evidence that they are.
• The cheap failure: the first molds to scrap and a long hour of lost cadence at the heart of the plant. The expensive one: a run with the wrong core or residual sand that passes visual inspection and ends up as a defective brake part at the customer.
• Changeovers keep growing. Differential cases, control arms and calipers for hybrid and electric vehicles run alongside traditional production, so the pattern change is now a daily routine.
• And there is no data: the changeover takes whatever it takes, different on every shift, with nobody knowing where the time goes.

How it fits the IRIS system

Edge with JIDOKA AI and SMED AI — the line does not start without evidence and a signature.

Edge combined with JIDOKA AI and SMED AI:

The cameras do not decide alone: each critical point is compared against the reference state for that order — this pattern plate, these cores, this chamber clear of residual sand — and it is the supervisor who signs on that evidence. In parallel, SMED AI times every phase, so the best changeover of the month becomes the standard. What changes is not who decides, but what they decide on. JIDOKA AI is the preventive block: if any point is not OK, the line does not start, and blind trust is replaced by a photo of the plate, the cores and the chamber, stored with the signature.

• Cameras at the N critical points compare the current state against the reference state for that order: this pattern plate, these cores, this molding chamber clear of residual sand.
• If any point is not OK, the line does not start. That is JIDOKA AI: preventive blocking replaces blind trust.
• The supervisor sees the evidence point by point and signs. Humans in command: the AI does not sign in their place.
• In parallel, SMED AI times each phase of the changeover and reports where the time goes, comparing across shifts, so the best changeover of the month becomes the standard.

See the full IRIS architecture →

Before and after

Today's changeover versus the validated changeover

Aspect	Today	With iLEAN Edge
Start-up authorization	A blind signature	Point-by-point visual evidence, then a signature
Wrong pattern plate or core	Found in the castings	Blocked before the first mold
Residual sand in the chamber	Passes visual inspection	Detected before start-up
First molds	To scrap as routine	A validated start-up
Changeover duration	Whatever it takes	Measured phase by phase, compared across shifts
Who signs	The supervisor, blind	The supervisor, on evidence

from a blind signature to a signature backed by visual evidence point by point · from first molds to scrap as a matter of routine to a validated start-up · from unknown changeover duration to a measured duration comparable across shifts.

Impact estimate

Impact estimate — to be validated with your 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, depending on the number of changeovers per week.
• A 20-40% reduction in changeover time through SMED AI, on the line that is the heart of the plant.
• Practical elimination of the risk of a run with the wrong pattern or core — which is mass-containment risk.
• And the first molds stop going to scrap as a matter of routine.

estimated payback 6 to 12 months, with a 20-40% reduction in changeover time through SMED AI and the practical elimination of the risk of a run with the wrong pattern or core, which is mass-containment risk. *Estimate to validate*.

And the fair question from the production manager

“What if the cameras block the line for no reason?” — that is the right objection, because the changeover sits on the critical path of the whole plant. Conformity is checked against the reference state for that specific order, and comparing a mounted pattern plate or a core set against its reference image is an anchored task where the best models drop below 1.5% error [1]. In case of doubt the system does not block silently: it escalates to the supervisor, who signs on the evidence. It stops on a verifiable discrepancy, not on model uncertainty.

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