Ready to restart, with visual evidence, not a guess

The problem

"Ready" is signed with no evidence behind it.

Today a supervisor signs off "ready" with no objective visual evidence. If the purge wasn't complete, the first batch comes out streaked or color-mixed.

• A color change in molding or in the paint booth — white to black, white to stainless-look — requires a full purge of the barrel, the hopper or the spray lines.
• Today a supervisor signs "ready" after a look. There is no objective visual evidence behind that signature, and the pressure to restart is always there.
• If the purge was incomplete, the first batch of washer lids or refrigerator door trims comes out streaked or color-mixed, and it is found at assembly. By then, dozens of parts have been molded and painted in the wrong shade. A black lid mixed into a white batch is easy to see; a slightly off-white trim is not, and it reaches the customer.
• And the same happens after a mold change, when a residue of the previous material stays in the runner or the nozzle.

How it fits the IRIS system

Edge with JIDOKA AI — no restart until the cameras and quality agree.

Edge + JIDOKA AI. Cameras with CNNs trained to recognize "ready vs. residue of the previous color/material". Restart lockout until OK + quality sign-off.

The camera does not sign. It checks each critical point against its reference image, and the quality supervisor signs on that evidence. What changes is not who decides, but what the decision rests on — and that the record of it stays with the batch.

See the full IRIS architecture →

Before and after

Changeover sign-off today versus sign-off with evidence

Aspect	Today	With iLEAN Edge
Basis for "ready"	A look and a signature	Images per critical point
Residue of previous color	Found in the first batch	Detected before restart
Paint booth after a change	Assumed clean	Checked against reference
Restart	On the supervisor's word	Locked until OK and sign-off
Streaked first batch	Scrapped or reworked	Not produced
Record of the changeover	A box ticked	Evidence stored with the batch

Blind sign-off → sign-off backed by visual evidence per critical point.

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.

• No fixed payback range: we do not set one because it depends directly on how many color and mold changes you run.
• The value comes from reduced startup scrap after each changeover — the streaked or mixed first batch that is simply not produced.
• Blind sign-off becomes sign-off backed by visual evidence for every critical point of the molder and the booth.
• To be estimated with your changeover frequency and the cost of the last streaked batch you had to scrap or rework. Plants with several color changes per shift see the value fastest.

Reduced startup scrap from color/mold changeovers. *Estimate to validate* based on changeover frequency.

And the fair question from the production manager

"What if the camera locks the line for nothing?" — that is the right worry, because a changeover sits on the critical path. Telling "clean" from "residue of the previous color" against a reference image is an anchored task, where the best models drop below 1.5% error [1]. When the image is doubtful, it does not block silently: it escalates to the quality supervisor, who decides on what they see. Every escalation is stored with its image, so the model learns which doubts were real.

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