Color changeovers, verified with visual evidence

The problem

A color changeover signed with no evidence can contaminate the first run.

Leftover glaze from the previous color can contaminate the first batch of the new color, forcing the whole startup run to be scrapped.

• Each color change on the glazing line needs the bell, the airless booths, the tanks and the inkjet heads cleaned and adjusted before the new glaze runs. Each point has its own way of hiding residue. A tank wall, a nozzle or an inkjet head can each carry the previous color into the new run.
• Today that is closed with a human sign-off on a checklist, with no visual evidence of each point. The signature is made under pressure, because the line is waiting.
• Leftover glaze from the previous color in a pipe or a tank can contaminate the first batch of the new color. It is often invisible until after firing. Glaze colors look similar raw and only separate in the kiln.
• When it happens, the whole startup run is scrapped, and the investigation finds a signed checklist and nothing else. It is not bad faith: a signature is not a photo. And with several color changes per week, the exposure adds up quickly.

How it fits the IRIS system

Edge with JIDOKA AI — the restart waits for OK at every point and a quality sign-off.

Edge + JIDOKA AI. Industrial cameras with CNNs trained to recognize "ready vs. leftover color". Startup blocked until OK + quality sign-off.

The cameras do not decide alone. They compare each critical point with its reference state, and the quality technician signs on that evidence. What changes is not who decides, but what the decision rests on. And the record of each changeover is already the evidence the audit will ask for. It feeds the evidence pack directly, with no extra work.

See the full IRIS architecture →

Before and after

Color changeover today versus changeover verified by camera

Aspect	Today	With iLEAN Edge
Closing the color changeover	A signature on a checklist	OK per critical point plus signature
Leftover glaze in tank or pipe	Found after firing	Detected before restart
Inkjet head adjustment	Trusted	Compared with the reference state
First run of the new color	At risk of scrap	Started on verified evidence
Restart while a point is not OK	Possible	Blocked by JIDOKA AI
Investigation after contamination	A signature and no proof	Photos of every point

Blind sign-off with contamination risk → sign-off based on 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.

• We do not fix a payback range, because it depends directly on how many color changeovers you run.
• Fewer scrapped startup runs caused by color contamination, which is where the return comes from. Each startup run saved is glaze, energy, kiln time and press time that are not lost.
• A blind sign-off with contamination risk becomes a sign-off based on visual evidence per critical point.
• With your changeover frequency and the cost of a lost startup run, the return can be estimated before deciding. We would rather calculate it with you than publish a range that does not fit your mix.

Reduced scrapped startups from color contamination. Payback depends on changeover frequency. *Estimate to validate*.

And the fair question from the production manager

“What if the cameras block the restart for no reason?” — the changeover sits on the line's critical path, so it is the right objection. Comparing a known point with its reference state — clean or with leftover color — is an anchored task, where the best models drop below 1.5% error [1]. And in case of doubt the system does not block in silence: it escalates to quality, who decides on the images. The block is always explained: which point, which image and why.

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