Visual setup validation before the run starts

The problem

After a part-number change, the first minutes decide the first lot.

the responsibility for a correct setup today rests on a human sign-off with no objective visual evidence.

• Switching part numbers on a multi-spindle lathe means resetting tool holders, inserts and cams across several stations, often under time pressure. On a high-mix fleet that can mean several changeovers per shift.
• If one is off, the first minutes produce out-of-tolerance parts, and on a safety part those parts are a containment risk, not just scrap. The scrap is visible; the risk of a part that passed is not.
• Today the setup is closed by a human sign-off with no objective visual evidence behind it, however experienced the setter is.
• When something goes wrong, the investigation finds a signature, not a reason, and the corrective action ends up being another line on the checklist.

How it fits the IRIS system

Edge plus JIDOKA AI — the run does not start until every camera and the quality lead agree.

Edge + JIDOKA AI. Industrial cameras with CNNs trained to recognize "correct setup vs. misplaced tooling"; the run is blocked until all cameras pass plus the quality lead's sign-off.

The cameras do not replace the setter's judgment: they put evidence next to it. What changes is not who signs, but what they sign on — and every signature comes with the images of each station at the moment of approval. For a setter, it is also proof that the setup they signed was right.

See the full IRIS architecture →

Before and after

A signed setup versus a camera-evidenced setup

Aspect	Today	With iLEAN Edge
Setup sign-off	A signature, no evidence	Cameras at every critical point, plus sign-off
Misplaced insert or holder	Found on the first parts	Detected before the run
First lot after changeover	Born under suspicion	Protected from minute one
Reference state per part number	In the setter's memory	Stored images per station
Run start	When the setter says so	Blocked until all cameras pass
Evidence for an investigation	None	Images of every station

blind sign-off with risk of out-of-tolerance parts from minute one → visual evidence at every critical setup point.

Impact estimate

Impact estimate — to be validated with your changeover calendar.

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.

• Payback depends on changeover frequency: we do not fix a range because it scales with how many part-number changes you run per week.
• Direct protection of the first lot after every changeover, which is where most setup-related scrap and risk concentrate.
• Blind sign-off is replaced by visual evidence at every critical setup point. The setter no longer carries the responsibility alone.
• And the investigation, if one is ever needed, starts with images instead of a tick on a sheet. Repeated discrepancies at one station also show where the setup procedure needs work.

direct protection of the first lot after every changeover, payback depending on changeover frequency. *Estimate to be validated*.

And the fair question from the production manager

“What if the camera blocks a correct setup and the lathe sits idle?” — comparing a station against its reference image is an anchored task, where the best models drop below 1.5% error [1]. And on doubt the system does not block silently: it escalates to the quality lead, who signs on the evidence, so the lathe stops on a real discrepancy, not on model uncertainty.

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