It doesn't start until the configuration is verified

The problem

A wrong busbar setup is an electrical safety risk signed off with a tick.

A wrong busbar configuration in medium-voltage equipment isn't a cosmetic defect, it's an electrical safety risk — and today it rests on a technician signing off "correctly configured" with no objective visual evidence.

• The same line assembles switchgear of different ratings in the same shift. Before each order, the cell has to be set up correctly: main busbars, joints, phase spacing, grounding. Each rating has its own drawing, and the differences between them are sometimes a few millimeters.
• A wrong busbar configuration in medium-voltage equipment is not a cosmetic defect. It is an electrical safety risk, and potentially an out-of-spec unit shipped as conforming.
• Today that risk rests on a technician signing “correctly configured” on a checklist, with no objective visual evidence behind the signature and the next order already waiting.
• When something goes wrong, the investigation finds a signed box and nothing else. It is not bad faith: a signature is not evidence, and under time pressure it never will be.

How it fits the IRIS system

Edge with JIDOKA AI — assembly does not start until the cameras read OK and quality signs.

Edge + JIDOKA AI. Industrial cameras with CNNs trained to recognize "correct configuration vs. deviation" for each rating. Assembly can't start until every camera reads OK + the quality lead signs off with evidence.

The cameras do not decide alone. They compare each critical point with the reference for that rating, and the quality lead signs on that evidence. What changes is not who approves the start, but what they approve it on — and that the evidence stays attached to the order.

See the full IRIS architecture →

Before and after

The rating changeover today versus the changeover checked by camera

Aspect	Today	With iLEAN Edge
Busbar configuration sign-off	A tick on the checklist	Every critical point compared with its reference
Phase spacing and joint layout	Checked by eye	Read by fixed cameras per rating
Grounding connection	Assumed from the drawing	Confirmed before startup
Start of assembly on a wrong setup	Possible under time pressure	Blocked until OK and signature
Evidence of the busbar setup afterwards	A signed box	Images stored with the order
Ratings mixed in one shift	More changeovers, more exposure	Each changeover verified the same way

A blind sign-off with configuration risk → a sign-off backed by visual evidence at every critical point.

Impact estimate

Configuration validation: impact we estimate with you, changeover by changeover.

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 varies with how often your configuration changes: we do not fix a range because it depends entirely on your ratings mix per shift and on what one incident would cost your plant.
• It protects against an electrical safety incident or an out-of-spec unit, whose cost far exceeds a few minutes of line downtime at each changeover, and against the investigation that follows either one.
• The sign-off goes from blind to backed by visual evidence at every critical point, and the quality lead signs knowing what they are signing.
• And when a customer or an auditor asks how a unit was configured, the images exist with its order instead of a tick on a sheet. The answer takes a minute, and it is evidence, not a recollection.

Variable payback depending on how often configuration changes occur; protects against an electrical safety incident or an out-of-spec unit, whose cost far exceeds a few minutes of line downtime. *Estimate to validate.*

And the fair question from the production manager

“What if the cameras block the line over nothing?” — that is the right objection, because changeovers sit on the shift's critical path. Comparing a busbar layout against its stored reference is an anchored task, where the best models drop below 1.5% error [1]. When the camera is unsure it does not block silently: it escalates to the quality lead, who decides on the evidence in front of them.

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