NH fuse breaking capacity test — reading the oscillogram by eye means expensive scrap and an expensive technician.

The problem

The test is sound; validating by eye breaks down at high volume.

The breaking capacity test on an industrial NH fuse admits no middle ground: the cartridge has to interrupt a real prospective current, without the arc escaping the enclosure, within the declared I²t and time. The standard — IEC 60269 — is clear. The laboratory knows how to run it. Where the chain breaks is in how each oscillogram is handled:

1. Every shot costs a sample. The NH fuse under test is destroyed. If the trace is doubtful and gets discarded to be safe, a sample is lost; if a doubtful trace is accepted under time pressure, the risk goes out into the field.
2. The qualified technician is spent on dumb work. Looking at 200 clean traces to find 5 doubtful ones is not what they were hired for. That is screening work, not judgment.
3. The oscillogram is stored in fragments. The bench oscilloscope, the results spreadsheet, the manager's report in Word. Rebuilding a batch dossier for an audit is a week in the archive.
4. The bench conditions never get tied together. Recovery voltage, frequency, room temperature, time of the shot: it all exists, but in separate files, and the cross-referencing is done by hand.

The laboratory has the technique and the people. What it lacks is the putty that binds the oscillogram to the standard, the conditions and the history — at second zero, not at month end.

How it fits the IRIS system

Edge on the bench, IEC 60269 in the agent, judgment in the person.

The breaking capacity test lives between three islands: the oscilloscope, the bench conditions sheet and the standard. iLEAN acts as the putty that seals the cracks between those three, without asking you to change the oscilloscope or rewrite the procedures.

Edge digitizes the oscillogram. The agent validates it against IEC 60269. The person signs — the batch report does not go out without a human seal.

The iLEAN pieces applied to NH fuse testing:

• Edge — a terminal next to the short-circuit bench. It reads the oscilloscope channels on every shot and digitizes the complete oscillogram. It applies a first screening (prospective current reached, arcing time inside the window, no visible re-ignition) in milliseconds. If the trace is clean, it seals it; if it is doubtful, it holds the sample and passes it to the technician. It works with no network.
• Connect — captures the room conditions (recovery voltage, frequency, ambient temperature) from the bench SCADA or the local panel, and cross-references them with the oscillogram. The bench configuration sheet — today a spreadsheet — becomes data bound to the shot.
• Agent — on top of what has been digitized, it validates the IEC 60269 criteria: total I²t, no re-ignition, let-through energy, arc waveform. It builds the batch dossier automatically, ready for an audit or for the customer running the approval.

See the full IRIS architecture →

Before and after

Validating by eye vs. validating with iLEAN Edge

Aspect	Laboratory today	With iLEAN Edge + Agent
Oscillogram capture	Oscilloscope + manual saving	Edge digitizes the complete oscillogram on every shot
Screening of clean traces	Technician looks at all 200	Agent filters; technician only looks at the doubtful ones
Validation against IEC 60269	By eye, against a mental checklist	Agent applies quantitative criteria
Bench conditions	Spreadsheet separate from the oscillogram	Cross-referenced with every shot, at second zero
Audit dossier	Rebuilt by hand, weeks	Per batch, automatic, with every trace
Operation with no network	n/a	Edge keeps capturing as long as the panel has power

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your laboratory. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• In-house laboratory of an NH fuse manufacturer or of an electrical testing operator, with its own short-circuit bench, several hundred shots a month and one qualified person mainly dedicated to validating traces.
• Edge pilot on one bench (terminal + integration with the oscilloscope + IEC 60269 adapter). First value expected within a few weeks: automatic screening of 80-90% of the clean traces frees technician hours from day one.
• Indicative payback between 4 and 9 months, depending on shot volume and on the cost of discarded samples.
• Expected reduction in samples discarded out of doubt ≥ 30% and reduction in technician time spent screening ≥ 30% — conservative.
• Extra levers (not counted in the floor): automatic audit dossier and complete batch traceability with no manual rebuilding.

And the quality manager's reasonable doubt

“What if the agent accepts a doubtful trace the technician would have held?” — hallucination is a problem of free generation, not of anchored tasks. Applying quantitative IEC 60269 criteria to a digitized trace is exactly an anchored task: the best models brought error below 1.5% ^[1]. And, once again, Edge does not sign — it proposes and holds. The seal on the report always belongs to the person.

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