MV switchgear heat rise test — the hot spot sits between two thermocouples.

The problem

Thermocouples see where they sit — heat gives no warning before it shows up.

The IEC 62271 heat rise test on metal-enclosed MV switchgear is designed with thermocouples placed by the standard on the critical connections. But the thermal reality of a loaded cubicle is the sum of hundreds of surfaces, bolted contacts and insulation zones. The classic setup leaks in three places:

1. The thermocouple measures where it sits — if the real hot spot is a few centimeters further along (a bolt torqued wrong, a badly tightened busbar), the thermocouple never sees it.
2. The IR photo arrives at the end — the engineer shoots a thermographic image when the test closes, hours after the fact, with no way to correlate it to a specific event.
3. The dossier is rebuilt afterwards — curves, IR photo, construction drawing and standard are stitched together by hand for the final report.

When a test comes back non-conforming close to the limit, you have to reassemble the cubicle, re-instrument it and energize again — days lost. The classic setup works 99% of the time. That 1% is a reopened test, and for the end customer waiting on the switchgear it is a delay that is hard to explain.

How it fits the IRIS system

iLEAN does not add a fourth system — it seals the cracks between the thermocouples and the construction drawing.

The problem is not a lack of data: thermography has been around for decades. The problem is that nobody can watch an IR map for six hours of testing. iLEAN acts as the putty that fills the gap between what the laboratory can instrument and what the engineer can sustain.

The IR camera sees the whole cubicle. iLEAN Edge analyzes the stream in real time against the type test pattern. The agent cross-references the construction drawing and flags the suspect connection. The person signs — never the other way round.

The iLEAN pieces applied to the MV switchgear heat rise test:

• iLEAN Edge — a terminal with a CNN trained to identify thermally anomalous zones in the IR camera stream. It compares against the learned pattern from the cubicle's type test and flags events in real time. It works with no network: it records and flags locally, then syncs later. What is critical does not depend on WiFi.
• Connect — captures the standard thermocouples (whether modern acquisition or legacy equipment), the cubicle's construction drawing from PLM/CAD, and the manufacturing batch of the busbars and connectors. The graduated capture covers all three cases you find in a laboratory.
• iLEAN Writer — generates the test dossier with curves, the evolving thermographic map, flagged events and the correlation with the construction drawing. The responsible engineer signs with their digital signature and the evidence stays linked to the cubicle's serial number. If a notified body asks for the file, it comes out in one click.

See the full IRIS architecture →

Before and after

Thermography at the end vs. real-time AI thermography with iLEAN

Aspect	Classic heat rise test	With iLEAN Edge + Writer
Coverage of the cubicle	Thermocouples where the standard says	The whole surface, by continuous thermographic stream
Hot spot detection	IR photo at the end of the test	Real-time thermal map, event flagged at the moment
Correlation with construction drawing	Manual, after the test	Automatic, while the test runs
IEC 62271 dossier	Rebuilt by hand for the report	Generated at second zero, human signature
Non-conforming test close to the limit	Reopen, re-instrument, repeat — days	Anomaly caught live, informed decision
Traceability of the connection	A note from the lab technician	Time stamp per event, linked to the serial number

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.

• Mid-sized manufacturer of metal-enclosed MV switchgear (IEC 62271-200), with its own laboratory running the rated-current heat rise test, instrumented with standard thermocouples and a final IR photo.
• Edge pilot with a fixed thermal camera over the cubicle + integration with the thermocouple acquisition system + construction drawing from PLM. First value expected within a few weeks: the first thermal anomaly flagged with a time stamp and correlated to a specific connection.
• Indicative payback between 4 and 9 months, depending on how often non-conforming tests get reopened and on the cost of each re-instrumentation.
• Expected reduction in reopened tests ≥ 30%, a defensible floor. The hard lever is one test not reopened: days recovered + delivery date to the end customer met.

And the lab engineer's reasonable doubt

“What if the AI flags something as a hot spot when it isn't?” — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI compares a real thermal map against a learned pattern from the type test, the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN flags the event and the engineer decides. The three safety rings exist precisely for this.

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