What load is needed before scanning an electrical panel with a thermal camera?

5 min read · Updated 30 July 2026

The physics is worth thirty seconds, because it does all the arguing. A fault — a loose or corroded connection — is extra electrical resistance, and the heat it produces equals the current squared times that resistance. Squared is the operative word: halve the load and the heat drops to a quarter; cut it to a third and the heat falls to about a ninth. The camera has not changed, the fault has not healed — the evidence has simply been turned down below the noise.

What "representative load" means in practice.

It means the plant doing what it normally does — production running, chillers on, the loads that matter actually drawing current. Not the lunch hour, not the weekend, not the convenient quiet window that made scheduling easy. The practical hierarchy runs: typical peak load (best — faults show at their worst), normal operating load (good, and usually achievable), about 40% of normal load (the common rule-of-thumb floor — developed faults still show, marginal ones vanish), and below that, a scan that mostly documents ambient temperature. This is why "no load, no findings" leads our list of the method's limits — it is the limit that gets violated by scheduling convenience rather than physics.

The false clean.

The dangerous output of a light-load scan is not a missed image — it is a confident, colourful, official-looking report that says everything is fine. That report then does real damage: it justifies skipping the next survey, it reassures an insurer about a system that was asleep during its examination, and it sits in the file contradicting the technician who later says the board smells hot. If the load was not recorded, nobody can even tell the report is weak. A clean result only means something when the report says what was running — which is why load comes first in the checks we publish for reading any thermography report.

What to do about it, practically.

Schedule the scan like production work, not like an inspection visit: pick the window when the site genuinely runs, and tell whoever surveys — us included — what will be operating. Where a circuit that matters would otherwise idle, stage it deliberately and give it time to heat: minutes under current, not seconds, because thermal mass lags. Then record the condition — what was on, and the measured current per phase where the panel metering allows — as one line in the scan record. That line is what turns this year's reading and next year's into a trend, and it costs nothing but the habit.