How often should electrical thermography be done?

6 min read · Updated 30 July 2026

"How often" is really a risk question wearing a scheduling costume. An electrical fault that produces heat — a loosening joint, a corroding lug, a slowly overloading circuit — develops over weeks to months, not years. So the interval you choose is a bet on how much development you are willing to let happen unseen. The industry has largely settled that bet at twelve months, and it helps to know why before deciding whether your site is the exception.

Where the annual default comes from.

Three forces converge on the same number. The physics: most heat-producing faults take months to progress from first detectability to serious, so a yearly pass catches the majority of them mid-development rather than post-failure. The insurers: thermographic evidence is a common request at policy renewal in Malaysia, and renewals come yearly — for many facilities this alone sets the rhythm. The standard: NFPA 70B, the reference document for electrical equipment maintenance, became a standard in its 2023 edition — no longer a "recommended practice" — and it ties the thermography interval to equipment condition: twelve months for equipment in good condition, tightening to six for equipment in poor condition.

Two honest cautions on that last one, because this is where marketing usually overreaches. NFPA 70B is American; it is not law in Malaysia, and no Malaysian regulation prescribes a thermography interval by name. The Malaysian duty is broader: Regulation 110 of the Electricity Regulations 1994 requires installations to be maintained in good working order, inspected and tested — with the Regulations separately requiring competent persons for that work — a continuing obligation that a scanning cadence helps discharge, not a rule that names a frequency. We cite the standard as good practice worth borrowing, nothing more.

What tightens the interval.

The condition-based logic in the 2023 standard matches what field experience says: the right question is not "what is normal?" but "what is this board's situation?" Six months — or tighter — earns its cost when any of these apply:

What ignores the calendar entirely.

Some scans are event-driven, whatever the schedule says. After a repair, to confirm the fault is actually gone — days after load is restored, not at the next annual pass. After major electrical work or switching changes. Before an insurer's renewal or an audit, when the evidence needs to be current. And on symptoms — unexplained tripping, warm panel doors, a smell of hot insulation — where the scan is a diagnostic, not an inspection.

One rule survives every cadence decision: the scan must happen under representative load. A survey timed for the quiet week because it was convenient reads cool everywhere and proves little; the calendar matters less than the current flowing when the camera arrives.

Why "once" is the only wrong answer.

Any of the honest intervals beats the common reality, which is a single scan — usually bought for an insurer, filed, and never repeated. The problem is not just that new faults develop unseen afterwards. It is that a single image is the weakest form of the evidence: every reading stands alone, with nothing to compare against. The second survey is disproportionately valuable, because it converts both scans into a trend — this joint was 2°C over its neighbour last year and is 9°C now, and that is a finding no single image can produce.

Declared interest, plainly: this is why we sell assessment as a year-on-year programme rather than as one-off surveys — the first year sets the baseline, and every later year is read against it, same points, same method. But the argument stands whoever you engage: pick an interval your equipment's condition justifies, hold it, and make each survey comparable with the last. A cadence is not a purchase schedule. It is what turns photographs of heat into a history of your switchboards.