Malaysia mandates the inspection. It never says what to look for.

14 min read · Updated 1 August 2026

There is a statistic that circulates in this trade, quoted in tenders and toolbox talks and sales decks, that loose connections cause somewhere between 30% and 40% of electrical failures. It is real, it has a source, and we will come back to it. But it is American, it counts insurance claims, and it says nothing about whether anyone was hurt.

Malaysia has its own numbers. The Energy Commission — Suruhanjaya Tenaga — publishes an annual electrical safety performance report drawn from the accidents it investigates here. It is not widely read. It is considerably more useful than the borrowed figure, and in one respect considerably less comfortable.

Malaysia also has its own rule, which is where most writing on this subject goes wrong, ours included until we read it properly. There is a mandatory periodic inspection of electrical installations in this country, and it has been law since 1994. The argument in this article is not that the requirement is missing. It is that the requirement never says what the inspection has to look for — and that the failure mode the trade has been quoting at each other for thirty years is precisely the one it does not ask anybody to find.

The rule exists. Read what it actually says.

Malaysia does not lack an inspection requirement, and this article is not going to pretend it does. Regulation 110 of the Electricity Regulations 1994 is explicit:

Regulation 110(2) puts that duty on the owner, management, licensee or occupier — the organisation whose switchboard it is. Regulation 110(4) gives protective relays and devices a shorter clock, two years. And it is enforced rather than decorative: the Commission's own 2019 report lists two of Tenaga Nasional's distribution divisions compounded for failing to comply with regulation 110(1).

That is not even the frequent part. Regulation 67 requires a competent person to visit and inspect far more often than every five years:

Regulation 68 then requires that competent person to report findings to both the Commission and the owner, to deliver a written report within two weeks of the inspection, and requires the owner to keep a book of every such record, open to the Commission at all times. Contravening any of this is an offence under regulation 122 — a fine up to RM5,000, up to a year's imprisonment, or both.

So Malaysia does not have a light-touch regime. It has a prescriptive one: who may inspect, how often, what they must file, how long the record is kept, and what happens if you skip it. The problem sits one level down from all of it.

Read any of it again and ask what it obliges anyone to actually look for. An installation shall be “checked and tested.” A competent person shall “visit and inspect.” Neither phrase is defined anywhere. Nothing says the installation must be examined under representative load, which is the only condition in which a deteriorating connection gives itself away. No method is named. No threshold is set for what counts as a finding. The regime prescribes who, how often, and what to file — and never once what to look for.

Now put that beside the failure data. The most quoted number in this trade — the one this article opened with — says loose connections are the single largest cause of electrical distribution failures. It is an insurer's claims table, and we will show it in full further down. The one practical way to find a loose connection on a live installation is to look at it with a thermal camera.

The words “thermography”, “infrared” and “thermal” do not appear anywhere in the Electricity Regulations 1994. We searched the current consolidated text, the one incorporating every amendment through 2014. The only hit is “geothermal”, in a definition about renewable generation. A monthly visit is mandatory. A five-yearly check is mandatory. Looking for heat is not mentioned.

Then the building runs. Loads get added. A line is reconfigured, a machine moves, somebody extends a board on a Saturday because production starts Monday. Connections cycle warm and cool through years of that, loosening exactly as the insurer's claims data describes, and dust settles on them in a switchroom nobody has opened since handover.

A competent person may well have walked past that switchroom twelve times in the year, filed twelve reports, and complied fully with every regulation named above without ever being required to point an instrument at the joint that is heating up. That is the gap this article is about. It is not a missing rule. It is a missing sentence inside the rules we already have.

And yes, we think the rule should name infrared thermography. NFPA 70B names it, because there is no other practical way to survey a live installation for thermal fault without shutting it down. We sell infrared thermography, so you are entitled to treat our enthusiasm with suspicion. The reply we would offer is that the NFPA sells nothing. It is a standards body with no equipment to move, it examined the same failure problem, and it wrote infrared into a mandatory document anyway.

What regulation 110 should borrow is not just the naming but the structure around it: intervals tied to the condition equipment is actually in rather than to the calendar, and the method for any given installation signed for by the competent person who is already required to be there. Infrared answers the thermal question. It does not answer every question — discharge testing, airborne ultrasound and contact resistance measurement each cover ground infrared cannot — and a rule that pretends otherwise would be a worse rule.

And the limit stands, because it is the reason this article exists. None of this follows from the accident data as proof. The Commission's figures record what caused accidents, not what would have prevented them, and no honest reading converts the one into the other. NFPA 70B is a US document with no force in Malaysian law, cited here as precedent and nothing more. The numbers in the sections above are checkable. This section is a position, and it is labelled as one.

What the regulator counts.

So much for the shape of the problem. Here is the size of it, from the only Malaysian source that counts: between 2002 and 2019, Suruhanjaya Tenaga investigated 1,026 electrical accident cases in Malaysia. Of those, 505 were fatal and 521 were not. That is an average of 57 investigated cases a year, and it means roughly half of the accidents serious enough to investigate involved someone dying.

Before any of that is used for anything, one boundary matters: these are cases investigated by the Commission. They are not every electrical accident in Malaysia. Incidents too minor to reach the regulator, or on installations outside its investigation practice, are not in the count. This is an enforcement dataset, and treating it as national incidence would overstate what it can carry.

The largest identified cause is maintenance.

The Commission classifies each investigated accident by cause. Across the full 2002–2019 series, the ranking is:

Identified cause of investigated electrical accidents, 2002–2019, as a count of cases out of 1,026 with each cause’s share beside it. Bars are scaled to the largest category. Source: Suruhanjaya Tenaga, Laporan Prestasi Keselamatan Elektrik 2019.

Read the category name carefully. It is pemasangan atau senggaraan tidak sempurna — installation or maintenance done imperfectly. The report does not split the two, so nobody can honestly say what share is maintenance alone. We have seen that figure quoted as a maintenance statistic. It is not one, and the source does not permit it.

What it does support is narrower and still significant: over eighteen years, the largest single identified cause is work that was not done properly — at installation or in the upkeep afterwards. Ahead of procedure failures, and ahead of equipment defects by a factor of eleven.

Where accidents happen, and to whom.

Location of investigated accidents, 2002–2019, as a count of cases out of 1,026. Bars are scaled to the largest category. Highlighted: factories and private commercial premises, the buildings where an owner is responsible for the switchroom. Source: Suruhanjaya Tenaga, Laporan Prestasi Keselamatan Elektrik 2019.

Utility infrastructure dominates, as you would expect. The part worth noticing is lower down: factories and private commercial premises together account for 209 cases, almost exactly one investigated accident in five, in the kind of building where somebody is responsible for the switchroom.

And the people hurt are mostly not the public. 46% of victims were contractors, 37% members of the public, 17% utility workers, with 21 to 30 the worst-affected age band. More often than not, it is the people sent to do the work.

The trend is good, and saying so costs us nothing.

A safety-services company quoting accident statistics has an obvious incentive to imply things are getting worse. In Malaysia they are not, and the regulator's own series shows it plainly.

2019 recorded 15 fatalities, against a peak of 38 in 2007. That is the second-lowest year in the record — the lowest is 2002, with 14, which is also the first year of the series and the one most likely to be undercounted. Taken as raw counts, then, the honest summary is not that fatalities have fallen; it is that they rose through the mid-2000s and have come back down to roughly where the record begins.

What changed underneath those counts is the size of the country they describe. The number of electricity users grew from 5.8 to 9.9 million over the same period, so the same raw count represents a materially lower rate. Deaths per million electricity users fell from 2.42 in 2002 to 1.52 in 2019, the lowest in the series, and total investigated cases per million users reached their series low in the same year.

One caveat belongs beside that, because leaving it out would flatter the trend. From 2018 the Commission stopped counting trespass cases, on the reasoning that trespass is deliberate. Part of the recent fall is therefore a change in what gets counted. The direction holds across the longer series; the size of the last step should be read knowing that.

What these numbers cannot tell you.

Two things, and the first is the one that matters most here.

They do not say an inspection would have prevented any of these accidents. The Energy Commission classifies causes; it does not evaluate detection methods, and it makes no finding about what a survey would or would not have caught. Nobody — including us — can convert “35.4% of investigated accidents involved imperfect installation or maintenance” into “inspection would have stopped a third of them.” The data does not reach that far, and the rest of this article is written knowing it.

And no single percentage in this trade means what it appears to mean. This is where the borrowed statistic from the top of this article comes back. Its source is Hartford Steam Boiler, the equipment-breakdown insurer, and the underlying table is public: loose connections and parts head the list of causes of electrical distribution system failures, at 30.3%.

Top causes of electrical distribution system failures, as a share of claims. Bars are scaled to the largest category. This is a different population from every other chart on this page: American insured equipment failures that generated a claim, not Malaysian accidents that hurt somebody. Source: Hartford Steam Boiler, The Locomotive — “Failures Based on Hartford Steam Boiler Claims Data”.

So the trade figure is real, and now you can see what it is a percentage of. That is the whole point. A seven-year infrared study at an Argentine utility found 48% of thermal anomalies in connections and bolted joints. An analysis of medium-voltage switchgear built on IEEE reliability data ranks moisture first instead, at 17.7%. All of them are sound. They count different things — failures that generated a claim, anomalies a camera could see, contributors in one class of switchgear — and the Malaysian figures above count something different again: accidents that hurt somebody. Stack them into a single number and you have produced nothing.

Which is worth carrying past this page. When someone quotes you a percentage about electrical failure, including the ones here, the useful question is what the denominator was.

Somebody else asked the same question.

They reached an answer in 2023. The National Fire Protection Association reissued NFPA 70B, its document on electrical equipment maintenance, as a standard rather than a recommended practice. That is not a cosmetic change. A recommended practice advises; a standard uses mandatory language and can be adopted, referenced in contracts, and enforced.

The substance moved with the wording. The earlier version suggested infrared scanning on particular equipment — outdoor substations, switchgear assemblies, motor control equipment, busway, UPS systems. The 2023 edition requires inspection of electrical equipment at defined intervals, with the interval tied to the condition the equipment is actually in: broadly annual, tightening to roughly six-monthly thermographic inspection for equipment whose condition has already been flagged.

The reasoning behind that shift is worth more than the rule itself. “Recommended” produces uneven practice — conscientious operators do it, others do not, and nobody can tell the difference from the outside. Making it a standard makes maintenance auditable: an insurer, a buyer or an investigator can ask what was inspected, when, and what was found, and expect a documented answer. And tying frequency to observed condition rather than to the calendar means the equipment that is deteriorating gets looked at more often than the equipment that is not.

What would actually have to change.

Very little, which is the strongest thing about this argument. No new statute, no new duty, no new profession, and no new paperwork. Malaysia already compels the inspection, already fixes it on the owner or occupier, already requires a licensed competent person, already sets the frequency, already demands a written report inside two weeks, and already makes the owner keep the record book. Every piece of machinery an inspection regime needs is in place. Two things would be added inside it:

Say what the inspection includes. For a live installation, a thermal survey under representative load, because a joint that is failing is only visible when current is flowing through it. This is the change that matters, and it is the one thing in this list that does not already exist in some form.

Let the interval follow the equipment, not the calendar. Five years for a switchboard in good condition is defensible. Five years for one already flagged as running hot is not. NFPA 70B tightens to roughly six-monthly thermographic inspection where condition warrants it, and there is no reason Malaysia could not do the same.

Notice how small that is. The record book that regulation 68 already obliges every owner to keep would simply have something in it worth reading.

And the cost would be real. A thermal survey is not free, it is a line in a maintenance budget that is usually already under pressure, and pretending otherwise would be dishonest — particularly from us, since it is our invoice.

One more thing belongs here, because it is the first objection an honest reader will raise. Most of the accidents in this record did not happen in anybody's factory. Utility substations, overhead lines and underground cable account for 561 of the 1,026 investigated cases — more than half. Factories and private commercial premises together are 209, about one in five. If you were allocating attention purely by case count, you would start with the network, not the switchroom.

We would answer that two ways. The utility network has a licensee, a maintenance obligation, an inspection regime and a regulator that demonstrably enforces it — the compounds under regulation 110(1) in that same report were issued against distribution divisions, not against factories. And this is an enforcement dataset: a utility accident is reportable, while a hot joint in a private switchroom that destroys a board and stops production without injuring anybody never enters these figures at all. The consumer side of the meter is not thinly represented here because it is safe. It is thinly represented because not one of those mandatory inspections has to look for a thermal fault, and because nothing obliges a building to report a failure that cost it money but hurt nobody.

Where that leaves it.

The change that would close the gap is not mysterious, and it is smaller than it sounds: say what the mandatory check must consist of, let the interval follow condition, and require the finding to be recorded. That is roughly what the United States adopted in 2023, and a shorter step here than it was there, because Malaysia already has the duty, the interval and the competent person written down. Only the substance is missing.

But you do not have to accept our version to test whether the gap is real. Ask your own facility manager when the main switchboard was last examined under load, and by whom, and what the record says. The law guarantees you that book exists and that somebody signed it. It guarantees you nothing whatsoever about whether anyone looked for the thing most likely to be going wrong — and if a year of monthly inspection reports contains no thermal record at all, that is the whole argument, sitting in a folder, fully compliant.

The figures above are Malaysian, dated, attributed and bounded. That does not make them the answer to your question — only your own installation answers that. It makes them checkable, which is the most any statistic can offer and the least you should accept from anyone quoting one at you, ourselves included.

Statistics reproduced with attribution — Sumber: Suruhanjaya Tenaga. Figures are drawn from the Laporan Prestasi Keselamatan Elektrik 2019 and describe cases investigated by the Commission over 2002–2019. Regulations 67, 68, 110 and 122 are quoted from the Electricity Regulations 1994 (P.U.(A) 38/94), made under the Electricity Supply Act 1990, as published by Suruhanjaya Tenaga incorporating the latest amendments — P.U.(A) 73/2014 and P.U.(A) 136/2014. Comparative figures are from Hartford Steam Boiler's published claims data, Martínez and Lagioia (CIRED 2007), and Paoletti and Baier's IEEE 493-derived analysis of medium-voltage switchgear.