A switchboard doesn’t usually fail all at once. It runs a little hot for weeks, sometimes months, while a connection loosens fractionally under normal vibration and load. Nobody notices, because nothing looks wrong.

Then it does fail, on a Tuesday afternoon, and a production line goes dark for three days while an electrician traces the fault back to a termination that was quietly cooking itself since spring.

That gap, between the point a fault becomes physically detectable and the point it becomes a problem, is where predictive maintenance lives.

Get the timing right and it’s a five-minute fix during a scheduled shutdown. Get it wrong and it’s an insurance claim.

What Predictive Maintenance Actually Means

There are really three approaches to keeping equipment running, and most facilities use some blend of all three without naming them. Reactive maintenance is fixing things once they break. Preventive maintenance is replacing or servicing parts on a calendar, regardless of their actual condition. Predictive maintenance is the newer, more useful idea: use real condition data to catch a fault while it’s still developing, so the fix happens exactly when it’s needed rather than too early or, more expensively, too late.

For electrical equipment, this approach earns its keep because so many failure modes are invisible to a walk-through inspection. Loose terminations, degrading insulation, unbalanced loads, none of these announce themselves to the eye. What they do announce, consistently and measurably, is heat. Long before a fault becomes audible or visible, it’s already showing up on a thermal camera.

The Two Data Sources That Actually Matter

Strip away the vendor jargon and a working predictive maintenance program rests on two kinds of information.

The first is continuous condition data: vibration analysis on rotating equipment, oil analysis on gearboxes and hydraulics, and increasingly a layer of IoT sensors streaming readings back to a maintenance platform around the clock. This builds a baseline over time, so a deviation actually means something instead of being noise.

The second is periodic, targeted inspection, and for electrical systems specifically, that means thermal imaging services. A proper thermal inspection captures infrared images of switchboards, cable terminations and transformers while they’re under normal operating load, and flags anomalies, hot spots that shouldn’t be there, that point to loose connections or insulation breakdown. Facilities running these scans annually or every two years tend to catch problems at the “tighten a bolt” stage. 

The ones that skip it find out at the “replace the switchboard” stage instead.

Neither source is much good on its own. Sensor data tells you something’s changed. Thermal inspection tells you where, and usually why. What connects the two is a review cycle someone actually owns, because a folder of reports nobody reads isn’t a maintenance program, it’s paperwork.

Building the Program, Step by Step

Asset criticality comes first, and it’s not the same thing as asset cost. A switchboard feeding a cold storage room and a lighting circuit in a stationery cupboard are not the same risk, even if they’re wired to a similar budget line. Rank by consequence of failure, not by replacement price, and the rest of the program falls into place around that ranking.

Inspection frequency follows from criticality rather than from habit. High-load, continuously running assets, manufacturing lines, data centres, cold storage, generally warrant annual or more frequent thermal inspection. A standard office electrical load can usually stretch to a two or three year cycle. None of this should be decided fresh each year by whoever happens to be free; it belongs in the documented plan, not in someone’s memory.

Then there’s ownership, which is where a surprising number of otherwise sound programs quietly fall apart. Reports get generated, findings get flagged, and then nothing happens because no single person is accountable for closing the loop. Someone needs to receive the report, action what it says, and log that it was done. 

That documentation, asset ID, load conditions at the time, thermal and visual imagery, severity rating, matters for another reason too: if something does go wrong later, it’s what shows an insurer or a regulator that the risk was being actively managed, not ignored.

The Budget Conversation

Here’s the part nobody enjoys. A thermal inspection program shows up as a line item. Downtime doesn’t show up anywhere, until it does, and by then it’s not a line item, it’s a crisis.

Faulty electrical equipment remains one of the leading causes of commercial fires in Australia, and insurers have noticed. 

QBE, for one, recommends periodic thermographic scans on switchboards as part of its guidance to commercial policyholders. Framed against that backdrop, the inspection cost isn’t really an added expense. It’s closer to a renewal fee on the continuity planning and insurance cover a business already relies on.

Where Programs Go Wrong

A single scan is a snapshot, not a program. The value is in the trend line that builds up over repeated inspections, so treating thermal imaging as a once-off tick-the-box exercise defeats the purpose before it’s even started.

Skipping the reporting loop is the quieter failure. Data that nobody reviews and nobody actions isn’t predictive maintenance; it’s an expensive photo library sitting on a shared drive.

And certification matters more than it might seem. 

Australian thermography standards define tiered competency levels for a reason, there’s a real gap between someone who can spot a warm patch on a screen and someone who can read that patch in the context of load, system design and actual risk. For anything beyond a basic screening pass, that gap decides whether the findings are actually usable.

None of this is complicated in principle. Rank the assets, match the inspection method and frequency to the risk, give someone the job of closing the loop. The facilities that treat it this way, as an ongoing discipline rather than an annual formality, are the ones that end up tightening a connection instead of explaining a fire.