Electrical Reliability, Explained Plainly.

What does an infrared electrical inspection actually find?

Overwhelmingly, loose and high-resistance connections — they are the majority of findings on almost every survey. A terminal that reads hot while the conductor feeding it stays cool is the signature, and it is nearly always a joint that was torqued wrong at installation or has loosened through thermal cycling.

Beyond that: overloaded circuits, which show as a whole conductor run reading warm rather than a single point; phase imbalance, where one phase is materially hotter across a length of bus; failing components such as pitted contactor tips, degraded fuse clips and breakers heating internally; load-side problems where the heat traces back to the equipment being fed; and installation defects like undersized lugs or mixed conductor materials without the right connector.

What it does not find is anything that is not carrying load at the time of inspection, or anything behind a cover the imager cannot see through. Both are worth knowing before commissioning a survey, and both should be stated in the report.

How often should industrial electrical equipment be inspected?

Annual infrared inspection is the common baseline for industrial distribution equipment, and for a lot of facilities it is the right answer. Where it is not enough is on equipment that is critical to production, operating close to its rating, in a dirty or hot environment, or already carrying a known condition — those warrant semi-annual or tighter.

The better way to set the interval is by criticality, loading, environment and condition history rather than a single plant-wide number. A switchboard in a clean, climate-controlled electrical room feeding non-critical load does not need the same attention as an MCC on a hot mezzanine feeding the main line.

What matters more than the exact interval is that it is consistent. Two inspections a year apart, at comparable load, produce a trend. Two inspections at random intervals under different conditions produce two unrelated snapshots.

What is power quality monitoring?

Connecting a three-phase analyser to the electrical system and logging what it is actually delivering — voltage, current, imbalance, harmonics, sags, swells, transients, flicker, frequency and neutral current — continuously over days or weeks.

The reason it takes days rather than minutes is that the problems worth finding are intermittent. A spot reading with a meter tells you the system looked fine at that moment, which you probably already knew. Continuous logging captures the event with its waveform and timestamp, which is what makes it diagnosable.

The output should answer a specific question: is power quality contributing to the problem you reported, and if so what is causing it? A monitoring report that presents graphs without answering that has not finished the job.

What causes industrial equipment to trip for no apparent reason?

The word “apparent” is doing the work in that question — there is a cause, it is just not visible from the panel. The usual candidates:

• Voltage sags from large motors starting across the line, dragging the bus down far enough to drop contactors or fault drives elsewhere in the plant.
• Harmonic distortion from drives, rectifiers and switching supplies, heating transformers and neutrals and distorting the voltage everyone else is fed.
• Voltage imbalance, which produces a much larger current imbalance in a motor than the voltage figure suggests.
• Transients from capacitor switching, utility events or inductive load switching.
• Excess neutral current from triplen harmonics or unbalanced single-phase loading.
• Loose connections producing intermittent high resistance and localised heating.
• Miscoordinated protection, where a downstream fault opens an upstream device because the settings were never reviewed.

The pattern is usually the clue. A fault that happens at the same time each day, or only when a particular machine cycles, is telling you where to put the analyser.

Why do electrical one-line diagrams become inaccurate?

Because drawing revision is the step that gets dropped, and it gets dropped for understandable reasons. The line had to run Monday. Updating a drawing does not make product. The change was small and someone marked up the print in the electrical room, intending to have it redrawn.

Twenty years of that produces a document that is reliable about the parts of the system that never changed and wrong about the parts that did — which is exactly backwards, because the parts that changed are the parts people need the drawing for.

The failure mode that matters is not the inaccuracy itself but the misplaced confidence. A drawing that is mostly right and silent about which parts are not gets trusted for switching decisions and study modelling. Explicitly marking unverified sections is what makes an imperfect drawing safe to use.

Transformer maintenance basics

For dry-type units, most of the value is in three things: keep the ventilation clear, keep the windings clean, and know the actual loading. Dust accumulation and blocked airflow are the common killers, and both are visible on a walkthrough. Loading is the one that surprises people — a unit sized correctly fifteen years ago is often feeding considerably more than it was specified for.

For oil-filled units, oil condition is the primary indicator. Dissolved gas analysis is the most informative test available, and its value is almost entirely in the trend: one sample gives you concentrations, a series gives you generation rate, and rate is what separates a unit that has always run slightly gassy from one that is actively degrading. Alongside that: leaks, bushing condition, gauge indications and cooling performance.

For both, infrared inspection under load and verified connection condition apply. And for both, the record matters more than any single reading.

What is the difference between preventive and predictive electrical maintenance?

Preventive is interval-based — the task happens because the interval elapsed. It is necessary, and on its own it means servicing some equipment that did not need it while missing problems that developed between intervals.

Predictive is condition-based — the task happens because measured data says the asset is changing. Thermography, power quality monitoring, oil analysis and loading trends are all predictive inputs.

A good program runs both. The preventive schedule covers what must be done regardless; the predictive data adjusts priority within it and tells you which scheduled items should be pulled forward. What makes that possible is the record, which is why documentation is not the administrative overhead of a maintenance program — it is the mechanism.

NFPA 70B and industrial electrical maintenance

NFPA 70B is the recommended practice for electrical equipment maintenance. Its 2023 edition moved from a recommended practice to a standard, which has raised its profile considerably with insurers and auditors.

In practice it gives a facility three useful things: a structure for what an electrical maintenance program should contain, a basis for setting intervals by equipment type and condition, and a documentation framework that produces evidence as a by-product of doing the work.

Worth being precise about what alignment means. Premier develops and documents programs aligned with the standard’s recommended practice. Premier does not certify compliance and is not an authority having jurisdiction — and anyone offering to certify your NFPA 70B compliance is worth a second question.

Signs an industrial transformer needs attention

• Operating temperature higher than the loading justifies — often the first measurable sign.
• Measured loading at or above nameplate, particularly where drive content has grown since installation.
• Blocked, dust-loaded or obstructed ventilation on dry-type units.
• Oil leaks at gaskets, valves, radiator seams or bushing bases.
• Cracked, chipped, contaminated or tracking bushings.
• Discolouration, charring or other evidence of past overheating.
• Unusual noise, or a change in the noise it has always made.
• A rising trend in dissolved gas analysis results on oil-filled units.
• Connection heating visible on thermal inspection.

The first two are the ones most often missed, because neither produces anything you can see or hear from across the room.

What an arc flash study tells a plant

It calculates the incident energy available at each equipment location and the resulting arc flash boundary. That determines the labelling and the PPE required to work on that equipment, which is the compliance-driven reason most studies get commissioned.

The more useful output is often incidental to that. A study frequently identifies locations where a protective device settings change would materially reduce the incident energy — reducing the hazard rather than just documenting it. It also tends to surface coordination problems and equipment whose interrupting rating is below the available fault current, both of which matter independently of labelling.

The caveat is that a study is only as good as the model, and the model is only as good as the field data. A study built from unverified drawings produces confident numbers that are wrong, on labels people then rely on.

How predictive electrical maintenance reduces downtime

The mechanism is simple: it converts unplanned outages into planned ones. A connection found hot in March is a scheduled repair during the next outage window. The same connection found in July, by failing, is an unplanned stoppage at whatever moment suits it least.

The economics are usually not close, because the comparison is not the cost of inspection against the cost of the repair — the repair is the same either way. It is the cost of inspection against the cost of unplanned downtime, plus the collateral damage a failure causes to equipment that a planned repair would not have touched.

The second effect is less obvious and often larger: better allocation. A plant with condition data spends its electrical budget on the equipment that is actually at risk, rather than on whatever is visible or whatever failed most recently.