The case for continuous monitoring is usually made in the abstract.
Periodic inspections only show a moment in time. As we know things happen between them. Monitoring closes the gap.
All true, and all fairly easy to nod along to without changing anything.
The more useful question is quantitative. If something is failing, how long before it fails could you have known?
If the answer is a few hours, continuous monitoring is an alarm system.
If it’s several months, it’s a planning tool, and those are very different propositions.
Cables: Weeks to months of warning
A study of live partial discharge monitoring on paper-insulated 11 kV cables in London found that 12 of 17 cable sections showed clearly rising PD activity between 5 and 150 days before failure.
Two things stand out.
- Nearly three quarters of these failures announced themselves in advance.
- The warning window was wide: five days at the short end is an emergency, 150 days is a planned outage during a scheduled shutdown.
Insulation type changes the arithmetic considerably.
Reviewing the field evidence, researchers publishing in IEEE Transactions on Dielectrics and Electrical Insulation, concluded that a weak spot in paper-insulated cable takes years, on average, to develop into a fault, while in XLPE it can be a couple of days to months.
That has a practical consequence.
On a site with mixed-vintage cabling, the same monitoring regime provides generously long warning on the older assets and a much tighter window on the newer ones. Measurement frequency should reflect that, and often doesn’t.
Motors and Drives: Months on the supply side, weeks on the machine
A 2023 IEEE paper on motor current signature analysis at a refinery set out detection lead times by failure mode. Broadly:
- Months of warning for supply-side problems (voltage and current unbalance, harmonic distortion, voltage drops) and for broken or loose rotor bars and rotor eccentricity.
- Weeks for stator winding shorts, winding looseness, misalignment, mechanical unbalance and bearing degradation.
The paper includes a case where a broken motor shaft was visible in the current signature approximately two weeks before it developed into a failure, and another where a rising deviation triggered an inspection of cable connections that prevented a motor failure outright.
The pattern is consistent: electrical precursors give more notice than mechanical ones, and supply-side problems give the most notice of all, which is fortunate, because they’re also the ones most likely to be damaging several assets simultaneously.
HV Switchgear: The warning exists but not during normal operation
Partial discharge activity inside HV switchgear, cable boxes and terminations develops over extended periods, and it is effectively undetectable while the asset is running normally.
A visual inspection finds nothing. The asset performs correctly. The deterioration continues.
The HSE’s guidance on switchgear safety is direct on the underlying risk: owners should not assume that historical reliability will continue indefinitely, and should have a strategy in place for when equipment reaches the limit of its design life, noting that manufacturers typically quote 20 years for vacuum switchgear, and that older equipment than that remains in service.
This is the category where periodic inspection is weakest. Not because inspections are poorly done, but because the failure mode doesn’t present visually and the asset gives no operational symptom until it’s close to the end.
What a snapshot cannot capture
- A voltage dip lasts a fraction of a second.
- Harmonic distortion may only appear when a particular load is running.
- A UPS alarm may be a response to something upstream that had resolved before the engineer arrived.
- Partial discharge activity varies with humidity and load.
An inspection window of a few hours every one to three years will capture almost none of that. It will confirm the asset was acceptable on the day, which is worth having, and is not the same as knowing the direction of travel. Which is the actual difference: an inspection gives you a state, and continuous measurement gives you a gradient. A single event means little. The same event twelve times in six weeks, always at shift changeover, is a diagnosis.
Two examples from our own deployments
A UK university, HV transformer under suspicion.
- From the initial call to live data on a monitoring dashboard took two days.
- A temporary nine-channel continuous partial discharge monitoring system was installed with no shutdown required, and ran for eight months.
- The dataset supported deferring a transformer replacement rather than committing capital on the basis of uncertainty, and gave the engineering team a defensible reason for the decision.
A UK food manufacturer, HV substation.
- Continuous environmental monitoring raised a high humidity alarm in the switchroom and a transformer cable box.
- High humidity in that setting increases partial discharge activity and accelerates equipment degradation, so the panel heater was activated.
- The follow-up measurement showed a 20% drop in average relative humidity.
The second example is deliberately small. It’s a heater switch, not an engineering project. But it’s the whole mechanism in miniature: continuous measurement identified a condition nobody would have seen on an inspection, a low-cost intervention followed, and the same measurement confirmed it worked.
What continuous monitoring shows that a snapshot does not
Depending on the assets covered, the parameters worth tracking over time are:
- load profile
- peak demand
- voltage quality
- transient events,
- phase imbalance
- harmonic distortion
- power factor
- transformer and UPS performance
- asset temperature
- partial discharge activity
Individually, each is useful. The value is in the correlation, and correlation is only available if the measurements are continuous and share a common clock.
Monitoring may show equipment resets coinciding, to the second, with voltage dips.
A changing load profile may reveal a transformer moving toward its practical limit.
Harmonic distortion may explain overheating on a board that has been investigated twice already.
It’s the reason we look at the system rather than the device: the single most useful piece of information about an event is usually what else was happening at the same instant.
What this changes in practice
Better timing, mostly, which is worth more than it sounds.
Manufacturers typically recommend maintenance on LV/MV equipment and transformers every three years in normal conditions. Condition evidence lets you depart from that interval deliberately in either direction: bringing work forward on an asset that’s deteriorating, or extending it with confidence on one that’s stable, scheduling around production rather than around the calendar.
For assets where deterioration is measurable months in advance, that turns an emergency into a line item in the next shutdown plan. That is the whole return.
See eight months of continuous partial discharge data used to defer a transformer replacement
Acteniq designs, installs and operates continuous monitoring on critical electrical infrastructure, then interprets the evidence and delivers the maintenance, remedial work or modernisation it points to. Work with engineers, not an account team.
Sources cited in this piece
- Walton, CIRED 2001, via Zhang et al., “Review on Detection and Analysis of Partial Discharge along Power Cables”, Energies 14:7692, 2021 (open access)
- Steennis et al., “Guarding MV cables on-line”, IEEE Transactions on Dielectrics and Electrical Insulation, 23, 2016
- Sadlon, Frey & Arnold, “Predictive Motor Failure Prevention in Refinery Using Cloud Based Motor Current Signature Analysis”, IEEE PCIC 2023, DOI 10.1109/PCIC43643.2023.10414329
- HSE, Keeping electrical switchgear safe, HSG230 2nd edition, 2015
- Schneider Electric, Benefits of shifting from traditional to condition-based maintenance in electrical distribution equipment, 998-22447106_GMA, 2022
- Acteniq deployments: UK university PD monitoring; UK food manufacturer environmental monitoring
