Electrical predictive maintenance monitors the condition of equipment in service, without de-energizing it, to detect a defect when it starts and decide when to intervene before the failure. Its main techniques are thermography, oil and dissolved gas analysis, partial discharge, ultrasound and online monitoring. It does not replace preventive maintenance: it guides it, so that outages are scheduled where they are needed.
Preventive, predictive and corrective
Electrical maintenance talks about three approaches:
- Corrective: repair after the failure. It is the most expensive when the failure stops production or damages a long-lead-time piece of equipment.
- Preventive: intervene on planned dates, almost always with the equipment de-energized, whether or not there are symptoms.
- Predictive: measure the condition of the equipment in service and intervene when the data show it is starting to deteriorate.
Predictive maintenance starts from a simple idea: most electrical failures give warning. A splice that heats up, an oil that begins to generate gases or insulation with partial discharges evolve over weeks or months before failing. The interval between the first detectable symptom and the functional failure is represented by the P-F curve, which the 2026 edition of NFPA 70B incorporates as a method for setting maximum maintenance intervals.
- Incipient defect. Not yet detectable with available techniques
- Point P. The defect can be detected: heat, gases, discharges
- Trend. Repeated measurements show it is progressing
- Planned intervention. Corrected in a scheduled outage
- Functional failure avoided. No unplanned outage or major damage
The practical consequence: for predictive maintenance to work, the measurement must be repeated more often than the P-F interval of that defect.
Techniques and what each detects
| Technique | Equipment | What it detects | Condition to measure |
|---|---|---|---|
| Infrared thermography | Switchboards, switchgear, transformers, connections, capacitor banks | Loose or corroded connections, overloads, imbalances, poor ventilation | With enough load and visual access |
| Dissolved gas analysis (DGA) | Oil-filled transformers | Overheating, partial discharge, internal arcing, paper degradation | Oil sample with the equipment in service |
| Physical-chemical oil analysis | Transformers, tap changers | Moisture, dielectric strength, acidity, aging | Oil sample |
| Online partial discharge | Switchgear, cables, transformers, GIS | Insulation defects before breakdown | Energized equipment; sensors depend on the equipment |
| Ultrasound | Switchgear, insulators, switchboards | Corona, surface tracking, discharges | Energized equipment |
| Online monitoring | Power transformers, bushings, batteries | Continuous trends of gases, moisture, bushing capacitance, temperature | Installed sensors |
| Power quality | Whole plant | Harmonics, imbalances, voltage sags that age equipment | Analyzer connected for a representative period |
| Relay records | Protective relays | Self-diagnostic alarms, operations, breaker contact wear | Event download |
Transformer techniques are explained in dielectric oil analysis and online transformer monitoring; DGA interpretation normally follows IEEE C57.104 or IEC 60599. Thermography has its own guide: thermography in substations and switchboards. For motors, generators and fans, vibration analysis is added; what portable vibration meters and balancers measure is explained in CEMB N130, N330 and N600.
Predictive maintenance in switchboards
For switchboards and motor control centers, the most used predictive technique is thermography under load, because most low-voltage failures start in a connection that heats up. In medium-voltage switchgear, ultrasound and partial discharge are added, because a badly made cable termination or a contaminated insulator can fail without heating up first. On busbars and critical connections, or in switchboards where opening doors while energized exposes personnel, permanent temperature sensors are used more and more, monitoring without having to open the equipment.
Predictive maintenance in transformers
The oil-filled transformer is the equipment where predictive maintenance pays off most: the oil keeps the history of what happens inside. The trend of dissolved gases tells a mild overheating from an arc, and the evolution of moisture and acidity shows the aging of the insulation. In large or critical transformers, online monitoring makes it possible to see the trend every day. In dry-type transformers, thermography, temperature probes and partial discharge play that role. See partial discharge test.
What predictive maintenance does not see
| Defect | Why predictive does not see it | How it is detected |
|---|---|---|
| Slow breaker | No symptom while it does not operate | Operating time test |
| Degraded internal contacts | Depends on load; may not heat up | Contact resistance |
| Relay with a wrong setting | Works normally until the fault | Secondary injection test and settings review |
| Winding deformed after a short circuit | May not generate gases immediately | Frequency response analysis (FRA), short-circuit impedance |
| Open trip coil | Not noticed without supervision | Trip circuit supervision and functional test |
That is why predictive maintenance and de-energized tests complement each other. Predictive says where to look and when; tests confirm and quantify.
How to start in a plant
A predictive program does not start by buying sensors: it starts by knowing which equipment is critical and what its most likely failure mode is. With that, the technique and frequency are chosen for each one, a baseline measurement is taken and it is compared over time. The frequency must be shorter than the time the defect takes to go from detectable to failure. NFPA 70B offers that condition and criticality framework; see NFPA 70B. In a solar plant, the equivalent framework is IEC 62446-2; see solar plant maintenance to IEC 62446-2.
Frequent mistakes
- Measuring without load. A thermography with the plant at half load may not show the connection that fails at full load.
- No baseline measurement. An isolated value of gases, discharges or temperature says little; the trend says a lot.
- Data with no owner. A monitoring system that generates alarms nobody reviews costs money and prevents no failures.
- Trusting predictive maintenance for what it does not see. A breaker or a relay that is never tested can fail on the day of the fault even if all indicators are green.
- Findings with no closure. Detecting the hot spot and not scheduling the correction turns predictive maintenance into a record of announced failures.
How we help
At Tiza Group our team combines predictive techniques with the tests that confirm the diagnosis: thermography, analysis of oil results, partial discharge, and tests on transformers, breakers, cables and protection in medium-, high- and extra-high-voltage substations and industrial plants. Field work is in Colombia, or elsewhere on request. When a finding points to a piece of equipment, we propose the test that confirms it and the intervention in the outage.
See our testing and commissioning and thermography service, or write to us to put together the predictive program for your plant.
Related guides
- Electrical substation maintenance: what it includes
- Online transformer monitoring
- Partial discharge test
- Dielectric oil analysis
- How often should a substation be maintained
- Plant transformer failures
- Solar plant maintenance to IEC 62446-2
- CEMB N130, N330 and N600: on-site vibration and balancing
References
- NFPA 70B-2026, Standard for Electrical Equipment Maintenance.
- IEEE C57.104-2019, Guide for the Interpretation of Gases Generated in Mineral Oil-Immersed Transformers.
- IEC 60599, Mineral oil-filled electrical equipment in service – Guidance on the interpretation of dissolved and free gases analysis.
- IEC 60270, High-voltage test techniques – Charge-based measurement of partial discharges.
Frequently asked questions
What is electrical predictive maintenance?
It is maintenance that measures the condition of equipment while it operates, with techniques such as thermography, oil analysis, partial discharge or online monitoring, to detect incipient defects and plan the intervention before they become a failure. It differs from preventive maintenance, which intervenes on fixed dates, and from corrective maintenance, which acts after the failure.
Which predictive maintenance techniques are used on electrical switchboards?
Infrared thermography under load, which shows loose or overloaded connections; ultrasound, which detects corona, surface tracking and discharges in medium voltage; partial discharge in medium-voltage switchgear; and permanent temperature sensors on busbars and critical connections.
Does predictive maintenance replace tests with the equipment de-energized?
No. Some defects can only be seen with the equipment out of service: the operating time of a breaker, contact resistance, the ratio of a transformer or the response of a relay to a simulated fault. Predictive maintenance indicates which equipment needs those tests and when, and allows spacing out the tests of equipment that is in good condition.