Preventive maintenance on an uninterruptible power supply is not a single annual service call. In well-run programs it is a layered discipline: operator walk-throughs, standards-based battery inspections, scheduled OEM-level service visits, firmware and alarm management, environmental control, and planned replacement of wear parts. The objective is always the same: keep the UPS able to carry the critical load during a real utility disturbance, not merely appear healthy while the utility is up.
This guide condenses what the primary sources actually say: the IEC 62040 equipment standards, the IEEE battery-maintenance standards, and the published maintenance scopes of Schneider Electric, Eaton, Vertiv and Mitsubishi Electric.
The standards map, in one minute
The landscape splits cleanly into two domains. Equipment safety and performance sit under IEC 62040: part 1 covers UPS safety during use, service and maintenance, and part 3 covers the performance and transfer requirements the system must keep meeting. Battery maintenance sits under the IEEE standards, and the right one depends on chemistry: IEEE 450 for vented lead-acid, IEEE 1188 for VRLA, IEEE 1106 for vented nickel-cadmium, and IEEE 2962 for stationary lithium-ion. For mixed fleets, one UPS PM schedule can apply to the power electronics, but the battery cadence and test methods should branch by chemistry.
How often should a UPS be maintained?
The manufacturer baseline is consistent in direction, not identical in detail. Eaton recommends at least one preventive maintenance visit per year under compliant conditions. Vertiv explicitly recommends two visits per year for many mission-critical installations: one online visit with infrared scanning and one annual offline inspection. Mitsubishi calls annual service the practical minimum and recommends semiannual service for many systems, moving to quarterly inspections as equipment ages.
For a site with no special constraints, the defensible program is risk-tiered:
| Cadence | Activity |
|---|---|
| Daily / weekly | Operator walk-through: display status, alarms, unusual noise or heat, room condition |
| Monthly | Review alarms, environmental conditions and battery status; check air filters in dusty sites |
| Quarterly | Battery and environmental trending review (impedance or conductance drift, temperature) |
| Annually | Full OEM-level PM: thermal scan, internal inspection, capacitor and fan checks, firmware evaluation, torque verification |
Move to semiannual UPS PM when any of these apply: higher criticality, an aging fleet, harsh or dusty environments, VRLA strings without continuous monitoring, or frequent transfers, alarms and generator events. Raise cadence on observed risk, not calendar dogma.
What a real PM visit covers
Operator checks and service PM are different activities. The annual service visit, done properly, includes live load and transfer-time verification, inspection of AC and DC capacitors for leakage or bulging, fan and cooling-path checks, connection torque and thermal imaging for hot spots, battery testing to the applicable IEEE practice, environmental verification, firmware and field-bulletin review, and a written report with trended data. Exact bypass sequences, breaker names and test limits always come from the model-specific manual: that part is never generic.
One nuance most programs get wrong: capacity testing. IEEE-based practice recommends periodic performance tests at intervals no greater than 25 percent of expected service life (for VRLA, no more than every two years or 25 percent of life, whichever is less). That does not mean every annual PM should include a full discharge test. Schneider explicitly warns that repeated battery testing and runtime calibration can shorten battery life. Trending impedance beats stress-testing marginal strings.
The five failure families, and the signs that show up first
- Battery degradation: the largest single risk, because a UPS can look normal on utility power after the battery has already lost its reserve time. Early signs: rising impedance, elevated battery temperature, swollen or leaking cells.
- Capacitor aging: visible leakage or bulging, unstable DC bus, nuisance alarms. Vertiv plans AC and DC filter capacitor replacement at roughly 12 to 15 years; Mitsubishi cites about 15.
- Cooling-path deterioration: dirty filters, fan alarms, chronic hot spots. Vertiv suggests filter attention as often as every two months in normal conditions, monthly in dusty ones.
- Connection and transfer-path problems: hot connections under thermal scan, repeated transfers to bypass.
- Control and firmware issues: mismatched firmware between parallel units, incompatible network cards, stale field bulletins. In modern fleets, firmware hygiene is a PM item, not an IT afterthought.
If your unit is already showing alarm codes, our troubleshooting guides decode them: APC UPS fault codes and Eaton UPS alarm codes.
Getting the longest life out of the batteries
Battery-life optimization is where PM pays off fastest, and the cross-manufacturer tactics are consistent. Keep the battery room in the proper band: Eaton points to about 25 °C for full battery life, Vertiv optimizes longevity at 22 to 25 °C, and the familiar rule holds that service life roughly halves for every 8 to 10 °C above 25 °C. Avoid unnecessary deep-discharge tests. Trend impedance rather than reacting to single data points. Keep airflow clean. Most importantly, plan replacement before end of life: for VRLA in 24/7 duty that means scheduling at year 3.5 to 4, and ordering replacement UPS batteries before the string starts cascading alarms. For APC fleets specifically, assembled RBC replacement cartridges make the swap a planned one-hour task instead of an emergency.
Records and the KPIs worth tracking
A PM program is only as good as its records. Keep a permanent asset history for each UPS (metered values, alarm history, mode of operation, filter dates, firmware level, recommendations) and a separate chemistry-aware battery log. The KPI set that matters: PM completion on schedule, overdue corrective actions, battery exceptions by severity, environmental excursions, repeat alarms, maintenance-bypass exposure time, firmware alignment, and the percentage of lifecycle replacements completed before their risk window closes.
Budgeting without surprises
OEMs do not publish universal PM prices because effort depends on size, redundancy, chemistry, bypass availability and travel. Budget in five buckets: scheduled labour, travel and after-hours premium, specialist test equipment, a planned wear-part reserve, and a corrective-action reserve. The last two are what keep inexpensive PM from turning into expensive emergency work: batteries, capacitors, fans and filters are expected lifecycle costs, not surprises.
Where to go from here
For the brand-by-brand view of the same discipline, see our essential UPS maintenance guide for APC, Eaton, Powerware, Tripp Lite and Liebert systems, and if you are validating whether your runtime still meets the requirement, the UPS runtime calculator gives a quick first answer. GDF Technologies delivers preventive maintenance programs on all major UPS brands across Canada, with the trending, reporting and lifecycle planning described above.
How often should a UPS be cleaned in a dusty plant?
In a clean server room, an internal clean once a year alongside the scheduled service visit is enough. In a food, bakery, flour, wood, cement or agricultural plant, intake filters need checking monthly, cleaning or replacement quarterly, and a full internal clean twice a year. The trigger is airflow restriction rather than the calendar: once filters load up, internal temperature climbs and battery and capacitor life fall with it.
Cleaning intervals by environment
| Environment | Filter check | Filter clean or replace | Internal clean |
|---|---|---|---|
| Office, clean server room, data centre | Annually | Annually | Annually |
| Warehouse, light industrial, mechanical room | Quarterly | Semi-annually | Annually |
| Food, bakery, flour, wood, cement, agricultural, mining | Monthly | Quarterly | Twice a year |
| Machining, welding, foundry (conductive dust) | Monthly | Quarterly | Twice a year, with insulation checks |
These are the intervals we schedule, not a published standard. Set them against your own site by measuring rather than guessing: log the intake and exhaust temperature at each visit, and shorten the interval when the gap between them starts widening. That delta is the earliest usable warning that airflow is being lost, and it moves well before any alarm does.
One practical note on method. Internal cleaning is done with HEPA vacuum extraction, not compressed air, which relocates contamination rather than removing it. The work requires the unit shut down or on maintenance bypass, so it belongs inside a scheduled service visit rather than a separate cleaning call.
