Preventive Maintenance for HVAC Systems

Ask a facilities manager where the HVAC maintenance budget goes and the answer usually starts with the chillers. Ask where the complaints come from and the answer is almost always somewhere else: a meeting room that never cools, a floor with a stale smell, a diffuser that blows cold air over one desk all day. The central plant is expensive and visible, so it gets the attention and the specialist contract. The air side, thousands of filters, coils, belts, dampers, drain pans and terminal boxes spread across every ceiling void in the building, is where the service actually reaches the occupant, and it is where most PM programmes are thin, generic and quietly non-compliant. This guide is about that half of the system.

The message up front: on the air side of an HVAC estate, filters are the single highest-value routine task and almost everyone gets the change trigger wrong. Fixed-interval filter changes either waste money or waste energy, and usually both in different parts of the same building. Move the trigger to differential pressure, get coils and drain pans onto a real schedule, and accept that fan coil units are a volume problem rather than an engineering problem. Do those three things and the rest of the HVAC PM programme becomes manageable.

1. Scoping the HVAC estate: the air side is the bigger half

Scoping the estate honestly reveals a lopsided asset register. A mid-size commercial building might have three chillers and forty pumps. The same building will have twelve to twenty air handling units, six hundred fan coil units, three hundred VAV boxes, a dozen extract systems and several kilometres of ductwork. The central plant is a handful of high-value assets. The air side is thousands of low-value assets that collectively consume more labour hours than everything else combined.

That asymmetry drives the design of the programme. Central plant maintenance is specialist, scheduled, and often contracted to the OEM. Air side maintenance is high-volume, repetitive, and executed by in-house or general FM technicians. The routing logic, the parts strategy and the quality assurance all have to be different. Treating an FCU like a small chiller, with the same PM detail and reporting expectations, is how programmes collapse under their own weight.

This guide covers the air side. For the central plant, the chillers, boilers, cooling towers, pumps, water treatment and refrigerant compliance, see the companion piece on boiler and chiller preventive maintenance. The categories to get into the register here:

• Air handling units (AHUs): the central air side plant. Filters, coils, fans, belts and drives, dampers and actuators, drain pans, bearings, controls.
• Fan coil units (FCUs): room-level terminal cooling, usually in ceiling voids, usually in the hundreds or thousands.
• VAV boxes and terminal units: airflow regulation at zone level, often with reheat.
• Split and package units: standalone systems serving server rooms, guard houses, retail units and anything the central system does not reach.
• Ductwork and air distribution: supply, return, extract, diffusers, grilles, fire and smoke dampers.
• Ventilation and extract systems: toilet extract, kitchen extract, car park ventilation, smoke extract.
• Humidification and dehumidification: where installed, and often the most neglected sub-system in the building.

Every one of these needs a criticality rating before it needs a schedule, because a 600-unit FCU population cannot all be equally important. The FCU serving the data room is not the FCU serving the third-floor corridor. The asset criticality classification approach applies directly, and on the air side it is the difference between a schedule you can staff and one you cannot.

2. Air handling units: the anchor asset of the air side

The AHU is where the air side programme earns most of its value, because one badly maintained unit degrades conditions for an entire floor. It is also a multi-discipline asset: filtration, heat transfer, mechanical drive, controls and drainage in one box, each sub-system with its own failure modes and its own sensible interval. Filters are the highest-value routine task on the whole air side and are covered in section 5; if nothing else in the AHU schedule survives a budget cut, that is the task to protect.

Coils. A dirty cooling coil is a slow, invisible performance loss. Air-side fouling reduces heat transfer, so the unit runs longer or colder to hit setpoint, and where cooling is constant that shows up directly on the electricity bill. Coil condition also affects air quality, because a wet, dirty coil is a biological growth surface sitting in the supply air stream. Inspect at least quarterly, clean at least annually, and more often where the outside air intake sits in a dusty or construction-heavy environment. The practical marker: if you cannot see clearly through the fin pack with a torch from the other side, it needs cleaning regardless of what the schedule says.

Belts and drives. On belt-driven units, tension and alignment are an energy item as well as a reliability item. A slipping belt wastes energy and eventually strands the unit. Check quarterly, and keep a matched spare set on site for every belt size in the estate rather than ordering per incident. On direct-drive EC fan units the task disappears entirely, which is one of the quieter operational benefits of an EC retrofit.

Dampers and actuators. The most commonly failed and least commonly noticed component in an AHU. A fresh air damper stuck closed silently starves the building of outside air. A damper stuck open in a hot climate pulls unconditioned 45-degree air across the coil all day and nobody sees it except the energy meter. Actuators seize, linkages loosen, and position feedback drifts. Stroke every damper through full range twice a year and verify that blade position matches what the BMS reports. Feedback agreeing with reality is not something to assume.

Drain pans and condensate. In a humid coastal climate the condensate load is heavy and the drain path is under constant biological load. Blocked drains cause ceiling damage, staining, complaints and, at worst, microbial growth in occupied space. Clean and flush pans and traps quarterly at minimum, monthly in high-humidity areas, and verify the trap seal depth suits the fan pressure. An undersized trap on a negative-pressure section pulls air through the drain instead of letting water out, and the pan overflows even though the drain is technically clear.

Bearings and vibration. On large or critical AHUs, a periodic vibration reading gives real early warning of bearing degradation. On small units it is not worth the cost, and listening for roughness during the routine visit is the honest level of monitoring. For where continuous condition monitoring genuinely pays, the predictive maintenance and failure prediction guide sets out the triage.

3. An AHU preventive maintenance checklist you can lift

Below is a working AHU schedule with acceptance criteria, the part most published checklists leave out. A task without an acceptance criterion cannot be verified, audited or enforced against a contractor. "Check filters" is not a task. "Measure differential pressure across the filter bank and record; replace if above the manufacturer's final resistance" is a task. Adjust intervals for your climate, filtration grade and criticality.

Two structural points. First, the filter tasks are deliberately split between a monthly measurement task and a condition-triggered replacement, which is the change that makes the whole programme work. Second, several acceptance criteria reference a commissioned baseline. If you have no commissioning data, capture the values once at the start and treat those as the baseline, because without a reference number "clean" is an opinion. For checklist structures that survive audit, see preventive maintenance checklists, templates and examples.

4. Fan coil units: a volume problem, not an engineering problem

An individual FCU is a simple device: a small fan, a coil, a filter, a drain and a control valve. There is almost nothing technically interesting about maintaining one. The difficulty is entirely in the number of them and where they are. A building with 800 FCUs, most above ceiling tiles in occupied offices, meeting rooms and hotel bedrooms, presents an access and logistics problem that dwarfs the technical one.

The failure pattern I see repeatedly: the schedule specifies quarterly FCU service, the team can physically reach perhaps half the units in the time allocated, the rest get closed as complete without being touched, and two years later the estate has a population of units with blocked drains, filthy coils and seized valves that nobody can account for. The schedule was never realistic, so the data became fiction, and the fiction was worse than an honest lower frequency would have been.

What actually works on a large FCU population:

• Tier by criticality and exposure. Units serving data rooms, kitchens, clinical spaces and high-occupancy meeting rooms get the full frequency; corridor and low-occupancy units get less. That is allocating a fixed labour budget where it changes outcomes, not cutting corners.
• Route by location, not by asset list. Generate work orders as floor or zone routes covering every unit in a ceiling void rather than one per FCU. A technician already up the ladder should do all four units in that room. Per-unit work orders on an 800-unit population create an unmanageable administrative load and encourage batch closure. The work order types guide covers the route-based PM pattern.
• Accept that access blocks a proportion of units. Some sit behind fixed ceilings, above tenant equipment, or in rooms unavailable during business hours. Record those as deferred with a reason, not as complete. An honest 12 percent access-blocked figure is something you can fix; 100 percent completion is not.
• Standardise filters and parts. Three FCU filter sizes instead of eleven changes the economics of the whole operation, and consolidating mixed models at replacement time pays back quietly for a decade.
• Treat drain condition as the primary risk. The most damaging FCU failure in a humid climate is a blocked condensate drain overflowing into occupied space: ceiling damage, disruption, reputational cost, sometimes a mould claim. Drain flushing is the task to refuse to drop.

5. Filter strategy: grade, differential pressure and the energy cost of dirt

Filters deserve a section of their own because they are simultaneously the most routine task in the building and the one most consistently done on the wrong logic. Three decisions matter: what grade to fit, when to change it, and whether anyone is measuring the cost of getting it wrong.

Grade selection. Filtration is now classified under ISO 16890, which rates filters by the fraction of particulate removed in the PM1, PM2.5 and PM10 ranges, replacing the older EN 779 classes many specifications still reference. The practical implication is that a filter is rated against the particle sizes that matter for health rather than an abstract test dust. Typical commercial practice is a coarse pre-filter protecting a fine filter appropriate to the space. Higher grade means better air quality and higher pressure drop, which means more fan energy, so this is a genuine engineering trade rather than a case of specifying the highest number available. Two constraints bound it: the minimum grade the design standard requires for the occupancy, and the maximum pressure drop the fan can take without losing design airflow. ASHRAE publishes the ventilation and filtration guidance most commercial specifications are built on, and it is worth checking what your design actually assumed before changing grade.

The change trigger: differential pressure, not the calendar. This is the single most valuable change most HVAC PM programmes could make. A fixed interval treats every filter as if it loads at the same rate, which is demonstrably false. A filter on a ground-floor intake facing a construction site loads in weeks. A filter on a sheltered high-level intake on a recirculation-heavy unit may still be within its clean pressure drop after a year. Changing both on the same six-month cycle means changing one far too late and one far too early: wasted money on one, wasted energy and air quality on the other.

The method is straightforward: fit a dP gauge or transmitter across each filter bank, record the clean pressure drop when a new filter goes in, take the final resistance from the manufacturer's data, and raise the change work order when the measured dP reaches that threshold. The PM task becomes "measure and record dP" monthly, and the replacement becomes a condition-triggered corrective. Where the AHU is on the BMS that trigger can be automated end to end, and the BMS to CAFM integration architecture is exactly the pattern that turns a dP alarm into a work order without a human retyping it.

Bypass is the failure nobody logs. A clean, correctly graded filter that is not sealed into its frame filters nothing, because air takes the path of least resistance and goes round it. Gaps at the frame, warped cells, missing clips and collapsed media all produce bypass, and bypass does not show on the dP reading. It lowers it, which means a bypassing bank can look healthier on the gauge than a properly sealed one. Visual inspection of seating and seals has to stay in the schedule alongside the dP measurement, because the two checks catch different failures.

6. Split, package and standalone units

Nearly every estate has standalone units sitting outside the central system and outside the main maintenance contract: splits serving server and comms rooms, package units on retail or warehouse space, cassette units in guard houses and site offices. Individually low value, collectively significant, and the assets most likely to be missing from the register entirely.

Two things matter here. The first is having a complete register, because these units get installed by tenants, projects and contractors without ever being handed over into maintenance, so a physical walk-down survey is usually the first productive piece of work. The second is recognising that some are critical out of proportion to their cost. The 3kW split serving a comms room is a low-value asset with a high-consequence failure, and it belongs on a higher frequency with a spares strategy, possibly with redundancy and a temperature alarm to the BMS. Treating it as a small air conditioner because that is what it looks like is a criticality error.

The task set is compact: filter clean or change, indoor coil clean, outdoor condenser coil clean (the one that degrades fastest in dusty environments and the one most often skipped), drain and pump check, refrigerant pressures and superheat where certification allows, electrical connection check and running current, control and thermostat operation. On condenser coils in a desert climate I would expect quarterly cleaning as a baseline and monthly where the unit sits in a sand-exposed or traffic-heavy location. A blocked condenser raises head pressure, raises consumption and shortens compressor life, all at once.

7. Ductwork, air distribution and indoor air quality

Ductwork gets inspected least and affects the occupant most. It is hidden, expensive to access, and the consequences of neglect accumulate slowly enough that no single year makes the case for action. The tasks that genuinely belong in the schedule:

• Internal duct cleanliness surveys. Periodic inspection via access panels and camera against a defined cleanliness standard. Frequency depends on upstream filtration quality and the nature of the space. Where filtration is well maintained, ductwork stays clean for a long time; where it is not, duct contamination is the downstream symptom.
• Kitchen extract duct cleaning. A fire safety task, not a hygiene task, on a strict frequency driven by cooking hours and volume. Grease accumulation is one of the more serious fire risks in a commercial building and one of the few HVAC tasks where deferral creates immediate life-safety exposure.
• Fire and smoke damper testing. Statutory in most jurisdictions, commonly annual, and commonly the task with the worst completion evidence in the estate because dampers are hard to access. Missing damper test records are a standard audit finding.
• Diffuser and grille cleaning. Quick, visually obvious to occupants, and a useful proxy indicator: dust streaking around a diffuser means upstream filtration is not doing its job.
• Duct leakage and insulation. Leaking supply ductwork delivers conditioned air to the ceiling void rather than the room, which is both an energy loss and a comfort complaint. Damaged insulation in a humid climate produces condensation and ceiling staining.

On indoor air quality, expectations changed permanently after 2020 and have not reverted. Occupants, tenants and corporate clients now ask about ventilation rates, filtration grade and CO2 levels in a way they rarely did before, and that changes what an HVAC PM programme has to be able to evidence. In practice it means three things: fresh air rates measured and recorded rather than assumed, filtration grade documented per unit and justified against design intent, and CO2 sensors calibrated as a scheduled task rather than trusted indefinitely. An uncalibrated CO2 sensor driving demand-controlled ventilation is worse than no sensor, because it will confidently under-ventilate a full meeting room. The smart building and IoT monitoring guide covers where continuous environmental sensing adds value on top of scheduled checks.

8. Ventilation, extract and humidification

Extract systems are the quiet backbone of a building's air quality and are almost universally under-maintained, because nobody complains about an extract fan until the smell is obvious. Toilet extract, car park ventilation, kitchen extract and smoke extract each carry a different consequence profile, and only the last two typically get attention.

The schedule essentials: fan operation and current draw, belt and bearing condition, damper and louvre operation, grille cleanliness, and for car park and smoke systems the control sequence and CO detection interlocks. Car park ventilation deserves particular attention because it is frequently CO-controlled, the sensors drift, and a drifted sensor either runs the fans permanently (an energy cost) or fails to run them when needed (a safety exposure). Annual calibration and a functional sequence test are the tasks that matter.

Humidification is the sub-system I most often find abandoned. Steam humidifiers scale up, electrode cylinders exhaust, evaporative media fouls, and where humidity control is not a strict requirement the units quietly get isolated and forgotten while remaining on the asset register. If humidification is genuinely not required, decommission it formally and remove it from the schedule. If it is required, for archives, laboratories, clinical or certain data environments, it needs a real schedule including water quality management, because a neglected humidifier is a microbial risk in the supply air stream rather than a neutral piece of idle equipment. What is not acceptable is the common middle answer: a live asset with a schedule everyone knows is not being done.

In the Gulf the more common requirement is the opposite. Dehumidification load is heavy, particularly in coastal locations and buildings with high fresh air requirements, and where dedicated outdoor air or desiccant units are installed they carry their own maintenance set: wheel condition, seals, regeneration heater and drive. These are specialist assets, often with the thinnest local support, which is worth knowing before the failure rather than after.

9. Controls, BMS and the tasks people forget are maintenance

A large proportion of HVAC performance problems that present as mechanical faults are control faults, and almost none of them appear in a standard PM schedule. The unit is mechanically perfect, the sequence is wrong, and the building runs badly while every PM closes green. The control-side tasks that belong in the programme:

• Sensor calibration. Temperature, humidity, pressure, CO2 and airflow sensors all drift, and every control decision in the building rests on them. Annual verification against a calibrated reference is the minimum and the highest-value control task.
• Setpoint audit. Setpoints drift over years through overrides, complaint response and forgotten temporary changes. An annual review against design intent regularly finds units running at setpoints nobody can justify, and it is usually an energy finding.
• Override and manual mode review. Units left in hand after a repair are one of the most common hidden energy wastes in a building. A monthly report of everything not in auto takes minutes and pays for itself.
• Schedule verification. Occupancy schedules that no longer match how the building is used mean plant running in empty space. Review annually and after any tenancy change.
• Alarm hygiene. A BMS with hundreds of standing alarms is a BMS nobody reads. Clearing, re-pointing or retiring nuisance alarms is maintenance work on the control system itself.

Getting BMS alarms and trends to flow into the maintenance system rather than sitting on a separate operator screen is what turns this from a monitoring activity into a maintenance one. The integration pattern, and the realistic limits of it, are covered in the BMS and CAFM integration reference architecture.

10. Seasonal rhythm, and what it means in a hot climate

Most HVAC maintenance literature is written from a temperate perspective, with a clear seasonal structure: pre-summer readiness on the cooling side in spring, pre-winter readiness on the heating side in autumn, and two shoulder seasons when plant can be taken offline for major work. That structure is useful and it should shape the schedule where it applies.

Pre-summer readiness on the air side means everything affecting cooling capacity and airflow: coils cleaned, filters fresh, belts tensioned, dampers stroked, condensate drains proven, control sequences verified, and airflow measured against design on the units serving the spaces that will suffer first. It is deliberately front-loaded so the estate enters peak load in known condition rather than discovering problems through complaints. Pre-winter readiness covers heating coils, reheat, humidification, frost protection and changeover sequences, and in mixed-mode buildings it is when to verify that economiser and free-cooling logic actually works, because it has not been exercised for months.

In the Gulf this structure largely does not apply, and pretending otherwise produces schedules that do not fit the operation. Cooling runs effectively year round. There is no shoulder season in which an AHU serving occupied space can be taken out for a week. Peak load is extreme and sustained for months, and the period that most resembles a low season, roughly December to February, is short and still carries real cooling load in many buildings. The practical adaptations I would recommend:

• Concentrate heavy intrusive work into the cooler months, and plan it as a campaign with a named list of assets rather than assuming it will fit around routine work. That window is the only realistic one and it closes fast.
• Front-load pre-summer readiness to finish well before April, not during it. Coil cleaning in May on a unit already struggling at peak load is a complaint response, not preventive maintenance.
• Shorten dust-driven intervals. Filter loading, condenser fouling and coil fouling all run faster in a sand-exposed environment than any generic schedule assumes. This is where the differential pressure trigger proves its worth, because it adapts automatically to conditions a fixed schedule cannot anticipate.
• Build redundancy and temporary cooling into the plan. If a unit cannot be taken offline in summer, either it has standby capacity, or the work does not happen, or someone hires a temporary unit at short notice at a bad price. Decide which, in advance.
• Treat condensate and humidity management as a year-round priority, not a summer task. Coastal humidity keeps drain pans and insulation under load for most of the year.

The broader point: a maintenance schedule imported from a temperate-climate template will be wrong in both directions here, over-specified on heating-related tasks that barely apply and under-specified on dust, condensate and continuous-duty items that dominate. For the general frameworks that these adaptations sit inside, see preventive maintenance strategies and the broader complete guide to preventive maintenance.

11. The energy business case: why HVAC PM funds itself

HVAC preventive maintenance is unusual because its business case does not rest solely on avoided failure. It rests on consumption. HVAC is typically the largest single energy load in a commercial building, and in a hot climate it dominates. Every degradation mechanism the programme addresses carries a continuous energy cost while it persists:

• A loaded filter raises fan power every hour it stays in.
• A fouled cooling coil reduces heat transfer, so the unit runs longer or the water temperature has to drop to compensate, pushing load back onto the chiller plant.
• A fouled condenser on a split or package unit raises head pressure and compressor power directly.
• A slipping belt wastes shaft power.
• A fresh air damper stuck open pulls unconditioned outside air across the coil continuously, which in a 45-degree ambient is a very expensive fault.
• Leaking ductwork delivers cooling to a ceiling void.
• A drifted sensor or a unit left in manual runs plant that should be off.
• An uncalibrated CO2 sensor either over-ventilates, importing hot outside air, or under-ventilates, creating an IAQ problem.

Framing the programme this way changes the budget conversation. A manager arguing for filter changes on reliability grounds is arguing for a cost. A manager arguing for a differential-pressure-triggered filter regime on energy grounds is arguing for a saving, and can point at fan power readings before and after. The measurement that makes it credible is already in the checklist above: fan motor current and power quarterly, dP monthly. Those two series over a year are the evidence base.

I would avoid quoting industry-average savings percentages, because they vary enormously with building type, climate and how degraded the estate was to begin with. Measure your own: take fan power and dP on a representative sample, run the changed regime for a season, compare. A number from your own estate survives challenge in a budget meeting in a way a vendor whitepaper figure does not. For turning that evidence into reported performance, the FM KPI framework covers how to structure the measures so they hold up.

12. The mistakes I see most often on HVAC programmes

• Generic imported schedules. A template written for a temperate climate applied unchanged to a Gulf estate, complete with heating-season tasks and filter intervals that are wrong by a factor of three.
• Fixed-interval filter changes. Covered at length above. The most expensive default setting in HVAC maintenance.
• FCU schedules that were never achievable. A frequency nobody costed against available labour hours, producing fictional completion data and a hidden backlog.
• Tasks with no acceptance criteria. "Check", "inspect" and "clean as required" cannot be failed, so they are never failed.
• Missing standalone units. Splits and packages installed by projects and tenants that never reached the asset register and are maintained by nobody until they fail.
• Ignoring the control layer. Perfect mechanical maintenance on units running wrong setpoints, wrong schedules and stuck in manual.
• Damper and actuator neglect. Invisible, energy-expensive, and almost never caught without a deliberate stroke test.
• No baseline. No commissioned airflow, no clean filter dP, no reference fan power, so no way to prove anything has degraded or improved.

None of these is technically difficult to fix. They are schedule design and data discipline problems, within the control of the maintenance organisation rather than dependent on capital spend. Structured task libraries such as SFG20 are a reasonable starting point for task content, provided the intervals are adapted to local climate and duty rather than adopted verbatim. For worked examples of how schedules differ by sector and operating environment, see preventive maintenance examples across industries.

The idea to walk away with

HVAC preventive maintenance succeeds or fails on the air side, not in the plant room. Filters, coils, drains, dampers and controls determine what the occupant experiences and what the building consumes, and they are the tasks most often written vaguely, scheduled unrealistically and closed without evidence. Fix three things and most of the programme follows: change filters on measured differential pressure rather than the calendar, write every task with an acceptance criterion a technician could fail, and set FCU frequencies against the labour budget you actually have rather than a frequency that sounds thorough.

The energy argument is what keeps the programme funded. HVAC degradation costs money continuously rather than at the moment of failure, which makes air side PM one of the few maintenance activities defensible on consumption as well as reliability. Measure fan power and filter dP, keep the series, and you will have the evidence when the budget question comes.

Final thoughts

The estates I have seen run well on HVAC are not the ones with the most detailed schedules. They are the ones where the schedule matches the labour available, the tasks have numbers attached, the standalone units are actually on the register, and somebody looks at the control layer as seriously as the mechanical one. Unglamorous, entirely achievable, and better than an elaborate schedule reporting 100 percent completion.

In a hot climate the margin for error is smaller. Cooling runs all year, dust loads faster than any generic interval assumes, and there is no quiet season to catch up in. That argues for condition-based triggers wherever they can be instrumented, for planning intrusive work deliberately into the short cooler window, and for being honest about what a fixed headcount can cover across several hundred terminal units. Honest coverage with good data beats optimistic coverage with fiction.

Related reading: Preventive maintenance: the complete guide, Boiler and chiller preventive maintenance, PM checklists, templates and examples, Preventive maintenance strategies, BMS and CAFM integration architecture, FM KPI framework.