Boiler and Chiller Preventive Maintenance

Walk into almost any central plant room and ask to see the PM schedule for the boilers and chillers, and you will usually be handed something that looks correct: quarterly service, annual shutdown, statutory inspection, a vendor contract. Then ask to see the operating log for the last ninety days, and the mood changes. Either the log does not exist, or it exists as a clipboard nobody has read since it was filled in. That gap is the whole subject of this article. Central plant preventive maintenance done on the calendar alone is a ritual. The same PM schedule plus disciplined log readings and trending becomes condition insight, and the difference shows up as fewer emergency call-outs, lower energy cost per unit of heat or cooling, and a plant that reaches its design life instead of being replaced early.

The message up front: the single highest-value preventive maintenance activity on a boiler or a chiller is not a task at all, it is a reading. Flue gas oxygen and stack temperature on a boiler, condenser and evaporator approach temperatures on a chiller, recorded consistently and trended over months, will tell you about developing problems long before any quarterly service finds them. Build the PM schedule around the readings, not the other way round.

1. Why central plant PM is a different discipline

Most of a facility's asset register can be maintained competently with a well-designed task list and a sensible frequency. Fire dampers, door closers, lighting, small pumps: inspect, clean, test, record. Central plant does not behave that way, for four reasons that are worth stating plainly before you design any schedule.

• Consequence is concentrated. A building typically has one or two boilers and two to four chillers carrying the entire load, with less redundancy than people assume at design conditions. When central plant fails, the whole facility feels it within hours.
• Degradation is gradual and invisible. Scale on a boiler tube, fouling in a condenser bundle, a slow refrigerant leak, a burner drifting out of tune: none of these announce themselves. They show up first as a two-degree shift in an approach temperature or a one percent change in flue oxygen, which is exactly why the log matters.
• Energy cost dwarfs maintenance cost. On a large chiller or a continuously fired boiler, the annual energy bill is usually an order of magnitude larger than the maintenance spend. A plant running a few percent below its achievable efficiency wastes more quietly every month than the PM contract costs.
• Parts of the work are statutory. Pressure vessel inspection, safety valve certification, refrigerant handling and water hygiene are regulated in most jurisdictions, with consequences that are legal rather than operational.

The practical implication is that central plant deserves a higher tier of PM design than the rest of the register: more frequent condition observation, tighter acceptance criteria, and a named engineer who actually reads the trends. If you are designing the wider programme, the complete guide to preventive maintenance covers the framework this sits inside, and asset criticality classification is what justifies giving plant that higher tier in the first place.

2. Boiler types and what each one asks of you

Boiler preventive maintenance is not one schedule. The type determines which tasks are critical, which are optional, and which are legally mandated. The main families you will meet in commercial and light industrial estates:

• Steam boilers (fire-tube or water-tube). The most demanding, because they are pressure vessels operating with a steam and water interface. Water quality, blowdown discipline, level control and safety valve integrity are life-safety matters, not efficiency matters. Fire-tube shell boilers dominate commercial estates; water-tube units appear where pressure or output is higher.
• Hot water boilers (low and medium temperature). A closed pressurised loop with no steam interface, so water chemistry is more stable and blowdown is usually not a daily operation. The dominant risks shift to loop corrosion, air ingress, magnetite accumulation, and pump and valve condition.
• Condensing boilers. Their efficiency advantage comes from condensing flue vapour and recovering the latent heat, which only happens when return water temperature stays low enough, so return temperature is itself a maintenance-relevant reading. They also produce acidic condensate, meaning a trap and neutraliser to inspect and service, and heat exchanger surfaces needing periodic cleaning.
• Electric and thermal fluid boilers. Electric units remove combustion entirely and concentrate the PM on element condition, contactors and controls. Thermal fluid systems shift the focus to fluid degradation testing, a chemistry regime of its own.

Before you write a single PM task, confirm the type, the fuel, the design pressure and the manufacturer's own service schedule. The manual is the baseline; everything here is the operating discipline that wraps around it.

3. Daily and weekly boiler operator checks

The most valuable boiler maintenance happens every day and takes ten minutes. This is operator-level work, not contractor work, and the pattern that works is a short structured round with recorded values rather than a tick sheet.

Daily, on a fired boiler in service:

• Water level and gauge glass. Confirm level is correct and the gauge glass is clear and readable. On steam plant, blow down the gauge glass to prove the level is live and not a stuck column.
• Operating pressure and temperature. Record the actual value, not "normal". A pressure that has drifted up over weeks is telling you something about the control or the load.
• Flue gas temperature. The single most useful daily boiler reading after level. A rising stack temperature at the same firing rate means heat is not transferring, which usually means waterside scale or fireside soot.
• Flame condition and burner behaviour. A visual check through the sight glass, plus listening for pulsation, delayed ignition or repeated lockouts.
• Feedwater and condensate return. Tank level, temperature and, on steam plant, whether return volumes look normal. A drop in condensate return means more raw make-up water, which means more chemical demand and more blowdown.
• Leaks, noise and general condition. Water on the floor, steam plumes, unusual noise from pumps or fans, insulation damage.

Weekly, adding to the daily round: test the low water cut-out by controlled evaporation or the manufacturer's prescribed method, exercise the safety valve if the manufacturer and your local regime permit manual easing, test the flame failure device by interrupting the flame signal, check the burner air inlet and filter, and perform bottom blowdown on steam plant to remove accumulated sludge.

4. Combustion tuning and where boiler efficiency actually goes

A fired boiler converts fuel into heat at an efficiency that is set on the day it is commissioned and drifts downward from then on. Combustion tuning is the periodic exercise of putting it back, and it is one of the few maintenance activities with a directly calculable payback.

The measurement that drives it is flue gas analysis: oxygen (or carbon dioxide) percentage, carbon monoxide in parts per million, and net stack temperature. The logic behind them is simple.

• Excess air. Combustion needs more air than the stoichiometric minimum to burn cleanly, but every extra unit of air is heated up and thrown out of the stack. Flue oxygen is the proxy for excess air. Too low and you get incomplete combustion, carbon monoxide and soot; too high and you are heating the atmosphere. There is a narrow band per burner and fuel where the trade sits correctly, and the commissioning report defines it for your unit.
• Carbon monoxide. The safety and completeness check. Rising CO alongside falling oxygen means you have tuned past the safe limit. CO should be low and stable across the firing range.
• Net stack temperature. Flue temperature minus combustion air temperature. This is the heat you failed to transfer into the water. A rise of tens of degrees over the commissioning baseline at the same firing rate is the fouling alarm, and the general engineering rule of thumb is that efficiency falls noticeably for every significant increment of stack temperature rise.
• Turndown behaviour. Check combustion at low fire, mid fire and high fire, not just at whatever rate the boiler happens to be at when the analyser arrives. Burners commonly drift at one end of the range while remaining acceptable at the other.

Diagnostically, the useful discrimination is between fireside and waterside. If stack temperature has risen and flue oxygen is roughly unchanged, suspect fouling: soot on the fireside, scale on the waterside. If stack temperature is stable but oxygen has climbed, suspect the air damper, linkage or fan. If CO is climbing, stop and investigate before doing anything else. Tuning frequency depends on duty, fuel and burner type, but a genuine combustion analysis at least annually, and semi-annually on a hard-run boiler, is a defensible baseline. Guidance on plant efficiency practice is available through ASHRAE .

5. Water treatment, blowdown and the cost of getting it wrong

More boilers are damaged by water than by fire. Scale, corrosion and carryover are all water chemistry failures, and all three are preventable with a treatment regime that is monitored rather than assumed.

• Scale forms when hardness salts precipitate onto hot heat transfer surfaces. It is a thermal insulator, so it simultaneously wastes fuel and overheats the metal underneath. A thin layer costs measurable efficiency; a thick one causes tube failure. Softening, dealkalisation or reverse osmosis on the make-up water is the control.
• Corrosion comes mainly from dissolved oxygen and carbon dioxide in the feedwater and from incorrect pH. Deaeration and oxygen scavenging handle the first, alkalinity control the second. Pitting in a boiler or a condensate return line is usually an oxygen control failure traced back weeks or months.
• Carryover is water and dissolved solids leaving with the steam, caused by high TDS, high alkalinity, contamination or operating above design. It damages steam plant downstream and is a direct consequence of insufficient blowdown.
• Condensate system chemistry is the part most often neglected. Returned condensate is your cleanest, hottest feedwater, so protecting the return lines is both a corrosion and an efficiency measure.

Blowdown keeps dissolved solids within limits and has two forms. Bottom or intermittent blowdown removes settled sludge from the lowest point and is typically a short, frequent operator action. Surface or continuous blowdown removes dissolved solids from just below the water line and is set to hold TDS at target, ideally under automatic conductivity control. Blowdown costs energy, because you are discharging hot treated water, so the correct rate is the minimum that holds chemistry in range, not a generous guess.

The testing regime should be routine: daily boiler water TDS or conductivity, alkalinity and pH, scavenger residual and feedwater hardness, with a fuller monthly analysis from the treatment provider. The failure mode I see most often is a treatment contract where samples are taken and reports issued, but nobody in the client organisation reads them or acts on an out-of-range result. Route the treatment report into the maintenance system as an inspection record with acceptance limits, so an out-of-range value raises a corrective work order rather than sitting in an inbox.

6. Safety devices, interlocks and statutory inspection

This is the part of boiler maintenance where "we do it annually with the service contract" is not a sufficient answer. Safety devices protect against events that are catastrophic rather than merely expensive, and they are tested on their own cycle.

• Low water cut-outs (primary and, where fitted, independent secondary). The single most important protection on a steam boiler. Tested on the manufacturer's prescribed frequency, which is commonly daily or weekly for the functional test and periodically for the full evaporation test.
• Safety and relief valves. Set pressure verification and certification are typically the domain of a competent person or an accredited inspection body, not the site team. Never adjust or gag a safety valve.
• Flame safeguard and burner management. Flame failure response time, purge cycle integrity, ignition sequence, and lockout behaviour. These are programmed safety functions and their test results should be recorded with values, for example measured flame failure response time, not a tick.
• Pressure and temperature limits. Operating control plus a separate high limit with manual reset. The high limit is a safety device and must be proven independently, not inferred from the fact that the operating control works.
• Gas train integrity. Valve proving, pressure switches, leak testing of the train, and combustion air proving.

A practical point on scheduling: statutory inspection usually requires the boiler to be shut down, cooled, drained and opened, and often requires isolation of fuel and electrical supplies. That makes it a permit-controlled activity with a long lead time, and it should be planned around, not squeezed in. The permit to work integration discipline is what keeps that from becoming an improvised shutdown.

7. A boiler PM schedule you can lift

The table below is a defensible starting schedule for a commercial fired boiler. Treat it as a template to be reconciled against the manufacturer's manual and your local statutory regime, which take precedence wherever they differ.

If you want the wider structure for turning tables like this into usable CMMS job plans, the PM checklists, templates and examples article covers task wording, frequency codes and acceptance criteria in detail.

8. Chiller types and what changes between them

Chiller preventive maintenance follows the same principle as boiler PM, readings first, but the machinery is different enough that the task list diverges considerably between types.

• Centrifugal chillers. The large-tonnage workhorse of district and campus cooling. Oil management, refrigerant charge integrity, bearing condition, surge behaviour and tube condition dominate the PM. They reward log discipline more than any other type, because approach temperatures and motor current tell you almost everything.
• Screw chillers. Positive displacement, generally more tolerant of part load and lift, with oil separation and oil condition as central concerns. Oil analysis matters here as much as it does on a gearbox.
• Scroll chillers. Usually modular and multi-circuit. Individual compressors are often replaced rather than overhauled, so the emphasis shifts to circuit-level refrigerant and electrical checks, staging behaviour and coil condition.
• Absorption chillers. Thermally rather than electrically driven, on a lithium bromide or ammonia cycle. An entirely different profile: solution concentration and chemistry, inhibitor levels, purge performance, vacuum integrity and crystallisation risk. Non-condensable gas removal is routine and critical.
• Air cooled versus water cooled. Air cooled machines remove the tower and condenser water system entirely, replacing them with coils to keep clean and fans to keep balanced. Water cooled machines are more efficient but bring the whole condenser loop, tower, pumps and water treatment into scope.

9. Chiller log readings and approach temperatures: the core diagnostic

If you take one thing from this article into your plant room, take this section. A chiller log, recorded consistently and plotted, is the closest thing to condition monitoring you can get without buying a single sensor, because the machine is already instrumented and the readings are already on the panel.

The readings to capture on every log round, at a recorded load percentage and with ambient or condenser water conditions noted:

• Chilled water supply and return temperature, and flow if metered
• Condenser water in and out temperature (or ambient air on air cooled)
• Evaporator refrigerant pressure and saturation temperature
• Condenser refrigerant pressure and saturation temperature
• Compressor motor current or percentage of full load amps, and kW where available
• Oil pressure, oil temperature and oil level
• Superheat and subcooling where the machine reports them
• Run hours, start count, and any alarm or fault codes since the last round

From those you calculate the two numbers that carry most of the diagnostic weight:

Approach temperature is the temperature difference the heat exchanger cannot close, and it is a direct measure of how well that heat exchanger is transferring. The absolute value is machine-specific and comes from the manufacturer's design data or the commissioning record, so the number to care about is the drift from your own baseline, not a universal figure.

• Condenser approach rising is the classic fouled-tube signature. Scale, biofilm or silt on the condenser tubes, or non-condensable gas in the condenser, or low refrigerant charge. It shows up as higher condensing pressure, higher compressor lift and higher power draw for the same tonnage, so it is an energy cost every hour the machine runs.
• Evaporator approach rising suggests fouling on the evaporator side, oil accumulation in the evaporator, or low refrigerant charge.
• Both approaches stable but power up points toward the compressor itself, the motor, or the operating regime rather than the heat exchangers.
• Condensing pressure high with normal approach points outside the machine: cooling tower performance, condenser water flow, or ambient conditions.

Trending this way is genuine condition-based maintenance using instrumentation you already own, and it is the natural on-ramp to the sensor-and-model approach described in the predictive maintenance and failure prediction guide. Start with the log. Earn the right to the sensors.

10. Refrigerant, oil analysis and tube condition

Three technical maintenance streams sit underneath the chiller log, and each has its own cycle and its own acceptance criteria.

Refrigerant management. A chiller losing refrigerant loses capacity and efficiency before it trips, so leak detection is both an environmental and an operating obligation. The regime covers routine leak checks at joints, seals, relief devices and purge discharge, monitoring of purge unit run time on low pressure machines (a purge running more than it used to is telling you air and moisture are getting in), and charge verification against the nameplate. Refrigerant handling, recovery and record keeping are regulated in most jurisdictions and typically require certified technicians. Confirm the applicable rules locally; refrigerant regulation has changed repeatedly and continues to change as phase-down schedules progress.

Oil analysis. Compressor oil carries the same evidence a gearbox sample does: wear metals, moisture, acidity and viscosity change. On centrifugal and screw machines, an annual laboratory oil sample is one of the highest-value low-cost PM tasks available. Moisture indicates a leak path or dryer problem; a rising acid number indicates refrigerant or thermal breakdown; wear metals indicate bearing or gear distress well before a vibration alarm would notice. Change oil and filter on condition or the manufacturer's interval, whichever comes first, and keep the trend rather than only the latest result.

Tube condition. The condenser and evaporator tube bundles are where most chiller capacity is lost and where the most expensive failures originate. Two distinct activities apply:

• Tube cleaning (mechanical brush cleaning, or chemical cleaning where deposits warrant it) restores heat transfer. The condenser side on a water cooled machine is the usual candidate and typically warrants annual cleaning, with frequency driven by the condenser approach trend and the quality of the condenser water treatment. Evaporator tubes in a closed chilled water loop foul far more slowly and are cleaned on condition. Some sites fit automatic tube cleaning systems; they reduce but do not eliminate the need for inspection.
• Eddy current testing is a non-destructive examination of tube wall thickness and integrity, detecting pitting, erosion, cracking and wall loss that visual inspection cannot see. It is not an annual task on most machines. A sensible pattern is a baseline test at a defined point in the machine's life, then periodic repeat testing on an interval informed by the results, the water chemistry history and the criticality of the machine, with a test triggered earlier by evidence of tube leakage or aggressive water conditions. Eddy current results are what convert "the tubes look fine" into a defensible remaining-life statement.

Compressor overhaul. Major overhaul intervals on centrifugal and screw machines are set by the manufacturer, usually in running hours, and are properly informed by oil analysis, vibration data and the machine's own operating history rather than by the clock alone. Treat the manufacturer interval as the outer bound and let condition evidence move the date earlier if the data says so.

11. A chiller PM schedule you can lift

As with the boiler table, this is a starting schedule for a water cooled machine, to be reconciled with the manufacturer's service manual and local regulation.

12. Cooling towers, water hygiene and the condenser water loop

On a water cooled plant, the chiller is only as good as the cooling tower feeding it. A tower that is delivering condenser water two or three degrees warmer than design pushes the chiller's condensing pressure up and its efficiency down every hour the plant runs, and the chiller log will show it as high condensing pressure with normal approach.

The mechanical scope is straightforward and mostly about airflow and water distribution:

• Fill condition. Scaled, fouled, collapsed or biologically blocked fill is the most common cause of lost tower performance. Inspect it, and replace it when degraded rather than nursing it indefinitely.
• Drift eliminators. These stop water droplets leaving with the air stream. Damaged or missing eliminators waste water and, more importantly, discharge aerosol from the basin into the surrounding area, which is a direct hygiene risk. Inspect them as a hygiene control, not merely a water-saving one.
• Water distribution. Blocked nozzles or an uneven hot water basin leave part of the fill dry and part flooded, and the tower loses capacity across its whole face.
• Fan, gearbox and drive. Blade condition and pitch, belt tension or coupling alignment, gearbox oil, bearing lubrication and vibration. A tower fan is a rotating asset in a wet, corrosive environment and deserves the same treatment as any other.
• Basin, strainer and structure. Silt accumulation, corrosion, strainer condition, make-up and bleed valve operation, level control.

Water treatment on the condenser loop controls three things at once: scale, corrosion and biological growth. Conductivity-controlled bleed sets the cycles of concentration, with inhibitor and biocide dosed to the provider's regime. Dip slides or equivalent microbiological testing are how you confirm the biocide programme is working rather than merely running.

Pumps. The chilled water and condenser water pumps are the forgotten members of the plant room. Each deserves its own PM line: seal condition and leakage, bearing lubrication and temperature, vibration survey, coupling alignment check, suction and discharge pressure logged against the pump curve, motor current, and strainer cleaning. A pump whose differential pressure has drifted down at constant speed is wearing, and it is another reading that belongs in the log alongside the plant readings.

13. Where this approach does not work, and what it costs

I would be doing the reader a disservice to present reading-led plant PM without the honest counterweight. There are conditions under which it does not deliver.

• It needs an operator presence. Daily log rounds assume someone competent is on site. Many buildings are now run with a roving technician covering multiple sites and no resident operator. In that model the answer is automated BMS trend logging, which is a capital and integration cost, not a free alternative.
• It needs someone to read the trend. This is the failure I see most often: logs captured diligently for months that nobody plots. If you cannot name the person reviewing the trend monthly, fix the ownership before starting.
• Baselines are often missing. Approach and stack temperatures only mean something against a baseline. On older plant with no commissioning records you have to build one from a period of known-good operation, which takes a season and some judgement.
• Instrumentation drifts. A two-degree sensor error produces a two-degree phantom approach change, which is why sensor calibration is a prerequisite task rather than an optional one.
• Some findings are expensive. Reading-led maintenance will eventually tell you tubes need plugging or a boiler needs retubing. That is the system working, but it produces unbudgeted capital conversations. Better condition information generates more, not fewer, difficult decisions in the first two years.

There is also a legitimate case for keeping some plant PM purely calendar-based. On a small, non-critical, standby or low-hours machine, the effort of building and maintaining a trending regime will not repay itself, and a competent annual service plus statutory compliance is the right answer. Criticality decides. The PM strategies article works through that time versus meter versus condition trade in general terms, and PM examples across industries shows how differently this lands in different sectors.

14. Setting this up properly in a CMMS

A schedule that lives in a spreadsheet does not survive a change of facilities manager. Some specifics on translating the above into a system, whether you are on IBM Maximo, SAP PM, Infor EAM, Hexagon EAM, Planon, or a mid-market tool like Fiix, Limble, eMaint, MaintainX or UpKeep:

• Model the plant properly in the hierarchy. The chiller is an asset; compressor, evaporator, condenser, starter and oil system are sensible child assets. Cooling towers and pumps are separate assets, not attributes of the chiller. That is what lets you track cost and failure history where decisions are made.
• Use meter or reading fields, not checkboxes. Every row above with a numeric acceptance criterion should be a numeric reading in the system so it is trendable. Where the tool supports meter-based triggering, drive overhaul-class tasks from running hours rather than the calendar.
• Put acceptance limits on the task so the technician sees the expected range on the work order, and out-of-range readings raise a corrective work order automatically where the system supports it. Link permits to the high-risk tasks as well: boiler shutdowns, refrigerant work, confined space entry into a water box and gas train work should carry a permit requirement on the job plan itself.
• Separate statutory PMs and flag them. Pressure vessel examination, safety valve certification, gas certification and water hygiene tasks should form an identifiable compliance class, with the governing document and issuing body recorded on the asset. When an auditor asks, you want one filtered report.
• Attach the baseline. Commissioning data, design approach temperatures, the service manual and the last combustion report belong on the asset record, so the person at the machine has the reference without hunting.

Finally, measure the programme rather than assuming it works. PM completion rate on plant, percentage of readings actually captured, number of out-of-range findings raised and closed, and emergency work orders on plant assets as a share of total plant work are a reasonable starting set. The FM KPI framework covers how to build that into a reporting pack that survives scrutiny.

The idea to walk away with

Boiler and chiller preventive maintenance has two halves, and most organisations only do one of them. The first half is the schedule: the services, the statutory inspections, the annual shutdown scope, the burner strip, the tube clean. That half is usually contracted out, usually reasonably well executed, and usually the only half anyone thinks about.

The second half is the observation: the daily log, the stack temperature against firing rate, the approach temperatures against load, the purge run time, the oil analysis trend. That half costs almost nothing, requires no procurement, and is where the condition insight actually comes from. It is also the half that decides whether the expensive interventions in the first half happen at the right time or a year too late. A plant maintained on the schedule alone is being maintained blind. The same plant, same schedule, with a log somebody reads, is being maintained on evidence. That is the whole difference, and it does not require a capital budget to start.

Final thoughts

If you are inheriting a plant room and want a defensible first ninety days, the sequence I would advise is narrow and unglamorous. Confirm the statutory position first: what the local regime requires for pressure vessel examination, gas safety, refrigerant handling and water hygiene, who is competent to do it, and when it was last done. Get that written against the asset records before anything else, because it is the part with legal consequence and it is very often the part that is quietly out of date.

Then establish the log: define the readings, put them on a daily or shift PM in the CMMS as numeric fields, record load and ambient context alongside them, and name the engineer who will review the trend monthly. Where commissioning and design data do not exist, accept that your first season of readings is the baseline. Only then reconcile the task schedule, manufacturer's manual against your current PM list, with the tables above as a cross-check for gaps. You will usually find a handful of missing tasks and a few being done more often than anyone can justify. Fix both, and let the trend data drive the frequency of the expensive interventions from there.

None of this is novel engineering. It is the ordinary discipline of reading the machine before servicing it, applied consistently to the two assets where the consequences of not doing so are largest.

Related reading: Preventive maintenance: the complete guide, PM checklists, templates and examples, Preventive maintenance strategies, Predictive maintenance and failure prediction, Asset criticality classification, FM KPI framework.