Chiller maintenance checklist for a data-center plant: every item by interval, the state the chiller must be in, and the reading that proves it

The annual on chiller 2 is booked for Thursday. The work order carries the manufacturer's annual inspection list, the plant is three chillers at N+1, and chiller 3's last oil sample came back with rising iron. Which of the items need chiller 2 stopped, and whether chillers 1 and 3 can carry the hall for the eight hours its waterboxes are open, is not written anywhere on the order.

A chiller maintenance checklist is the manufacturer's list of daily log readings, quarterly and semi-annual service items and an annual shutdown inspection, with a non-destructive tube test every three to five years and ASHRAE Standard 180 as the published floor. On a data-center plant every item also carries the state the chiller must be in to do it. Running items are readings from the controller and cost the plant nothing, stopped items cost the plant its spare for the length of the visit, and opened items add the EPA card and the lock. Every item closes on a reading from the chiller's own controller rather than on a tick.

• The intervals are the manufacturer's guides, and the manufacturers say so. Carrier's 19XR manual calls its time intervals guides to service only, to be set on load, run hours and water quality, and ASHRAE 180 calls its Section 5 frequencies the minimum required for compliance.
• A data-center chiller runs about 8,760 hours a year. Trane's first oil change falls at six months or 1,000 hours, whichever comes first, and at 24 hours a day the 1,000 hours arrive in about six weeks.
• Roughly a third of the items need the chiller stopped, and the annual inspection needs it opened. Those are the items an N+1 plant has to gate behind a confirmed spare, because a second chiller out in the same window is how a PM becomes an incident.
• Every item can be proven from the log: approach temperature against the new-unit baseline, oil differential pressure, discharge temperature, motor amps and, on a low-pressure chiller, the purge pump-out minutes. The work order should close on those readings.

What is a chiller maintenance checklist?

It is the list of readings, inspections, tests and cleaning tasks a water-cooled chiller receives on a fixed cycle, so that it fails less often and so that a developing fault is caught while the spare can still carry the load. A chiller is the plant's cold source, and every computer room air handler (CRAH) on its loop depends on it, which is why its list is longer and its shutdown costlier than any hall unit's. The list divides into what an operator reads every day, what a technician services every few months, and what a certified technician does once a year with the machine open.

Four documents set the intervals, and they do not agree with each other. The manufacturer's operation and maintenance manual is the binding one, ASHRAE 180 is the reference the sector cites, the EPA's refrigerant rules fix the leak-repair clock, and the site's own method of procedure (MOP) decides how a chiller is taken out of the plant and put back. The table below is what each of the opened documents actually says.

The words that matter in those manuals are the ones about guides and minimums. Carrier tells the reader to establish a regular maintenance schedule based on your actual chiller requirements such as chiller load, run hours, and water quality, with the listed intervals offered as guides to service only. ASHRAE's 2018 errata restates clause 4.3.2.2 as the Section 5 frequencies being the minimum required for compliance. A plant that runs the manufacturer's annual once a year and nothing between is below both.

Why is a data-center chiller's checklist different from a building's?

A building chiller has a cooling season and an off-season, and the annual is done in the off-season with the machine cold and nobody depending on it. A data-center chiller has neither. Kaltra's checklist is the one page-one result that says so, in a clause: inspect twice a year for seasonal use and four times a year in continuous operation. The load does not fall in January, and there is no afternoon on which the plant can lose a chiller for free.

The manufacturers write some intervals in hours, and hours arrive faster on a plant that never stops. Trane's guide sets the first oil and filter change at six months of accumulated operation or 1,000 hours, whichever comes first, and at 24 hours a day 1,000 hours is about 42 days. A building reads that sentence as six months. A data center reads it as six weeks, and should read every calendar interval in the manuals with the same arithmetic in mind.

On a building the checklist is a list of tasks with a season to do them in. On a data-center plant it is a list of tasks, each with a question in front of it: can the plant spare this chiller right now?

Standard 180 points the other way for buildings. Its 2012 text gives the example of an old chiller retained as a backup whose inspection may be adjusted to reflect fewer operating hours, which is a relief a data-center spare cannot take, because a rotated N+1 chiller runs the same hours as the duty machine and carries the hall the moment one is stopped. The Uptime Institute's own explainer names chillers among the redundant components a Tier II site holds for select maintenance opportunities, and defines a Tier III site as one that requires no shutdowns for equipment replacement and maintenance. Which of those two sentences describes the plant decides whether a stopped-chiller item is a task or a risk.

What is on the checklist, daily to every three years, and what proves each item?

The table below is the list, assembled from the Trane, Carrier and York manuals and ASHRAE 180's Table 5-6, with the interval each source gives. Two columns are added to what any of those documents carries: the state the chiller has to be in to do the item, and the reading on its own controller that proves the item afterwards.

Running means the log is read with the machine on, and stopped means the compressor is off with the refrigerant in place, so the plant is carrying the hall on N. Opened means the refrigerant circuit or a waterbox is open, which adds the EPA card and the lock.

York's maintenance-requirements table ends with a footnote worth keeping: the certified-technician items must be performed at the specified interval by an industry-certified technician, and a record of the procedure being carried out must be kept on file by the owner should proof of adequate maintenance be required for warranty. That is the manufacturer saying the tick is not the record. The reading column above is what a record should contain.

Carrier's manual shows how much of that record the controller already keeps. Its Service Ontime counter is to be reset to zero by the service person or the operator each time major service work is completed, so that the time between service events can be tracked, and Trane's log sheet carries the purge pump-out minutes for the last 24 hours, the last seven days and the life of the purge. A counter reset without the work, or the work without the reset, is the kind of disagreement a PM record exists to catch.

Which readings on the daily log say a chiller needs work before the calendar does?

The daily log is the front of the checklist, not a chore that precedes it. York's manual asks for a permanent daily record of temperatures and pressures at regular intervals throughout each 24-hour operating period, and says that readings taken when the system is new establish the normal conditions later readings are compared with. Every manufacturer opened for this page gives at least one threshold for that comparison, and the table below collects them.

The purge row deserves its own arithmetic, because it turns a daily reading into the leak-rate figure the regulation wants. Trane's guide gives the formula: percent annual leakage rate equals daily purge minutes times 0.0001 pounds of refrigerant a minute, divided by the initial charge, times 100, and it tells the operator to leak-check at the earliest convenience in the small-leak region and immediately in the large-leak one. The purge guide adds the reason the number is worth trending: the daily pump-out time indicates how the hermetic integrity of the chiller compares to historic pump-out times for the same chiller, and frequent filter-drier changes on the purge can indicate significant air or tube leaks.

Carrier gives the high-pressure equivalent as a rule rather than a reading. Its manual says ASHRAE recommends that a chiller be taken off line immediately and repaired if the refrigerant leak rate for the entire chiller is more than 10 percent of the operating charge per year, and Carrier itself recommends repairing leaks above 0.1 percent a year at the annual or whenever the refrigerant is transferred. Ten percent is also the legal threshold, and the section below says which rule carries it for which refrigerant.

Which items need the chiller stopped, and what has to be true of the plant first?

Eleven of the eighteen items stop or open the machine, and the annual inspection is the largest of them. Trane's Table 22 says to shut the chiller down once each year to check the items on its annual inspection list, and both Trane and Carrier put the electrical work behind a lockout: Trane's warning is to disconnect all power including remote disconnects, discharge the capacitors, verify with a meter rated per NFPA 70E and follow proper lockout and tagout procedures, and Carrier's is to shut off the chiller and open and tag all disconnects before working on any starter. On a building that is an electrical decision. On a plant it is a redundancy decision first.

The plant's N is the number of chillers the hall needs at its present load, and the arithmetic of the spare decides whether a stopped chiller is a task or an exposure. Uptime's Tier III definition is the sentence a stopped-chiller work order has to satisfy: every component needed to support the IT load can be shut down and maintained without impacting IT operations. On a Tier II plant, with redundant components but one distribution path, the same item is one of the select maintenance opportunities, and the window is the whole hall's.

• Confirm the spare is ready, not just present: in automatic, no active alarm, no manual override, its own daily log clean, and start-tested since its last PM. A spare with a rising-iron oil report is not the spare it was last quarter.
• Confirm the hall's load against N with this chiller stopped, at the condenser water temperature the plant will actually see that day, and with any other chiller already down or derated.
• Take a baseline before the machine is touched: every reading in the daily-log table, the chilled-water supply temperature and delta-T at the plant header, and the inlet temperatures in the rows the loop serves.
• Isolate under the site's written energy control procedure. 29 CFR 1910.147 requires procedures developed, documented and utilized for the control of hazardous energy, and the lock comes off only by the employee who applied it.
• Do the stopped and opened items in one visit, so the chiller comes out of the plant once per cycle and not four times.
• Restart under the manufacturer's sequence, return the chiller to its place in the staging order, and put the spare back to standby.
• Watch the plant through the pull-down against the ASHRAE limit of 5 °C in any 15 minutes at the inlets, and close the work order when the post-visit readings match the proof column, not when the last box is ticked.

The reason the gate is stricter for a chiller than for a hall unit is what a stopped chiller does to the time the hall has. A CRAC out of rotation costs one unit's airflow. A chiller out of the plant costs every CRAH on its loop their cold source the moment the remaining machines cannot hold the header, and the ride-through the hall then has is the water in the pipes and whatever storage the plant was built with. The pull-down after restart is bounded the same way, by ASHRAE's 5 °C in any 15-minute period at the IT inlet.

Which leak-repair rule and which EPA card does your chiller's refrigerant put it under?

The refrigerant rows depend on what the chiller holds, and page one gives one answer for every chiller. The EPA's Emissions Reduction and Reclamation rule put appliances charged with an HFC, or a substitute with a global warming potential over 53, under leak-repair requirements at 15 pounds and up, from January 1, 2026. The older rule, 40 CFR 82.157, has applied since April 2020 only to appliances with 50 or more pounds of a class I or class II refrigerant, and the agency's fact sheet adds that appliances using a substitute with a GWP of 53 or below are not covered by the leak-repair requirements at all. Which of the three a chiller sits in is decided by its nameplate.

The card follows the pressure, not the rule. 40 CFR 82.152 defines a low-pressure appliance as one using a refrigerant with a liquid-phase saturation pressure below 45 psia at 104 °F, with R-123 as its example, and 82.161 requires anyone who maintains, services or repairs a low-pressure appliance to be certified as a Type III technician. A chiller on R-514A or R-1233zd(E) is outside the leak-repair rule and inside the certification rule at the same time, and which cards the person on shift actually holds is a record on the site, not a job title. The GWP figures in the table are from EPA's own reference table.

What the new section asks of an R-134a or R-513A plant is a set of records, each of which is a checklist row. The leak rate is calculated every time refrigerant is added, so every top-up on the work order carries a rate against the full charge. A rate over 10 percent starts a 30-day repair clock with two verification tests, the second after the chiller returns to normal operating characteristics, and an annual leak inspection until a year of calculations shows the machine under the rate. The full charge on the order comes from the nameplate or the methods 84.102 lists, not from memory.

What do the chiller controller, BMS, DCIM and CMMS already hold about a PM?

Each of them holds one column of the checklist table, and none of them holds the join between the item, the state the plant is in, and the reading before and after. The gap is not that the systems fail to communicate. It is that the PM record lives in the one system that never sees the log.

The chiller controller (Trane's Tracer AdaptiView, Carrier's PIC 5, York's OptiView)

Every reading in the daily-log table, the alarm history, the Service Ontime or run-hours counter, and on a low-pressure machine the purge pump-out minutes for 24 hours, 7 days and the life of the purge. Not whether the plant can spare the chiller, and not a before-and-after across a visit.

BMS

The trend of every point the controller exposes, the plant header temperatures and flows, and each chiller's status and staging order. Not the checklist, the interval, or which item a change in a trend belongs to.

DCIM

The hall's load, the installed plant capacity and therefore N. Nothing inside the chiller.

CMMS or EAM

The PM schedule, the item list, the parts, the technician, the oil report as an attachment, and the tick. Rarely a single point value on a closed order.

The MOP

The written sequence for taking the chiller out of the plant and putting it back. Not the live state on the day.

Standard 180 quietly allows the join. Its 2012 text says inspection may include data provided by sensors or a building management system, and that if unacceptable condition indicators are found on two successive inspections the owner shall investigate, while three successive acceptable inspections allow the frequency to be reduced and documented. A plant that closes each PM on a reading has that two-and-three history for free. A plant that closes on a tick has nothing to reduce or investigate from.

How to turn the checklist into a work order that closes on a reading this week

The table above becomes a site document in one afternoon per chiller model, and it changes what a closed work order means. The steps are in the order a plant lead would actually do them.

• Copy the manufacturer's list for each chiller model you run, keep its intervals, and add the two columns: the state the chiller must be in and the reading that proves the item. Where your model's manual has no interval, use the nearest OEM band above and note the source.
• Re-read every hours-based interval at 24 hours a day, and every calendar interval against the continuous-duty clause, before you copy it. Write the result on the order, not the manual's number.
• Mark the stopped and opened items and group them into one visit per chiller, so each machine leaves the plant once per cycle.
• Write the plant gate as three lines at the top of every stopped-chiller order: spare ready in automatic and start-tested, load against N with this chiller out at the day's condenser water temperature, and any other chiller already down.
• Take the first-month baseline for every chiller that has none, from the controller's log, and store it on the asset. Without it the approach rule has nothing to compare with.
• For every chiller, write the refrigerant, the full charge from the nameplate, the pressure class and card, and the leak-repair rule that follows from the refrigerant table above. Add the leak-rate calculation as a mandatory field on any addition.
• On a low-pressure chiller, set the purge daily pump-out limit from a week's baseline and log the 24-hour and 7-day figures as daily items.
• Close the order only when the post-visit readings sit inside the proof column, and hold it open when they do not.

What closes a chiller PM item is a reading that agrees with the tick. The tube clean shows as the condenser approach back at the new-unit difference, the oil change shows as the next laboratory report, the electrical work shows as balanced amps on restart, and the purge service shows as pump-out minutes back at the baseline. A visit that leaves none of those behind was a visit, not a maintenance record.

What this cannot do

A checklist assembled from three manufacturers' manuals is not your chiller's manual. The intervals here are the ones those documents give, Trane's normal pressures are for a low-pressure machine, York's approach rule compares with the new-unit value rather than a fixed number, and the refrigerant rule is the regulation as it stands on the eCFR this month. The Thursday scene on chiller 2 is a composite assembled from those lists, not a site. Your model, your refrigerant and your site's reading of the lockout procedure decide the rest.

Nothing here takes control of a chiller or a plant. A decision layer that reads the log can say that the spare is in alarm and the order should not release, or that the Service Ontime counter was reset while the condenser approach never moved, and it does not start, stop, stage, isolate or reset anything. A named human stops the chiller, a named human applies the lock, a certified technician opens the circuit, and the layer records what was read before and after and whether the two agreed.

Which guides sit beside this one?

• CRAC and CRAH preventive maintenance checklist: the hall units' list on the same three-column pattern, and the full set of 84.106 obligations for an R-410A unit.
• Cooling systems ranked by failure time: the chiller restart and the stored water a stopped chiller spends.
• Cooling redundancy, N+1 versus 2N: the arithmetic behind the plant gate, and why the tier does not tell you the topology.
• CRAC vs CRAH, which failure buys you time: what a chiller failure does to every unit on its loop.
• Legionella compliance for cooling towers: the condenser water and the tower this page leaves to the tower checklist.

Answered