Cooling tower maintenance checklist for a data center: what each item needs the cell to be doing, and the reading that says it is due

The quarterly on tower cell 2 is booked for Tuesday. The plant is three cells at N+1 serving two chillers, the wet-bulb has sat at 74 °F all week, and the order reads inspect and clean fill, drift eliminators, nozzles and basin. Nothing on it says the cell has to be drained and its fan locked out first, or what the remaining two cells will do to condenser-water temperature while it is out.

A cooling tower maintenance checklist is the recurring cycle of inspection, cleaning, drive service and water-chemistry control that keeps a tower rejecting its design heat load, with ASHRAE Standard 180-2012 as the published floor and the manufacturer's manual as the binding document. On a data-center plant every item carries two things the generic lists leave out: the state the cell must be in to perform it, and the reading that says it is due. Approach temperature, conductivity, free-chlorine residual, oil level and vibration are what close a tower PM. A date only says when someone looked.

• The published floor and the binding document are not the same thing. ASHRAE 180-2012 Table 5-10 covers cooling towers and evaporative-cooled devices at monthly, quarterly, semiannual and annual frequencies, while Marley's NC Class manual sets weekly observation, monthly inspection, semi-annual belt and oil work and a five-year gearbox oil interval for that specific tower.
• Roughly a third of the list cannot be done with the cell running. Marley is blunt about the gate: do not attempt any service unless the fan motor is locked out, which puts most of the quarterly list behind 29 CFR 1910.147 and behind a confirmed spare cell.
• Approach temperature is the one number that says a fill or airflow item is due. It is the cold water leaving the tower minus the ambient wet-bulb, and it widens when the fill fouls or the airflow drops, at the same load and the same wet-bulb.
• The standard's own relief clause runs backwards here. ASHRAE 180-2012 lets a facility stretch an interval, and both worked examples in §4.2.2.d are conditions a year-round plant with idle N+1 cells never gets.

What is a cooling tower maintenance checklist?

It is the fixed list of inspections, cleaning tasks, drive services and water tests an open-circuit cooling tower receives on a repeating cycle. The tower is the plant's heat-rejection stage, so every chiller on the condenser-water loop and every computer room air handler (CRAH) behind those chillers depends on it staying clean and moving air.

The list divides into three kinds of work that behave very differently in a live plant. Observation and chemistry happen with the cell wet and running, drive and structural service need the fan locked out, and the clean-down needs the cell drained and off the loop entirely.

Fill

The media that presents water to air as a film or splash. Where the heat transfer actually happens, and the first thing to foul.

Drift eliminator

The baffle the leaving air passes through, stripping entrained droplets. It is the emission path, which is why it is inspected rather than assumed.

Cold-water basin

The sump under the tower holding the circulating water and the pump suction. Sludge collects here and the strainer blinds because of it.

Blowdown or bleed

The controlled discharge of concentrated water that holds cycles of concentration where the treatment programme can work.

Approach

Cold water leaving the tower minus the ambient wet-bulb temperature. The tower's performance number, and the trigger reading this page is built on.

Cell

One tower module with its own fan, fill and basin. An N+1 plant has a spare cell the way it has a spare chiller, and it is isolated the same way.

Why does a data-center tower need a different checklist from a building's?

A commercial building's tower runs a cooling season and shuts down for winter, which is the assumption almost every published checklist is written against. A data-center tower runs about 8,760 hours a year, so an interval expressed in months is an interval expressed in a quarter more run hours than the manual assumed.

The second difference is the idle cell. A building rarely carries a spare, while an N+1 plant carries one by design and that cell spends most of its life wet and still, which is the condition a treatment programme exists to prevent.

Read together they agree on the component list and disagree on almost everything else. The practical resolution is the one the manufacturers state themselves: the manual is binding, the standard is the floor you can be audited against, and the site's own run hours decide where inside that range the interval sits.

What is on the checklist, weekly to annual, and what reading proves each item?

The table below is the list, with the two columns the result set leaves out. Cell state is what the tower has to be doing for the item to be performed safely, and the closing reading is the measurement a technician records to show the item moved something.

Three states matter, and they cost the plant very differently. Running costs nothing, fan locked out costs the cell its airflow while keeping water in the basin, and drained and isolated costs the plant that cell's whole capacity for the length of the visit.

Which reading says a tower item is due before the calendar does?

Approach temperature is the tower's performance number, and it is the reading a calendar-driven checklist never mentions. It is the cold water leaving the tower minus the ambient wet-bulb, so a cell producing 85 °F cold water at a 78 °F wet-bulb is running a 7 °F approach.

The useful property is that the approach is bounded by physics rather than by the equipment, because no tower can cool water below the wet-bulb of the air. Design approaches generally fall between 5 and 20 °F and no manufacturer guarantees performance below 5 °F, which means a cell's own commissioned approach is a stable baseline to read against for its whole life.

Recording the wet-bulb alongside the cold-water temperature is what makes the whole scheme work, because a cold-water temperature on its own proves nothing. The same 85 °F supply is excellent at a 78 °F wet-bulb and a sign of a fouled cell at a 68 °F one.

A date says when someone looked at the tower. An approach that came back to its commissioned value at the same load and wet-bulb says the cleaning actually moved heat transfer.

Which items need the cell drained and the fan locked out, and what has to be true of the plant first?

The manual states the gate in one line, and it applies to most of the quarterly list rather than to the annual alone. Marley's instruction is that no service should be attempted unless the fan motor is locked out, which places the work under the site's hazardous-energy procedure before any question of scheduling arises.

Draining the cell is the more expensive state, because it removes that cell's heat rejection from the condenser-water loop for the length of the visit. On an N+1 plant that is the spare being spent, and the three conditions below are what a method of procedure (MOP) should require before it is.

• The remaining cells can hold the condenser-water supply setpoint at the current load and the forecast wet-bulb, with margin. Wet-bulb is the variable that decides this and it is the one nobody checks, because the plant that coped at a 68 °F wet-bulb on Monday may not at 78 °F on Thursday.
• No other cell, condenser-water pump or chiller is already out. A second loss in the same window is how a scheduled PM becomes an incident, which is the same gate the chiller checklist applies to an opened machine.
• The isolation is written down. Which valves close, which equaliser is broken between adjacent cells, how the basin is drained and in what order the cell comes back are a standard operating procedure (SOP), not a decision taken at the tower.

The size of that cost is worth stating, because it is the part a tower checklist never quantifies. PNNL's assessment of variable-speed chiller plants records that centrifugal chiller efficiency rises by about 0.4 percent for every 1 °F reduction in entering condenser-water temperature, quoting Thumann, so the same relationship runs in reverse when a cell comes out and the remaining cells give up a few degrees. That is an inference about your plant rather than a measurement of it, and the number to trust is the one your own plant meter shows during the last cell isolation you did.

Why does ASHRAE 180's frequency-adjustment clause run backwards on a data-center tower?

The standard does not treat its own intervals as fixed. Section 4.2.2.d of ASHRAE 180-2012 permits inspection and maintenance frequency to be adjusted for climate-related or operational reasons, provided each adjusted frequency is documented with the reason for it.

The clause then gives two worked examples, and both of them reduce the frequency. One is a cooling tower shut down during the winter, where inspection and maintenance may be suspended for the shutdown period, and the other is a chiller retained as a backup, where the frequency may be adjusted to reflect fewer operating hours.

The reading that falls out of this is not that the standard is wrong, but that its relief clause is written for a building and documented adjustment is the mechanism, not the direction. Applied honestly to a year-round plant, §4.2.2.d is an argument for shortening the drive and chemistry intervals on the duty cells and for treating the idle cell as a higher-attention asset than the running ones.

What must the circulating water stay inside, and where does that collide with disinfection?

The tower's own materials set the envelope, and on a galvanized-steel tower that envelope is narrower than most operators expect. Marley defines a normal environment for the NC Class in terms a treatment supplier can work to, and it is the manual rather than the treatment contract that decides what the metal will tolerate.

The routine number that keeps the tower inside the envelope is cycles of concentration, the ratio of dissolved solids in the circulating water to the makeup. Blowdown follows from it directly, since the bleed rate is the evaporation rate divided by the cycles minus one, which is why a bleed valve that has stopped stroking shows up as drifting conductivity long before it shows up as scale.

What do the BMS, the water-treatment controller, DCIM and the CMMS already hold about a tower PM?

All four hold something true, and none of them holds the join. That is the honest description of the situation, and it is different from saying the systems do not talk to each other.

The gap that matters is the smallest one. The CMMS knows the quarterly was completed and the BMS knows the approach afterwards, and nothing in the standard stack asks whether the second changed because of the first.

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

None of this needs a new system to start. It needs the work order to carry two fields it probably does not have today, and one baseline per cell.

• Record a baseline approach for every cell, at a known load and a measured wet-bulb, on a day the cell is clean. Without it every later reading is an opinion.
• Add a cell-state field to the tower PM templates in the CMMS, with three values — running, fan locked out, drained and isolated — and populate it from the table above rather than from memory.
• Add a closing-reading field, and make it required. The quarterly fill and nozzle item closes on an approach at the same load and wet-bulb, the strainer item on restored condenser-water flow, the bleed item on a measured bleed rate against the conductivity setpoint.
• Put the plant gate in the MOP for any drained-cell item: remaining cells hold the setpoint at the forecast wet-bulb, nothing else in the loop is already out, and the isolation and return sequence is written. Treat the forecast wet-bulb as a scheduling input, not weather.
• Log every emergency or hyperhalogenation event against the cell, and inspect eliminator media and sealants at the next quarterly after one.
• Give the idle N+1 cell its own schedule rather than the duty cells' schedule, and document the adjustment the way §4.2.2.d requires. Circulation during standby is the item that belongs on it.

The change that produces the most in the first month is the baseline, because it converts every later approach reading from a number into evidence. The change that produces the most over a year is the cell-state field, since it is what stops a drained-cell item being scheduled on a week when the plant cannot afford it.

What this cannot do, and what we do not claim

A widening approach does not diagnose anything. It says the cell is transferring less heat than it did, and fouled fill, a blocked air inlet, a slipping belt and a gearbox losing oil all produce the same movement in the same direction.

The product boundary is the same one this site states everywhere. The operating layer reads the loop, the chemistry log and the work record, and it does not start a fan, close an isolation valve or move a condenser-water setpoint — a person with the authorisation and the lock does that, and a person signs the return to service.

Two further limits are worth naming rather than hiding. The intervals here are one manufacturer's for one tower family and a 2012 edition of the standard, so your manual wins where they differ, and the 0.4 percent per °F figure is a published relationship for centrifugal chillers rather than a measurement of your plant.

Which guides sit beside this one?

The compliance half of the tower question, the chiller the tower serves and the redundancy arithmetic behind a drained cell each have their own page. This one stops at the mechanical list and the readings that close it, and the redundancy guide is the one to read before deciding what a spare cell is actually for.

Answered