Chilled water hydro test package field guide for data centers

The chilled-water hydro test package, and what it is

A chilled-water hydrostatic test, the hydro test, fills a section of piping with water, pressurizes it above its working pressure, and holds that pressure while you watch for any leak or pressure decay. The test package is the paperwork around it: the marked-up drawing of what was tested, the calculation of the test pressure, the calibrated gauge record, the temperature log, the witnessed result, and the pass or fail. The test proves the pipe and its joints hold. The package proves the test happened and lets someone defend it a year later.

On a data center this is not a formality you check off. The chilled-water loop is what keeps the hall inside the ASHRAE envelope, and most of it runs over, around, and above electrical gear and racks that a single weeping flange can take out. You prove the piping holds with water, in a controlled test, before there is anything underneath it worth ruining and before the insulation hides every joint you would want to inspect.

The scope is the same physics across chilled water, condenser water, and the process-water and makeup loops, and it is one loop upstream of the direct-to-chip liquid cooling work. The cold-plate loop gets its own proof test for the same reason this one does. This guide stays on the building chilled-water and process-piping package that the cooling plant rides on.

What does a hydrostatic pressure test prove?

A hydrostatic pressure test proves two things: that the piping has the strength to hold a pressure above its working pressure, and that every joint, weld, flange, and threaded connection in the tested section is tight enough not to leak under that pressure. It is both a strength test and a tightness test in one, which is why the test pressure sits above the design pressure rather than at it.

What it does not prove is just as worth naming. It does not prove the system is clean, because debris and mill scale survive a pressure test fine. It does not prove the flow is balanced or the pumps are right. And it does not prove the joints will never fail under thermal cycling and water hammer over twenty years of service. The hydro test is a snapshot that says the piping, as built and as it sits today, holds water above its rated pressure. The flush, the balance, and the functional tests are separate scopes that come after.

The reason it carries weight is timing. A leak found at the test is a joint you redo on a dry, empty system with nothing under it. The same leak found after the system is filled, insulated, and cooling a live hall is a shutdown, a soaked ceiling, and a conversation with the owner about what got wet.

Why test before insulating and before the system goes critical?

You test before insulation because the insulation hides the joints. Chilled-water pipe gets wrapped in closed-cell or fiberglass insulation with a vapor barrier to stop condensation, and once that jacket is on, you cannot see a weeping flange or a sweating weld until the water finds its way out somewhere far from the source. Insulate first and you are betting the joints are good with no way to look. Test first and you walk every connection dry-eyed before anything covers it.

You test before the system is critical for the obvious reason and a quieter one. The obvious reason is that a leak over live racks or energized switchgear is a loss event, not a punch-list item. The quieter reason is access. Before turnover the section can be isolated, drained, and opened without coordinating an outage, without escorts, and without a change-control board. After the facility is live, the same repair is a project. Pressure testing is front-loaded in the mechanical sequence precisely because every week you wait makes the fix more expensive and more dangerous.

The sequence that holds up is pressure test, then flush and clean, then insulate, then the functional and integrated tests. Skip the order and you either insulate over an unproven joint or you flush debris through a loop you have not yet shown will hold.

Hydrostatic vs pneumatic: why water and not air?

The default is hydrostatic, water, and the reason is stored energy. Water is nearly incompressible, so when a fitting lets go on a water-filled test the pressure collapses almost instantly and the water barely moves. Compressed air or gas is a coiled spring. It stores something on the order of a couple of thousand times more energy per unit volume than water at the same pressure, so a failure on a pneumatic test does not weep, it lets go, and it can throw a cap, a blind, or a section of pipe like a projectile. The pressure-piping codes treat hydrostatic as the standard test and pneumatic as the substitute you justify, exactly because of that hazard.

Pneumatic testing has a place. Where a hydrostatic test is impractical, when the system cannot tolerate the water, cannot carry the fill weight, or cannot be fully dried afterward, the ASME B31 codes allow a pneumatic or a combined test, taken with the stored-energy hazard explicitly in mind and run to a written procedure with exclusion zones and a lower test factor. Verify the exact provisions against the applicable B31 section and edition, because the allowance, the test factor, and the required precautions are spelled out there and they differ by code.

On a building chilled-water loop you almost never need air. The pipe holds water, the structure holds the weight, and the system gets flushed and filled anyway, so hydrostatic is both safer and closer to how the system will actually live. Reach for pneumatic only when there is a real reason the water cannot go in, and when you do, treat it as the more dangerous test it is.

What pressure do you test chilled-water piping to?

Chilled-water piping is commonly hydrostatically tested to 1.5 times the design or working pressure of the section. For building services piping, which is where most data center chilled-water, condenser-water, and HVAC loops fall, the governing reference is ASME B31.9, and it sets the hydrostatic test pressure at not less than 1.5 times the design pressure. The same 1.5 times factor shows up in ASME B31.1 power piping and in B31.3 process piping for hydrostatic tests, with code-specific adjustments. The applicable code section, the edition, and the project specification control the exact number, not this rule of thumb.

Pin down which code applies before you set a number, because it changes the rigor, not usually the factor. B31.9 covers building services piping, the lower-pressure and lower-temperature water, steam, and condensate systems in commercial and institutional buildings, and that is the usual home for a data center chilled-water loop. B31.1 covers power piping, the higher-pressure and higher-temperature systems in generating stations and central plants, and a large district or campus plant can land there instead. B31.3 covers process piping. The factor is similar across them, but the design rules, examination, and acceptance behind it are not, so the code call matters.

Two practical cautions on the number. The test pressure is referenced to the design pressure at test temperature, and the section under test is only as strong as its weakest component, so a single Class 125 fitting or a gauge glass rated below the test caps what you can safely apply. And where a pump, a chiller barrel, or an expansion tank cannot take 1.5 times the system design pressure, you isolate it rather than test through it. Confirm the design pressure of the section from the drawings and the equipment ratings before you ever touch the pump on the test rig.

How long do you hold the test?

The code minimum is short and the spec acceptance is usually longer, and people confuse the two. ASME B31.9 and B31.1 require the test pressure to be held for a minimum on the order of 10 minutes so the examiner can walk the system and look for leaks, after which the pressure may be reduced to the design pressure for the close inspection. That 10 minutes is the floor for the examination, not the acceptance hold a data center specification asks for.

Project specifications routinely call for a longer held test as the acceptance criterion, commonly in the range of 2 hours, and some owners and tougher specs push to 4 hours or more with a stated maximum allowable pressure drop, sometimes zero loss after temperature is accounted for. The longer hold is what catches the slow weep that ten minutes hides, and it is also what drags thermal drift into the picture, which is the next problem. Verify the required hold and the allowable drop against the project specification, because that number, not the code minimum, is what you pass or fail against.

A clean way to think about it: the 10-minute code hold proves the system did not fail under pressure, and the multi-hour spec hold proves it is actually tight. Hold the section, log the pressure and the water temperature at intervals, and do not call a pass until the temperature has settled and the gauge has settled with it. A test that ends while the water is still warming or cooling is a test you cannot read.

Thermal drift: telling a leak from a temperature change

On any hold longer than a few minutes, the gauge moves with temperature, and the move can look exactly like a leak. A water-filled section of steel pipe is a nearly rigid container full of a nearly incompressible fluid, so a small change in water temperature makes a surprisingly large change in pressure. A drop of a few degrees in the test water can pull the gauge down by several psi in a closed section, with no leak anywhere. The same swing warming up pushes the gauge up. This is thermal drift, and it is the single most misread result on a long hold.

The fix is to log both. Record the test water temperature and the ambient temperature alongside the pressure at every reading, so when the gauge falls you can ask the first question: did the water cool. A falling pressure that tracks a falling temperature, and recovers when the temperature recovers, is thermal, not a leak. A falling pressure with steady temperature is a leak. You cannot tell the two apart from the gauge alone, which is exactly why a thermometer is part of the test rig, not an afterthought.

The classic trap is the overnight hold. You pressurize in a warm afternoon, the building cools through the night, the morning gauge reads low, and a crew condemns a joint that is perfectly tight. Run the long holds when the temperature is stable, fill with water near the ambient temperature so it is not chasing equilibrium for hours, shade an exposed gauge and pipe from direct sun, and read the drop against the temperature log before you blame the pipe. The water and the steel are telling you the truth. You just have to read both instruments.

Test medium, glycol, and freeze protection

Test with clean water, and know which water. Most specs call for clean, potable, or filtered water for the hydro test, because the test water can carry the same debris and chemistry problems you are about to flush out, and a dirty fill seeds the system you are trying to keep clean. Some specs require treated or demineralized water, particularly where the test fill will become part of the operating charge or where chlorides in the water can attack stainless. Confirm the water quality the specification calls for rather than running the yard hose into a stainless loop.

Glycol is a question, not a default. The operating chilled-water loop may run a propylene-glycol mix for freeze protection or for a low-temperature design, but the hydro test is usually done on water and the glycol charged later, unless the spec says otherwise. The reason to test on water is cost and cleanup. The reason a spec might call for testing on the operating fluid is to prove the system tight on the chemistry it will actually hold.

Freeze protection is the part that bites in cold weather. Water left in a tested section that drops below freezing will split a fitting or a pipe and hand you a leak you created after the test passed. In freezing conditions, either keep the section heated and circulating, add temporary freeze protection, or drain the section completely and promptly after the test. Draining a complex loop fully is harder than it sounds, the low points and dead legs hold water, so plan the drain points and the blow-down before you fill, not after the forecast turns. A burst from a missed pocket of trapped water is a self-inflicted callback.

Isolating the test section

Half of a safe hydro test is deciding what not to pressurize. The section under test gets bounded with blinds, blanks, or test plugs, and everything in that boundary has to be able to take the test pressure. The components that usually cannot are the ones you isolate or remove: chiller evaporator and condenser barrels, pumps and their seals, expansion tanks, control valves with limited body ratings, instruments, sight glasses, and anything rated Class 125 or below in a section you intend to push to 1.5 times a higher design pressure.

Relief valves are the item people forget, and the consequences run both ways. A relief valve left in the section will lift at its set pressure, which is at or below the working pressure and well under the test pressure, so it dumps your test the moment you try to build pressure. Remove the relief valve and blank the connection, or gag it only if the procedure and the manufacturer allow, and put it back the instant the test is done. A relief valve left gagged and forgotten is a removed safety on a live system, which is a far worse problem than a failed test.

Isolation valves are not blinds. A closed gate or ball valve seats against pressure but it is rated to seal at the system pressure, not necessarily to hold 1.5 times it as the only boundary of a test, and a valve can pass enough across the seat to confuse a long hold. Where the boundary matters, use a blind or a blank, not a shut valve, and treat any valve you do rely on as a suspect if the pressure will not hold. Mark up the test section on the drawing with every blind, every removed component, and every boundary valve, so the reinstatement at the end puts the system back exactly the way it has to be.

The gauge and instrumentation

The gauge is the witness, and an uncalibrated or wrong-range gauge invalidates the whole test. Size the gauge so the test pressure lands near the middle of the scale, commonly somewhere around the middle third, because a gauge reads most accurately in its mid-range and worst near the ends. A test at 225 psig on a 0 to 300 psig gauge is crowding the top; the same test on a 0 to 400 psig gauge sits where you can actually read tenths. Too large a range and small drops disappear in the needle width. Too small and you risk over-ranging and damaging the gauge.

The calibration is the part the witness checks first. The gauge carries a current calibration sticker traceable to a standard, with a date inside the calibration interval, and the calibration record goes in the package. An expired or missing calibration is a finding on its own, regardless of what the needle showed. On longer or higher-value tests, a calibrated pressure recorder or a datalogger trends the pressure continuously and timestamps it, which beats hand-logged readings for proving a flat hold and for catching a drift the eye misses between readings.

Add the thermometer to the rig and treat it as test equipment, not a convenience. You are logging water temperature and ambient temperature to read thermal drift, so the temperature instrument needs to be readable and, on a tight acceptance, calibrated too. The full instrumentation set for a defensible test is a mid-scale calibrated gauge, a temperature reading at the water and the ambient, and on the bigger tests a recorder that timestamps both.

Flushing and cleaning the system

Pressure testing and cleaning are two jobs, and the order between them matters. The hydro test proves the piping holds. The flush removes the construction debris, weld slag, pipe scale, cutting oil, and thread sealant that every field-built loop comes full of. A pressure test on a dirty system passes fine, because debris does not leak, so the test does not clean anything and the flush is a separate hold point that comes after the system is proven tight.

The cleanliness target is set by what the water has to touch downstream. Chiller barrels, CRAH and CRAC coils, control valves, and the narrow passages in a coolant distribution unit heat exchanger all foul or clog on the grit a new system sheds. Strainers go in at the pumps and the equipment, and the start-up strainer screens, the fine temporary screens, get pulled and cleaned repeatedly through the flush until they come out clean. The flush circulates at a velocity high enough to scour the pipe walls and carry the debris to the strainers rather than letting it settle in the low points.

This is the same discipline the liquid cooling loop demands one step closer to the silicon, where a cold plate micro-channel will choke on debris a chilled-water coil would shrug off. On the building loop the consequence is a fouled coil that quietly loses capacity. On the cold-plate loop it is a throttled chip. Either way, a loop that was pressure-tested but never properly flushed and cleaned is only half done, and the half that got skipped is the one that shows up as lost capacity months later.

How do you fill and vent before the test?

Fill from the low point and vent from every high point, because trapped air ruins the test. Air in a water-filled section makes the test spongy: the trapped gas compresses, so the pressure builds and bleeds in a way that masks a small leak and makes the gauge wander on its own. A pocket of air at a high point also hides a leak right at that pocket, since the air cushions the pressure change a weep would otherwise show. The goal is a water-solid section with the air pushed out ahead of the fill.

So you open the high-point vents and fill slowly from the bottom, letting the rising water push the air up and out the vents until each one runs solid water with no spitting. High points on a real loop are more than the obvious tops of risers. They are the tops of coils, the crowns over equipment, the up-and-over around structure, and the dead legs, and every one of them needs a vent or a way to bleed. Fill too fast and you trap air the vents cannot catch; fill from the top and you trap it for certain.

Bring the section up to test pressure in stages, not in one shot. Pressurize part way, pause, check for obvious leaks and recheck the vents, then step up to the test pressure. Stepping up lets you catch a gross problem at low pressure instead of finding it at 1.5 times design, and it gives any last trapped air a chance to show itself. A section that will not hold a steady pressure at the first step usually has air, a passing valve, or a leak, and it is cheaper to find out which at low pressure than high.

Finding the leaks

When the gauge will not hold and the temperature is steady, you walk the section and find the leak by hand and eye. The places to look are the connections, in roughly the order they fail: threaded joints first, then flanges and their gaskets, then valve bonnets and stems, then mechanical couplings and grooved fittings, then the welds. A field weld weeps less often than a threaded or flanged joint, so the threads and flanges get the first and closest walk. Pipe wall failures are rare on new pipe and usually mean a damaged section or a bad batch.

Look, feel, and listen. A real leak under test pressure usually shows as water, a bead at a thread, a damp gasket edge, a drip building under a flange, so a dry walk of every joint with good light catches most of them. Run a dry hand or a paper towel around the underside of a suspect joint where a drip you cannot see collects. A high-pressure pinhole can hiss or spray fine enough to feel as a mist before you see it, so listen at the quiet joints. On a long hold, the leak that drained your pressure may have left a telltale wet spot or a mineral track even after the pressure is gone.

Mark what you find, drop the pressure to repair, and retest the section, do not just nip up the joint under pressure and call it good. A joint tightened under load may seal for the test and weep in service, and a repair made under pressure is a safety problem. Depressurize, fix it dry, and run the hold again from the start. The retest is not optional. A section is only proven if it passed the full hold after the last repair, not before it.

The test package contents

The test package is what turns a test into a record someone can trust without having been there. It is built before the test, filled in during it, and signed at the end. The marked-up drawing or P&ID is the heart of it: the section under test outlined, every blind and blank located, every isolated or removed component noted, and the gauge and fill points shown. Without that drawing, no one can later say exactly what was and was not proven.

Around the drawing sit the supporting documents. The test-pressure calculation ties the 1.5 times factor to the section's design pressure and to the governing code and spec. The gauge calibration certificate shows the instrument was traceable and in date. The completed test record carries the readings, the start, hold, and end pressures with the water and ambient temperatures at each, the hold duration, and the pass or fail against the spec acceptance. And the signatures close it: the contractor who ran it, and the commissioning agent, owner, or authority who witnessed it.

Build the package as a checklist of these items and the test gets run right, because the blanks force the steps. A package missing the cal cert, the marked-up drawing, or the temperature log is not a passed test with paperwork to follow. It is an unproven section, because the parts that make the result defensible are the parts that are missing.

Witnessing, signoff, and turnover

A hydro test that nobody witnessed is a hydro test you may get to run again. The acceptance test is usually witnessed by the commissioning agent, the owner's representative, the engineer, or the authority having jurisdiction, depending on the project and the code, and the witness wants to see the calibrated gauge, the in-date cal sticker, the section boundaries on the drawing, and the hold from start to finish, not just the final number on a form. Schedule the witness for the start of the hold, not the end, because a witness who arrives to a passed reading on a sheet has witnessed paperwork, not a test.

The witnessed hydro is a hold point in the larger commissioning sequence, and it gates what comes next. You do not insulate, flush, or charge the system on an unwitnessed or failed test. Once the section passes and the package is signed, it rolls up into the mechanical turnover and feeds the cooling commissioning: the flush and clean, the test and balance, the functional tests on the chillers, pumps, and valves, and finally the integrated systems test that proves the plant rides through a failure at load. The piping proof is the first link in that chain.

It also ties to the power side, because a chilled-water plant that cannot ride through a utility event is not commissioned, and the cooling integrated test runs alongside the electrical integrated test on a real facility. The hydro test does not prove any of that. It proves the piping holds, which is the precondition for everything downstream. Turn it over clean and the cooling commissioning can build on a loop it knows is tight.

Field example: an 8 in supply riser at a 2-hour hold

An 8 in chilled-water supply riser section was tested before insulation. The section design pressure was 150 psig, the governing code was ASME B31.9, and the spec called for a hydrostatic test at 1.5 times design, 225 psig, held 2 hours with a stated allowable drop after temperature correction. The chiller barrel and the base-mounted pumps at the bottom of the riser were isolated with blinds because their bodies and seals were not rated for 225 psig, and the relief valve on the section was removed and the connection blanked.

The section was filled from the low point with clean water near 68 F while the high-point vents at the top of the riser and over the coil bled air until each ran solid. Pressurized in two steps to 225 psig on a 0 to 400 psig calibrated gauge, with the test water at 68 F and the ambient at 70 F at the start. Two hours later the gauge read 219 psig and the water had cooled to about 65 F as the afternoon faded.

Six psi down looked like a leak until the temperature log answered it. The water had dropped about 3 F, and in a rigid, water-solid steel section that accounts for a drop of a few psi per degree, so the pressure change tracked the temperature change and the joints walked dry. The crew topped the section back to 225 psig, let the temperature settle, and the gauge then held flat within the allowable drop. Logged as a pass, with the temperature correction noted in the record so the witness and the next reader could see the 6 psi was thermal, not a weep. The psi-per-degree relationship varies with the system and the trapped-air content, so the temperature log, not a fixed conversion, is what reads the drift.

What to document

The hydro-test package is the only proof the section was ever pressurized. Without it, the next engineer has to assume the work was never done. The record is what answers the question a year out when a joint weeps over a live hall and the owner asks whether the section was ever proven. Capture enough that someone who was not there can see exactly what was tested, to what pressure, for how long, and how the result was read.

Record the system and section identity tied to the marked-up drawing, the design pressure and the test pressure with the factor and code, the test medium, the start, hold, and end pressures, the water and ambient temperature at each reading, the hold duration, the gauge identity and calibration date, the witnesses, and the pass or fail against the spec acceptance. If a drop was thermal, note the temperature correction that explains it. If a repair was made, note it and that the retest ran from the start after the repair.

Common mistakes

• Running a pneumatic test without a written justification and procedure, ignoring the stored-energy hazard of compressed gas.
• Filling without venting the high points, so trapped air makes the test spongy and hides a leak.
• Reading a pressure drop on a long hold as a leak when the water cooled, with no temperature log to tell the difference.
• Testing through a chiller barrel, pump, expansion tank, or instrument that is not rated for the test pressure.
• Leaving a relief valve in the section so it lifts and dumps the test, or gagging it and forgetting to restore it.
• Using an uncalibrated gauge, an out-of-date cal, or a range that puts the test pressure at the top or bottom of the scale.
• Insulating or charging the system before the witnessed test passes, hiding the joints you needed to inspect.
• Tightening a weeping joint under pressure and calling it fixed instead of depressurizing, repairing dry, and retesting.
• Leaving water in a section through freezing weather with no heat or drain, creating a leak after the test passed.
• Treating a closed isolation valve as a test boundary when a passing seat confuses the hold.

Field checklist

Standards and references

The governing reference for most data center chilled-water, condenser-water, and HVAC piping is ASME B31.9, Building Services Piping, which sets the hydrostatic test at not less than 1.5 times the design pressure and a minimum hold for the leak examination. Where the piping is higher pressure or temperature, as in a central or district plant, ASME B31.1, Power Piping, may govern instead, with a similar 1.5 times factor and its own design and examination rigor. Process and specialty fluids fall under ASME B31.3, Process Piping, which carries the same hydrostatic factor and the explicit treatment of pneumatic testing and its stored-energy hazard. Confirm which B31 section the project specifies and the adopted edition before citing a pressure, hold, or acceptance, because the provisions and the exact paragraph numbers move between editions.

The mechanical code adopted by the jurisdiction, commonly a version of the International Mechanical Code or a local equivalent, sits over the installation and references the pressure-piping standards. The manufacturer pressure ratings on the chillers, pumps, valves, expansion tanks, and instruments govern what gets isolated rather than tested through, and they override any general rule when they are lower. The project specification sets the acceptance hold and the allowable pressure drop, which are commonly stricter than the code minimum, and the contract documents control where they differ from the code.

Two practices keep the citation honest on a submittal. Name the B31 section that actually applies to the system in front of you rather than the most familiar one, and confirm the test pressure, the hold time, and the acceptance against the current edition and the project spec instead of from memory. ASME B31 gives the framework and the 1.5 times factor; the equipment ratings and the project documents control the limits you test to.

Units, terms, and acronyms

Pressure testing carries vocabulary from piping fabrication, from commissioning, and from the codes, and the same idea reads differently across a test record, a piping spec, and a B31 section. The terms below travel across the whole test package.

Hydrostatic test

A pressure test using water (or another liquid), the safe default because water stores little energy and a failure does not let go violently

Pneumatic test

A pressure test using air or gas, allowed only when hydrostatic is impractical, because compressed gas stores far more energy and fails dangerously

psi / psig

Pounds per square inch; psig is gauge pressure, above atmospheric, which is what the test gauge reads

Design / working pressure

The pressure the system is rated and designed to operate at; the basis for the test pressure

Test pressure

The raised pressure held during the test, commonly 1.5 times the design pressure for a hydrostatic test

Thermal drift

Pressure change on a long hold caused by the test water warming or cooling, read against the temperature log so it is not mistaken for a leak

Blind / blank

A solid plate isolating the test section at a boundary, used instead of relying on a valve seat

ASME B31.9

Building Services Piping, the usual code for data center chilled-water and HVAC piping; sets the hydrostatic test at 1.5 times design pressure

ASME B31.1

Power Piping, for higher-pressure and higher-temperature systems such as central plants

Hold time

How long the test pressure is maintained; a code minimum around 10 minutes for examination, often 2 hours or more for spec acceptance

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is bypass airflow in a data center?
• What is direct-to-chip liquid cooling?
• What is thermal energy storage in a data center?
• What is an as-built drawing?
• Why do battery rooms need ventilation?