The Heat Exchanger Leak That Leaves No Puddle
Every hose is dry. The floor is spotless. Still, the coolant inventory or chemistry has started to move.
Most leaks announce themselves by escaping the system. This one can disappear into another loop instead.
Inside a cooling distribution unit, a heat exchanger lets heat pass between the facility water system and the technology cooling system while keeping the two fluids apart.
A microscopic hole changes that agreement. Fluid can cross internally, leaving no puddle and no obvious leak location.
The hardest leak to spot may be the one that never reaches the floor.
The border that only heat should cross
Picture a brazed plate heat exchanger as a tightly compressed deck of corrugated stainless steel cards.
Alternating passages carry facility water, or FWS, and technology coolant, or TCS.
The plates create a large heat-transfer area inside a compact device.
Hot technology coolant and cooler facility water pass on opposite sides of each thin plate. Heat crosses the metal. The liquids should not.
It is a border with one approved traveler: thermal energy.
The two circuits may use different materials, treatment programs, glycol concentrations, cleanliness standards, and operating histories.
ASHRAE water classes describe temperature envelopes for liquid-cooled equipment, not chemistry grades.
The actual FWS chemistry still comes from the facility specification and treatment program.
If a hole forms, the labels primary and secondary do not decide the direction. The higher local static pressure at the breach pushes fluid toward the lower-pressure side.
Change the pressure relationship below and watch the same opening tell two different stories.
One hole, two possible stories
When FWS pressure is higher, facility water enters the technology loop.
If the TCS contains glycol and the incoming fluid contains less, glycol concentration can fall.
The incoming water also brings its own dissolved ions, treatment chemistry, particles, and corrosion history.
Conductivity does not have to rise. It may rise or fall depending on the two fluids.
The stronger clue is a chemistry departure from known-good operation, read alongside glycol concentration and the fill record.
Think of two cups holding different recipes. A spoonful transferred between them changes more than the liquid level. It changes the mixture.
That is why concentration and chemistry belong on the same timeline.
When TCS pressure is higher, coolant moves into the facility loop. The technology loop then loses inventory while every rack-side hose remains dry.
Correct premix may replace the loss, keeping TCS chemistry surprisingly steady while makeup consumption climbs.
The receiving facility loop can dilute that coolant into a much larger water volume.
The clearest evidence may therefore be the missing fluid on one side, not a dramatic chemistry alarm on the other.
A normal top-up leaves a receipt
A legitimate top-up has a time, a volume, a known fluid, and an expected sample response.
An internal cross-leak is more like a bank balance changing without a transaction in the ledger.
- Recorded top-up. The fluid movement agrees with the service record.
- Possible FWS ingress. Glycol and chemistry move with no matching TCS addition.
- Possible TCS egress. Makeup rises without a visible external leak, while TCS chemistry may remain near baseline.
Compare the three fingerprints below. Move through the timeline to see how several ordinary signals become meaningful when they move together.
These are diagnostic patterns, not universal alarm limits and not proof by themselves.
Build the case before ordering the test
No single instrument can look through a brazed plate and announce a pinhole. The case becomes credible when independent evidence agrees.
- Pressure on both circuits. Measure close enough to the exchanger to understand the local differential across relevant pump and valve states.
- Reservoir level and makeup volume. Track actual volume, not only a low-level alarm or pump run command.
- Read glycol and chemistry together. Compare approved measurements with known-good operation and both source fluids.
- Samples from both sides. The receiving circuit can carry evidence the source circuit cannot reveal alone.
- Maintenance and fill records. Reconcile known additions, drains, vents, and sampling events before blaming a hidden transfer.
A small internal breach may move chemistry long before it moves enough heat to trigger a thermal alarm.
Once the pattern is credible, isolate and test the exchanger using the manufacturer or qualified service provider procedure.
Manufacturer guidance describes pressure and integrity tests for internal plate leaks and microscopic cross-contamination paths.
The pinhole is the final frame
The opening is the end of a longer physical story.
Its root cause may sit in water chemistry, plate or braze compatibility, pressure transients, temperature cycling, freeze exposure, debris, cleaning practice, or several factors acting together.
This is why a universal chloride cutoff can mislead. Manufacturer corrosion guidance ties risk to material, pH, temperature, oxidizing conditions, and the full water composition.
Copper brazing can also be vulnerable to ammonia and sulfur compounds.
Open each failure mechanism below and move its control.
The scenes explain the physical route; the actual root cause still belongs to inspection, operating history, testing, and material evidence.
Four ways the barrier can be challenged
- Corrosion. An incompatible water environment can attack stainless plate, copper braze, or a vulnerable local site.
- Cycling. Pressure spikes, water hammer, valve movement, and thermal cycling can repeatedly flex the same plate or braze joint.
- Freeze exposure. Ice expansion inside a narrow water channel can deform the plate pack and damage its pressure boundary.
- Debris. Particles can restrict passages, disturb flow, raise pressure drop, and encourage fouling or corrosion.
How often does it happen?
There is no credible public fleet-wide failure rate for pinholes in data center brazed plate heat exchangers.
Manufacturers document the failure mode, causes, test methods, and real cases.
They do not publish the installed-unit years needed for a defensible fleet-wide percentage.
Two statements can therefore be true.
Internal cross-leaks are not expected operating behavior, and they are real enough that major manufacturers publish guidance for finding and preventing them.
The risk earns attention through consequence, not a made-up frequency.
One breach can expose an entire technology coolant charge to fluid whose chemistry and cleanliness were never selected for rack-side materials.
A failure does not need a fictional percentage to deserve a detection plan.
What to do when the pattern appears
- Preserve the timeline. Capture pressure, flow, makeup, level, alarms, chemistry, and interventions before changing the system.
- Verify the instruments and the ledger. Rule out sensor drift, sampling error, an unlogged fill, and known drain or vent events.
- Establish the likely pressure direction. Compare both circuits at the exchanger across relevant pump and valve states.
- Sample both fluids. Use approved methods and compare with commissioned baseline samples and known source fluids.
- Escalate to an integrity test. Follow the manufacturer and site safety procedure for isolation, pressure testing, or qualified leak detection.
- Decide the fate of the affected charge. Follow the approved fluid and equipment plan for cleaning, flushing, replacement, and return to service.
A chemistry trend is not a pressure-boundary test. Its job is to tell you when that test has become worth doing.
Make the baseline a living operating picture
A baseline is not one sample collected on day one and forgotten.
It starts with commissioned, known-good operation and grows into a record of normal loads, pump states, seasons, and service events.
That living history is what makes an unusual change visible.
- Watch the fluid and the loop together. Trend approved fluid-condition signals beside makeup, pressure, flow, level, temperature, and filtration.
- Record the operating context. Match readings to pump and valve states, workload, maintenance, additions, drains, vents, and samples.
- Use predictive analytics to find agreement. Compare current behavior with known-good operation and look for independent signals moving together.
- Protect the meaning of normal. Update the baseline only after the team has explained the change and verified recovery.
- Consider purpose-built leak detection. When consequence justifies it, double-wall plates can route an internal failure to a visible path.
Equipment selection still depends on system requirements and supplier guidance.
This is where continuous fluid monitoring and predictive analytics earn their place.
The goal is not to collect more readings. It is to show what changed, whether the change makes physical sense, and which evidence the team should gather next.
Turn scattered clues into one operating story
A cross-leak rarely arrives as one perfect alarm.
It appears as a few quiet disagreements.
Chemistry moved without a fill. Makeup rose without a puddle. The pressure direction makes an unexplained transfer physically possible.
Reliability Engine connects coolant health, pressure, flow, reservoir behavior, makeup, samples, alarms, and interventions on one timeline.
It compares the current pattern with known-good operation so teams can notice meaningful drift earlier and investigate it with context.
The platform does not call a trend a confirmed hole. It helps the right integrity test start with a better question.
The goal is simple: know when heat is the only thing crossing the exchanger.
Operating or commissioning liquid-cooled AI infrastructure? Reliability Engine helps teams turn cooling-loop telemetry and fluid evidence into decisions they can defend.
Practical reads on coolant health, GPU thermal margin, and what to check next.
References
- Open Compute Project: Advanced Cooling Solutions Connection Guidance
- Open Compute Project: Cold Plate Cooling Loop Requirements
- ASHRAE: AI Data Center Framework and Water Classes
- Alfa Laval: Plate Heat Exchanger Troubleshooting
- Alfa Laval: Water Characteristics to Avoid Corrosion in Brazed Heat Exchangers
- SWEP: Refrigeration Handbook for Brazed Plate Heat Exchangers
- Kelvion: Brazed Plate Heat Exchanger Operating Manual
- Alfa Laval: Finding Internal Leaks in Brazed Heat Exchangers
- Alfa Laval: Double-Wall Brazed Plate Heat Exchanger