Why Your Chiller Works but Your Process Won't Get Cold: 119 Secondary Refrigerant Failures, Sorted by Cause

Your compressor is doing its job. Suction pressure sits where it always has. But the jacket on your reactor is six degrees warmer than spec, and nobody on shift can tell you why.

We get this call a lot at Glacier Coolant. Over the past three years we logged the root cause of 119 industrial service calls on secondary refrigerant loops — the circulating heat-transfer fluid that carries cold from your refrigeration unit out to the process. Almost half of them were not refrigeration problems at all.

Here is the breakdown, what each failure actually looked like on site, and how we work through them.

What 119 service calls actually showed

Failure category	Share of cases	Typical symptom you'd see first
Poor cooling performance	45%	Process temperature drifts up, compressor runs longer
System corrosion	23%	Leaks, discoloured fluid, sludge in the tank
Ice blockage and restriction	18%	Pump seizes intermittently, pressure swings, then recovers
Other operational faults	14%	Foaming, pressure spikes, odour, deposits

None of these are evenly distributed across industries, by the way. Pharmaceutical batch reactors fail differently from cold stores, and cold stores fail differently from data-centre loops. But the four categories hold up every time.

The cost side is worth stating plainly. Across the loops we surveyed, energy consumption ran 15–25% above design once any of these faults took hold. Pumps and heat exchangers aged faster. In the worst cases the line stopped.

Cause 1: The loop was never designed for this fluid

This is the one nobody wants to hear, because it's expensive to fix. It accounted for a large slice of that 45%.

A flower cold-storage facility in Kunming had roughly 152 tonnes of fluid in the system, mixed from about 51 tonnes of concentrate. The pipework was undersized and the tank had no baffle. Flow distributed unevenly between branches, so some evaporators starved while others short-circuited. Energy use climbed and temperature never stabilised. Two hardware changes — bigger lines and a baffled tank — solved what no amount of fluid tuning could.

Pump-to-exchanger mismatch shows up the same way. On a heat-source tower project in Beijing the pump was undersized for the design flow. No fluid change was going to produce the duty that the heat exchanger needed.

Then there is air. Three separate cases traced corrosion back to air entrainment:

• One aerospace project had no water seal on the make-up tank. Air mixed in continuously and drove electrochemical corrosion in the exchangers.

• A Beijing data-centre loop had no vent at the pump discharge. Cavitation pitted the impeller and casing.

• At a Huanghua facility, the return line sprayed into the tank and dragged oxygen down with it. Those exchangers failed twice in one year.

If you take one thing from this section: the tank design matters as much as the pipe sizing. The multi-discipline check we now insist on during design review covers water seals, nitrogen blanketing and vent points as line items. They used to be left to the installer.

Cause 2: Someone swapped the fluid without redoing the math

Fluid substitution caused another chunk of failures, and this one is entirely preventable.

A customer in Taizhou replaced their charge with a deep-cold grade, expecting better performance at low temperature. It got worse. The replacement had lower specific heat and lower thermal conductivity, so it simply moved less energy per litre. The machine wasn't broken. The fluid was carrying less heat than the system was designed around.

It cuts both ways. Swapping a salt-based fluid for a hydrocarbon-based deep-cold grade means you may need DN40 or DN50 line sizes, roughly double the flow, and about 1.5× the heat-exchange area. Nobody budgets for that six months after commissioning.

Viscosity deserves its own note, because it's the number people panic about unnecessarily. One chemicals customer was worried about a fluid measuring 17.59 cSt against 12.93 cSt for the conventional grade at -50 °C. They assumed pressure drop would balloon. When we ran it, the effect on flow resistance was 7.5% and on heat transfer about 1.5% — comfortably inside design tolerance. High viscosity at low temperature is normal. It is not automatically disqualifying.

The other side of the coin is concentration. A cold store in Linyi specified -35 °C and ran the fluid too rich. At that concentration the viscosity rose to where the pump simply could not move it. The best temperature they ever achieved was -28 °C. Adding concentrate past the optimum does not buy you extra cold. It costs you flow, and flow is what actually delivers cooling.

Cause 3: Water got into a system that can't tolerate it

Ice blockage accounted for 18% of cases, and almost all of them trace back to water ingress.

Here's the distinction that matters. Some products are water-miscible — they take up water and the freezing point rises gradually. Others, particularly deep-cold hydrocarbon grades, don't mix with water at all. In those systems water stays as a separate phase, migrates to the coldest point, and freezes there.

Typical sequence: cooling gets progressively harder, the pump binds intermittently, pressure gauges swing, and everything recovers once things warm back up.

Two separate plants hit this on the same deep-cold grade. One of them, run by a refrigeration contractor in Guangzhou, had a pump seize within ten days of commissioning.

The three-stage response we use:

1. Immediate. Bring fluid above 0 °C to melt the ice, let it stand so the phases separate, and drain the water off the bottom. Fit a fine strainer ahead of the pump.

2. Deep treatment. Circulate through a 3A molecular sieve to pull dissolved water out. Below -50 °C this isn't optional. If contamination is severe, the charge has to go.

3. Prevention. Nitrogen blanket the tank — 1,500 Pa is what we recommend. Pressure-test after any installation or repair with dry nitrogen and purge thoroughly, not water.

One habit worth building into your specification: for water-free systems, pressure-test with dry nitrogen. Looking for leaks with water is how these systems get contaminated in the first place. In an emergency some plants have dosed alcohol to depress the freezing point and get through a shift. It works short-term and it is not a fix.

Cause 4: Operation and process drift

The last group is uncomfortable because usually somebody made a decision that seemed reasonable at the time.

At a Taizhou pharmaceutical plant in late 2025 an operator closed the inlet and outlet valves on a reactor jacket while the loop was running. Thermal expansion had nowhere to go. Pressure climbed until the reactor bulged, and it was scrapped. A closed valve on a running loop is a bomb with a slow fuse.

Six months earlier, at Chongqing Changyu Electromechanical, a high-temperature silicone-based fluid was heated toward 106 °C. Residual moisture boiled, the charge heaved, and it overflowed. Any process change that pushes toward the boiling point needs a residual-water check and a small-scale trial first.

Neither of these was a fluid defect. Both were unmanaged change.

The five loops we run on every call

1. Diagnose the problem. Record observable facts only. No guessing at causes yet.

2. Analyse root cause. Not stop at the nearest explanation.

3. Build the plan.

4. Implement and verify.

5. Summarise and archive.

If the result falls short, go back to step 1. That sounds obvious. It is also the step most people skip, and it's why the same site often sees the same failure twice.

For step 2 we use a fishbone across five headings:

• People — qualifications, induction training, operating errors, attitude.

• Machine — pump/valve/exchanger match, wear, whether maintenance actually got done.

• Material — is the grade right for the duty, is it contaminated, has the chemistry degraded, is concentration on spec.

• Method — are procedures written down, followed, and reviewed.

• Environment — humidity accelerates water ingress, acidic gases pull pH down, stray electromagnetic fields from transformers can drive electrochemical corrosion, dust and oil films interfere with heat transfer.

That environment line surprises people. We've seen it, though, and it's worth checking before replacing expensive hardware.

Fixes land in four directions: modify the loop (piping, pump, exchanger, vents, pressure-holding feed), treat the fluid (change grade, correct concentration, dose performance additive or antifoam, dewater), change operations (revise SOP, train, schedule maintenance), or adjust the process (reset parameters, avoid the critical band).

Symptom triage: what to check first

What you observe	Check in this order
Temperature won't come down	1. Measure freezing point — is concentration on spec? 2. Is evaporation temperature approaching the freezing-point limit? 3. Is low-temperature viscosity cutting pump flow? 4. Has end load exceeded design? 5. Is flow matched to exchanger and pump? 6. Ice or debris restriction? Try warming above 0 °C, circulate, then step down to locate the critical point.
Pressure climbing unexpectedly	Usually thermal expansion with no relief path. Fit a pressure-holding feed unit. If using head nitrogen, calculate partial pressure or you'll get flashing at height.
Temperature differential shrinking	Often lower end load or excessive pump flow. Reduce frequency and flow. If production isn't affected, this may actually help you.
Foaming	Sample it. Dose antifoam — verified on a Beijing data-centre loop. Foaming leads to cavitation and pipe damage.
Fluid turned black or smells sharp	Sample and send in. Check for ammonia, refrigeration oil, or cross-contamination.
Performance dropped after topping up	Cold dosing can precipitate additives and block plate exchangers. Warm to ambient and circulate.
Considering a grade change	Recalculate specific heat, thermal conductivity and viscosity, and keep the record. Never swap directly.

That last row is the one that would have prevented most of Cause 2