The Loop That Froze With Antifreeze In It

Before anyone opens the loop

A split glycol loop is a pressurized, possibly hot, chemical-bearing system with ice still in it. Isolate the section at its shutoff valves and relieve pressure through a drain valve into a container or a floor drain receptor before loosening any joint, because a bulged tube that has not torn through yet can open under your hands. Confirm the fluid is below scald temperature before you crack anything: burn-prevention literature puts a serious scald at around 140 F in a few seconds and at 120 F in a few minutes, and glycol at operating temperature carries no visual warning. If a section is still frozen, warm it with heat trace or warm air and never with an open flame, and vent the section first, because thawing between two ice plugs traps expanding water in a closed volume and ruptures it at the point you are standing beside.

Read the safety data sheet for the specific fluid before you handle it and take your glove class and any respiratory control from that sheet, not from habit. Ethylene glycol is acutely toxic if swallowed and its breakdown products are too; propylene glycol has far lower acute oral toxicity, and neither belongs in a potable line. If the loop's fill connection ties to building water, treat that as the first thing you inspect, not the last.

The call, and what was left of the evidence

The call was a rooftop unit with no preheat and water running down an interior wall. By the time anyone arrived, the branch had been drained by the building engineer and the split coil was already off the roof. What was left was a torn tube, a drained branch, and a folder.

The tube told the first half of the story. Freeze damage bulges before it tears: the tube swells over a length, then splits longitudinally along the swell, and the fracture edges are clean and drawn rather than eaten. A corrosion pinhole is round, sits in a pit, and has oxide around it. This was a freeze split, not a corrosion failure, so the question was never "what attacked the metal" but "how did water at that spot get to its freeze point."

The building's answer was the obvious one: the glycol must be weak. That answer is available, cheap, and wrong often enough that it is worth the hour it takes to disprove.

Reconstructing the fill from three records

Three documents existed, and together they are enough.

The commissioning fill log recorded a calculated system volume of 400 gallons and 160 gallons of inhibited propylene glycol concentrate added at fill, which is 40 percent by volume.

Last season's PM sheet recorded a refractometer reading of 34 percent taken at the pump discharge, with no note about what was done with it.

The makeup water meter on the fill line read 46 gallons since commissioning.

Work the meter against the fill log. Every gallon that leaves a mixed loop leaves at loop concentration and comes back as plain water. Forty-six gallons out at roughly 40 percent carries about 18 gallons of glycol out of the system, so the remaining glycol is about 142 gallons in a 400 gallon volume, or roughly 35 percent. A continuous-loss model lands slightly above that, because later losses carried less glycol than earlier ones. The point is the size and the direction, not the third digit: the records predict mid-30s, and the refractometer read 34 percent. Two independent records agreeing is what lets you stop testing the bulk fluid and start looking somewhere else.

The night in question bottomed at 11 F. Read the freeze point for a given concentration off the fluid manufacturer's own table, because inhibitor packages shift it and the curve is not linear: for propylene glycol the freeze point falls steeply between roughly 20 and 50 percent by volume, flattens, and reaches its minimum near 60 percent, above which it starts rising again. At mid-30s percent the table put this fluid's freeze point well below 11 F. The bulk fluid was not the failure.

The makeup meter is the only honest witness

That 46 gallon reading did more than confirm a concentration. It bounded what could have happened. A loop that has taken 46 gallons of makeup in three winters has a slow loss, not an event: a weeping pump seal or an air vent passing a little. A loop that had taken 200 gallons would have been a different article, because that is dilution fast enough to walk the freeze point up into the weather.

Most loops carry no meter, and the ones that do rarely have anyone reading it. Fit one, or fill manually from a marked drum and log the volume. An automatic makeup valve left open on a glycol loop is a device whose entire job is to quietly replace your freeze protection with tap water, and it does that job well. If you close the automatic fill to stop that, fit or verify a low-pressure alarm or a pump cutout first, because a loop with no automatic makeup and no low-pressure protection answers a real leak by running the pump dry.

The branch nobody re-opened

The last document in the folder was a work order from 18 months earlier: the same preheat coil had been replaced, the branch isolated at two shutoff valves, and the coil filled from a hose to check the joints. A control problem was noted as outstanding. Nobody wrote that the valves were reopened, because nobody reopened them.

That branch held roughly 30 gallons behind two closed valves, filled with plain water, in an outdoor air stream. It had no glycol, no flow, and no route for expansion. It did not need an unusual night. It needed one night below 32 F, and it had already had many.

This is the load-bearing point of the case, and it generalizes past glycol: freeze protection lives where the fluid is, not where you added it. A refractometer reading taken at the pump tells you about the fluid that reaches the pump. Any leg that has been isolated, dead-headed, valved off, or refilled separately carries its own concentration and answers to nothing you did at the fill point.

The direction check confirms it. If weak bulk fluid had been the cause, the damage would have appeared on the coils seeing the lowest fluid temperature and the lowest flow, plural, and the measured concentration would have had to sit near the freeze point of that night rather than far below it. One split, on the one branch with no circulation, on a loop whose two records agree, points at isolation rather than dilution.

Sampling so the number means something

Three habits separate a reading you can act on from a number in a box.

Sample where the fluid is coldest and slowest, not where it is easiest. A pump discharge sample is the best-mixed fluid in the building. The sample that matters comes from the far end of the outdoor run, taken while the loop is circulating.

Cross-check the instrument. A refractometer reads refractive index and converts it on a scale printed for one fluid family. Read ethylene glycol on a propylene scale and the number is wrong in a direction that flatters you. Confirm the scale matches the fluid, zero the instrument on distilled water, and read at the instrument's calibration temperature or use its correction.

Read the chemistry, not just the concentration. Glycol oxidizes into organic acids when it is run hot or exposed to air, the inhibitor package is consumed defending the metal, and pH falls as that happens. A loop at the right concentration with spent inhibitor and a pH heading down is a corrosion problem wearing a freeze-protection number. Send a sample to the fluid supplier's lab annually and act on the reserve alkalinity and pH, not only the percent.

The response that would have made it worse

The building's first instinct after the split was to raise the whole loop to 50 percent and add a margin on top of that. Two things push back.

Concentration buys freeze protection and costs pumping and heat capacity. Glycol solutions carry less heat per gallon than water and get markedly more viscous as they get colder, so the coil that was designed on water needs more flow to move the same heat and the pump sees more head to deliver it, and both effects are worst at exactly the temperature you added the glycol for. Near room temperature the penalty is modest; at the cold end it is not.

And the freeze curve turns around. Past the eutectic near 60 percent by volume, adding more glycol raises the freeze point again, so a shop that keeps topping up "for safety" can walk itself back toward the weather from the other side while every drum it adds feels like progress. Set the target concentration from the design low temperature plus a stated margin, confirm it with a table from the fluid maker, and hold it. Do not treat concentration as a dial that only goes one way.

Freeze point and burst point are also two different numbers and the fluid's table lists both. Freeze point is where the first ice crystals form; burst point is the lower temperature at which the resulting slush expands enough to split the pipe. Protecting to burst but not to freeze is a real design choice for a drainable line that carries no load, and it is the wrong choice for a coil with outdoor air moving across it, because the moment flow stops the coil goes to air temperature and the next hour is the failure.

Closing the record

The repair was the easy part. The fix was three lines of paperwork.

Every isolation valve closed during service gets a tag with the date and the tech, and the work order does not close until the tag is removed or the open item is written into the PM. The fill line gets a meter, the meter gets read at every PM, and the reading goes on the sheet next to the refractometer number so a rising makeup volume and a falling concentration are visible on the same page. And the annual sample goes to the fluid supplier's lab for concentration, pH and inhibitor reserve, because two of those three were never being measured at all.

References

• OSHA 29 CFR 1910.1200, hazard communication, for the requirement that the safety data sheet for the specific glycol product is available and is the source for glove class and respiratory control
• OSHA 29 CFR 1910.147, control of hazardous energy, for isolation and stored energy on the pumps and pressurized sections before opening a loop
• The adopted plumbing code in your jurisdiction for the backflow assembly class required on a fill connection between building water and a fluid classed as a health hazard
• Fluid manufacturer freeze point and burst point tables, and the manufacturer's own concentration limits and inhibitor reserve targets
• See related: The Water Chemistry That Attacks a System; Commissioning the Water Side of a New System