How to Purge a System and Know It Worked

Why this matters

Purging is the job most often declared finished on the wrong evidence. Bubbles stop coming out of the hose, the tech shuts the valve, and the callback arrives four days later with one cold zone. Bubbles stopping means air stopped arriving at that vent, which is not the same as air being gone: air collects at high points and in slow-moving dead legs, and it only travels to a vent if the water is moving fast enough to drag it there. Treat purging as a velocity problem with a proof at the end and the callback rate drops to near zero.

Scope note: this procedure is for liquid systems, hydronic and domestic water. Do not apply any of it to fuel gas piping. Purging gas lines creates an explosive atmosphere, must be vented to outdoors clear of ignition sources and intakes, and is governed by the fuel gas code adopted in your jurisdiction. Different procedure, different hazard, different card.

Gate 1: make the conditions survivable before any water moves

Each step below is a gate. You do not move to the next one until its own indicator passes, and a later gate that fails sends you back to a specific earlier one rather than to the beginning.

Scald is the first hazard and the action is to bring the loop below scald temperature before you open a purge valve, by running the circulator with the heat source off until the supply gauge reads down. If the schedule will not allow that, secure the hose end in a floor drain or a bucket that cannot tip, keep your body out of the plane of the valve outlet, and use a face shield rather than safety glasses, because a purge cock throws sideways.

Water near live electrical parts is the second, and the action is to open and lock the appliance disconnect before you route a hose past an open control panel. Work inside that panel is electrical work under 29 CFR 1910.333(b)(2), and if you meter anything in there, the live-dead-live proving sequence comes from NFPA 70E-2021, 120.5. If the panel does not need to be open, close it, which is faster than proving it dead.

Never restrain, plug or gag a relief valve to stop it weeping during a fill, and do not leave a manual fill bypass open unattended. A relief discharging is telling you the fill pressure is wrong; a relief that cannot discharge is a pressure vessel with no way out.

If the fluid is a glycol mix or an inhibited boiler water, read the SDS first. The route that matters when a hot line sprays is inhalation of mist rather than skin contact, so ventilate the space, and where the SDS calls for respiratory protection the employer's respirator program at 29 CFR 1910.134 applies, including fit testing before first use and at least annually under 1910.134(f)(2).

Gate passes when: the loop reads below scald temperature or the discharge is secured and you are out of its path, the appliance disconnect is open where a hose passes an enclosure, and the relief is unobstructed.

Gate 2: give the water exactly one path

Air moves where the water moves. If three zones are open, the fill flow divides between them by relative resistance and every one of them may end up below the speed that carries a bubble.

Close every path except the one you are purging. On a manifold, that means one loop at a time. On a zone-valve system, that means powering one valve open and confirming the others are shut, not assuming. On a system with a bypass or a primary-secondary tee arrangement, close the bypass, because it is a low-resistance shortcut that will take most of your flow and purge nothing useful.

Gate passes when: you can trace, by hand on the pipe or by valve position, that the water leaving the fill has one route to the purge discharge.

Gate 3: reach entrainment velocity in that path

This is the gate people skip, and it is the one that decides whether the purge works.

A bubble in a horizontal run wants to sit at the top of the pipe. It only travels with the water if the water is moving fast enough to drag it along. A commonly used purge target is at least 2 ft/s in the segment being purged, and that figure is for carrying bubbles along horizontal runs; a vertical drop where flow runs downward against the bubble's buoyancy needs more, which is why down-flowing risers and inverted traps are the last places to clear.

Velocity is flow divided by the pipe's actual inside area, so the flow you need depends on the real inside diameter, not the nominal size. For type L copper:

PEX, PE-RT and other materials carry different inside diameters at the same nominal size, so pull the actual dimension for what is in the wall rather than reusing this table.

Gate passes when: the flow you are actually achieving through the single open path meets or beats the figure for that pipe's real inside diameter. If it does not, the fix is more supply pressure, a larger fill path, or a purge pump, not more time.

Gate 4: run clear past the last bubble

Bubbles stopping is the start of the gate, not the end of it. Once the discharge runs clear, keep it running for a full minute at the same flow. Air that was sitting in a dead leg needs time to be swept into the moving stream, and the last of it typically arrives in a short burst well after the stream looks clean.

Listen as much as you watch. A stream that pulses or gulps still has slugs coming, even if you cannot see them in a bucket.

Gate passes when: 60 continuous seconds of clear, steady, non-pulsing discharge at the flow from Gate 3.

Gate 5: hold pressure with the fill isolated

Close the fill isolation and watch the system gauge for ten minutes. A system that is genuinely full and tight holds its fill pressure. A system that drifts down has a leak; a system that climbs has the heat source firing or a fill valve passing.

Gate passes when: the needle is stable for ten minutes with the fill isolated.

Gate 6: prove it at design flow, on an independent indicator

Open everything, run the system at its normal operating condition, and confirm on evidence that has nothing to do with what came out of the hose:

• No flow noise at any terminal. Air makes a distinct trickle or rush that water alone does not.
• Supply-to-return temperature difference comparable across parallel zones, measured at the same supply temperature and the same firing condition.
• Circulator running at its expected point, with no surging on the pressure gauge.
• Automatic vents quiet after the first few minutes.

Gate passes when: all four hold. A failure here sends you back to Gate 2, one zone at a time, not back to the start.

Worked run: three zones, one that would not clear

A three-zone residential loop, 3/4 in type L copper on all three zones. The fill assembly can deliver about 6 gpm at the available street pressure.

First attempt, all three zones open at once. The 6 gpm divides by resistance and lands at roughly 2.8, 1.9 and 1.3 gpm. Against the 3 gpm that 3/4 in copper needs for 2 ft/s, those are about 1.85, 1.26 and 0.86 ft/s. None of the three reaches entrainment velocity, so the discharge ran clear on the easiest zone while the other two never moved their air. Gate 3 failed and nobody checked it.

What Gate 6 caught. At design flow, zones one and two showed an 18 F drop. Zone three showed a 31 F drop at the same supply temperature and the same firing condition, and the far end of its baseboard was cold to the hand.

That combination is the low-flow signature and it is worth reading carefully, because a bigger temperature difference sounds like more heat and is the opposite. Using the sensible heat relationship for plain water near 60 F, roughly 500 times gpm times the temperature difference in degrees F, zones one and two at 3 gpm and 18 F carry about 27,000 Btu/h. Zone three at 1.3 gpm and 31 F carries about 20,150 Btu/h, which is about 75 percent of what its neighbours deliver, despite the larger drop. On a glycol mix that 500 constant is lower and you use the fluid's own specific heat and density.

Second attempt, zone three alone. With the other two isolated, the full 6 gpm went through one 3/4 in path, about 4 ft/s, comfortably past entrainment. The discharge ran clear in under a minute, then produced a short burst of air at about the ninety-second mark, which is exactly the dead-leg air that the first attempt never had the speed to move. Held clear for the full minute after that.

Re-run of Gate 6. Zone three came back at a 19 F drop against the other two zones' 18 F, both read at the same supply temperature and firing condition as before, so the comparison is like for like. The far baseboard was hot to the hand along its full length.

Confirming it stayed purged

Entrained air is what you just removed. Dissolved air is different: water holds gas in solution, and it holds less of it as temperature rises and as pressure falls. That means dissolved gas comes out of solution at the hottest, lowest-pressure point in the loop, which is usually at the heat source outlet, and it does so over days rather than minutes.

So the honest confirmation is not the same day. Leave the automatic air vents in service, and on the next visit check whether the vent at that point has been passing air. If it has, the system is still shedding dissolved gas, which is normal for a newly filled loop and abnormal a month later. Air still arriving after a month means air is still getting in, and the place to look is anywhere the loop runs below atmospheric pressure: a pump inlet on the wrong side of the expansion tank, a fill pressure set too low for the height of the building, or a fitting that leaks air in without leaking water out.

Write the purge flow, the zone-by-zone temperature differences and the date on the equipment record. The next tech comparing a 31 F drop against a documented 19 F baseline diagnoses this in five minutes instead of re-purging the whole system.

References

• 29 CFR 1910.333(b)(2) for electrical work at an appliance panel or disconnect, with NFPA 70E-2021, 120.5 for the live-dead-live proving sequence
• 29 CFR 1910.134, respiratory protection, including fit testing under 1910.134(f)(2), where an SDS calls for it on a glycol or inhibited fluid
• The fuel gas code adopted in your jurisdiction for gas-line purging, which this procedure does not cover
• Manufacturer documentation for fill pressure, relief setting and air separator placement on the specific appliance
• See related: Why Velocity Matters as Much as Volume; How a Branch Steals From the Branch Next to It; Why Balancing Is Not Optional