How a Thermostatic Trap Fails and What You See

Why this matters

A thermostatic trap has no opinion about steam. It opens when the fluid touching its element is some margin below the boiling point, and that boiling point moves every time the system pressure moves. So a thermostatic trap can only be judged against the saturation temperature at the pressure it is seeing at that moment, and an absolute temperature reading tells you nothing on its own. Two identical readings taken on two identical traps in the same building can mean "this one is blowing steam" and "this one is flooding the equipment," and the only thing separating them is a gauge. Techs who carry a remembered number in their head for what a good trap reads get this backwards on a regular basis, and the two errors have opposite consequences.

Before you read or pull one

Never judge a trap by hand. Use a contact temperature probe on a cleaned spot with your hand and the lead clear of the pipe, and take the pressure from a gauge you read from beside rather than in front of any plug, cap or gauge fitting, wearing face protection under 29 CFR 1910.133 with sleeves down.

If there is no gauge close to the trap and you have to fit one, that is not a live-line job: isolate the inlet and the outlet, lock and tag both under 29 CFR 1910.147, vent to atmosphere through a drain you operate from the side, confirm zero gauge pressure on a readable gauge, and let the body fall below 120 F, verified with a non-contact reading taken at arm's length rather than assumed from the clock, before you break the fitting. The same sequence applies to pulling the trap, and condensate held between two closed valves flashes out of the first joint you open.

If you find a thermostatic trap wrapped in insulation, do not pull that lagging off, whatever the building's age. Thermal system insulation on steam piping is presumed asbestos-containing under 29 CFR 1926.1101 for construction work and 29 CFR 1910.1001 in general industry until it has been sampled; the presumption attaches to buildings constructed no later than 1980, and a later repair does not clear it, only sampling does. Thermal system insulation there is presumed asbestos-containing under 29 CFR 1926.1101 for construction work and under 29 CFR 1910.1001 in general industry, the route is inhalation, and the control is a trained abatement contractor with respiratory protection under a written program meeting 29 CFR 1910.134. Record it as a finding and have it removed properly.

What the element is actually doing

The element is a temperature-sensitive part that closes a valve when it gets hot and opens it when it cools. Because condensate sitting at saturation is exactly as hot as the steam above it, the element cannot open at saturation without passing steam. So it is built to stay shut until the fluid has cooled a margin below saturation, and it opens on that margin. Everything about how this family behaves and how it fails is a statement about that margin.

Balanced-pressure bellows. A sealed bellows holds a volatile liquid whose own vapour pressure rises along roughly the same curve saturated steam does. Because the inside and the outside of the bellows track each other, the element opens at roughly the same margin below saturation whether the system is at 2 psig or 60 psig. That self-adjustment is the whole reason this type exists.

Bimetallic. Stacked strips of two metals with different expansion rates deflect as they heat. That deflection follows the metals' behaviour, not the steam saturation curve, so the manufacturer stacks and preloads discs to approximate the curve, and the approximation drifts as pressure changes. A bimetallic element generally holds condensate back by a larger margin than a bellows, and the size of that margin at a given pressure lives in the manufacturer's published offset curve for that specific trap, which is the only place it lives. Do not carry a number for it.

The two default failure directions

A bellows element defaults to failing open. Rupture it or let it lose its charge and the closing force is gone, so the valve sits open and passes steam continuously. The usual killers are water hammer and an overpressure event such as a pressure reducing valve failing open, both of which crush or overstretch a thin sealed part. That said, two other failures of the same trap run the other way, and they are worth knowing because they are less common rather than impossible: a distorted element or debris can jam the valve shut, and a heavily scaled seat can restrict it.

A bimetallic element defaults toward holding more back. Scale and corrosion build on the element stack and the seat, the travel available shrinks, the margin below saturation grows, and the trap subcools further and further until the equipment upstream is running with a column of cool water in it. Two counterpart failures pull against that default: a seat wire-drawn by dirt leaks through continuously, and an element that has been overheated takes a permanent set that shifts the margin in either direction.

So the two types are not mirror images the way the two mechanical families are; each one has a default and a genuine secondary. What makes this family distinctive is not the direction, it is that the verdict depends on a pressure you have to measure at the same moment.

The gate

Take the inlet surface temperature and the system pressure at the same moment, convert the pressure to a saturation temperature from a steam table, and judge the difference between them against the element's published margin. The absolute temperature is not evidence.

Case one: 226 F at 5 psig

A balanced-pressure trap on a two-pipe radiator. Inlet surface reads 226 F. The gauge on that branch reads 5 psig, and saturation at 5 psig is about 227 F. The offset is about 1 F below saturation.

A balanced-pressure element is designed to hold shut until the condensate is a real margin below saturation, so an element sitting essentially at saturation is not holding anything back. That is a fail-open reading, and the mechanism that fits it is a bellows that has lost its charge or split. In a radiator system the confirming detail is usually easy: the return runs hot where it should run warm, and the radiator itself heats all the way across instead of leaving its far end cooler.

But do not stop at the temperature. One degree is a small number, and the next section explains why on this branch it is too small to stand alone. Confirm acoustically: hold an ultrasonic probe against the trap body through the instrument's own handle and listen for a continuous unvarying rush rather than discrete discharges with quiet between them. Temperature and sound agreeing is the call; temperature alone at this offset is not.

Case two: 226 F at 20 psig

An identical reading on a bimetallic trap on a small heat exchanger, taken with the same probe. Same 226 F. The gauge here reads 20 psig, and saturation at 20 psig is about 259 F. The offset is 259 minus 226, which is 33 F below saturation.

Thirty-three degrees is a large subcool even for a bimetallic element, and it means condensate is being held well past the point where this trap was meant to release it. The picture upstream is a heat exchanger part-full of cooling water, which shows up as output that sags at part load and recovers when the load goes to full and the extra pressure pushes the column out. The mechanisms that fit are scale on the element and seat restricting travel, or an element that has taken a set.

Same reading, opposite verdicts. Nothing about the trap decided this. The pressure did.

Before you condemn the trap, ask whether anything moved the pressure. A bimetallic element's margin diverges from the saturation curve, so a trap that behaved correctly on a 5 psig main and is now on a main somebody raised to 20 psig has not failed at all - it is doing exactly what its curve says it does at a pressure nobody re-checked it against. Pull the manufacturer's offset curve for that model before you replace it, because a replacement of the same model will do the same thing.

Reading the offset without fooling yourself

The offset is a difference between a number that came from a probe and a number that came from a gauge and a steam table. That is a different error situation from comparing two probe readings, and it matters here because the whole verdict rests on the difference.

The probe's systematic offset does not cancel. When you take an inlet and an outlet temperature with one probe, that instrument's fixed bias is present in both and largely cancels in the difference. Here only one of the two numbers comes from the probe, so its bias survives into the answer at full size.

The gauge error arrives multiplied by the slope of the saturation curve, and that slope is not constant. The curve is steep at the bottom and flat at the top: around 5 psig it runs near 3 F per psi, and around 20 psig nearer 1.8 F per psi. A dial gauge specified as a percentage of full scale is worth a fixed pressure anywhere on the face, so a 1 percent of full scale gauge on a 0 to 60 psi dial is worth about 0.6 psi at any reading. That 0.6 psi becomes about 1.8 F of saturation temperature at 5 psig and about 1.1 F at 20 psig.

Combine them by character, not by adding them. The probe's random spread and the gauge's are independent, so they combine in quadrature. If the probe's spread is on the order of 2 F, the low-pressure case combines 2 F and 1.8 F to about 2.7 F, and the high-pressure case combines 2 F and 1.1 F to about 2.3 F. Any fixed systematic bias you know about is added separately as a bias, not folded into that spread.

Now re-read the two cases against those figures. The 1 F offset in case one sits inside a 2.7 F combined spread, so the temperature reading alone cannot call it and the acoustic confirmation is doing the real work. The 33 F offset in case two sits more than ten times outside a 2.3 F spread, so that one is decisive on temperature alone. Same method, two levels of confidence, and the honest report says which is which.

The uncomfortable part worth stating plainly: this test is least trustworthy at low pressure, which is exactly where the design margins are smallest and where most thermostatic traps live. On low-pressure work, get a gauge close to the trap rather than reading a header gauge two hundred feet away, and never call a thermostatic trap on a remembered temperature.

The failure that is not a failure

A temperature-test trap uses its own surroundings as part of the mechanism, because the element has to lose heat to fall below saturation. Three installation conditions therefore produce a trap that reads as failed and is not:

Somebody insulated it. A lagged thermostatic trap cannot shed heat, so the element stays hot, the valve stays shut, and condensate backs up. The trap is fine.

It sits in a hot pit or a closed enclosure. Same mechanism, slower onset, and it appears seasonally, which is why it gets diagnosed as an intermittent fault.

It is mounted where the inlet leg is too short. A thermostatic trap needs a length of uninsulated inlet pipe ahead of it for the condensate to cool in. Fit it hard against the equipment with no cooling leg and it holds condensate inside the equipment instead of in the leg, which is a design fault that will follow every replacement trap you fit.

Check all three before you condemn an element, because each of them will reproduce itself on the new part.

References

• Steam trap manufacturer published offset-versus-pressure curves and margin below saturation for the specific trap, which is the only authoritative source for the margin an element should hold
• ASME Steam Tables (IAPWS industrial formulation) for the saturation pressure and temperature pairs and the slope of the curve used above
• 29 CFR 1910.147 for isolation of pressure and thermal stored energy; 29 CFR 1910.133 for eye and face protection
• 29 CFR 1926.1101 (construction) and 29 CFR 1910.1001 (general industry) for presumed asbestos-containing thermal system insulation, with respiratory protection under 29 CFR 1910.134
• See related: What a Steam Trap Is Actually Deciding; How a Mechanical Trap Fails and What You See; The Trap Families and What Each One Is Good At