How to Survey a Trap Population and Rank What to Fix

Why this matters

Most trap surveys end with a list of failures and no order to work them in, so the list gets worked by trap size, by access, or by whoever asks loudest. Trap size is the worst of those three, because the loss a failed-open trap actually inflicts is its flow rate multiplied by the hours it runs, and hours vary across a plant by a factor of six or more, and absolute pressure by four, while orifice sizes rarely vary by half that. The ranking measure needs three inputs. Two of them you read at the trap; the third, the one that swings widest, you have to go and ask for.

The hazards this walk creates

A survey means walking a plant touching hot piping, and the hazards belong to the walk rather than to any one trap.

• Do not brush, cut or lift pipe insulation to reach bare metal for a reading. Thermal system insulation on steam and condensate lines is presumed asbestos-containing material under 29 CFR 1926.1101 for construction work, with general industry duties at 29 CFR 1910.1001, until it has been sampled. Take readings at fittings and trap bodies that are already bare, or have removal done by a trained crew under the applicable standard. This is an inhalation route, so gloves and glasses are not the control.
• Do not enter a below-grade condensate pit, vault or tunnel to reach a trap until that space has been evaluated as a permit-required confined space under 29 CFR 1910.146. Steam and hot condensate in a below-grade space are part of why it is one.
• Where a blowing trap or an open blowdown puts you in continuous high noise, that exposure falls under your hearing conservation program per 29 CFR 1910.95, and it will happen repeatedly across a survey day.
• Never trace a leak by hand. A jet at main pressure is invisible for the first stretch out of the seat and cuts before you feel heat. Work from the side of any discharge and keep a body length off it.
• Where you brush an already-bare fitting to seat a contact probe, use a long-handled brush, heat-resistant gloves rated for the surface temperature, and eye and face protection under your PPE hazard assessment per 29 CFR 1910.132, brushing across the pipe rather than toward yourself.

The measure, stated before anything is sorted

For a trap that has failed open and is discharging into a return well below its inlet pressure, flow through the seat is choked, and the standard choked-flow approximation for saturated steam makes mass rate proportional to the open flow area and to the absolute upstream pressure. That proportionality holds only while downstream absolute pressure is below roughly 58 percent of upstream absolute, the critical pressure ratio for steam, and it is an upper bound rather than a measurement, because the actual open area at a worn seat is unknown and is often partly obstructed.

Total heat per pound of saturated steam changes by only a few percent across the pressure range a field service plant runs, so ranking on mass rate ranks on energy rate to within that few percent. That lets the whole measure stay dimensionless:

Loss index = (open area, relative to a chosen reference trap) x (absolute inlet pressure, relative to the reference) x (annual operating hours, relative to the reference)

Pick one failed trap as the reference and set its index to 1.00. Everything else is a ratio against it, which keeps the arithmetic honest and keeps currency out of it entirely.

Step 1: Tag the population before you test any of it

Every trap gets a permanent, legible tag with a unique number, fixed to a bracket or to the pipe support rather than to anything that will be replaced with the trap. Attach it without leaning on hot pipe.

Without tags there is no year-over-year comparison, and year-over-year comparison against a trap's own history is far stronger evidence than any single reading, because it removes model, pressure and installation differences at once.

Step 2: Capture the three ranking inputs, two of which are not readings

This is the step that makes or breaks the survey, and standard survey sheets routinely omit two of the three fields.

• Open area proxy. Read the trap body and orifice designation off the nameplate and convert it to a relative area using the manufacturer's data for that body. Where the nameplate is unreadable, record that fact rather than guessing, because a guessed orifice propagates straight into the ranking.
• Absolute inlet pressure. Read a gauge at or near the trap and add atmospheric pressure. Gauge pressure alone will invert your ranking at low pressures, because 15 psig is a quarter of 100 psig on a gauge and closer to a quarter of it in absolute terms only by coincidence.
• Annual operating hours of the equipment served. This comes from operations or from a run-time meter, not from the trap. It is the input with the widest spread across a plant and the one most often left off the sheet.

Step 3: Test and record state, not just pass or fail

Use the three-signal method covered in the article on testing a trap without taking it apart, and record the state as one of failed open, failed closed, good, or out of service with no load. Record the closed interval in seconds for any intermittent trap, and record the two pressures and the ambient condition next to it.

Skipping the out-of-service category is the common shortcut and it corrupts the population count, because a cold silent trap on idle equipment gets logged as failed closed and generates a work order for nothing.

Step 4: Compute the index, then sort against it and read the sort backwards

Compute the index for every trap in the failed-open group. Sort descending. Then read the sorted list from the bottom up before you commit it, because the bottom is where an ordering goes stale and where a mis-keyed operating-hours figure hides without looking wrong.

Two entries whose indices sit within the per-trap uncertainty are not ordered by this measure and should not be presented as ordered. Break those ties on access cost, on whether the equipment is already scheduled down, or on whether one of them is on the header that feeds the other.

Step 5: Rank the failed-closed group separately, on damage exposure

Failed-closed traps do not belong in this ranking at all, because their loss index is zero. They rank on what the flooded equipment is exposed to: a coil on sub-freezing entering air first, a drip leg on a steam main second because backed-up condensate in a main is the source of a hammer slug, and everything else after. That ordering and its reasoning belong to the article on why failed-closed traps are found first.

Worked example: sixty traps, three candidates, and an inverted answer

A survey of 60 traps returns 9 failed open, 4 failed closed, 2 out of service with no load, and 45 good. Those four counts sum to 60, which is worth checking on your own sheet before anything else, because a population that does not add up means a trap was tested twice or missed.

Take three of the nine failed-open traps. The reference is trap A.

Trap A. Drip trap on a 100 psig main, so 114.7 psia absolute. Reference orifice, area factor 1.0. The main runs year round, 8,760 hours. Index 1.00 by definition.

Trap B. Identical body and orifice to A, also on 100 psig, but serving a seasonal process that runs 1,400 hours a year. Index is 1.0 times 1.0 times 1,400 divided by 8,760, which is 0.160.

Trap C. A physically larger trap, 2.5 times A's orifice area, on a 15 psig coil, so 29.7 psia absolute, running 2,000 hours a year. Index is 2.5 times (29.7 divided by 114.7) times (2,000 divided by 8,760). That is 2.5 times 0.259 times 0.228, or 0.148.

Check the choked-flow condition on each before trusting any of it. The return header sits at 2 psig, or 16.7 psia. For A and B the ratio is 16.7 over 114.7, which is 0.146, comfortably below the 0.577 critical ratio, so the proportionality holds. For C the ratio is 16.7 over 29.7, which is 0.562. That is below 0.577 but barely, so C is only just choked, and if that return climbs at all C's flow stops being proportional to inlet pressure and the index overstates it. C therefore carries a wider bound than the other two on top of ranking last.

The ranking, read bottom up. C at 0.148 sits below B at 0.160, and the two are within about 8 percent of each other. Per-trap uncertainty here is far larger than that, so B and C are a tie and get presented as one, with access breaking the order. A at 1.00 sits above B by a factor of about 6.3, which is well outside any plausible uncertainty, so that part of the ordering is a real finding.

What the naive ranking would have done. Sorted by trap size, C goes to the top of the work list and A goes below it. The measure inverts that completely: the small trap on the year-round high-pressure header outranks the large trap on the low-pressure seasonal coil by more than six to one, and every input driving that inversion is one of the two fields most survey sheets leave out.

Now the population total, and why it is a bound. Each per-trap figure carries a wide relative spread, because the open area at a worn seat is not measured. Take that spread as roughly plus or minus 50 percent per trap and treat the traps as independent. Independent spreads combine in quadrature, so summing 9 of them gives the total a random relative component of about 50 divided by the square root of 9, which is about 17 percent. The sum is a much better number than any single trap in it, and that is the honest reason to report the population figure and not the individual ones.

What does not shrink is the systematic part. Assuming every failed-open trap passes at full orifice capacity is one assumption applied 9 times, so it biases the total in one direction and stays at full strength no matter how many traps you add. That makes the population figure an upper bound, and it gets written down as a bound rather than as a value with a plus-or-minus, because those are two different claims.

How to verify the survey before you hand it over

• Re-add the state counts against the register count. If they do not sum, a trap was double-tested or skipped, and both corrupt the ranking.
• Re-check every operating-hours figure that came from memory rather than from a meter or from operations. It is the highest-leverage input in the measure and the one nobody verifies.
• Re-check the choked-flow condition on every low-pressure trap in the list, not just the one in your example. Low inlet pressure against a loaded return is where the proportionality quietly stops holding.
• Re-read the sorted list from the bottom. Confirm the last entry really does belong last and that its three inputs are the ones on its sheet.
• Confirm no failed-closed trap ended up in the loss ranking, and that the freeze-exposed ones are at the top of their own list.
• Re-test ten traps a month after the survey. Several state changes in that sample means the interval, not the ranking, is the thing to fix first.

References

• 29 CFR 1926.1101 and 29 CFR 1910.1001, asbestos in construction and in general industry, covering presumed asbestos-containing thermal system insulation on steam and condensate piping
• 29 CFR 1910.146, permit-required confined spaces, for below-grade condensate pits, vaults and tunnels
• 29 CFR 1910.95, occupational noise exposure, and 29 CFR 1910.132, PPE hazard assessment, for survey-day exposures
• Trap manufacturer data for orifice designation and relative flow area, and steam tables for saturation conditions at the measured pressures
• See related: How to Test a Steam Trap Without Taking It Apart; Why a Failed-Open Trap Costs More Than a Failed-Closed One