How to Find Where Water Hammer Is Being Generated

Why this matters

The bang is almost never where the event is. A pressure transient in a steam line arrives at both ends of the building at effectively the same instant, and what you hear is the pipe striking a hanger, a sleeve or a joist somewhere it happens to be loose. Techs walk toward the loudest point, find a fitting there, and change it. The generation point can be in another room, on another branch, at the far end of a main. The method below finds the event by what triggers it, because the trigger is tied to the mechanism and the mechanism is tied to a place.

Stop it before you study it

If the line is hammering repeatedly right now, shut that branch down before diagnosing. Move everyone out of the plane of the flanges, threaded joints and the outboard side of the elbows, since that is the direction a failed joint discharges, and close the supply slowly from beside the valve rather than in front of the bonnet. A fast closure adds a transient to a line already producing them.

Nothing in this procedure involves finding a steam leak by hand. Steam at usable pressure is invisible for the first stretch out of a leak and it cuts skin before it feels hot. Look for plumes against a dark background from a distance.

Step 1: get the trigger, not the location

Before touching an instrument, find out what operation precedes each bang. Ask the operator, then confirm it yourself by standing where you can see a control while somebody else listens.

The candidates are short and each one points somewhere different: a manual or motorised isolation valve opening, a modulating control valve stroking on a call, a condensate pump starting, a burner cycling, the morning start-up sequence, night setback recovery, or nothing identifiable at all. That last answer is itself a finding, because an event with no operational trigger is usually tied to a temperature crossing rather than to a command, and it moves suspicion toward a line that cools between calls.

This step comes first because it is the only one that eliminates whole subsystems for the cost of a conversation. Skip it and you instrument the wrong branch, which produces readings that are all normal.

Step 2: name the mechanism from the trigger

The sibling article on what water hammer actually is in a steam line sets out the two mechanisms; do not re-derive them here, just place your case.

A bang on admission of steam to a line that was cold or drained, occurring once or a few times and then quietening as the line heats, is condensate being picked up and driven as a slug. Look for a place where condensate could have been lying: a sag, a reverse grade, a drip point that is not draining.

A bang repeated during steady operation, worse at low load, is steam collapsing on subcooled water. Look for a place condensate is being held: a flooded shell, a coil below its drain point, a lift the system cannot make at that moment.

The two need different searches, and running the wrong one wastes the whole visit.

Step 3: find the cold metal

Collapse needs water below saturation temperature. Slug propulsion needs water at all. Both leave the same signature: metal colder than the saturation temperature for the gauge pressure in that line.

Read pressure at the nearest gauge, convert to saturation temperature from a steam table, then walk the suspect branch reading pipe surface temperature. On bare, shiny steel an infrared thermometer reads low because the surface emissivity is poor, sometimes by tens of degrees, so use a contact probe or put a patch of high-emissivity tape on the pipe and read the tape. Take non-contact readings from arm's length and keep bare skin off the pipe at any pressure.

Where insulation blocks access, read at existing gaps, at valve bonnets and at flange faces rather than opening the lagging. Thermal system insulation on steam and condensate piping is presumed asbestos-containing under 29 CFR 1926.1101 for construction work, with general industry duties at 29 CFR 1910.1001, until it has been sampled. The presumption attaches to thermal system insulation in buildings constructed no later than 1980, and a later repair does not clear it, so do not cut, scrape or pull it to reach metal. This is an inhalation route, which means the control is sampling and qualified removal, or a respirator under a 29 CFR 1910.134 program, not gloves.

A stretch reading well under saturation on a live line is either flooded, air-bound or not receiving steam. All three are findings.

Step 4: map the drainage on that stretch only

Now, and only now, walk the drainage. Doing this first on a whole system is a day of work that mostly documents traps that are fine.

For the suspect stretch, record every drip point, whether its leg is full size, whether the trap is discharging, and whether the leg has a dirt pocket below the take-off. Check discharge from outside the discharge path; a listening device on the trap body or a temperature comparison across it tells you what it is doing without opening anything. A trap failed closed backs condensate into the main, which is the precondition for both mechanisms.

Step 5: time the delay, and know what your stopwatch is worth

For triggers you can command, measure the delay from the trigger to the bang. This is not an acoustic measurement. Sound in a steel pipe travels on the order of 16,000 ft per second, so across any building the report arrives everywhere at once and loudness tells you where the pipe is loose, not where the event happened. The delay you are timing is a transport time in seconds, set by how far a steam front or a body of condensate has to move, and it separates a generation point near the valve from one a long way down the main.

Before that number does any work, know its basis and its character. A hand-timed delay carries a spread dominated by your own reaction, not a manufacturer's specification. Measure it: time the same known event ten times and take the spread of your own readings, which commonly lands in the range of a few tenths of a second. That spread is an independent random error, not a fixed offset, and the difference matters. If you always start late by the same amount, that is a fixed offset and it cancels when you subtract two readings taken the same way. The random part does not cancel; two independent readings differenced combine in quadrature, so a spread of 0.3 seconds on each reading gives about 0.3 times the square root of two, or roughly 0.4 seconds, on the difference.

The practical consequence: with a hand stopwatch, a one-second difference between two delays is not evidence of anything, and a three-second difference is. If you need better resolution than that, the timing has to come off a controller trend or a datalogger, not off a thumb.

Step 6: change one thing and re-trigger

Confirmation is one change at a time, and the change should be reversible and diagnostic rather than a repair: open a manual drain at the suspect drip point to a discharge routed where nobody stands, opening it slowly from the side and wearing hearing protection where the discharge is above the action level in 29 CFR 1910.95; hold a modulating valve at a fixed position instead of letting it throttle; delay a pump start. If the hammer stops when condensate is given an escape route at one point, that point is the generation site.

If nothing you change makes any difference, go back to step 1. An unchanged symptom under a genuine change of condition means the trigger was misidentified.

Worked example: the converter that only banged in mild weather

A steam-to-water converter fed from a 15 psig main, so about 250 F, banged in the mechanical room through October and November and was quiet in January. Two techs had already tightened hangers and one had quoted a larger trap.

Trigger. Watching the control, every bang followed the modulating steam valve stroking open on a call. Nothing happened when the valve was already open.

Mechanism. Repeated during steady operation, worse at low load. That is collapse, so something was holding subcooled water where steam could reach it.

Cold metal. The shell read well below the saturation temperature for the main's pressure, and the bottom of the shell read colder than the top. The lower part of the shell was full of water.

Drainage. The trap was discharging in bursts and looked healthy. The interesting reading was two gauges rather than one: the shell gauge and a gauge on the return main downstream of the trap. On a cold day, with the valve near wide open, the shell read 12 psig against a return back pressure of 5 psig, a positive differential of 7 psi across the trap. On a mild day, with the valve throttling to hold the same water temperature at a fraction of the design load, the shell read 2 psig against the same 5 psig return. The differential was negative by 3 psi.

That is the whole answer, and it is measured rather than inferred. A trap is a valve; with a negative differential across it, no trap of any size or type discharges, because there is nothing to push the condensate out. The shell floods, the pooled water cools below saturation, and the next time the valve opens, steam meets it and collapses. The larger trap that was quoted would have changed nothing.

Timing. Delay from valve stroke to bang averaged about 2.5 seconds across six events, with individual readings spread a few tenths either side. Nobody drew a distance conclusion from that, because the differences between events were inside the stopwatch spread from step 5. It was used only to confirm the events were tied to the stroke rather than coincidental.

Confirmation. A manual drain on the shell, piped to a floor point clear of walkways and opened slowly from the side, was cracked with the area barriered and a second person attending, once. Water first, then a flash plume as the shell cleared, at which point it was shut, because past that point it is a live steam discharge into an occupied room rather than a diagnostic. The hammer stopped for that call and came back on the next one with the drain closed.

Direction check. The mechanism explains both ends of the range, which is how you know it is right. At design load, the valve is near wide open, shell pressure sits above the return back pressure, the trap discharges, and there is no pool to collapse on. At light load, the valve throttles the shell below the return, the trap stalls, and the pool forms. The complaint was seasonal because the load was seasonal, and January was quiet for exactly the reason October was not.

Repair. A gravity drain from the shell to a vented receiver below the converter, which is the part that actually removes the 5 psig of back pressure the shell could not beat. A vacuum breaker goes on with it, but know which part did what: the shell measured 2 psig, above atmospheric, so a breaker alone would have changed nothing. It covers the separate overnight case where the shell goes sub-atmospheric on cooldown. Where the geometry does not allow a fall to a receiver, the alternative is a pressure-powered pump-trap arrangement, which moves condensate on motive steam rather than on differential pressure. The sibling article on what a condensate receiver and pump set is doing covers the receiver side.

Verifying the fix held

Silence on the day of the repair proves very little, because most of these faults are load-dependent and you are unlikely to be standing there at the load that produced them.

Two checks are worth booking. First, repeat the two-gauge reading at the lightest load the equipment sees, which for a heating converter means a mild shoulder-season day, and confirm the differential across the trap stays positive at that load. Second, put a temperature reading on the bottom of the shell into whatever log the site already keeps; a shell bottom that drifts below saturation over a few weeks says the drain is silting or the vacuum breaker is stuck, and that shows up before the noise does.

And leave the hangers alone unless they are genuinely damaged. Inspect them and replace what is bent, but do not add restraint to quieten a transient you have not eliminated. The noise is the only free warning the system gives.

References

• ASME Steam Tables, or an equivalent saturated-steam table, for the saturation temperature at the gauge pressures you read
• 29 CFR 1926.1101 (construction) and 29 CFR 1910.1001 (general industry) for asbestos in thermal system insulation, with respiratory protection under a 29 CFR 1910.134 program where one is required
• 29 CFR 1910.95, occupational noise exposure, for hearing protection at open steam or condensate discharges
• Trap and converter manufacturer literature for stall conditions, vacuum breaker selection and drainage arrangement
• See related: articles in this library on what water hammer actually is in a steam line, and on what a condensate receiver and pump set is doing