What Surge Does in a Pumped Liquid Line and Why It Is Not Steam Hammer

Why this matters

The pressure spike that splits a fitting on a pumped water main has nothing to do with how much head the pump makes. It is decided by how fast the liquid was moving and how quickly you stopped it, and on a long line it can be several times the working pressure of the system. That is why a 60 psi system breaks 200 psi components, and why the fix is almost never a stronger pipe. The controllable variable is time, measured against a clock the pipe itself sets, and most people who think they installed a slow-closing valve did not, because a valve's stroke time and its effective closure time are different numbers.

The library already owns water hammer in a steam line. These are two different mechanisms and the fix for one is not the fix for the other, so this card starts by separating them.

Two events, one nickname

In a steam line, the bang is condensation-induced collapse. Steam meets accumulated condensate, a pocket of vapour condenses almost instantly, the void it leaves closes at high speed, and slugs of water are accelerated into fittings. The cure is drainage: traps, slopes, drip legs, warm-up procedure. The sibling card on water hammer in a steam line owns that mechanism in full.

In a pumped liquid line, the bang is momentum arrest. A column of liquid with real mass and real velocity is stopped, and its momentum has nowhere to go except into pressure. Nothing condenses, no vapour pocket has to collapse for the bang to happen, and no amount of drainage helps. The cure is time, or a device that gives the momentum somewhere to go.

A reader who carries the steam explanation into a pumped water main will look for condensate that is not there. A reader who carries this one into a steam main will slow a valve that was never the cause.

Before you go looking for it

Do not reproduce a surge to confirm it. Closing a valve quickly on a live main to hear the bang is running the failure mechanism on purpose, and the joint that is about to fail is the one you are standing beside. Diagnose it from the evidence it leaves and from a recorded pressure trace instead.

Do not stand under or beside an unsupported suspended run while valves are being cycled. A surge event moves pipe. Broken hangers, scarred insulation and pipe that has walked out of a clamp are the physical record of forces large enough to injure someone standing in the way.

Where a pressure tapping has to be added for a transducer, isolate and bleed the section to zero on a gauge first, and lock and tag the pump's energy isolating device under 29 CFR 1910.147 so a level or pressure control cannot restart it while the line is open. On a hot line, let it cool below 120 F or drain through a hose to a floor drain while standing to the side in a face shield and heat-rated gloves.

Where the pressure comes from

Stop a moving column of liquid and its kinetic energy has to go somewhere. Liquid is very nearly incompressible and the pipe is only slightly elastic, so what it goes into is a pressure wave that travels back up the line at the speed of sound in that pipe and that fluid.

Two things follow immediately, and both are counter-intuitive.

The pump's head is not in the calculation. A booster making 40 psi and a booster making 120 psi produce identical surge for the same velocity and the same closure. The surge rides on top of whatever the line pressure happened to be, which is why the static pressure appears as a separate term in the arithmetic below rather than as a factor in it.

Pipe length does not change the size of the spike, it changes how long you have. A longer line does not surge harder. It gives you a longer round trip for the wave, which makes slow closure easier to achieve, which is why long transmission mains are often better behaved than a short, fast, high-velocity branch.

The bound, and the conditions it was derived under

The classic instantaneous-closure result gives the maximum possible rise as the fluid density times the wave speed times the velocity change. In field units for cold water in restrained steel pipe with no entrained air, that works out to roughly 50 to 60 psi for every 1 ft/s of velocity arrested, where the wave speed runs about 3,900 to 4,400 ft/s across ordinary schedules. Wall thickness sets it: a thin-wall, large-diameter steel line falls to nearer 3,200 ft/s and 43 psi per ft/s, and a thick-wall small-diameter one sits at the top of the range. In PVC, where the wave speed is nearer 1,400 ft/s, the same arithmetic gives roughly 19 psi per ft/s. Compute it for the pipe in front of you rather than carrying one figure across materials and schedules, because this number is a ceiling and a low one is the unsafe direction to be wrong in.

Four conditions are attached to that figure and each one matters:

• Instantaneous closure, meaning faster than the round-trip time below. Slower closure gives less.
• No entrained air. A small percentage of undissolved gas can cut the wave speed by half or more, which cuts the rise proportionally. That makes the figure conservative, and it is not a design strategy, because entrained air brings its own problems and its quantity is not something you control.
• Restrained pipe of that material. Wave speed depends on the pipe's stiffness and wall thickness as well as the fluid, so a thin-wall or a flexible material is lower and a thick-wall steel is higher.
• No friction damping and no branch, relief or air vessel taking energy out of the wave.

Because of those four, the result is an upper bound on the first peak, and it is written that way: peak pressure <= static pressure + the rise per ft/s for that pipe, times the ft/s arrested. One inequality sign, no plus-or-minus, because a bound is not an interval and quoting it as one invites somebody to design to the middle of it.

The clock the pipe sets

The wave leaves the valve, travels to the nearest large reservoir, pressurised main or open tank, reflects, and comes back. That round trip is the pipe's own time constant: twice the length divided by the wave speed.

If the valve finishes closing before the reflected wave gets back, the valve never gets the relief the reflection provides, and the full bound applies. If closure takes longer than the round trip, the reflection arrives while the valve is still closing and cancels part of the rise, roughly in proportion to how much longer the closure took.

The trap is what "closure time" means. A butterfly or a ball valve does almost nothing to the flow through the first two thirds of its travel; nearly all of the flow reduction happens in the last portion. So the effective closure time - the time over which the velocity actually falls - can be a third of the stroke time or less. A specification that says three seconds and a valve that does its work in the last eight tenths of a second are not the same valve, and only one of them is protected.

Worked example: a transfer main and a "slow" actuator

A 6 in. steel transfer main, 900 ft from the pump station to a tank, water at ambient temperature, flowing at 5.0 ft/s, with the line sitting at 60 psi static. An automated butterfly valve at the tank end is specified with a 1.5 second stroke. A Class 125 cast iron valve body in the line is the lowest-rated element in the path at about 200 psi, established the same way a relief valve's set point is established, by inventory of the path rather than by the pump's rating.

The bound, with each term on its own line:

• Wave speed, cold water in restrained 6 in schedule 40 steel: 4,400 ft/s
• Rise per unit velocity at that wave speed: 59 psi per ft/s
• Velocity arrested: 5.0 ft/s
• Maximum rise: 59 x 5.0 = 295 psi
• Static pressure the rise sits on top of: +60 psi
• Peak <= 355 psi

The pipe's clock:

• Round trip: 2 x 900 ft / 4,400 ft/s = 0.41 seconds

The credit the stroke time appears to earn:

• Stated stroke 1.5 s is longer than 0.41 s, so the rise reduces roughly in proportion: 295 x (0.41 / 1.5) = 81 psi
• Predicted peak on that basis: 60 + 81 = 141 psi, comfortably under the 200 psi body

The correction that reverses the answer. The valve was stroked and timed with the disc angle recorded. It passed through the first 70 percent of travel in 1.15 seconds while flow barely changed, and completed the last 30 percent - where essentially all of the flow reduction happens - in 0.35 seconds. The effective closure time is 0.35 s, which is shorter than the 0.41 s round trip, so no reflection credit applies at all and the full bound stands: peak <= 355 psi, against a 200 psi body.

That is the whole finding. The specified stroke time was not wrong as a number and was irrelevant as a protection. The recorded trace during a controlled, instrumented closure peaked at 244 psi, which sits under the bound as it must and well over the body's rating.

What was done. Not a stronger valve body, and not a gas-charged arrestor, which is a fixture-scale device with nowhere near the volume to absorb a 900 ft column. The actuator was re-configured with a two-speed profile so the last 30 percent of travel takes 2.5 seconds, putting effective closure at more than four times the round trip and bringing the predicted rise under about 50 psi. Line velocity was also a candidate: dropping from 5.0 to 3.5 ft/s would have cut the RISE by 30 percent by itself, taking the bound from 355 to 267 psi, and on a system being designed rather than repaired that is the cheaper lever.

The other half: what happens when the pump trips

Valve closure is the case everyone knows. The pump trip is the one that breaks things, and it runs in the opposite direction first.

When a running pump loses power, the column downstream keeps moving and the pressure immediately behind it falls. On a line with any elevation, it can fall to the vapour pressure of the liquid, at which point the column separates and a vapour cavity opens. The column then decelerates, reverses, and rejoins. A cavity does open in that case, but it is opened by the pressure drop of the departing column and closed by that column's own momentum, not by a steam interface condensing on it. Same word, different machine. The pressure spike from a column rejoining can exceed the upsurge from any valve closure on the same line, and it arrives seconds after the trip, when the room has gone quiet and everybody has relaxed.

A check valve makes it worse or better depending on how it behaves. A swing check that stays open while flow reverses and then slams shut at high reverse velocity adds its own Joukowsky event on top of the rejoining column. A spring-assisted or non-slam check that closes as flow decelerates through zero avoids that contribution entirely, and on a station where trips are routine that is the single highest-value component change.

This is why a surge complaint that only appears on power failures is not diagnosed by cycling valves, and why "it only bangs when the plant loses power" is a specific finding rather than an incidental detail.

How to verify you got this right

A standard bourdon gauge cannot see this. The event lasts a fraction of a second and the mechanism cannot follow it; a needle-damped gauge is designed specifically not to. A gauge reading 65 psi during a surge is evidence about the gauge, not about the line. Use a fast pressure transducer with a logger sampling many times per second, mounted at the point of interest, and read the trace.

Without instruments, read the physical record instead, and it is a good record: hangers pulled or broken, pipe walked out of a clamp, insulation scarred at a support, repeated failures at the same fitting rather than at random ones, and a check valve with a hammered seat. Repeat failures clustered at one joint on a long straight run point at a wave, because a wave loads the whole run and finds the weakest point in it every time.

Then close the loop by timing the valve rather than trusting the specification. Stroke it, record the time in the last third of travel, and compare that number to twice the pipe length divided by the wave speed. If the first is smaller than the second, nothing has been slowed down.

References

• Hydraulic Institute and pipe manufacturer guidance on transient pressures and wave speed for the specific pipe material, wall thickness and restraint condition, in the edition your specification references
• Valve and actuator manufacturer documentation for the closure characteristic, which owns the relationship between travel and flow and therefore the effective closure time
• 29 CFR 1910.147 for isolation and pressure relief before adding a tapping or opening a line
• See related: What Water Hammer Actually Is in a Steam Line; Water Hammer Control; What a Relief Valve on a Positive Displacement Pump Is Protecting