Check Valves And Water Hammer How Non-Slam Designs Stop The Surge Before It Starts

1. The water hammer starts when the pump stops, not when the valve fails

Ask a plant what caused the pipe that ruptured at the pump discharge and the answer is usually a failed check valve. The check valve is the suspect because it is the component that broke, but the cause is almost always upstream of the component: somewhere in the piping design there is more stored kinetic energy than the system can absorb when the pump stops delivering it. The check valve's only job in that moment is to interrupt the reverse flow before it accelerates, and every check valve design is really an answer to one question: how fast can the disc get to the seat, and how much damage does the arrival do?

This article works through that question the way it is decided in practice: what the Joukowsky equation actually promises and what it leaves out, what happens in the first ten seconds after a power failure, why the conventional swing check valve is both the default choice and the default problem, how the common closing designs compare, why deceleration rather than flow rate is the parameter that selects the valve, and what has to be on the specification before a supplier can be held to a non-slam claim. The check valve range covers most of the families discussed here, but the selection logic matters more than the catalogue.

2. Joukowsky in one line, and what it leaves out

The headline number comes from the Joukowsky equation: the pressure rise equals the fluid density multiplied by the wave speed multiplied by the change in velocity. In metric terms, a change of 1 m/s in velocity produces roughly a 10 to 14 bar pressure rise in a typical steel pipeline carrying water, depending on pipe stiffness and entrained air. Stop 3 m/s of forward flow instantaneously and the surge pressure is 30 to 40 bar on top of the static head. That number alone is enough to explain most burst pump-discharge manifolds.

What the equation leaves out is the word instantaneous. Joukowsky assumes the velocity change happens faster than the pressure wave can travel the pipe and return. If the flow decelerates more slowly than that, over a time longer than twice the pipe length divided by the wave speed, the peak surge is lower than the Joukowsky value, and how much lower depends on the deceleration profile. This is the entire engineering basis of non-slam check valves: they do not change the physics, they change the timing. A disc that reaches the seat while the reverse velocity is still small converts a hard stop into a soft one, and the surge that would have been 30 bar becomes 5.

Two corollaries follow. First, in short lines, where the wave return time is short, almost any check valve looks adequate because the system never lets reverse flow develop much velocity; the problem is characteristically a long-line problem, typically lines above a few hundred metres. Second, in systems with large elevation differences or long suction lift, the column of water can separate at a high point and rejoin later, producing a second and often larger surge than the first, which no check valve alone can prevent.

3. The first ten seconds after power failure

The sequence is worth walking through because the specification language used later refers to it. When the pump trips, its head decays quickly and the flow at the pump begins to fall while the column further down the line keeps moving on its inertia. In the first phase the flow reverses while the pump still spins forward, decelerating against the reversed flow. In the second phase the pump stops and starts to run backwards, accelerating as the column drives it; this is the reverse rotation that destroys thrust bearings and throws impellers off shafts on pumps not designed for it. In the third phase the pump reaches runaway reverse speed, and in the fourth, if the line remains connected, the reverse flow settles toward a steady value determined by the static head and the system resistance.

A check valve is a device for cutting this sequence short, and the useful metric is the reverse velocity at the moment of disc contact. That single number, which some manufacturers publish and many do not, maps directly to the surge: the pressure rise on closure is the Joukowsky value for that reverse velocity, softened by whatever cushioning the disc and damper provide. Swing check valves commonly slam shut against reverse velocities of 1.5 to 2.5 m/s in long water lines, which is precisely the 20 to 30 bar band where pipework fails. Spring-loaded designs and nozzle-type valves reach the seat below 0.3 to 0.5 m/s, an order of magnitude less stored energy.

4. Why the swing check is the default and the default problem

The swing check valve survives in specifications for good reasons: it is simple, cheap, full-bore, low pressure drop, repairable, and available in every material and pressure class from every maker. For short lines, low head, and infrequent pump starts, it is often the correct answer. Its failure mode is geometric. The disc must swing through roughly 70 to 90 degrees between full open and closed, driven only by reverse flow, with the pivot offset providing the closing torque. Nothing about that geometry is fast, and nothing about it is damped. The disc travels at whatever speed the reverse flow dictates and arrives at the seat at full reverse velocity, which is why a swing check closing against a reversing column produces the single sharp metallic bang that gives the family its reputation.

The bang is not cosmetic. The disc-to-seat impact at 2 m/s of reverse flow hammers the seat face, and in the largest sizes the disc and arm inertia is enough to bend shafts and crack body webs. A second, quieter cost is the flutter: in lines with modest pulsation the disc hovers part open, chattering against the stop, wearing the pin and bushings until the disc no longer seats at all. Both failures trace back to the same root, which is that the swing check has no stored closing energy of its own. It only closes because the flow pushes it shut, and it can only be fast if the flow is fast, which is exactly when fast is the worst possible thing.

5. Closing dynamics by design: the families compared

Every non-slam design is an attempt to close the disc faster or more gently than the reverse flow would do it. The table summarises how the common families behave and where each earns its keep.

Two readings of the table are worth making explicit. The first is that travel distance matters as much as mechanism: a nozzle check disc moves millimetres along the axis, so even a moderate spring closes it before the column has reversed, while a swing disc must travel the full bore diameter. The second is that lighter is faster: a dual plate valve has two short, light discs against small torsion springs, which is why a DN600 dual plate outperforms a DN600 swing check despite looking fragile beside it. The ball valve family sits upstream of all of this as the isolation valve, and the pairing of a low Cv high-recovery isolation valve with a fast-closing check is a separate sizing conversation about the Cv and recovery topics covered in the sizing article in this series.

6. Choose by deceleration, not by size

The mistake that produces slammed swing checks is selecting on line size and pressure class alone. The parameter that actually selects the valve is the flow deceleration the pump failure produces, expressed in m/s², and it is estimable before anyone buys anything: the pump's own deceleration after power loss is on the order of 2 to 6 m/s² for typical centrifugal sets, and the column deceleration after flow reversal is set by the static head against system friction. A rough practical screen: if the static head across the line is small relative to friction losses, the column decelerates gently and almost any check works; if the static head is large, the reversal is violent and the check valve must close on the order of 0.2 seconds or faster, which eliminates the swing family in anything above DN150.

The second selection input is what the valve must survive after closing, because in many pump-failure events the check valve is asked to hold a large static head continuously, and in some it takes a second surge when the reflected wave returns. The third is the pump: if the pump is not rated for reverse rotation, the check valve is its only protection, and the reliability of the closure moves from a nuisance question to a machine-protection question. Spring-loaded axial checks and dual plate checks both close fast enough to protect the pump; a conventional swing check does not, and specifying a reverse-rotation-rated pump to accommodate it is usually the more expensive of the two options.

7. System-side measures a check valve cannot replace

A faster check valve reduces the surge from flow reversal, but three other surge mechanisms are untouched by it, and honest specifications say so. Column separation at high points, where the pressure falls below vapour and the column breaks and rejoins, produces surges governed by pipe profile, not by the check valve; the fixes are profile changes, air valves, or one-way surge tanks. Slow-operating isolation valves downstream, particularly motorised gate or butterfly valves closing against a running pump, generate their own surge on their own timeline and need their own closing curves. And trapped-pump or dead-headed pump trips generate surges inside the pump itself that only the control system can mitigate.

Where the check valve alone cannot carry the duty, the standard companions are: a surge vessel or air chamber near the pump discharge, which supplies and absorbs fluid to stretch the deceleration; controlled-closure motorised valves acting in sequence with the trip; and air release valves at the summits sized for both the air discharge and vacuum cases. None of these is exotic; the failure pattern is that the check valve gets selected in isolation, in a documents package that never shows the transient study, and the pipework finds the gap at commissioning.

8. What to put on the specification

A non-slam claim is only enforceable if it is expressed as a number against a defined event. The minimum data set is short, and a request that omits it will get quotations that are not comparable.

The last practical point is the one most often missed: reverse velocity at seat contact is measurable, and a supplier confident of the claim will state it per size without being pressed. Where the figure is missing from an offer, it is usually because the valve would not meet it, and the comparison the buyer needed was hidden by the omission. Asking for it costs one line in the enquiry and is the cheapest transient protection in the whole package, a point that applies equally when the same line is being reviewed for the seat leakage classes and test standards covered elsewhere in this series.

Why does a swing check valve slam? 

A swing check has no stored closing energy of its own: the disc must swing 70 to 90 degrees on a pin, driven only by the reverse flow. By the time it reaches the seat the column has usually built up 1.5 to 2.5 m/s of reverse velocity in a long line, and stopping that velocity produces the sharp pressure spike and the metallic bang. Faster designs shorten the travel or add spring energy so the disc contacts the seat while the reverse velocity is still small.

What is the difference between a silent check valve and a dual plate check valve? 

Both are fast-closing spring-assisted designs, but the mechanism differs. A silent or nozzle check moves its disc axially along the flow axis over a few millimetres and typically contacts the seat at 0.1 to 0.3 m/s of reverse velocity, the fastest closure available. A dual plate valve uses two half-discs on torsion springs and closes at about 0.3 to 0.6 m/s, which is far better than a swing check and much more compact, but slightly slower than the axial design in critical long lines.

How much pressure rise does water hammer actually create? 

The Joukowsky equation gives about 10 to 14 bar of pressure rise per 1 m/s of velocity change in a typical steel water pipeline, depending on pipe stiffness. Stopping 3 m/s instantaneously therefore adds roughly 30 to 40 bar on top of the static head. The rise applies only if the stop is faster than the pressure wave can travel the line and return; slower deceleration softens the peak, which is exactly the principle non-slam check valves exploit.

How do I know if my check valve is the cause of the surging? 

The signature is a single sharp knock at the pump discharge immediately after a pump stop, accompanied in larger sets by visible pump backspin. If the line is long, the static head is large relative to friction, and the installed valve is a conventional swing check, the closure surge is the most likely cause. A transient calculation of the reverse velocity at seat contact, compared against the valve's stated figure, confirms it.

Can a non-slam check valve replace a surge vessel?

No. A fast-closing check valve only limits the surge caused by flow reversal at the pump. Column separation at high points, slow-closing isolation valves and pump-internal transients are untouched by it and still need air valves, surge vessels or controlled closure sequences. The check valve and the surge vessel solve different parts of the same event and are usually specified together in long or high-head lines.