Water Hammer Calculator — Joukowsky Surge Pressure and Wave Speed

Returns the pressure wave speed a corrected for pipe-wall elasticity, the instantaneous Joukowsky surge rise ΔP = ρ·a·ΔV/(144·g), and the critical closure period t_c = 2L/a — the round-trip time below which a valve closure develops the full surge. Bulk modulus K, pipe modulus E and the restraint coefficient c₁ are user inputs.

Method last updated 2026-07-30 (calculation changelog) · fixture-verified on every build — most recently 2026-07-31.

Water hammer is what happens when you take momentum away from a liquid faster than the liquid can be told about it. Close a valve on a flowing line and the fluid immediately upstream stops; the fluid behind it does not yet know, arrives, compresses slightly, and the pipe wall stretches. The result is a pressure rise that propagates back up the line as a wave at acoustic speed. The magnitude has a famously simple form — Joukowsky's — and the simplicity is deceptive, because two of its three ingredients are easy to get wrong.

The first thing engineers underestimate is the size. A 5 ft/s velocity change in a water line is unremarkable, and it produces a surge on the order of 275 psi on top of whatever the line was already running at. That is not a trim adjustment to a design pressure — for a line operating at 150 psi it is roughly a tripling, and it is the reason valve closure is a governing load case on liquid pipelines rather than a footnote.

The second is that the surge magnitude does not depend on the length of the line. Length changes the timing, not the amplitude. What length determines is t_c = 2L/a, the time for the wave to travel to the far boundary and return — and that sets whether the closure counts as fast or slow. A closure completed within that window develops the full Joukowsky rise regardless of how gently the operator thought they were being; a closure much slower than it develops less, because relief arrives before the valve is fully shut. On a 2,000 ft line the window is about one second, which means a great many valves that feel slow to a human are instantaneous to the pipe.

The third is the wave speed itself. The acoustic speed in water is roughly 4,700 ft/s, but a pipe is not rigid: the wall stretches under the pressure rise, adding compliance to the system and slowing the wave. This calculator applies the standard elastic correction, which for a steel line of ordinary proportions brings the wave speed down to around 4,100 ft/s and reduces the surge in the same proportion. Skipping that correction overstates the surge by roughly 15% on steel — and understates it badly if you assume steel numbers on a plastic line, where the wall is far more compliant and wave speeds can fall to a quarter of the rigid value.

Method

Three results, computed in sequence. Pressure wave speed first, with the elastic pipe-wall correction:

a = √( K·144·g / ρ ) ⁄ √( 1 + c₁·(K/E)·(D/t) )

The numerator is the rigid-pipe acoustic speed in the fluid; the denominator is the compliance correction. Note what drives it: the ratio K/E (how soft the pipe material is relative to the fluid) times D/t (how thin-walled the pipe is). A thin-walled plastic line is soft on both counts and its wave speed collapses; a heavy-wall steel line is close to rigid.

Joukowsky surge — the instantaneous pressure rise for a velocity change ΔV:

ΔP = ρ · a · ΔV / (144 · g)

Critical closure period — the wave round-trip time:

t_c = 2 · L / a

Units are US customary throughout: K and E in psi, ρ in lb/ft³, ΔV in ft/s, L in ft, g = 32.174 ft/s². The 144 converts ft² to in² so ΔP lands in psi. The restraint coefficient c₁ depends on how the pipe is anchored — whether it is free to move axially, anchored at one end, or fully restrained against longitudinal movement — and is your input; approximately 1.0 is typical for a thin-walled anchored line.

Entering L = 0 suppresses the critical-time output and returns wave speed and surge only, which is the right mode when you want the surge magnitude for a valve or pump-trip screen and the line length is not yet fixed. Every run carries a standing warning that ΔP is the instantaneous worst case: it is the rise on top of the operating pressure, and taking credit for a slower closure requires a transient analysis, not a rule of thumb.

Limitations — what this calculator is not

• Instantaneous closure only — this is the upper bound. The Joukowsky equation gives the surge for a velocity change completed within the critical period. It does not model a closure profile, and it cannot give you credit for a slower valve: that requires a transient (method-of-characteristics) analysis of the actual system.
• Single pipe, single velocity change, no reflections tracked. Real systems have branches, changes of section, pump characteristics, air vessels, surge relief valves and check valves that slam. Wave reflections at each of those can superpose and produce local pressures above the single-pipe Joukowsky value — this card does not see any of them.
• Line pack, friction and column separation are outside the model. On long lines the friction gradient adds to the surge on closure (line packing), and if the pressure falls to vapour pressure anywhere the liquid column can separate and rejoin, producing a rejoining shock that can exceed the original surge. Neither is captured here, and both are reasons a long-line screen must escalate to a transient study.
• Liquids only. The derivation assumes a slightly-compressible liquid filling the pipe. It does not apply to gas or two-phase flow, and entrained gas — even a small volume fraction — dramatically reduces the effective bulk modulus and the wave speed, which the card has no way to detect from the inputs you give it.
• ΔP is a rise, not an absolute. The number returned is added to the operating pressure at that point, and the pressure that matters is the peak at the worst location — typically immediately upstream of the closing valve, and at low points where static head adds. The card does not know your operating pressure or your elevation profile.
• c₁ is a user input and the answer is sensitive to it. The restraint coefficient depends on the anchoring arrangement and on Poisson effects; picking it carelessly moves the wave speed and therefore the surge. It is exposed rather than assumed so that the assumption is visible on the face of the calculation.

Quick reference — is your closure fast or slow?

Everything about how seriously to take a valve closure follows from comparing the closure time against t_c = 2L/a. The bands below are for screening only; the linear approximation in the third column is the classical one and is not computed by this card.

Two traps worth naming. First, the effective closure time of a valve is not its stroke time — most of the flow is shut off in the last part of the travel, so a 10-second valve can behave like a 2-second one. Second, a check valve slamming shut on flow reversal is an uncontrolled closure with no stroke time at all; it belongs in the rapid row regardless of what the rest of the system does.

Worked example — fixture-verified

A water line — NPS 12 Schedule 40 carbon steel, 12.0 in bore on a 0.375 in wall — running 2,000 ft from a supply vessel to a valve. Water at 62.4 lb/ft³ with a bulk modulus of 300,000 psi, steel at 29.5×10⁶ psi, thin-wall anchored restraint (c₁ = 1.0). The valve closes on a flow of 5 ft/s.

Step by step

1. Rigid-pipe acoustic speed: a_rigid = √(K·144·g/ρ) = √(300,000·144·32.174/62.4) = √22,274,308 = 4,719.57 ft/s.
2. Elastic correction term: c₁·(K/E)·(D/t) = 1.0·(300,000/29,500,000)·(12/0.375) = 0.0101695·32 = 0.325424.
3. Correction divisor: √(1 + 0.325424) = √1.325424 = 1.151271 — the pipe wall costs about 13% of the wave speed.
4. Wave speed: a = 4,719.57/1.151271 = 4,099.44 ft/s.
5. Joukowsky surge: ΔP = ρ·a·ΔV/(144·g) = 62.4·4,099.44·5/(144·32.174) = 1,279,025/4,633.06 = 276.065 psi.
6. Critical closure period: t_c = 2L/a = 2·2,000/4,099.44 = 4,000/4,099.44 = 0.9757 s.

Read the two numbers together. A 276 psi rise on a 5 ft/s water line is large enough to govern the pressure design of the pipe — and the critical period says any closure completed inside 0.98 seconds develops all of it. That is the number to take to the valve specification: a quarter-turn valve on an actuator, or a check valve slamming on reversal, closes well inside a second and gets the full surge. Take ΔP into the <a href="/calculators/b313-pipe-wall-thickness/">B31.3 wall thickness</a> or <a href="/calculators/pipeline-maop/">MAOP</a> check as an addition to the operating pressure, not as a replacement for it.

Worked example 2 — surge magnitude without a line length

The same fluid, pipe and velocity change, but the line length is left at zero — the screening mode you use when you want the surge magnitude for a valve or pump-trip case before the routing is fixed. The point of running it is to make explicit what length does and does not affect.

Step by step

1. Wave speed is unchanged at 4,099.44 ft/s — it depends on the fluid and the pipe, not on how long the pipe is.
2. The surge is also unchanged at 276.065 psi. This is the substantive point: ΔP does not scale with line length. A 200 ft line and a 20,000 ft line of the same pipe and fluid produce the same Joukowsky rise for the same velocity change.
3. With L = 0 the critical period is suppressed and returned as 0 rather than a misleading number.
4. What has been lost is the timing question. Without L you know how big the surge is, but not whether your valve is fast enough to cause it — so this mode screens magnitude, and the length must come back before the closure can be classified.

The practical use of this mode is early screening: given a fluid, a candidate pipe and the velocity you intend to run, you get the surge that any rapid closure will produce, before a single foot of routing exists. If that number already breaks the pressure design, the fix is a velocity or a pipe decision — and it is far cheaper to discover at that stage than after the isometrics are issued. Length matters when you come back to ask whether the specified valve is fast enough to cause it.

Fixture case “no-length mode skips critical time” (tolerance 0.000001) — locked in the same release gate as the example above.

Sources & citations

• N. E. Joukowsky, 'Über den hydraulischen Stoss in Wasserleitungsröhren' (1898) — the momentum-based derivation of ΔP = ρ·a·ΔV.
• Standard elastic pipe-wall wave-speed correction, a = a_rigid / √(1 + c₁·(K/E)·(D/t)), with c₁ set by the pipe restraint condition — user-supplied.
• ASME B31.4, Pipeline Transportation Systems for Liquids and Slurries — surge pressure provisions; codes commonly permit a bounded overpressure above MAOP for surge, evaluated against the code of record.
• Constant: g = 32.174 ft/s²; the factor 144 converts lb/ft² to psi. Physical and unit constants, not code values.

Per the source & citation policy, allowable-stress and factor table values are user-supplied. Where a page does reproduce specific ASME data (the B16.5 ratings, the quick-reference tables), it does so under ASME authorization and states the source table and conditions inline.

FAQ

Does the surge get bigger on a longer line?

No — and this is the most common misconception about water hammer. The Joukowsky rise depends on fluid density, wave speed and the velocity change, none of which involve length. A 200 ft line and a 20,000 ft line of identical pipe and fluid produce the same ΔP for the same closure. What length changes is the critical period 2L/a: a longer line has a longer window, so a closure of given duration is more likely to count as rapid and develop the full rise. Length also brings in effects this equation excludes — line packing and column separation both grow with length, and on long lines those can push the peak pressure above the Joukowsky value rather than below it.

My valve takes 10 seconds to close. Am I safe?

Compare it against t_c, not against your intuition — and be careful what you mean by 10 seconds. Two things routinely defeat this reasoning. First, the effective closure time is not the stroke time: most valve types shut off the great majority of the flow in the final portion of travel, so a 10-second stroke can present the line with an effective closure of a second or two. Second, the stroke time only applies to the controlled case. A check valve slamming on flow reversal after a pump trip has no stroke time at all, and pump trip is usually the governing transient on a liquid line precisely because nothing is controlling it. If the margin matters, model it — this card gives you the bound and the window, not permission to take credit.

Why is the calculated wave speed lower than the speed of sound in water?

Because the pipe is not rigid. Sound travels through unconfined water at roughly 4,700 ft/s, but in a pipe the wall stretches slightly under the pressure wave, adding compliance to the system and slowing the wave down. The correction term c₁·(K/E)·(D/t) quantifies that: it grows with a softer pipe material (larger K/E) and with a thinner wall relative to diameter (larger D/t). In the worked example the correction takes 4,720 ft/s down to 4,099 ft/s — about 13%. On a thin-walled plastic line the effect is far larger; wave speeds of 1,000–1,500 ft/s are ordinary for HDPE, with the surge reduced in the same proportion. That is a genuine and useful advantage of plastic pipe in surge-prone service.

How do I use ΔP in a code check?

As an addition to the operating pressure at the location where the surge peaks, which is normally immediately upstream of the closing valve, and then combined with static head at low points. Most codes treat surge as an occasional or transient condition and permit a bounded overpressure above the design pressure or MAOP — commonly on the order of 10% for liquid pipelines — but read that allowance carefully: it is an allowance against MAOP, not a licence to add the surge on top of the allowance. If the peak fits inside it, the line is qualified for the transient; if it does not, the answer is a pressure design change, a surge mitigation device, or a slower closure demonstrated by transient analysis.

What about air chambers, surge tanks and relief valves?

They work, and this calculator does not model any of them. Accumulators, surge vessels, air chambers and surge-relief valves all act by giving the decelerating column somewhere to go, and a properly sized one can cut the peak substantially. But sizing them is a transient-analysis problem — the device has to respond within the same fraction of a second the wave is travelling in, and an undersized or slow-acting device provides much less relief than its nameplate suggests. Use this card to establish the unmitigated surge and to size the problem; use a method-of-characteristics model to design the mitigation and to prove it works.

Can I use this for a gas line or a two-phase line?

No. The derivation assumes a slightly-compressible liquid completely filling the pipe, and neither condition holds. Gas is far too compressible for the acoustic-wave treatment used here, and transient analysis of gas lines is a different problem entirely. Two-phase and gas-entrained liquids are worse than merely inapplicable — they are misleading, because even a small volume fraction of entrained gas collapses the effective bulk modulus and can cut the wave speed by a large factor, so this card would return a surge substantially higher than reality. If you know or suspect there is gas in the line, the honest answer is that the inputs no longer describe the fluid.

Related calculators

• Pipeline MAOP Calculator — Maximum Allowable Operating Pressure (ASME B31.8 / B31.4) — The steady-state ceiling the surge peak has to fit under
• ASME B31.3 Pipe Wall Thickness Calculator — Add the surge rise to design pressure in the wall check
• Pipe Support Spacing Calculator — Maximum Deflection-Limited Span — Unbalanced surge thrust loads the supports the wave travels past
• Slurry Critical Deposition Velocity Calculator (Durand–Condolios) — Slurry lines are liquid lines — they surge on closure too