Hydrotest vs Pneumatic Test — the Stored Energy You Cannot See

The site situation

Friday afternoon. A 100 m run of 12″ Sch 40 gas line, design 20 barg, is mechanically complete and the hydrotest is due. The subcontractor's supervisor comes to you with a proposal that sounds entirely reasonable:

"There is no water on site yet, the rack was never checked for a water-filled line, and the client wants the line bone dry for start-up anyway. Let's just do an air test. Same pressure, no water to source, no water to dispose of, no drying afterwards. It's only air."

Guess first, before reading on: at the same test pressure and the same pipe, how much more energy is stored in the air than in the water? Twice? Ten times? Write your number down.

The answer for this line, at 22 barg, is about 3,000 times. The water holds 7.9 kilojoules — roughly a brick dropped from a third-floor window. The air holds 24.5 megajoules — about 5.3 kg of TNT, and a personnel exclusion radius near 26 metres.

Nothing about the test is "easier". The plumbing is easier. The consequence of a failure is in a different universe.

The misconception, stated plainly, then dismantled

"An air test is easier and safer than filling the line with water."

The first half is true and it is exactly why the second half gets believed. Air weighs nothing, needs no source, no disposal permit, no drying, no freeze protection, and the compressor is already on site. Every logistical signal says air is the easy option — and people generalise "easy" into "safe".

Now the physics. A pressure test is a deliberate act of storing energy in a container you are not yet sure about. The only question that matters is: when it lets go, what comes out?

Water is nearly incompressible. To raise 7.2 m³ of water to 22 barg you must squeeze roughly 7.2 litres of extra water into the line (ΔV = p·V/K = 22e5 × 7.22 / 2.2e9) — that is the entire "spring". When a crack opens, those 7.2 litres escape in milliseconds, the pressure collapses to zero essentially instantly, and the crack stops running because its driving force has vanished. Hydrotest failures are loud, wet, embarrassing, and locally dangerous within a metre or two of the jet. They are almost never fatal at a distance.

Air is a spring the length of the pipe. The same 7.2 m³ at 22 barg contains about 160 m³ of free air. When a crack opens, the gas behind it keeps pushing — it expands, drives the crack along the pipe, launches fragments, and dumps its energy into a blast wave. The pressure does not collapse; it feeds the failure. That is why a pneumatic failure is an explosion and a hydrostatic failure is a leak.

Water (liquid spring):   E = p² · V / (2·K)                   K ≈ 2.2 GPa
Air (gas spring, isentropic expansion to atmosphere):
                         E = p₁·V/(k−1) · [1 − (pₐ/p₁)^((k−1)/k)]      k = 1.4

Look at the shape of those two expressions. Water's energy is proportional to p²/K, and K is two thousand two hundred megapascals — the divisor is enormous, so the result is tiny. Air's energy is proportional to p·V directly, with no big divisor anywhere. That structural difference, not a detail of the numbers, is the whole lesson.

Put your own line into it — size, length, pressure, medium — and watch both energies, the ratio, the TNT-equivalent and the indicative exclusion radius move together: ▶ open the interactive: construction hydrotest energy calc

Symbol key — every symbol on this sheet

• p — gauge test pressure (the pressure above atmosphere the gauge reads) · bar, Pa
• p₁ — absolute test pressure = p + pₐ; gas laws only ever take absolute · Pa
• pₐ — atmospheric pressure, 101.325 kPa · Pa
• V — internal volume of the system under test · m³
• K — bulk modulus of the liquid: pressure rise per unit fractional volume squeeze · Pa
• K_eff — K corrected for pipe-wall stretch, 1/K_eff = 1/K + D/(E·t) · Pa
• k — ratio of specific heats, c_p/c_v; 1.4 for air, ≈ 1.3 for most process gases · –
• E — stored energy released if the boundary fails · J, MJ
• W — TNT-equivalent mass = E / 4.6 MJ per kg · kg
• R — indicative exclusion radius from Hopkinson–Cranz scaling, R = Z·W^(1/3) · m
• Z — scaled distance; Z ≈ 15 m/kg^⅓ puts you at the ≈ 1 psi (6.9 kPa) injury threshold · m/kg^⅓
• P_T — code test pressure · barg
• P — internal design pressure of the piping · barg
• S_T / S — allowable stress at test temperature ÷ at design temperature (B31.3 correction) · –
• β — volumetric thermal expansion coefficient of water, ≈ 2.1 × 10⁻⁴ per K · 1/K
• D, t, E_steel — pipe OD, wall thickness, Young's modulus (200 GPa) · mm, mm, Pa

Test pressure, exclusion zones and the rest of the pneumatic package

Code test pressures — and why the multipliers differ

ASME B31.3 §345.4.2  hydrostatic:  P_T = 1.5 · P · (S_T / S)      (ratio need not exceed 6.5)
ASME B31.3 §345.5.4  pneumatic:    P_T = 1.1 · P                  (no temperature correction)
ASME B31.1           hydrostatic:  1.5 · P     pneumatic: 1.2 · P (cap 1.5 · P)

The naïve reading is "the code trusts water more". The real reading is the opposite — the code is buying different things with the two multipliers.

• With water, the overpressure margin is nearly free. 1.5 × costs you nothing in consequence, so the code takes a generous proof margin and, because allowable stress is higher at the cool test temperature, it lets you scale the test pressure up by S_T/S so the test still proves the hot design condition.
• With gas, every extra bar of test pressure is extra stored energy and extra chance of a running fracture. So the code spends the margin the other way: 1.1 × is barely a proof test at all, and it deliberately declines the temperature bonus. B31.3 additionally caps the pneumatic test so that the resulting stress stays below 90 % of yield, and requires a preliminary check at the lesser of 25 psi (1.7 barg) or 50 % of test pressure, then stepwise increase in 10 % increments with a hold and a leak check at each step.

Read the step-wise requirement as what it is: the code does not believe the line, so it makes you approach the energy level slowly with everybody standing far away. Nothing in the hydrostatic procedure looks like that.

When pneumatic testing is genuinely justified — and what it costs you

There are real cases. None of them is "we did not want to fetch water".

1. Traces of water are not tolerable — cryogenic service, instrument air/dry gas systems, chlorine, sulphuric acid, catalyst beds, lines where a residual droplet freezes into a hydrate or an ice plug, or where the drying cost after hydrotest exceeds the test cost.
2. The structure cannot take the water weight — an existing rack or platform never designed for a flooded large-bore line, a vessel on legs, an FPSO topside.
3. Freezing conditions — ambient below ≈ 5 °C with no heated water and no glycol, where the test itself would split the line.
4. Lining or internals would be damaged — refractory, some coatings, desiccant.

The moment you choose pneumatic, a package of extra requirements comes with it, and a good construction manager quotes all of it in the same breath:

• A written risk assessment / stored-energy calculation (ASME PCC-2 Article 5.1 gives the method) signed off before the compressor is connected.
• Higher NDE: typically 100 % radiography or UT of all pressure-retaining welds before the test, because you are no longer allowed to discover the bad weld by testing it.
• Stepwise pressurisation with holds, as above, with everyone outside the exclusion zone during every rise and only approaching during a held, stable step.
• A calculated and physically barricaded exclusion zone, with the calculation on the permit.
• Remote or shielded gauge reading, pressure relief on the test manifold set just above P_T, and a defined depressurisation rate.
• A check that the material is not at risk of brittle fracture at the test temperature — carbon steel at 2 °C during a gas test is a different problem than at 25 °C.

See the two failures side by side — fill the spool with water or air, then break it, and watch the pressure gauge collapse instantly in one case and the blast front expand past the barricade in the other: ▶ open the interactive: construction hydrotest energy 3d

Worked example — the Friday-afternoon line, both ways

12″ Sch 40 (ID 303.2 mm), 100 m, design 20 barg at 300 °C, A106 Gr B.

V   = π/4 · 0.3032² · 100                       = 7.22 m³
P_T(hydro) = 1.5 × 20 × (137.9 / 120.7)         = 34.3 barg      S_T/S ≈ 1.14
P_T(pneum) = 1.1 × 20                           = 22 barg

Hydrostatic at 34.3 barg

E = (34.3e5)² × 7.22 / (2 × 2.2e9)   = 19.3 kJ   ≈ 0.0042 kg TNT   R ≈ 2.4 m

Pneumatic at 22 barg — note this is the lower pressure, and it still loses by three orders of magnitude:

p₁ = 22e5 + 1.013e5 = 23.01e5 Pa
E  = 23.01e5 × 7.22 / 0.4 × [1 − (1.013/23.01)^0.2857]
   = 4.153e7 × 0.5903 = 24.5 MJ    ≈ 5.3 kg TNT    R ≈ 26 m

Ratio 1,270 × — at the code pressures each medium is actually allowed. Nineteen kilojoules versus twenty-four and a half megajoules. A 2.4 m "stand clear" versus a 26 m barricaded circle that will not fit between this rack and the next one.

And the weight side of the ledger, which is the real argument for the air test:

12″ Sch 40 pipe steel          79.7 kg/m
water in the bore              72.2 kg/m
test weight (pipe + water)    151.9 kg/m   =  1.9 × the empty/operating weight of a gas line

For a 24″ Sch 20 flare header it is worse: 141 kg/m of steel carrying 274 kg/m of water — a 2.9 × weight increase, all of it applied to supports and a rack that were sized for a pipe full of vapour. That is a genuine engineering problem, and it belongs in the piping stress model as a hydrotest weight case (W + water, no thermal), not in a conversation on site on a Friday. It is solved with temporary supports, not by changing the test medium.

Practical points that decide whether the test goes well

• Vent every high point while filling. A trapped air pocket is a pneumatic test inside your hydrotest, and it also gives a spongy, slowly-settling gauge that people misread as a leak.
• Support the water. Check the hydrotest weight case. Provide temporary supports for large-bore, and check that the rack and the foundations were designed for it.
• Pin the spring hangers. Variable and constant-effort springs must have their travel stops installed and pinned before filling — a spring set for the operating load will bottom out under the test weight, and worse, the pins must be removed and logged before start-up. A spring left pinned is a rigid support in a hot line, and it will show up later as a cracked nozzle. Keep the removal register.
• Test blinds are designed items, not scrap plate. They are sized for the test pressure with the correct material, numbered, registered and removed against the register.
• Isolate what must not see the test pressure — PSVs (remove or gag, and log), control valves, instruments, rotating equipment, expansion joints, in-line check valves that will not let you drain backwards.
• Never leave a system liquid-locked. Water trapped between two closed valves and warmed by the sun develops roughly 4.5 bar per °C (ΔP = K·β·ΔT). A 10 °C afternoon rise on a locked spool is 45 bar of nowhere-to-go. Depressurise, drain, and leave a vent cracked open.
• Drain and dry properly, respecting dead legs; for austenitic stainless, use test water with a controlled chloride limit (commonly ≤ 50 ppm) and dry promptly, or you have swapped a testing problem for a stress-corrosion-cracking problem.
• Hold times and gauges. B31.3 wants at least 10 minutes at pressure before examination; use a calibrated gauge with the test pressure in the middle of its range, and record ambient temperature — pressure drift on a long line is very often just the sun.

Common pitfalls

• Choosing pneumatic for convenience and only then discovering that the exclusion radius does not fit inside the site, so the zone quietly shrinks to what is available.
• Quoting "200 ×" as if it were the answer; at plant test pressures it is ten times better than reality.
• Testing pneumatically at the hydrostatic multiplier (1.5 × instead of 1.1 ×) because the procedure was copy-pasted — that is a 2 × increase in stored energy over what the code allows.
• Forgetting the water weight case entirely, then being surprised when a spring hanger bottoms out or a rack member visibly sags during filling.
• Leaving spring travel stops pinned after the test. Silent, common, and expensive.
• Discovering trapped air only from the gauge behaviour, after the crew has been standing next to the line for an hour.
• Applying blast/exclusion thinking to the test and forgetting that the same stored energy exists during a service leak test with process gas at operating pressure.

Outcome

• A pressure test stores energy deliberately. Water stores p²V/2K, which is tiny because K is 2.2 GPa; gas stores ≈ pV/(k−1) × [1 − (pₐ/p₁)^((k−1)/k)], which is not divided by anything large.
• At plant test pressures the pneumatic/hydrostatic energy ratio is 10²–10³, not 200; a 12″ × 100 m line at 22 barg holds 5.3 kg TNT-equivalent of air and needs a ≈ 26 m exclusion radius.
• Water's failure mode is self-arresting (pressure collapses with the crack); gas's failure mode is self-feeding (the crack can outrun its own decompression wave). That is why the code multipliers are 1.5 × and 1.1 ×, and why only the pneumatic procedure has step-holds and barricades.
• Pneumatic is justified by water intolerance, weight, freezing or lining damage — never by convenience — and it triggers a risk assessment, more NDE, stepwise pressurisation and an exclusion zone.
• Hydrotest weight is a real load case (a gas line can weigh ~2–3 × its operating weight when flooded): model it, support it, pin the springs, and unpin them afterwards.
• Interactive: ▶ open the interactive: construction hydrotest energy calc — energies, ratio, TNT, radius, weights. 3D: ▶ open the interactive: construction hydrotest energy 3d — the two failures, side by side.

Open items

• Add a nitrogen/helium option (k ≈ 1.4 / 1.66) and a real-gas correction for high-pressure tests.
• Replace the Hopkinson–Cranz indicative radius with the ASME PCC-2 Article 5.1 tabulated method and compare the two.
• Worked example of a test-blind thickness calculation, and a spring-stop removal register template.

Educational teaching material — simplified and illustrative. Not for engineering design use.
© EPiC Circle · provided by PipeAxis