Hydrostatic vs Pneumatic Leak Testing — ASME B31.3 and B31.1

ASME B31.3 makes the hydrostatic leak test the default and the pneumatic test a conditional alternative, and the reason is energy: a gas-filled system at test pressure stores orders of magnitude more energy than the same system full of water. This guide covers the hydrostatic formula P_T = 1.5·P·(S_T/S) and the ¶345.2.1 ceilings that bound it, when pneumatic testing is permitted and what it costs you in proof, a fixture-verified worked example, and the component limits that routinely make the calculated piping pressure the wrong number to actually apply.

Content last reviewed 2026-09-11 · page regenerated 2026-09-14 at build.

Why hydrostatic is the default

The choice between test media is a stored-energy decision before it is a code decision. Water is very nearly incompressible: pressurising a water-filled system to 500 psi stores almost no energy, because almost no volume change was needed to get there. When a hydrostatic test fails, the pressure collapses essentially instantly, and what you have is a leak, a wet floor, and a repair.

A gas-filled system at the same pressure is a different object entirely. The gas has been compressed, and that compression is stored energy waiting for a release path. The energy stored in a pneumatically pressurised system exceeds the hydrostatic case by two to three orders of magnitude for ordinary geometries. A pneumatic failure is not a leak — it is a rupture that accelerates fragments and generates a blast wave, and the exclusion zones that accompany pneumatic testing exist because that risk is real and has killed people.

This is why the codes do not present the two as equivalent options. B31.3 permits a pneumatic test where a hydrostatic test is impracticable, and it pairs that permission with a substantially lower test pressure, a preliminary low-pressure check, and procedural requirements that hydrostatic testing does not carry. The engineering question is never "which would I prefer" — it is "can I justify that hydrostatic is impracticable," and if you cannot, the decision is already made.

The usual legitimate justifications are specific and physical: the structure or the supports cannot carry the weight of the water (large-diameter gas lines and elevated runs are the classic case); the process cannot tolerate any residual moisture (cryogenic service, high-purity systems, catalyst-bearing systems, some instrument and analyser lines); or the fluid service reacts with water. "The schedule is tight" and "we do not have a water source" are not on that list, though they are frequently the actual reason.

The hydrostatic test pressure formula, and the cap inside it

ASME B31.3 ¶345.4.2 sets the minimum hydrostatic test pressure with Eq. (24):

P_T = 1.5 · P · ( S_T / S )

where P is the internal design gage pressure, S is the allowable stress at the design temperature, and S_T is the allowable stress at the test temperature. The 1.5 is the familiar proof factor. The ratio is the part engineers most often drop, and dropping it produces an unconservative test.

Why the ratio is there. The test is almost always performed cold, and allowable stress is higher cold than hot. A cold test at 1.5 × P therefore does not demonstrate the same margin against the hot design condition that 1.5 × P would demonstrate if the system were at design temperature. The S_T/S ratio scales the test pressure up to recover the equivalent proof. For a line whose design temperature is modest, S_T/S is close to 1 and the correction is small. For a line running hot, S collapses, the ratio grows, and the required test pressure climbs steeply.

And what bounds it. Left alone, that behaviour runs away: a design temperature at which S has fallen to a small fraction of its cold value would demand a test pressure many times the design pressure — a test that would yield the pipe rather than prove it. The current B31.3 does not cap the ratio itself; Eq. (24) is written without a ceiling on S_T/S. What bounds the test is ¶345.2.1(a): where the test pressure would produce a circumferential or longitudinal stress above the yield strength at test temperature, or exceed 1.5 times a component rating at test temperature, the test pressure may be reduced to the maximum that stays within the lesser of those two limits. On a hot line that provision, not the ratio, sets the number, and the code has effectively said that beyond it a proof test is no longer the right instrument.

Many owner specifications add a ceiling of their own on the ratio — 6.5 is a common figure, and it is the default on the calculator's cap input — but that is a specification limit, not a paragraph of the code. Set the cap to whatever your specification states, or to a large value where none applies; the ¶345.2.1(a) limits govern either way. Both are ceilings on the test, not on the design.

ASME B31.1 takes the same shape with different numbers, and the hydrostatic requirement in ¶137.4.5 is written around 1.5 × design pressure. Because the factors, the cap and both allowable stresses all move between codes and editions, the calculator treats every one of them as a user input rather than a constant — you enter the values your code of record actually specifies, and the arithmetic is done exactly.

When pneumatic is permitted — and what you give up

Where a pneumatic test is justified, B31.3 ¶345.5 sets the test pressure at 1.1 × the design pressure. B31.1 ¶137.5.5 sets not less than 1.2 × and not more than 1.5 ×. Notice the direction: the pneumatic factor is lower than the hydrostatic 1.5, not higher. That is deliberate — the code is capping stored energy, and it accepts a weaker proof in exchange.

That trade is the thing to be clear-eyed about. Run the numbers on an ordinary 300 psi line and the hydrostatic test lands above 500 psi while the pneumatic test lands at 330 psi. The pneumatic test is a genuine leak test; it is a much weaker proof test. Defects that a hydrostatic test would find, a pneumatic test at 1.1 × P can pass over.

Pneumatic testing also brings procedural requirements that hydrostatic testing does not:

• A preliminary low-pressure check before stepping to test pressure — commonly cited as 25 psi gage — to find gross leaks while the stored energy is still trivial. Confirm the value and paragraph against your edition.
• Staged pressurisation with holds. Pressure is raised in increments with a pause at each step to allow the system to equalise and to be examined, rather than run straight to the target.
• An exclusion zone and a written procedure. Personnel are kept clear during pressurisation, and access is controlled while the system is at pressure.
• Attention to the relief path. A pressure-relieving device set appropriately above test pressure, so an error on the pressurising side cannot walk the system past its limit.

One more consideration that rarely makes the decision but should: a pneumatic test at ambient temperature on carbon steel raises a brittle-fracture question that hydrostatic testing largely sidesteps, because the stored energy is available to drive a crack. Where low test-metal temperature and heavy wall coincide, that deserves its own look rather than a footnote.

Worked example — a 300 psi line, both media

A process line with an internal design gage pressure of 300 psi. The allowable stress at the design temperature is 17,500 psi; at the test temperature it is 20,000 psi. B31.3 factors apply: hydrostatic 1.5, pneumatic 1.1, with the calculator's stress-ratio cap left at its 6.5 default (a specification-style ceiling — the code sets none).

Step by step

1. Stress ratio: S_T/S = 20,000 / 17,500 = 1.14286.
2. Check it against the cap input: 1.14286 is well under 6.5, so no cap engages and the computed ratio is used as-is.
3. Hydrostatic test pressure: P_T = 1.5 · 300 · 1.14286 = 514.286 psi.
4. Pneumatic test pressure: 1.1 · 300 = 330 psi.

The 184 psi gap between the two media is the whole argument in one number. Note also how modest the ratio correction is here — 14% — because the design temperature is not extreme. Had the same line been specified for high-temperature service with S down at, say, 2,500 psi, the ratio would compute to 8.0 and Eq. (24) would ask for 1.5 · 300 · 8.0 = 3,600 psi — a 6.5 specification cap would trim that to 2,925 psi — and either way that is the regime where the ¶345.2.1(a) yield check and the component ratings take over the conversation entirely.

The trap: the piping minimum is not the system maximum

Everything above computes a minimum test pressure for the pipe. Neither word is incidental, and conflating them is the most common way a correctly calculated test pressure damages equipment.

The system contains components with their own limits: flanges with pressure-temperature ratings, valves with seat and shell test limits, expansion joints and bellows with test pressures well below what the adjacent pipe will take, instruments, sight glasses, rupture-disc holders, and pump and compressor casings. A test pressure that is correct and conservative for the pipe can be destructive for a bellows and can unseat a valve.

The practical sequence is:

• Compute the piping minimum from the formula above.
• Identify the limiting component in the test boundary and establish its permitted test pressure from the manufacturer's data or the applicable component standard.
• If the component limit is below the piping minimum, either isolate, blind or remove that component for the test, or subdivide the system so that each test boundary contains only components that will take its pressure.
• Only if neither is practical does the conversation turn to testing at a reduced pressure with a documented engineering justification — and that is a decision with a signature on it, not a field adjustment.

The calculator raises this as a standing warning on every run for exactly that reason: it returns the piping number, and the piping number is not the whole answer. It is also worth remembering that a component standard's own hydrostatic shell test — a flange proof-tested by its manufacturer, for instance — is a manufacturing test at the factory, not a licence to hold your installed system at that pressure.

The alternatives, and where they fit

B31.3 provides routes other than a full hydrostatic or pneumatic test, and each is tied to a fluid-service category rather than being freely available:

Two habits keep this defensible. Record the code edition alongside the test pressure on the test package, because the factors and the cap are edition-dependent and a reviewer should not have to guess which set was used. And record the allowable stresses and their source, not just the ratio — S_T/S = 1.14286 tells a reviewer nothing about whether the right material line was read from the right table.

When you have the numbers, the leak test pressure calculator runs the whole comparison free and in your browser. It sits alongside the B31.3 wall thickness check that establishes the design pressure the test scales from, and the allowable pressure calculator when you need to rate the wall you actually installed rather than the one you specified.

FAQ

Why does the hydrostatic test pressure include the ST/S ratio at all?

Because the test is normally run cold and the design condition is normally hot. Allowable stress is higher at the lower test temperature, so a cold test at 1.5 × design pressure demonstrates less margin against the hot design condition than the same factor would at temperature. Multiplying by S_T/S scales the test pressure up to recover the equivalent proof. Omitting the ratio — testing at a flat 1.5 × P — under-tests any line whose design temperature is meaningfully above the test temperature, and the hotter the service the larger the shortfall. The size of the shortfall is easy to underestimate: at S_T/S = 1.3, a flat 1.5×P test runs 23% light — on a 300 psi line, that is proving 450 psi where the code asked for 585 — a 135 psi hole in the demonstrated margin.

Does B31.3 cap the S_T/S ratio?

No — not in the current edition. ¶345.4.2 writes Eq. (24) without a ceiling on the ratio, and the 2016, 2022 and 2024 texts all read the same way. What bounds a hot line's test is ¶345.2.1(a): if the test pressure would push the circumferential or longitudinal stress past yield at test temperature, or exceed 1.5 times a component's rating at test temperature, the test pressure may be reduced to the maximum that stays within the lesser of the two. A ratio cap — 6.5 is the figure many owner specifications use, and the default on the calculator's cap input — is a specification limit layered on top, and the report shows the ratio actually applied so a capped case is visible as capped. Reaching either limit is information in its own right: a line hot enough to need it is deep in the regime where creep, not proof pressure, governs integrity, and the code is telling you the pressure test has stopped being the meaningful qualification.

Can I choose a pneumatic test because it is faster and cleaner?

Not on those grounds. B31.3 permits pneumatic testing where hydrostatic testing is impracticable — the recognised justifications are physical: the structure or supports cannot carry the water weight, the process cannot tolerate residual moisture, or the fluid service reacts with water. Schedule pressure and water availability are not justifications, though they are often the real motivation. The reason the code is strict is stored energy: a pneumatic test at the same pressure holds two to three orders of magnitude more energy than a hydrostatic one, and a failure is a rupture rather than a leak. The documentation matters as much as the decision: “hydrostatic impracticable — rack never rated for flooded weight, calculation attached” survives an audit; “schedule pressure” does not, and writing the justification down forces the check of whether it is actually true.

The pneumatic test pressure is lower than the hydrostatic one. Is that backwards?

It is deliberate, and it is the trade at the centre of the decision. B31.3 sets pneumatic at 1.1 × design against hydrostatic at 1.5 × design × the stress ratio, because the code is limiting the energy stored in the system rather than maximising the proof obtained. You are accepting a weaker demonstration in exchange for a survivable failure mode. Be explicit about that when the choice is made: a pneumatic test is a competent leak test and a substantially weaker proof test, and defects a hydrostatic test would have found can survive it. The stored-energy arithmetic also drives every downstream precaution — the exclusion zone, staged pressurization, the low-pressure preliminary check — so treat those as the price of admission for the weaker proof, not as bureaucracy layered onto a routine test.

Which pressure do I actually apply — the calculated one or the component limit?

Whichever is lower, after you have deliberately set the test boundary. The formula gives the minimum the piping requires; the components inside the boundary impose maxima. If a bellows, a valve, an instrument or a pump casing cannot take the piping minimum, the correct move is to isolate, blind or remove it, or to subdivide the system so each boundary contains only components that will take its pressure — not to quietly lower the test pressure. Testing below the code minimum is an engineering decision requiring documented justification and a signature, not a field adjustment made because a gauge would not hold. One boundary resolution worth knowing: a component that cannot take the piping minimum often can take its own — a vessel tested separately at its code pressure, then tied in with a final closure weld examined in lieu of test, is a resolution the code itself provides for.

Does the calculator embed the allowable stresses?

No, and that is intentional. S and S_T are code table values for your material at your two temperatures, and reproducing those tables is not something this site does — the same data firewall applies across every calculator here. You enter both allowables, along with the hydrostatic and pneumatic factors and the ratio cap, from the edition your project actually invokes. What the tool guarantees is that the arithmetic is exact, repeatable, and locked by a regression fixture; selecting the right values from the right table remains the engineer's responsibility, and it is worth recording their source on the test package. One check worth a minute per project: confirm S and S_T landed in the right fields, because a transposed pair under-tests silently — the ratio inverts below 1.0 and the shortfall runs in exactly the direction nobody catches at the manifold.