Water and air behave very differently when a vessel fails. A tank full of water that splits releases very little stored energy, because water is nearly incompressible. A tank full of compressed gas releases everything that went into compressing it, in a fraction of a second, and the fragments travel. That single difference governs how the two tests are treated.
Hydrostatic testing is therefore the default for proving strength. Pneumatic testing is a leak-location tool and, on some assemblies, the only practical option. Using one where the other belongs is not a matter of preference. It is a decision about what happens to the people standing nearby if the part gives way.
A water test proves that the vessel holds shape and holds its joints under load, and that no leak path exists through wall or weld at the test condition. It stresses the whole structure rather than the surface, which is why it catches what a surface test cannot: an undercut that became a through-wall defect under load, a root that opened as the shell strained.
It also has a useful side effect. Because the medium is nearly incompressible, a failure tends to produce a weep or a small split rather than an explosion. The part deforms, water escapes, pressure falls. The operator has time to step back, which is why the method is trusted on fabricated tanks where the volume is large.
Air has one advantage water cannot match: it finds the leak. Gas escaping through a small path can be detected with a bubble solution at the exact point, with no need to drain and dry the part. On an assembly that must ship dry, or a shape that traps water in a cavity it cannot drain, that is decisive.
Air is also practical on small assemblies where filling with water is awkward and the stored energy is genuinely limited. The judgement is not whether gas is acceptable in principle. It is how much volume and pressure are involved, and what the part is surrounded by when the test runs.
Preparation matters more than the test. Every opening needs a plug or blank rated for the duty, and each one restrained so it cannot become a projectile. A blank held by a single clamp is not a closure; it is a hazard with a test attached. Venting goes at the highest point so the vessel fills completely, because trapped air cushions the pressurisation and hides a leak.
The fill should be slow. A fast fill sends a shock through a baffle or a weld that the test pressure never would, and the damage gets blamed on the material. Once full, the vessel should settle so remaining air can be bled off, and the outside should be dried and marked so a weep shows against a clean surface.
A water test leaves water behind, and in a fuel tank or an intake part that residue is a problem in its own right. It promotes corrosion in the places the test was meant to prove, and it contaminates whatever flows through the part later. Drain points, baffle geometry and filler position decide how much stays inside.
Drying is part of the procedure, not an afterthought. Air through the vent, a dry warm environment, a period of drainage with the part inverted: the method matters less than having one and recording that it was done. A tank sealed wet and shipped is a warranty claim waiting to happen, and it will read as a material fault.
When the question is where rather than whether, gas wins. A water weep on a vertical surface runs down and appears some distance from the defect, and on a coated part the trace is hard to follow. Bubble solution applied to a joint shows the exact point and can be wiped and re-applied as the operator works along the seam.
That is why many shops run a two-stage approach: water to prove the structure, then a low-energy gas check on the joints where a leak is most likely. The second stage does not need to be a strength test, so stored energy stays small while diagnostic value stays high.
Stored energy scales with volume and pressure, and it rises steeply. A small cavity at modest pressure is manageable. The same pressure in a large tank is not. The common mistake is to treat pressure as the only variable and ignore volume, which is how a procedure that was safe on a compact manifold gets copied onto a full-size tank.
Practical controls follow from that: barriers, exclusion zones, remote gauges, and a rule that nobody stands in line with a blank during pressurisation. None of it is exotic, and all of it gets skipped when the test is treated as a formality at the end of a long day.
The choice belongs in the job before fabrication starts, because it affects where drain points go, how baffles are arranged and which surfaces get finished. A tank designed without a drain cannot be water tested and dried properly, and that constraint usually surfaces too late to fix cheaply.
On our aluminium fuel tanks that reasoning is settled at the drawing stage. Tanks are built with drainage and venting thought through, because the test is part of the design rather than an inspection bolted on at the end.
A certificate that records only a pass tells a buyer almost nothing. What matters is the method, the medium, the condition of the part, how long it was held, what was observed and who decided. Those fields separate a record from a receipt.
Buyers should also ask where the test sits in the sequence. A tank tested before the final fittings are welded has not been tested at all. A part tested after finishing cannot be reworked without destroying the finish. The sequence shows whether the shop understands what it is doing.
The same discipline applies to assemblies that hold no fluid at all. On fabricated titanium components the question is usually a sealed cavity rather than a tank, and the reasoning transfers: know the volume, control the energy, define the sequence, record the decision.
Work of this kind, from welded water tanks to general fabrication, is described on our company profile, which sets out the plant where it is carried out. The test is not there to prove the welder was right. It is there to find the part that is wrong before a customer does. How the method is agreed for a particular job is set out under custom work.
