Pressure testing puts one question to a finished part: does it hold what it carries where the metal and the joints say it should. On a welded tank, a catch can, a charge pipe or a fabricated exhaust section, the test looks past how the part appears and loads it the way service will, so that a pinhole, an unfused joint or a badly seated fitting reveals itself before the part leaves the bench.
A part can look flawless and still fail this step. A weld with an internal void, a crack that closed as the metal cooled, or a threaded boss that was never fully seated are all invisible to the eye and obvious to a gauge. That is the whole reason the check exists rather than being folded into a visual inspection.
Air is compressible, so a charged part stores energy and even a small leak escapes quickly. That makes air convenient, and it is why leak location with soapy water or a trace of smoke works so well. It also means a failure releases that stored energy, so the pressure used and the way the part is restrained have to be sensible.
Water is very nearly incompressible, so a small loss of volume shows as an unmistakable drop at the gauge. Liquid testing suits parts built to hold fluid, such as a welded tank or reservoir. A part tested wet then has to be dried properly, particularly where moisture could sit inside a seam or behind a bracket on stainless or aluminium.
Welds fail at the ends of a run before they fail anywhere else. The start and stop of a weld, the point where a tube meets a flange, and the narrow heat-affected zone beside the bead are where a defect most often hides. A leak there is small at first and grows as the part is used.
Joints come second: a threaded fitting, a folded seam, a spigot meant to take a hose, or a drain boss welded into a flat wall. A bracket that pulled on the parent metal, or a wall thinned by grinding, is also worth looking at before the gauge is connected.
A hold test compares the gauge at the start of the period with the reading at the end of it. A slow decay is not automatically a leak in the part. A weeping valve, a gauge that crept with temperature, or a seal compressed into a fitting can lose pressure with no defect present at all.
Conditions matter for the same reason. A sealed part standing in sun warms and the reading rises; the same part cooling in a workshop drops. Recording the medium and the conditions turns a test into a result someone else can interpret, rather than an impression only the person holding the gauge remembers.
It says nothing about how a part behaves under vibration, heat cycling or repeated load. A part that passes cold can still crack in service if its support is wrong, if it is bolted to something that moves, or if the wall is thinner than the duty demands. Pressure proves soundness at one moment.
It also does not confirm that the design suits the application. Wall thickness, internal flow, fitment on the car and clearance to the parts around it are separate questions with separate answers. A passing pressure result should be read as one line in a set of checks, not as the conclusion of them.
On custom work there is no service history to lean on, so testing carries more weight than it does on a catalogue part. The sequence follows the same path every time: negotiation, then a mould, then samples, then measurement before production. Knowing how that path runs helps you judge what has been checked and when, which is set out on the custom work page.
For one-off pieces such as a welded tank, a bracket set or a pipe run, a pressure check and a careful look at the welds catch nearly everything this style of fabrication produces. That pairing is standard practice for the titanium fab components side of the work, where parts are assembled by hand and each joint is made once.
