Pressure decay leak testing: how it works, and how to size the test for your part

The part is sealed, filled with air to the test pressure and then isolated from the supply. From that moment the air inside is a fixed amount in a fixed volume. If air escapes through a leak, the pressure falls; if nothing escapes, it holds. The machine measures how far the pressure falls in a fixed test time and compares it with a limit.

• Fill — bring the part to test pressure. If it cannot get there in time, it is a gross leak: no point testing further.
• Stabilise — let the air settle. Compressing air warms it; as it cools back to the part's temperature, pressure falls even in a perfect part. Seals bed in and thin walls stretch in the same seconds.
• Test — isolate and measure the drop over a fixed time.
• Exhaust — vent the part before it is unclamped.

A leak rate in standard cubic centimetres per minute (sccm) is a volume of air at atmospheric pressure leaving per minute. Over the test time, that air leaving the trapped volume produces a pressure drop of:

Worked example. The limit is 2 sccm, so Q = 2 ÷ 60 = 0.0333 cm³/s. The part and its test circuit hold 150 cm³ and the test time is 10 s:

So a part leaking exactly at the limit loses about 2.25 mbar in 10 seconds. That is the number the machine has to see reliably — at a test pressure that may be 3 bar, or 3000 mbar. The same limit gives a very different drop as volume and time change:

Two lessons follow. Volume is the enemy: every cm³ of dead volume in the fixture and tubing shrinks the signal, so good leak fixtures fill empty cavities with blocks and keep the valve close to the part. Time buys signal, but it is paid for in cycle time — which is why the test time is a design decision, not a setting to guess at commissioning.

Trapped air changes pressure with temperature as well as with leaks. At a test pressure of 3 bar gauge (about 4013 mbar absolute) and room temperature (293 K), the change is:

In the example above the limit was 2.25 mbar. If the air inside cools by just 0.2 °C during the test, the pressure falls by about 2.7 mbar — more than the leak limit, from a part that does not leak at all. That is why:

• the stabilise phase exists, and cutting it to save cycle time causes false rejects;
• parts that arrive warm from washing or machining test differently from parts at room temperature;
• a machine that passes everything in the morning and rejects good parts after lunch usually has a temperature problem, not a leak problem;
• a long test time does not help if the temperature is still moving — it just measures the drift for longer.

Because decay is a difference between two readings of the same sensor, a fixed offset in the transmitter cancels out. What does not cancel is noise and drift. The useful question is not “how accurate is the transmitter?” but “how small a change can it show, steadily, as installed on this machine?” — and that needs to be several times smaller than the drop at your limit.

• Drops of tens of mbar: a good gauge-pressure transmitter per station, read by the PLC, is usually enough.
• Drops of a few mbar: careful transmitter selection, filtering and a stable fixture; test the noise on the real machine.
• Drops below that, or tight limits on large volumes: a dedicated leak test instrument with differential measurement and its own compensation.

Before judging the drop, a good machine checks the fill: pressure too low at the end of fill means a gross leak or a missing seal; pressure outside its window means a supply problem or a blocked port. Those parts are rejected with their own reason instead of being given a misleading leak value.

Finally, a leak tester that has never been shown a bad part has not been proven. Keep a master good part and a master part fitted with a calibrated reference leak at the limit. The good one must pass and the leaking one must fail, every time — at commissioning, and as a routine check.

• The allowed leak rate, in sccm, and the test pressure it applies at.
• The part drawing, so the internal volume and the sealing faces can be worked out.
• Parts per shift, or the cycle time you need.
• Whether parts arrive warm — from washing, machining or welding.
• Whether you need to know only that a part leaks, or also where (a water dip).

See how we build these machines on the leak testing machines page.