Any condition that could take the system above its safe working pressure - most often a failed regulator or a trapped volume.

The commonest scenario is a regulator failing open, so a downstream system designed for a lower pressure sees full supply pressure. Without relief, whatever is weakest downstream fails.

Other cases are thermal: a volume of air isolated between two closed valves and then heated - by sunlight, by a nearby process, or simply by the compression that put it there - rises in pressure with nowhere to go.

On a compressor and receiver, relief protects against a failed pressure switch allowing the compressor to run on past its cut-out.

The valve is sized so it can pass the full flow that could be delivered into the system, not merely crack open - a relief valve that lifts but cannot pass enough flow does not prevent the pressure continuing to rise, which is a sizing error that looks like protection.

Because the two loads act in opposite directions and a thin shell is far weaker in external compression than in internal tension.

Internal pressure puts a cylindrical or spherical shell into tension. Steel is strong in tension, and the shell simply carries the load - which is why a vessel can hold many times atmospheric pressure inside.

External pressure, which is what a vacuum inside produces, puts the shell into compression. A thin shell in compression does not fail by crushing; it fails by buckling, and buckling occurs at a far lower load than the material's strength suggests. A small dent, an out-of-round section or a weld distortion reduces it further.

So a tank rated for substantial internal pressure can be collapsed by a vacuum well under one atmosphere.

That is why vacuum breakers are fitted to tanks that are drained, steamed, or cooled after being heated - all of which create vacuum quietly, with no alarm and no visible warning until the vessel deforms.

By the spring loading, set at or below the protected system's maximum allowable pressure - and on statutory applications it is set and sealed rather than adjusted in service.

The valve opens when system pressure produces enough force on the disc to overcome the spring. Adjusting the spring compression changes the set pressure.

The set point must be at or below the maximum allowable working pressure of whatever it protects, with allowance for the fact that a relief valve does not open fully at its set pressure - it needs some overpressure to reach full flow, and that accumulation has to fit within the equipment's margin.

On pressure vessels and other statutory applications, relief valves are set by a competent person, tested, and sealed so the setting cannot be altered without breaking the seal. Adjusting one in service is not a maintenance task.

On general pneumatic applications adjustment is more routine, but the setting should still be recorded and verified rather than turned by feel.

Relief valves wherever overpressure is possible and the equipment cannot tolerate it; vacuum breakers on any vessel that can develop internal vacuum.

For relief: on air receivers, downstream of regulators feeding lower-rated equipment, on any volume that can be isolated while a source is still connected, and on trapped volumes that could be heated. The valve must be on the protected side, with no isolating valve between it and what it protects - an isolation valve upstream of a relief valve defeats it entirely.

For vacuum: on tanks that are drained or pumped out, on vessels that are steamed and then cool, on systems where a liquid can siphon, and on any closed vessel whose contents can be removed faster than air can enter.

Both should be positioned where their discharge or intake is unobstructed and safe - a relief valve venting into a confined space or towards an operator position is a hazard in its own right.

By lifting them and confirming they operate at the correct pressure, at an interval set by the application and by any statutory regime.

The failure mode that matters is a valve that has corroded, gummed or been painted shut. It gives no indication at all - the system looks normal, and the protection is simply absent until the day it is needed.

Many relief valves have a manual lift lever or ring specifically so they can be operated periodically to confirm they are free. That proves the valve is not stuck; it does not prove the set pressure, which requires a test against a calibrated source.

On statutory pressure systems the testing regime, interval and competence requirements are prescribed and documented, and the valve is typically removed for bench testing or replaced with a tested unit.

On general pneumatic applications, include a periodic manual lift in the maintenance schedule - it takes seconds and it is the only check most such valves ever get.

The valve must pass the full flow that could enter the system, at an acceptable accumulation above the set pressure.

Sizing is not about the connection thread. The question is: if the worst-case failure occurs - a regulator fails wide open, a compressor runs on - how much air enters the system, and can the relief valve discharge that much without the pressure continuing to climb?

A relief valve begins to open at its set pressure but needs a further rise to reach full lift. That accumulation must stay within the protected equipment's margin.

So the sizing calculation uses the worst-case inflow, the set pressure and the permitted accumulation - and an undersized valve produces the worst outcome: it opens, discharges audibly, and the pressure rises anyway, while everyone believes the system is protected.

For vacuum breakers the equivalent question is how fast vacuum can form, which depends on the drain or pump-out rate, and the breaker must admit air at least that fast.

Combined pressure and vacuum relief valves exist and are widely used on storage tanks, where both conditions can occur.

A combined unit has two elements in one body: a pressure pallet that lifts outward when internal pressure rises, and a vacuum pallet that lifts inward when internal pressure falls below atmospheric. Each is set independently.

They are standard on atmospheric and low-pressure storage tanks, where filling raises the pressure and emptying creates vacuum, and where temperature changes do both over the course of a day.

For compressed air work the two functions are more often separate, because the pressure relief duty is at a much higher pressure and the vacuum case arises in different equipment.

Where a combined valve is used, both settings must be verified and both elements tested - it is easy to check one and assume the other, and a vacuum pallet that has stuck is exactly as invisible as a stuck relief valve.