Brittle fracture: the one damage mechanism you cannot inspect for

Corrosion gives you years. Creep gives you a bulge. Fatigue gives you a crack that grows at a rate you can measure. Brittle fracture gives you nothing at all — a component that passed every thickness survey and every visual splits from end to end in the time it takes the pressure wave to travel along it.

Because there is nothing to find, Part 3 of API 579 is not built like the other parts. It does not ask how much metal is left. It asks a different question, and it is a yes or no: is this component ever cold enough, and stressed enough, to fracture in a brittle manner?

The three things that have to coincide

Brittle fracture needs all three at the same moment, and removing any one of them prevents it:

• Low toughness — the steel is below its transition temperature, so it fails by cleavage instead of tearing. Toughness falls with temperature, rises with better steelmaking, and falls again with thickness, because a thick section constrains the metal at a crack tip and stops it deforming its way out of trouble.
• Enough stress — usually pressure, but residual stress from welding counts too, and in an as-welded joint that residual stress can sit at yield magnitude before you add anything.
• A flaw to start from — a weld defect, a crack, a sharp notch, a gouge. It does not need to be large.

CET and MAT: the two temperatures

The assessment turns on comparing two numbers.

The critical exposure temperature (CET) is the lowest metal temperature the component actually experiences while a significant stress is on it. Note both halves of that sentence. A vessel that goes to −30 °C empty and unpressurised is not at its CET; a vessel at −30 °C with pressure in it is.

The minimum allowable temperature (MAT) is the lowest metal temperature at which the material is permitted to carry that stress. It comes from the material, its product form, its heat treatment and — critically — its thickness.

The acceptance criterion is as blunt as it sounds: CET must be at or above MAT.

Where MAT comes from

For a Level 1 assessment you read it off an exemption curve. Carbon and low-alloy steels are sorted into four curves, conventionally lettered A to D, by how the steel was made and treated. Curve A holds the least tough materials — older as-rolled plate, unknown material. Curve D holds normalised, killed, fine-grain steels. Each curve plots exemption temperature against governing thickness, and every one of them slopes the wrong way: the thicker the section, the warmer it has to be.

The number that surprises people the first time they see it is how warm Curve A gets. A 25 mm as-rolled plate on the least tough curve lands somewhere near room temperature. Not below freezing — around +20 °C. Read the curve for the material in front of you rather than trusting that figure, but understand what it implies: an old thick vessel of unknown pedigree may have no margin at all on a cool morning.

Buying margin with stress

Because stress is one of the three ingredients, lowering it earns you temperature. If the applied stress is a small fraction of what the material is allowed, the assessment permits the MAT to be reduced — the steel does not need to be tough if there is not enough stress to drive a crack. The pressure-vessel codes carry the same logic; below a stress ratio of roughly a third, a substantial reduction becomes available without any impact testing at all.

This is the reason a de-rate is a legitimate answer to a brittle fracture problem. You are not patching the steel. You are removing one of the three ingredients permanently.

When Level 1 says no

Level 2 refines the same picture — more credit for stress reduction, for PWHT, for the actual loading history. Level 3 abandons exemption curves entirely and does fracture mechanics: you need a real toughness value, from Charpy data or a CTOD test, and a real flaw size. At that point the problem has stopped being a Part 3 problem and become a Part 9 one, and the whole answer rests on how well somebody sized a crack.

What to take onto the plant

• Ask what the CET actually is, not what the design minimum temperature says. Start-ups, depressurisation and auto-refrigeration produce metal temperatures nobody wrote down.
• Thick and old is the exposed combination. Thin modern normalised plate is rarely the problem; a 60 mm shell of unknown origin is.
• Treat every weld flaw in a low-MAT component as more serious than its size suggests. In a tough material it is a defect to monitor. Here it is an initiation site.
• Put a minimum water temperature in the hydrotest procedure and refuse to start below it.

The uncomfortable truth of this mechanism is that a component can pass a full inspection, with every reading in tolerance and no relevant indications anywhere, and still be one cold start-up away from failure. Nothing you measure will tell you. Only the temperature will.