Metal loss asks how much section is left. A crack asks something else entirely — will it run? The answer depends on two competing failure modes at once, which is why the assessment is a chart rather than a number, and why the depth your inspector reported matters more than anything else on it.
Hand a Part 4 assessment a patch of general thinning and it does something intuitive: it works out how much wall is left and asks whether the remaining wall can hold the pressure. Hand the same assessment a crack and it is useless, because a crack barely removes any metal at all. A 10 mm deep crack in a 20 mm wall takes away a sliver of cross-section and yet it can put the component in two pieces before the end of the shift.
What a crack does is concentrate. The stress at a sharp tip is not the stress in the wall, and the question is not how much section remains but whether the tip will advance.
A cracked component can fail by either of two completely different routes, and which one gets there first depends on the material and the load.
A single acceptance number cannot cover both, because the thing that makes a material safe from one does not help with the other. So the assessment plots them against each other.
The FAD is a graph with a curve on it and your component plotted as a single point. Inside the curve is acceptable. Outside is not.
| Axis | What it is | What it measures |
|---|---|---|
| Vertical, Kr | Applied stress intensity divided by material toughness | How close you are to brittle fracture |
| Horizontal, Lr | Reference stress divided by yield strength | How close you are to plastic collapse |
A point high on the left is a brittle problem: low stress, but the toughness cannot cope. A point far to the right along the bottom is a collapse problem: plenty of toughness, not enough remaining ligament. Most real assessments land somewhere in between, failing partly by each, which is exactly why neither simple check on its own is safe.
The curve is cut off vertically at a maximum Lr, past which the component is yielding wholesale regardless of toughness. That cut-off is conventionally set from the flow stress — the average of yield and tensile, expressed as a ratio to yield — so a material with a large gap between yield and tensile earns a little more room before the cut-off bites.
Four inputs, and they are not equally reliable.
K scales with the square root of crack depth. That sounds forgiving until you look at where the assessments actually sit. Report a crack as 4 mm when it is 6 mm and the stress intensity is understated by roughly a fifth — enough to move a point from inside the curve to outside it when the component was marginal, which is precisely the situation in which somebody bothered to do the assessment.
Detection is not the hard part of crack assessment. Height sizing is. A method that reliably says "there is a crack here" and vaguely says "about three or four millimetres deep" has not given you an input; it has given you a range that spans the answer.
Techniques that size crack height directly and repeatably — time of flight diffraction, and phased array run specifically for depth rather than for detection — produce inputs an assessment can use. Techniques that confirm a crack exists but not how deep it goes are the start of the job, not the end. Radiography deserves its own warning: a tight planar crack lying across the beam can be effectively invisible on film, so a clean radiograph is not evidence that there is no crack.
The last thing to understand is that Part 9 assesses the crack you have today. If the crack got there by a mechanism that is still running — stress corrosion cracking in a service that has not changed, fatigue on a machine that is still cycling, corrosion fatigue in a line still being cycled wet — then an acceptable result buys you nothing on its own. You need a growth law and a re-inspection interval, or you need to remove the driver.
An assessment that says "acceptable" for a crack in live SCC service, with no interval attached to it, is not a result. It is a date nobody has calculated.
Depth sizing is the input the whole assessment rests on. The detection methods list what each technique finds, what it misses, and which ones give height rather than just presence.
Open the detection methodsWho writes this. A mechanical engineer with twelve years in oil and gas — in-line inspection, fired heater and furnace inspection, and pipeline integrity. What is here comes from the published standards and from what those years in the field actually looked like. It is not written by an API-certified inspector.
This is not an assessment. Nothing on this site may be used to justify a decision about real equipment. Assessing plant requires the current editions of the applicable codes, data from a licensed source, and a competent engineer who signs for the answer. · Integrity Field Guide