Calculating corrosion rate: short term, long term, and which one governs

Two rates are calculated from the same thickness history, and they answer different questions.

The convention in API 510 and 570 is to calculate both and use the one that gives the shorter remaining life — unless there is a documented engineering reason to do otherwise.

Why they disagree, and what it means

Short term much higher than long term

Something has changed. A process upset, a feed change, a chemical injection that stopped working, a coating that failed, water in a line that used to be dry. This is the situation the convention is designed to catch, and it should trigger investigation rather than just a shorter interval.

Short term much lower than long term

Either the damage has genuinely slowed — corrosion inhibitor working, feed cleaned up, temperature reduced — or, much more often, the difference is measurement noise. Be sceptical before you extend an interval on the strength of it.

A negative short-term rate

The wall appears to have got thicker. Since that does not happen, it tells you the measurement uncertainty is comparable to the corrosion over that interval. It is a useful diagnostic: it means you cannot resolve the rate at this interval, not that corrosion has stopped.

The noise problem, with numbers

A UT thickness reading is typically quoted to ±0.1 mm under good conditions. Two readings differenced carry the error of both.

Suppose a wall genuinely loses 0.3 mm over three years — a real rate of 0.10 mm/yr. Two readings, each ±0.1 mm, could measure that loss as anything from 0.1 mm to 0.5 mm.

A remaining life calculated from the low end is three times longer than one calculated from the high end, on identical equipment. And that is before accounting for the probe not landing on precisely the same spot, surface condition changing, or a different technician.

Which reading do you use?

A grid gives several readings at one location. Three defensible choices, and they are not equivalent:

• The minimum at each inspection. Conservative, but it may be a different point each time, so the "rate" is partly the difference between two locations rather than change over time.
• The same point each inspection. This is the one that actually measures a rate, and it requires the points to be marked and recorded properly — which is the real argument for permanent CML marking.
• The average of the grid. Smooths noise, and is appropriate for genuinely general corrosion. Meaningless if the damage is localised.

The failure mode to avoid is mixing them: minimum this time, average last time, gives a number that means nothing at all.

What the rate is used for

Remaining life = (current thickness − minimum required thickness) ÷ corrosion rate.

And the inspection interval follows from it — conventionally the lesser of half the remaining life or a code maximum. Halving it is the margin that covers everything above: that the rate is uncertain, that it may accelerate, and that a single measurement may not have found the worst point.

Where it feeds fitness-for-service

Every API 579 assessment includes a remaining life step, and it uses this rate. Three points worth carrying:

• The future corrosion allowance in a MAWP calculation is the rate multiplied by the intended interval. Get the rate wrong and the MAWP is wrong.
• An accepted flaw keeps growing if the mechanism is active. The assessment is valid at the thickness assessed, for as long as the remaining life calculation says — not indefinitely.
• Localised mechanisms have localised rates. A rate averaged over a vessel says nothing about the patch that is going three times faster. The rate that sets your interval should come from the worst location, not the mean.