Wall Penetration and Remaining Life

A corrosion rate is a number about metal. A remaining life is a number about a decision, and this division is where one becomes the other. Take the wall you have, subtract the wall you are required to keep, and divide what is left by the rate at which it is disappearing.

The retirement thickness is not zero and it is not a matter of taste. It is the wall the component needs to hold its pressure, its bending load and its own weight with the code's margin intact, and it is calculated from the design conditions before any corrosion is considered. Everything above it is the corrosion allowance — metal deliberately bought so that it can be lost. A component at its retirement thickness has not failed; it has run out of the margin it was sold with, and it goes to repair, replacement, or a formal fitness-for-service assessment that may well let it keep running at a reduced rating.

The unit trap here is the one that costs the most money in this whole subject. Metric practice quotes penetration in millimetres per year; North American corrosion practice quotes it in mils per year, a mil being one thousandth of an inch. The two differ by a factor of 25.4, and 0.25 mm/yr and 25 mpy are so close in appearance and so far apart in meaning that reading one as the other is a routine error. A steel that is "corroding at 5" is either comfortably fine at 5 mpy or in real trouble at 5 mm/yr, and nothing in the sentence tells you which.

Now the honest caveat, which matters more than the arithmetic. This calculation assumes the metal is leaving evenly. Where it is not — pitting, crevice attack, under-deposit corrosion, microbially influenced corrosion, weld-line attack — the average is worthless and can be reassuring in exactly the situation that should alarm you. A tank that loses 0.05 mm a year on average and carries pits growing at 2 mm a year will leak in five years while the calculation promises fifty. Localised attack is an inspection problem, not a computation: the number you need is the depth of the deepest pit found, and finding it takes coverage rather than cleverness.

Finally, treat a rate derived from two thickness readings with suspicion in proportion to how close together they were taken. An ultrasonic thickness gauge is good to perhaps ±0.1 mm, so two readings a year apart on a wall losing 0.05 mm a year are measuring their own uncertainty four times over. Long baselines, repeatable measurement locations and three or more readings are what make a trend line mean something; two points always make a perfect straight line, whatever they are actually telling you.