The inspection codes compute corrosion two ways, and the difference is not academic. Long-term rate divides total metal loss by total service time: the initial or earliest reliable thickness minus the current thickness, over the years between them. Short-term rate does the same arithmetic on the last two readings only. Remaining life and the next interval are then evaluated against whichever rate produces the more conservative result. Every audit of a thickness program eventually lands on this calculation, so it is worth understanding what each rate actually measures and how both get fooled.
What each rate is for
The long-term rate is the smoother. It averages across process campaigns, chemistry changes, and measurement error, and it describes what the equipment has done over its life. Its weakness is inertia: a corrosion regime that changed two years ago is diluted by fifteen years of history, and the long-term number can stay comfortable long after conditions turned aggressive.
The short-term rate is the alarm. It responds to the most recent interval, which is exactly where a process change, a chemistry upset, or a new mechanism first shows up. Its weakness is noise: it is built from two readings, and two readings carry all the error of gauge calibration, temperature correction, surface condition, and probe placement, divided by a short time base.
The codes' answer to this trade is to compute both and let the worse one govern. That rule is simple to state and constantly mishandled in practice, usually by programs that quietly default to the long-term rate because it is calmer.
The noise problem, concretely
Ultrasonic thickness measurement has real scatter. When the interval between surveys is short and the true metal loss over that interval is comparable to the measurement scatter, the short-term rate is mostly measuring the gauge, not the corrosion. Two symptoms give this away:
- Negative rates. Metal does not grow. A reading thicker than its predecessor means error somewhere: placement, calibration, temperature, or a broken pairing between reading and location.
- Rates that flip violently between surveys. A location that reads high rate, then low, then high again on a stable process is showing measurement variation, not corrosion variation.
The disciplined responses are procedural, not mathematical: fix CML identity so consecutive readings truly share a location, control gauge calibration and temperature correction, and where short intervals matter, take multiple readings per location so scatter can be seen rather than guessed at.
What the record must support
An auditor testing a rate calculation walks a specific path: which two readings produced the short-term rate, which readings anchored the long-term rate, which rate governed, and how it flowed into remaining life and the interval. That walk requires reading-level traceability with dates, and it requires the calculation to run from the stored history rather than from a value someone once copied into a spreadsheet.
The classic failure is the frozen rate: a corrosion rate computed years ago, carried forward through successive plans while new readings accumulated unexamined. The program has the data to know better and a record showing it did not look. That is a worse audit position than having sparse data.
Rates are also decisions. Excluding an outlier reading, overriding a noisy short-term rate with an engineering judgment, adopting a similar-service rate for a new circuit: all legitimate, all needing a name, a date, and a reason in the file.
intelliSPEC computes both rates from the stored reading history and keeps the governing choice traceable, so the math an auditor checks is the math the program actually ran. Bring a noisy circuit to a demo and test it.
