Towards robust estimation of minimum thickness from sample data

carbon steel pressure systems subject to corrosion. As this approach obtains data from a sample of the overall surface area, it introduces sampling uncertainty to the measured minimum thickness values. This and other epistemic uncertainties associated with estimated minimum thickness values should be understood when performing safety-critical condition and remaining useful life assessments. Statistical methods can be used to estimate the true minimum thickness by extrapolating from the measured part of the thickness distribution into the unmeasured lower tail. This paper gives an example application of a method based on Extreme Value Theory (EVT). The true minimum thickness of a piping loop is estimated from minimum measured values from 84 thickness monitoring locations. Limitations of the approach are discussed, including that extreme value methods rely on assumptions that may not be appropriate for the data, and that other forms of epistemic uncertainty are not considered. The paper concludes by suggesting more avenues for more robust minimum thickness estimation, including distribution-free quantile estimation methods that rely on fewer assumptions, and the use of imprecise probability methods to represent uncertainty.