Review of statistical methods for assessing the probability of detecting defects in non-destructive testing

Abstract

The probability of detection (POD) depends on defects size and is an integral part of calculating the resource during non-destructive testing of parts. This article provides an overview of well-established statistical methods for estimating PODs, with a little historical insight into their emergence. An overview of new advances in POD calculation in recent years is given: three- and four-parameter models; nonparametric models; planning the experiment and sampling of defects; applying defect modeling to reduce the number of samples; the application of the Box-Cox transformation; the influence of the variability of the initial data on the result; application of Bayesian statistics. An overview of the tasks that POD specialists still have to solve in the future: nonlinear models, modeling in conjunction with Bayesian statistics, etc