Numerical model based approach for estimating probability of detection in NDE applications

A method for using a numerical model in a probabilistic framework for estimating the probability of detecting a flaw in the presence of measurement variabilities is presented. The approach used centers around a measurement model which is based on the finite-element analysis of the eddy current nondestructive testing (NDT) phenomenon. The probability of detection (POD) model consists of modeling the different sources of measurement variabilities by a multivariate Gaussian distribution about a mean vector and a diagonal covariance matrix. The effects of various sources of noise in a practical eddy current NDT situation have been studied. Results obtained show that the probability of detecting a flaw is decreased as a consequence of the measurement variabilities.<>