Assessment of Components of Variance in NDE Data

There is an important need to quantify and improve the probability of detection (POD) for NDE inspection. To improve POD, it is important to identify and quantity sources of variability. A standard NDE inspection assessment method uses a specimen containing seeded flaws of known size and character. The specimen is inspected according to an experimental design that will capture the important sources of variability. The commonly used NDE data analysis/modeling method, known as â versus a, uses a linear regression to relate the NDE signal response to the flaw size. The model behind this method contains only one component of variance for the response. There are, however, many random factors causing variability in NDE inspection. In this paper, we use a Bayesian hierarchical model to identify and quantify the inspection variance components in the presence of data censoring. We use Markov Chain Monte Carlo simulation to estimate the model parameters, including the variance components. We demonstrate the effecti...