Beyond the single predictor variable POD model

摘要

A single predictor variable model is typically used to introduce the underlying regression methods behind the hit/miss and a versus a Probability of Detection (POD) models. The typical example assumes that POD is related to a single characteristic of the indication, such as length. Often, though, it is more than one characteristic of an indication that affects POD. For example, POD for a Fluorescent Penetrant system is affected by indication length as well as depth and width. Hence, if possible, a model with multiple predictor variables is more appropriate and will result in a more accurate estimate of the capability of a non-destructive testing (NDT) system. The hit/miss POD model will be used to show that the single variable model can easily be extended to more than one predictor variable. In addition, a hit/miss POD analysis will be performed on simulated data, the results of which will be used to illustrate how the analysis yields an NDT system capability range rather than a single capability value when more than one predictor variable is included in the POD model.