An analytical review of perforation statistics – using aluminium as witness plates in perforation experiments

This report documents the statistical methods that are used in perforation experiments at the Norwegian Defence Research Establishment (FFI). We also introduce new (or at least unused) statistical methods, which have certain advantages over the ones we currently employ.

A concrete example is the receiver operating characteristic curves (ROC curves) and corresponding classification tables, which provide a much more direct method to evaluate the goodness of fit compared to the more conventional pseudo-𝑅² measures. The ROC curves take multiple additional details into account (such as true positives and negatives and false positives and negatives) that 𝑅² measures do not when evaluating model performance. It is therefore a more reliable goodness of fit metric. We recommend using both 𝑅² and ROC methods to supplement each other.

As an example of the application of the various statistical methods, we evaluate whether a 0.5 mm aluminium plate is suitable as a witness plate in perforation experiments. We find that the aluminium plate is suitable as a witness plate for shots against skin simulants modelling the abdomen and back. For other target locations, such as the thorax, thigh, or buttocks, we find that it is necessary to use plates that are more easily perforated. Both the 𝑅² measures and the ROC curves indicate that the experimental model has a high goodness of fit.