Abstract
A method is developed for experimental assessment of reliability of a system with a stringent safety requirement: explosive material. The focus is on analysis and management of both statistical variability of measurements and non-probabilistic uncertainty in probability distributions (distributional uncertainty). Info-gap theory is used to model the distributional uncertainty in the pdf of the threshold for actuation of the explosive material. The quantitative analysis and the qualitative judgments which accompany the certification of safety are studied. A proposition is proven asserting that the info-gap robustness function, for the class of problems examined, is independent of the experimental design over virtually all of its range.