Uncertain reliability of structural design standards

Abstract

A limitation of current reliability-based design Standards is their dependence on imprecise estimates of the probability of failure. These imprecise estimates do not, by themselves, provide a dependable basis for explicit risk-based decision-making. Therefore reliability-based Standards depend on Code calibration to determine acceptable levels of the estimated nominal probabilities of failure, which hinders further rationalisation and innovation. The paper discusses uncertain failure probabilities and defines a related characteristic failure probability that provides a dependable basis for risk-based decision-making. Uncertain probability estimates arise from epistemic uncertainty and they can be described by their own probability distributions. The paper discusses probability distributions of uncertain probabilities of failure, with examples of uncertainty arising from the sampling variability for design based on prototype testing in accordance with provisions in Australian Standards and an AISI Standard, and also sampling variability in the estimation of a design wind load based on wind speed data. It is shown that the characteristic failure probability provides a dependable basis for assessing the acceptability of extremely uncertain probabilities of structural failure.