Uncertainties that influence the estimated probability of failure (PoF) are of two types: uncertainty due to inherent randomness of the physical properties and measurement error (termed ‘aleatory uncertainty’), and uncertainty due to lack of knowledge or simplification of physical processes represented in the empirical models (termed ‘epistemic uncertainty’). The objective of this study is to demonstrate the effects of epistemic uncertainty, represented as model error related to 1) reliability-based decision making for single features; 2) development basis of model error distributions; and 3) system wide reliability assessment.

In this study, the effect of model error is demonstrated with two approaches. First, the model error is estimated for various subsets of population features to demonstrate the effect of sub-populations on the model error distributions. Second, the epistemic uncertainty due to model error is separated in the estimation of failure probability. The effect of model error is represented by estimating the failure probability as an interval instead of a point estimate. The effect of using the range of failure probability, instead of a point estimate, on the reliability-based integrity management of both individual features and the system-wide decision-making is demonstrated.

This content is only available via PDF.
You do not currently have access to this content.