Notes on Computational Uncertainties in Probabilistic Risk/Safety Assessment

Abstract

In this article, we study computational uncertainties in probabilistic risk/safety assessment resulting from the computational complexity of calculations of risk indicators. We argue that the risk analyst faces the fundamental epistemic and aleatory uncertainties of risk assessment with a bounded calculation capacity, and that this bounded capacity over-determines both the design of models and the decisions that can be made from models. We sketch a taxonomy of modelling technologies and recall the main computational complexity results. Then, based on a review of state of the art assessment algorithms for fault trees and event trees, we make some methodological proposals aiming at drawing conceptual and practical consequences of bounded calculability.