On when to base event trees and fault trees on probability models and frequentist probabilities in quantitative risk assessments

Abstract

This paper discusses the analysis approach when using event trees and fault trees in a quantitative risk assessment context. The basic question raised is when to introduce probability models and frequentist probabilities (chances) instead of using direct probability assignments for the events of the trees. We argue that such models should only be used if the key quantities of interest of the risk assessment are frequentist probabilities and when systematic information updating is important for meeting the aim of the analysis. An example of an event tree related to the analysis of an LNG (Liquefied Natural Gas) plant illustrates the analysis and discussion.