ABSTRACT
This study investigated the treatment of uncertainty in quantitative offshore fire risk assessment. We investigated the main sources of the limitations of the existing Quantitative Risk Assessment (QRA) procedure-assumptions and simplifications-as well as the employed method of selecting representative scenarios and the use of probabilistic and statistical approaches. We found that QRA was highly governed by two types of uncertainties: parameter uncertainty and completeness uncertainty. Accordingly, the QRA of fire accidents during low-frequency, high-consequence disasters posed a challenge to the selection of representative scenarios and frequency analysis due to the randomness of these accidents and a lack of statistical data. To treat these uncertainties, a sampling technique was proposed that would enable the analyst not only to select representative scenarios for frequency and consequence analysis but also to estimate uncertainties simultaneously. As a result, a distribution map of frequency uncertainty was obtained for credible scenarios. This map of uncertainty indicated the inaccuracy and lack of credibility of previous risk-informed decision-making efforts, which had used only one line diagram of an exceedance curve or an FN curve, a technique frequently employed in fire safety engineering.
