Probabilistic risk assessment considering parameter and model uncertainties

Probabilistic risk assessment (PRA) has become a widely used and accepted tool for managing risk in several industry sectors. The present paper focuses on parameter and model uncertainties within PRA that combine event trees and fault trees. To this end, a PRA is applied to a case study from oil and gas activities. Parameter uncertainties concern frequencies of initiating events, failure rates of safety barriers, factors of common cause failures, test coverages, and conditional probabilities. To perform these uncertainty analyses, a classical approach based on probability density functions and Monte Carlo simulations is used. On the other hand, model uncertainties concern effectiveness and architecture of safety barriers. To perform these uncertainty analyses, an approach based on fictitious events is proposed, which aims at transferring model uncertainties to parameter uncertainties. Resulting frequencies of occurrence of each accidental scenario are then assessed considering both parameter and model uncertainties. The impact of these results on risk management, notably in terms of risk acceptability, are then discussed, considering the selection of test ing policy (including proof and partial tests) as example.