ABSTRACT
Networks (BBNs) (Jensen & Nielsen, 2007) are increasingly being explored, developed and applied for HRA. Use of BBNs in risk and reliability analysis is relatively established, while their use for HRA is quite recent (mostly from the last five years). In addition to the features already mentioned, BBNs are especially suitable for HRA for the following reasons. First, BBNs handle both causal and evidential reasoning: this allows not only the estimation of Human Error Probabilities (HEPs) given the contributing factors but also the analysis of the most dominant factors given a human failure. Second, BBNs can incorporate information of different nature from different types of evidence sources-a key feature in case comprehensive data is not available. Third, because of their straightforward graphical representation of cause-effect relationships, BBNs are easy to use for both domain experts and analysts, and the results are easy to interpret for non-experts. Finally, the probabilistic nature of BBNs allows them to be integrated with other mathematical models such as fault trees and event trees; therefore the results can be used in Probabilistic Safety Assessment (PSA).
