Seismic risk quantification for multi-unit probabilistic safety assessment based on the partial binary decision diagram approach

• Two approaches are proposed for risk quantification of seismic multi-unit PSA. • The approaches can generate minimal cut sets (MCSs) as well as near-exact risk. • Fault tree approach has limitations for PSA models with a large number of units. • MCS post-processing approach is applicable to PSA models with a large number of units. • The obtained near-exact risks are compared with those from the Monte Carlo approach. Many efforts have been made to quantitatively evaluate seismic risk in multi-unit probabilistic safety assessment (PSA) models in light of the fact that the success units cannot be neglected due to the seismic events with large failure probabilities. Previously, a partial binary decision diagram (BDD) approach was shown to generate minimal cut sets (MCSs) with near-exact single-unit risk. To extend this approach to multi-unit risk, this study explores two approaches: one based on fault tree modeling, and one based on the post-processing of MCSs. The first approach reflects the fault trees of success units by converting into BDD logic for risk-significant cut sets of single-unit risk. A multi-unit PSA model with two units was evaluated, and the results showed near-exact risk. However, this approach can produce inaccurate risk based on the cut-off value and MCS information that is difficult to understand for a multi-unit PSA model with many units. The proposed post-processing approach regards the MCSs of success units as different super events according to what shared events are included in each cut set. To implement this approach, a software tool was developed, with the results of a multi-unit site having six units showing near-exact risk along with the MCSs.