Seismic analyses within a Probabilistic Safety Assessment (PSA) differ from typical internal event analyses in several important ways. One key distinction arises from the nature of strong seismic events. Earthquakes with high peak ground acceleration occur infrequently, but they likely cause the simultaneous unavailability of many components. As a result, the minimal cut set list associated with such events is characterized by a low initiating event frequency and a high conditional core damage probability (CCDP). A conservative approach sets the CCDP to one, effectively equating the result to the initiating event frequency. However, this can be overly pessimistic and prevents meaningful comparison of the relative importance of safety measures for high-intensity seismic scenarios.
A more realistic estimate can be obtained by quantifying the minimal cut set list using a Binary Decision Diagram (BDD)–based method. The algorithm implemented in RiskSpectrum can process large minimal cut set lists through multiple heuristics and approximations. One such heuristic divides the list into two parts: a dominant subset representing nearly the entire top value, which is quantified using the BDD, and a remaining subset with insignificant contribution to the top value, quantified using the Min-Cut Upper Bound method. For internal events, the BDD-treated portion is typically much smaller than the remainder. Seismic analyses, however, challenge this assumption. If we limit the number of cut sets allocated to the BDD, the CCDP of the remaining part often remains high. This issue becomes particularly disadvantageous when both subsets contain mutually exclusive events, requiring their high CCDP contributions to be summed.
We present a new heuristic for the BDD-based quantification algorithm based on a new way to create modules that mitigates this problem and enables more realistic quantification of high-intensity seismic events.