Toward improvement of sampling-based seismic probabilistic safety assessment method for nuclear facilities using composite distribution and adaptive discretization

As a seismic probabilistic safety assessment (SPSA) method for nuclear facilities, direct quantification of fault tree using the Monte Carlo simulation (DQFM) was developed to accurately consider the partial dependency between components. However, since this is a sampling-based method, there is a disadvantage in that a large number of samples must be extracted for accurate seismic risk estimation. Accordingly, this study develops an efficient SPSA method by improving the existing DQFM method. We replace the method of extracting samples from both seismic response and capacity at each component by that of taking samples only from a single response distribution with a composite deviation. Also, a method of adaptive discretization for seismic intensity (ADSI) is devised by linking the seismic intensity subdivision with the convergence of the final seismic risk. The Monte-Carlo sampling technique is replaced by the Latin hypercube sampling. As an application result to nuclear facilities, the proposed method requires only half samples for every seismic intensity than the existing DQFM method, while the accuracy of results was almost similar. Besides, through the ADSI method, the proposed method was able to secure approximately three times efficiency more than the existing DQFM method, without losing the accuracy of the results.

Automated summary

A new efficient seismic probabilistic safety assessment method was developed by improving the existing DQFM method, reducing sample size and improving accuracy of results. Through the ADSI method, the proposed method achieved three times efficiency improvement compared to the existing DQFM method.