Evaluation of Common Cause Failure by an Initiating Event for Multi-unit using Bayesian Belief Network

Abstract

Focused attention about multi-unit probabilistic safety assessment (PSA) after the Fukushima Daiichi accident in March 2011 has been maintained and methodology of PSA in the multi-unit site has been developing in many countries. In this study, the probability which represents that if one unit has core damage then the other one unit in the same site also has core damage is estimated using Bayesian Belief Network (BBN). For the simplicity and optimization of the developed model, only common cause failures (CCFs) of the components in a single unit PSA model are considered. Since different mitigation systems are considered depending on initiating events, a different correlation factor between units needs to be identified according to the given initiating events. In the step of establishing the model, the top node of the model is defined as the correlation factor of inter-unit in the site which represent the probability of core damage on two or more units by an initiating event. The leaf nodes are component level CCFs related to the initiating event. The relations of component CCFs to system CCFs and system CCFs to the inter-unit correlation factor are represented in the node probability tables (NPTs) of the BBN model. In this work, a method was proposed to estimate the quantitative correlation representing the probability of core damage in two units at a site for a specific initiating event. Since only CCFs are considered in the proposed model, the size of model is more compact than that of the current PSA model and it is able to illustrate the correlation more intuitively. Furthermore, as taking advantage of the BBN model, components or systems which have a relatively great impact on the correlation can be identified and complemented to improve the safety of the site.