The Implementation of Fault Tree Analysis Approaches in Nuclear Power Plant Probabilistic Safety Assessment

Probabilistic safety assessment (PSA) has been extensively implemented to assess the performance of nuclear power plant (NPP) safety systems. One well-known modeling approach in NPP PSA is a fault tree analysis (FTA). A fault tree is a graphical representation of possible failure scenarios of the system being evaluated. To estimate the top event failure probability, a quantitative analysis needs to be performed based on those scenarios. Prior to performing quantitative analysis, basic events' failure probabilities of the system fault tree need to be provided well in advance. Conventional FTA assumes that basic events always have precise probability distributions characterizing their lifetime to failure. However, in practical applications, this is not the case. For example, a new system will not have sufficient operating experiences to probabilistically estimate reliabilities of their components. To deal with this limitation, a number of approaches has been developed and proposed. Each approach offers advantageous but also has disadvantageous. Since the results of FTA will be used to verify NPP designs, it is necessary to select the most suitable approach. It is, therefore, essential to clearly understand the strengths and weaknesses of each approach. The purpose of this study is to review the implementation of various FTA approaches in NPP PSA. The strengths and weaknesses of each approach are also discussed. To achieve research objectives, this study classified those FTA approaches into conventional FTA and fuzzy FTA. Fuzzy FTA is further grouped into fuzzy hybrid FTA and fuzzy based FTA. This study concludes that safety analysts need to, firstly, confirm the type of reliability data at hands. Secondly, if epistemic uncertainty is essential and need to be considered in the study being performed, fuzzy based FTA should be applied. Otherwise, safety analysts should apply conventional FTA or fuzzy hybrid FTA depending on how the basic events' failure probabilities are generated.