Application of a Bayesian network to quantify human reliability in nuclear power plants based on the SPAR-H method

Abstract

Human error is an important factor leading to nuclear power plant (NPP) accidents. Human reliability analysis (HRA) is considered an effective method to reduce human error. Therefore, this article proposes a method to quantify human reliability based on the standardized plant analysis risk–human reliability analysis (SPAR-H) method. Firstly, the method used the performance shaping factors of SPAR-H to build a human reliability model. Secondly, the triangular fuzzy number was used to quantify the qualitative information of root nodes, and the fuzzy IF–THEN rule was used to determine the prior probability distribution of intermediate nodes. Finally, Bayesian reasoning was used to quantify human reliability based on the human reliability model. The result of the developed method is consistent with the result of cognitive reliability and error analysis methods (CREAM). The developed method can be used as a tool to quantify human reliability in the NPP system.