Use of Bayesian networks and improved SPAR-H for quantitative analysis of human reliability during severe accidents mitigation process in nuclear power plant

Abstract

ABSTRACT Human reliability analysis (HRA) for severe accidents is an important component of a Level 2 probabilistic risk assessment (PRA) for nuclear power plants. In this study, a Bayesian network (BN) is used to construct a qualitative causal model of human error during severe accidents by identifying the main cognitive functions (MCFs), crew failure modes (CFMs), and performance influencing factors (PIFs) for emergency personnel implementing severe accident management guidelines (SAMGs) after core damage. This qualitative model is used to develop a quantitative HRA method for severe accidents in nuclear power plants by improving the definition of performance-shaping factors (PSFs), levels, and multiplier criteria of the standardized plant analysis risk human reliability analysis (SPAR-H) method and applying BNs to quantify the failure probability of human performance. The usability of the proposed method is verified using a case study of emergency personnel following SAMGs to depressurize a reactor cooling system.