Development of a BN framework for human reliability analysis through virtual simulation

Abstract

Humans are an integral part of complex systems such as nuclear power plants and have to play a significant role in ensuring the safety and reliability of these systems. Failure to perform the intended task within the stipulated time by the operator can challenge the safety of the system. Human reliability analysis (HRA) is a widely practiced methodology to estimate the contribution of operator error towards the overall risk to the facility. HRA methods quantify this contribution in terms of human error probability (HEP) accounting for various psychological and physiological factors that influence the performance of the operator. These factors are referred to as human factors (HF), which enhance or degrade the human performance. The paper discusses the use of virtual simulation as a tool to generate the HF data from the virtual model of an in-house experimental facility. This paper also demonstrates the use of multi-attribute utility theory to determine a suitable HRA method amongst several HRA methods to quantify the HEP based on the desired set of HRA attributes. As classical HRA methods, generally, do not address the interactions among the HFs, the Bayesian network technique has been employed in this study to account for HF interactions.