In the last decade, Bayesian networks (BNs) have been identified as a powerful tool for human reliability analysis (HRA), with multiple advantages over traditional HRA methods. In this paper we illustrate how BNs can be used to include additional, qualitative causal paths to provide traceability. The proposed framework provides the foundation to resolve several needs frequently expressed by the HRA community. First, the developed extended BN structure reflects the causal paths found in cognitive psychology literature, thereby addressing the need for causal traceability and strong scientific basis in HRA. Secondly, the use of node reduction algorithms allows the BN to be condensed to a level of detail at which quantification is as straightforward as the techniques used in existing HRA. We illustrate the framework by developing a BN version of the critical data misperceived crew failure mode in the IDHEAS HRA method, which is currently under development at the US NRC (Xing et al., 2013). We illustrate how the model could be quantified with a combination of expertprobabilities and information from operator performance databases such as SACADA. This paper lays the foundations necessary to expand the cognitive and quantitative foundations of HRA.
Discrete Bayesian networks (BNs) can be effective for risk-and reliability assessments, in which probability estimates of (rare) failure events are frequently updated with new information. To solve such reliability problems accurately in BNs, the discretization of continuous random variables must be performed carefully. To this end, we develop an efficient discretization scheme, which is based on finding an optimal discretization for the linear approximation of the reliability problem obtained from the First-Order Reliability Method (FORM). Because the probability estimate should be accurate under all possible future information scenarios, the discretization scheme is optimized with respected to the expected posterior error. To simplify application of the method, we establish parametric formulations for efficient discretization of random variables in BNs for reliability problems based on numerical investigations. The procedure is implemented into a software prototype. Finally, it is applied to a verification example and an application example, the prediction of runway overrun of a landing aircraft.
A Bayesian network (BN) model for predicting wildfire spreading was developed. From the available indicator variables related to weather, topography and land cover, the most informative were selected with the help of automatic structure learning algorithms. A final BN model was then constructed from these indicators using phenomenological reasoning. Automatic structure learning of the complete model was found to have severe limitations due to large number of variables in combination with limited number of observations. The BN model was learned and validated with data from the Mediterranean island of Cyprus. The final BN was compared to a Naïve Bayesian Classifier (NBC), which serves as a benchmark, and it was shown to be applicable for prediction purposes.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.