BibTeX

@inproceedings{GrothPSA2011,
title = {Development and Use of a Bayesian Network to Estimate Human Error Probability},
author = {Katrina Groth and Ali Mosleh},
year = {2011},
date = {2011-03-01},
booktitle = {Proceedings of the ANS International Topical Meeting on Probabilistic Safety Assessment and Analysis (PSA 2011)},
address = {Wilmington, NC},
organization = {American Nuclear Society},
abstract = {In Human Reliability Analysis (HRA), Performance Influencing Factors (PIFs) are used to represent the various factors that influence individual behavior and to predict the outcome of human cognitive processes. PIFs have been used in many HRA methods as a means to estimate Human Error Probability (HEP). Recently there has been an interest in replacing “linear models” of accounting for the impact of PIF on estimates for HEPs with model-based approach that include the interdependencies among PIFs. Addressing the PIFs in a model is expected to provide more refined HEP estimates and reduce the amount of information required to assess HEPs.
A previous paper [1] has proposed a Bayesian Network (BN) model of the relationships among PIFs. The model structure and probabilities were developed based on analysis of available data. The BN provides a natural framework to assess the impact of different combinations of the same PIFs. This paper describes an extension of the original model to estimate HEPs. This paper discusses how to the model was modified and how it can be used to make inferences in the BN. It also demonstrates how to integrate the PIF model into traditional PRA.},
keywords = {Bayesian Networks, causal models, human error, Human Reliability Analysis (HRA), nuclear power, Performance Shaping Factors (PSFs)},
pubstate = {published},
tppubtype = {inproceedings}
}

Abstract

In Human Reliability Analysis (HRA), Performance Influencing Factors (PIFs) are used to represent the various factors that influence individual behavior and to predict the outcome of human cognitive processes. PIFs have been used in many HRA methods as a means to estimate Human Error Probability (HEP). Recently there has been an interest in replacing “linear models” of accounting for the impact of PIF on estimates for HEPs with model-based approach that include the interdependencies among PIFs. Addressing the PIFs in a model is expected to provide more refined HEP estimates and reduce the amount of information required to assess HEPs.
A previous paper [1] has proposed a Bayesian Network (BN) model of the relationships among PIFs. The model structure and probabilities were developed based on analysis of available data. The BN provides a natural framework to assess the impact of different combinations of the same PIFs. This paper describes an extension of the original model to estimate HEPs. This paper discusses how to the model was modified and how it can be used to make inferences in the BN. It also demonstrates how to integrate the PIF model into traditional PRA.