The prediction of human error probability based on Bayesian networks in the process of task

With the increasing of the reliability of the equipment itself, more and more accidents are caused by human error so that human reliability analysis (HRA) becomes more and more crucial. The paper constructs a prediction method of human error probability (HEP) by integrating performance shaping factors (PSFs) and Bayesian networks theory. Firstly, conditional probability table (CPT) of each node of Bayesian networks is determined through simulation data, field operation or expert knowledge. Secondly, the value of HEP is acquired by Junction Tree inference algorithm based on the given network structure and the known evidence, while PSF is a variable during task, the new HEP should be gotten by updating Bayesian networks based on new evidence, According to this, curve of HEP change over time can be drawn. Finally, the paper takes pilots as example to present application process and validity of this method.