Comparing the treatment of uncertainty in Bayesian networks and fuzzy expert systems used for a human reliability analysis application

摘要: The use of expert systems can be helpful to improve the transparency and repeatability assessments in areas risk analysis with limited data available. In this field, Human Reliability Analysis (HRA) is no exception, and, particular, dependence an HRA task strongly based on analyst judgement. among Failure Events refers assessment effect earlier human failure probability subsequent ones. This paper analyses compares two systems, Bayesian Belief Networks Fuzzy Logic (a Expert System, FES), respectively. comparison shows that a BBN approach should preferred all cases characterized by quantifiable uncertainty input (i.e. when distributions assigned describe parameters uncertainty), since it provides satisfactory representation its output directly interpretable for within PSA. On other hand, very knowledge, may feel constrained probabilistic framework, which requires assigning describing uncertainty. these cases, FES seems lead more transparent