Dependence assessment in Human Reliability Analysis based on canonical representation on fuzzy numbers and AHP

Dependence assessment is an important issue in human reliability analysis (HRA) in large and complex systems such as nuclear power plant. It mainly includes assessment of human task dependence and its influence on the final human error probability (HEP). In this paper, a new dependence assessment method in HRA is proposed. Firstly, the dependence influencing factors and their weights are derived by domain experts. Secondly, judgments of the status of the influencing factors are given by the analyst according to the anchors and qualitative tags. Thirdly, the judgments are converted into fuzzy numbers, and further represented in the canonical form. Finally, the overall dependence degree and the conditional human error probability (CHEP) are calculated based on a computational model. The proposed method has the merits of reasonable and flexible representation of the analysts' judgments, and less subjectivity and computational complexity in the calculation of CHEP. Examples are illustrated to show the use and effectiveness of the proposed method.