Modeling Epistemic and Aleatory Uncertainty in Bayesian Networks for Dependability Analysis

Presented at the 5th Annual BayesiaLab Conference in Paris, September 25–October 4, 2017.

Abstract

The question of modeling epistemic uncertainty is a long-lasting debate in the community. In dependability analysis, it is a more recent question. If yesterday we have much data from monitored systems or well-known sub-system reused in well-known conditions, it is not the case today. The studied system becomes large, complex and not all data are completely known.

Bayesian network in reliability analysis is interesting because of the ease of the modeling tool which can encode reliability models in the same way as well handled tools. But, the question of epistemic uncertainty arises. How to encode the different forms of uncertainty in the Bayesian network tool.

This plenary talk explains the way we consider aleatory and epistemic uncertainty and how we encode it in a Bayesian network for assessing system reliability and driving maintenance actions.

Presentation Video

Presentation Slides

About the Presenters

C. Simon, P. Weber, B. Iung