Handling epistemic uncertainty in fault trees: New proposal based on evidence theory and Kleene Ternary decision diagrams

Fault tree (FT) is the most used approach in reliability and safety studies. In most cases, the quantification of the FT top event is carried out either (i) without considering uncertainties associated with the basic events probability distribution parameters (assuming single-valued parameters) or (ii) using Monte Carlo analysis (MC) to account for that uncertainties (using probability density function (pdf)). However, MC approach may be inappropriate to characterize parameter uncertainties (epistemic uncertainty) for the case where the available data are poor. For that case, interval-valued information (supplied by experts) related to the considered parameters is more suitable than the MC approach. Within this framework, the present paper propose a new approach addressing epistemic uncertainty in FT based on coupling Dempster-Shafer Theory (DST, also known as Evidence Theory) and the Kleene Ternary Decision Diagrams (Kleene-TDDs). Indeed, DST is used to characterize epistemic uncertainty at basic events level, whereas Kleene-TDDs make it possible to propagate that uncertainty through the fault tree gates up to the top event.