Uncertainty Handling with RBDs and FTs

This chapter deals with uncertainty propagation handling with fault tree models (FTs). The same principles can be applied with reliability block diagrams (RBDs). The uncertain parameters have no longer point values but have to be considered as random variables with probabilistic distributions: the log-normal law (Chap. 38 ) proves very useful to do that. Analytical calculations are no longer tractable to propagate input data uncertainties toward the top event of an FT and Monte Carlo simulations (Chap. 32 ) have to be used instead. This allows to calculate the average value as well as the 90% confidence interval of the results at system level (e.g. system unavailability). The average values provided by the simulations are generally more optimistic than results without uncertainty and must be cautiously used. The 90% confidence interval is an indicator of the impact of input parameters uncertainties on the overall result uncertainty. Transformed into a pseudo error factor (see Chap. 38 ), it can be used in a relative way for comparison purpose. The difference between correlated (e.g. components from the same provider) and non-correlated (e.g. components from different providers) input parameters is brought to light and the way to model them properly is explained.