Closed-Form Approach for Epistemic Uncertainty Propagation in Analytic Models

Abstract

AbstractSystem dependability or performance is often studied using stochastic models. These models capture the natural uncertainty in the system being studied, known as aleatory uncertainty. Randomness in events of interest like times to failure/recovery of components, ability to detect failures, ability to perform recovery action, inter-arrival time, service time, etc., are taken into account in the models, by means of their distributions. The models are usually solved at fixed parameter values. However, the model input parameter values have uncertainty associated with them as they are derived either from a finite number of observations (from lifetime determining experiments or field data) or are based upon expert guesses. This uncertainty in model input parameter values, known as epistemic uncertainty, is not normally taken into account by the stochastic aleatory model.KeywordsEpistemic UncertaintyUncertainty PropagationModel Input Parameter ValuesPerform Recovery ActionsExpert GuessesThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.