Comparison between two modern uncertainty expression and propagation approaches

Two different uncertainty expression and propagation approaches are presented and compared. In particular, an implementation of the known probabilistic approach and a new Random-Fuzzy Variable (RFV) method based on the theory of Evidence. Both approaches use an explicit time correlation of input quantities to take into account systematic contributions. Numerical results show that both the type of uncertainty contribution (random or systematic) and the owned level of knowledge (Probability density Function PDF or simply a limited interval) must be carefully evaluated in uncertainty analysis. The new RFV approach allows to seamlessly deal with PDFs and limited intervals. This advantage is not present in the probabilistic approach, which yields questionable results in complete ignorance situations.