Using the Uncertainty Propagation Study Type

In an uncertainty propagation study, probability distribution for the input parameters is propagated to probability distributions for the QoI. Another, less statistical, way of expressing this: In a uncertainty propagation study, the variation for the outputs is computed from given variations for the inputs. The distribution (or variation) for the inputs is well defined through the input parameter specification for the UQ study, but the output distribution is in general far from simple to compute or estimate, in particular for nonlinear QoIs. The method used for the Uncertainty Quantification Module is to first train a generally valid surrogate model, and then use Monte Carlo methods to get raw data for a KDE. The kernel density estimation itself is performed by a new dataset under Results and a new plot type, both automatically added by this analysis type.