SC²S Colloquium - January 20, 2016
|Date:||January 20, 2016|
|Time:||3:00 pm, s.t.|
Alex Ziller: Uncertainty Quantification of decoupled multi-component systems
In the growing field of uncertainty quantification (UQ), an important class of problems are the ones in which the underlying model has a composite structure. To efficiently quantify the uncertainty in such systems, we break the uncertainty analysis in two phases, i.e. an online and an offline phase. In the online phase, we decouple the system's components and propagate the uncertainty in each subsystem, in parallel. In the offline phase, based on the obtained simulation, we update the proposal distribution and afterwards, perform the post-processing of the quantities of interest. In this work, we approach the problem using several approaches based on Monte Carlo sampling and polynomial chaos approximation. To gain insights about the capabilities of these approaches, we apply them in several test scenarios, using multiple proposal distributions.