Abstract:
|
Uncertainty quantification (UQ), and especially Bayesian calibration, help obtain reliable predictions for multi-scale problems. These models often must be constrained to experimental bench-scale data. The calibration framework requires discrepancy term between the model and reality, and often a stochastic emulator trained to limited output from a computationally expensive model. Here we present advancements in statistical methodology in UQ; in particular, the discussion will be about dealing with model discrepancy in submodels and upscaling uncertainty or calibration at the large scale system level. Other possible challenges include accounting for due functional inputs and outputs, an intrusive dynamic discrepancy approach, upscaling of uncertainty for a multi scale system, and calibration problems with a large number of potential parameters. We used flexible and computationally approaches for the emulator and/or discrepancy, including a Bayesian Smoothing Spline (BSS) ANOVA Gaussian Process. The methodology presented may have far-reaching impact in many areas of science where multiscale modeling is used. This work will include applications to carbon capture systems.
|