Model Uncertainty Quantification and Objective-Oriented Sampling in Simulation-Based Design Under Uncertainty
Various sources of uncertainty exist in simulation-based design under uncertainty. Quantifying the uncertainty of a model and the resulting probabilistic predictions is essential for robust and reliable decision making. In physics-based engineering modeling, the two primary sources of model uncertainty, which account for the differences between computer models and physical experiments, are parameter uncertainty and model discrepancy. Distinguishing the effects of these two sources of uncertainty can be challenging. For situations in which identifiability cannot be achieved using only a single response, we propose the use of multiple responses that share a mutual dependence on the common set of calibration parameters to improve identifiability. In this talk, we will present a multi-response modular Bayesian approach and demonstrate that using multiple responses can substantially enhance identifiability. In addition, we will also introduce a newly developed, sequential objective-oriented sampling approach to robust design. We will present the need and the techniques developed for sampling control and noise variables separately towards optimizing a robust design objective that incorporates both the mean and variance of design performance.