An HPC framework for Bayesian uncertainty quantification of flows across scales (INVITED)

Abstract eng:
Predictive computational science implies the integration of models driven by physical understanding, discretised with appropriate numerical algorithms with experimental/real world data that can take advantage of extreme scale HPC architectures. Here we discuss large scale flow simulations across scales and their integration with experimental data through a Bayesian framework. Bayesian inference stands amongst the prevalent Uncertainty Quantification and Propagation (UQ+P) techniques. It is used for quantifying and calibrating engineering models, to achieve data-driven, robust predictions of system performance, reliability and safety. We assess the role of imperfections (numerical, experimental, observational) in developing and selecting models with predictive capabilities while incorporating both expert knowledge and experimental data. We demonstrate the capability to efficiently harness extreme scale computational resources for the proposed Bayesian UQ+P framework, ushering its applicability to large scale engineering problems.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket