Open Collections

Uncertainty quantification in multiphase computational fluid dynamics

Banff International Research Station for Mathematical Innovation and Discovery

Talk: Regular Abstract: Predictive simulations based on multiphase computational fluid dynamics (CFD) have become indispensable for scaling up multiphase devices used in energy technologies. The predictive accuracy is subject to various uncertainties in the simulation. Quantifying them is essential for making engineering decisions based on simulations, which reduces the time and cost of scaling up new multiphase devices. Uncertainty quantification (UQ) and sensitivity analysis also help engineers to plan validation experiments and reduce the risk in scale up. UQ in multiphase CFD is currently a major research focus at National Energy Technology Laboratory (NETL). This presentation will review the methods explored at NETL and the recent progress. A validation methodology is applied to the simulations of a circulating fluidized bed. The overall pressure drop is the quantity of interest; the solids circulation rate and the gas velocity are the uncertain input quantities. From the known uncertainties in the input quantities, surrogate model, spatial discretization and time averaging, the uncertainty in the pressure drop is calculated, and the model-form uncertainty is determined with the help of validation data. The UQ results are expressed as a p-box plot, which can provide answers to various design questions. A Bayesian calibration methodology is applied to conduct predictive simulations of a carbon capture device. This methodology handles situations where many parameters describing different physical phenomena appear in a complex multi-physics problem, and a hierarchy of experiments is required for model validation. It enables progressively improving the model by learning from new information provided by the model validation hierarchy. Sensitivity and grid convergence analyses have been used to understand a gasifier model. Calculating the uncertainty resulting from grid convergence has been identified as an issue. An error transport equation formulation is being developed as an alternative to standard methods for calculating that uncertainty.

Mathematics; Fluid mechanics; Mechanics of particles and systems; Fluid dynamics

video/mp4

Author affiliation: NETL

2017-02-08

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/60536

Unreviewed

http://creativecommons.org/licenses/by-nc-nd/4.0/