A Validation and Uncertainty Quantification Framework for Eulerian-Eulerian Two-Fluid-Model based Multiphase-CFD Solver. Part II: Applications

TL;DR

This paper demonstrates a framework for validating and quantifying uncertainties in multiphase CFD simulations, applied to boiling heat transfer and bubbly flow, enhancing confidence in predictive modeling.

Contribution

It applies a comprehensive VUQ framework to multiphase CFD, integrating sensitivity analysis, Bayesian uncertainty quantification, and validation metrics for complex flow scenarios.

Findings

Framework effectively quantifies uncertainties in multiphase CFD predictions.

Sensitivity analysis identifies key closure parameters influencing QoIs.

Validation metrics confirm the framework's applicability to real-world problems.

Abstract

This paper is the second part of a two-part series, which introduces and demonstrates a Validation and Uncertainty Quantification (VUQ) framework that serves two major purposes: i). quantify the uncertainties of the closure relation parameters and predictions of the Multiphase Computational Fluid Dynamics (MCFD) solver; ii). evaluate the agreement between the solver predictions and the experimental measurements. The framework, with the corresponding theory and method, are outlined in the first part paper. In this paper, the workflow of the framework is implemented and demonstrated for two relevant case studies: the wall boiling heat transfer in subcooled boiling flow and the adiabatic bubbly flow. The influential closure relation parameters for multiple quantities of interest (QoIs) are identified through two different global sensitivity analysis (GSA) methods: Morris screening and Sobol indices. The model form uncertainty and model parameter uncertainty of relevant closure relations are evaluated using the modular Bayesian approach. The uncertainties of QoIs are quantified by propagating the obtained uncertainties through the solver. The agreement between solver predicted QoIs and the experimental measurement are evaluated using two different validation metrics: confidence interval and area metric. The results demonstrate the applicability of the framework.