Uncertainty quantification using Gaussian processes for topographic speed-up factors from CFD simulations

Published

July 31, 2024

Author(s)

Yunjae Hwang, Adam L. Pintar, DongHun Yeo

Abstract

Computational fluid dynamics (CFD) simulation has become increasingly popular for evaluating the topographic effects on wind fields due to its relative advantage over experimental techniques in accurately generating approach flows and capturing complex flow near the ground over topography at a small scale. Despite the popularity of the CFD methods, however, concerns about its accuracy and reliability remain. Therefore, the current study focuses on developing a calibration and uncertainty quantification (UQ) framework that integrates topographic CFD simulations with Gaussian process (GP) regression models. This framework aims to address various sources of uncertainties in both simulations and experiments. The UQ framework can be further extended to various wind engineering applications, providing a valuable tool for improving the predictive capability and reliability of CFD simulations.

Proceedings Title

Special Issue of the Journal of Wind Engineering and Industrial Aerodynamics

Conference Dates

July 29-August 2, 2024

Conference Location

Birmingham, GB

Conference Title

9th International Colloquium on Bluff Body Aerodynamics and Applications

Pub Type

Conferences

Keywords

Computational fluid dynamics(CFD), topography, uncertainty quantification(UQ), Gaussian process(GP)

Citation

Hwang, Y. , Pintar, A. and Yeo, D. (2024), Uncertainty quantification using Gaussian processes for topographic speed-up factors from CFD simulations, Special Issue of the Journal of Wind Engineering and Industrial Aerodynamics, Birmingham, GB (Accessed November 22, 2024)