This article discusses the use of Bayesian Neural Networks (BNNs) for uncertainty quantification in data-driven closure models for reacting turbulence. It explores the incorporation of epistemic and aleatoric uncertainties in modeling the progress variable scalar dissipation rate. The study demonstrates the efficacy of BNN models in providing unique insights into the structure of uncertainty in data-driven closure models. The article also proposes a method for incorporating out-of-distribution information in BNNs and evaluates the model's performance on a dataset with various flame conditions and fuels.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Graham Pash,... at arxiv.org 03-01-2024
https://arxiv.org/pdf/2402.18729.pdfDeeper Inquiries