Unsupervised Deep Learning of Turbulent Heat Transfer via Generative Adversarial Networks

Read full text Summarise with VIRAT
Authors: Nicholas J. Ward
Cited by: 0
References: 37

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Machine learning and deep learning can be useful in providing insight to near-wall turbulence. The research question that this study seeks to answer is how to effectively predict the turbulent heat transfer of a 2D channel flow using a generative adversarial network (GAN). The first objective of this study is to predict the heat transfer using the wall variables p, du/dy, and dw/dy only. The second objective is to extrapolate the trained GAN model to temperature gradients at higher Reynolds numbers. Direct numerical simulations (DNS) were performed to determine the training data, and machine learning algorithm was developed to effectively capture the flow physics in a turbulent channel flow. The predictions from the proposed GAN correlated well with the DNS data at Re_tau = 180 . The predictions were effectively captured within 2% of the DNS training data. The temperature was not required to measure the temperature gradients using the GAN model proposed in this study. Additionally, the GAN was also able to capture the flow physics for the test data at higher Reynolds numbers. To accomplish this, an appropriate normalization technique was required for the predictions. This work will be useful in future work to be applied to other problems in fluid mechanics, such as drag reduction, capturing near-wall flow physics, and super-resolution.

Fields of Study & Subjects

Model Reduction and Neural Networks Fluid Dynamics and Turbulent Flows Heat Transfer Mechanisms

Publication details

Journal / source
International Journal of Renewable Energy Research
Publication date
1 January 2026
Language
EN