.png){kind=icon}
Return
A Predictive Model for Turbulence Evolution and Mixing Using Machine Learning
.png){kind=icon}
OA
AI
DOI：10.1109/ACCESS.2024.3442189.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The high cost associated with high-fidelity computational fluid dynamics (CFD) is one of the main challenges that inhibit the design and optimisation of new fluid-flow systems. In this study, we explore the feasibility of a physics-informed deep learning approach to predict turbulence evolution and mixing without requiring a classical CFD solver. The deep learning architecture was inspired by integrating U-Net with inception modules for capturing the multi-scale nature of turbulent flows. In addition, a physics-constrained loss function was designed to enforce the mass and pressure conservation of the predicted solution. After trained, the optimised model was validated in the large eddy simulation (LES) of a forced turbulent mixing layer at two distinct Reynolds numbers ( $\mathrm {Re} =3000$ and 30000). The results demonstrate that the proposed approach achieves a promising solution accuracy and extrapolation ability with a significant reduction in computing time when compared to those obtained using a classical LES flow solver. The success in developing such a physics-informed deep learning approach not only justifies the potential of ML-based surrogate solvers for fast prototyping and design of generic fluid-flow systems but also highlights the key challenges arising from data-driven surrogate solver development for turbulence modelling.
Keywords:
Mathematical models
Computational modeling
Predictive models
Convolutional neural networks
Computer architecture
Feature extraction
Accuracy
Fluid dynamics
Machine learning
Computational fluid dynamics
data-driven approaches
machine learning
physics-informed neural networks
turbulence modeling
turbulent flows
Journal
IF:
3.6
Papers:
9.8W
Citations:
29.4W
Organization
Cited Papers
ShareSaveShareSaveThe aluminum-in-olivine thermometer for mantle peridotites — Experimental versus empirical calibration and potential applications.png)
Lithos
LithosIF0
PREAI
ShareSave
View more