Machine-Learning-Based Subgrid Modelling for the Atmospheric Boundary Layer
Abstract
A proof-of-concept study of machine-learning-based subgrid modelling for large-eddy simulation (LES) is performed. An artificial neural network is trained for the prediction of residual stresses and scalar fluxes based on the filtered velocity and scalar fields from large-eddy simulation. One half of the dataset is used for training and the other half for validation representing a priori testing of the subgrid model. More comprehensive testing and optimisation for evaluation speed is necessary before practical use in an LES model.
Fields of Study & Subjects
Fluid Dynamics and Turbulent Flows
Model Reduction and Neural Networks
Meteorological Phenomena and Simulations
Publication details
- Journal / source
- Topical Problems of Fluid Mechanics
- Pages
- 64–67
- Publication date
- 1 January 2026
- Language
- EN