Physics-Constrained Convolutional Autoencoders for 2D Cylinder Flow at Re=100
Abstract
This dataset and code repository contains the implementation of convolutional autoencoders (CAE) and physics-constrained autoencoders (PO-CAE) trained on 2D incompressible cylinder flow simulations at Reynolds number 100. The models incorporate soft constraints on kinetic energy, vorticity, enstrophy, and temporal energy increments. The repository includes training scripts, hyperparameter sweeps, evaluation tools, Pareto analysis, and latent space diagnostics.
Fields of Study & Subjects
Model Reduction and Neural Networks
Neural Networks and Applications
Fluid Dynamics and Turbulent Flows
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 17 February 2026
- Language
- EN
- Document type
- Dataset