KH-PINN

Physics-informed neural networks for the inverse problems (field reconstruction and parameter inference) of Kelvin-Helmholtz instability (KHI) flows.

Codes for the paper: Physics-informed neural networks for Kelvin-Helmholtz instability with spatiotemporal and magnitude multiscale.

Preprint version: https://arxiv.org/abs/2411.07524.

Some DNS data can be downloaded from the supplementary materials of the paper.

Framework

Variable density:

Constant density:

Typical results

Variable density, Re = 10000

Variable density, Re = 10000

Constant density, Re = 10000

Constant density, Re = 10000