7tl7qns7ch / PINN4___

physics-informed neural networks parts for ___

PINN4___

Physics-informed neural network parts for ___. Codes are based on https://github.com/neuraloperator/physics_informed (Z. Li, et. al. 2022.) and https://github.com/BaratiLab/Diffusion-based-Fluid-Super-resolution (D. Shu, et. al. 2023) with some modifications for further research.

Conda Environment

conda create -n pinn4 python=3.9

conda activate pinn4

pip install --upgrade pip

pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2

conda install -c conda-forge deepxde

pip install pyDOE

Navier Stokes Data Generator

NS_generator.ipynb (above figure can be obtained with changing args.Re=100, 1000, 10000 in the second code block)

Previous Navier Stokes Dataset

• spatial domain: $x\in (0, 2\pi)^2$
• temporal domain: $t \in [0, 0.5]$
• forcing: $-4\cos(4x_2)$
• Reynolds number: 500

Data of shape (N, T, X, Y) where N is the number of instances, T is temporal resolution, X, Y are spatial resolutions.

1. NS_Re500_s256_T100_test.npy: 100x129x256x256

• spatial domain: $x\in (0, 2\pi)^2$
• temporal domain: $t \in [0, 10]$
• forcing: $-4\cos(4x_2) -0.1\omega(x, t)$
• Reynolds number: 1000

Data of shape (N, T, X, Y) where N is the number of instances, T is temporal resolution, X, Y are spatial resolutions.

1. (kf_2d_re1000_256_40seed): 40x320x256x256

- PINN loss for Navier Stokes

• Fourier neural operator with pinn loss in spectral space.

python train_pdeloss.py --tqdm