_   _______ ______
    ____  __ ___  / | / / ___// ____/
   / __ \/ / __ \/  |/ /\__ \/ __/   
  / /_/ / / / / / /|  /___/ / /___   
 / .___/_/_/ /_/_/ |_//____/_____/   
/_/                                  

Setup

Virtual Environment

Windows (Powershell)

$ python -m venv .venv
$ .\.activate.ps1

Linux

$ python -m venv .venv
$ source ./venv/bin/activate

Dependencies

$ python -m pip install --upgrade pip
$ python -m pip install wheel
$ python -m pip install torch --index-url https://download.pytorch.org/whl/cu124
$ python -m pip install -r requirements.txt

Usage

• Wikipedia

usage: nse [-h] -E {empty,step,slalom,block,slit,cylinder,wing} [--inlet <u>] [--nu <nu>] [--rho <rho>] [--id <id>] [-N <train>] [-L <layers>] [-D {cpu,cuda}] [--supervised] [--dry] [-P] [-F] [-G] [-R]

options:
  -h, --help            show this help message and exit
  -L <layers>           size of layers seperated by colon (default: 100:100:100)

initialization:
  -E {empty,step,slalom,block,slit,cylinder,wing}
                        choose experiment
  --inlet <u>           set intake (default: 1.0)
  --nu <nu>             set viscosity (default: 0.1)
  --rho <rho>           set density (default: 1.0)

optimization:
  --id <id>             identifier / prefix for output directory (default: timestamp, example: 2024-10-28_09-26-09)
  -N <train>            number of optimization steps (default: 0)
  -D {cpu,cuda}         device used for training (default: cpu)
  --supervised          set training method to supervised approach (requires -F)
  --dry                 dry run

output:
  -P                    plot NSE in output directory
  -F                    initialize OpenFOAM experiment
  -G                    grade prediction (requires -F and -P)
  -R                    plot NSE with high resolution grid in output directory (requires -P)

Examples

$ python -m src.nse -E step --inlet 5 --nu .08 -N 100

$ python -m src.nse -E wing --id wing -L 100:100:100:100 --inlet 1 --nu .01 -D cuda -PRFGN 30000

$ python -m src.nse -E block

References

• Raissi, M. et al.: Physics Informed Deep Learning (Part II)