Skip to content

rishiiyer01/Thermal-PINO

Repository files navigation

Thermal-PINO

This is a fully executable training code for an augmented Fourier Neural Operator for 2D conjugate heat transfer problems. See channel.py to train a model. If you already have velocity data, you can use heatgen.py to generate temperature data from the heat convection-diffusion equation. If not, please use some of the OpenFOAM functionality to generate data.

image Full model was trained on a single NVIDIA p100 15 gb node. Inference takes less than 0.1s, this type of architecture shows potential to automate ChT simulations of varying similar geometries.

See inference_testing_notebook.ipynb for test-set examples.

Paper: https://rishiiyer.com/THESIS_RISHABH_IYER.pdf

About

Physics informed neural operator that solves the navier stokes equations and the heat advection equation for conjugate heat trasnfer problems on a channel. The goal is to simulate and generalize across different geometries and reynolds numbers with the same architecture. Research with Prof. Hongwei Sun.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors