/soil-net

This project used convolutional neural networks to predict the steady-state concentration of 3D porous media, and subsequently calculates the tortuosity. This package includes data generation, processing, training, and post-processing functions. The loss function includes a Laplacian loss, which is a physics-informed loss.

Primary LanguagePythonMIT LicenseMIT

soil-net

Data generation and processing pipeline as well as deep learning utilities to train neural networks to predict tortuosity of porous media