A Physics Informed Neural Network that predicts the a two phase flow in porous media.
Prerequisites:
- Julia version 1.3.1
- ADCME
- FwiFlow
- Python 3.7 +
This repository contains the full cycle of training a neural network. It contains code to do the following
- DNN to solve analytical Buckley-Leverett
- DNN to solve Forward Problem based on simulation data
- Calibration tests to solve inverse problem for unknown Permeability and Porosity
- DNN to solve homogeneous inverse flow
- DNN for homogeneous approximation of heterogeneous flow + invert permeability and porosity values
- DNN to solve heterogeneous flow + invert permeability and porosity values