HeatModelMDFM

The codes are for the paper

Modeling Heat Conduction with Two-Dissipative Variables: A Mechanism-Data Fusion Method

Setup

All codes were developed and tested with Python 3.8, and were implemented by pytorch 1.10.1.

Training

Code and default options can be found in train.py. You can simply modify these default options to benefit training.

python train.py

Trained parameters will be stored in directory like ./BTEdataU01N10/net_params_U01N10Coarse1steps1.pkl.

Download training and testing data and make your working directory like this:

traindata/
testdata_075_085/
testdata_discontinuous_cases/
train-predict/
    train.py
    predict.py

Testing

python predict.py

You can change the default options in predict.py to test different settings. Remember to change the CDFfile_path.