Temperature field reconstruction implementation package

Introduction

This project provides the implementation of the paper "A Machine Learning Modelling Benchmark for Temperature Field Reconstruction of Heat-Source Systems". [paper] [data generator]

Requirements

• Software
  • python >= 3.6
  • cuda (only GPU is required)
  • pytorch
• Hardware
  • GPU with at least 16GB (recommended)
  • CPU

Environment construction

1. Install required packages followed requirements.txt.

pip install -r requirements.txt

2. Install torch-cluster, torch-scatter, torch-sparse package (matching the version of torch, cuda)

   • Automatic installation [install instruction]

     pip install torch-scatter -f https://pytorch-geometric.com/whl/torch-${version}+${CUDA}.html
     pip install torch-sparse -f https://pytorch-geometric.com/whl/torch-${version}+${CUDA}.html
     pip install torch-geometric
     pip install torch-cluster -f https://pytorch-geometric.com/whl/torch-${version}+${CUDA}.html
     pip install torch-spline-conv -f https://pytorch-geometric.com/whl/torch-${version}+${CUDA}.html
     

     $version describes the torch version and should be replaced by 1.4.0,1.5.0,1.6.0,1.7.0, 1.8.0,1.9.0,1.7.1,1.8.1.

     $CUDA should be replaced by cpu, cu101, cu102, cu111, cu92.

   • Manual installation [download]

Running

All the methods for TFR-HSS task can be accessed by running main.py file.

All the parameters are defined in two forms, namely the yaml file (default config\config.yml) and command line parameters. Priority: yaml < command line

Image-based and Vector-based methods

The image-based and vector-based methods are following the same command.

• Training

  python main.py -m train
  

  or

  python main.py --mode=train
  

• Test

  python main.py -m test --test_check_num=21
  

  or

  python main.py --mode=test --test_check_num=21
  

  or

  python main.py -m=test -v=21
  

  where variable test_check_num is the number of the saved model for test.

• Prediction visualization

  python main.py -m plot --test_check_num=21
  

  or

  python main.py --mode=plot --test_check_num=21
  

  or

  python main.py -m=test -v=21
  

  where variable test_check_num is the number of the saved model for plotting.

Point-based methods

Only testing is permitted for point-based methods.

• Testing

  python main.py
  

• Testing with prediction visualization

  python main.py --plot
  

Project architecture

• config: the configuration file
  • config.yml describes configurations
    • model_name: model for reconstruction
    • backbone: backbone network, used only for deep surrogate models
    • data_root: root path of data
    • train_list: train samples
    • test_list: test samples
    • others
• samples: examples
• outputs: the output results by test and plot module. The test results is saved at outputs/*.csv and the plotting figures is saved at outputs/predict_plot/.
• src: including surrogate model, training and testing files.
  • test.py: testing files.
  • train.py: training files.
  • plot.py: prediction visualization files.
  • point.py: Model and testing files for point-based methods.
  • DeepRegression.py: Model configurations for image-based and vector-based methods.
  • data: data preprocessing and data loading files.
  • models: interpolation and machine learning models for the TFR-HSS task.
  • utils: useful tool function files.

• docker: start with docker.
• lightning_logs: saved models.

One tiny example

One tiny example for training and testing can be accessed based on the following instruction.

• Some training and testing data are available at samples/data.
• Based on the original configuration file, run python main.py directly for a quick experience of this tiny example.

Citing this work

If you find this work helpful for your research, please consider citing:

@article{gong2021,
    Author = {Xiaoqian Chen and Zhiqiang Gong and Xiaoyu Zhao and Weien Zhou and Wen Yao},
    Title = {A Machine Learning Modelling Benchmark for Temperature Field Reconstruction of Heat-Source Systems},
    Journal = {arXiv preprint arXiv:2108.08298},
    Year = {2021}
}