Project: Efficient upscaling of geologic model based on theory-guided convolutional encoder-decoder
- This is a course project for Deep Generative Models.
- This project trys to use the deep learning models to deal with the engineering problems in geological modeling.
- This project trys to incorporate the physical laws into the training process of encoder-decoder and achieve unsupervised training.
- 1_TgCNN_2D_hete_upscaling_tx.py: mapping construction with theory-guided training.
- 2_up_calculate_stack.py: the implementation of upscaling with the proposed method.
- 3_up_results_plot.py: results visualization.
- KLE.py: geological model generation tool.
- MyConvModel.py: the structure of the convolutional network.
- fun_P5_periodic.py: the numerical solution tool for each patch of geological model.
- [1] Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
- [2] Analytical solution for upscaling hydraulic conductivity in anisotropic heterogeneous formations
- [3] Efficient analytical upscaling method for elliptic equations in three-dimensional heterogeneous anisotropic media