fast-eng-super-resolution

Fast super-resolution for engineering simulations on 3D unstructured meshes

Introduction

This repository contains the code for the paper "Fast super-resolution analysis of low-pressure duct air flow through adaptive domain decomposition" by Wenzhuo Xu, Akibi Archer, Mike McCarrell, Scott Hesser, Noelia Grande Gutiérrez and Christopher McComb.

Quick Start

Use the following command to configure the environment:

pip install -r requirements.txt

Training and prediction on the framework can be done by running the python script in the project folder. An example using the duct geometry dataset, and neural operator (NO) is shown below:

python run_DS_3D.py --dataset=duct --model=neuralop --mode=train

After excuting the above command, the model will be trained on the duct dataset. To predict the results, run the following command:

python run_DS_3D.py --dataset=duct --model=neuralop --mode=predict

All results will be saved to the ./logs folder.