Overview

We use M-GAN to predict unavailable parameters (u and p) from available parameters (v and w) of turbulent channel flow. The sample code, data, and tutorial are provided below.

Dependencies

Python 3.6-3.8

tensorflow >=2.2.0 <2.4.0 (cuDNN=7.6, CUDA=10.1 for tensorflow-gpu)

Numpy <1.19

Data

Training data: 1000 snapshots of in/output and label data are given and mixed randomly.

Testing data: 50 snapshots of each channel flow at Reτ = 180 and 500 are provided for testing.

Training

Use code mgan.py to train the model. Then, get architecture and weights files.

Testing

Use code predict_test.py with architecture and weights files to test. Then, get predicted flow fields.