README.md

cnn_spallart_allmaras

A project concerned with the prediction of turbulent viscosity field from an input velocity field using patch-based CNN architectures. If you exploit this work for your own research, please consider citing the article or the pre-print. This work was produced during the PhD thesis of A. Pattil.

Abstract:

From the simplest models to complex deep neural networks, modeling turbulence with machine learning techniques still offers multiple challenges. In this context, the present contribution proposes a robust strategy using patch-based training to learn turbulent viscosity from flow velocities, and demonstrates its efficient use on the Spallart-Allmaras turbulence model. Training datasets are generated for flow past two-dimensional obstacles at high Reynolds numbers and used to train an auto-encoder type convolutional neural network with local patch inputs. Compared to a standard training technique, patch-based learning not only yields increased accuracy but also reduces the computational cost required for training.

Install Requirements

cd transport
pip install -r requirements.txt
cd train
pip install -r requirements.txt

Make sure you install TF2.0 with GPU support.

Pre-process the training data

Make sure the training data is stored at case_dir

cd transport
python3 preprocess.py

The current preprocess.py looks for .vtu files in case_sample/resultats/2d/bulles*.vtu . Modify the case_dir, resultats_dir, fileListVTU to accomodate your training dataset containing .vtu files of interest.

Train the model

cd train
make train

You may clean the previous training data, if any, by

cd train
make clean

The file train.py is self-explanatory: We first load the system and user-defined libraries, set the training parameters, load the pre-processed dataset, make a patched-data, load the model architectures, define training and validation steps to suit TF2.0, and then perform the training. Make sure cuda-capabale devices and drivers are visible to Tensorflow, you may need to module load cudaxxx depending on the machine configuration.