Subsampled Bi-Transformer Surrogate Models for Flow Estimation Arround Airfoil Geometries

This repository shows an original model adapted from Transformers [1] for solving the RANS equations, based on the LIPS framework [2] and the Airfrans Dataset [3].

The study provided here is part of the ML4physim challenge hosted by IRT-Systemx (see Codabench page). CFD simulations being very costly, the use of data-driven surrogate models can be useful to optimize the shape of airfoils without paying the cost of expensive simulations.

Subsampled Bi-Transformer (SsBT) models:

SsBT are modified version of transformer networks, where for each simulations, the query tokens are only attended to a subsampled set of key tokens from the pointcloud of the simulation which we call the skeleton of the mesh. It uses a hybrid approach on decoder-only models where both the input to be regressed and the tokens to attend to are jointly transformed. The best model's implementation is locateed in the subsampled_bi_transformers/bi_transformer folder, and can be ran using the run.py file.

This model got us the $4^{\text{th}}$ place in this challenge!

---

Installation

Install the LIPS framework

Setup an Environment

conda create --name ml4science python=3.9

Create a virtual environment

Enter virtual environment

conda activate ml4science

Install from source

Download the LIPS repository in the src folder

git clone https://github.com/IRT-SystemX/LIPS.git

Then remove the numpy and scipy requirement from the setup.py file to avoid conflicts.

cd LIPS
pip install -U .
cd ..

Install pytorch

Checkout https://pytorch.org/get-started/locally/

Install the Airfrans library and install the datasets

Install the library

pip install airfrans

Download the dataset

import os
import airfrans as af

directory_name='Dataset'
if not os.path.isdir(directory_name):
    af.dataset.download(root = ".", file_name = directory_name, unzip = True, OpenFOAM = False)

Running the model

Before running, make sure your repository has the following structure:

+---Dataset/
+---LIPS/
+---src/
    +---models/
        +---bi_transformer/
            +---my_custom_transformer.py
            +---parameters.json
    +---my_ingestion.py
    +---run.py

to run the model, simply run python run.py [model_directory].

The team

• Anthony Kalaydjian, Master student @ ENSTA/EPFL - [email protected]
• Anton Balykov, Master student @ EPFL - [email protected]
• Adrien Chan-Hon-Tong, Researcher in ML @ Onera Université Paris Saclay – [email protected]

References

[1] Attention Is All You Need, A. Vaswani et al. (2017).

[2] LIPS - Learning Industrial Physical Simulation benchmark suite, M. Leyli Abadi et al. (2022).

[3] AirfRANS: High Fidelity Computational Fluid Dynamics Dataset for Approximating Reynolds-Averaged Navier-Stokes Solutions, F. Bonnet et al. (2023).