Conditional Flow Matching for Generative Modelling of Near-Wall Turbulence

This repository contains the official implementation of the paper: "Conditional flow matching for generative modelling of near-wall turbulence with quantified uncertainty".

Read the paper

Methodology

Environment Setup

To set up the environment, follow these steps:

1. Create the Conda environment:

   conda env create -f environment.yml

2. Activate the environment:

   conda activate fm-swag

3. Configure Environment Variables: Create a .env file in the root directory of the repository with the following content. Specify the GPU device ID you wish to use (leave empty if not applicable or to use default).

   CUDA_VISIBLE_DEVICES=
   PYTHONPATH=$PYTHONPATH:.

   Then, source the environment file using the following command:

   set -o allexport && source .env && set +o allexport

Usage

Run scripts using the following syntax:

python scripts/<script_name>.py scripts/<config>.yaml

Training

• Flow Matching Model: For training the flow matching model to learn the prior distribution of velocity fluctuations:

  python scripts/training/train_unet_fm.py scripts/training/unet_fm.yaml

• Baseline (FCN): For the baseline training, you can use either the Wall-Measurement $\rightarrow$ Velocity-Fluctuations (WMVF) or Velocity-Fluctuations $\rightarrow$ Wall-Measurement (VFWM) configuration.

  python scripts/training/train_unet_baseline_fwd_op.py scripts/training/unet_bb_wmvf.yaml
  # OR
  python scripts/training/train_unet_baseline_fwd_op.py scripts/training/unet_bb_vfwm.yaml

• SWAG Model: For training SWAG forward operator, used for conditional inference:

  python scripts/training/train_unet_swag_fwd_op.py scripts/training/unet_swag_fwd_op.yaml

Inference

• Conditional Inference:

  python scripts/inference/ours_prod.py scripts/inference/ours.yml

• Unconditional Inference:

  python scripts/inference/ours_uncond_prod.py scripts/inference/ours_uncond.yml

• Baseline Inference:

  python scripts/inference/baseline_prod.py scripts/inference/baseline.yml

Issues

Please open an issue if you have any questions.

Citation

If you find this code useful in your research, please cite:

@article{Parikh_Fan_Wang_2026,
 title={Conditional flow matching for generative modelling of near-wall turbulence with quantified uncertainty},
 volume={1029},
 DOI={10.1017/jfm.2026.11193},
 journal={Journal of Fluid Mechanics},
 author={Parikh, Meet Hemant and Fan, Xiantao and Wang, Jian-Xun},
 year={2026},
 pages={A32}
}