CFD–ANN Surrogate Modelling for Fixing-Angle Optimisation of a Darrieus VAWT

This repository contains the Artificial Neural Network (ANN) surrogate model code accompanying the paper:

CFD and ANN surrogate modelling for fixing angle optimisation of a Darrieus vertical axis wind turbine A. Tamaddon, Ma Mohin, Y. I. Mogul, J. D. Quadros, M. Ndiaye Engineering Applications of Computational Fluid Mechanics, 2026, 20(1), 2654918. DOI: 10.1080/19942060.2026.2654918

📄 Read the paper

• Download the PDF: paper.pdf (open access, CC BY 4.0)
• Publisher / DOI: https://doi.org/10.1080/19942060.2026.2654918

Overview

The notebook trains a feed-forward neural network that predicts the power coefficient (C_P) of a three-bladed Darrieus VAWT from two inputs — Tip-Speed Ratio (TSR) and blade Fixing Angle (FA). Trained on URANS-based CFD data, the surrogate maps the TSR–FA design space at negligible computational cost and locates the optimum configuration (FA = +1.5°, TSR ≈ 2.0).

Repository contents

File	Description
paper.pdf	The published open-access article (CC BY 4.0)
ANN_surrogate_model.ipynb	Jupyter notebook: data loading, preprocessing, ANN training, evaluation, and prediction plots
VAWT.csv	CFD dataset used to train the model (TSR, FA, C_P) — see note below
requirements.txt	Python package dependencies

Method summary

• Inputs: TSR (0.5–3.0), FA (−1.5° to +1.5°)
• Output: Power coefficient C_P
• Dataset: 42 CFD operating points, 80/20 train/test split, fixed random seed (42)
• Preprocessing: feature normalisation with StandardScaler
• Architecture: 2 fully connected hidden layers (ReLU) + 1 linear output neuron
• Training: Adam optimiser, MSE loss, early stopping (patience 200), up to 4000 epochs, batch size 8
• Metrics: R², MAE, MSE

How to run

1. Install the dependencies:

   pip install -r requirements.txt

2. Make sure VAWT.csv is in the same folder as the notebook.
3. Open and run the notebook:

   jupyter notebook ANN_surrogate_model.ipynb

Data availability

Per the paper, the full CFD simulation results and ANN training data are available from the corresponding author upon request. If you do not include VAWT.csv in this repository, please remove that row from the table above and note here how the data can be obtained.

Citation

If you use this code, please cite the paper:

@article{tamaddon2026cfdann,
  title   = {CFD and ANN surrogate modelling for fixing angle optimisation of a Darrieus vertical axis wind turbine},
  author  = {Tamaddon, Amir and Mohin, Ma and Mogul, Yakub Iqbal and Quadros, Jaimon Dennis and Ndiaye, Mamadou},
  journal = {Engineering Applications of Computational Fluid Mechanics},
  volume  = {20},
  number  = {1},
  pages   = {2654918},
  year    = {2026},
  doi     = {10.1080/19942060.2026.2654918}
}

▶️ View / run the notebook

License

Released under the MIT License (or another license of your choice).

💾 Cite the code / data

This repository is archived on Zenodo. If you use the code or dataset, please cite both the paper (above) and this DOI: 10.5281/zenodo.20479467