---

Explainable Fuzzy GNNs for Leak Detection in Water Distribution Networks

This repository contains the implementation of an Explainable Fuzzy Graph Neural Network (FGNN) framework for detecting and localizing leaks in Water Distribution Networks (WDNs). By combining Graph Neural Networks (GNNs) with fuzzy logic, the model achieves both accurate predictions and interpretable, rule-based explanations.

This work is based on the Hanoi Benchmark Network dataset (LeakDB) and accompanies the paper:

"Explainable Fuzzy GNNs for Leak Detection in Water Distribution Networks" Qusai Khaled, Pasquale De Marinis, Moez Louati, David Ferras, Laura Genga, Uzay Kaymak Submitted to the 2025 IFSA World Congress NAFIPS

---

🔍 Overview

The project investigates multiple GNN architectures and proposes a fuzzy-enhanced model that integrates mutual information and fuzzy rules to support explainable leak detection.

Main tasks:

• Leak Detection (Graph-level classification): Detect whether a leak exists in the network.
• Leak Localization (Node-level classification): Identify the exact node or pipe with the leak.

Example explanation:

"IF pressure at Node 1 is high AND pressure at Node 2 is low, THEN leak probability at Node 5 is high."

This rule-based reasoning supports better decision-making for engineers and water utility managers.

---

📁 Repository Structure

.
├── notebooks/                # Interactive Jupyter notebooks
│   ├── 01_Reshape_data.ipynb
│   ├── 01_Reshape_oldata.ipynb
│   ├── 05_GCN.ipynb
│   ├── 06_GAT.ipynb
│   ├── 11_AnomalyDetectionScenario.ipynb
│   ├── 12_FeatureExtraction.ipynb
│   └── Explain.ipynb
├── scripts/                  # Python source files
│   ├── data.py
│   ├── preprocess.py
│   ├── model.py
│   ├── train.py
│   ├── test.py
│   ├── grid.py
│   ├── loss.py
│   ├── tracker.py
│   └── logger.py
├── .gitignore
└── README.md

---

⚙️ Setup

1. Install dependencies:

Ensure Python 3.8+ is installed. Then install dependencies:

pip install -r requirements.txt

ℹ️ You may need to install PyTorch Geometric separately depending on your system.

2. Prepare the dataset:

• Download the Hanoi Benchmark Network dataset (LeakDB).
• Place it in the data/ directory or another appropriate path.
• Use the preprocessing notebooks (01_Reshape_data.ipynb, etc.) to convert and prepare the data.

---

🚀 Usage

Train and Test

python train.py
python test.py

Configuration options can be modified directly in the script or through arguments.

Run Experiments

Explore different GNN architectures and fuzzy logic features

Hyperparameter Tuning

Use:

python grid.py

Or refer to Grid search py for custom tuning setups.

---

📊 Results

Performance metrics and visualizations are provided in the paper. Experimental results (e.g., accuracy, explainability) are saved in the results/ directory after training.

---

📚 Citation

If you use this repository in your research, please cite the following:

@article{khaled2025explainable,
  title={Explainable Fuzzy GNNs for Leak Detection in Water Distribution Networks},
  author={Khaled, Qusai and De Marinis, Pasquale and Louati, Moez and Ferras, David and Genga, Laura and Kaymak, Uzay},
  journal={2025 IFSA World Congress NAFIPS},
  year={2025}
}

---

🙏 Acknowledgments

This work is supported by the ILUSTRE project, funded in part by the Dutch Research Council (NWO).

---

📬 Contact

For questions or suggestions, feel free to open an issue. or email qusai.khaled@ieee.org

---

Let me know if you'd like a requirements.txt template or badge support (e.g., license, build status) for the README as well.