cs.LGMar 25, 2026

The impact of sensor placement on graph-neural-network-based leakage detection

Authors: J. J. H. van Gemert, V. Breschi, D. R. Yntema, K. J. Keesman, M. Lazar

Organizations: Control Systems Group, Dept. of Electrical Engineering, Eindhoven University of Technology, 5600MB Eindhoven, The Netherlands · Wetsus, Centre of Excellence for Sustainable Water Technology, 8900MA Leeuwarden, The Netherlands. · Mathematical and Statistical Methods – Biometris, Wageningen University, 6708PB Wageningen, the Netherlands

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Sensor placement for leakage detection in water distribution networks is an important and practical challenge for water utilities. Recent work has shown that graph neural networks can estimate and predict pressures and detect leaks, but their performance strongly depends on the available sensor measurements and configurations. In this paper, we investigate how sensor placement influences the performance of GNN-based leakage detection. We propose a novel PageRank-Centrality-based sensor placement method and demonstrate that it substantially impacts reconstruction, prediction, and leakage detection on the EPANET Net1.

Figures & tables

Figure 1: Illustration of EPANET Net1 WDN consisting of two reservoirs/tanks (black icons), one pump (black circle), and junctions (blue circles) connected through pipes (blue lines).
Figure 2: Processing pipeline from WDN topology to Chebnet input.
Figure 3: ChebNet architecture for pressure reconstruction and prediction in water distribution networks (WDNs).
Node102332133121121122
pp0.0560.0920.0920.0930.0930.1320.1320.1390.1699
Table 1: PageRank values for Net1 in Fig. 1 , computed using AA in ( 16 ) with αp=0.85\alpha_{p}=0.85 .
GNNTraining MSEValidation MSE
Reconstructor (PageRank)0.14520.1358
Reconstructor (Arbitrary)3.9104.407
Predictor (PageRank)0.40920.6831
Predictor (Arbitrary)2.8483.582
Table 2: Training and validation mean squared errors (MSE) for the four GNNs on Net1.
Figure 4: Training and validation loss on a logarithmic scale for the reconstructor (a) and predictor (b), using PageRank-Centrality-based sensor placement.
Figure 5: Residual distributions for the four GNNs on Net1, shown as violin plots over all junctions in (a) and over non-sensed junctions in (b).
Figure 6: True vs. reconstructed pressures in (a) and true vs. one-step-ahead predictions in (b).
Figure 7: Leakage simulation in Net1: (a) pressure trajectories with the leaky junction highlighted, (b) pressure differences relative to leak-free conditions.
Figure 8: Comparison leak alarms for PageRank-Centrality-based and arbitrary sensor placements on Net1.

Explore similar work

CardsList

  1. Joint Sensor Deployment and Physics-Informed Graph Transformer for Smart Grid Attack Detection

    Feb 16, 2026Mariam Elnour, Mohammad AlShaikh Saleh, Rachad Atat +6Power SystemsPhysics-Informed GNNs

    department of Electrical and Computer Engineering, Texas A&M University, College Station, TX 77840, USA · department of Electrical and Computer Engineering, Texas A&M University at Qatar, Doha, Qatar · department of Computer Science and Mathematics, Lebanese American University, Beirut 1102-2801, Lebanon +4

  2. Optimizing Sensor Placement for Hydrogen Leak Detection in Enclosed Infrastructure: A Comparative Study Using CFD-informed Genetic Algorithm and DeepSets Neural Surrogate

    Jul 20, 2026Fangnian Wang, Nicholas Tan Jerome, Thomas Jordan +1Surrogate-Assisted OptimizationDeep Sets

    Karlsruhe Institute of Technology, Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany.

  3. GENIE: Watermarking Graph Neural Networks for Link Prediction

    Jun 7, 2024Venkata Sai Pranav Bachina, Aaryan Ajay Sharma, Ankit Gangwal +1Graph Neural NetworksRobust Watermarking

    IIIT Hyderabad, India