Sensor Placement for Early Contamination Detection in Water Distribution Networks Using Grid Search and Bayesian Optimization

Fadelis Sukya; Ary Mazharuddin Shiddiqi; Ervin Nurhayati

doi:10.1109/SIML69834.2026.11621545

Sensor Placement for Early Contamination Detection in Water Distribution Networks Using Grid Search and Bayesian Optimization

Fadelis Sukya^*
, Ary Mazharuddin Shiddiqi
, Ervin Nurhayati

^*Corresponding author for this work

Departement of Informatics
Institut Teknologi Sepuluh Nopember
Department of Technology Information
State Polytechnic of Malang

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Optimal sensor placement for contamination detection in water distribution networks (WDNs) is critical for early warning and public health protection. Deep Reinforcement Learning (DRL), particularly Deep Q-Network (DQN), provides a flexible framework for learning sensor placement policies, but its performance is highly sensitive to hyperparameter settings. This paper investigates Bayesian Optimization (BO) as a sampleefficient approach for tuning DQN hyperparameters in WDN sensor placement. Experiments on four benchmark networks using EPANET simulations show that BO reduces tuning time by approximately 24-77% while maintaining detection coverage and detection time comparable to Grid Search (GS). These results demonstrate that BO provides an effective strategy for reducing computational cost in DRL-based sensor placement without compromising detection performance.

Original language	English
Title of host publication	SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9798331552237
DOIs	https://doi.org/10.1109/SIML69834.2026.11621545
Publication status	Published - 2026
Event	2026 International Conference on Smart Computing, Internet of Things, and Machine Learning, SIML 2026 - Hybrid, Surakarta, Indonesia Duration: 10 Jun 2026 → 11 Jun 2026

Publication series

Name	SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning

Conference

Conference	2026 International Conference on Smart Computing, Internet of Things, and Machine Learning, SIML 2026
Country/Territory	Indonesia
City	Hybrid, Surakarta
Period	10/06/26 → 11/06/26

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Keywords

Deep Q-Network
Sensor placement
contaminant detection
hyperparameter optimization
water distribution networks

Access to Document

10.1109/SIML69834.2026.11621545

Cite this

Sukya, F., Shiddiqi, A. M., & Nurhayati, E. (2026). Sensor Placement for Early Contamination Detection in Water Distribution Networks Using Grid Search and Bayesian Optimization. In SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning (SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/SIML69834.2026.11621545

Sukya, Fadelis ; Shiddiqi, Ary Mazharuddin ; Nurhayati, Ervin. / Sensor Placement for Early Contamination Detection in Water Distribution Networks Using Grid Search and Bayesian Optimization. SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning. Institute of Electrical and Electronics Engineers Inc., 2026. (SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning).

@inproceedings{61e1db52057f4373a86f657f5d246f4a,

title = "Sensor Placement for Early Contamination Detection in Water Distribution Networks Using Grid Search and Bayesian Optimization",

abstract = "Optimal sensor placement for contamination detection in water distribution networks (WDNs) is critical for early warning and public health protection. Deep Reinforcement Learning (DRL), particularly Deep Q-Network (DQN), provides a flexible framework for learning sensor placement policies, but its performance is highly sensitive to hyperparameter settings. This paper investigates Bayesian Optimization (BO) as a sampleefficient approach for tuning DQN hyperparameters in WDN sensor placement. Experiments on four benchmark networks using EPANET simulations show that BO reduces tuning time by approximately 24-77\% while maintaining detection coverage and detection time comparable to Grid Search (GS). These results demonstrate that BO provides an effective strategy for reducing computational cost in DRL-based sensor placement without compromising detection performance.",

keywords = "Deep Q-Network, Sensor placement, contaminant detection, hyperparameter optimization, water distribution networks",

author = "Fadelis Sukya and Shiddiqi, \{Ary Mazharuddin\} and Ervin Nurhayati",

note = "Publisher Copyright: {\textcopyright} 2026 IEEE.; 2026 International Conference on Smart Computing, Internet of Things, and Machine Learning, SIML 2026 ; Conference date: 10-06-2026 Through 11-06-2026",

year = "2026",

doi = "10.1109/SIML69834.2026.11621545",

language = "English",

series = "SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning",

address = "United States",

}

Sukya, F, Shiddiqi, AM & Nurhayati, E 2026, Sensor Placement for Early Contamination Detection in Water Distribution Networks Using Grid Search and Bayesian Optimization. in SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning. SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning, Institute of Electrical and Electronics Engineers Inc., 2026 International Conference on Smart Computing, Internet of Things, and Machine Learning, SIML 2026, Hybrid, Surakarta, Indonesia, 10/06/26. https://doi.org/10.1109/SIML69834.2026.11621545

Sensor Placement for Early Contamination Detection in Water Distribution Networks Using Grid Search and Bayesian Optimization. / Sukya, Fadelis; Shiddiqi, Ary Mazharuddin ; Nurhayati, Ervin.
SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning. Institute of Electrical and Electronics Engineers Inc., 2026. (SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Sensor Placement for Early Contamination Detection in Water Distribution Networks Using Grid Search and Bayesian Optimization

AU - Sukya, Fadelis

AU - Shiddiqi, Ary Mazharuddin

AU - Nurhayati, Ervin

PY - 2026

Y1 - 2026

N2 - Optimal sensor placement for contamination detection in water distribution networks (WDNs) is critical for early warning and public health protection. Deep Reinforcement Learning (DRL), particularly Deep Q-Network (DQN), provides a flexible framework for learning sensor placement policies, but its performance is highly sensitive to hyperparameter settings. This paper investigates Bayesian Optimization (BO) as a sampleefficient approach for tuning DQN hyperparameters in WDN sensor placement. Experiments on four benchmark networks using EPANET simulations show that BO reduces tuning time by approximately 24-77% while maintaining detection coverage and detection time comparable to Grid Search (GS). These results demonstrate that BO provides an effective strategy for reducing computational cost in DRL-based sensor placement without compromising detection performance.

AB - Optimal sensor placement for contamination detection in water distribution networks (WDNs) is critical for early warning and public health protection. Deep Reinforcement Learning (DRL), particularly Deep Q-Network (DQN), provides a flexible framework for learning sensor placement policies, but its performance is highly sensitive to hyperparameter settings. This paper investigates Bayesian Optimization (BO) as a sampleefficient approach for tuning DQN hyperparameters in WDN sensor placement. Experiments on four benchmark networks using EPANET simulations show that BO reduces tuning time by approximately 24-77% while maintaining detection coverage and detection time comparable to Grid Search (GS). These results demonstrate that BO provides an effective strategy for reducing computational cost in DRL-based sensor placement without compromising detection performance.

KW - Deep Q-Network

KW - Sensor placement

KW - contaminant detection

KW - hyperparameter optimization

KW - water distribution networks

UR - https://www.scopus.com/pages/publications/105047220857

U2 - 10.1109/SIML69834.2026.11621545

DO - 10.1109/SIML69834.2026.11621545

M3 - Conference contribution

AN - SCOPUS:105047220857

T3 - SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning

BT - SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2026 International Conference on Smart Computing, Internet of Things, and Machine Learning, SIML 2026

Y2 - 10 June 2026 through 11 June 2026

ER -

Sukya F, Shiddiqi AM , Nurhayati E. Sensor Placement for Early Contamination Detection in Water Distribution Networks Using Grid Search and Bayesian Optimization. In SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning. Institute of Electrical and Electronics Engineers Inc. 2026. (SIML 2026 - International Conference on Smart Computing, Internet of Things, and Machine Learning). doi: 10.1109/SIML69834.2026.11621545

Sensor Placement for Early Contamination Detection in Water Distribution Networks Using Grid Search and Bayesian Optimization

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this