Role

First Author

Team

Yeji Kim, Matthew Bartos

Platforms

Python, EPANET, AWS data pipeline, Bayesian inference

Date

Jan 2024 – Dec 2025

Status: Manuscript in preparation — Kim, Y. & Bartos, M. (2026). Probabilistic parameter-estimation framework for discovery of pre-existing leaks in water distribution systems (target: Water Research).

Overview

This work develops a probabilistic framework for leak detection, localization, and system diagnostics in water distribution networks operating under uncertainty in pipe roughness, demand variability, and hydraulic losses. The framework is validated on a pilot-scale water distribution network in Unalakleet, Alaska (4 loops, ~740 population).

Approach

• Python-based hydraulic modeling and data assimilation using EPANET for network-wide state estimation
• Bayesian inference for leak detection and localization, propagating uncertainty in model parameters and field conditions
• SCADA API integration with wireless pressure sensors and adaptive sampling, on an AWS-based data pipeline
• Decision-support tools for pump scheduling and valve control validated against pilot deployment data

Expected Contributions

• Continuous monitoring and anomaly detection under realistic operating uncertainty
• Quantitative leak localization with confidence bounds
• Data-driven operational decisions for utility operators in resource-constrained networks