Análisis estadístico de factibilidad para sistemas automáticos de localización de fugas mediante sensores de presión
Loading...
Publication date
Advisors
Research group
Center
Abstract
Este trabajo analiza la factibilidad del problema de localización de fugas en redes hidráulicas basados en datos de presión. Se
comparan redes de agua potable (malladas) y de riego (jerárquicas) ante la aplicación de ruido uniforme y gaussiano. Se generan
firmas de fuga mediante EPANET y se aplican técnicas de agrupamiento y simulación Monte Carlo para estudiar el solapamiento
entre regiones. Se propone una métrica probabilística de indistinguibilidad que cuantifica la ambigüedad en la detección. Los resultados
muestran que las redes jerárquicas presentan mayor vulnerabilidad al ruido, con regiones altamente ambiguas incluso bajo
niveles peque˜nos de perturbaci´on. El histograma de solapamientos revela diferencias estructurales entre redes y permite clasificar
su robustez. Se sugiere que fragmentar redes extensas en subregiones independientes puede mitigar el solapamiento y mejorar la
discriminabilidad bajo incertidumbre
This work analyzes the feasibility of locating leaks in hydraulic networks under measurement uncertainty using simulated pressure. Potable water (meshed) and irrigation (hierarchical) networks are compared under uniform and Gaussian noise. Leak signatures are generated using EPANET, and clustering and Monte Carlo methods are applied to study the overlap of response regions. A probabilistic indistinguishability metric is proposed to quantify ambiguity. Results show that hierarchical networks are more vulnerable to noise, with highly ambiguous regions even under low perturbation levels. The overlap histogram reveals structural differences between networks and enables classification of their diagnostic robustness. It is suggested that fragmenting large networks into independent sub-regions can mitigate overlap and improve discriminability under uncertainty
This work analyzes the feasibility of locating leaks in hydraulic networks under measurement uncertainty using simulated pressure. Potable water (meshed) and irrigation (hierarchical) networks are compared under uniform and Gaussian noise. Leak signatures are generated using EPANET, and clustering and Monte Carlo methods are applied to study the overlap of response regions. A probabilistic indistinguishability metric is proposed to quantify ambiguity. Results show that hierarchical networks are more vulnerable to noise, with highly ambiguous regions even under low perturbation levels. The overlap histogram reveals structural differences between networks and enables classification of their diagnostic robustness. It is suggested that fragmenting large networks into independent sub-regions can mitigate overlap and improve discriminability under uncertainty
Unesco Subjects
Bibliographic citation
Neto, F.M., Manzano, J.M., Orihuela, L. 2025. Heuristic feasibility analysis for automatic leak location systems using pressure sensors. XX Simposio CEA de Control Inteligente, Huelva (Spain)







