Scholarlink

<p>This study examines the fundamental limits of pressure-based leak detection in water networks operating under pressure-driven demand (PDD) conditions. While prior research has emphasized improving detection performance through advanced models and feature engineering, the intrinsic detectability of leak events remains understudied. Here, the focus is shifted from performance optimization to analyzing the information content of hydraulic responses. A three-part framework is developed to evaluate detectability through data-scaling behavior, feature-structure analysis, and the probabilistic characteristics of model outputs. Results show that detection performance rapidly saturates with increasing data, that predictive information is diffusely distributed across variables, and that model outputs exhibit substantial overlap between leak and non-leak states. These patterns support the concept of hydraulic indistinguishability, indicating that leak-induced pressure variations cannot be reliably separated from demand-driven fluctuations under PDD conditions. The findings highlight inherent constraints in pressure-only monitoring and underscore the need for detectability-aware modeling and multi-source sensing approaches. This work offers a refined perspective on the structural limits of leak detection in hydraulic systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

On the Fundamental Limits of Pressure-Based Leak Detection under Pressure-Driven Demand Conditions

  • 期刊论文

  • 发表日期：

    2026年5月23日

查看全文
  • Ali Maleki,
  • Mohammadreza Alizadeh Tataki Afshar,
  • Amin Nejat

摘要

This study examines the fundamental limits of pressure-based leak detection in water networks operating under pressure-driven demand (PDD) conditions. While prior research has emphasized improving detection performance through advanced models and feature engineering, the intrinsic detectability of leak events remains understudied. Here, the focus is shifted from performance optimization to analyzing the information content of hydraulic responses. A three-part framework is developed to evaluate detectability through data-scaling behavior, feature-structure analysis, and the probabilistic characteristics of model outputs. Results show that detection performance rapidly saturates with increasing data, that predictive information is diffusely distributed across variables, and that model outputs exhibit substantial overlap between leak and non-leak states. These patterns support the concept of hydraulic indistinguishability, indicating that leak-induced pressure variations cannot be reliably separated from demand-driven fluctuations under PDD conditions. The findings highlight inherent constraints in pressure-only monitoring and underscore the need for detectability-aware modeling and multi-source sensing approaches. This work offers a refined perspective on the structural limits of leak detection in hydraulic systems.