Search for a command to run...
Urban water distribution networks (WDNs) are intended to reliably supply safe drinking water to users, while ensuring an economically and environmentally sustainable management of the supply infrastructure. However, WDNs are prone to water losses, mostly due to ageing components and insufficient maintenance. In the last decades, substantial research efforts have focused on developing methods for water loss reduction, targeting an optimal management of the WDNs infrastructure and quantifiable economic and water savings. Proposed approaches include programmed replacement of pipes, pressure management by means of optimally deployed pressure reduction valves, and prompt leak detection and subsequent localisation by means of sensor data and automated algorithms.Beyond their direct impact on water supply efficiency, leakages in WDNs also contribute to an underestimated and, so far, understudied problem, namely the hydrogeological instability in urban environments, which can manifest as ground subsidence, surface deformation, or the formation of sinkholes. Such processes can cause substantial damage to infrastructure, disrupt traffic circulation, damage vehicles, compromise underground utilities, weaken the structural integrity of buildings, and, in the most severe cases, pose a threat to public safety. Incorporating planning and management strategies aimed at reducing hydrogeological risks associated with pipe leaks in WDNs is thus key to fostering the resilience of WDNs within the urban environment, besides water supply efficiency and reliability. This study presents an optimal sensor placement framework for mitigating the risk of Hydrogeological Disruption from Leakage (HDL).The framework sequentially combines a spatial risk zonation approach with an optimal pressure sensor placement for accurate leak localisation, where sensor placement is driven by the objective of maximising leak localisation accuracy in the most vulnerable and exposed areas of the city. The optimisation process makes use of an evolutionary algorithm (GA from the package Pymoo) where candidate pressure sensor configurations are evaluated across different leak scenarios. The objective function combines the minimum hydraulic paths between actual and predicted leaks with weights representative of different risk levels, while the leak scenarios are produced with the hydraulic model included in the WNTR library under the assumption that a single active leak occurs along pipes. Localisation is based on a sensitivity matrix that maps the pressure response of the network nodes to specific leak scenarios, characterised by their location and magnitude. To avoid the introduction of binarisation thresholds, which would burden the optimisation process, a threshold-independent cosine similarity measure is adopted to evaluate the directional consistency between the pressure residual vector and the sensitivity vectors.Our framework is tested on the L-Town benchmark WDN, a realistic WDN inspired by a real-world infrastructure, for which risk zonation is assessed considering exposure (qualitative exposed value of real estate assets, population density, and road network importance) and hazard (operational and pipe intrinsic factors). The numerical results demonstrate that the approach is effective in choosing the set of sensors that reduces the distance between the predicted leak localisation and the actual leak point, rewarding urban areas characterised by a higher potential HDL risk.