A topology-first and demand-driven sensor placement framework for hydraulic-state reconstruction in urban drainage networks.

Monitoring data from urban drainage sensor networks are fundamental for system-state perception and risk-management decision-making. However, existing sensor placement (SP) studies primarily focus on detection accuracy for specific monitoring tasks while overlooking the contribution of SP to system-wide perceptibility, and typically rely on full-node hydraulic state data that is rarely available in practice. To address these limitations, we proposed a topology-first and demand-driven end-to-end framework spanning sensor placement, perceptibility assessment, and monitoring applications. The framework introduces the ideal perceptual domain (IPD) to quantify the theoretical perceptibility of sensor configurations and guide demand-driven placement optimization, while a graph-based reconstruction model is used to estimate the actual perceptual domain (APD) and support monitoring tasks. In the case study, the sensor network designed to achieve a target IPD coverage of 90% required sensors at only 3.0% of the network nodes. Under a representative rainfall event, this configuration achieved an APD coverage of 86.7%, an IPD-APD consistency coefficient of 0.96, and flooding-node identification precision and recall values of 0.95 and 0.91, respectively. The proposed framework provides a practical and interpretable solution with potential engineering applications in sensor-network design for urban drainage systems under data-limited conditions, thereby supporting urban flood-risk monitoring and decision-making for drainage-infrastructure management.