Pipe roughness assessment of water distribution networks based on physics-data fusion

As one of critical urban infrastructures, the basic function of water distribution networks (WDNs) is to provide essential water for residential and industrial use. Accurate assessment of pipe roughness coefficient is crucial for analyzing the functional performance of WDNs under operation and disasters, such as earthquakes. Existing assessment methods for pipe roughness coefficient face challenges including: (1) physically, governing laws are not fully understood; (2) data-wise, lacking direct pipe roughness data, and head/flow monitoring does not cover all nodes. Therefore, relying solely on physical or data-driven methods is insufficient to assess pipe roughness coefficient. To this end, this paper proposes a novel method to assess pipe roughness coefficient based on physics-data fusion. Specifically, this method is achieved by embedding physical constraint mechanisms into data expression models. This approach leverages the strengths of both physical and data methods, effectively overcoming issues of insufficient monitoring data and underdetermined problems in assessment of pipe roughness coefficient. It enables accurate and efficient assessment even when only total water consumption is known, without requiring individual nodal consumption. The proposed method is applied to a small-scale WDN and a real-world WDN for case study. The results show that the proposed method can achieve accurate and efficient assessment of pipe roughness coefficient of the case WDNs.