Rapid quantification of the surface overflow and underground infiltration in sewer pipes based on computer vision and continuous optimization

Abstract

The overloading of the sewer network caused by unwarranted infiltration of stormwater may lead to waterlogging and environmental pollution. The accurate identification of infiltration and surface overflow is essential to predict and reduce these risks. To retrieve the limitations of infiltration estimation and the failure of surface overflow perception using the common stormwater management model (SWMM), a surface overflow and underground infiltration (SOUI) model is proposed to estimate the infiltration and overflow. First, the precipitation, water level of the manhole, surface water depth and images of the overflowing point, and volume at the outfall are collected. Then, the surface waterlogging area is identified based on computer vision to reconstruct the local digital elevation model (DEM) by spatial interpolation, and the relationship between the waterlogging depth, area and volume is established to identify the real-time overflow. Next, a continuous genetic algorithm optimization (CT-GA) model is proposed for the underground sewer system to determine the inflow rapidly. Finally, surface and underground flow estimations are combined to perceive the state of the urban sewer network accurately. The results show that, compared with the common SWMM simulation, the accuracy of the water level simulation is improved by 43.5% during the rainfall period, and the time cost of the computational optimization is reduced by 67.5%. The proposed method can effectively diagnose the operation state and overflow risk of the sewer networks in real time during rainfall seasons.