Abstract
Urban sewer detection is important for the proper conveyance of sanitary water to wastewater treatment plant prior to environmental discharge. An effective approach to address this important process still needs to be developed. This study introduced a novel data-driven approach to sewer detection utilizing in-sewer distributed temperature sensing (DTS) measurement combined with wavelet-based denoising of DTS data. It underlines that the effective denoising of DTS data, and consequently the accurate determination of DTS noise threshold, is pivotal to reliable sewer detection. DTS background noise is chiefly influenced by the threshold rescaling. A reliable DTS background noise threshold was found to be ± 0.25 °C in a field study, established with the threshold rescaling of a level-dependent estimation of level noise, and the associated threshold selection rule of heuristics threshold or minimum maximum variance. Deviation from this threshold could hamper the identification of true inflow or infiltration points. Applying the established threshold to the study site, our study identified two sewer problematic points including a groundwater infiltration point, and a clean water inflow point based on generated three-value image. Further interpretation of the three-value image revealed that both groundwater infiltration and clean water inflow into the sewer exhibited intermittent instead of constant behavior, which was due to time-variable water head difference associated with sewage discharge variation over the daily period and rainfall events. Thus, the methodology offers considerable potential for urban sewer detection, especially for its performance to capture intermittent sewer infiltrations and inflows without draining sewers.