Artificial Intelligence of Things (AIoT)-oriented framework for blockage assessment at cross-drainage hydraulic structures

Abstract

Blockage of cross-drainage hydraulic structures is a key factor to be recursively assessed within flood management domain because of its involvement in originating flash floods. However, this issue has been least addressed in literature because of highly complex nature of blockage formulation and unavailability of relevant data to investigate the hydraulic impacts of blockage. Given the success of data-driven approaches in dealing with complex real-world problems within water resources management domain (e.g. water depth estimation, ground water prediction, water demand forecasting, drainage pipe detection, sewer fault detection), this paper proposes the idea of an Artificial Intelligence of Things (AIoT)-oriented framework for the assessment of blockage at cross-drainage hydraulic structures to facilitate the flood management agencies in better managing the blockage related issues. The proposed framework makes use of multiple AI approaches (e.g. image classification, object detection, object segmentation, regression, end-to-end deep learning) to assess visual and hydraulic blockage at a given structure using both visual and hydraulic data coming from sensors (i.e. camera, water level sensors, inlet discharge sensor, surface velocity sensor). As the output, the framework provides the information about the blockage status, estimation of percentage visual blockage and estimation of percentage hydraulic blockage. This information will be used by the flood management agencies in maintaining the hydraulic structures and to incorporate the blockage in hydraulic structures design process.