An AIoT‐Based Edge Intelligence System for Drainage Pipe Blockage Identification Using Acoustic Emission Sensing
Year: 2026
Type: article
Abstract: ABSTRACT This paper presents an Artificial Intelligence of Things (AIoT) framework for identifying current drainage pipe blockage conditions using acoustic emission sensing and edge‐oriented machine learning. The proposed system integrates a resonant acoustic‐emission sensor, signal conditioning electronics, and an STM32‐based data‐acquisition platform. Frequency‐domain features extracted from the acquired signals were used to develop lightweight classifiers using Edge Impulse and an offline Python‐based evaluation pipeline. Model development and optimisation were supported by the Edge Impulse platform. The approach was evaluated on a public benchmark dataset and laboratory‐scale drainage testbed as an embedded proof‐of‐concept under controlled single‐flow conditions. The offline Python‐based four‐class dataset was evaluated using four‐fold collection‐group validation, with all frames from the same experimental collection retained within the same partition. Experimental results demonstrate reliable discrimination between empty, partially blocked, and near‐fully blocked pipe conditions, under the tested laboratory conditions. Real‐time on‐device operation of the three‐class Edge Impulse model using 266 features was validated using 30 consecutive live frames, with all inference cycles completing successfully, and a mean capture‐to‐result latency of 41.395 ms. The proposed AIoT framework enables condition monitoring for drainage infrastructure by shifting from reactive inspection to continuous, data‐driven monitoring at the network edge.
DOI: 10.1049/aie2.70029
Institution: Hong Kong University of Science and Technology
Language: English
Subfield: Civil and Structural Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7229602866
