Year: 2024
Type: conference-paper
Abstract: This paper presents a novel approach to detecting small residential gas leaks by developing and testing a leakage detection algorithm, termed the Weeg algorithm. Our proposed methodology combines a Seq2Seq generative model with a CNN-based classification network, enabling a robust and efficient approach to this task. The evaluation was performed on a dataset containing 86 leakage samples and 1103 non-leakage samples. The Weeg algorithm demonstrated significant improvements, achieving an accuracy of 90.2% and an F1 score of 0.458, while reducing false positive rates compared to conventional logic-based algorithms.
Authors: Alex Da Yin, Weiping Ni, Fangcun Shu
Language: English
Subfield: Civil and Structural Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4411747384