Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Building and Real Estate | en_US |
| dc.contributor.advisor | Zayed, Tarek (BRE) | en_US |
| dc.creator | Liu, Rongsheng | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/13621 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | Smart acoustic leak diagnosis in water distribution networks : leveraging machine learning approaches | en_US |
| dcterms.abstract | Water leakage in the water distribution network (WDN) is a significant issue leading to infrastructure damage, economic loss, and potential health hazards, emphasizing effective solutions. Among the various leak detection and localization methods, acoustic-based approaches are widely employed for their comprehensive capabilities. However, their effectiveness depends heavily on signal quality and is susceptible to external factors, requiring substantial prior knowledge. Given the remarkable performance of machine learning (ML) techniques, they have been introduced into leak diagnosis, offering significant benefits while also introducing new challenges. | en_US |
| dcterms.abstract | This study aims to propose an acoustic ML leak diagnosis framework and complete the following objectives: 1) Propose a generative approach that augments the leak detection dataset, addressing the data scarcity-related problems for WDNs. 2) Establish an explainable leak detection model, enhancing the interpretability and entailing comprehensive analysis of collected signals. 3) Develop an effective and robust time-series leak detection model for WDN, and facilitate smart leak detection in real scenarios. 4) Develop a robust time-delay estimation deep learning leak localization model for WDN. | en_US |
| dcterms.abstract | The key findings can be concluded as four points. First, an innovative data augmentation approach has been proposed to enhance the vibroacoustic datasets. The generated samples have been demonstrated to have similar acoustic features to real samples, contributing to improving leak detection accuracy. Second, an explainable deep learning framework has been proposed to enhance interpretability during leak detection modeling and deepen the understanding of the decision-making mechanism of deep learning models. | en_US |
| dcterms.abstract | Third, the Time-Transformer leak detection model has been proposed to enhance detection accuracy. The proposed model utilizes the attention mechanism, capturing the temporal patterns inherent in signals. The empirical results demonstrate that the Time-Transformer outperforms alternative models, achieving 88.46% accuracy in out-of-sample validation. Fourth, a time-delay-based leak localization model has been proposed. The model harnesses the complex pattern recognition capabilities of deep learning techniques to deduce the time delay of signal pairs. The deep learning leak localization model exhibits reduced prediction error compared to basic cross-correlation, particularly in low signal-to-noise ratio (SNR) conditions. | en_US |
| dcterms.abstract | This study contributes to the development of ML-based acoustic leak diagnosis. Theoretically, it expands the knowledge of water leak diagnosis by providing a comprehensive review of ML applications in acoustic leak detection and proposing advanced data augmentation techniques for acoustic data. Additionally, this study is the first to reveal the underlying mechanisms of acoustic leak detection models, thereby enhancing their interpretability. The experiments validate the effectiveness of one- and two-dimensional data for leak detection and introduce a novel deep-learning model to estimate the time delay of signal pairs for leak localization. Practically, this study contributes to developing a smart acoustic leak detection system, improving the accuracy and reliability of leak diagnosis, reducing maintenance costs and resource waste, advancing the understanding and acceptance of ML techniques for leak detection, and facilitating the progress of smart leak management systems. | en_US |
| dcterms.extent | xiv, 227 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2025 | en_US |
| dcterms.educationalLevel | Ph.D. | en_US |
| dcterms.educationalLevel | All Doctorate | en_US |
| dcterms.accessRights | open access | en_US |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/13621