ATT-LCNN: An Attention-Enhanced Lightweight CNN for Acoustic Signal-Based Municipal Pipeline Leak Detection
Abstract
This study focuses on the recognition of acoustic signals from municipal pipeline leaks. An original end-to-end recognition algorithm (ATT-LCNN) based on the fusion of an attention mechanism and an improved lightweight convolutional neural network (CNN) is proposed, targeting the bottlenecks of traditional methods in minor leak detection and poor generalization. The algorithm employs a preprocessing flow of "DC component removal → improved Kalman filtering → signal normalization → segmented windowing", constructs a multi-dimensional fusion feature set across time, frequency, and time-frequency domains, and embeds an improved SE-Net attention module with depth-wise separable convolution for lightweight and high-accuracy feature learning. Experiments are conducted on a self-built simulated pipeline dataset (1250 samples, covering 3 pipe diameters, 3 leakage levels, 5 propagation distances, 5 noise intensities, with 750 leak samples and 500 non-leak samples), with 5- fold cross-validation to ensure statistical robustness. Performance is benchmarked against traditional machine learning methods (SVM, Random Forest) and standard deep learning models (vanilla CNN, LSTM). The results show that the ATT-LCNN algorithm achieves a 98.7% detection rate (recall), 98.2% precision, 98.4% F1-score for minor leaks, maintains an overall accuracy of over 97.2% at a propagation distance of 20m, and reaches 97.8% accuracy under 60 dB strong noise conditions. The model has only 1.2M parameters, with a single-sample inference time of 2.3ms on CPU, meeting embedded deployment requirements. The repeated detection accuracy fluctuation range is only 0.8%, with stable performance across various pipe diameters and materials. Compared with baseline methods, ATT-LCNN improves minor leak detection accuracy by 6.3%-12.1% under high-noise conditions, balancing recognition accuracy, anti-interference capability, and real-time performance, providing technical support for intelligent leak detection in municipal pipeline networks.References
[1]Li, S., Dai, Z., Liu, M., & Cai, M. (2022). Leak identification method of water supply pipeline based on compressed sensing and least squares twin support vector machine. IEEE Sensors Journal, 23(7), 7115-7128. doi: 10.1109/JSEN.2022.3211343.
[2]Han, Y., Feng, X., & Todd, M. D. (2023). A novel methodology for quantitative identification of pipeline leakage and negative pressure wave velocity. Structural Health Monitoring, 22(4), 2267-2279. https://doi.org/10.1177/14759217221123403
[3]Ullah, N., Siddique, M. F., Ullah, S., Ahmad, Z., & Kim, J. M. (2024). Pipeline leak detection system for a smart city: leveraging acoustic emission sensing and sequential deep learning. Smart Cities, 7(4), 2318-2338. https://doi.org/10.3390/smartcities7040091
[4]Zhao, S. L., Liu, S. G., Zhao, D., Qiu, B., Hong, Z., & Shi, X. L. (2024). Experimental study on leakage location of branched pipelines based on acoustic propagation characteristics. Proceedings of the Institution of Mechanical Engineers, Part E: Journal of Process Mechanical Engineering, 238(5), 2186-2196. https://doi.org/10.1177/09544089231158953
[5]Feng, C., Zhao, J., Ran, Q., Qu, M., & Guo, Z. (2024). Acoustic-based approach for micro-leakage detection and localization in water supply pipelines. Environmental Science: Water Research & Technology, 10(8), 1881-1889. https://doi.org/10.1039/D3EW00686G
[6]Zhu, Y., Lang, X., Zhang, L., & Cai, Z. (2023). Leak localization method of jet fuel pipeline based on second-generation wavelet transform and short-time energy time delay estimation. IEEE Sensors Journal, 23(3), 2823-2832. doi: 10.1109/JSEN.2022.3233660.
[7]Kothandaraman, M., Law, Z., Ezra, M. A., Pua, C. H., & Rajasekaran, U. (2022). Water pipeline leak measurement using wavelet packet-based adaptive ICA. Water Resources Management, 36(6), 1973-1989. https://doi.org/10.1007/s11269-022-03119-y
[8]Saravanabalaji, M., Sivakumaran, N., Ranganthan, S., & Athappan, V. (2023). Acoustic signal based water leakage detection system using hybrid machine learning model. Urban Water Journal, 20(9), 1123-1139. https://doi.org/10.1080/1573062X.2023.2239782
[9]Chen, C., Hao, P., Liu, J., Ni, L., Jiang, J., Diao, X., & Gu, W. (2023). Pipeline leak AE signal denoising based on improved SSA-K-α index-VMD-MD. IEEE Sensors Journal, 23(21), 26177-26194. doi: 10.1109/JSEN.2023.3314166.
[10]Debroy, P., Smarandache, F., Majumder, P., Majumdar, P., & Seban, L. (2025). OPA-IF-Neutrosophic-TOPSIS Strategy under SVNS Environment Approach and Its Application to Select the Most Effective Control Strategy for Aquaponic System. Informatica, 36(1), 1-32. doi:10.15388/24-INFOR583
[11]Long, Y., Zhang, J., Huang, S., Peng, L., Wang, W., Wang, S., & Zhao, W. (2022). A novel crack quantification method for ultra-high-definition magnetic flux leakage detection in pipeline inspection. IEEE Sensors Journal, 22(16), 16402-16413. doi: 10.1109/JSEN.2022.3190684.
[12]Priyanka, E. B., & Thangavel, S. (2022). Multi-type feature extraction and classification of leakage in oil pipeline network using digital twin technology. Journal of Ambient Intelligence and Humanized Computing, 13(12), 5885-5901. https://doi.org/10.1007/s12652-022-03818-9
[13]Lyu, C., Zhang, M., Li, B., Liu, Y., & Lin, X. (2022). High reliability pipeline leakage detection based on machine vision in complex industrial environment. IEEE Sensors Journal, 22(21), 20748-20760. doi: 10.1109/JSEN.2022.3206456.
[14]Abdelmageed, S., Tariq, S., Boadu, V., & Zayed, T. (2022). Criteria-based critical review of artificial intelligence applications in water-leak management. Environmental Reviews, 30(2), 280-297. https://doi.org/10.1139/er-2021-0046
[15]Cervenka, M., Kohout, J., & Lipus, B. (2024). A Novel Radial Basis Function Description of a Smooth Implicit Surface for Musculoskeletal Modelling. Informatica, 35(4), 721-750. doi:10.15388/24-INFOR571
[16]Sousa, D. P., Du, R., Mairton Barros da Silva Jr, J., Cavalcante, C. C., & Fischione, C. (2023). Leakage detection in water distribution networks using machine-learning strategies. Water Supply, 23(3), 1115-1126. https://doi.org/10.2166/ws.2023.054
[17]Lu, S., Zhou, T., Wang, C., Lin, Z., & Yi, G. (2023). An internal detector positioning method in oil pipelines using vibration signal. IEEE Sensors Journal, 23(12), 13411-13421. doi: 10.1109/JSEN.2023.3273534.
[18]Huynh, C., Hibert, C., Jestin, C., Malet, J. P., Clément, P., & Lanticq, V. (2022). Real‐time classification of anthropogenic seismic sources from distributed acoustic sensing data: Application for pipeline monitoring. Seismological Society of America, 93(5), 2570-2583. https://doi.org/10.1785/0220220078
[19]Goyal, N., Nain, M., Singh, A., Abualsaud, K., Alsubhi, K., Ortega-Mansilla, A., & Zorba, N. (2022). An anchor-based localization in underwater wireless sensor networks for industrial oil pipeline monitoring. IEEE Canadian Journal of Electrical and Computer Engineering, 45(4), 466-474. doi: 10.1109/ICJECE.2022.3206275.
[20]Lu, H., Xi, D., Xiang, Y., Su, Z., & Cheng, Y. F. (2025). Vehicle–canine collaboration for urban pipeline methane leak detection. Nature Cities, 2(4), 336-343. https://doi.org/10.1038/s44284-024-00183-w
[21]Abuhatira, A. A., Salim, S. M., & Vorstius, J. B. (2023). CFD-FEA based model to predict leak-points in a 90-degree pipe elbow. Engineering with Computers, 39(6), 3941-3954. https://doi.org/10.1007/s00366-023-01853-4
[22]Gemeinhardt, H., & Sharma, J. (2023). Machine-learning-assisted leak detection using distributed temperature and acoustic sensors. IEEE Sensors Journal, 24(2), 1520-1531. doi: 10.1109/JSEN.2023.3337284.
DOI:
https://doi.org/10.31449/inf.v50i12.14020Keywords:
Municipal pipeline network, leak detection, acoustic signal, ATT-LCNN algorithm, attention mechanism, machine learningDownloads
Published
How to Cite
Issue
Section
License
Authors retain copyright in their work. By submitting to and publishing with Informatica, authors grant the publisher (Slovene Society Informatika) the non-exclusive right to publish, reproduce, and distribute the article and to identify itself as the original publisher.
All articles are published under the Creative Commons Attribution license CC BY 3.0. Under this license, others may share and adapt the work for any purpose, provided appropriate credit is given and changes (if any) are indicated.
Authors may deposit and share the submitted version, accepted manuscript, and published version, provided the original publication in Informatica is properly cited.
