Attention-Module-Guided Time-Lapse Leakage Plume Imaging Driven by LeakInv-CUNet GPR Inversion Framework
Accurate characterization of pipeline leakage by capturing time-lapse features and pathway orientations using ground penetrating radar (GPR) is crucial for optimizing the operational efficiency of water supply system and reducing water resource losses. However, achieving spatiotemporal leakage imagi...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11079582/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Accurate characterization of pipeline leakage by capturing time-lapse features and pathway orientations using ground penetrating radar (GPR) is crucial for optimizing the operational efficiency of water supply system and reducing water resource losses. However, achieving spatiotemporal leakage imaging remains challenging for current deterministic or probabilistic inversions due to the signal complexity, environmental interference, and computational burden. This paper develops LeakInv-CUNet, a novel attention-guided GPR inversion framework, to enable refined imaging of leakage plumes and their temporal-spatial evolution. To enhance network training, extensive GPR datasets are generated by augmenting simulated data and experimentally measured data, accounting for variations in injection orientation, plume dynamics, and subsurface media properties. By leveraging the dual advantages of the Convolutional Block Attention Module (CBAM) and U-Net architecture, the developed LeakInv-CUNet framework effectively extracts subtle leakage-induced response features, enabling refined imaging of leakage plumes and their orientations. Specifically, the training process utilizes the inversion result of GPR datasets corresponding to different leakage permittivity distributions, with feedbacks provided through functional mapping based on the Topp equation for water content distribution imaging. Special emphasis is placed on leakage features influenced by injection orientation and time-lapse characteristics. Simulations and on-site experiments demonstrate the framework’s superior noise robustness and practicality compared to conventional U-Net and Enc-Dec networks. With a mean absolute percentage error (MAPE) below 0.707%, structural similarity (SSIM) exceeding 0.924, and peak signal-to-noise ratio (PSNR) above 45.416, the attention-guided leakage imaging framework exhibits high efficiency and accuracy, showcasing its potential applications in early leakage warning and pipeline time-lapse monitoring. |
|---|---|
| ISSN: | 2169-3536 |