journal article·2025-07Open access

Automatic GPR detection of grouting defects behind the tunnel shield segments based on wavelet coherence analysis combined with modified Res-RCNN

Dengyi Wang, Ming Peng, Liu Liu, Xiongyao Xie, Zhenming Shi, Yaoying Liang, Jian Shen, Qiyu Wu

Abstract

Ground penetrating radar (GPR), a widely used non-destructive testing technique for detecting grouting defects behind tunnel shield segments, faces challenges like steel rebar interference, low working efficiency, and expert interpretation reliance. To address these, this paper introduces an automated approach using wavelet coherence and a modified Res-RCNN. The approach employs wavelet coherence to transform the time-series GPR profile into the time-frequency images and reveal the weak defect reflections. Then, a modified Res-RCNN is applied to automatically extract the defect features from the wavelet coherence images. Finally, the post-processing and visualization automatically give an intuitive clear feature map that shows the location and probability of the grouting defects along the tunnel. The proposed methods are verified through full-size model tests with the aid of synthetic experiments to quantify their performance. The results show that wavelet coherence analysis improves the visibility of weak signals in (GPR) profiles, enabling their identification in the time-frequency domain by leveraging local coherence between adjacent signals and using phase information. The wavelet coherence analysis enables the observation of grouting defects behind tunnel shield segments with interferences of steel rebars. It can be applied even when the defect reflection is very weak, such as when the SNR is less than −40 dBs. The modified multi-task Res-RCNN, combined with post-processing and visualization, generates defect features including location and probability of existence. The network demonstrates superior training convergence and prediction accuracy due to information sharing between different task heads, compared to a two-classification network with the same Res-Net backbone. Through quantitative experiments in both model and synthetic tests, we recommend a trace interval of 15 to avoid the high coherence amplitude caused by two reflections out of same individual rebar.
Source: openalex_inverted_index · Quality: baseline

Research context

OpenAlex topics: Geophysical Methods and Applications · Seismic Imaging and Inversion Techniques · Rock Mechanics and Modeling

SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.

References (58)

Show source-provided references
1. Predicted and observed settlements induced by the mechanized tunnel excavation of metro line C near S. Giovanni station in Rome10.1016/j.tust.2019.01.022 ↗
2. Recent advances in the GPR detection of grouting defects behind shield tunnel segments10.3390/rs13224596 ↗
3. Application of ground penetrating radar in grouting evaluation for shield tunnel construction10.1016/j.tust.2009.09.006 ↗
4. Evaluation of grout behind the lining of shield tunnels using ground-penetrating radar in the Shanghai Metro Line, China10.1088/1742-2132/4/3/S03 ↗
5. Detection of grouting defects behind shield segments via shear wave ultrasound arrays: numerical and full-scale model experiments10.1016/j.tust.2024.106287 ↗
6. On the use of combined geophysical methods to assess water content and water conductivity of near-surface formations10.1016/S0022-1694(01)00588-1 ↗
7. Soil water content measurements at different scales: accuracy of time domain reflectometry and ground-penetrating radar10.1016/S0022-1694(01)00336-5 ↗
8. Application of ground penetrating radar methods in soil studies: a review10.1016/j.geoderma.2019.02.024 ↗
9. Wave-equation datuming before stack10.1190/1.1441620 ↗
10. Wave-equation datuming10.1190/1.1441010 ↗
11. Depth migration from irregular surfaces with depth extrapolation methods10.1190/1.1442947 ↗
12. Implicit static corrections in prestack migration of common-source data10.1190/1.1442887 ↗
13. Full-resolution 3D GPR imaging10.1190/1.1852780 ↗
14. 3-D ground-penetrating radar applied to fracture imaging in gneiss10.1190/1.1444026 ↗
15. Reconstruction method for extended depth-of-field optical diffraction tomography10.1016/j.ymeth.2017.10.005 ↗
16. Low-dose electron diffraction tomography (LD-EDT)10.1016/j.ultramic.2019.02.010 ↗
17. GPR trace analysis for coastal depositional environments of southern Brazil10.1016/j.jappgeo.2019.01.002 ↗
18. GPR imaging criteria10.1016/j.jappgeo.2019.04.008 ↗
19. Ground penetrating radar imaging of water leaks from buried pipes based on back-projection method10.1016/j.ndteint.2011.12.008 ↗
20. Image reconstruction: Part 1 – understanding filtered back projection, noise and image acquisition10.1016/j.jcct.2019.04.008 ↗
21. Maxwell curl equation datuming for GPR based on the Kirchhoff integral solution and application in a tunnel grouting test10.3997/1873-0604.2012029 ↗
22. Multiple suppression in GPR image for testing back-filled grouting within shield tunnel
23. Distributed fiber optic sensors for tunnel monitoring: A state-of-the-art review10.1016/j.jrmge.2024.01.008 ↗
24. GPR signal enhancement using band-pass and K–L filtering: a case study for the evaluation of grout in a shielded tunnel10.1088/1742-2132/10/3/034003 ↗
25. Anomaly detection via improvement of GPR image quality using ensemble restoration networks10.1016/j.autcon.2024.105552 ↗
26. Noise reduction algorithm of GPR data based on MMSE-PDS10.1002/nsg.12279 ↗
27. Identifying NDT methods for damage detection in concrete elements reinforced or strengthened with FRP
28. GPR detection localization of underground structures based on deep learning and reverse time migration10.1016/j.ndteint.2024.103043 ↗
29. Fast concrete crack depth detection using low frequency ultrasound array SH waves data10.1016/j.jappgeo.2024.105530 ↗
30. Designing a Distributed Sensing Network for Structural Health Monitoring of Concrete Tunnels: A Case Study10.1155/2024/6087901 ↗
31. Detection and localization of rebar in concrete by deep learning using ground penetrating radar10.1016/j.autcon.2020.103279 ↗
32. Simultaneous tunnel defects and lining thickness identification based on multi-tasks deep neural network from ground penetrating radar images10.1016/j.autcon.2022.104633 ↗
33. Tunnel lining crack detection model based on improved YOLOv510.1016/j.tust.2024.105713 ↗
34. Deep learning for geophysics: current and future trends10.1029/2021RG000742 ↗
35. Deep-learning inversion: A next-generation seismic velocity model building method10.1190/geo2018-0249.1 ↗
36. U-Net: Convolutional Networks for Biomedical Image Segmentation
37. Radar target detection with multi-task learning in heterogeneous environment
38. A practical guide to wavelet analysis10.1175/1520-0477(1998)079<0061:APGTWA>2.0.CO;2 ↗
39. A statistical analysis of morse wavelet coherence10.1109/TSP.2009.2033645 ↗
40. Deep residual learning for image recognition
41. Deep residual learning for image recognition: a survey10.3390/app12188972 ↗
42. Faster R-CNN: towards real-time object detection with region proposal networks10.1109/TPAMI.2016.2577031 ↗
43. Background removal, velocity estimation, and reverse-time migration: a complete GPR processing pipeline based on machine learning10.1109/TGRS.2023.3300276 ↗
44. Multi-task learning on the edge for effective gender, age, ethnicity and emotion recognition10.1016/j.engappai.2022.105651 ↗
45. Multi-task learning in natural language processing: an overview
46. Multi-task learning for simultaneous retrievals of passive microwave precipitation estimates and rain/no-rain classification10.1029/2022GL102283 ↗
47. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics
48. Multi-task learning for dense prediction tasks: a survey
49. A CUDA-based GPU engine for gprMax: open source FDTD electromagnetic simulation software10.1016/j.cpc.2018.11.007 ↗
50. Modelling ground penetrating radar by GprMax10.1016/j.conbuildmat.2005.06.007 ↗
51. gprMax: open source software to simulate electromagnetic wave propagation for Ground Penetrating Radar10.1016/j.cpc.2016.08.020 ↗
52. Machine learning for data-driven discovery in solid Earth geoscience10.1126/science.aau0323 ↗
53. U.Bin Waheed, T. Alkhalifah, E. Haghighat, C. Song, J. Virieux, PINNtomo: Seismic tomography using physics-informed neural networks, 2021.10.1016/j.cageo.2021.104833 ↗
54. Physics-informed neural networks (PINNs) for fluid mechanics: a review10.1007/s10409-021-01148-1 ↗
55. Physics-informed neural networks (PINNs) for wave propagation and full waveform inversions
56. Solving the frequency-domain acoustic VTI wave equation using physics-informed neural networks10.1093/gji/ggab010 ↗
57. Application of the cross wavelet transform and wavelet coherence to geophysical time series10.5194/npg-11-561-2004 ↗
58. A comparative study of exchange rates and order flow based on wavelet transform coherence and cross wavelet transform10.1016/j.econmod.2019.09.006 ↗