A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Gradient optimization method for tunnel resistivity and chargeability joint inversion based on deep learning
https://orcid.org/0000-0002-1798-868X
https://orcid.org/0000-0002-1848-8134
Abstract The water inrush hazards have become one of the bottleneck problems that constrain tunnel construction. Ahead geological prospecting is the major tool to avoid geo-hazards and ensure safe, economical and efficient tunnelling. This work proposes a novel deep learning-based electrical method, which jointly inverses resistivity and chargeability to estimate water-bearing structures and water volume. Specifically, we design an encoder–decoder network, with one shared encoder to extract features from input data, two encoders to output resistivity and chargeability models, respectively, and an elaborate collinear regularization on the two outputs to reduce solution multiplicity. Moreover, our input is first transformed to the gradient domain to address the spatial incorrespondence issue between observation data and the electrical model. Compared with traditional linear inversion methods, our proposed method demonstrates superiority in locating and delineating anomalous bodies. The ablation study also shows that joint inversion of two parameters could benefit and outperform independent parameter inversion.
Gradient optimization method for tunnel resistivity and chargeability joint inversion based on deep learning
https://orcid.org/0000-0002-1798-868X
https://orcid.org/0000-0002-1848-8134
Abstract The water inrush hazards have become one of the bottleneck problems that constrain tunnel construction. Ahead geological prospecting is the major tool to avoid geo-hazards and ensure safe, economical and efficient tunnelling. This work proposes a novel deep learning-based electrical method, which jointly inverses resistivity and chargeability to estimate water-bearing structures and water volume. Specifically, we design an encoder–decoder network, with one shared encoder to extract features from input data, two encoders to output resistivity and chargeability models, respectively, and an elaborate collinear regularization on the two outputs to reduce solution multiplicity. Moreover, our input is first transformed to the gradient domain to address the spatial incorrespondence issue between observation data and the electrical model. Compared with traditional linear inversion methods, our proposed method demonstrates superiority in locating and delineating anomalous bodies. The ablation study also shows that joint inversion of two parameters could benefit and outperform independent parameter inversion.
Gradient optimization method for tunnel resistivity and chargeability joint inversion based on deep learning
https://orcid.org/0000-0002-1798-868X
https://orcid.org/0000-0002-1848-8134
Abstract The water inrush hazards have become one of the bottleneck problems that constrain tunnel construction. Ahead geological prospecting is the major tool to avoid geo-hazards and ensure safe, economical and efficient tunnelling. This work proposes a novel deep learning-based electrical method, which jointly inverses resistivity and chargeability to estimate water-bearing structures and water volume. Specifically, we design an encoder–decoder network, with one shared encoder to extract features from input data, two encoders to output resistivity and chargeability models, respectively, and an elaborate collinear regularization on the two outputs to reduce solution multiplicity. Moreover, our input is first transformed to the gradient domain to address the spatial incorrespondence issue between observation data and the electrical model. Compared with traditional linear inversion methods, our proposed method demonstrates superiority in locating and delineating anomalous bodies. The ablation study also shows that joint inversion of two parameters could benefit and outperform independent parameter inversion.
Gradient optimization method for tunnel resistivity and chargeability joint inversion based on deep learning
Jiang, Peng (author) / Liu, Benchao (author) / Tang, Yuting (author) / Liu, Zhengyu (author) / Pang, Yonghao (author)
2023-11-18
Article (Journal)
Electronic Resource
English
Deep Learning Joint Inversion of Electrical Data for Ahead-Prospecting in Tunneling
| DOAJ | 2023
Nonlinear Inversion of Electrode-Type Resistivity Measurements
| Online Contents | 1994