3-D modeling of coupled geophysical fields for hidden hazards in the embankment dam using YOLO convolutional network model

Geophysical techniques are an efficient method for identifying hidden hazards in embankment dams due to the presence of significant physical differences in dam hazards. However, there is still a lack of sufficient understanding of the coupling relationship between different geophysical fields of different hazards, which hinders the detection accuracy of geophysical methods. By combining the theories of seepage field, stable electric field, electromagnetic wave field, and elastic wave field, a multi-physics coupling equation and boundary conditions for the hidden hazard model of embankment dams are established. Based on different geophysical methods, the geophysical responses of dam models with different water levels, hazard types, and sizes were modeled and used as the library of training samples. These samples were thoroughly trained using the YOLO convolutional network model, and training metrics like recall, accuracy, and loss curve were used to assess the quality. The results indicate that the GPR and seismic images are more accurate in identifying the hazard of the cavity, ant nest, and fracture, whereas the ERT is more successful in identifying the leakage risks. In addition, the location of the submerged surface can be accurately determined by the ERT, which is more sensitive to the water level.

 

This work was funded by the Science and Technology Project of Jiangxi Province (2022SKLS04, 2023KSG01008) and the National Natural Science Foundation of China (42374097)