Answers are grounded in this paper and cite their sources.
Demonstrates rapid prediction of electromagnetic responses in core-wall dams, suggesting efficient hazard detection techniques.
Noninvasive geophysical methods, such as the transient electromagnetic method (TEM), are increasingly used to detect hazards in earth‐rock dams. However, analytical solutions for the TEM are only available for homogeneous half‐ and full‐space models; the boundary conditions of earth‐rock dams are nearly trapezoidal with one side adjacent to water. This unique structural configuration affects both the diffusion of the TEM field and the response signals. In this study, finite‐element forward modelling was applied to obtain representative observation curves from core‐wall dams. The influence patterns of various geometric parameters (dam height, slope ratio, core‐wall slope ratio and upstream water level) and resistivity parameters (dam body, foundation, core wall and upstream water resistivities) on the TEM response curves were analysed. Multiscale sensitivity analysis methods based on the cumulative area index and permutation approaches were applied to elucidate the dominant and temporal effects of different boundary condition parameters on the response curves. A feedforward neural network prediction model incorporating logarithmic time and dam boundary conditions was constructed, enabling highly accurate and rapid prediction of the TEM response curves. The results showed that among the geometric parameters, the core‐wall slope ratio significantly influenced the early‐stage response, whereas the upstream water level dominated the late‐stage response. Among the resistivity parameters, the dam body resistivity was the primary controlling factor in the early stage, dam foundation resistivity played a prominent role in the late stage, and significant parameter interactions were observed in the middle stage. The proposed neural network model produced rapid and highly accurate predictions of the observation curves using dam boundary‐condition inputs alone, reducing the forward modelling time from hours to seconds while maintaining relative errors below 0.6%. This research provides important theoretical support and technical references for the detection of core‐wall dam defects.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: