Fault-slips cause seismic events or earthquakes and are simulated in underground mines using the Mohr–Coulomb criterion or the linear slip-weakening law, which are inefficient representations of dynamic fault rupture. The residual shear strength and post-failure frictional weakening of a fault are not considered in Mohr–Coulomb friction, and the linear slip-weakening (LSW) lacks experimental support and assumes a linear reduction of the friction coefficient. Both rupture models have been widely deployed due to the ease of numerical implementation. To investigate a more robust fault-rupture model, we employed an advanced computational seismology solver, SPECFEMD3, to simulate the rate-and-state friction (RSF) law. The rate-and-state dependent friction studies are based on Ruina’s slip variable. We find that the linear slip-weakening model overestimates rock deformation’s magnitude: peak particle velocity is 0.81 m/s compared to 0.31 m/s for the rate-and-state model at the tunnel wall. Moreover, the nucleation mechanism of the LSW fault-rupture undergoes a velocity-jump, whereas the RSF model smoothly starts from zero. The LSW model also exhibits shorter risetime, associated with abrupt energy release during fault-slip. The seismograms of the body waves show that the rate-and-state model is better resolved and less sensitive to viscous damping. Additionally, the increase in velocity weakening condition increases induced deformation. Finally, the radiated seismic waves induced peak particle velocity attenuates when the effective normal stress acting on the strike-slip fault increases. Insights gained from this 3D-physics-based study improve understanding of the nucleation of dynamic fault models and their application in deep underground mines.
Rockburst is a typical brittle failure that occurs under high in situ stress, and accurate intensity prediction is essential for the safety of underground engineering. This study proposes an interpretable and high-performance intelligent framework for rockburst intensity prediction. A dataset of 410 real rockburst cases was compiled using seven key features. DBSCAN clustering was used to identify structural characteristics of different rockburst intensities, and a radial oversampling strategy was integrated to alleviate class imbalance. A broad learning system (BLS) served as the base classifier, while a supervised attention mechanism enabled adaptive feature weighting, improving the representation of multi-factor coupling mechanisms. Hyperparameters were optimized using the asynchronous successive halving algorithm (ASHA), resulting in the integrated ASHA–Attention–BLS model. The model achieved an accuracy of 93.90
The study investigates the directional dependence of wave propagation in a rock mass with gradient-distributed micropores. A discontinuous characteristics method was developed by discretizing the rock mass with gradient-distributed micropores into a series of layers. The velocity gradient G was introduced to quantify the spatial variation in wave velocity associated with the gradient distribution of micropores. The proposed method was used to analyze stress and strain waves propagating in two opposite directions. The effects of frequency and propagation distance on the transmission coefficients were also investigated. The results demonstrate that the transmitted stress and strain waves exhibit significant directional dependence. In the low-frequency range, the particle stress transmission coefficient exceeds 1.0 for propagation from a high-to-low micropore density region (G > 0), whereas it falls below 1.0 for propagation in the opposite direction (G < 0). The particle strain transmission coefficient shows the opposite trend. The directional dependence is negligible at short propagation distances but becomes pronounced as the propagation distance increases. The energy transmission coefficients are approximately equal for waves propagating in opposite directions through the same rock mass, consistent with elastodynamic reciprocity in linear viscoelastic media. This study provides a theoretical basis for analyzing the directional dependence of wave propagation induced by the gradient distribution of micropores in rock masses.
Heat recovery from superhot rock is a transformative, high-density energy source with strong baseload potential. However, in situ geomechanical characterization remains impractical, while conventional sampling and laboratory testing are intrusive, costly, and alter material properties. We compile a curated database of 1711 specimens spanning eight lithologies tested under high-pressure, high-temperature conditions to examine the use of index properties recoverable from drill cuttings—rock type, porosity, and mineralogy—to predict stiffness, strength, and brittle–ductile response at in situ stress and temperature. Physics-guided machine learning (PGML) embeds semi-empirical baseline equations as physical constraints while extracting high-dimensional patterns to derive explicit multiplicative formulations. This structure preserves phenomenological consistency beyond the calibration range and enables systematic refinement as new data emerge. The model predicts stiffness and strength with a normalized mean absolute error NMAE 0.3, and the brittle-to-ductile transition confining stress with 85
In tunnel structures, the ceiling is a primary structural component that directly resists vertical loads, and its settlement can significantly affect overall stability. Box-type tunnels have flat ceiling slabs supported by columns and sidewalls, making them more vulnerable to vertical loads than arched tunnels. In particular, the central area of the ceiling, which is distant from supporting elements, is highly susceptible to localized settlement. This study developed the cloud-to-plane displacement (C2PD) analysis, a point-cloud-based distance estimation technique, to assess ceiling settlement in a box-type tunnel in Incheon, South Korea. C2PD establishes a reference plane from the geometric characteristics of the measured ceiling point cloud and calculates the vertical distance of each point from this plane. In this study, downward vertical deviation from the C2PD-derived reference plane is defined as relative ceiling settlement. This approach enables the magnitude and spatial distribution of relative ceiling settlement to be quantitatively evaluated from a single-epoch TLS dataset, while providing visualizations that support maintenance and structural assessment.
