Damage Evolution Characterization and Failure Time Prediction of Heterogeneous Sandstone upon Uniaxial Compression Based on Acoustic Emission Activity
摘要
Heterogeneous sandstone under uniaxial compression is commonly affected by mesoscopic compositional differences and their spatial distribution during damage evolution and unstable failure. However, under conditions of comparable macroscopic mechanical properties, how mesoscopic heterogeneity modifies crack propagation paths, damage mode evolution, and final failure morphology remains to be further clarified. To address this issue, three sandstone specimens collected from the same area, with the same lithology and identical specimen preparation and loading conditions, were selected as comparative experimental objects. X-ray diffraction (XRD), uniaxial compression tests, and acoustic emission (AE) monitoring were conducted to obtain information on mineral composition, mechanical response, AE evolution, and failure morphology. Meanwhile, heterogeneous numerical models based on stochastic assignment of elastic modulus following a normal distribution were established. Specifically, while keeping the model geometry, boundary conditions, and mean elastic modulus unchanged, a Python script was used in ABAQUS to randomly assign elastic modulus values following a normal distribution to each mesh element. The coefficient of variation Cv was used to control the dispersion of material parameters, thereby equivalently representing the material heterogeneity caused by mesoscopic compositional differences in sandstone. This approach enables the spatial combination of elements with different stiffnesses to be simulated under controlled conditions, allowing the effects of heterogeneity degree on local stress concentration, crack initiation locations, and crack propagation paths to be analyzed. The principal findings are as follows: a) Although the three sandstone specimens exhibited broadly comparable peak strength and elastic modulus, the XRD results revealed differences in mineral composition, which corresponded to different failure modes. Specimens S1 and S3 mainly failed by axial splitting, whereas S2 showed a tendency toward conjugate inclined shear failure. AE damage mode identification and heterogeneous numerical analysis further indicate that mesoscopic compositional differences and the resulting local stiffness heterogeneity can modify stress concentration and crack propagation paths. In particular, the more pronounced acceleration of shear damage in S2 near instability is an important reason for its shear-dominated failure b) PCA was used to reduce the dimensionality of multi-parameter AE features, and K-means + + was then applied for unsupervised clustering. The clustering results were physically mapped using the RA–AF relationship, identifying three damage modes: tensile, shear, and mixed. More importantly, this approach further revealed the dynamic evolution differences of these damage modes during loading. In particular, the conjugate shear failure of specimen S2 was found not to be governed by the highest cumulative proportion of shear damage over the entire loading process, but to be closely associated with the accelerated growth of shear damage near instability. c) Based on the power-law singularity theory, AE ring-down count and duration were selected as response variables for failure time prediction. Under the optimal parameter combination, the predicted failure time tended to stabilize and approach the actual failure time within an early-warning window of approximately 5.5%–17.4% before failure. This indicates that ring-down count and duration can effectively capture the critical acceleration of AE activity near instability and are suitable for trend-based early warning of heterogeneous sandstone failure. This work establishes an integrated framework linking damage characterization, mesoscopic mechanism interpretation, and failure time prediction, offering methodological insights for instability mechanisms and engineering early warning in heterogeneous rock masses.