Quantitative characterization of fracture surfaces of granite specimens under triaxial extension using SEM and deep learning

The mesoscopic failure mechanism of rock under triaxial extension (axial tension/unloading-tension in confining pressure) remained still unclear and required further investigation. This study conducted quantitative characterization on fracture surfaces of granite specimens under triaxial extension to quantify the mesoscopic failure mechanism. By combining scanning electron microscope and deep learning, the shear morphology on fracture surfaces of granite under triaxial extension with different confining pressures was quantificationally identified. The results showed: (1) Under zero confining pressure (pure tension), part of the fracture surface of granite exhibits shear morphologies, and the proportion of shear morphologies is 3.9 %. (2) Under low confining pressure (2.5 MPa), the proportion of shear morphologies on fracture surface of granite is lower than that observed in pure tension (2.61 %), indicating that its tensile characteristics are more dominant. (3) As confining pressure varies in a range of 0–70 MPa, the proportion of shear morphologies on fracture surface of granite increases with increasing confining pressure, but the tensile morphology remains dominant. (4) Under confining pressures greater than 25 MPa, the granite specimens undergo axial unloading tension in confining pressure and fail in triaxial compression state. In this case, the Mohr-Coulomb (MC) criterion does not accurately predict the fracture angle of the granite. This is because the granite failure is dominated by mesoscopic tensile fractures, while the MC criterion is fundamentally applicable to shear-dominated failure. The findings of this study offer new insights into the link between mesoscopic fracture mechanisms and macroscopic failure patterns of rock under triaxial extension.