Integrated micro-macro investigation of permeability evolution and failure modes in limestone: Combining acoustic emission monitoring with digital rock physics characterization

Quantitative characterization of permeability evolution and failure modes in deep rock engineering is essential for predicting seepage risks. This study investigated the mechanical behavior and cross-scale failure mechanisms of limestone using multi-stress path experiments. Acoustic emission (AE) monitoring and digital rock physics (DRP) techniques were used to evaluate the spatiotemporal evolution of cracks and their influence on permeability evolution, and post-failure pore structure characteristics and seepage behavior were revealed. Results demonstrated that unloading confining pressure (UCP) reduced axial bearing capacity by 12.8% compared to constant confining pressure (CCP), leading to stepwise increases in permeability. Under UCP, permeability of matrix-type, filled-crack-type, and crack-type specimens increased by 123.4%, 240.0%, and 42.9% respectively. Shear cracks were predominant under CCP, whereas UCP primarily induced mixed-mode tensile-shear cracks. This distinction caused a transition in macroscopic failure modes, from single shear slip (CCP) to an evolving coordinated crack network (UPC). Further investigations identified dual regulatory effects of crack type on permeability. Compression-dominant shear cracks induced flow channel closure, reducing permeability by 20.6%-87.5%, while mixed-mode tensile-shear cracks enhanced flow network connectivity, increasing permeability by 208.6%-934.1%. The fractal dimensions of 3D reconstructed pore models were 1.49–2.01 for CCP specimens (S1-S3) and 1.97–2.15 for UCP specimens (S4-S6). Correspondingly, average flow velocities in UCP specimens (S4-S6) were 2.1% to 91.6% higher than their CCP counterparts (S1-S3), confirming that UCP-induced heterogeneity improved seepage capacity. Numerical simulations aligned well with AE signatures and observed failure modes, validating the reliability of this cross-scale methodology.