Prediction of continuous–discontinuous failure of granite subjected to direct shearing using acousto-optical-mechanical scheme

Abstract

The catastrophic failure of deep rocks is difficult to predict, which generally exhibits a continuous–discontinuous behavior. A multimodal acoustic–optical–mechanical (AOM) monitoring system integrating acoustic emission (AE) analysis, digital image correlation (DIC), and mechanical loading was developed to address this limitation. Direct shear tests on granite were conducted under six different normal stresses to investigate the AOM responses during shear rupture. The shear failure was divided into three stages: Stage I—compaction and stable crack propagation, Stage II—from crack damage stress to catastrophic failure, and Stage III—post-peak residual shear and structural reconfiguration. The acoustic quiet period occurred consistently between Stages II and III, suggesting that the high heterogeneity induced by crack damage significantly disrupted the acoustic acceleration pattern preceding failure. This disruption explains the low predictive capacity of AE methods. An inverted U-curve was observed between the AE event rate and the loading time. An effective method was proposed for indirectly identifying crack damage stress. AE source classification and localization revealed that shear events were dominant (exceeding 74%), whereas tensile events were higher under higher normal stress, indicating a transition in microcrack evolution from shear-dominated to shear–tensile dominated mechanisms. The DIC-based principal strain fields showed local strain concentration near the nominal shear plane, evolving into discontinuous fracture zones highly consistent with the AE spatial distributions. The log-periodic power law singularity (LPPLS) model accurately captured the pre-failure acceleration and oscillation. The LPPLS model provided more accurate failure time predictions than the conventional power law singularity (PLS) model, with smaller residuals for optical and mechanical parameters. The AE parameters were more suitable for early warning of acceleration onset. Finally, we established an AOM-based framework to characterize and model coupled failure, offering valuable insights into early warning and the accurate prediction of deep rock failure.

Topics & Concepts