Karst limestone in Southwest China commonly contains vertically developed dissolution conduits and fissure networks, which may weaken surrounding-rock integrity and induce water inrush, tunnel collapse, and local instability in underground space engineering. However, how vertical conduit diameter and number differentially control progressive crack interaction and instability remains insufficiently understood. In this study, uniaxial compression tests combined with acoustic emission (AE), digital image correlation (DIC), and particle image velocimetry (PIV) monitoring were performed on karst limestone specimens containing simplified vertical cylindrical holes with different diameters and numbers. Within the tested parameter range, hole diameter mainly affects local stress concentration and crack initiation, whereas hole number promotes crack interaction, damage-channel organization, and chain-type coalescence. The peak strength decreased by about 4.5% as hole diameter increased from 0 to 18 mm, and by about 9.0% as hole number increased from 0 to 5. Two core contributions are highlighted in this study. First, a multiscale collaborative crack-chain (MCC-chain) evolution framework is proposed to interpret the transition from local damage activation to dominant fracture-path coalescence. Second, an adaptive Gaussian mixture model and support vector machine (GMM-SVM) model is developed for intelligent tensile-shear crack identification. These findings provide experimental evidence for conduit-controlled instability and support damage identification and stability assessment in karst-affected underground engineering.
This study proposes an enhanced discrete particle tracking (DPT)-aided global digital volume correlation (G-DVC) method to characterize the full-field displacements and strains in a completely decomposed granite (CDG) soil sample with irregular boundaries. The method consists of two main steps: (1) matching extracted particles using the 3D SHOT++ descriptor-based DPT methodology to obtain an initial displacement field (u0); (2) performing finite element method (FEM)-based G-DVC computation on unstructured meshes generated from the sample’s actual geometry, with u0 serving as the initial guess. The SHOT++ tracking method is capable of tracking a large number of CDG particles even under severe crushing conditions, while the G-DVC approach efficiently computes full-field deformations within domains with irregular boundaries. An in-situ X-ray micro-computed tomography (μCT) undrained triaxial test was performed on a miniature saturated CDG sample, and the DPT-aided G-DVC analysis was made on the acquired μCT images to capture the dynamic deformation evolution during shearing. The accuracy of the proposed method is quantitatively evaluated, and the resulting full-field deformations are presented and discussed in detail.
