Accurate classification of rock fracture modes is essential for understanding rock mass instability mechanisms. To address the limitation of traditional acoustic emission (AE) classification methods that treat a single AE signal as a single fracture event, overlooking its composite nature from multiple fracture events and leading to misclassification, this study proposes a novel rock fracture mode classification method based on AE spectral analysis. This study details the development framework, theoretical model, classification criteria, application process, and experimental validation of the new rock fracture mode classification method. Uniaxial compression tests on granite, marble, and limestone, along with rockburst simulation tests on granite, were conducted to validate the classification of fracture modes. In rockburst simulations, shear fracture signals accounted for 48% on average, composite signals 40%, and tensile signals 12%. The method effectively distinguishes multiple fracture events within a single AE signal, accurately classifies fracture modes, and elucidates the dynamic evolution of fracture modes during the rockburst precursor stage, offering significant advantages for rock fracture mode classification and mechanistic insight.

To address the challenges of small-sample rock fine classification—such as overfitting caused by limited sample size and the increased complexity resulting from high inter-class similarity—this study proposes a Stacking ensemble method tailored for small-sample rock image classification. Using a dataset of seven rock categories provided by the BdRace platform, 38 features were extracted across three dimensions—color, texture, and grain size—through grayscale thresholding, HSV color space analysis, gray-level co-occurrence matrix computation, and morphological analysis. The interrelationships among features were evaluated using Spearman correlation analysis and hierarchical clustering, while a voting-based fusion strategy integrated Lasso regularization, gray correlation analysis, and variance filtering for feature dimensionality reduction. The Whale Optimization Algorithm (WOA) was employed to perform global optimization on the base learners, including Random Forest (RF), K-Nearest Neighbors (KNN), Naive Bayes (NBM), and Support Vector Machine (SVM), with Logistic Regression serving as the meta-classifier to construct the final Stacking ensemble model. Experimental results demonstrate that the Stacking method achieves an average classification accuracy of 85.41%, with the highest accuracy for black coal identification (97.16%). Compared to the single models RF, KNN, NBM, and SVM, it improves accuracy by 7.27%, 8.64%, 6.79%, and 6.94%, respectively. Evidently, the Stacking model integrates the strengths of individual models, significantly enhancing recognition accuracy. This research not only improves rock identification accuracy and reduces exploration costs but also advances the intelligent transformation of geological exploration, demonstrating considerable engineering application value.

To characterize the three-dimensional evolution of creep damage in deep roadway surrounding rock, this study establishes a COMSOL-based (Multiphysics 6.3) integrated analysis framework coupling creep damage, seismic velocity response and three-dimensional imaging. The framework adopts the Norton–Bailey creep constitutive model, Weibull-distributed material heterogeneity and an elastic damage model to simulate the 210-day creep damage evolution of surrounding rock. The three-dimensional P-wave velocity field is then reconstructed via a regional correlation approach using active-source P-wave first-arrival times. The results show that with progressive damage accumulation, low-velocity zones gradually expand from the intersection of the fault and roadway roof towards both the fault zone and the surrounding roadway. From 90 d to 210 d, these anomalies develop into a large-scale regional low-velocity belt with significantly enhanced spatial connectivity. The fault weak plane exerts a controlling effect on damage localization and the formation of low-velocity zones. Numerical results demonstrate a distinct spatiotemporal correspondence between P-wave low-velocity zones and damage concentration areas. Furthermore, the predicted damage distribution agrees well with the field failure morphology of surrounding rock, verifying that the proposed method can effectively track creep damage evolution in rock masses and provide an analytical basis for dynamic stability evaluation and early warning of deep roadway surrounding rock.