Rockbursts pose a significant threat to human safety and environmental stability. This paper aims to predict rockburst intensity using a machine learning model. A dataset containing 344 rockburst cases was collected, with eight inducing features as input and four rockburst grades as output. In the preprocessing stage, missing feature values were estimated using a regression imputation strategy. A novel approach, which combines feature selection (FS), t-distributed stochastic neighbor embedding (t-SNE), and Gaussian mixture model (GMM) clustering, was proposed to relabel the dataset. The effectiveness of this approach was compared with common statistical methods, and its underlying principles were analyzed. A voting ensemble strategy was used to build the machine learning model, and optimal hyperparameters were determined using the tree-structured Parzen estimator (TPE), whose efficiency and accuracy were compared with three common optimization algorithms. The best combination model was determined using performance evaluation and subsequently applied to practical rockburst prediction. Finally, feature sensitivity was studied using a relative importance analysis. The results indicate that the FS + t-SNE + GMM approach stands out as the optimum data preprocessing method, significantly improving the prediction accuracy and generalization ability of the model. TPE is the most effective optimization algorithm, characterized simultaneously by both high search capability and efficiency. Moreover, the elastic energy index W_et, the maximum circumferential stress of surrounding rock σ_θ, and the uniaxial compression strength of rock σ_c were identified as relatively important features in the rockburst prediction model.

The Sejila tunnel, part of the Sichuan-Tibet railway, is in a complex geostress environment because of its deep burial depths and tectonic movement. Based on the measured stress data combined with the structural history and features, the stress characteristics of the Sejila area are preliminarily identified. Then, a three-dimensional numerical model that can provide real topographic features is established, and a distribution law of stress boundary conditions is proposed according to compilations of much measured stress data. By means of support vector regression (SVR), the stress field of the whole Sejila region is determined and finds a reasonable accordance with the measured stress data. Results show that the vertical stress in deep buried stratum can be approximately regarded as one of the principal stresses, and it is reasonable to apply the lateral stress to the model boundary according to a linear function with burial depth. The in situ stress in the tunnel site exhibits that σ_H >σ_V > σ_h, and the σ_H direction deflects when it encounters faults or strata interfaces; the larger that the intersecting angle between fault strike and σ_H is, the smaller the deflection. Compared to the entrance, the rear of the tunnel is subjected to a high maximum principal stress with a high angle; moreover, most sections of the tunnel are estimated to suffer from severe rockbursts, except for a range of 5 km away from the tunnel entrance and 2.5 km away from the exit, according to the Russense criterion. This paper can provide the basis for the prediction and prevention of rockbursts in the Sejila tunnel.

Rocks present complex deformation behaviours and damage processes under triaxial cyclic loading—a subject not yet sufficiently researched. This paper performed triaxial multistage constant-amplitude cyclic loading experiments under different confining stresses on carbonaceous phyllite. The degradation process is analysed by investigating the variation of elastic modulus E_S, Poisson’s ratio υ, irreversible strain ε^irr and energy. Moreover, the rock’s failure mode is explored from both macro and micro perspectives. The results showed that the increase in stress level caused the decrease of E_S in a step-like form, and the constant-amplitude cyclic loading in each stress level caused a slow decrease of E_S, while the υ increased with stress level and constant-amplitude cycles in a similar form. ε^irr accumulated rapidly at first and then slowly at each stress level; the stress level and irreversible axial strain are related by an exponential function. In terms of energy evolution analysis, the damage to rock can be represented by the cumulative damage energy, there were deceleration accumulations and stability accumulation stages of damage at all stress levels, and an acceleration accumulation stage occurred when the rock was close to failure. The failures of rock under cyclic loading are mainly shear failures, accompanied by grain crushing.