Optimizing mine fan operations in underground coal mines is important for ensuring proper ventilation, enhancing safety, and improving operational efficiency. A single main ventilation fan is insufficient to meet the ventilation demands of the entire mine. Therefore, it is necessary to consider the addition of booster fans to ensure effective ventilation. However, the selection of booster fans involves multiple influencing factors, and the complex interrelationships among fans remain unclear, making solution selection and risk assessment more challenging. To address this issue, this study proposes an optimization and risk analysis method for booster fan selection based on an improved analytic hierarchy process. This method leverages spherical fuzzy sets to handle uncertainty in the ventilation parameters and cloud models to facilitate probabilistic decision making. Through this model, the important relationships of the influencing factors for fan selection can be systematically determined, allowing for a rational assessment of the performance scores of candidate solutions. It provides a ranking of the alternatives based on their superiority, along with the risk indicators and optimization potentials of the selected solution. Ultimately, the reliability of the chosen model was verified through comparison and validation. This method not only enhances the scientific and rational basis for booster fan selection, reducing the complexity of the selection process, but also provides theoretical support for the optimization of coal mine ventilation systems. This study demonstrates the model’s effectiveness at improving ventilation safety and cost efficiency, making it a valuable tool for modern underground mining operations.

Microbially-induced calcite precipitation (MICP) has emerged as a promising bio-geotechnical technique for sustainable soil improvement, yet accurate prediction of treatment effectiveness remains challenging due to complex multi-factor interactions. This study develops an ensemble learning framework (LARO-EnML) for predicting the unconfined compressive strength (UCS) of MICP-treated sand. A comprehensive database containing 402 experimental datasets was utilised in the study, consisting of unconfined compression test results from bio-cemented sands with eight key input parameters considered. The performance evaluation demonstrates that LARO-EnML achieves superior predictive accuracy, with RMSE of 0.5449, MAE of 0.2853, R² of 0.9570, and OI of 0.9597 on the test data, significantly outperforming other models. Model interpretability analysis reveals that calcite content serves as the most influential factor, with a strong positive correlation to strength enhancement, while urease activity exhibits complex, staged influence characteristics. This research contributes to advancing the practical implementation of MICP technology in geotechnical engineering by offering both accurate predictive capability and enhanced process understanding through interpretable ML approaches.

With the advancement of mining technologies, the evaluation of rock diggability has become a critical research topic for ensuring both safety and efficiency in mining operations. This study establishes a comprehensive evaluation system for mine rock diggability and proposes corresponding grading criteria. For the determination of indicator weights, a combination of subjective and objective methods is employed, integrating expert knowledge and data characteristics to identify optimal weights, thereby providing a reliable basis for comprehensive evaluation. The single-indicator cloud model effectively mitigates the difficulties associated with defining transitional values between adjacent intervals. The multidimensional cloud model, by considering the interactions among indicators, enables the optimization of indicator interactions and enhances the interpretability of diggability grades. Comparison with the Diggability Index (DI) method shows a high consistency between the two approaches (R² = 0.991). The absolute accuracy of diggability levels reaches 74%, while the accuracy based on cloud model fuzzy evaluation reaches 100%, demonstrating the effectiveness of the cloud model in handling transitional intervals and capturing uncertainty. This study provides a novel methodology and theoretical foundation for the scientific evaluation of mine rock diggability, offering practical guidance for reasonable grading, optimization of mining parameters, and interpretation of diggability levels in engineering practice.

Rock drillability classification provides an important basis for drilling-parameter optimization, equipment selection, and improved mining efficiency. Existing drillability evaluation methods often rely on fixed empirical weights and rigid grade boundaries, making it difficult to capture fuzzy transitions between adjacent grades under multi-indicator geological conditions or to explain classification deviations for boundary samples. This study proposes a quantitative rock drillability classification method that integrates FDAHP-based subjective weighting, objective weighting, and a multidimensional cloud model. A 12-indicator evaluation system is first established by considering rock physicomechanical properties, rock-mass structural conditions, and drilling-response characteristics. FDAHP is then used to derive subjective weights from judgment matrices provided by five experts, while the entropy weight method, CRITIC method, and coefficient of variation method are used to obtain objective weights. These weights are combined into a subjective–objective weighting scheme and incorporated into a multidimensional cloud model to represent the fuzziness and randomness of drillability grade boundaries. For incomplete-indicator samples, the comprehensive weights are projected onto the available indicator subset and renormalized, avoiding forced imputation of missing indicators. Validation using 15 complete samples and seven incomplete-indicator samples from the Sungun copper mine shows that the proposed combined weighting method achieves an accuracy of 93.33% for complete samples and correctly classifies six of seven incomplete-indicator samples, with an accuracy of 85.71%. The cloud-model analysis of the misclassified sample indicates that it lies near an adjacent-grade boundary, providing an interpretable explanation for its classification uncertainty. These results suggest that the proposed method can provide interpretable quantitative drillability classification for the Sungun case study and may support preliminary field drillability assessment under incomplete-information conditions.

Tunnel large deformation is governed by coupled effects of rock mass strength, groundwater, and discontinuity conditions. In practice, blurred grade boundaries and inconsistent indicator definitions often undermine the stability and interpretability of conventional threshold-based grading and single-weighting methods, especially in cross-project applications. This study develops a multi-dimensional cloud-model-based fuzzy grading framework for nonlinear large deformation. Under predefined grading intervals, asymmetric cloud models are established to derive membership (association) intervals of each indicator across grades; the upper–lower envelope bandwidth of these intervals is further extracted to quantify sample dispersion and variability, thereby mitigating boundary sensitivity. A subjective–objective integrated weighting strategy is adopted: subjective weights are obtained from an expert scoring matrix, while objective weights are computed using EWM, CRITIC, PCA, and TOPSIS. The representative objective scheme is selected based on grading accuracy and then fused with the subjective weights to form comprehensive weights. Moreover, a same-grade coupling mechanism is introduced to construct pairwise coupling cloud models, which quantify the joint support of indicator combinations for grade discrimination and provide interpretable explanations for misclassifications with actionable optimization directions. Validation on Telmo, Huangjiagou, and Muzhailing datasets yields accuracies of 88%, 90%, and 85%, outperforming single-weighting schemes. Aggregating the three projects produces a global ranking of 13 indicators, offering a unified, robust, and interpretable basis for grading, early-warning prioritization, and mitigation optimization.