Answers are grounded in this paper and cite their sources.
The uniaxial compressive strength (UCS) of cemented paste backfill (CPB) is a key indicator for assessing the mechanical performance of mine backfill materials, goaf stability, and the quality control of backfilling operations. However, conventional UCS testing is time-consuming and labor-intensive, making it difficult to meet the demands of rapid mixture optimization and engineering decision-making in mining operations. To improve the accuracy and interpretability of UCS prediction for CPB, this study proposes a data-driven prediction framework integrating an improved particle swarm optimization (PSO) algorithm with an ensemble tree model. Using backfill materials from the Duancun-Leigou mining area, a database comprising 136 CPB specimens with different mixture proportions and curing ages was established. Cement content, fly ash content, gravel content, water content, moisture content, and curing age were selected as input variables. The predictive performance of eight ensemble tree models was systematically evaluated, and CatBoost was selected as the base regressor because of its favorable overall predictive performance and generalization stability among the candidate models. Subsequently, the key hyperparameters of CatBoost were further optimized using standard PSO, LinWPSO, RandWPSO, and CosNCPSO. The results demonstrate that the CosNCPSO-optimized CatBoost model (CosNCPSO-CatBoost) achieves the best predictive performance, with a coefficient of determination (R²) of 0.949, a mean squared error (MSE) of 0.163 MPa², and a root mean squared error (RMSE) of 0.404 MPa, outperforming the other hybrid models. Error analysis and generalization tests further confirm that the proposed model exhibits effective error control, robustness against input perturbations, and stable generalization performance. Sensitivity analysis based on Shapley Additive Explanations (SHAP) indicates that curing age (36.00%), cement content (17.07%), and moisture content (16.58%) are the three most influential factors governing UCS prediction, collectively accounting for 69.65% of the total feature importance. These findings demonstrate that the proposed CosNCPSO-CatBoost framework provides an accurate and interpretable data-driven approach for rapid UCS prediction, mixture design optimization, and quality control of mine backfill materials, particularly under small-sample conditions.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: