Abstract: Accurate prediction of equipment operating power and product particle size is fundamental to the intelligentization of crushing circuits. The crushing circuit of the Sanxin Gold-Copper Mine was taken as the research object to achieve this goal, and simulation research was carried out by adopting BPS software. Meanwhile, combined with machine learning methods on the basis of the simulation results, prediction indicators of 9 different algorithms were compared, and a predictive metamodel for crushing process parameters was finally built. The results show that the crushing process parameters obtained by simulation agree well with the process survey data, with relative errors ranging from 0.190 % to 7.349 %. The optimal predictive metamodels for primary, secondary, and tertiary crushing operating power and final product P₈₀ are Lasso, Ridge, GBDT, and GBDT, respectively. The relative errors between the predicted values and simulated values are 0.889 %, 0.702 %, 1.990 %, and 0.361 % respectively, indicating that all the metamodels for the crushing process parameters have sound prediction accuracy. Interpretability analysis results show that the primary crushing operating power is mainly influenced by feed F₈₀, the ore A*b value and throughput, the secondary crushing operating power is governed by feed F₈₀, secondary crusher closed side setting and throughput, and the tertiary crushing power is dominated by tertiary crusher closed side setting and the ore A*b value, with the final product P₈₀ mainly controlled by the apertures of the primary and secondary screens. The findings provide data-driven decision support for the prediction of crushing process indicators, adjustment of process parameters, and optimization of equipment energy consumption in complex and varying feed conditions, thereby promoting the intelligent development of crushing circuits.