As a typical screening apparatus, the elliptically vibrating screen was extensively employed for the size classification of granular materials. Unremitting efforts have been paid on the improvement of sieving performance, but the optimization problem was still perplexing the researchers due to the complexity of sieving process. In the present paper, the sieving process of elliptically vibrating screen was numerically simulated based on the Discrete Element Method (DEM). The production quality and the processing capacity of vibrating screen were measured by the screening efficiency and the screening time, respectively. The sieving parameters including the length of semi-major axis, the length ratio of two semi-axes, the vibration frequency, the inclination angle, the vibration direction angle and the motion direction of screen deck were investigated. Firstly, the Gradient Boosting Decision Trees (GBDT) algorithm was adopted in the modelling task of screening data. The trained prediction models with sufficient generalization performance were obtained, and the relative importance of six parameters for both the screening indexes was revealed. After that, a hybrid MACO-GBDT algorithm based on the Ant Colony Optimization (ACO) was proposed for optimizing the sieving performance of vibrating screen. Both the single objective optimization of screening efficiency and the stepwise optimization of screening results were conducted. Ultimately, the reliability of the MACO-GBDT algorithm were examined by the numerical experiments. The optimization strategy provided in this work would be helpful for the parameter design and the performance improvement of vibrating screens.

作为典型的筛选设备，椭圆振动筛被广泛用于颗粒材料的尺寸分类。在筛分性能的改善方面做出了不懈的努力，但是由于筛分过程的复杂性，优化问题仍然困扰着研究人员。本文基于离散元法（DEM）对椭圆振动筛的筛分过程进行了数值模拟。分别通过筛选效率和筛选时间来测量振动筛的生产质量和处理能力。筛分参数包括半长轴的长度，两个半轴的长度比，振动频率，倾斜角度，研究了筛板的振动方向角和运动方向。首先，在数据筛选的建模任务中采用了梯度提升决策树算法。获得了具有足够泛化性能的训练好的预测模型，并揭示了这两个筛选指标的六个参数的相对重要性。在此基础上，提出了一种基于蚁群优化算法的混合MACO-GBDT算法来优化振动筛的筛分性能。进行了筛选效率的单目标优化和筛选结果的逐步优化。最终，通过数值实验验证了MACO-GBDT算法的可靠性。