302 / 2024-03-15 09:57:43

Robust Mechanical Assessment of Mining Paste Backfill Materials with Ensemble Adversarial Model

Mining waste, Backfill material, Ensemble learning,Adversarial perturbation,Mechanical properties

1、资源开发与利用 > 10、地学大数据集成与应用

The waste rocks produced from mining can pollute the environment, while converting waste rocks into backfill materials can not only reduce pollution but also mitigate surface subsidence. The mechanical properties of backfill materials are crucial for surface protection. Therefore, this study established a large-scale dataset based on coal gangue and tailings through experiments and data collection. An ensemble learning model was developed to assess the nonlinear effects of 29 dimensions on the mechanical properties. Various factors such as different backfill materials, preparation methods, and measurement errors can lead to significant variations in mechanical performance, thus potentially affecting the accuracy of assessments. Hence, this paper proposes an ensemble adversarial approach to enhance the model's robustness to disparate data and investigates the performance of the model under different defense levels and attack levels. The study reveals that the model can adaptively assess the mechanical properties of backfill materials in both coal and non-coal mines. Compared to a single machine learning model and conventional ensemble models, the ensemble adversarial model exhibits a mean square error of 1.28 and a correlation coefficient of 0.91, demonstrating superior robustness on unstable datasets. This study contributes to the advancement of large-scale models in the field of mining waste disposal, facilitating environmentally friendly intelligent mining practices.