摘要
针对传统公式对爆破振动预测精度不高的问题,构建了基于Bi-LSTM(双向长短期记忆网络)算法的露天矿山爆破振动速度预测模型。该模型可以在两个方向上处理时间序列数据,同时捕获过去和未来的上下输入信息与输出数据之间的依赖关系。以马钢集团高村铁矿露天矿山爆破开采监测数据为依据,选取相关数据为输入参数,并将Bi-LSTM预测结果与萨道夫斯基公式预测结果进行对比。结果表明:萨道夫斯基公式预测的爆破振动速度平均误差为26.87%,Bi-LSTM算法预测的爆破振动速度平均误差为8.95%;Bi-LSTM模型预测结果与实测结果具有较高的吻合度。后期将以其他矿山的监测数据为依托对模型进行训练,以提高Bi-LSTM模型的泛化能力,并通过迁移学习植入矿山安全实时监测预警平台。
The traditional formula for prediction of blast-induced vibration has low accuracy,thus a prediction model for blast-induced vibration velocity in open-pit mines was constructed based on bidirectional long-short-term memory network(Bi-LSTM).This model can process time series data in both directions while capturing the dependency between inputs of the past and future information at upper and lower layers and the outputs.From the monitoring data of blasting operation in Gaocun Iron Mine of Maanshan Iron and Steel Group,the relevant data were selected as the inputs,and the prediction results by Bi-LSTM were compared with those based on Sadaovsky formula.The results show that the blast-induced vibration velocity predicted based on Sadaovsky formula has a mean error of 26.87%,and the blast-induced vibration velocity predicted by Bi-LSTM algorithm has a mean error of 8.95%.It is shown that the Bi-LSTM model can have the prediction results in a high degree of agreement with the measured results.In the future,this Bi-LSTM model will be trained with the monitoring data of other mines to improve its generalization ability,and also will be implanted by transfering learning into a real-time safety monitoring and early warning platform for mines.
作者
张伟
倪彬
王立
谢伟
魏士钰
ZHANG Wei;NI Bin;WANG Li;XIE Wei;WEI Shiyu(China Nonferrous Metals Industry Xi′an Survey and Design Institute Co.,Ltd.,Xi′an 710000,Shaanxi,China;Lanzhou Nonferrous Metallurgy Design and Research Institute Co.,Ltd.,Lanzhou 730000,Gansu,China;School of Civil Engineering and Mapping&Surveying Engineering,Jiangxi University of Science and Technology,Ganzhou 341000,Jiangxi,China)
出处
《矿冶工程》
北大核心
2025年第1期21-26,共6页
Mining and Metallurgical Engineering
基金
江西省教育厅科学技术研究项目(GJJ210859)。
