现代矿业 ›› 2024, Vol. 40 ›› Issue (03): 48-52.

• • 上一篇    下一篇

基于贝叶斯优化XGBoost模型的岩爆等级短期预测研究*

为进一步提高岩爆灾害短期预测的准确性，选取最大主应力与岩石单轴抗压强度比 值、支护条件、开挖跨度、地质条件、微震震级和峰值粒子速度作为岩爆等级预测的动静态指标，运用 贝叶斯优化后的 XGBoost 算法构建岩爆等级预测模型，探寻各指标参数对于岩爆等级预测的敏感 性，并将优化后的XGBoost模型同XGBoost、随机森林、神经网络和决策树模型进行比较分析。结果 表明：贝叶斯优化后的XGBoost模型具有更高的准确性和可靠性，预测短期岩爆的准确率为74.67%， 其中峰值粒子速度、开挖跨度及应力与强度比值等指标对岩爆等级预测较为敏感。研究结果能够对 现场岩爆等级短期预测提供一定的理论依据和决策工具。   

Short-term Rock Burst Grade Prediction Based on Bayesian Optimized XGBoost Model

In order to further improve the accuracy of rock-burst short-term prediction，the ratio of to⁃ tal maximum principal stress to uniaxial compressive strength of rock，supporting conditions，excavation span，geological conditions，micro seismic magnitude and peak particle velocity were selected as the dynam⁃ ic and static indicators for rock burst grade prediction. XGBoost algorithm under Bayesian optimization is employed to build prediction model，and the sensitive analysis is studied. The predictive results of Bayesian optimized XGBoost model is then compared with XGBoost，Random forest，Neural network and Decision tree models. The results show that the Bayesian optimized XGBoost model has higher accuracy and reliabili⁃ ty，and the accuracy for short-term rock burst grade prediction is 74.67%. Peak particle velocity，excava⁃ tion span and the ratio of total maximum principal stress to uniaxial compressive strength of rock are more sensitive to rock-burst grade. The results can provide certain theoretical basis and decision tool for short⁃ term rock burst grade prediction in site.   

PDF (PC)

27

可视化

0