基于秦岭输水隧洞微震参数的机器学习岩爆预测模型
Rockburst prediction model using machine learning based on microseismic parameters of Qinling water conveyance tunnel
马克 1申青青 1孙兴业 1马天辉 1胡晶 2唐春安1
作者信息
- 1. State Key Laboratory of Coastal and Offshore Engineering(Dalian University of Technology),Dalian 116024,China;Institute of Rock Instability and Seismicity Research,Dalian University of Technology,Dalian 116024,China
- 2. Key Laboratory of Simulation and Regulation of Water Cycle in River Basin,China Institute of Water Resources and Hydropower Research,Beijing 100038,China
- 折叠
摘要
随着地下工程不断向深部发展,岩爆的频次和强度日益增加.为了实现对岩爆烈度等级的预测,本文引入6种机器学习算法,建立了6个岩爆预测模型.以秦岭输水隧洞工程289天的微震监测数据与岩爆案例为基础,构建岩爆烈度等级预测数据集.在建模过程中,首先讨论了数据不均衡对模型性能的影响,得出Borderline-SMOTE1是消除数据不均衡最有效的方法.其次,对6个模型性能指标进行分析,发现Adaboost算法的岩爆预测模型性能最好,其精度、宏F1值、微F1值均最高,分别为0.938、0.937和0.938.最后将Borderline-SMOTE1-Adaboost岩爆预测模型应用于秦岭输水隧洞工程2020年6月1日至2020年6月10日的岩爆烈度等级预测,10次强烈岩爆均被准确预测,验证了通过微震参数来预测岩爆烈度等级的有效性,表明了Borderline-SMOTE1-Adaboost岩爆预测模型可为深埋隧道施工过程中岩爆灾害的预警提供参考.
Abstract
The frequency and intensity of rockburst in underground engineering have increased with excavation depth. In order to predict rockburst intensity grade, this paper introduces six machine learning algorithms to establish six rockburst prediction models. Based on 289-day microseismic monitoring data and rockburst events of Qinling water conveyance tunnel, the rockburst intensity grade prediction dataset is constructed. In the process of model establishment, the impact of data imbalance on model performance is discussed first, and it is concluded that Borderline-SMOTE1 is the most effective method to eliminate data imbalance. Secondly, the analysis of six models ' performance indicators shows that the rockburst prediction model based on the Adaboost algorithm has the best performance, with the highest accuracy, macro-F1, and micro-F1, which are 0.938, 0.937, and 0.938, respectively. Finally, the Borderline-SMOTE1-Adaboost model was applied to the prediction of the rockburst intensity grade of Qinling water conveyance tunnel from June 1, 2020 to June 10, 2020. All ten strong rockbursts are accurately predicted, which verifies the effectiveness of predicting rockburst intensity grade through microseismic parameters. The results show that the Borderline-SMOTE1-Adaboost rockburst prediction model can provide a reference for the early warning of rockburst disasters during the construction of deep-buried tunnels.
关键词
微震/岩爆预测/机器学习/过采样Key words
microseismic/rockburst prediction/machine learning/over sampling引用本文复制引用
基金项目
National Natural Science Foundation of China(51974055)
National Natural Science Foundation of China(42122052)
Joint Fund of Natural Science Basic Research Program of Shaanxi Province,China(2021JLM-11)
Yunnan Fundamental Research Projects,China(202001AT070150)
Fund of China Petroleum Technology and Innovation(2020D-5007-0302)
出版年
2023
NETL
NSTL
维普
万方数据