机器学习实现崩塌落石微震信号自动化检测和分类

Machine Learning for Automated Detection and Classification of Microseismic Signals from Avalanche Rockfalls

ES评分 0

DOI 10.14075/j.jgg.2024.08.369
刊名
Journal of Geodesy and Geodynamics
年,卷(期) 2025, 45(6)
作者
作者单位

1 赣州地震监测中心站, 江西省赣州市贺兰山路55号, 341000;
2 广东省大宝山矿业有限公司, 广东省韶关市沙溪镇, 512128;
3 中国科学院、水利部成都山地灾害与环境研究所, 成都市群贤南街189号, 610041;
4 中国科学院大学, 北京市玉泉路19号, 100049;
5 江西理工大学资源与环境工程学院, 江西省赣州市客家大道156号, 341000

摘要
基于意大利中部佩鲁贾市某大理石矿山微震监测数据和49个微震特征参数,对比分析神经网络、决策树和随机森林3种机器学习模型在微震信号自动化分类中的精度,通过网格搜索算法优化模型参数提高模型性能,讨论特征参数的选择和分类集的大小对分类结果的影响。结果表明,随机森林分类器在微震信号分类中获得的准确度最高,优化后分类准确度为98.11%,同时减少部分权重小的特征参数,分类计算效率大幅提高,准确度下降较少,训练集与测试集的最佳分配比例为8 ∶2。
Abstract
In this paper, based on microseismic monitoring data and 49 microseismic feature parameters of a marble mine in Perugia, central Italy, we compare and analyze the accuracy of three machine learning models, namely, neural networks, decision trees and random forests, in the automated classification of microseismic signals, and optimize the model parameters to improve the performance of the model through the grid searcher algorithm, with an emphasis on discussing the effects of the feature parameter selection and the size of the classification set on the classification results. The results show that the random forest classifier obtains the highest accuracy in microseismic signal classification, with 98.11% classification accuracy after optimization, meanwhile, by reducing some of the feature parameters with small weights, the efficiency of the classification computation is greatly improved and the accuracy decreases less, and the optimal allocation ratio of the training set to the test set is 8 ∶2.
关键词
崩塌落石; 微震监测; 机器学习; 监测预警
KeyWord
avalanche rockfall; seismic monitoring; machine learning; early warning
基金项目
页码 654-660
  • 参考文献
  • 相关文献
  • 引用本文

谢斌, 高桂山, 黄志勇, 冯亮. 机器学习实现崩塌落石微震信号自动化检测和分类 [J]. 大地测量与地球动力学. 2025; 45; (6). 654 - 660.

  • 文献评论

相关学者

相关机构