基于深度学习的地震震级分类

刘涛 戴志军 陈苏 傅磊

刘涛,戴志军,陈苏,傅磊. 2022. 基于深度学习的地震震级分类. 地震学报,44(4):656−664 doi: 10.11939/jass.20210046
引用本文: 刘涛,戴志军,陈苏,傅磊. 2022. 基于深度学习的地震震级分类. 地震学报,44(4):656−664 doi: 10.11939/jass.20210046
Liu T,Dai Z J,Chen S,Fu L. 2022. Earthquake magnitude classification based on deep learning. Acta Seismologica Sinica,44(4):656−664 doi: 10.11939/jass.20210046
Citation: Liu T,Dai Z J,Chen S,Fu L. 2022. Earthquake magnitude classification based on deep learning. Acta Seismologica Sinica44(4):656−664 doi: 10.11939/jass.20210046

基于深度学习的地震震级分类

doi: 10.11939/jass.20210046
基金项目: 国家重点研究发展计划(2018YFE0109800)和国家自然科学基金(51738001,U1839202)共同资助
详细信息
    作者简介:

    刘涛,在读硕士研究生,主要从事地震动记录特征分析的相关研究,e-mail: 844613078@qq.com

    通讯作者:

    戴志军,博士,研究员,主要从事强地震动特性分析与合成、地震动过程数值模拟、计算机视觉方法在工程地震方面的应用和深度学习相关应用研究,e-mail:dzj@cea-igp.ac.cn

  • 中图分类号: P315.9

Earthquake magnitude classification based on deep learning

  • 摘要: 为了探索地震加速度时程记录的震级信息,训练卷积神经网络基于地震震级大小对地震记录进行分类,将K-NET和KiK-net中将近12万个地震记录作为样本,对其进行信息筛选和归一化,之后将地震加速度时程记录用作输入,训练卷积神经网络模型以M5.5为分类界限来区分大震和小震。结果显示,在训练集中基于该模型的分类准确率为93.6%,在测试集中的准确率为92.3%,具有良好的分类效果,这表明大震记录与小震记录之间存在一些根本的区别,即可通过地震动加速度时程记录获取一定的震级信息。

     

  • 图  1  预处理阶段采样图

    红线部分是采样频率为100 Hz的五个采样位置,每个位置采样时长为4 s,五个部分共20 s

    Figure  1.  Pre-processing sampling

    The red boxes delineate the five sampling positions with a sampling frequency of 100 Hz. The sampling time of each position is 4 s,and the five parts are 20 s in total

    图  2  神经网络模型从输入数据到获取分类结果的流程图

    (a) 模型的流程说明;(b) 模型的架构

    Figure  2.  Flow chart of neural network model from data inputting to classfication result acquirement

    (a) Process description of the model;(b) Model architecture

    图  3  不同学习率下训练集(a)和测试集(b)的准确率随训练次数增加的变化

    Figure  3.  The change in the correct rates of the trainings set (a) and the test set (b) with the training time increasing on the condition of different learning rate

    图  4  不同批量大小下训练集(a)和测试集(b)的准确率随训练次数增加的变化

    Figure  4.  The change in the correct rates of the trainings set (a) and the test set (b) with the training time increasing on the condition of different batch size

    图  5  (a) 计算模型精度的流程;(b) 准确率;(c) 模型对于部分记录的识别结果

    Figure  5.  (a) The flowchart of calculating model accuracy;(b) Training accuracy;(c) The recognition results of the model for some recordings

  • [1] 李安,杨建思,彭朝勇,郑钰,刘莎. 2020. 基于样本增强的卷积神经网络震相拾取方法[J]. 地震学报,42(2):163–176.
    [2] Li A,Yang J S,Peng C Y,Zheng Y,Liu S. 2020. Seismic phase identification using the convolutional neural networks based on sample enhancement[J]. Acta Seismologica Sinica,42(2):163–176 (in Chinese).
    [3] 隗永刚,杨千里,王婷婷,蒋长胜,边银菊. 2019. 基于深度学习残差网络模型的地震和爆破识别[J]. 地震学报,41(5):646–657. doi: 10.11939/jass.20190030
    [4] Wei Y G,Yang Q L,Wang T T,Jiang C S,Bian Y J. 2019. Earthquake and explosion identification based on Deep Learning residual network model[J]. Acta Seismologica Sinica,41(5):646–657 (in Chinese).
    [5] 谢礼立,翟长海. 2003. 最不利设计地震动研究[J]. 地震学报,25(3):250–261. doi: 10.3321/j.issn:0253-3782.2003.03.003
    [6] Xie L L,Zhai C H. 2003. Study on the severest real ground motion for seismic design and analysis[J]. Acta Seismologica Sinica,25(3):250–261 (in Chinese).
    [7] 张肇诚,张炜. 2016. 地震预报可行性的科学与实践问题讨论[J]. 地震学报,38(4):564–579. doi: 10.11939/jass.2016.04.004
    [8] Zhang Z C,Zhang W. 2016. Discussion on scientific and practical problems of feasibility of earthquake prediction[J]. Acta Seismologica Sinica,38(4):564–579 (in Chinese).
    [9] 周本伟,范莉苹,张龙,李珀任,房立华. 2020. 利用卷积神经网络检测地震的方法与优化[J]. 地震学报,42(6):669–683.
    [10] Zhou B W,Fan L P,Zhang L,Li P R,Fang L H. 2020. Earthquake detection using convolutional neural network and its optimization[J]. Acta Seismologica Sinica,42(6):669–683 (in Chinese).
    [11] 朱晓炜. 2011. 强震记录的选择与缩放方法研究[D]. 哈尔滨: 中国地震局工程力学研究所: 4–5.
    [12] Zhu X W. 2011. The Selection and Scaling of Strong Motion Records[D]. Harbin: Institute of Engineering Mechanics, China Earthquake Administration: 4–5 (in Chinese).
    [13] Arias A. 1970. A measure of earthquake intensity[G]//Seismic Design for Nuclear Power Plants. Cambridge: Massachusetts Institute of Technology Press: 438–483.
    [14] Baker J W,Cornell C A. 2005. A vector-valued ground motion intensity measure consisting of spectral acceleration and epsilon[J]. Earthq Eng Struct Dyn,34(10):1193–1217.
    [15] Baker J W,Cornell C A. 2006a. Correlation of response spectral values for multi-component ground motions[J]. Bull Seismol Soc Am,96(1):215–227.
    [16] Baker J W,Cornell C A. 2006b. Spectral shape,epsilon and record selection[J]. Earthq Eng Struct Dyn,35(9):1077–1095.
    [17] Dysart P S,Pulli J J. 1990. Regional seismic event classification at the NORESS array:Seismological measurements and the use of trained neural networks[J]. Bull Seismol Soc Am,80(6B):1910–1933.
    [18] Goulet C A,Haselton C B,Mitrani-Reiser J,Beck J L,Deierlein G,Porter K A,Stewart J P. 2004. Evaluation of the seismic performance of a code-conforming reinforced-concrete frame building:From seismic hazard to collapse safety and economic losses[J]. Earthq Eng Struct Dyn,36(13):1973–1997.
    [19] Goulet C. 2005. Improving the Characterization of Seismic Hazard for Performance-Based Earthquake Engineering Design[D]. Los Angeles, California: Environmental and Civil Engineering, University of California: 246–268.
    [20] Goulet C A, Haselton C B, Mitrani-Reiser J, Deierlein G, Stewart J P, Taciroglu E. 2006. Evaluation of the seismic performance of a code-conforming reinforced-concrete frame building: Part I: Ground motion selection and structural collapse simulation[C]//8th National Conference on Earthquake Engineering (8NCEE). San Francisco, California: Earthquake Engineering Research Institute: 456–489.
    [21] Huang L Q,Li J,Hao H,Li X B. 2018. Micro-seismic event detection and location in underground mines by using convolutional neural networks (CNN) and deep learning[J]. Tunnel Underground Space Technol,81:265–276. doi: 10.1016/j.tust.2018.07.006
    [22] Jordan M I,Mitchell T M. 2015. Machine learning:Trends,perspectives,and prospects[J]. Science,349(6245):255–260. doi: 10.1126/science.aaa8415
    [23] Krizhevsky A,Sutskever I,Hinton G E. 2017. ImageNet classification with deep convolutional neural networks[J]. Commun ACM,60(6):84–90.
    [24] Murphy K P. 2012. Machine Learning: A Probabilistic Perspective[M]. Cambridge: Massachusetts Institute of Technology Press: 1–16.
    [25] Qian Y M,Bi M X,Tan T,Yu K. 2016. Very deep convolutional neural networks for noise robust speech recognition[J]. IEEE/ACM Trans Audio Speech Language Process,24(12):2263–2276.
    [26] Ross Z E,Yue Y S,Meier M A,Hauksson E,Heaton T H. 2019. Phaselink:A deep learning approach to seismic phase association[J]. J Geophys Res:Solid Earth,124(1):856–869. doi: 10.1029/2018JB016674
    [27] Sainath T N, Mohamed A R, Kingsbury B, Ramabhadran B. 2013. Deep convolutional neural networks for LVCSR[C]//2013 IEEE International Conference on Acoustics, Speech and Signal Processing. Vancouver B C: IEEE: 8614–8618.
    [28] Sercu T, Puhrsch C, Kingsbury B, LeCun Y. 2016. Very deep multilingual convolutional neural networks for LVCSR[C]//2016 IEEE International Conference on Acoustics, Speech and Signal Processing. Shanghai: IEEE: 4955–4959.
    [29] Tiira T. 1999. Detecting teleseismic events using artificial neural networks[J]. Comput Geosci,25(8):929–938. doi: 10.1016/S0098-3004(99)00056-4
    [30] Tóth L. 2013. Convolutional deep rectifier neural nets for phone recognition[C]//Proceedings of the Annual Conference of the International Speech Communication Association, Interspeech. Lyon: International Speech Communication Association (ISCA): 1722–1726.
    [31] Ursino A,Langer H,Scarfì L,Di Grazia G,Gresta S. 2001. Discrimination of quarry blasts from tectonic microearthquakes in the Hyblean Plateau (Southeastern Sicily)[J]. Ann Geophys,44(4):703–722.
    [32] Wiszniowski J,Plesiewicz B M,Trojanowski J. 2014. Application of real time recurrent neural network for detection of small natural earthquakes in Poland[J]. Acta Geophys,62(3):469–485. doi: 10.2478/s11600-013-0140-2
    [33] Yu D, Xiong W, Droppo J, Stolcke A, Ye G L, Li J Y, Zweig G. 2016. Deep convolutional neural networks with layer-wise context expansion and attention[C]//Proceedings of the Annual Conference of the International Speech Communication Association, Interspeech. Lyon: International Speech Communication Association (ISCA): 17–21.
    [34] Zhang G Y,Wang Z Z,Chen Y K. 2018. Deep learning for seismic lithology prediction[J]. Geophys J Int,215(2):1368–1387.
    [35] Zhu L J,Peng Z G,McClellan J,Li C Y,Yao D D,Li Z F,Fang L H. 2019. Deep learning for seismic phase detection and picking in the aftershock zone of 2008 MW7.9 Wenchuan earthquake[J]. Phys Earth Planet Inter,293:106261.
  • 加载中
图(5)
计量
  • 文章访问数:  83
  • HTML全文浏览量:  26
  • PDF下载量:  21
  • 被引次数: 0
出版历程
  • 收稿日期:  2021-04-01
  • 修回日期:  2021-05-31
  • 网络出版日期:  2022-07-14
  • 刊出日期:  2022-08-16

目录

    /

    返回文章
    返回