基于短时傅里叶变换和卷积神经网络的地震事件分类

张帆 杨晓忠 吴立飞 韩晓明 王树波

张帆,杨晓忠,吴立飞,韩晓明,王树波. 2021. 基于短时傅里叶变换和卷积神经网络的地震事件分类. 地震学报,43(4):1−11 doi: 10.11939/jass.20200128
引用本文: 张帆,杨晓忠,吴立飞,韩晓明,王树波. 2021. 基于短时傅里叶变换和卷积神经网络的地震事件分类. 地震学报,43(4):1−11 doi: 10.11939/jass.20200128
Zhang F,Yang X Z,Wu L F,Han X M,Wang S B. 2021. Classification of seismic events based on short-time Fourier transform and convolutional neural network. Acta Seismologica Sinica,43(4):1−11 doi: 10.11939/jass.20200128
Citation: Zhang F,Yang X Z,Wu L F,Han X M,Wang S B. 2021. Classification of seismic events based on short-time Fourier transform and convolutional neural network. Acta Seismologica Sinica43(4):1−11 doi: 10.11939/jass.20200128

基于短时傅里叶变换和卷积神经网络的地震事件分类

doi: 10.11939/jass.20200128
基金项目: 地震科技星火计划项目(XH20014)和中国地震局震情跟踪定向任务(2020020103)联合资助
详细信息
    通讯作者:

    杨晓忠,e-mail:yxiaozh@ncepu.edu.cn

  • 中图分类号: P315.01

Classification of seismic events based on short-time Fourier transform and convolutional neural network

  • 摘要: 本文选用内蒙古区域地震台网记录到的417个爆破事件和519个天然地震事件的观测资料,对其进行截取和滤波等预处理后,经过短时傅里叶变换转换为时频域的对数振幅谱,使用含有3个卷积层的卷积神经网络作为分类器,实现地震事件自动分类。5折交叉验证结果显示,本文所使用算法的平均准确率达到97.33%,测试集的准确率达到98.03%,本文采用的模型应用了较完整的原始信息,因此获得了较高的准确率和较好的稳定性。

     

  • 图  1  地震和爆破事件的空间分布图

    Figure  1.  Spatial distribution of natural earthquakes and explosions

    图  2  天然地震和爆破的震级频次分布

    Figure  2.  Magnitude-frequency distribution of natural earthquakes and explosions

    图  3  数据预处理过程

    Figure  3.  The process of data preprocessing

    图  4  天然地震与爆破的波形(a)及时频域对数振幅谱(b)对比

    Figure  4.  Comparison of waveform (a) and time-frequency log amplitude spectrum (b) between earthquake and explosion

    图  5  天然地震与爆破的功率谱对比

    Figure  5.  Power spectrum comparison between natural earthquakes and explosions

    图  6  卷积神经网络结构

    Figure  6.  CNN network structure

    图  7  学习率测试

    (a) 损失函数曲线; (b) 准确率曲线; (c) 准确率随学习率变化

    Figure  7.  Learning rate test

    (a) Loss curve ;(b) Accuracy curve ;(c) Accuracy rate changes with learning rate

    图  8  5折交叉验证结果

    Figure  8.  Five fold cross validation results

    表  1  模型的超参数设置

    Table  1.   Super parameter setting of the model

    序号类型参数输出
    1 输入层 50×100×1
    2 卷积层 16个 3×3卷积核 50×100×16
    3 批量正则化层 50×100×16
    4 RELU激活层 50×100×16
    5 最大池化层 2×2 25×50×16
    6 卷积层 32个3×3×16卷积核 25×50×32
    7 批量正则化层 25×50×32
    8 RELU激活层 25×50×32
    9 最大池化层 2×2 12×25×32
    10 卷积层 64个3×3×32卷积核 12×25×64
    11 批量正则化层 12×25×64
    12 RELU激活层 12×25×64
    13 最大池化层 2×2 6×12×64
    14 Dropout层 比例50% 6×12×64
    15 全连接层 单元数128 128
    16 批量正则化层 128
    17 Relu激活层 128
    18 全连接层 单元数2 2
    19 Softmax激活层 2
    20 输出层
    下载: 导出CSV
  • [1] 毕明霞,黄汉明,边银菊,李锐,陈银燕,赵静. 2011. 天然地震与人工爆破波形信号HHT特征提取和SVM识别研究[J]. 地球物理学进展,26(4):1157–1164. doi: 10.3969/j.issn.1004-2903.2011.04.004
    [2] Bi M X,Huang H M,Bian Y J,Li R,Chen Y Y,Zhao J. 2011. A study on seismic signal HHT features extraction and SVM recognition of earthquake and explosion[J]. Progress in Geophysics,26(4):1157–1164 (in Chinese).
    [3] 边银菊. 2002. 遗传BP网络在地震和爆破识别中的应用[J]. 地震学报,24(5):516–524. doi: 10.3321/j.issn:0253-3782.2002.05.009
    [4] Bian Y J. 2002. Application of genetic BP network to discriminating earthquakes and explosions[J]. Acta Seismologica Sinica,24(5):516–524 (in Chinese).
    [5] 蔡杏辉,张燕明,陈惠芳,巫立华. 2020. 基于小波特征和神经网络的天然地震与人工爆破自动识别[J]. 大地测量与地球动力学,40(6):634–639.
    [6] Cai X H,Zhang Y M,Chen H F,Wu L H. 2020. Automatic identification of earthquake and explosion based on wavelet transform and neural network[J]. Journal of Geodesy and Geodynamics,40(6):634–639 (in Chinese).
    [7] 陈润航,黄汉明,柴慧敏. 2018. 地震和爆破事件源波形信号的卷积神经网络分类研究[J]. 地球物理学进展,33(4):1331–1338. doi: 10.6038/pg2018BB0326
    [8] Chen R H,Huang H M,Chai H M. 2018. Study on the discrimination of seismic waveform signals between earthquake and explosion events by convolutional neural network[J]. Progress in Geophysics,33(4):1331–1338 (in Chinese).
    [9] 范晓易,曲均浩,曲保安,刘方斌,山长仑,周少辉. 2019. 支持向量分类机LIBSVM方法识别天然地震、爆破与塌陷[J]. 大地测量与地球动力学,39(9):916–918.
    [10] Fan X Y,Qu J H,Qu B A,Liu F B,Shan C L,Zhou S H. 2019. Support vector machine LIBSVM method for identifying natural earthquakes,blasting and collapse[J]. Journal of Geodesy and Geodynamics,39(9):916–918 (in Chinese).
    [11] 郝国成,谈帆,程卓,王巍,冯思权,张伟民. 2019. 强鲁棒性和高锐化聚集度的BGabor–NSPWVD时频分析算法[J]. 自动化学报,45(3):566–576.
    [12] Hao G C,Tan F,Cheng Z,Wang W,Feng S Q,Zhang W M. 2019. Time–frequency analysis of Bgabor–Nspwvd algorithm with strong robustness and high sharpening concentration[J]. Acta Automatica Sinica,45(3):566–576 (in Chinese).
    [13] 黄汉明,边银菊,卢世军,蒋正锋,李锐. 2010. 天然地震与人工爆破的波形小波特征研究[J]. 地震学报,32(3):270–276. doi: 10.3969/j.issn.0253-3782.2010.03.002
    [14] Huang H M,Bian Y J,Lu S J,Jiang Z F,Li R. 2010. A wavelet feature research on seismic waveforms of earthquakes and explosions[J]. Acta Seismologica Sinica,32(3):270–276 (in Chinese).
    [15] 刘建伟,赵会丹,罗雄麟,许鋆. 2020. 深度学习批归一化及其相关算法研究进展[J]. 自动化学报,46(6):1090–1120.
    [16] Liu J W,Zhao H D,Luo X L,Xu J. 2020. Research progress on batch normalization of deep learning and its related algorithms[J]. Acta Automatica Sinica,46(6):1090–1120 (in Chinese).
    [17] 刘莎,杨建思,田宝峰,郑钰,姜旭东,徐志强. 2012. 首都圈地区爆破、矿塌和天然地震的识别研究[J]. 地震学报,34(2):202–214. doi: 10.3969/j.issn.0253-3782.2012.02.007
    [18] Liu S,Yang J S,Tian B F,Zheng Y,Jiang X D,Xu Z Q. 2012. Discrimination between explosions,mine collapses and earthquakes in capital region of China[J]. Acta Seismologica Sinica,34(2):202–214 (in Chinese).
    [19] 王凤瑛,张丽丽. 2006. 功率谱估计及其MATLAB仿真[J]. 微计算机信息,22(31):287–289. doi: 10.3969/j.issn.1008-0570.2006.31.102
    [20] Wang F Y,Zhang L L. 2006. Power spectrum density estimation and the simulation in MATLAB[J]. Control &Automation,22(31):287–289 (in Chinese).
    [21] 王婷婷,边银菊,张博. 2013. 地震和爆破的综合识别方法研究[J]. 地球物理学进展,28(5):2433–2443. doi: 10.6038/pg20130522
    [22] Wang T T,Bian Y J,Zhang B. 2013. The comprehensive identification methods between earthquakes and explosions[J]. Progress in Geophysics,28(5):2433–2443 (in Chinese).
    [23] 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
    [24] 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).
    [25] 杨选辉,沈萍,刘希强,郑治真. 2005. 地震与核爆识别的小波包分量比方法[J]. 地球物理学报,48(1):148–156. doi: 10.3321/j.issn:0001-5733.2005.01.020
    [26] Yang X H,Shen P,Liu X Q,Zheng Z Z. 2015. Application of method of spectral component ratio of wavelet–packets to discrimination between earthquakes and nuclear explosions[J]. Chinese Journal of Geophysics,48(1):148–156 (in Chinese).
    [27] 余凯,贾磊,陈雨强,徐伟. 2013. 深度学习的昨天、今天和明天[J]. 计算机研究与发展,50(9):1799–1804. doi: 10.7544/issn1000-1239.2013.20131180
    [28] Yu K,Jia L,Chen Y Q,Xu W. 2013. Deep learning:Yesterday,today,and tomorrow[J]. Journal of Computer Research and Development,50(9):1799–1804.
    [29] 曾融生,陈运泰,吴忠良. 2000. 探测地球内部的“雷达”:地震波(续)[J]. 城市防震减灾,3(6):12–14.
    [30] Zeng R S,Chen Y T,Wu Z L. 2000. Radar for detecting the earth interior:Seismic wave[J]. Ruban Earthquake Disaster Prevention and Reduction,3(6):12–14 (in Chinese).
    [31] 张博,边银菊,王婷婷. 2014. 用逐步代价最小决策法识别地震与爆破[J]. 地震学报,36(2):233–243. doi: 10.3969/j.issn.0253-3782.2014.02.008
    [32] Zhang B,Bian Y J,Wang T T. 2014. Discrimination of earthquakes and explosions by SAMC decision method[J]. Acta Seismologica Sinica,36(2):233–243 (in Chinese).
    [33] 张帆. 2006. 地震波时频谱分析及其在爆破识别中的应用[D]. 合肥: 中国科学技术大学: 57.
    [34] Zhang F. 2006. Seismic Signal Time–Frequency Analysis and Its Application to Blast Distinguishing[D]. Hefei: University of Science and Technology of China: 57 (in Chinese).
    [35] 张帆,韩晓明,郝美仙,张晖. 2016. 内蒙古阿拉善地区爆破和地震自动识别研究[J]. 华南地震,36(3):98–103.
    [36] Zhang F,Han X M,Hao M X,Zhang H. 2016. Auto identification of explosion and seismic events in Alashan area[J]. South China Journal of Seismology,36(3):98–103 (in Chinese).
    [37] 张号逵,李映,姜晔楠. 2018. 深度学习在高光谱图像分类领域的研究现状与展望[J]. 自动化学报,44(6):961–977.
    [38] Zhang H K,Li Y,Jiang Y N. 2018. Deep learning for hyperspectral imagery classification:The state of the art and prospects[J]. Acta Automatica Sinica,44(6):961–977 (in Chinese).
    [39] 赵永,刘卫红,高艳玲. 1995. 北京地区地震、爆破和矿震的记录图识别[J]. 地震地磁观测与研究,16(4):48–54.
    [40] Zhao Y,Liu W H,Gao Y L. 1995. Distinguishing earthquake,explosion and mine earthquake in Beijing area[J]. Seismological and Geomangnetic Observation and Research,16(4):48–54 (in Chinese).
    [41] Beyreuther M,Barsch R,Krischer L,Megies T,Behr Y,Wassermann J. 2010. ObsPy:A python toolbox for seismology[J]. Seismol Res Lett,81(3):530–533. doi: 10.1785/gssrl.81.3.530
    [42] Kingma D P, Ba J L. 2015. Adam: A method for stochastic optimization[C]//3rd International Conference on Learning Representations, ICLR 2015: Conference Track Proceedings. San Diego: IEEE Press: 1–15.
    [43] Krizhevsky A,Sutskever I,Hinton G E. 2017. ImageNet classification with deep convolutional neural networks[J]. Commun ACM,60(6):84–90. doi: 10.1145/3065386
  • 加载中
图(8) / 表(1)
计量
  • 文章访问数:  76
  • HTML全文浏览量:  32
  • PDF下载量:  7
  • 被引次数: 0
出版历程
  • 收稿日期:  2020-07-28
  • 修回日期:  2020-11-08
  • 网络出版日期:  2021-08-16

目录

    /

    返回文章
    返回