A Real-Time Seismic Intensity Prediction Model Based on Long Short-Term Memory Neural Network
-
摘要: 实时烈度预测可在破坏性地震波到达前，根据P波估计地震可能造成的最大影响.预警对象可以采取措施，降低可能造成的损失.P波位移幅值是一种有效估计地震动峰值的参数，然而单个或多个参数难以全面表征地震动中的信息.同时，参数的计算需要确定时间窗大小，无法实现连续预测.为了解决上述问题，提出了一种基于长短期记忆网络的实时地震烈度预测模型.基于2010-2021年K-NET数据构建模型，并选取2022年3月MJMA7.3地震事件作为案例验证模型.结果表明，P波到达后可以在记录的每个时间步预测烈度，P波到达3 s时在测试集中准确率为96.47%.提出的LSTM模型改善了烈度预测的准确性和连续性，可为地震预警、应急响应等提供科学依据.Abstract: Real-time intensity prediction can estimate the maximum possible impact of an earthquake based on P-wave before the arrival of destructive seismic waves. Earthquake early warning targets can take measures to reduce the potential damage. Peak P-wave displacement amplitude is a parameter that effectively estimates the peak ground motion, however, it is difficult to fully characterize the information in ground motion by a single or multiple parameters. Meanwhile, the calculation of the parameter requires the determination of the time window size, and continuous prediction cannot be achieved. To solve the above problems, a prediction model based on long short-term memory network is proposed in this paper. The model is constructed based on K-NET data from 2010‒2021, and the MJMA 7.3 earthquake event in March 2022 is selected as a case to validate the model. The results show that the intensity can be predicted at each time step of the record after the P-wave arrival, and the accuracy in the test set is 96.47% at 3 seconds after P-wave arrival. The LSTM model proposed in this paper improves the accuracy and continuity of intensity prediction and can provide a scientific basis for earthquake early warning and emergency response.
-
Key words:
- seismic intensity /
- real time /
- neural network /
- deep learning /
- earthquake early warning /
- engineering geology
-
图 1 台站在训练集、测试集（a）和案例集（c）以及震中位置在训练集、测试集（b）和案例集（d）上的分布
Fig. 1. Distribution of stations on training set, test set (a) and case set (c), and location of the epicenters on training set, test set (b) and case set (d)
图 7 P波到达后MSE（a）和准确率（b）随时间变化的关系
Fig. 7. Relationship of MSE (a) and accuracy (b) with time after P-wave arrival
图 8 3 s时间窗时的预测值和观测值
a. LSTM模型；b. Pd模型的对比
Fig. 8. Comparison of predicted and observed values at 3-second time window
图 9 案例中震源（a）和台站的位置以及烈度的空间分布（b）和数据分布（c）
Fig. 9. Location of the epicenter and stations(a), and spatial distribution (b) and data distribution (c) of intensities on case set
图 10 台站被触发后烈度估计的误差与时间之间的关系
Fig. 10. Relationship between the intensity estimation errors and the time after the station is triggered
图 11 （a）~（l）分别为不同烈度下仪器地震烈度的预测值和观测值随时间变化的关系
Fig. 11. (a) to (l) are the predicted and observed values of the instrumental seismic intensity with time for different intensities
表 1 不同LSTM网络结构的MSE和训练时间（验证集）
Table 1. MSE and training time for different LSTM network structures (validation set)
No. 网络层数 隐层单元数量 MSE 训练时间 1 1 8 828.350 6 13 477.373 6 2 1 16 667.354 3 14 354.786 8 3 1 32 569.153 7 16 753.753 1 4 1 64 448.534 1 13 578.841 5 5 1 128 392.617 7 6 252.108 5 6 2 8 355.553 5 22 833.113 7 7 2 16 328.543 4 23 788.3574 8 2 32 267.564 8 22 879.357 4 9 2 64 228.820 8 20 708.898 1 10 2 128 224.943 3 22 326.148 8 11 3 8 297.358 7 32 443.092 9 12 3 16 276.548 6 34 512.365 7 13 3 32 257.789 3 33 574.789 3 14 3 64 234.638 5 32 876.654 7 15 3 128 238.049 9 29 957.053 3 16 4 8 295.915 5 39 464.159 2 17 4 16 268.465 1 40 873.546 1 18 4 32 243.165 7 37 453.312 8 19 4 64 225.267 7 33 316.395 8 20 4 128 226.192 1 38 112.503 2 注：其他超参数的值为：Learning rate=10-3 Batch size=50; Maximum iterations=100 (Early stop=30). 
下载: 导出CSV
表 2 不同阶段模型的计算效率
Table 2. Computational efficiency of models at different stages
阶段 时间 输入信息计算时间 忽略 训练模型时间(共47 304样本) 1.78×101 s 测试模型时间(共9 166样本) 3.153 3×101 案例数据集上测试模型时间(每个样本) 3.440×10-3 s
下载: 导出CSV
表 3 3秒时间窗时Pd和LSTM模型的评价指标
Table 3. Evaluation metrics for Pd and LSTM models at 3-second time windows
评价指标 Pd方法 LSTM方法 准确率(%) 90.17 96.47 MSE 0.379 5 0.369 1
下载: 导出CSV
-
Carranza, M., Buforn, E., Colombelli, S., et al., 2013. Earthquake Early Warning for Southern Iberia: A P Wave Threshold‐Based Approach. Geophysical Research Letters, 40: 4588-4593. https://doi.org/10.1002/grl.50903 Chen, Y. L., Jin, X., 2016. A Continuous Real⁃Time Method for Seismic Intensity Prediction. Earthquake Engineering and Engineering Dynamics, 36(6): 22-29 (in Chinese with English abstract). Chu, D. P., Wan, B., Li, H., et al., 2021. Geological Entity Recognition Based on ELMO⁃C0NN⁃BiLSTM⁃CRF Model. Earth Science, 46(8): 3039-3048 (in Chinese with English abstract). doi: 10.12090/j.issn.1006-6616.2021.27.03.035 Deng, L., Yu, D., 2014. Deep Learning: Methods and Applications. Foundations and Trends in Signal Processing, 7: 197-387. https://doi.org/10.1561/2000000039 Festa, G., Zollo, A., Lancieri, M., 2008. Earthquake Magnitude Estimation from Early Radiated Energy. Geophysical Research Letters, 35(22): L22307. https://doi.org/10.1029/2008GL035576 Graves, A., Fernández, S., Gomez, F., et al., 2006. Connectionist Temporal Classification: Labelling Unsegmented Sequence Data with Recurrent Neural Networks. In: Proceedings of the 23rd International Conference on Machine Learning. Association for Computing Machinery, New York, 369-376. https://doi.org/10.1145/1143844.1143891 Hochreiter, S., Schmidhuber, J., 1997. Long Short⁃Term Memory. Neural Computation, 9: 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735 Hoshiba, M., Kamigaichi, O., Saito, M., et al., 2008. Earthquake Early Warning Starts Nationwide in Japan. EOS, Transactions American Geophysical Union, 89: 73-74. https://doi.org/10.1029/2008EO080001 Kamigaichi, O., Saito, M., Doi, K., et al., 2009. Earthquake Early Warning in Japan: Warning the General Public and Future Prospects. Seismological Research Letters, 80: 717. https://doi.org/10.1785/gssrl.80.5.717 Kanamori, H., 2005. Real⁃Time Seismology and Earthquake Damage Mitigation. Annu. Rev. Earth Planet. Sci. , 33: 195-214. https://doi.org/10.1146/annurev.earth.33.092203.122626 Kingma, D. P., Ba, J., 2014. ADAM: A Method for Stochastic Optimization. https://doi.org/10.48550/arXiv.1412.6980 Kunugi, T., Aoi, S., Nakamura, H., et al., 2013. An Improved Approximating Filter for Real⁃Time Calculation of Seismic Intensity. Zisin, 2: 223-230. https://doi.org/10.4294/zisin.65.223 Liu, R. S., Xiong, M. P., Ma, Q., et al., 2021. Vulnerability Study for High Voltage Electrical Equipment in Substations Based on Instrumental Seismic Intensity. Journal of Natural Disasters, 30(2): 14-23 (in Chinese with English abstract). Nazeri, S., Shomali, Z. H., Colombelli, S., et al., 2017. Magnitude Estimation Based on Integrated Amplitude and Frequency Content of the Initial P Wave in Earthquake Early Warning Applied to Tehran, Iran. Bulletin of the Seismological Society of America, 107: 1432-1438. https://doi.org/10.1007/978⁃3⁃642⁃55903⁃7_92 Otake, R., Kurima, J., Goto, H., et al., 2020. Deep Learning Model for Spatial Interpolation of Real‐Time Seismic Intensity. Seismological Research Letters, 91: 3433-3443. https://doi.org/10.1785/0220200006 Peng, C. Y., Yang, J. S., Zheng, Y., et al., 2017. New Τc Regression Relationship Derived from all P Wave Time Windows for Rapid Magnitude Estimation. Geophysical Research Letters, 44: 1724-1731. http://www.researchgate.net/profile/Peng_Chaoyong/publication/313653194_New_t_c_regression_relationship_derived_from_all_P-wave_time_windows_for_rapid_magnitude_estimation/links/59d21862a6fdcc181ad5e238/New-t-c-regression-relationship-derived-from-all-P-wave-time-windows-for-rapid-magnitude-estimation.pdf Reimers, N., Gurevych, I., 2017. Optimal Hyperparameters for Deep LSTM⁃Networks for Sequence Labeling Tasks. Geophysical Research Letters, 44(4): 1724-1731. https://doi.org/10.1002/2016GL071672 Sheen, D. H., Park, J. H., Chi, H. C., et al., 2017. The First Stage of an Earthquake Early Warning System in South Korea. Seismological Research Letters, 88: 1491-1498. https://doi.org/10.1785/0220170062 Sherstinsky, A., 2020. Fundamentals of Recurrent Neural Network (RNN) and Long Short⁃Term Memory (LSTM) Network. Physica D: Nonlinear Phenomena, 404: 132306. https://doi.org/10.1016/j.physd.2019.132306 Srivastava, N., Hinton, G., Krizhevsky, A., et al., 2014. Dropout: A Simple Way to Prevent Neural Networks from Overfitting. The Journal of Machine Learning Research, 15: 1929-1958. http://pdfs.semanticscholar.org/6c8b/30f63f265c32e26d999aa1fef5286b8308ad.pdf Suárez, G., Espinosa Aranda, J. M., Cuéllar, A., et al., 2018. A Dedicated Seismic Early Warning Network: The Mexican Seismic Alert System (SASMEX). Seismological Research Letters, 89: 382-391. https://doi.org/10.1785/0220170184 Sun, D. Z., Zhang, R. P., Sun, B. T., 2018. Discussion on the Application of Instrumental Intensity for Seismic Intensity Assessment. Building Structure, 48(S2): 279-283 (in Chinese with English abstract). http://en.cnki.com.cn/Article_en/CJFDTotal-JCJG2018S2057.htm Wang, S., Jiang, J., 2015. Learning Natural Language Inference with LSTM. In: Proceedings of the Human Language Technologies: The 2016 Annual Conference of the North American Chapter of the Association for Computational Linguistics. Association for Computational Linguistics. Association for Computational Linguistics, Stroudsburg. https://doi.org/10.48550/arXiv.1512.08849 Wang, Y., Li, X., Li, L., et al., 2022. New Magnitude Proxy for Earthquake Early Warning Based on Initial Time Series and Frequency. Seismological Research Letters, 93: 216-225. https://doi.org/10.1785/0220210106 Wang, Z., Zhao, B., 2018. Method of Accurate⁃Fast Magnitude Estimation for Earthquake Early Warning: Trial and Application for the 2008 Wenchuan Earthquake. Soil Dynamics and Earthquake Engineering, 109: 227-234. https://doi.org/10.1016/j.soildyn.2018.03.006 Wu, Y., Kanamori, H., 2005. Rapid Assessment of Damage Potential of Earthquakes in Taiwan from the Beginning of P Waves. Bulletin of the Seismological Society of America, 95: 1181-1185. https://doi.org/10.1785/0120040193 Wu, Y. M., Zhao, L., 2006. Magnitude Estimation Using the First Three Seconds P‐Wave Amplitude in Earthquake Early Warning. Geophysical Research Letters, 33(16): L16312. https://doi.org/10.1029/2006GL026871 Xie, Y., Ebad Sichani, M., Padgett, J. E., et al., 2020. The Promise of Implementing Machine Learning in Earthquake Engineering: A State⁃of⁃the⁃Art Review. Earthquake Spectra, 36: 1769-1801. https://doi.org/10.1177/8755293020919419 Xu, Y., Lu, X., Cetiner, B., et al., 2021. Real‐Time Regional Seismic Damage Assessment Framework Based on Long Short‐Term Memory Neural Network. Computer‐Aided Civil and Infrastructure Engineering, 36: 504-521. https://doi.org/10.1080/13632469.2020.1826371 Yamamoto, S., Rydelek, P., Horiuchi, S., et al., 2008. On the Estimation of Seismic Intensity in Earthquake Early Warning Systems. Geophysical Research Letters, 35(7): L07302. https://doi.org/10.1029/2007GL033034 Yu, Y., Si, X., Hu, C., et al., 2019. A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures. Neural Computation, 31: 1235-1270. https://doi.org/10.1162/neco_a_01199 Zhang, W., Li, H., Li, Y., et al., 2021. Application of Deep Learning Algorithms in Geotechnical Engineering: A Short Critical Review. Artificial Intelligence Review, 54: 5633-5673. https://doi.org/10.1007/s10462⁃021⁃09967⁃1 Zhang, W., Phoon, K., 2022. Editorial for Advances and Applications of Deep Learning and Soft Computing in Geotechnical Underground Engineering. Elsevier, 14: 671-673. https://doi.org/10.1016/j.jrmge.2022.01.001 Zhao, Z., Zheng, P., Xu, S., et al., 2019. Object Detection with Deep Learning: A Review. IEEE Transactions on Neural Networks and Learning Systems, 30: 3212-3232. https://doi.org/10.1109/TNNLS.2018.2876865 Zhu, J., Li, S., Song, J., et al., 2021. Magnitude Estimation for Earthquake Early Warning Using a Deep Convolutional Neural Network. Frontiers in Earth Science, 9: 341. https://doi.org/10.3389/feart.2021.653226 Zollo, A., Amoroso, O., Lancieri, M., et al., 2010. A Threshold⁃Based Earthquake Early Warning Using Dense Accelerometer Networks. Geophysical Journal International, 183: 963-974. https://doi.org/10.1111/j.1365⁃246X.2010.04765.x Zuo, R. G., Peng, Y., Li, T., et al., 2021. Challenges of Geological Prospecting Big Data Mining and Integration Using Deep Learning Algorithms. Earth Science, 46(1): 350-358 (in Chinese with English abstract). 陈以伦, 金星, 2016. 一种持续实时预测仪器地震烈度的方法. 地震工程与工程振动, 36(6): 22-29. https://www.cnki.com.cn/Article/CJFDTOTAL-DGGC201606003.htm 储德平, 万波, 李红, 等, 2021. 基于ELMO⁃CNN⁃BiLSTM⁃CRF模型的地质实体识别. 地球科学, 46(8): 3039-3048. doi: 10.3799/dqkx.2020.309 刘如山, 熊明攀, 马强, 等, 2021. 基于仪器地震烈度的变电站高压电气设备易损性研究. 自然灾害学报, 30(2): 14-23. https://www.cnki.com.cn/Article/CJFDTOTAL-ZRZH202102002.htm 孙得璋, 张仁鹏, 孙柏涛, 2018. 浅谈仪器烈度在地震烈度评定中的应用. 建筑结构, 48(S2): 279-283. https://www.cnki.com.cn/Article/CJFDTOTAL-JCJG2018S2057.htm 左仁广, 彭勇, 李童, 等, 2021. 基于深度学习的地质找矿大数据挖掘与集成的挑战. 地球科学, 46(1): 350-358. doi: 10.3799/dqkx.2020.111 -
点击查看大图
图(11) / 表(3)
计量
- 文章访问数: 618
- HTML全文浏览量: 513
- PDF下载量: 62
- 被引次数: 0
