基于深度学习算法的地震动重要持时预测模型

PDF(4540 KB)

振动与冲击 ›› 2023, Vol. 42 ›› Issue (19) : 249-259.

论文

贾佳1,2，公茂盛1,2，赵一男1,2

作者信息 +

Models for predicting ground motion significant duration based on deep learning algorithm

JIA Jia1,2, GONG Maosheng1,2, ZHAO Yinan1,2

Author information +

文章历史 +

摘要

地震动持时对工程结构地震反应有重要影响，发展可靠的地震动持时预测模型是地震工程一个重要课题。本文基于日本K-NET和KiK-net在1997-2021年间获得的67813条地震动记录，将地震动记录按照震源类型分为浅地壳地震、俯冲带板间地震、俯冲带板内地震和上地幔地震四种类型，采用深度学习算法，建立了四种震源类型地震的地震动重要持时预测模型，并与传统预测方程进行了对比。结果表明：采用深度学习算法建立的地震动重要持时预测模型具有合理性和可靠性，能够取得良好的预测效果。四种类型地震的地震动重要持时的预测结果差异明显，尤其在震级较大的情况下，所以预测不同类型地震的地震动持时不应采用同一种预测模型。研究结果和结论可供地震动参数预测、地震区划、结构抗震设计和地震危险性分析等工作参考。

Abstract

The development of a reliable prediction model is important in earthquake engineering. Based on the 67,813 ground motion records obtained by Japan K-NET and KIK-net from 1997 to 2021, the ground motion records are divided into four categories according to the source type: shallow crustal, subduction interface, subduction slab and upper mantle earthquakes. The duration prediction model for four types of earthquakes are established by using deep learning algorithm and compared with the traditional prediction equation. The results show that deep learning algorithm is reasonable and reliable, and can achieve accurate prediction results. There are obvious differences in the prediction results, especially in the case of large magnitude. The different prediction model should be used to predict the significant duration for different type of earthquakes. The results and conclusions can be used as reference for prediction of ground motion parameters, seismic zoning, seismic design and probabilistic analysis of seismic risk.

导出引用

贾佳1,2，公茂盛1,2，赵一男1,2. 基于深度学习算法的地震动重要持时预测模型[J]. 振动与冲击, 2023, 42(19): 249-259

JIA Jia1,2, GONG Maosheng1,2, ZHAO Yinan1,2. Models for predicting ground motion significant duration based on deep learning algorithm[J]. Journal of Vibration and Shock, 2023, 42(19): 249-259

参考文献

[1]. Bommer J J, Marytínezpereira A. The effective duration of earthquake strong motion[J]. Journal of Earthquake Engineering, 1999,3(2): 127-172.
[2]. 钱向东,程玉瑶.地震动持时预测方程的最新研究进展[J].三峡大学学报(自然科学版), 2013, 35(02): 42-46.
Qian Xiangdong, Cheng Yuyao. State of Art Review on Earthquake Ground Motion Duration[J]. J of China Three Gorges Univ (Natural Sciences), 2013, 35(02): 42-46. (in Chinese)
[3]. Bommer J J, Stafford P J, Alarcón J E. Empirical equations for the prediction of the significant, bracketed, and uniform duration of earthquake ground motion[J]. Bulletin of the Seismological Society of America, 2009, 99(6): 3217-3233.
[4]. Lee J, Green R A. An empirical significant duration relationship for stable continental regions [J]. Bulletin of Earthquake Engineering, 2002,39(4): 255-271.
[5]. 徐培彬, 温瑞智. 基于我国强震动数据的地震动持时预测方程[J]. 地震学报, 2018, 40(06): 809-819, 832.
Xu Peibin, Wen Ruizhi. The prediction equations for the significant duration of strong motion in Chinese mainland[J]. Acta Seismologica Sinica, 2018, 40(06): 809-819, 832. (in Chinese)
[6]. Du W, Wang G. Prediction equations for ground-motion significant durations using the NGA-West2 database[J]. Bulletin of the Seismological Society of America, 2017, 107(1): 319-333.
[7]. Bahrampouri M, Rodriguez-Marek A, Green R A. Ground motion prediction equations for significant duration using the KiK-net database[J]. Earthquake Spectra, 2021, 37(2): 903-920.
[8]. Kim T, Song J, Kwon O S. Pre- and post-earthquake regional loss assessment using deep learning [J]. Earthquake Engineering and Structural Dynamics, 2020, 49(7): 657-678.
[9]. 高经纬,涂建维,刘康生,李召.基于GA-LSTM的高层建筑结构地震响应的分散控制研究[J].振动与冲击,2021, 40(10):114-122.
Gao Jingwei, Tu Jianwei, Liu Kangsheng, Li Zhao. Decentralized control for the seismic response of high-rise building structures based on GA-LSTM[J].Journal of Vibration and Shock, 2021, 40(10):114-122. (in Chinese)
[10]. 刘春城,刘佼.基于支持向量机的大跨度拱桥损伤识别方法研究[J].振动与冲击, 2010, 29(07):174-178+244.
Liu Chuncheng, Liu Jiao. Damage identification of a long - span arch bridge based on support vector machine[J]. Journal of Vibration and Shock, 2010, 29(07):174-178+244. (in Chinese)
[11]. 陈伏彬,唐宾芳,蔡虬瑞,李秋胜.大跨平屋盖风荷载特性及风压预测研究[J].振动与冲击, 2021, 40(03):226-232.
Chen Fubin, Tang Binfang, Cai Qiurui, Li Qiusheng. Wind load characteristics and wind pressure prediction of long-span flat roof[J]. Journal of Vibration and Shock, 2021, 40(03):226-232. (in Chinese)
[12]. Ji D F, Liu J, Wen W, et al. Prediction of cumulative absolute velocity based on refined second-order deep neural network[J]. Journal of Earthquake Engineering, 2021, 0(0):1-20.
[13]. 姚兰, 李爽. 工程输入地震动持时的人工智能预测方法[J]. 哈尔滨工业大学学报, 2022, 54(04): 74-81.
YAO Lan, LI Shuang. Prediction of earthquake ground motion duration based on artificial intelligence method[J]. Journal of Harbin Institute of Technology, 2022, 54(04): 74-81.
[14]. Arjun C R, Kumar A. Neural network estimation of duration of strong ground motion using Japanese earthquake records[J]. Soil Dynamics and Earthquake Engineering, 2011, 31(7): 866-872.
[15]. Leonardo A N, Silvia G, Efraín Ovando-Shelley M A M C. A new model for the prediction of earthquake ground-motion duration in Iran[J]. Geofísica Internacional, 2014, 53(3): 221-239.
[16]. Withers K B, Moschetti M P, Thompson E M. A. Machine Learning Approach to Developing Ground Motion Models From Simulated Ground Motions[J]. Geophysical Research Letters, 2020, 47(6): 1-9.
[17]. 余聪,宋晋东,李山有.基于支持向量机的现地地震预警地震动峰值预测[J].振动与冲击, 2021, 40(03):63-72+80.
Yu Cong, Song Jindong, Li Shanyou. Prediction of peak ground motion for on-site earthquake early warning based on SVM[J]. Journal of Vibration and Shock, 2021, 40(03): 63-72+80. (in Chinese)
[18]. Zhao J X, Zhou S, Gao P, et al. An Earthquake Classification Scheme Adapted for Japan Determined by the Goodness of Fit for Ground‐Motion Prediction Equations[J]. Bulletin of the Seismological Society of America, 2015, 105(5): 2750-2763.
[19]. Hancock J, Bommer J J. Using spectral matched records to explore the influence of strong-motion duration on inelastic structural response[J]. Soil Dynamics and Earthquake Engineering, 2007, 27(4): 291-299.
[20]. Raghunandan M, Liel A B. Effect of ground motion duration on earthquake-induced structural collapse[J]. Structural Safety, 2013, 41: 119-133.
[21]. Chandramohan R, Baker J W, Deierlein G G. Quantifying the influence of ground motion duration on structural collapse capacity using spectrally equivalent records[J]. Earthquake Spectra, 2016, 32(2): 927-950.
[22]. Trifunac M D, Brady A G. A study on the duration of strong earthquake ground motion[J]. Bulletin of the Seismological Society of America, 1975, 65(3): 581-626.
[23]. Somerville P G, Yoshimura J. The effect of critical Moho reflections on strong ground motions recorded in San Francisco and Oakland during the 1989 Lorna Prieta earthquake[J]. Geophysical Research Letters, 1990, 17(8): 1203-1206.
[24]. Bahrampouri M, Rodriguez-Marek A, Shahi S, et al. An updated database for ground motion parameters for KiK-net records[J]. Earthquake Spectra, 2021, 37(1):505-522.
[25]. Boore D M, Thompson E M. Cadet Héloïse. Regional Correlations of VS30 and Velocities Averaged Over Depths Less Than and Greater Than 30 Meters[J]. Bulletin of the Seismological Society of America, 2011, 1019(6): 3046-3059.
[26]. Abadi M, Barham P, Chen J, et al. TensorFlow: A system for large-scale machine learning[J]. USENIX Association, 2016,0(0):265-283
[27]. Kingma D P, Ba J L. Adam:A method for stochastic optimization[C]\ In Proceedings of the International Conference on Learning Representations (ICLR), 2015.
[28]. Han J, Moraga C. The influence of the sigmoid function parameters on the speed of backpropagation learning[M]. Mira, J., Sandoval F.: From Natural to Artificial Neural Computation, 1995.
[29]. Baltay A S, Hanks T C, Abrahamson N A. Uncertainty, variability, and earthquake physics in ground-motion prediction equations[J]. Bulletin of the Seismological Society of America, 2017, 107(4): 1754-1772.
[30]. Afshari K, Stewart J P. Physically Parameterized Prediction Equations for Significant Duration in Active Crustal Regions[J]. Earthquake Spectra, 2016, 32(4):2057-2081.