基于卷积神经网络的预警震级分段估算方法  

Segmented Estimation Method for Early Warning Magnitude Based on Convolutional Neural Network

作　　者：任涛[1] 刘昕靓 陈宏峰[2] 马延路[2] REN Tao;LIU Xin-liang;CHEN Hong-feng;MA Yan-lu(School of Software,Northeastern University,Shenyang 110169,China;China Earthquake Networks Center,Beijing 100029,China)

机构地区：[1]东北大学软件学院,辽宁沈阳110169 [2]中国地震台网中心,北京100029

出　　处：《东北大学学报（自然科学版）》2024年第8期1073-1079,共7页Journal of Northeastern University(Natural Science)

基　　金：国家自然科学基金资助项目(62276058,61902057);中央高校基本科研业务费专项资金资助项目(N2217003);辽宁省自然科学基金机器人国家重点实验室项目(2020-KF-12-11);星火计划公关项目(XH21042)。

摘　　要：针对地震预警震级估算问题,提出一种基于卷积神经网络(convolutional neural network,CNN)的震级分段估算方法,该方法以单台站的P波初至后3 s时间的波形作为输入,输出结果为地震波形所属的震级区段(大地震,近震震级M_(L)≥5.0;小地震,M_(L)<5.0).如果波形属于大地震区段,直接发出警报;如果波形属于小地震区段,再进行具体震级的估算.对于震级区段估算,CNN模型的准确率可达98.04%.根据震级估算参数τ_(c)和P_(d)估算的小地震震级平均绝对误差(mean absolute error,MAE)分别为0.20和0.31.结果表明,预警震级分段估算方法可以准确预警大地震,减少大地震漏报率;同时使得小地震震级估算结果更为准确.Aiming at magnitude estimation in earthquake early warning,a segmented estimation method based on convolutional neural network(CNN)is proposed.The input is the waveform starting from the P wave onset and lasting 3 s.The output is the estimated magnitude range(local magnitude M_(L)≥5.0 for large earthquake and M_(L)<5.0 for small earthquake).If the waveform belongs to the large earthquake range,the alarm will be sent directly;if the waveform belongs to the small earthquake range,the specific magnitude value will be estimated.For the estimation of magnitude range,the accuracy of the CNN model can reach 98.04%.The mean absolute errors(MAE)of estimating small earthquake magnitudes based on parameters τ_(c) and P_(d) are 0.20 and 0.31,respectively.The results demonstrate the efficacy of the proposed segmented magnitude estimation method in accurately early warning large earthquakes and reducing the probability of missed warnings.Additionally,it enhances the precision of small earthquake magnitude estimation.

关 键 词：地震预警 震级预警 分段估算 卷积神经网络 震级估算参数 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]