U型神经网络结合BiLSTM网络的地震相识别系统  

Research on Seismic Phase Recognition System Using U-shaped Neural Network Combined with BiLSTM Network

作　　者：王天哲 WANG Tianzhe(China University of Geosciences(Beijing),Beijing 100083,China)

机构地区：[1]中国地质大学(北京),北京100083

出　　处：《华北地震科学》2024年第4期15-20,共6页North China Earthquake Sciences

基　　金：国家自然科学基金面上项目(41974060)。

摘　　要：以BiLSTM网络为基础识别框架,通过在BiLSTM网络结构中引入U型卷积神经网络进行改进,并采用改进的BiLSTM网络对地震相进行识别,实现了地震P波和S波震相的精确识别。仿真结果表明:该方法可有效、准确地识别地震P波和S波震相,平均识别正确率为90.01%,平均漏检率和均方根误差分别为11.00%和0.23,相较于BiLSTM网络以及常用地震相识别MEA-BP神经网络模型和CNN模型,该方法对地震相的识别精度更高,具有明显的优越性,为实现地震相的精确识别提供了参考。To improve the accuracy of seismic phase identification,a deep learning-based seismic phase identification method is proposed.This method is based on the BiLSTM network recognition framework,improved by introducing a U-shaped convolutional neural network into the BiLSTM network structure.The improved BiLSTM network was used to identify seismic phases,achieving accurate identification of seismic P-wave and S-wave phases.The simulation results show that this method can effectively and accurately identify seismic P-wave and S-wave phases,with an average recognition accuracy of 90.01%,an average missed detection rate of 11.00%,and a root mean square error of 0.23.Compared with BiLSTM network,commonly used seismic phase recognition MEA-BP neural network models,and CNN models,this method has higher recognition accuracy for seismic phases and obvious advantages,providing a reference for achieving accurate identification of seismic phases.

关 键 词：深度学习 地震相识别 BiLSTM网络 U型卷积神经网络 

分 类 号：P315.7[天文地球—地震学]