基于卷积神经网络的微地震初至拾取  

Microseismic first arrival picking based on convolutional neural network

作　　者：李政超 王维波[1] 高明[1] 盛立[1] LI ZhengChao;WANG WeiBo;GAO Ming;SHENG Li(College of Control Science and Engineering,China University of Petroleum(East China),Qingdao 266580,China)

机构地区：[1]中国石油大学(华东),控制科学与工程学院,青岛266580

出　　处：《地球物理学进展》2022年第3期1060-1069,共10页Progress in Geophysics

基　　金：国家自然科学基金项目(62173343,62073339,62033008);山东省自然科学基金项目(ZR2020YQ49)联合资助。

摘　　要：微地震初至拾取是微地震领域的核心问题之一,有效信号初至拾取的正确与否,直接影响到后续数据处理的精度.考虑到人工拾取初至费时、费力,而传统的自动拾取方法又非常依赖于自身参数的设置,受主观因素影响较大,因此将深度学习中卷积神经网络(CNN)结构引入到微地震数据处理中.针对微地震波形信号的特点和初至拾取的要求,搭建合适的CNN模型,以已标注好初至波到时的合成和实际微地震信号作为数据集样本,对模型进行训练测试,得到拾取效果最优的网络模型.该方法只需输入预处理后的原始微地震波形信号,就可以直接输出信号初至时间.与传统自动拾取方法相比,该方法具有拾取流程简单、运算时间短、拾取效率高等特点.使用训练好的网络模型分别拾取待检测的合成微地震信号和实际监测信号的初至波到时,结果表明基于CNN的初至拾取方法可以在短时间内较为准确地拾取微地震信号初至.Microseismic first arrival picking is one of the core issues in the field of microseismic.Whether the first arrival picking is correct or not directly affects the accuracy of subsequent processing.Considering that manual picking is time-consuming and laborious,and the traditional automatic picking method is very dependent on the setting of its own parameters,which is greatly affected by subjective factors,so the Convolutional Neural Network(CNN)structure in deep learning is introduced into microseismic data processing.According to the characteristics of microseismic waveform signals and the requirements of first arrival picking,a suitable CNN model is built.Taking the microseismic signal data marked with the arrival time of the first arrival as samples,the model is trained and tested,and the network model with the best picking effect is obtained.The method only needs to input the original microseismic waveform signal after preprocessing,and then it can output the first arrival time of the signal directly.Compared with the traditional automatic picking method,this method has the advantages of simple picking process,short operation time and high picking efficiency.The trained network model is used to pick up the first arrival time of the synthetic microseismic signal and the actual monitoring signal to be detected,and the results show that the first arrival picking method based on CNN can pick up the first arrivals of microseismic signals relatively accurately in a short time.

关 键 词：微地震 初至拾取 卷积神经网络 模型训练测试 实际信号拾取 

分 类 号：P631[天文地球—地质矿产勘探]