摘要
通过积分变换,可将满足扩散方程的瞬变电磁场转换成虚拟波场.传统的波场转换方法对正则化因子依赖性较强,导致计算结果稳定性差.为了克服正则化因子的干扰,本文提出采用深度学习的方法将瞬变电磁场转换到虚拟波场.首先分别计算相同电阻率模型下的瞬变电磁场和拟地震波场,构建训练集与对应的标签.然后设计复合型神经网络C-Unet,相比于传统的Unet,该网络进一步提升了模型特征的学习能力.模型测试结果表明,C-Unet在迭代误差和预测结果精度均优于Unet网络.最后利用实测数据进一步验证了本文提出的方法的有效性.
Through integral transformation,transient electromagnetic fields that satisfy the diffusion equation can be transformed into virtual wave fields.However,traditional wave field transformation methods have a strong dependence on regularization factors,resulting in poor stability of the calculation results.This article uses deep learning methods to convert transient electromagnetic fields into virtual wave fields.Firstly,the transient electromagnetic fields and pseudo seismic wave fields under the same resistivity model are calculated separately,and a training set and corresponding labels are constructed.In order to further enhance the learning ability of neural networks,this paper constructs a composite neural network C-Unet,which adds a convolutional network module before inputting data into the Unet network.Finally,the effectiveness of the proposed method was verified through model testing and actual measurement data.
作者
薛俊杰
周楠楠
常江浩
余传涛
鲁凯亮
范克睿
XUE JunJie;ZHOU Nannan*;CHANG JiangHao;YU ChuanTao;LU KaiLiang;FAN KeRui(Key Laboratory of Deep Petroleum Intelligent Exploration and Development,Institute of Geology and Geophysics Chinese Academy of Sciences,Beijing 100029,China;College of Earth and Planetary Sciences,University of Chinese Academy of Sciences,Beijing 100049,China;Key Laboratory of Intelligent Detection and Equipment for Underground Space of Beijing-Tianjin-Hebei Urban Agglomeration,Ministry of Natural Resources,Hebei GEO University,Shijiazhuang 050031,China;College of Mining Engineering,Taiyuan University of Technology,Taiyuan 030024,China;China University of Mining and Technology,Xuzhou Jiangsu 221116,China;College of Civil Engineering and Architecture,Shandong University of Science and Technology,Qingdao Shandong 266590,China)
出处
《地球物理学报》
北大核心
2025年第8期3282-3290,共9页
Chinese Journal of Geophysics
基金
国家青年科学基金(42404146)
河北省自然科学基金(D2023403003)
陕西投资集团有限公司科技研发项目(SIGC-2024-KY-02)
陕西省煤田地质集团有限公司2023年科技研发重大专项(SMDZ-2023ZD-8)
陕西省自然科学基金(2024JC-YBQN-0317)资助.
关键词
波场转换
深度学习
卷积神经网络
瞬变电磁场
虚拟波场
Wave field conversion
Deep learning
Convolutional neural networks
Transient electromagnetic fields
Virtual wave field
