Transient electromagnetic numerical simulation based on the transformer neural network

Authors

WANG Yunhong, CCTEG Xi’an Research Institute (Group) Co. Ltd., Xi’an 710077, ChinaFollow

Abstract

Objective The deep learning-based transient electromagnetic (TEM) forward and inverse modeling methods are data-driven, requiring considerable numerical simulation results as supervisory data to train and assess neural networks. The conventional finite-difference time-domain (FDTD) method for TEM numerical simulation necessitates an iterative solution of time-domain Maxwell's equations. Therefore, this method is time-consuming and computationally intensive, failing to meet the data demand of deep learning-based TEM inversion. Methods This study introduced deep learning for TEM numerical simulation. Based on the transformer neural network architecture, a neural network for deep learning-based TEM numerical simulation was designed using an encoder-decoder structure. This neural network comprised a 3D gridding module, a patch embedding module, a transformer encoder module, and a linear decoder module. With geoelectric parameters as inputs, this neural network output corresponding TEM responses at the center of the loop source. It was trained for over 200 epochs using the optimization strategy of stochastic gradient descent with momentum combined with adaptive moment estimation (Adam)—an adaptive learning rate algorithm—on a server equipped with four NVIDIA V100 GPUs. Results and Conclusions The trained network for deep learning-based TEM numerical simulation was employed to predict the TEM responses of the loop sources of four geoelectric models in real time: a homogeneous halfspace model, a layered model, a plate model, and a 3D volume model. The test results of the validation set reveal that the numerical simulation results after the turn-off time derived using the neural network exhibited low mean relative errors (MREs) of less than 2% compared to analytical solutions, linear digital filtering solutions, and FDTD numerical simulation results. Meanwhile, the TEM numerical simulation results were obtained within 1 s. Therefore, the proposed neural network exhibited fast and accurate calculations. This study will provide a theoretical foundation and data support for research on deep learning-based TEM inversion.

Keywords

transient electromagnetic(TEM), deep learning, neural network, numerical simulation

DOI

10.12363/issn.1001-1986.24.07.0499

Recommended Citation

W. (2024) "Transient electromagnetic numerical simulation based on the transformer neural network," Coal Geology & Exploration: Vol. 52: Iss. 11, Article 14.
DOI: 10.12363/issn.1001-1986.24.07.0499
Available at: https://cge.researchcommons.org/journal/vol52/iss11/14

Reference

[1] 刘树才，刘志新，姜志海. 瞬变电磁法在煤矿采区水文勘探中的应用[J]. 中国矿业大学学报，2005，34(4)：414−417.

LIU Shucai，LIU Zhixin，JIANG Zhihai. Application of TEM in hydrogeological prospecting of mining district[J]. Journal of China University of Mining & Technology，2005，34(4)：414−417.

[2] 于景邨，刘志新，刘树才，等. 深部采场突水构造矿井瞬变电磁法探查理论及应用[J]. 煤炭学报，2007，32(8)：818−821.

YU Jingcun，LIU Zhixin，LIU Shucai，et al. Theoretical analysis of mine transient electromagnetic method and its application in detecting water burst structures in deep coal stope[J]. Journal of China Coal Society，2007，32(8)：818−821.

[3] 于景邨，刘志新，岳建华，等. 煤矿深部开采中的地球物理技术现状及展望[J]. 地球物理学进展，2007，22(2)：586−592.

YU Jingcun，LIU Zhixin，YUE Jianhua，et al. Development and pospect of geophysical technology in deep mining[J]. Progress in Geophysics，2007，22(2)：586−592.

[4] 白登海，Maxwell A Meju，卢健，等. 时间域瞬变电磁法中心方式全程视电阻率的数值计算[J]. 地球物理学报，2003，46(5)：697−704.

BAI Denghai，MEJU M A，LU Jian，et al. Numerical calculation of all-time apparent resistivity for the central loop transient electromagnetic method[J]. Chinese Journal of Geophysics，2003，46(5)：697−704.

[5] 戚志鹏，李貅，郭文波，等. 瞬变电磁水平分量视电阻率定义[J]. 煤炭学报，2011，36(增刊1)：88−93.

QI Zhipeng，LI Xiu，GUO Wenbo，et al. Definition of apparent resistivity of transient electromagnetic horizontal component[J]. Journal of China Coal Society，2011，36(Sup.1)：88−93.

[6] 杨海燕，邓居智，张华，等. 矿井瞬变电磁法全空间视电阻率解释方法研究[J]. 地球物理学报，2010，53(3)：651−656.

YANG Haiyan，DENG Juzhi，ZHANG Hua，et al. Research on full-space apparent resistivity interpretation technique in mine transient electromagnetic method[J]. Chinese Journal of Geophysics，2010，53(3)：651−656.

[7] 薛国强，李貅，底青云. 瞬变电磁法理论与应用研究进展[J]. 地球物理学进展，2007，22(4)：1195−1200.

XUE Guoqiang，LI Xiu，DI Qingyun. The progress of TEM in theory and application[J]. Progress in Geophysics，2007，22(4)：1195−1200.

[8] 薛国强，李貅，郭文波，等. 大回线源瞬变电磁场响应特性[J]. 石油地球物理勘探，2007，42(5)：586−590.

XUE Guoqiang，LI Xiu，GUO Wenbo，et al. Characters of response of large-loop transient electro-magnetic field[J]. Oil Geophysical Prospecting，2007，42(5)：586−590.

[9] 范亮，钱荣毅. 瞬变电磁法在煤矿采空区的应用研究[J]. 工程地球物理学报，2011，8(1)：29−33.

FAN Liang，QIAN Rongyi. Application of transient electromagnetic method to coal mine gobs[J]. Chinese Journal of Engineering Geophysics，2011，8(1)：29−33.

[10] 孙怀凤，程铭，吴启龙，等. 瞬变电磁三维FDTD正演多分辨网格方法[J]. 地球物理学报，2018，61(12)：5096−5104.

SUN Huaifeng，CHENG Ming，WU Qilong，et al. A multi-scale grid scheme in three-dimensional transient electromagnetic modeling using FDTD[J]. Chinese Journal of Geophysics，2018，61(12)：5096−5104.

[11] PUZYREV V. Deep learning electromagnetic inversion with convolutional neural networks[J]. Geophysical Journal International，2019，218(2)：817−832.

[12] LIU Bin，GUO Qian，LI Shucai，et al. Deep learning inversion of electrical resistivity data[J]. IEEE Transactions on Geoscience and Remote Sensing，2020，58(8)：5715−5728.

[13] SHAHRIARI M，PARDO D，RIVERA J A，et al. Error control and loss functions for the deep learning inversion of borehole resistivity measurements[EB/OL]. 2020：2005.08868. https://arxiv.org/abs/2005.08868v3.

[14] OH S，NOH K，SEOL S J，et al. Cooperative deep learning inversion of controlled-source electromagnetic data for salt delineation[J]. Geophysics，2020，85(4)：E121−E137.

[15] 范涛，薛国强，李萍，等. 瞬变电磁长短时记忆网络深度学习实时反演方法[J]. 地球物理学报，2022，65(9)：3650−3663.

FAN Tao，XUE Guoqiang，LI Ping，et al. TEM real-time inversion based on long-short term memory network[J]. Chinese Journal of Geophysics，2022，65(9)：3650−3663.

[16] 李瑞友，白细民，张勇，等. 基于小波包分解与GA优化BP神经网络的瞬变电磁反演[J]. 吉林大学学报(地球科学版)，2024，54(3)：1003−1015.

LI Ruiyou，BAI Ximin，ZHANG Yong，et al. Using wavelet packet denoising and BP neural network based on GA optimization for transient electromagnetic inversion[J]. Journal of Jilin University (Earth Science Edition)，2024，54(3)：1003−1015.

[17] 燕帅，殷长春，苏扬，等. 基于卷积神经网络的瞬变电磁数据快速成像算法研究[J]. 地球物理学报，2023，66(10)：4290−4300.

YAN Shuai，YIN Changchun，SU Yang，et al. Fast imaging of transient electromagnetic data based on convolutional neural network[J]. Chinese Journal of Geophysics，2023，66(10)：4290−4300.

[18] 古瑶，解海军，周子鹏，等. 基于Attention机制的CNN-BiLSTM瞬变电磁实时反演方法[J]. 煤田地质与勘探，2023，51(10)：134−143.

GU Yao，XIE Haijun，ZHOU Zipeng，et al. An Attention mechanism-based CNN-BiLSTM real-time transient electromagnetic method[J]. Coal Geology & Exploration，2023，51(10)：134−143.

[19] 程久龙，王慧杰，徐忠忠，等. 基于全卷积神经网络的钻孔瞬变电磁法岩层富水性预测研究[J]. 煤田地质与勘探，2023，51(1)：289−297.

CHENG Jiulong，WANG Huijie，XU Zhongzhong，et al. Research on aquifer water abundance evaluation by borehole transient electromagnetic method based on FCNN[J]. Coal Geology & Exploration，2023，51(1)：289−297.

[20] 游希然，张继锋，石宇. 基于人工神经网络的瞬变电磁成像方法[J]. 物探与化探，2023，47(5)：1206−1214.

YOU Xiran，ZHANG Jifeng，SHI Yu. Artificial neural network-based transient electromagnetic imaging[J]. Geophysical and Geochemical Exploration，2023，47(5)：1206−1214.

[21] 冼锦炽，蔡红柱，熊咏春，等. 基于深度学习的地面拖曳式瞬变电磁快速成像方法[J]. 工程地球物理学报，2022，19(4)：536−545.

XIAN Jinchi，CAI Hongzhu，XIONG Yongchun，et al. Ground-based towed transient electromagnetic imaging method based on deep learning[J]. Chinese Journal of Engineering Geophysics，2022，19(4)：536−545.