基于改进残差网络的大地电磁反演研究

doi:10.11720/wtyht.2023.0186

Abstract
Figure/Table
References
Related Citation (15)

Download: PDF (13460 KB) HTML (1 KB)
Export: BibTeX | EndNote (RIS)

Abstract

Traditional inversion techniques rely on initial models and exhibit prolonged inversion times. This study proposed a magnetotelluric inversion method based on an improved residual network. Specifically, geoelectric models of varying shapes and resistivity values were established, and apparent resistivity data were obtained using the TM mode, forming a dataset. Then, a novel inversion network-iResNet (an improved residual network)-was established by improving classic residual network ResNet, and the new network was trained using the afore-mentioned data set. Finally, the apparent resistivity data were input to trained network, directly producing inversion results. The experimental results demonstrate that the method proposed in this study can accurately determine the positions, shapes, and resistivity values of the geoelectric models through swift inversion, suggesting high generalization and anti-noise capabilities. Therefore, this method can effectively deermine measured magnetotelluric data.

Key words： residual network magnetotelluric inversion

Received: 04 May 2023 Published: 23 January 2024

P631.1

	Service

	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors
	Si-Ping LI
	Cai-Yun LIU
	Jie XIONG
	Hui-Xiao TIAN
	Fang WANG

Cite this article:

Si-Ping LI,Cai-Yun LIU,Jie XIONG, et al. Magnetotelluric inversion based on an improved residual network[J]. Geophysical and Geochemical Exploration, 2023, 47(6): 1508-1518.

URL:

https://www.wutanyuhuatan.com/EN/10.11720/wtyht.2023.0186 OR https://www.wutanyuhuatan.com/EN/Y2023/V47/I6/1508

13]
">

Sructure diagram of convolutional neural network^[13]

General structure of ResNet

Two typical residual structures

Inversion process

9]
">

iResNet network structure^[9]

Schematic diagram of models

分类	参数设置
数据集	训练集	8088
数据集	测试集	899
网络设置	学习率	$η = 0.001$
	激活函数	ReLU
	优化器	Adam
	L2正则化	$λ$ =0.003
	Dropout	0.5
训练过程	Epochs	2000
训练过程	Batch size	500

iResNetFc network parameter settings

Inversion results of part of the validation set samples

Inversion results of CNN

Inversion results of iResNet

Inversion results of experiment one

Inversion results of experiment two

Inversion results of experiment three

Apparent resistivity after adding white Gaussian noise

Inversion results after adding 3% Gaussian white noise

Inversion results after adding 5% Gaussian white noise

19]
">

Detailed geological map of parts of southern Africa^[19]

Apparent resistivity map of Northern zone

iResNet inversion results for Northern zone

18]
">

KIM+ETO line regularization inversion results^[18]

[1]	童孝忠. 大地电磁测深有限单元法正演与混合遗传算法正则化反演研究[D]. 长沙: 中南大学, 2008.
[1]	Tong X Z. Research of forward using finite element method and regularized inversion using hybrid genetic algorithm in magnetotelluric sounding[D]. Changsha: Central South University, 2008.
[2]	姜奋勇, 叶益信, 陈海文, 等. 基于非结构网格的带地形MT二维Occam反演及应用[J]. 物探与化探, 2022, 46(2):482-489.
[2]	Jiang F Y, Ye Y X, Chen H W, et al. Application of 2D inversion of magnetotelluric data bearing terrain information based on an unstructured mesh[J]. Geophysical and Geochemical Exploration, 2022, 46(2):482-489.
[3]	谭捍东. 大地电磁法三维快速松弛反演[J]. 地球物理学报, 2003, 46(6):850-854.
[3]	Tan H D. Three-dimensional rapid relaxation inversion fir the magnetotelluric method[J]. Chinese Journal of Geophysics, 2003, 46(6):850-854,
[4]	阮帅, 汤吉, 陈小斌, 等. 三维大地电磁自适应L1范数正则化反演[J]. 地球物理学报, 2020, 63(10):3896-3911.
[4]	Ruan S, Tang J, Chen X B, et al. Three-dimensional magnetotelluric inversion based on adaptive L1-norm regularization[J]. Chinese Journal of Geophysics, 2020, 63(10):3896-3911.
[5]	熊杰, 孟小红, 刘彩云, 等. 基于差分进化的大地电磁反演[J]. 物探与化探, 2012, 36(3):448-451.
[5]	Xiong J, Meng X H, Liu C Y, et al. Magnetotelluric inversion based on differential evolution[J]. Geophysical and Geochemical Exploration, 2012, 36(3):448-451.
[6]	Liu S, Hu X, Liu T, et al. Ant colony optimization inversion of surface and borehole magnetic data under lithological constraints[J]. Journal of Applied Geophysics, 2015, 112(1):115-128.
[7]	Conway D, Alexander B, King M, et al. Inverting magnetotelluric responses in a three-dimensional earth using fast forward approximations based on artificial neural networks[J]. Computers & Geosciences, 2019, 127(6):44-52.
[8]	Noh K, Yoon D, Byun J. Imaging subsurface resistivity structure from airborne electromagnetic induction data using deep neural network[J]. Exploration Geophysics, 2020, 51(2):214-220.
[9]	Liu Z, Chen H, Ren Z, et al. Deep learning audio magnetotellurics inversion using residual-based deep convolution neural network[J]. Journal of Applied Geophysics, 2021, 188(6433):104309.
[10]	Wang H, Liu W, Xi Z Z. Nonlinear inversion for magnetotelluric sounding based on deep belief network[J]. Journal of Central South University, 2019, 26(9):2482-2494.
[11]	廖晓龙, 张志厚, 姚禹, 等. 基于卷积神经网络的大地电磁反演[J]. 中南大学学报:自然科学版, 2020, 51(9):2546-2557.
[11]	Liao X L, Zhang Z H, Yao Y, et al. Magnetotelluric inversion based on convolutional neural network[J]. Journal of Central South University:Science and Technology, 2020, 51(9):2546-2557.
[12]	范振宇. 基于卷积神经网络的大地电磁深度学习反演研究[D]. 北京: 中国地质大学(北京), 2020.
[12]	Fan Z Y. Magnetotelluric deep learning inversion based on convolutional neural network[D]. Beijing: China University of Geosciences(Beijing), 2020.
[13]	刘倩. 基于深度学习的重力异常反演[D]. 荆州: 长江大学, 2021.
[13]	Liu Q. Inversion of gravity anomaly based on deep learning[D]. Jingzhou: Yangtze University, 2021.
[14]	He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE conference on computer vision and pattern recognition. 2016:770-778.
[15]	Zhang X Y, Zou J H, He K M, et al. Accelerating very deep convolutional networks for classification and detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(10):1943-1955
[16]	王蓉, 熊杰, 刘倩, 等. 基于深度神经网络的重力异常反演[J]. 物探与化探, 2022, 46(2) :451-458.
[16]	Wang R, Xiong J, Liu Q, et al. Inversion of gravity anomalies based on a deep neural network[J]. Geophysical and Geochemical Exploration, 2022, 46(2):451-458.
[17]	Begg G C, Griffin W L, Natapov L M, et al. The lithospheric architecture of Africa:Seismic tomography,mantle petrology,and tectonic evolution[J]. Geosphere, 2009, 5(1):23-50.
[18]	Khoza T D, Jones A G, Muller M R, et al. Lithospheric structure of an Archean craton and adjacent mobile belt revealed from 2-D and 3-D inversion of magnetotelluric data:Example from southern Congo craton in northern Namibia[J]. Journal of Geophysical Research:Solid Earth, 2013, 118(8):4378-4397.
[19]	Singletary S J, Hanson R E, Martin M W, et al. Geochronology of basement rocks in the Kalahari Desert,Botswana,and implications for regional Proterozoic tectonics[J]. Precambrian Research, 2003, 121(1/2):47-71.
[20]	Goscombe B, Hand M, Gray D, et al. The metamorphic architecture of a transpressional orogen:The Kaoko Belt,Namibia[J]. Journal of Petrology, 2003, 44(4):679-711.