基于U-Net网络的FWI地震低频恢复方法

doi:10.11720/wtyht.2023.1125

摘要
图/表
参考文献
相关文章
Metrics

全文: PDF(5479 KB) HTML
输出: BibTeX | EndNote (RIS)

摘要

由于实际地震资料中缺乏低频数据,使得全波形反演易陷入局部极小值,导致反演质量差,结果不可靠。为解决这一问题,本文利用数据驱动低频恢复映射思想,分别利用高通和低通滤波器从原始数据中分离出高频数据和低频数据,对其进行一系列数据预处理操作,将处理后的数据作为模型的训练集;利用U-Net网络为基础构建模型,建立高低频之间的映射关系。为了有效防止模型过拟合,本文在U-Net模型基础上添加了Dropout层和批处理层。利用训练后的模型从高频数据中预测对应低频数据并进行逆数据预处理,对比分析逆数据预处理后的预测低频数据和真实低频数据之间误差,并利用多尺度全波形反演在洼陷模型和Marmousi模型进行有效性验证。实验结果表明:训练与测试数据的预测低频与真实低频数据的平均相对误差为5.02%和13.32%,误差较小,数据吻合良好;洼陷模型、Marmousi模型以及实际数据的反演结果表明加入预测低频后反演质量得到显著提高,并且对处理含较大噪声的数据也有很好的效果。

	服务

	把本文推荐给朋友
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	王莉利
	杜功鑫
	高新成
	王宁
	王维红

关键词 ：全波形反演, 低频恢复, 局部极小值, 能量均衡, U-Net

Abstract：

The lack of low-frequency data in actual seismic data makes the full waveform inversion (FWI) tend to fall into the local minimum,resulting in poor inversion quality and unreliable results.In view of this,data-driven low-frequency recovery mapping was adopted in this study.First,high-pass and low-pass filters were employed to separate high-frequency and low-frequency data from raw data,respectively,and then data preprocessing was carried out.The processed data were used as the training set of the model.Then,the model was built based on the U-Net to establish the mapping relationship between high and low frequencies.To effectively prevent the model from overfitting,the dropout layer and batch processing layer were added based on the U-Net model.Finally,the trained model was used to predict the corresponding low-frequency data from the high-frequency data and conduct inverse data preprocessing.The errors between the predicted low-frequency data after inverse data preprocessing and the real low-frequency data were compared and analyzed,The effectiveness of multi-scale FWI was verified using the depression and Marmousi models.The experimental results show that the average relative errors between the predicted low-frequency data and the real low-frequency data were 5.02% and 13.32%,respectively for training and test data,indicating small errors and high data coincidence.The inversion results of the depression model,the Marmousi model,and actual data show that the prediction of low-frequency data significantly improved the inversion quality and delivered a great performance in the processing of data with much noise.

Key words： full waveform inversion low frequency recovery local minimum energy equalization U-Net

收稿日期: 2022-03-22 修回日期: 2023-02-01 出版日期: 2023-04-20

ZTFLH:

P631.4

基金资助:国家重点自然科学基金项目“粘声介质最小二乘逆时偏移及全波形反演研究”(41930431);国家自然科学基金项目“基于数据驱动的逆散射级数层间多次波压制方法”(41974116);黑龙江省自然科学基金联合引导项目“深部储层衰减补偿逆时偏移成像研究”(LH2021D009);东北石油大学引导性创新基金项目“基于自适应卷积神经网络的地震速度建模方法研究”(2020YDL-03)

引用本文:

王莉利, 杜功鑫, 高新成, 王宁, 王维红. 基于U-Net网络的FWI地震低频恢复方法[J]. 物探与化探, 2023, 47(2): 391-400.
WANG Li-Li, DU Gong-Xin, GAO Xin-Cheng, WANG Ning, WANG Wei-Hong. FWI seismic low frequency recovery method based on the U-Net. Geophysical and Geochemical Exploration, 2023, 47(2): 391-400.

链接本文:

https://www.wutanyuhuatan.com/CN/10.11720/wtyht.2023.1125 或 https://www.wutanyuhuatan.com/CN/Y2023/V47/I2/391

Fig.1 改进后的U-Net模型网络结构

Table 1 实验环境

Table 2 有限差分正演参数

Fig.2 观测数据频谱(部分)

Fig.3 训练集高频数据与对应低频标签可视化展示
a—6张大小64×64的训练集输入高频数据;b—图a中高频数据对应低频标签数据

Fig.4 真实低频与预测低频道集图像对比
a—缺失低频数据;b—预测低频数据;c—真实低频数据;d—预测和真实低频数据之间的误差

Fig.5 真实低频与预测低频频谱对比

Fig.6 原始与补充低频数据道集图像对比
a—原始数据;b—缺失低频数据;c—预测低频数据;d—补充预测低频后的数据

Fig.7 观测数据与补充低频数据的频谱对比

Fig.8 第250道的波形对比

Fig.9 测试集高频数据与对应预测低频结果可视化展示
a—6张大小为64×64的测试高频数据;b—图a中数据对应预测低频数据

Fig.10 网络模型提取特征过程可视化展示
a~e—编码过程下采样提取特征;f~i—解码过程上采样恢复特征

Fig.11 洼陷模型试验反演结果
a—洼陷速度模型;b—初始速度模型;c—原始数据的反演结果;d—缺失低频的反演结果;e—补充预测低频

Fig.12 洼陷模型某一炮原始数据及其预测低频数据道集

Fig.13 Marmousi模型试验反演结果
a—Marmousi速度模型;b—初始速度模型;c—原始数据的反演结果;d—缺失低频的反演结果;e—补充预测低频的反演结果

Fig.14 Marmousi某1炮原始数据及其预测低频数据道集

Fig.15 实际数据的反演结果

Fig.16 加入预测低频的实际数据的反演结果

Fig.17 洼陷原始数据及加入噪声后的数据道集

Fig.18 加入噪声的洼陷原始数据及其预测低频数据的道集

Fig.19 加入噪声的洼陷原始数据反演结果

Fig.20 补充低频的噪声洼陷原始数据反演结果

[1]	Tarantola A. Linearized inversion of seismic reflection data[J]. Geophysical Prospecting, 1984, 32(6):998-1015. doi: 10.1111/gpr.1984.32.issue-6
[2]	Wu R S, Luo J, Wu B. Seismic envelope inversion and modulation signal model[J]. Geophysics, 2014, 79(3):WA13-WA24. doi: 10.1190/geo2013-0294.1
[3]	Boßmann F, Ma J. Asymmetric chirplet transform for sparse representation of seismic data[J]. Geophysics, 2015, 80(6):WD89-WD100. doi: 10.1190/geo2015-0063.1
[4]	Shin C, Cha Y H. Waveform inversion in the Laplace domain[J]. Geophysical Journal International, 2008, 173(3):922-931. doi: 10.1111/gji.2008.173.issue-3
[5]	Li Y E, Demanet L. Full-waveform inversion with extrapolated low-frequency data[J]. Geophysics, 2016, 81(6):R339-R348. doi: 10.1190/geo2016-0038.1
[6]	Hu W. FWI without low frequency data-beat tone inversion[C]// SEG Technical Program Expanded Abstracts 2014,Society of Exploration Geophysicists, 2014:1116-1120.
[7]	刘张聚, 童思友, 方云峰, 等. 基于时域加权的拉普拉斯—频率域弹性波全波形反演[J]. 石油地球物理勘探, 2021, 56(2):302-312.
[7]	Liu Z J, Tong S Y, Fang Y F, el at. Full waveform inversion of elastic wave in Laplace frequency domain based on time domain weighting[J] Oil Geophysical Prospecting, 2021, 56(2):302-312.
[8]	Jin Y, Hu W, Wu X, et al. Learn low wavenumber information in FWI via deep inception based convolutional networks[C]// 2018 SEG International Exposition and Annual Meeting,OnePetro, 2018.
[9]	Sun H, Demanet L. Low frequency extrapolation with deep learning[C]// SEG Technical Program Expanded Abstracts 2018,Society of Exploration Geophysicists, 2018:2011-2015.
[10]	Ovcharenko O, Kazei V, Kalita M, et al. Deep learning for low-frequency extrapolation from multioffset seismic data[J]. Geophysics, 2019, 84(6):R989-R1001. doi: 10.1190/GEO2018-0884.1
[11]	Ovcharenko O, Kazei V, Peter D, et al. Transfer learning for low frequency extrapolation from shot gathers for FWI applications[C]// 81^st EAGE Conference and Exhibition 2019,European Association of Geoscientists & Engineers, 2019(1):1-5.
[12]	Sun H, Demanet L. Deep learning for low frequency extrapolation of multicomponent data in elastic full waveform inversion[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60:1-11.
[13]	Fang J, Zhou H, Elita Li Y, et al. Data-driven low-frequency signal recovery using deep-learning predictions in full-waveform inversion[J]. Geophysics, 2020, 85(6):A37-A43. doi: 10.1190/geo2020-0159.1
[14]	Hu W, Jin Y, Wu X, et al. Progressive transfer learning for low-frequency data prediction in full waveform inversion[J]. Geophysics, 2021, 86(4):1-82.
[15]	Mousavi A, Baraniuk R G. Learning to invert:Signal recovery via deep convolutional networks[C]// 2017 IEEE International Conference on Acoustics,Speech and Signal Processing (ICASSP),IEEE, 2017:2272-2276.
[16]	Ronneberger O, Fischer P, Brox T. U-net:Convolutional networks for biomedical image segmentation[C]// International Conference on Medical Image Computing and Computer-Assisted Intervention,Springer,Cham, 2015:234-241.
[17]	Zeiler M D. Adadelta:an adaptive learning rate method[J]. Computer Science, 2012:1-6.
[18]	Bunks C, Saleck F M, Zaleski S, et al. Multiscale seismic waveform inversion[J]. Geophysics, 1995, 60(5):1457-1473. doi: 10.1190/1.1443880

[1]	陈子龙, 王海燕, 郭华, 王光文, 赵玉莲. 地震全波形反演研究进展与应用现状综述[J]. 物探与化探, 2023, 47(3): 628-637.
[2]	姚含, 徐海. 基于梯度投影法的全变差正则化全波形反演[J]. 物探与化探, 2022, 46(4): 977-981.
[3]	国运东. 基于组合震源编码的多尺度全波形反演方法[J]. 物探与化探, 2022, 46(3): 729-736.
[4]	周斯琛, 李振春, 张敏, 张凯. 基于截断牛顿法的频率域全波形反演方法[J]. 物探与化探, 2017, 41(1): 147-152.
[5]	芮拥军. 一种常规检波器的低频恢复方法[J]. 物探与化探, 2016, 40(6): 1198-1202.
[6]	苗永康. 叠前地震全波形反演实际应用的关键影响因素[J]. 物探与化探, 2016, 40(5): 947-954.
[7]	赵永辉, 吴健生, 万明浩. 多次叠加技术在探地雷达资料处理中的应用[J]. 物探与化探, 2000, 24(3): 215-218.

Viewed

Full text

Abstract

Cited

Shared

Discussed