Research on noise suppression of magnetotelluric signal based on recurrent neural network
-
摘要:
由于天然电磁场源信号微弱，观测数据极易受到噪声干扰，严重影响反演和解释结果.传统去噪方法依赖于人工对时间序列和功率谱的筛选，去噪效率低，主观性强.本文提出利用循环神经网络对大地电磁时域信号进行特征噪声的识别和提取，进而重构出去噪后的大地电磁信号.在对大地电磁时域信号进行大量分析的基础上，对噪声进行分类并搭建含噪信号数据库，利用该数据库训练了两个循环神经网络，并选取长短时记忆单元优化循环神经网络结构，分别实现含噪数据段筛选和噪声形态提取.对仿真和实测数据分别进行了测试，循环神经网络均能准确筛选出大地电磁信号中的噪声段，本方法在避免人为操作主观性的同时提高了工作效率，视电阻率和相位曲线质量得到明显改善.
Abstract:As the natural electromagnetic signal is weak, the observation data is highly susceptible to noise interference, which seriously affects the inversion and interpretation results. Traditional denoising methods rely on manual screening of time series and power spectra, resulting in low denoising efficiency and strong subjectivity. In this paper, Recurrent Neural Network (RNN) is used to identify and extract the characteristic noise of magnetotelluric time domain signal, and then reconstruct the magnetotelluric signal. On the basis of a lot of analysis of magnetotelluric time-domain signals, noise is classified and a noisy signal database is built. We trained two neural networks using this database, and choose long-short term memory units to optimize the networks, respectively to screen noisy data segments and extract noise patterns. The simulated and measured data are tested respectively, and the RNN can accurately screen out the noise segment in the magnetotelluric signal. This method avoids the subjectivity of manual operation and improves the work efficiency, and the quality of apparent resistivity and phase curve is significantly improved.
-
Key words:
- Magnetotelluric /
- Recurrent Neural Network (RNN) /
- Strong interference /
- Denoising /
- Machine learning
-
-
图 6
不同幅值的方波去噪效果
Figure 6.
Noise removal effect of square wave with different amplitude
图 9
不同幅值的工频噪声去噪效果
Figure 9.
Noise removal effect of power frequency noise with different amplitude
图 10
工频干扰去噪效果对比图
Figure 10.
Comparison of power frequency interference denoising effect
图 14
神经网络处理前后的时间序列段对比
Figure 14.
Comparison of time series before and after neural network processing
图 15
循环神经网络去噪前后视电阻率和相位曲线对比图
Figure 15.
Comparison diagram of apparent resistivity and phase curve before and after noise removal of cyclic neural network
图 16
去噪前后二维反演结果对比
Figure 16.
Comparison of two-dimensional NLCG inversion results before and afterdenoising
表 1
仿真噪声参数
Table 1.
Simulated noise parameters
噪声类型 幅值 宽度 相位 频率/Hz 方波 1~5倍的纯净信号数据段的峰峰值 随机 无 无 三角波 1~5倍的纯净信号数据段的峰峰值 随机 无 无 工频 1~5倍的纯净信号数据段的峰峰值 覆盖整个数据段 -π~π 49~51 阶跃波 1~5倍的纯净信号数据段的峰峰值 随机 无 无 尖脉冲 6~10倍的纯净信号数据段的峰峰值 单点 无 无 
下载: 导出CSV
表 2
基于含噪数据段筛选的网络模型参数
Table 2.
Network model parameters based on noisy data segment filtering
参数名称 参数值 输入层维度 1 隐藏层数 2 隐藏层节点数 128 隐藏层激活函数 Tanh 优化器 Adam 损失函数 Binary_crossentropy 全连接层输出维度 1 全连接层激活函数 Sigmoid
下载: 导出CSV
-
An Z G, Di Q Y, Qu W Z, et al. 2022a. Shallow crustal electrical structure of the Qingchengzi orefield in Liaodong area revealed by three-dimensional inversion of Magnetotelluric data. Journal of Applied Geophysics, 202: 104650, doi: 10.1016/j.jappgeo.2022.104650.
An Z G, Di Q Y, Wang Z X, et al. 2020. Deep geological structures associated with terrestrial volcanic hydrothermal metallogenic system: Evidence from geophysical survey in Taohemu superlarge silver-polymetallic deposit, Inner Mongolia. Earth and Space Science, 7(10): e2019EA000939, doi: 10.1029/2019ea000939.
An Z G, Zhang Y, Dong Y H, et al. 2022b. Modeling the crustal electrical structure of the Zhangye Basin via three-dimensional inversion of magnetotelluric data-Implications for future geothermal development. Tectonophysics, 842: 229590, doi: 10.1016/j.tecto.2022.229590.
Bengio Y, Simard P, Frasconi P. 1994. Learning long-term dependencies with gradient descent is difficult. IEEE Transactions on Neural Networks, 5(2): 157-166, doi: 10.1109/72.279181.
Cai J H, Xiao X. 2015. Method of processing magnetotelluric signal based on the adaptive threshold wavelet. Progress in Geophysics (in Chinese), 30(6): 2433-2439, doi: 10.6038/pg20150601.
Cao X L, Yan L J, Jiang T. 2018. Application of blind source separation algorithm based on DWT-EEMD in removal of power line interference for MT. Coal Geology & Exploration (in Chinese), 46(2): 164-172. doi: 10.3969/j.issn.1001-1986.2018.02.025
Chouliaras G, Rasmussen T M. 1988. The application of the magnetotelluric impedance tensor to earthquake prediction research in Greece. Tectonophysics, 152(1-2): 119-135. doi: 10.1016/0040-1951(88)90032-7
Deng Y, Tang J. 2019. Advances in magnetotelluric data processing. Progress in Geophysics (in Chinese), 34(4): 1411-1422, doi: 10.6038/pg2019CC0188.
Egbert G D, Booker J R. 1986. Robust estimation of geomagnetic transfer functions. Geophys. J. Roy. Astr. Soc. , 87(1): 173-194. doi: 10.1111/j.1365-246X.1986.tb04552.x
Fan T, Xue G Q, Li P, et al. 2022. TEM real-time inversion based on long-short term memory network. Chinese Journal of Geophysics (in Chinese), 65(9): 3650-3663, doi: 10.6038/cjg2022P0572.
Gamble T D, Goubau W M, Clarke J. 1979. Magnetotellurics with a remote magnetic reference. Geophysics, 44(1): 53-68. doi: 10.1190/1.1440923
Goubau W M, Gamble T D, Clarke J. 1978. Magnetotelluric data analysis: removal of Bias. Geophysics, 43(6): 1157-1166. doi: 10.1190/1.1440885
Graves A, Mohamed A R, Hinton G. 2013. Speech recognition with deep recurrent neural networks. arXiv: 1303.5778.
Graves A, Schmidhuber J. 2005. Framewise phoneme classification with bidirectional LSTM networks. //2005 IEEE International Joint Conference on Neural Networks. Montreal, QC, Canada: IEEE, 2047-2052.
Han Y, An Z G, Qu W Z. 2021. Research status of magnetotelluric time domain data processing based on machine learning. Progress in Geophysics (in Chinese), 36(5): 1975-1987, doi: 10.6038/pg2021EE0581.
He L F, Di Q Y, Wang Z X, et al. 2023. Crustal structures of the Qimantagh Metallogenic Belt in the northern Tibetan Plateau from magnetotelluric data and their correlation to the distribution of mineral deposits. Minerals, 13(2): 225, doi: 10.3390/min13020225.
Hermance J F. 1973. Processing of magnetotelluric data. Physics of the Earth and Planetary Interiors, 7(3): 349-364. doi: 10.1016/0031-9201(73)90060-5
Hochreiter S, Schmidhuber J. 1997. Long short-term memory. Neural Computation, 9(8): 1735-1780. doi: 10.1162/neco.1997.9.8.1735
Hu Q X, Xu Y. 2022. Review of machine learning and application of geophysical signal feature recognition and interpretation. Progress in Geophysics (in Chinese), 37(6): 2395-2407, doi: 10.6038/pg2022FF0568.
Hua S H, Han L G. 2023. Seismic data denoising method based on deep convolutional auto-encoder network. Progress in Geophysics (in Chinese), 38(2): 654-661, doi: 10.6038/pg2023GG0290.
Huang N E, Shen Z, Long S R, et al. 1998. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. Proceedings of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences, 454(1971): 903-995. doi: 10.1098/rspa.1998.0193
Huang N E, Wu M L C, Long S R, et al. 2003. A confidence limit for the empirical mode decomposition and Hilbert spectral analysis. Proceedings of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences, 459(2037): 2317-2345. doi: 10.1098/rspa.2003.1123
Liu J M, Zhang R, Zhang Q Z. 2004. The regional metallogeny of Da Hinggan Ling, China. Earth Science Frontiers (in Chinese), 11(1): 269-277. doi: 10.3321/j.issn:1005-2321.2004.01.024
Liu L J, Hasterok D. 2016. High-resolution lithosphere viscosity and dynamics revealed by magnetotelluric imaging. Science, 353(6307): 1515-1519. doi: 10.1126/science.aaf6542
Mallat S G. 1989. A theory for multiresolution signal decomposition: the wavelet representation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 11(7): 674-693. doi: 10.1109/34.192463
Manoj C, Nagarajan N. 2003. The application of artificial neural networks to magnetotelluric time-series analysis. Geophysical Journal International, 153(2): 409-423. doi: 10.1046/j.1365-246X.2003.01902.x
Puzyrev V. 2019. Deep learning electromagnetic inversion with convolutional neural networks. Geophysical Journal International, 218(2): 817-832. doi: 10.1093/gji/ggz204
Qiu Y. 2018. Research on fetal electrocardiogram extraction based on recurrent neural network[Ph. D. thesis](in Chinese). Hangzhou: Zhejiang University.
Shen H Y, George D, Huerta E A, et al. 2019. Denoising gravitational waves with enhanced deep recurrent denoising auto-encoders. arXiv: 1903.03105.
Tang J, Wang JJ, Chen X B, et al. 2005. Preliminary investigation for electric conductivity structure of the crust and upper mantle beneath the Aershan volcano area. Chinese Journal of Geophysics (in Chinese), 48(1): 196-202. doi: 10.3321/j.issn:0001-5733.2005.01.026
Tang J T, Hua X R, Cao Z M, et al. 2008. Hilbert-Huang transformation and noise suppression of magnetotelluric sounding data. Chinese Journal of Geophysics (in Chinese), 51(2): 603-610. doi: 10.3321/j.issn:0001-5733.2008.02.034
Vozoff K. 1972. The magnetotelluric method in the exploration of sedimentary basins. Geophysics, 37(1): 98-141. doi: 10.1190/1.1440255
Wang H, Campanyà J, Cheng J L, et al. 2017. Synthesis of natural electric and magnetic Time-series using Inter-station transfer functions and time-series from a Neighboring site (STIN): Applications for processing MT data. Journal of Geophysical Research: Solid Earth, 122(8): 5835-5851. doi: 10.1002/2017JB014190
Wang H, Cheng J L, Teng X Z, et al. 2016. Source effect on magnetotelluric data due to mining area and its suppression. Progress in Geophysics (in Chinese), 31(3): 1358-1366, doi: 10.6038/pg20160359.
Wang H, Jiang H, Wang L, et al. 2015. Magnetotelluric inversion using artificial neural network. Journal of Central South University (Science and Technology) (in Chinese), 46(5): 1707-1714.
Wang J Y. 1997. New development of magnetotelluric sounding in China. Acta Geophysica Sinica (in Chinese), 40(S1): 206-216.
Wang J Y, Wang Z G, Chen Y M, et al. 2023. Deep artificial neural network in seismic inversion. Progress in Geophysics (in Chinese), 38(1): 298-320, doi: 10.6038/pg2023FF0467.
Williams R J, Zipser D. 1989. A learning algorithm for continually running fully recurrent neural networks. Neural Computation, 1(2): 270-280. doi: 10.1162/neco.1989.1.2.270
Wu X, Xue G Q, Xiao P, et al. 2019. The removal of the high-frequency motion-induced noise in helicopter-borne transient electromagnetic data based on wavelet neural network. Geophysics, 84(1): K1-K9. doi: 10.1190/geo2018-0120.1
Wu X, Xue G Q, Zhao Y, et al. 2022. A deep learning estimation of the Earth resistivity model for the airborne transient electromagnetic observation. Journal of Geophysical Research: Solid Earth, 127(3): e2021JB023185, doi: 10.1029/2021JB023185.
Yi J D, Zhang M, Li Z C, et al. 2023. Review of deep learning seismic data reconstruction methods. Progress in Geophysics (in Chinese), 38(1): 361-381, doi: 10.6038/pg2023GG0048.
Zhang Y, An Z G, Dong Y H. 2021. Study of the crust electrical structure of Zhangye Basin. Progress in Geophysics (in Chinese), 36(4): 1477-1486, doi: 10.6038/pg2021EE0351.
Zhu F, Cheng Q, Li S S, et al. 2022. Research on noise suppression of seismic migration based on deep learning. Progress in Geophysics (in Chinese), 37(2): 746-755, doi: 10.6038/pg2022FF0121.
蔡剑华, 肖晓. 2015. 基于小波自适应阈值去噪的MT信号处理方法. 地球物理学进展, 30(6): 2433-2439, doi: 10.6038/pg20150601.
曹小玲, 严良俊, 蒋涛. 2018. 基于DWT-EEMD的盲源分离算法在MT工频干扰消除中的应用. 煤田地质与勘探, 46(2): 164-172. doi: 10.3969/j.issn.1001-1986.2018.02.025
邓琰, 汤吉. 2019. 大地电磁测深方法数据处理进展. 地球物理学进展, 34(4): 1411-1422, doi: 10.6038/pg2019CC0188.
底青云, 薛国强, 王中兴等. 2021. 利用自制SEP电磁观测系统研究祁幔塔格山和西柴达木盆地岩石圈结构. 中国科学: 地球科学, 51(12): 2197-2204. https://www.cnki.com.cn/Article/CJFDTOTAL-JDXK202112014.htm
范涛, 薛国强, 李萍等. 2022. 瞬变电磁长短时记忆网络深度学习实时反演方法. 地球物理学报, 65(9): 3650-3663, doi: 10.6038/cjg2022P0572.
韩盈, 安志国, 屈文璋. 2021. 基于机器学习的大地电磁时域数据处理的研究现状. 地球物理学进展, 36(5): 1975-1987, doi: 10.6038/pg2021EE0581.
胡广书. 1997. 数字信号处理——理论、算法与实现. 北京: 清华大学出版社.
胡琪鑫, 徐亚. 2022. 地球物理信号特征识别与解释的机器学习方法及应用综述. 地球物理学进展, 37(6): 2395-2407, doi: 10.6038/pg2022FF0568.
滑世辉, 韩立国. 2023. 基于深度卷积自编码网络地震数据去噪方法. 地球物理学进展, 38(2): 654-661, doi: 10.6038/pg2023GG0290.
刘建明, 张锐, 张庆洲. 2004. 大兴安岭地区的区域成矿特征. 地学前缘, 11(1): 269-277. doi: 10.3321/j.issn:1005-2321.2004.01.024
内蒙古国土资源勘查开发院. 2016. 内蒙古自治区科尔沁右翼前旗巴尔陶勒盖-复兴屯银铅锌多金属矿普查报告. 全国地质资料馆.
仇悦. 2018. 基于循环神经网络的胎儿心电提取方法研究[博士论文]. 杭州: 浙江大学.
汤吉, 王继军, 陈小斌等. 2005. 阿尔山火山区地壳上地幔电性结构初探. 地球物理学报, 48(1): 196-202. doi: 10.3321/j.issn:0001-5733.2005.01.026
汤井田, 化希瑞, 曹哲民等. 2008. Hilbert-Huang变换与大地电磁噪声压制. 地球物理学报, 51(2): 603-610. doi: 10.3321/j.issn:0001-5733.2008.02.034
王鹤, 蒋欢, 王亮等. 2015. 大地电磁人工神经网络反演. 中南大学学报(自然科学版), 46(5): 1707-1714. https://www.cnki.com.cn/Article/CJFDTOTAL-ZNGD201505019.htm
王辉, 程久龙, 腾星智等. 2016. 矿区近场源噪声对大地电磁测深数据的影响及其压制方法. 地球物理学进展, 31(3): 1358-1366, doi: 10.6038/pg20160359.
王家映. 1997. 我国大地电磁测深研究新进展. 地球物理学报, 40(S1): 206-216. http://www.geophy.cn/article/id/cjg_7448
王竟仪, 王治国, 陈宇民等. 2023. 深度人工神经网络在地震反演中的应用进展. 地球物理学进展, 38(1): 298-320, doi: 10.6038/pg2023FF0467.
易继东, 张敏, 李振春等. 2023. 深度学习地震数据重建方法研究综述. 地球物理学进展, 38(1): 361-381, doi: 10.6038/pg2023GG0048.
张优, 安志国, 董艳辉. 2021. 张掖盆地壳内电性结构研究. 地球物理学进展, 36(4): 1477-1486, doi: 10.6038/pg2021EE0351.
诸峰, 程前, 李帅帅等. 2022. 基于深度学习的地震偏移噪声压制研究. 地球物理学进展, 37(2): 746-755, doi: 10.6038/pg2022FF0121.
-
图(16)
表(2)
计量
- 文章访问数: 1551
- PDF下载数: 186
- 施引文献: 0