A seismic data denoising method based on complete ensemble empirical mode decomposition with adaptive noise and Hurst exponent
-
摘要: 在地震观测中，地震数据中普遍包含有噪声信号。由于噪声信号的干扰，地震分析的效率会受到不同程度影响。传统的去噪方法通常需要噪声的先验知识，并且滤波时会造成部分有效数据丢失。针对这一问题，本文提出一种将自适应噪声完全集合经验模态分解（CEEMDAN）算法与Hurst指数相结合的地震数据去噪方法。首先通过CEEMDAN方法将信号分解为一系列本征模函数（IMF），然后利用Hurst指数对滤波后的IMF分量进行识别，最后对地震数据IMF分量进行重构，从而实现数据去噪。与传统方法的去噪效果对比表明，本文方法对低信噪比波形的滤波性提高32%，对高信噪比波形的滤波性提高6倍。同时对地磁数据的去噪结果表明，本文方法能够较完整地将地铁噪声从地磁信号波形中滤除。
-
关键词:
- 地震数据去噪 /
- 地磁数据去噪 /
- 自适应噪声完全集合经验模态分解 /
- Hurst指数
Abstract: In seismic observation, seismic data generally contain background noise, which reduces the efficiency of seismic analysis. Traditional denoising methods usually need a priori knowledge of noise, and some effective data will be lost when filtering. To solve this problem, this paper proposes a seismic data denoising method based on complete ensemble empirical mode decomposition with adaptive noise （CEEMDAN） and Hurst exponent. Firstly, the magnetic detection signal is decomposed into a series intrinsic mode functions （IMF） by CEEMDAN method. Secondly, the Hurst exponent is used to identify the filtered IMF component. Finally, the IMF component of seismic data is reconstructed to realize data denoising. Compared with the denoising effect of traditional methods, the filtering ability of this method for low SNR waveform is improved by 32%, and the filtering ability for high SNR waveform is 6 times higher. At the same time, on the surface of geomagnetic data denoising results, this method can completely filter subway noise from geomagnetic signal waveform.-
Key words:
- Seismic data denoising /
- Geomagnetic data denoising /
- CEEMDAN /
- Hurst exponent
-
图 3 CEEMDAN算法分解结果
（a）百色地震事件波形；（b）灵川地震事件波形；（c）南宁爆破事件波形；（d）桂林坍塌事件波形
Figure 3. Decomposition results of CEEMDAN
（a） Baise earthquake waveform；（b） Lingchuan earthquake waveform； （c） Nanning blasting waveform；（d） Guilin collapse waveform
图 4 使用本文方法所得四个事件的去噪结果
左侧为原始波形，右侧为去噪后波形（a）百色地震事件波形；（b）灵川地震事件波形；（c）南宁爆破事件波形；（d）桂林坍塌事件波形
Figure 4. Denoising results of the four events by the method proposed in this paper
The left column stands for original waveforms and the right column stands for denoising waveforms （a） Baise earthquake waveform；（b） Lingchuan earthquake waveform；（c） Nanning blasting waveform；（d） Guilin collapse waveform
图 5 不同方法对灵川地震事件（a）和桂林坍塌事件（b）的去噪结果
Figure 5. Denoising results of Lingchuan earthquake （a） and Guilin collapse （b） by different methods
图 6 邕宁地磁台2021年1月1日00时至5日24时观测数据波形
Figure 6. Observation waveform at Yongning geomagnetic station from 00:00 on January 1 to 24:00 on January 5，2021
表 1 本文所选事件的信息
Table 1. The information of events selected in this paper
事件序号 事件类型 发震日期 震中位置 记录台站 震中距/km ML a 天然地震 2021-04-28 广西百色 FUN 73 1.3 b 天然地震 2017-11-25 广西灵川 YOF 148 1.9 c 人工爆破 2021-04-04 广西南宁 DAQ 88 2.0 d 山体坍塌 2020-06-23 广西桂林 GUL 36 2.3 
下载: 导出CSV
表 2 不同事件IMF分量的Hurst指数
Table 2. The Hurst exponents of different IMF components
分量序号 2021年4月28日广西
百色地震事件2017年11月25日广西
灵川地震事件2021年4月4日广西南宁
人工爆破事件2020年6月23日广西桂林
人工山体坍塌事件IMF1 0.302 5 0.294 6 0.258 9 0.359 4 IMF2 0.322 6 0.407 5 0.425 5 0.325 4 IMF3 0.307 0 0.287 5 0.356 1 0.466 3 IMF4 0.418 9 0.333 7 0.306 3 0.487 1 IMF5 0.483 8 0.384 6 0.515 9 0.361 1 IMF6 0.655 8 0.640 8 0.612 2 0.647 8 IMF7 0.686 0 0.686 2 0.658 9 0.683 1 IMF8 0.763 8 0.792 9 0.796 6 0.653 0 IMF9 0.806 2 0.877 7 0.870 9 0.864 5 IMF10 0.969 3 0.916 5 0.992 9 0.908 0 IMF11 0.973 3 0.966 8 1.007 0 0.955 1 IMF12 0.985 1 1.009 2 1.004 0 IMF13 0.994 9
下载: 导出CSV
表 3 不同类型去噪方法的质量因子Q值和均方根误差RMSE
Table 3. Quality factor Q and root-mean-square error RMSE of different denoising methods
去噪方法 灵川地震事件 桂林坍塌事件 Q值 RMSE Q值 RMSE 带通滤波器 0.723 9 0.380 1 1.360 1 0.076 1 小波包变换 0.718 5 0.375 6 1.193 9 0.073 3 EMD方法 0.854 9 0.374 7 7.344 2 0.040 0 本文方法 0.955 9 0.373 9 9.606 9 0.025 7
下载: 导出CSV
表 4 地磁数据IMF分量的Hurst指数
Table 4. The Hurst exponent of different IMFs from geomagnetic data
序号 Hurst指数 序号 Hurst指数 IMF1 0.284 6 IMF7 0.702 9 IMF2 0.323 2 IMF8 0.861 3 IMF3 0.208 0 IMF9 0.954 7 IMF4 0.352 4 IMF10 0.989 9 IMF5 0.497 9 IMF11 0.997 1 IMF6 0.583 0 IMF12 1.006 3
下载: 导出CSV
-
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[J]. 地球物理学进展,30(6):2433–2439. doi: 10.6038/pg20150601 Cai J H,Xiao X. 2015. Method of processing magnetotelluric signal based on the adaptive threshold wavelet[J]. Progress in Geophysics,30(6):2433–2439 (in Chinese). 陈学华,贺振华,黄德济. 2008. 广义S变换及其时频滤波[J]. 信号处理,24(1):28–31. doi: 10.3969/j.issn.1003-0530.2008.01.007 Chen X H,He Z H,Huang D J. 2008. Generalized S transform and its time-frequency filtering[J]. Signal Processing,24(1):28–31 (in Chinese). 韩卫雪,周亚同,池越. 2018. 基于深度学习卷积神经网络的地震数据随机噪声去除[J]. 石油物探,57(6):862–869. doi: 10.3969/j.issn.1000-1441.2018.06.008 Han W X,Zhou Y T,Chi Y. 2018. Deep learning convolutional neural networks for random noise attenuation in seismic data[J]. Geophysical Prospecting for Petroleum,57(6):862–869 (in Chinese). 牛永效. 2017. 基于频率及f-k域联合应用的地震波去噪技术研究[J]. 铁道标准设计,61(2):47–49. Niu Y X. 2017. Research on seismic data noise reduction based on combined application of frequency and F-K domain filter[J]. Railway Standard Design,61(2):47–49 (in Chinese). 孙月. 2012. 基于广义S变换和阈值函数的地震信号去噪研究[D]. 长春: 吉林大学, 19–26. Sun Y. 2012. Seismic Signal Denoising Research Based on Generalized S Transformation and Threshold Function[D]. Changchun: Jilin University, 19–26 （in Chinese）. 万光南. 2014. f-k滤波在压制面波噪声中的应用[J]. 中州煤炭,(2):99–101. Wan G N. 2014. Application of f-k filtering in noise suppression of surface wave[J]. Zhongzhou Coal,(2):99–101 (in Chinese). 杨凯,刘伟. 2012. 基于改进EMD的地震信号去噪[J]. 西南石油大学学报(自然科学版),34(4):75–82. Yang K,Liu W. 2012. Random noise attenuation of seismic signal based on improved EMD[J]. Journal of Southwest Petroleum University (Science &Technology Edition) ,34(4):75–82 (in Chinese). 张杏莉,卢新明,贾瑞生,阚淑婷. 2018. 基于变分模态分解及能量熵的微震信号降噪方法[J]. 煤炭学报,43(2):356–363. doi: 10.13225/j.cnki.jccs.2017.4153 Zhang X L,Lu X M,Jia R S,Kan S T. 2018. Micro-seismic signal denoising method based on variational mode decomposition and energy entropy[J]. Journal of China Coal Society,43(2):356–363 (in Chinese). Chen K,Sacchi M D. 2015. Robust reduced-rank filtering for erratic seismic noise attenuation[J]. Geophysics,80(1):V1–V11. Huang N E,Shen Z,Long S R,Wu M C,Shih H H,Zheng Q A,Yen N C,Tung C C,Liu H H. 1998. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J]. Proc Roy Soc A Math Phys Eng Sci,454(1971):903–995. doi: 10.1098/rspa.1998.0193 Hurst H E. 1951. Long-term storage capacity of reservoirs[J]. Trans Am Soc Civ Eng,116(1):770–799. doi: 10.1061/TACEAT.0006518 Liu W,Cao S Y,Chen Y K. 2016. Seismic time–frequency analysis via empirical wavelet transform[J]. IEEE Geosci Remote Sens Lett,13(1):28–32. doi: 10.1109/LGRS.2015.2493198 Lu Y,Huang Y M,Xue W,Zhang G B. 2019. Seismic data processing method based on wavelet transform for denoising[J]. Cluster Comput,22(3):6609–6620. Torres M E, Colominas M A, Schlotthauer G, Flandrin P. 2011. A complete ensemble empirical mode decomposition with adaptive noise[C]//2011 IEEE International Conference on Acoustics, Speech and Signal Processing （ICASSP）. Prague: IEEE: 4144–4147. Wang Y Q, Peng Z M, He Y M. 2016. Time-frequency representation for seismic data using sparse S transform[C]//2016 2nd IEEE International Conference on Computer and Communications （ICCC）. Chengdu: IEEE: 1923-1926. Zhang C,Li Y,Lin H B,Yang B J. 2015. Signal preserving and seismic random noise attenuation by Hurst exponent based time-frequency peak filtering[J]. Geophys J Int,203(2):901–909. doi: 10.1093/gji/ggv340 -
查看大图
计量
- 文章访问数: 33
- HTML全文浏览量: 16
- PDF下载量: 2
- 被引次数: 0