基于Tikhonov正则化的地磁反演建模与实验分析
Tikhonov Regularization for Geomagnetic Inversion: Modeling and Experimental Analysis
DOI: 10.12677/ag.2026.1611139, PDF,    科研立项经费支持
作者: 李 博:河北红山巨厚沉积与地震灾害国家野外科学观测研究站,河北 邢台;保定地震监测中心站,河北 保定
关键词: Tikhonov正则化;地磁反演;深度加权;L曲线;广义交叉验证;Tikhonov Regularization; Geomagnetic Inversion; Depth Weighting; L-Curve; Generalized Cross-Validation
摘要: 在测点数远少于待求物性单元数的极端稀疏条件下,地表总场磁异常反演地下磁化率分布的性能边界与参数敏感性直接决定结果可信度。本文以棱柱体偶极核构建灵敏度矩阵,构建含数据残差项与深度加权正则项的Tikhonov目标泛函,基于广义奇异值分解稳定计算L曲线与广义交叉验证(GCV)函数,并以真实磁测数据与已知真模型的合成基准开展系统实验。结果表明:深度加权能以噪声量级的拟合代价消除系统性浅部聚集偏差,其有效标定参数β与z0使质心深度沿参数网格单调变化、存在质心无偏等值线,据此可先验校正标定过冲,参考深度zref则仅为会被正则化参数吸收的规范自由度;巨大零空间使GCV分母中的迹trace(I − AA+)趋于常数、GCV函数近乎平坦,极小点漂移至噪声内插或过度平滑区,而利用噪声水平先验的偏差原理选参保持稳定;模型相对误差在全参数网格内几乎不变,逐点重建不可达,深度加权改善的是深度定位的方向性偏差。据此为稀疏磁测反演提炼出规避浅部聚集、参数过冲与过拟合三类风险的实践准则。
Abstract: Under extremely sparse conditions where the number of survey points is far smaller than the number of model cells, the performance boundary and parameter sensitivity of recovering the subsurface susceptibility from surface total-field magnetic anomalies directly determine the credibility of the inversion. The sensitivity matrix is built with a prism/dipole kernel, a Tikhonov objective functional comprising a data-misfit term and a depth-weighted regularization term is formulated, and the L-curve and generalized cross-validation (GCV) functions are computed stably via the generalized singular value decomposition; systematic experiments are conducted on real magnetic survey data and on a synthetic benchmark with a known true model. The results show that depth weighting removes the systematic shallow-clustering bias at a data-misfit cost on the noise level, and that its effective calibration parameters β and z0 move the center-of-gravity depth monotonically across the parameter grid with an unbiased-centroid contour, so that calibration overshoot can be corrected a priori, while the reference depth zref is merely a gauge freedom absorbed by the regularization parameter. A huge null space renders the denominator trace of GCV nearly constant and the GCV function nearly flat, so that its minimum drifts into the noise-interpolation or over-smoothing region, whereas discrepancy-principle-based selection with a reliable noise-level prior remains stable. The model relative error is nearly invariant over the whole calibration grid, indicating that pointwise reconstruction is unattainable and that depth weighting improves only the directional bias of depth positioning. Practical guidelines are accordingly distilled for users of sparse magnetic inversion to avoid shallow clustering, calibration overshoot, and overfitting.
文章引用:李博. 基于Tikhonov正则化的地磁反演建模与实验分析[J]. 地球科学前沿, 2026, 16(11): 1575-1588. https://doi.org/10.12677/ag.2026.1611139

参考文献

[1] 管志宁. 地磁场与磁力勘探[M]. 北京: 地质出版社, 2005.
[2] Hansen, P.C. (1998) Rank-Deficient and Discrete Ill-Posed Problems. Society for Industrial and Applied Mathematics.
https://doi.org/10.1137/1.9780898719697
[3] Tikhonov, A.N. and Arsenin, V.Y. (1977) Solutions of Ill-Posed Problems. Winston & Sons.
[4] Zhdanov, M.S. (2002) Geophysical Inverse Theory and Regularization Problems. Elsevier.
[5] Tikhonov, A.N. (1963) On the Solution of Ill-Posed Problems and the Method of Regularization. Doklady Akademii Nauk SSSR, 151, 501-504.
[6] Parker, R.L. (1994) Geophysical Inverse Theory. Princeton University Press.
[7] Li, Y. and Oldenburg, D.W. (1996) 3-D Inversion of Magnetic Data. Geophysics, 61, 394-408.
https://doi.org/10.1190/1.1443968
[8] Li, Y. and Oldenburg, D.W. (1998) 3-D Inversion of Gravity Data. Geophysics, 63, 109-119.
https://doi.org/10.1190/1.1444302
[9] Tarantola, A. (2005) Inverse Problem Theory and Methods for Model Parameter Estimation. Society for Industrial and Applied Mathematics.
https://doi.org/10.1137/1.9780898717921
[10] 李泽林, 姚长利, 郑元满, 等. 数据空间磁异常模量三维反演[J]. 地球物理学报, 2015, 58(10): 3804-3814.
[11] 郭良辉, 孟小红, 石磊. 磁异常ΔT三维相关成像[J]. 地球物理学报, 2010, 53(2): 435-441.
[12] 孟小红, 刘国峰, 陈召曦, 等. 基于剩余异常相关成像的重磁物性反演方法[J]. 地球物理学报, 2012, 55(1): 304-309.
[13] 陈召曦, 孟小红, 郭良辉. 重磁数据三维物性反演方法进展[J]. 地球物理学进展, 2012, 27(2): 503-511.
[14] 冯杰, 刘天佑, 杨宇山, 等. 3D井地磁测联合反演技术及应用[J]. 地球物理学进展, 2010, 25(5): 1685-1691.
[15] 姚长利, 郝天珧, 管志宁, 等. 重磁遗传算法三维反演中高速计算及有效存储方法技术[J]. 地球物理学报, 2003, 46(2): 252-258.
[16] 李金朋, 张英堂, 范红波, 等. 基于χ2准则的磁梯度张量3D聚焦反演方法[J]. 武汉大学学报(信息科学版), 2018, 43(2): 255-261.
[17] 朱裕振, 强建科, 王林飞, 等. 深埋铁矿磁测数据三维反演分析与找矿靶区预测[J]. 物探与化探, 2019, 43(6): 1182-1190.
[18] 管志宁, 侯俊胜, 黄临平, 等. 重磁异常反演的拟BP神经网络方法及其应用[J]. 地球物理学报, 1998, 41(2): 242-251.
[19] 姚长利, 郝天珧, 管志宁. 重磁反演约束条件及三维物性反演技术策略[J]. 物探与化探, 2002, 26(4): 253-257.
[20] Fang, Y., Li, S., Liu, G., Meng, X., Wang, J. and Zheng, S. (2024) An Efficient Three-Dimensional (3D) Inversion Algorithm for Recovering Magnetic Targets When Remanence Exists. Acta Geophysica, 72, 4027-4043.
https://doi.org/10.1007/s11600-024-01292-9
[21] Liu, F. and Yan, B. (2024) Local Magnetic Field Inversion Based on the Modified L-Curve Model Function Method. IEEE Sensors Journal, 24, 29232-29242.
https://doi.org/10.1109/jsen.2024.3410988
[22] Gschwentner, A., Roppert, K. and Kaltenbacher, M. (2024) Determination of Local Magnetic Material Properties Using an Inverse Scheme. IEEE Transactions on Magnetics, 60, 1-4.
https://doi.org/10.1109/tmag.2023.3299620
[23] Blakely, R.J. (1995) Potential Theory in Gravity and Magnetic Applications. Cambridge University Press.
https://doi.org/10.1017/cbo9780511549816
[24] Oldenburg, D.W. and Li, Y. (2005) 5. Inversion for Applied Geophysics: A Tutorial. In: Butler, D.K., Ed., Near-Surface Geophysics, Society of Exploration Geophysicists, 89-150.
https://doi.org/10.1190/1.9781560801719.ch5
[25] Constable, S.C., Parker, R.L. and Constable, C.G. (1987) Occam’s Inversion; a Practical Algorithm for Generating Smooth Models from Electromagnetic Sounding Data. Geophysics, 52, 289-300.
https://doi.org/10.1190/1.1442303
[26] Hansen, P.C. (1992) Analysis of Discrete Ill-Posed Problems by Means of the L-Curve. SIAM Review, 34, 561-580.
https://doi.org/10.1137/1034115
[27] Golub, G.H., Heath, M. and Wahba, G. (1979) Generalized Cross-Validation as a Method for Choosing a Good Ridge Parameter. Technometrics, 21, 215-223.
https://doi.org/10.1080/00401706.1979.10489751
[28] Cella, F. and Fedi, M. (2012) Inversion of Potential Field Data Using the Structural Index as Weighting Function Rate Decay. Geophysical Prospecting, 60, 313-336.
https://doi.org/10.1111/j.1365-2478.2011.00974.x
[29] Portniaguine, O. and Zhdanov, M.S. (2002) 3-D Magnetic Inversion with Data Compression and Image Focusing. Geophysics, 67, 1532-1541.
https://doi.org/10.1190/1.1512749