Electrical resistivity imaging inversion based on support vector regression
DONG Li1,2, JIANG Fei-Bo2, 3, DAI Qian-Wei2, FU Yu-Hang4
1. Department of Information Science and Engineering, Hunan International Economics University, Changsha 410205, China; 2. School of Geosciences and Info-Physics, Central South University, Changsha 410083, China; 3. College of Physics and Information Science, Hunan Normal University, Changsha 410081, China; 4. Hunan Power Transmission and Distribution Survey and Design Consulting Co., Ltd., Changsha 410114, China
Abstract:Support Vector Regression is a Learning Machine based on statistic learning theory. It has better performance of generalization and fitting precision than traditional neural network inversion under the condition of small samples learning. Under the application background of electrical resistivity imaging, SVR inversion method based on limited learning samples was studied in this paper. The key issues of sample division and data preprocessing, inversion flow and evaluation indicators were analyzed. A multi-parameter optimization method based on cross validation was presented. The optimized SVR inversion model by comparing the influence of RBF kernel functions with different ε values with the inversion results was established. Data simulation and model inversion show that the proposed inversion method has better inversion accuracy and imaging quality than traditional least squares inversion and RBFNN inversion, and is equivalent to BPNN, but it has disadvantage of only one output dimension.
[1] Shima H.2-D and 3-D resistivity image reconstruction using crosshole data[J].Geophysics,1992,57(10):1270-1281. [2] Loke M H,Barker R D.Least-squares deconvolution of apparent resistivity pseudosections[J].Geophysics,1995,60(6):1682-1690. [3] Zohdy A A R.A new method for the automatic interpretation of Schlumbeger and Wenner sounding curves[J].Geophysics,1989,54(2):245-253. [4] Lesur V,Cuer M,Straub A.2-D and 3-D interpretation of electrical tomography measurements,Part2:The inverse problem[J].Geophysics,1999,64(2):396-402. [5] Singh U K,Tiwari R K,Singh S B.Neural network modeling and prediction of resistivity structures using VES Schlumberger data over a geothermal area[J].Computers & Geosciences,2013,52:246-257. [6] Singh U K,Tiwari R K,Singh S B.One-dimensional inversion of geo-electrical resistivity sounding data using artificial neural networks—a case study[J].Computers & geosciences,2005,31(1):99-108. [7] El-Qady G,Ushijima K.Inversion of DC resistivity data using neural networks[J].Geophysical prospecting,2001,49(4):417-430. [8] 徐海浪,吴小平.电阻率二维神经网络反演[J].地球物理学报,2006,49(2):584 589. [9] Neyamadpour A,Taib S,Abdullah W A T W.Using artificial neural networks to invert 2D DC resistivity imaging data for high resistivity contrast regions:a MATLAB application[J].Computers & Geosciences,2009,35(11):2268-2274. [10] Zhang L Y,Liu H F.The application of ABP method in high-density resistivity method inversion[J].Chinese Journal of Geophysics.(in Chinese),2011,54(1):22764-23371. [11] Ho T L.3-D inversion of borehole-to-surface electrical data using a back-propagation neural network[J].Journal of Applied Geophysics,2009,68(4):489-499. [12] Neyamadpour A,Wan Abdullah W A T W,Taib S,et al.3D inversion of DC data using artificial neural networks[J].Studia Geophysica et Geodaetica,2010,54(3):465-485. [13] 戴前伟,江沸菠,董莉.基于汉南-奎因信息准则的电阻率层析成像径向基神经网络反演[J].地球物理学报,2014,57(4):1335-1344. [14] 史峰,王小川,郁磊.Matlab神经网络案例分析[M].北京:北京航空航天大学出版社,2010. [15] Pidlisecky A,Knight R.FW2_5D:A MATLAB 2.5-D electrical resistivity modeling code[J].Computers & Geosciences,2008,34(12):1645-1654. [16] Wang H S,Wang Y N,Wang Y C.Cost estimation of plastic injection molding parts through integration of PSO and BP neural network[J].Expert Systems with Applications,2013,40(2):418-428. [17] Vapnik V.Vapnik.Statistical learning theory[ M].New York,NY:Wiley,New York,NY,1998. [18] 闫国华,朱永生.支持向量机回归的参数选择方法[J].计算机工程,2009,35(14):218-220. [19] 邴萍萍,曹思远,路交通.基于支持向量机的非线性AVO反演[J].地球物理学报, 2012,55(3):1025-1032. [20] 易世华,刘代志,何元磊,等.变化地磁场预测的支持向量机建模[J].地球物理学报,2013,56(1):127-135. [21] 江沸菠,戴前伟,董莉.基于主成分—正则化极限学习机的超高密度电法非线性反演[J].地球物理学报,2015,58(9):3356-3369.