基于支持向量机的电磁波传播测井反演参数约束条件提取方法

Constraint Extraction of Inversion Parameters Based on Support Machine Vector for Electromagnetic Propagation Logging

多参数阵列电磁传播电阻率测井(EPRL)资料的反演问题中,对每个待反演参数取值范围施加不同、恰当的限制条件,即所谓的约束条件,不仅可以有效地克服反演的多值性,而且可以提高反演的收敛速度。提出了一种基于支持向量机(SVM)的约束条件提取方法,首先,对实际问题可能出现地层进行分类,其次通过测井资料正演模拟得到训练样本,进而利用SVM进行地层类型识别,最后将识别的地层类型转化为反演参数的约束条件。将该方法应用于阵列电磁传播电阻率测井资料径向一维反演参数约束条件提取中,取得令人鼓舞的结果,由此可见这种方法是可行和有效的。

In inversion of multi-parameter array electromagnetic propagation resistivity logging （EPRL）, constraints, i.e. some conditions for properly restricting every inversion parameter within different ranges according to logging characters, could not only effectively overcome multiplicity, but also improve convergence speed of the inversion. Introduced is an extraction method of the constraints based on support vector machine （SVM）. Firstly, the possible strata are classi： fled; in sequence, the training samples are gotten by forward simulation of logging, and the types of strata are identified with SVM; lastly, the identified stratum types are transformed to the constraints of inversion parameter. In several stratum models of EPRL, the application of the method in the constraint extraction of inversion parameter of radial one-dimensional EPRL shows that the inversion convergence speed of the method is evidently faster than that of nonrestraint inversion for the stratum models whose type is able to be identified correctly.