Support Vector Machine In Logging Curve Hierarchical Application

Lithology identification, log facies analysis, reservoir classification, reservoirparameters and so on, first of all need to layering logging curve,to facilitatesubsequent studies with different characteristics determine research objectives andkey strata, unification of different borehole research scope.Layering logging curveuses artificial methods of interpretation, not only time-consuming and laborious, butalso by many subjective factors and environmental factors. Computer automaticstratification can avoid artificial stratification of the subjectivity and randomness, andlargely improve work efficiency.This paper first introduces the related theory of support vector machine, supportvector machine is based on statistical machine learning method. Support vectormachine not only has a solid theoretical foundation, intuitive geometricalinterpretation, simple mathematical expressions, in the small sample, nonlinear, highdimension pattern recognition also showed a higher advantages, and in many areasexhibit good generalization performance, artificial intelligence and machine learninghas become a new hotspot in the field of. This paper also introduced the supportvector machine method, compares and analyses the advantages and disadvantages ofthe methods and properties of different. Support vector machine SVM trainingprocess is a solution of two problems, the number of samples will influence the twoplanning complexity。Then introduced the related theory and the nature of the kernel function, and thecommonly used kernel function construction method, analyzes and discusses thekernel function and parameters selection of support vector machine and its advantagesand disadvantages of the effect. By1wells using different kernel functions andparameters as well as punishment scope of hierarchical experiment, obtained fordifferent kernel functions and parameters have different influence on the well logging curves stratified, with the same punishment scope, different kernel functionscorresponding to the support vector machine performance in time are very different,illustrates the selection of kernel function for support vector machine performancethere is a certain effect.Finally, in this paper the application of support vector machine to a developedarea of1～13wells experiment, first of all on the No.1well establishedclassification model, proposed in view of geological structure order continuous thischaracteristic, the application of support vector machine for layering logging curve,according to the needs of the sequentially ordered learning multi class classificationsupport vector machine model, the establishment of a number of adjacent twoconsecutive classification model of support vector machine to solve the logging curvestratification problems, and with artificial stratification method carries on the contrastanalysis, can see the application of support vector machine can get good stratificationresults, basic and artificial layered anastomosis, greatly improving the work efficiencyof layering logging curve. Then the2-7well automatic layered, stratified resultsobtained with artificial stratification results were compared and analyzed by usingsupport vector machine to logging curve layering of the advantages and disadvantagesand improvement space, finally the area from8to13wells were automaticallystratified, obtained certain achievements.