摘要:
庐枞矿集区科学钻探钻遇岩性复杂,岩性亚种类别繁多,利用测井资料识别岩性存在较大的技术难度. 笔者采用网格搜索法、粒子群优化和遗传算法三种方法优选支持向量机的核函数参数γ和惩罚因子C,其中基于遗传算法优选的支持向量机参数准确率最高. 利用测井,结合岩芯、录井等资料,基于遗传算法建立支持向量机岩性自动识别模型,该模型实际数据预测总体符合率为86.86%,优于BP神经网络,全井岩性识别与岩芯录井相符,取得了好的应用效果.
Abstract:
On the basis of SinoProbe,scientific drilling ZK01 of the Luzong ore district located in the Yangtze River basin of eastern China is an integrated geophysical logging study.This study aims at establishing the physical property of lithologic section and revealing the vertical distribution law of metals in the lower crust.For the purpose of detecting the lithologic distribution of the Luzong ore district and providing the information concerning the distribution of metallic ores and evaluation of reserves,the authors chose Support Vector Machine (SVM) to established automatic lithologic identification model for all of the wells.Three methods,i.e.,Grid Search (GS),Par-ticle Swarm Optimization (PSO) and Genetic Algorithm (GA),were applied to find the best parameters C and γ.GA was the best method becasuse it took 34 seconds to obtain the best parameters as (151.9852,9.1105),and its accuracy was up to 98.6364%.Com-pared with BP neural network identification results,the GA-SVM model achieved better accuracy of 86.86%.The lithologic identification and automatic zonation results are similar to the core data and artificial lithologic section,and the rationality and feasibility of GA-SVM are verified.
下载: