【摘要】 记录不同井深岩屑元素含量信息的岩屑录井工作是油气勘探开发的必要过程之一。本研究将西南某钻井现场的岩屑样品分为训练集和测试集,使用自行研制的小型台式激光诱导击穿光谱仪(LIBS-Trace)采集各岩屑样品谱图,对8种元素(Si、Al、Ca、K、Mg、Mn、Ti、Fe)优化建立基于支持向量机回归(SVR)的元素定量分析模型。将测试集数据用于算法模型评估验证,结果表明,8种元素预测的平均百分比误差(MPE)分别为Si 5.68%、Al 7.22%、Ca 7.45%、K 9.76%、Mg 8.79%、Mn 11.9%、Ti 11.4%和Fe 10.4%,可以满足录井工作的要求。此外,依据录井工作的流程与需求,对仪器软硬件进行修改,将优化确立的定量模型集成于仪器之中,使之成为录井专用型号。在录井工作现场进行测试,结果表明,算法模型结合仪器可快速完成岩屑元素定量分析,在油气勘探开发领域具有潜在的应用价值。更多还原

【Abstract】 It is one of the necessary work in oil and gas exploration and development to record the logging of different depth cuttings. In this study,a laser induced breakdown spectrometer( LIBS-Trace) developed by our research group was used for the lithology logging in the field of oil and gas exploration and development. A total of 60 samples from a drilling site in the Southwest China were divided into training set and test set,and then element quantitative analysis model of eight elements of Si,Al,Ca,K,Mg,Mn,Ti and Fe was established based on support vector machine regression( SVR). The results showed that the mean percentage prediction errors( MPE) predicted by this method were Si 5.68%,Al 7.22%,Ca 7.45%,K 9.76%,Mg 8.79%,Mn 11. 9%,Ti 11. 4% and Fe 10. 4%,which met the requirements of logging work. In addition,according to the logging process and demand,the instrument hardware and software were modified,and the quantitative model integrated in the instrument was optimized. The results showed that the algorithm model combined with the instrument could quickly complete the quantitative analysis of rock debris samples,and exhibited potential application value and broad application prospect in oil and gas exploration and development.更多还原