基于激光诱导击穿光谱技术的录井岩屑多元素定量分析方法(特邀)

Quantitative Analysis Method of Multi-elements for Logging Cuttings Based on LIBS Technology(Invited)

利用自行研制的台式激光诱导击穿光谱仪,采集了岩石样品的光谱图,将数据划分为训练集和测试集,然后结合主成分回归算法对Si、 Al、 Ca、 Mg和K元素建立了定量分析模型,将测试集数据应用于模型评估,五种元素预测的决定系数(R^(2))分别为0.974 5、0.941 1、0.984 3、0.990 1、0.811 4。最后,将定量模型集成到仪器上,对97个来自我国西南某钻井现场的岩屑样品进行了分析,并与该批样品的实验室X射线荧光检测结果进行了对比。结果表明,五种元素的激光诱导击穿光谱预测值与X射线荧光检测值之间的决定系数(R^(2))分别达到0.971 9、0.960 4、0.947 9、0.965 2和0.924 5,整体趋势基本一致,说明集成了定量模型的仪器能够实现快速、有效的岩屑元素含量分析,在油气勘探开发领域具有良好的应用潜力。

It is one of the basic works to determine the elemental contents of cuttings’ samples from different depths in the process of oil and gas exploration and development. An accurate element logging map can provide significant information for identifying lithology, predicting the formation to be drilled, selecting suitable drilling parameters and reducing drilling risks. With the disadvantages of complex operation processes, complicated sample pretreatment and long analysis time, most traditional element analysis technologies are difficult to meet the requirements of rapid analysis for cuttings at thelogging site. In recent years, X-ray fluorescence spectrometry technology has been widely used in the logging field for real-time online element analysis due to its portability, low sample demand and fast detection speed. However, based on its technical principles, XRF has a certain radiation risk, and it is poor to detect light elements(atomic number < 11). As a promising analytical technology, laser-induced breakdown spectroscopy technology has the advantages of real-time, in-situ, simple structure and multi-element simultaneous analysis, which meets the actual need of rapid analysis for cuttings logging. The experimental system used in this study was a self-developed benchtop laser-induced breakdown spectrometer, mainly consisting of a Nd: YAG laser, three spectrometers, a digital delay generator and several optical fibers and lenses. A total of 1320 LIBS spectra of 66 rock samples, including 49 Chinese national standard rock samples as training samples and 17 supplementary samples as test samples, were collected by the instrument. The characteristic peaks including 288.16 nm(Si Ⅰ), 308.22 nm(Al Ⅰ), 445.48 nm(Ca Ⅰ), 516.73 nm(Mg Ⅰ) and 769.90 nm(K Ⅰ) were selected as target peaks for quantitative analysis. To improve the data stability and repeatability, a part area normalization was proposed. Different from total area normalization, only considering 18 characteristic spectral lines of 9 main elements in rocks, part area normalization could take into account the mapping relationship between the spectral line intensity and content of the element, reduce the spectral difference caused by the inconsistent ablative amount and weaken the influence of matrix effect. Based on the training data set, the appropriate data pretreatment method was selected from three normalization methods(including min-max normalization, total area normalization and part area normalization) for each element.To improve the quantitative accuracy, a coefficient correction method based on a principal component regression algorithm was proposed to correct the characteristic peak intensity. On the basis of the calibration curve of peak intensity after normalization and content established for each element, the ideal peak intensity and correction coefficient between peak intensity corresponding to different content could be calculated.PCR algorithm was used to establish models for calculating correction coefficients with the input of the full spectrum data. Then, the quantitative analysis models for five target elements were established upon the corrected peak intensity data. The test data set was applied to evaluate the quantitative models. The coefficients of determination(R^(2)) of the Si, Al, Ca, Mg and K between LIBS predicted values and known contents were 0.974 5, 0.941 1, 0.984 3, 0.990 1 and 0.811 4, respectively, which proved the effectiveness of quantitative models. Finally, to verify the feasibility of application in logging site, the quantitative models were integrated into the self-developed instrument to analyze 97 cuttings samples from the same well at different depths, which were collected from a drilling site in southwest China and provided by Chengdu Aliben Science & Technology Co., LTD. The predicted results were compared to the detection values obtained by using the laboratory XRF. The coefficient of determination(R^(2)) between LIBS predicted values and XRF detected values of Si, Al, Ca, Mg and K were 0.971 9, 0.960 4, 0.947 9, 0.965 2 and 0.924 5, respectively, which indicates that the instrument integrated with the quantitative models can rapidly and effectively analyze cuttings element content. In the light of the analysis results, the variation curves of contents for five target elements were plotted, which demonstrated the overall variation trends of element contents obtained by LIBS and XRF were basically consistent. Based on the variation trends of LIBS predicted values, the changes of lithology and formation could be analyzed, which were consisted with the real situation at the logging site. The research results showed that this instrument integrated with the developed quantitative models is expected to achieve rapid and effective quantitative element analysis for cuttings, and has good potential application value in the field of oil and gas exploration and development.