石油钻井过程中的岩屑录井无论对油气勘探开发还是钻井工程都是极为关键的技术, 而岩屑描述工作是岩屑录井过程中的一个重要的环节。 利用激光诱导击穿光谱(LIBS)技术结合判别偏最小二乘法(PLS-DA)对来自录井现场的粉砂岩、 石英砂岩、 青绿色泥岩、 黑色泥岩四种岩屑样品进行识别。 在获取LIBS光谱数据后分别建立了全谱模型和特征模型, 其中特征模型的识别正确率为86.7%, 略低于全谱模型的88.3%, 但通过特征提取使得模型中变量数由24 041个减少到27个, 极大缩减了数据量, 提高了运算效率, 更加符合岩屑录井现场快速分析的要求。 结果表明LIBS技术结合一定的化学计量方法能够对不同种类的岩屑进行快速、 有效的识别, 在岩屑录井现场中有很大的应用潜力。

Cutting identification is one of the most important links in the course of cutting logging which is very significant in the process of oil drilling. In the present paper, LIBS was used for identification of four kinds of cutting samples coming from logging field, and then multivariate analysis was used in data processing. The whole spectra model and the feature model were built for cuttings identification using PLS-DA method. The accuracy of the whole spectra model was 88.3%, a little more than the feature model with an accuracy of 86.7%. While in the aspect of data size, the variables were decreased from 24041 to 27 by feature extraction, which increased the efficiency of data processing observably. The obtained results demonstrate that LIBS combined with chemometrics method could be developed as a rapid and valid approach to cutting identification and has great potential to be used in logging field.