论文部分内容阅读
激光诱导击穿光谱(LIBS)已经被证明是极具潜力的物质定性、定量分析工具之一。将激光诱导击穿光谱结合自组织映射神经网络技术,引入到石油勘探录井领域,对五类岩心样品(火山灰岩、泥岩、页岩、砂岩、白云岩)进行了岩性自动分类,为以后在录井现场实现岩性在线快速识别奠定基础。使用构造特征变量和主成分分析两种方法对原始光谱进行特征提取,相应的特征参量和主成分分别作为自组织映射神经网络的输入变量。两种输入方式下,神经网络对全部44块岩心样品岩性分类的准确率分别为75%和86%。其中以主成分作为网络输入变量,对火山灰岩、砂岩、白云岩的分类准确率可达100%。实验分析表明:在进一步提高对泥岩和页岩的区分能力后,LIBS有望成为录井领域新的岩性快速识别技术。
Laser Induced Breakdown Spectroscopy (LIBS) has proven to be one of the most promising qualitative and quantitative analytical tools. Combined with laser-induced breakdown spectroscopy and self-organizing map neural network technology, it was introduced into the field of petroleum exploration and logging and lithology automatic classification of five types of core samples (volcanic limestone, shale, shale, sandstone and dolomite) Laying the foundation for the rapid identification of lithology on the site of mud logging. The features of the original spectrum were extracted by using the method of constructing characteristic variables and principal component analysis. The corresponding characteristic parameters and principal components were used as the input variables of self-organizing neural network respectively. The neural network accuracy of lithology classification for all 44 core samples was 75% and 86%, respectively, for both inputs. Among them, the main component as the network input variable, the classification accuracy of volcanic limestone, sandstone and dolomite can reach 100%. Experimental analysis shows that LIBS is expected to become a new lithology rapid identification technology in mud logging after further improving the discrimination ability between shale and shale.