近红外光谱技术结合主成分聚类分析判别海面溢油种类

Oil Spill identification by near-infrared spectroscopy coupled with clustering analysis based on principle components

鉴于目前没有一种方法能独立解决溢油鉴别的所有问题,本文提出了近红外光谱技术结合主成分聚类分析鉴别溢油种类的方法。通过有机溶剂萃取出自行配制的汽油、柴油和润滑油模拟样品中的溢油后记录其近红外光谱,对5800～6200cm-1区段范围内的谱图经多元散射校正(MSC)、Norris一阶导数平滑预处理处理后求其主成分,并在主成分的基础上引入Ward聚类分析法(离差平方和法)对样品分类。结果表明近红外光谱技术结合聚类分析能对体积分数在0.4～0.8mL/L间的海面溢油样品正确、快速分类,近红外光谱技术结合主成分聚类可作为溢油鉴别的一种辅助方法。

Owing to none of the existing method can do it all for oil identification, but a method by near-infrared spectroscopy (NIR) coupled with clustering based on principle components is proposed. Petroleum oil stimulating samples of gasoline, diesel fuel and lubricating oil were employed to develop the method. NIR spectra ranges of 5800～6200cm^(-1) were recorded with the multiplicative signal correction and Norris first derivative filter pretreatments. The clustering analysis by Ward's method base on the different principle components was discussed. The results show that the method provides the rapid and accurate classification for spilled oil with concentration between 0.4～0.8 μL/L.