Origin based classification of crude oils by infrared spectrometry and chemometrics

Crude oil samples from different Iranian petrol resources in both, raw and mixture forms have been characterized by attenuated total reflectance mid infrared spectroscopy. Obtained spectra were classified by chemometric techniques to propose a method for geological based classification of crude oil samples. Totally 251 samples from 7 petrol fields and 3 mixtures were analyzed. Mean centering and principal component analysis (PCA) supported - leverage value based outlier detection were used as preprocessing approaches. PCA, cluster analysis and soft independent modeling of class analogy (SIMCA) were utilized to classify the spectra. Obtained results confirmed that SIMCA is a robust chemometric technique for origin classification of crude oil samples based on their IR spectra, while the mixture samples were also classified satisfactory in some cases. Root mean square error, method precision and regression coefficient for the prediction of origin of an independent validation set of 111 samples were 1.41%, 96.7% and 0.957 respectively.