Development of a Novel Multivariate Chemometric Platform for Discrimination and Authentication of Nigerian Crude Oil Blends Using Infrared Spectroscopy and Machine Learning

The authentication and provenance determination of crude oil is of significant economic and regulatory importance for Nigeria, which produces multiple distinct crude oil grades, yet a rigorous chemometric platform for rapid discrimination of Nigerian crude blends is absent from published literature. This dissertation develops a novel multivariate chemometric platform integrating attenuated total reflectance Fourier-transform infrared spectroscopy with advanced machine learning algorithms for the discrimination, blend ratio quantification, and provenance attribution of Nigerian crude oil samples. A reference library of 180 certified crude oil samples representing Bonny Light, Qua Iboe, Forcados, Brass River, and Antan crude grades was assembled from terminal operators and analysed by ATR-FTIR, gas chromatography with flame ionisation detection, and stable carbon isotope ratio analysis as complementary reference methods. Spectral preprocessing including standard normal variate, Savitzky-Golay derivatives, and extended multiplicative scatter correction was systematically evaluated. Classification models built with support vector machines, random forest, and artificial neural network algorithms were benchmarked against partial least squares discriminant analysis. Quantitative blend ratio prediction models were developed by partial least squares regression and compared with artificial neural network regression. The optimised support vector machine model achieved 99.1 percent correct classification of five crude grades using ATR-FTIR alone, outperforming all benchmark models. Blend ratio quantification root mean square error of prediction was 0.8 percent for binary blends. The platform was successfully applied to detecting adulteration of Bonny Light with condensate fractions at concentrations as low as 3 percent. This dissertation represents an original contribution to petroleum analytical chemistry and oil trade fraud detection. Keywords: chemometrics, crude oil authentication, FTIR spectroscopy, machine learning, Nigerian crude grades