Quantitative and qualitative assessment of petroleum products by spectroscopic and chemical data analytics / Abd Rahim Othman

Abd Rahim , Othman (2022) Quantitative and qualitative assessment of petroleum products by spectroscopic and chemical data analytics / Abd Rahim Othman. Masters thesis, Universiti Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (214Kb)
	PDF (Thesis M.A.) Download (2539Kb)

Abstract

This research study describes the use of Fourier Transform Near Infra-Red (FT-NIR) technology with various chemometric methods in petroleum products. Current practice in the refinery’s quality assurance and quality control laboratories relies on the use of conventional method measurements in the quantitative and qualitative assessment of their petroleum products. These conventional methods consume high workforce, longer analysis time and high operating expenditure. A new innovative way of qualitative and quantitative measurement and application were explored to improve and address the pain points and provide a total solution. This study aims to evaluate and assess the quality of petroleum products through spectroscopic and chemical data analytics, both qualitative and quantitative assessments. In this work, coupled near infra-red spectroscopy and chemometrics techniques for calibration models development, i.e. Partial Least Square (PLS) and Principal Component Regression (PCR) and Near Infra-red (NIR) spectra pre-treatment or re-processing, i.e. Multiplicative Scatter Correction (MSC) and Savitzky Golay Second Derivative (SGSD) were employed accordingly for rapid and simultaneous determination of chemical and physical properties of petroleum hydrocarbons. The coupled NIR and chemometrics methods are an alternative to the existing laboratory reference methods to address the refinery pain points. FT-NIR spectroscopy has been successfully utilized to rapidly identify and discriminate three types of petroleum products (gasoline, diesel, and kerosene) using Principal Component Analysis (PCA). More than 95% of each product was accurately identified and differentiated. This qualitative multivariate measurement is important when fast results are required at the operation site, such as during product transfer cross-contamination and adulteration or illegal product blending. In addition, qualitative measurement by PCA was used to differentiate gasoline and diesel fuels directly sourced from refineries without additive. In contrast, additives were added to the gasoline and diesel fuels, such as corrosion inhibitors, detergency, and lubricity improvers, to enhance the engine's performance and protection of the engine components. Diesel with and without palm methyl ester (PME) blend were also determined qualitatively using PCA based on significant the presence of fatty acid methyl ester (FAME) in diesel. This work demonstrates the multivariate calibration strategy for the simultaneous near-infrared spectrometric determination of the physical and chemical properties of the petroleum products, namely the boiling point at 95% recovery (T95%), flash point (FP), cloud point (CP) and cetane index (CI) which include the spectral region selection, calibration/validation set partition, data pre-processing, and regression. Based on the results, the calibration constructed on the combination region of 4800-4000 cm-1 using the randomly selected calibration set managed to deliver excellent predictive performance in terms of coefficient of determination, root mean square error of cross-validation, root mean square error of prediction and the ratio of performance deviation. Moreover, all the developed models satisfied the reproducibility requirement of respective American Society for Testing and Materials (ASTM) standard methods regardless of the employment of multiplicative scattering correction/Savitzky-Golay second-derivatization and principal component regression/partial least square regression. This revealed that the fitness of the model relies upon every single calibration component. It was also realized that data pre-treatment is crucial in delivering predictive-performing predictions.

Actions (For repository staff only : Login required)