Determination Of Bitumen Concentration by Analyzing Laser-Induced Breakdown Spectra (LIBS) Using Partial Least Squared Regression (PLSR) Method

Abstract

The determination of bitumen content in oil sands is essential for the characterization of tailings and enhancement of oil extraction efficiency. However, traditional bitumen content measurement methods, such as Dean-Stark Extraction, necessitate extensive lab-based sample preparation, increasing processing time. In-field bitumen content evaluation would substantially assist decision-making in order to run more effective mining and recovery operations. Here, we demonstrate the rapid determination of bitumen concentration using laser-induced breakdown spectra (LIBS) and four different normalization approaches in partial least squares regression (PLSR) analysis of wet and dry tailing sands consisting of 0.5-12% bitumen. The results reveal that wet bitumen samples require fewer PLSR components to explain 90% of the variance in LIBS spectra than dry bitumen samples, demonstrating the suitability of LIBS for analyzing samples with little preparation. A good correlation (R2 > 0.96) between LIBS spectra and bitumen content for both wet and dry samples was obtained, however the prediction standard deviation (SD < 4%) was improved for wet samples. The chemical composition of bitumen was also determined by computing the percentage abundance of each atomic element from the PLSR coefficient of the LIBS spectra.