PEFTEC - Abstract

Abstract Information :

Effective data visualisation/ analysis strategies are essential for exploring trends and features of complicated datasets. As the bio-omic and data science fields advance so too do the tools to analyse and visualise data. From an analytical perspective, petroleum samples are one of the hardest samples to analyse and to generate useful information from. In this poster we demonstrate some practical strategies for analysing petroleum samples and discerning useful information. High-resolution field ionisation data was collected for a series of lubricant base oils without separation; some are refined from crude oil and some are synthetic in origin. The data was pre-processed within python, where components had their molecular formula assigned within a ppm error limit. The data were analysed further within a python workspace using libraries such as scikit-learn, matplotlib, and pandas. The processed data was collated into a dataset, where it could be easily filtered, sorted and manipulated. From this dataset, it is possible to apply different statistical and numerical analyses to scout for trends. Using visualisations such as principal component analysis with scatterplots and other statistical visualisations we identified meaningful correlations in the data set. One such example is the classification of samples by base oil API group. This is industrially relevant as the API group can help both product formulation and the application of a fully formulated final product. The strategies described in this poster were applied to base oils, however, the selective ionisation of hydrocarbons that field ionisation offers allows this process to be applied to formulated lubricants with minimal methodological changes. Future work aims to predict physical properties of base oils and to explore the applicability of multi-technique approaches.