Petroleum is an economically and industrially important resource. Crude oil must be refined before use to ensure suitable properties of the product. Among the processes used in this refining is distillation and desulfurization. In order to optimize these processes, it is essential to understand them.
Comprehensive two-dimensional gas chromatography (GCGC) is a method
for analyzing the volatile parts of a sample. It can separate hundreds or thousands of compounds based on their boiling point, polarity and polarizability. This makes it ideally suited for petroleum analysis.
The number of separated compounds makes the analysis of GCGC chromatograms tricky, as there are too much data for manual analysis , and automated analysis is not always trouble-free: Manual checking of the results is often necessary.
In this work, I will investigate the possibility of another approach to analysis of
GCGC chromatograms. Instead of evaluating the chromatograms individually and determining the content of single compounds or groups of compounds, the raw data of all the samples are analyzed simultaneously.
Instrument variation negatively impacts this approach to analysis. It is important to remove non-sample contributions to variation before analysis. In the thesis, I present methods for handling some of these contributions.
Another problem with this approach is that models tend to only be one of
many that could possible describe the data. The underlying phenomena is another of these models, but it is often impossible to find it. For a special class of models, multi-way models, unique solutions often exist, meaning that the underlying phenomena can be found. I have tested this class of models on GCGC data from petroleum and conclude that more work is needed before they can be automated.
I demonstrate how to analyze GCGC data of petroleum with several examples, and how this analysis can be used to investigate the properties of petroleum and refinery processes.