For
the efficient real-time monitoring of reaction chemistry in
a complex mixture using online spectroscopy, it is essential to develop
a mathematical tool that can automatically resolve the spectra so
that either the spectral or the concentration profile of the changing
species can be tracked easily. While self-modeling multivariate curve
resolution (SMCR) is a well-suited tool when initial profiles are
known beforehand, it is not straightforward to use when dealing with
complex mixtures. In this study, a multivariate data analysis algorithm
was designed for use with online infrared spectroscopy to provide
an instant best estimate of the reaction chemistry of a complex mixture
with no additional user input. The investigated process is thermal
conversion of oil sands bitumen, and the study employed 43 infrared
spectra from samples, collected offline, of products treated at different
temperatures and time periods. The resolved spectral and concentration
profiles can be used to understand the reaction mechanism
of the system. In addition to the concentration and spectral profile,
simple parameters were devised to monitor the changes in the key regions
of the spectral profiles. In general, the results described the possible
reaction mechanism of the investigated system and were consistent
with other experimental findings in the literature. Computationally,
the algorithm requires only a few seconds to converge and is therefore
suitable for online monitoring.