Detection of unexpected species in soft modelling of vibrational spectra
Spectral soft modelling is a straightforward approach to analyse vibrational spectra obtained from process monitoring using IR or Raman spectroscopy. However, when unexpected species, i.e. species that have not been included in the calibration, contribute to the signal, the data evaluation may lead to significant errors in the determined mixture composition. We present a simple procedure that allows detection of such unexpected species and in many cases their identification as well. The evaluation approach is based on a systematic and dynamic piecewise spectral fitting algorithm facilitating the quantification of the calibrated species and the detection of unexpected ones.
Fast and accurate analytical technology is the key to allow effective monitoring and control of industrial processes. While technologies based on sampling, for example gas chromatography (GC) and high performance liquid chromatography (HPLC) are still the workhorses of most operators, spectroscopic methods have experienced an ever increasing interest over the past decades (1-4). The main reasons for this development are: 1) spectroscopic techniques can be applied inline, for example using fibre technology, and hence they do not require sampling, 2) spectroscopic techniques do not require any kind of time consuming downstream processing like for example the separation column in chromatographic instruments, hence they can provide short acquisition tim