Abstract
Many industrial processes use in-process analytics to achieve reproducibly high product quality. During a batch distillation, for example, the quality of the product goes through an optimum and the main fraction can be collected when the concentration of all by-products drops below a critical value. The NIR spectra we obtain from such distillations show a systematic change with product quality. A quantitative model would require tens of samples for each by-product. We find that changes of concentrations of many by-products are highly correlated and there is only little variability in the data. This is not enough to extract the linearly independent contributions of each by-product. Therefore a qualitative analysis is made.
Get full access to this article
View all access options for this article.