Spectra calibration modeling and statistical analysis for cumulative quality interpretation and prediction

Abstract

An improved calibration modeling and statistical analysis algorithm is proposed for spectra quality interpretation and prediction is presented here. In a previous work, the frequency-band varying characteristics of the underlying spectra over the entire wavelength were treated by separately analyzing the spectra in each sub-band. Following that, the current major task lies in how to further comprehend and model the cumulative effects of different sub-bands on qualities from the inter-sub-band viewpoint. It reveals that one part of the underlying variation in each sub-band stays invariable over sub-bands, whereas the other part changes with the alternation of sub-bands. The original variation in each sub-band can thus be separated into two different parts, the common and specific ones. They reveal sub-band-similar and dissimilar contributions on quality interpretation, respectively, which are referred to “repetitive” and “complementary” cumulative effects in this approach. Correspondingly, different calibration modeling and analyses are performed to explore their respective and joint roles in quality interpretation. The feasibility of the proposed calibration analysis algorithm is verified through both simple numerical data and real spectra data. © 2011 American Institute of Chemical Engineers AIChE J, 2012