CORDIS - Forschungsergebnisse der EU
CORDIS

Advances in Spectroscopic Monitoring of Styrene Emulsion Polymerization

Final Report Summary - ASMOSEP (Advances in spectroscopic monitoring of styrene emulsion polymerization)

Acquisition of accurate and robust calibration models and thus applicability of spectroscopic methods in process analytics for quantitative NIR analysis of turbid samples is hindered by nonlinear particle effects such as multiple light scattering that degrade conventional multivariate linear calibration models and make extraction of chemical information from such samples challenging. There are essentially two ways to deal with undesirable particle effects in NIR measurements: remove / minimise them by means of empirical pre-processing or separate them from absorption using first principles i.e. by invoking light propagation theory such as the radiative transfer theory. In either case, the goal is to obtain a measure of absorption per unit length, which is independent from variations in path length of photons that occur due to multiple scattering and linearly proportional to concentrations of constituents.

Although a considerable progress has been made in the empirical scatter correction techniques they are not expected to solve the problem of multiple light scattering completely because they are intrinsically approximate. And while they may be sufficient in some practical situations they may not be able to accommodate the whole complexity of nonlinear multiple scattering effects in many situations. Therefore, approaches based on extracting the absorption part independent from scatterer (i.e. particle) effects using fundamentals of light propagation are gaining more and more impetus in the field of NIR quantitative analysis of highly scattering samples. The aim of this project was advancing and testing one of such approaches, the concept of which was initially proposed and developed in the previous FP6 project (INTROSPECT).

The scatter correction methodology presented there was based on eliminating multiple light scattering effects, which is the major source of nonlinear variation in NIR measurements, by extracting bulk absorption coefficient via inversion of radiative transfer equation. However, although a significant improvement in the prediction performance was achieved by decoupling absorption and scattering in comparison with using empirical pre-processing techniques, it was still not up to the level that can be achieved with the transparent media where the measured absorption is linearly proportional to chemical concentrations, because, nonlinear particle effects on the bulk absorption coefficient were not taken into account. Hence, one of the main goals of this project was to develop and test the full correction approach. With increasing availability of accurate, versatile and cheap spectroscopic measurements this approach lends itself very well to online monitoring of turbid process streams.

Three key issues have been addressed in this project: the scatter correction and calibration method (i.e. accuracy and robustness of the extraction of chemical information), the measurement setup and the analysis time. The main result - a new improved scatter correction method for the quantitative NIR analysis of turbid samples has been created / developed, which gives significantly better accuracy and robustness than the state of the art empirical scatter correction techniques and is fast enough for applying it to online analysis or monitoring of chemical concentrations.

This method allows us to enjoy high accuracy and more importantly - good robustness in the quantitative analysis of highly scattering streams / systems over an unlimited range of scatterer sizes, which is of great benefit to process and quality control in processes such as styrene emulsion polymerisation where accurate monitoring of concentrations of chemicals is very important. The developed new analytical technique was tested on samples typical to styrene emulsion polymerization reaction medium. It is worth noting, however, that the performance of calibration models which we have with transparent liquids has not been achieved yet and there is some scope for future work.