Online monitoring of polymerization reactions is not only important due to the high exothermic nature of most polymerization systems; the data gained also provides information on product composition and quality and control possibilities thereof. Many inline and online methods are still in development and are better suited for application on the laboratory or pilot-plant scale. Industrial polymerization plant environments pose additional technical and financial challenges and constraints for the use of such systems. Available methods and current developments are reviewed with regard to their practicability and usefulness under these aspects.

Monitoring and control of resin polymerizations is essential for high process safety, high product quality, and competitive production costs. Vinyl acetate resins created by bulk and solution polymerization usually have a high molecular weight and viscosity, making sample extraction for analysis a cumbersome process. In-process analytical methods, like Raman spectroscopy, enable not only the measurement of monomer and polymer composition during the reaction without complex mathematical calibrations but also the determination of final product properties. The latter is also possible in conjunction with other process data like temperatures and feed rates and with a multivariate approach. An overview of challenges, necessary considerations, and results is given. Graphical abstract Prediction of product quality parameter viscosity using online-Raman spectroscopy data vs. reference data (Hoeppler viscosity measured in the lab after sample extraction) using partial least squares modelling.

Summary: Monitoring and control of polymerization reactions is essential for high process safety, high product quality and competitive production costs. Ideally the entire process chain is regarded, starting with raw material analysis and the polymerization reaction up to the measurement of polymer‐ and application‐ properties. Process data like temperatures and pressures can be used to monitor reaction trajectories in a cost effective way, e.g. using calorimetric evaluations. Additional sensors can provide chemical or morphological information but must be robust and inexpensive for commercial applications (e.g. NIR‐ or Raman spectroscopy). Data from these different sources can be used for multivariate data analysis, delivering additional insights that might not be obtained by direct measurement.

Measuring product properties during vinylacetate‐ethylene (VAE) polymerizations can be difficult due to economic and technical limitations. The use of soft sensors, mathematical methods using available process data (current and historic), can be an effective solution to infer the sought after property and calculate properties not directly measurable. The setup of a soft sensor in an industrial environment for a long term, successful application needs to take into account several additional hurdles. These include the variability of the process, soft factors leading to the acceptance of this tool in production, and the effort needed for long‐term support and maintenance among others. These points are illuminated using the successful development and implementation of a VAE dispersion process based tile adhesive tensile strength soft sensor.

No abstract