control of chemical reactors by ml


The control of chemical reactors is a well-studied problem but is especially challenging in the case of polymerization reactors because many of the important characteristics of the polymer evolve continuously with time and cannot be measured directly online.

Aided by first-principles mathematical models of the polymerization process, CINEMA aims to address this problem by using machine learning to accelerate the prediction of the evolution of the polymerization such that reactors can be controlled on-line.

polymer design by ml

Polymeric systems often have elaborate structure-property relationships that are challenging to accurately establish using only first-principles models.

In CINEMA this issue will be overcome by combining first principles and machine learning approaches such that they can be trained with comparatively small amounts of experimental data.

This “science-informed” machine learning approach will be applied to industrially relevant products such as waterborne coatings and adhesives.