Adding Machine Learning to the Polymer Reaction Engineering Toolbox

Abstract

Mathematical modeling has long played a crucial role in the development of macromolecular systems, offering a framework for designing polymeric materials to achieve specific targets. Traditionally, these models have been grounded in first-principles knowledge of the underlying physical and chemical processes. However, in recent years, data-driven approaches, particularly those based on machine learning (ML), have gained significant traction. Unlike conventional models, which are constrained by predefined assumptions, ML models offer greater flexibility, which can have both positive and negative consequences. On the positive side, the flexibility of machine learning models makes them particularly useful for analyzing complex systems, such as those common to polymeric materials, which are often challenging to fully capture with traditional approaches. However, a well-known drawback is that their lack of physical grounding can sometimes result in unrealistic predictions. In this review, recent advances in the use of machine learning in the field of polymer reaction engineering are discussed, with a particular focus on how to incorporate the strengths of both first-principles and data-driven mathematical models. The review begins with an overview of the key machine learning techniques currently available and then explores specific scenarios where their application has proven beneficial in modelling of polymeric systems. Following an in-depth discussion of the state-of-the-art with respect to polymer reaction engineering applications, the article concludes with a perspective on the future of this nascent field, outlining key challenges and opportunities for further research.