Machine Learning for Polymer Development

Machine learning enables the quantitative establishment of complex relationships, such as those present in polymerisation systems, although it requires large amounts of data for training. However, in many areas of the physical sciences, and especially in polymers in dispersed media, data are limited due to the complexity and time required for their characterisation.
Our research seeks to go beyond exclusively data-driven approaches and demonstrate the potential of knowledge-guided machine learning. In this approach, scientific fundamentals are directly integrated into the models, enabling the exploitation of algorithmic capacity to analyse data without the need for massive information datasets. Among the main applications are reactor control, formulation design and the development of new polymeric materials.