AI-Based Forecasting of Polymer Properties for High-Temperature Butyl Acrylate Polymerizations

Area of Science:

• Polymer Chemistry
• Computational Chemistry
• Materials Science

Background:

• High-temperature polymerizations with self-initiating monomers offer advantages like faster rates and lower viscosity.
• However, these processes yield complex polymer microstructures, including branching and macromonomer formation.
• Molecular-level understanding requires computationally intensive simulations.

Purpose of the Study:

• To apply AI-based forecasting for predicting polymer properties in self-initiated polymerizations.
• To utilize kinetic Monte Carlo simulations for generating training and testing data for machine learning models.
• To investigate the data requirements for reliable AI predictions in polymer science.

Main Methods:

• Employed machine learning models, specifically random forest and kernel density (KD) regression.
• Utilized kinetic Monte Carlo simulations to generate synthetic polymerization data.
• Applied explainability methods to validate model predictions against expert knowledge.

Main Results:

• Achieved excellent predictive performance (R² > 0.99, MAE < 1% for KD regression) for monomer concentration, macromonomer content, and molar mass distributions.
• Demonstrated accurate prediction of average branching levels over time.
• Confirmed that AI model variable importance aligns with established chemical principles.

Conclusions:

• AI-based forecasting offers a computationally efficient alternative to traditional simulations for understanding complex polymerizations.
• Machine learning models can reliably predict key polymer characteristics with high accuracy.
• Explainable AI reinforces the validity of these predictive models in polymer chemistry research.