Forecast of Glass Transition Zone of Thermoset Polymers Using a Multiscale Machine Learning Approach

Area of Science:

• Polymer Science
• Materials Science
• Computational Materials Science

Background:

• Traditional machine learning (ML) models predict a single glass transition temperature (T_g), neglecting the transition's dynamic nature.
• A comprehensive understanding of the glass transition zone is crucial for advanced polymer applications.

Purpose of the Study:

• To develop a novel ML approach for predicting the storage modulus change across the entire glass transition zone in thermoset polymers.
• To offer a more holistic prediction of polymer phase transitions beyond a single T_g point.

Main Methods:

• Utilized a multiscale fingerprinting technique with microscopic, mesoscopic, and macroscopic features as input.
• Employed support vector regression (SVR), artificial neural network (ANN), and Gaussian process (GP) models for predicting the glass transition zone.
• Identified four key features essential for accurately modeling modulus change with temperature.

Main Results:

• The artificial neural network (ANN) model demonstrated superior performance in predicting the glass transition zone compared to SVR and GP models.
• The developed model successfully predicted glass transition zone curves for three classes of new polymers.
• Experimental validation confirmed that the model captured the essential characteristics of the experimental curves.

Conclusions:

• The novel ML approach provides a more comprehensive prediction of polymer glass transitions.
• This method advances ML applications in polymer design and material innovation.
• The approach enhances the ability to manipulate polymer properties for future material science advancements.