Polymer Composites Informatics for Flammability, Thermal, Mechanical and Electrical Property Predictions

TL;DR

This paper tackles the challenge of predicting polymer composite properties by extending Polymer Informatics to composites through a large, curated database and ML models for 15 flame-resistant, mechanical, thermal, and electrical properties. It compares Gaussian process regression and deep learning in single-task and physics-informed multi-task settings, deploying five Pi/MT models within PolymRize and using descriptor-based representations to handle incomplete material information. The physics-informed MT approach consistently outperforms single-task models, achieving $R^2$ near or above 0.9 for most properties and delivering strong performance on unseen data, though the electric-strength prediction remains challenging with $R^2$ around 0.54. These results highlight the potential of MT, physics-informed learning to leverage related property data, and point to NLP-driven data curation and richer chemical representations as avenues to scale AI-assisted polymer composite design into practical, industrial workflows.

Abstract

Polymer composite performance depends significantly on the polymer matrix, additives, processing conditions, and measurement setups. Traditional physics-based optimization methods for these parameters can be slow, labor-intensive, and costly, as they require physical manufacturing and testing. Here, we introduce a first step in extending Polymer Informatics, an AI-based approach proven effective for neat polymer design, into the realm of polymer composites. We curate a comprehensive database of commercially available polymer composites, develop a scheme for machine-readable data representation, and train machine-learning models for 15 flame-resistant, mechanical, thermal, and electrical properties, validating them on entirely unseen data. Future advancements are planned to drive the AI-assisted design of functional and sustainable polymer composites.

Figures (6)

• Figure 1: (Center panel) polymer composites, formed by implanting reinforcement fibers, fillers, or functional additives in a polymer matrix, and (surounding panels) their applications in different sectors of human life.
• Figure 2: Two sources of polymer composites data curated for this work are (a) research articles and (b) technical datasheets/brochures provided by the manufacturers/distributors of commercialized products. Panel (a) was adapted from Ref. yen2012synergistic with permission while panel (b) was taken from a product brochure obtained from https://www.albis.com.
• Figure 3: Top ten base polymer matrices in four group of polymer composite datasets curated and used for this work.
• Figure 4: Visualization of 5 physics-informed MT models developed (and deployed in PolymRize™) for 15 properties of polymer composites. For each of them, $R^2$ and aRMSE are provided. Each of 5 physics-informed MT models is marked by a distinct color.
• Figure 5: Predictions of the deployed models for tensile modulus $E$, stress at break $\sigma_{\rm break}$, melting temperature $T_{\rm m}$, and longitudinal coefficient of thermal expansion $\alpha_{\rm long}$ on the unseen validation datasets curated completely independently. For each target property, the base polymer matrix of the validating materials are distinguished by colors.