Explore Thermal and Mechanical Properties of Biobased Polyurethane Elastomers Through Machine Learning Models

Area of Science:

• Materials Science
• Polymer Chemistry
• Data Science

Background:

• Biobased polyurethane elastomers (BPUEs) are crucial for sustainable applications.
• Understanding their mechanical and thermal properties is essential for determining application suitability.
• Predictive models can accelerate the design and optimization of BPUEs.

Purpose of the Study:

• To develop accurate predictive models for six core mechanical and thermal properties of BPUEs.
• To identify key features influencing these properties for rational material design.
• To provide data-driven insights for creating customized BPUEs.

Main Methods:

• Compiled a dataset of over 1500 BPUE samples with detailed composition, process, structure, and property information.
• Employed domain-knowledge augmented feature engineering to select 26 predictive features.
• Utilized multi-target regression models to predict Young's modulus (YM), tensile strength (TS), elongation at break (EB), glass transition temperature (Tg), decomposition temperature (Td5), and energy dissipation factor (tanδ).

Main Results:

• Multi-target regression models achieved R² > 0.70 for YM, TS, and EB, and R² > 0.80 for Tg, Td5, and tanδ in validation and blind tests.
• Features related to chemical structure and formulation dominated, explaining over 70% of property variations.
• Identified biomass feedstocks, polyol molecular weights, and hard segment content as key regulatable variables.

Conclusions:

• Data-driven insights enable the rational design of BPUEs with specific mechanical and thermal properties.
• Formulation parameters and processing conditions (stretching/heating rates) are critical for achieving desired and repeatable BPUE performance.
• The developed models and identified features offer a pathway for efficient development of fitting-for-purpose BPUEs.