Data-Driven Process–Structure–Property Framework for the Design, Understanding, and Prediction of Sustainable Polymers

Data-driven methodologies are increasingly recognized as pivotal tooks for the efficient development of materials and mechanistic understanding. However, existing approaches are often limited to isolated tasks, hindering their ability to address the complex, multi-process, multitask, and hierarchical nature of sustainable polymers. In this study, we propose an integrated data-driven framework that models the process–structure–property (PSP) relationships of sustainable polymers through three synergisitic components: (1) Design – Bayesian optimization of process conditions to balance multiple material properties using surrogate objective variables derived from structure data; (2) Understanding – extraction of key crystalline features related to single-property and multi-objective axes using dimensionality reduction and XAI on X-ray scattering profiles; and (3) Prediction – hierarchical PSP modeling to regenerate structural X-ray information and predict material properties accurately from processing parameters. This unified framework enables a more systematic and automated approach to sustainable polymer development, reducing dependence on traditional trial-and-error methods and accelerating innovation.