Physics-informed framework for long-term polymer degradation modeling

Purpose Predicting the long-term degradation of polymeric materials remains challenging because conventional accelerated-aging experiments often require months or years to generate reliable lifetime data. This study develops a physics-informed framework for predicting long-term thermal–oxidative degradation behavior from short-duration aging data while preserving physical consistency and interpretability. Design/methodology/approach A physics-informed neural network (PINN) framework is developed to model the thermal–oxidative degradation of polyether ether ketone. The approach embeds Arrhenius reaction kinetics, oxygen diffusion and a continuous property–damage relationship directly into the learning architecture. A transparent synthetic benchmark dataset generated from polymer aging equations is used to evaluate the framework under controlled conditions. The model jointly learns spatiotemporal degradation fields and physically meaningful material parameters while enforcing governing reaction–diffusion physics. Findings The trained PINN recovers the activation energy within 1.2% of the reference value, achieves a short-term test root mean square error of 0.008 and demonstrates stable long-horizon extrapolation behavior under coupled reaction–diffusion physics. The learned solution remains consistent with governing equations across spatial and temporal domains, enabling physically interpretable and numerically stable extended-time predictions. Compared with purely data-driven models, the framework maintains smoother and more physically consistent degradation trajectories beyond the training window. Originality/value This study introduces a physics-informed degradation modeling methodology integrating parameter identification, physics-constrained learning and stable extrapolation within a unified framework. The work focuses on methodological validation using a controlled synthetic benchmark dataset rather than deployment-level industrial digital twin implementation. The framework is intended as a foundation for future experimentally validated degradation prediction and lifetime-assessment systems for polymer-based components.