A Physics-Informed Manifold Neural Operator Framework for Multi-Parameter Prediction of Polymer Aging in HTPB Solid...

Area of Science:

• Materials Science
• Polymer Chemistry
• Computational Mechanics

Background:

• Predictive modeling of thermal aging in HTPB solid propellants is complex due to their particulate-reinforced polymer nature, limited data, and unclear aging mechanisms.
• Existing models struggle with nonlinear parameter coupling and sparse experimental data, hindering accurate life assessment.

Purpose of the Study:

• To develop a novel Physics-Informed Manifold Neural Operator (PIMANO) framework for accurate multi-parameter prediction of polymer aging in HTPB solid propellants.
• To establish a physically interpretable bridge-state variable (crosslinking density) linking aging conditions to viscoelastic responses.
• To enhance the prediction accuracy and stability for material life assessment.

Main Methods:

• Utilized accelerated thermal aging, stress relaxation, and swelling experiments to gather data on aging temperature, time, crosslinking density, and viscoelastic parameters.
• Reconstructed a continuous aging-state field using radial basis function (RBF) interpolation and introduced crosslinking density as a bridge-state variable.
• Developed the PIMANO framework integrating manifold learning, DeepONet operator learning, and physical constraints (non-negativity, evolution direction).

Main Results:

• The modified Arrhenius-Avrami model accurately described crosslinking density evolution (R² = 0.988).
• PIMANO achieved high accuracy in predicting viscoelastic responses, with R² = 0.9995 and significantly reduced errors (RMSE, MAE, MRE) compared to traditional models.
• Validation on unseen aging temperatures demonstrated PIMANO's robustness and generalization capability (average R² of 0.9469-0.9647).

Conclusions:

• PIMANO offers a highly accurate, stable, and physically interpretable framework for multi-parameter aging prediction in HTPB solid propellants.
• The framework successfully integrates physics-informed constraints with advanced machine learning for improved material life assessment.
• This approach addresses the challenges of sparse data and complex aging mechanisms in energetic materials.