Volumetric Ablation of Phenolic Resin in Extreme Environments: A Variable-Property Physics-Informed Neural Network Framework.

Phenolic resin ablation is a complex multi-physics process characterized by intense pyrolysis, strong thermal-chemical coupling, and drastic evolution of material properties. Predicting the through-thickness thermal response is computationally challenging, primarily due to the intricate coupling and the non-linear evolution of properties with temperature and pyrolysis. This study proposes a variable-property integrated Physics-Informed Neural Network (PINN) framework to model the high-temperature ablation of phenolic resin. Specifically, by constructing a loss function that strictly embeds the physics of heat conduction and pyrolysis, the framework explicitly captures the non-linear evolution of five thermophysical parameters, including thermal conductivity, specific heat, and density, throughout the ablation process. Numerical verifications across multiple scenarios demonstrate that the proposed variable-property PINN framework can simultaneously predict the sharp transient temperature and complex volumetric pyrolysis fields with a maximum relative error of less than 8.0% compared to the high-fidelity FEM baseline solutions. This mesh-free, efficient approach offers a novel paradigm for solving strongly coupled, variable-property ablation problems. It effectively overcomes the computational stiffness of conventional techniques, thereby providing valuable insights for the precise optimization of thermal protection systems.