A Hybrid ANN–Physics Framework for Inverse Heat-Flux Estimation in Thermal Ablation
Year: 2026
Type: article
Abstract: Thermal ablation plays a critical role in the design and performance of thermal protection systems (TPS) used in aerospace applications. This study addresses the inverse heat-transfer problem in non-charring ablation, where unknown surface heat flux must be reconstructed from limited in-depth temperature measurements. Such problems are inherently ill-posed and challenging under sparse sensing and measurement noise. A hybrid modeling framework is developed by coupling an artificial neural network (ANN) with a physics-based numerical solver (PNS). The computational domain is divided into an accessible region, where sparse temperature data are used to train the ANN, and a non-accessible region, where the governing transient heat-conduction equations are enforced through the numerical solver. This hybrid ANN–PNS approach maintains physical consistency while reducing dependence on dense sensor data. Results demonstrate heat-flux reconstruction errors below 5%, even with up to 7.5% Gaussian noise and limited sensor configurations. The method remains stable under poor initial guesses, whereas conventional PINNs exhibit convergence issues and reduced accuracy. Overall, the proposed framework provides a robust and efficient alternative for inverse thermal ablation analysis under practical constraints.
Authors: Islam, Md Shariful, Dutta, Prashanta
Institution: Washington State University
Language: English
Subfield: Statistical and Nonlinear Physics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7160311529