Physics-informed adaptive characterisation of interfacial debonding in thermal protection structures using nonlinear guided waves

Interfacial debonding in thin-walled thermal protection system (TPS) composites is a critical threat to reusable aerospace vehicles. Detecting incipient debonding is challenging due to coupled structural dispersion, multimodal propagation, and inherent material nonlinearity. Conventional linear ultrasonic methods struggle to distinguish defect-induced scattering from boundary reflections, while standard data-driven approaches lack interpretability and robustness under evolving damage states. To address this, we propose a physics-guided nonlinear guided wave characterisation framework. Integrating time–frequency mechanics with a prototype-based learning architecture, the method captures contact acoustic nonlinearity (CAN) features—such as higher-order harmonics and waveform distortion—associated with intermittent debonding contact. A Compact-Uniform-Geometric (CUG) constraint maps these signatures into an interpretable metric space, enabling the adaptive tracking of evolving damage configurations with limited labelled data. Experiments on TPS structures demonstrate that the extracted features align with wave propagation physics, improving the discrimination of debond-induced nonlinearities within mixed responses. This method achieves superior accuracy and physical consistency over existing baselines, providing a robust health monitoring framework for thin-walled composites.