Simulation-assisted Physics-based Age Prediction of Real-Serviced Thermal Barrier Coated Samples Using Infrared Thermography

Thermal barrier coatings (TBCs) are protective layers applied to metal components exposed to extreme temperatures, such as gas turbine blades and aero-engine parts. Over time, exposure to extreme operational environments results in the formation of a thermally grown oxide (TGO) layer, leading to the degradation of TBCs and affecting operational efficiency. Accurate service life prediction of TBCs at all stages is essential for predictive maintenance and extending component lifespan, yet current age estimation models rely less accurate empirical models or destructive testing. In this study, proposed a novel simulation-assisted, physics-based AI framework for predicting the service age of TBC coatings using pulsed infrared thermography. A finite-element model is developed to simulate the thermal response of multilayer TBCs under pulsed heating and generating synthetic dataset which is validated against pulsed infrared experiment dataset. A multi-fidelity 1D-convolutional neural network (1D-CNN) trained based on simulation and experiment data is developed to predict thermal diffusivity of the TBC coatings from surface temperature profiles. The predicted thermal diffusivities are used as labels for the real-serviced samples correlated with service hours and fed into age classification models including long short-term memory (LSTM) and gated recurrent units (GRU), achieving an accuracy of 95.01% and F1-score of 95.13%. The approach bridges the physics-based interpretation of thermal behavior and data-driven learning of thermal profiles for reliable age estimation with limited real experiment data, which offers a practical solution for the TBC life assessment.