Simulation-assisted Multimodal Deep Learning (Sim-MDL) Fusion Models for the Evaluation of Thermal Barrier Coatings using Infrared Thermography and Terahertz Imaging

Thermal Barrier Coatings (TBCs) are critical components in high-temperature applications, such as gas turbines and aerospace engines, where they protect the underlying substrate metals from extreme thermal stress and extend component life. Accurate evaluation of TBCs is essential to improve operational efficiency, optimize predictive maintenance strategies, and extend component life. Popular non-destructive evaluation (NDE) techniques such as infrared thermography (IRT) and terahertz (THz) imaging have been widely used for TBC inspection with limitations when used independently, including sensitivity to surface conditions, limited penetration depth, and challenges in inspecting multi-layer coatings and detecting subsurface defects. To address these challenges, our study proposes a novel framework called simulation-assisted multimodal deep learning (Sim-MDL) that integrates the strengths of IRT and THz imaging for a comprehensive evaluation of TBCs. To generalize the study to a range of thermophysical properties of TBCs, the study uses simulation-generated data along with experimental data for training deep learning models. The data from IRT and THz modalities are fused in the Sim-MDL models to enable characterization of the TBC topcoat layer. IRT and THz experimental data, together with simulations, form a large dataset that is used to train deep learning models. The framework is tested and optimized for multimodal data fusion using two DL architectures based on convolutional neural networks (CNN) and long short-term memory (LSTM), allowing the model to learn correlations and complex patterns across the IRT and THz modalities. The study is conducted on four newly coated samples ranging in thickness from 24 to 120 µm. An attention-based LSTM model trained on both simulation and experimental data shows high prediction accuracy with MAPE values ranging from 2.06–4.43% for thermal conductivity, 2.05–3.57% for heat capacity, 11.53–1.75% for topcoat thickness, and 0.27–1.05% for refractive index, respectively, for the topcoat layers of four samples. Our model outperformed the single-modality models and conventional parameter estimation methods in terms of accuracy and robustness, highlighting the potential of multimodal data for automated analysis of TBC in industrial settings.