The terahertz (THz)-based thickness measurement of thermal barrier coatings (TBCs) demands smart determination of THz peaks instead of human intervention for efficiency and accuracy, especially in the case of uneven microstructure due to imperfect spraying conditions. To this end, in this article, a long short-term memory (LSTM) framework driven by a theoretical model of THz signals is proposed to locate peaks automatically and location errors are eliminated by finding the maximum around the originally located results. Then, the local extrema are used to extract peaks for calculating refractive index online to remove the influence of the change in refractive index on thickness measurement. Finally, the simulations and experiments were carried out to verify the proposed method. The results show that our method generated identical results with those from the manual intervention, while the efficiency is significantly improved.

The liftoff point of intersection (LOI) is a desirable signal feature in the pulsed eddy current (PEC) testing. It is immune to liftoff effect caused by the change of liftoff distance between a probe and a sample. However, adjusting LOI remains an open problem for enhancement of PEC evaluation or extends the application of LOI to thicker samples. In this paper, LOI point could be adjusted by varying the rising time of pulse excitation to extend the measurement range or increase sensitivity for enhancement of the PEC evaluation. The simulations and experiments demonstrate that the difference PEC signal increases and decreases significantly slowly when the rising time of pulse excitation increases. When the rising time of pulse excitation increases, the measurement range of LOI amplitude increases, and the measurement range of LOI time decreases, whereas the sensitivity of LOI amplitude decreases, and the sensitivity of LOI time increases. This investigation helps us to optimize the rising time for the improvement of specific PEC evaluation.

The decay of pulsed eddy currents (PEC) with depth in diffusion decreases resultant signal changes, thus bringing a challenge for reliable and accurate evaluation of deep subsurface defects. In this article, a novel cross-correlation inspired residual network (ResNet), termed CCResNet, is proposed to improve the capability for smart evaluation of subsurface defects. It consists of a cross-correlation layer, a ResNet, and a novel loss function, namely, focal-probability of detection (Focal-POD) loss. The customized Gaussian wavelet basis enables us to derive weak features from heavily noised PEC signals due to the similarity by cross-correlation operation, which is the origin of the constructed cross-correlation layer. Then, a Focal-POD loss is proposed to address class imbalance and endow CCResNet with powerful capability for detection of deep subsurface defects by increasing their loss values. Finally, a semisupervised framework is built to retrain CCResNet using pseudo and labeled dataset to obtain classified results as imaging features. The experimental results show that the developed CCResNet is featured as better imaging resolution, more accurate evaluation, and intelligence in detection of deeper subsurface defects.

The uneven and complicated porose microstructure of thermal barrier coatings brings a big challenge for accurate and reliable thickness measurement due to the resultant change of refractive index. In this article, a novel insight into the physics, the first two peaks are originally adopted as the preferred signal, and a physics-inspired two-stream deep learning framework, termed as the terahertz (THz) residual network (THzResNet) is proposed to accurately infer topcoat thickness by extracting refractive index online. Specifically, the division and reciprocal layers are figured out and incorporated into the first-stream, yielding the reciprocal of refractive index. Then, a product layer connects it with time-of-flight from the second-stream to generate topcoat thickness, which endows the THzResNet in accord with the THz physics. Finally, the experiments demonstrate the advantages of THzResNet in accuracy and efficiency. The presented approach allows for addressing the challenge for accurate thickness measurement of coatings with complicated microstructures.