Terahertz based Thickness Measurement of Thermal Barrier Coatings Using Hybrid Machine Learning
Yunli Gong, Binghua Cao, Hong Zhang, Fengshan Sun, Mengbao Fan
Abstract
Thickness of thermal barrier coatings (TBCs) is closely related to the performance of hot-section components in aero-engine. In this paper, a model of Elman neural network optimised by whale optimisation algorithm (WOA) with principal component analysis (PCA) based on singular value decomposition (SVD) was proposed for terahertz (THz)-based thickness measurement of TBCs. First of all, the theoretical model of THz propagation in TBC is employed to generate simulated signals to meet the demand of sample size for Elman neural network training. Second, PCA based on SVD is used to reduce the dimension of each signal. In order to decrease the possibility of falling into local optimisation and improve the output accuracy of neural network, the weights and biases of network are optimised by WOA. Finally, the performance of the models was evaluated by statistical assessments. Our results show that the thickness measurement method combined with hybrid machine learning adopted in this paper have improved the accuracy of thickness measurement and occupied great potential applications on thickness measurement of TBCs.