Three-dimensional nondestructive characterization of delamination in GFRP by terahertz time-of-flight tomography with sparse Bayesian learning-based spectrum-graph integration strategy

Recently, terahertz time-of-flight tomography (THz TOFT) has attracted attention for the localization and quantitative characterization of internal defects in composite materials due to its superior time and spatial resolution and high penetration for various non-metallic materials. However, the dispersion in the THz optical constants and a lack of prior knowledge of defect types in a composite sample significantly degrade the temporal and spatial resolution in THz TOFT measurements, which hinders the super-resolution characterization of the defects internal the material. Here, a sparse Bayesian learning-based (SBL) spectrum-graph integration method is proposed to enable super-resolution characterization of multi-ply glass fiber reinforced polymer (GFRP) composites with delamination defects. The core of this strategy includes spectrum localization and graph quantification. The spectrum-localization process is first implemented to accurately obtain the depth location of delamination defects by SBL, in which dispersion compensation based on double parametric dictionaries is performed to address the dispersion and overlapping issues of the dispersive THz echoes reflected from various interfaces within the sample. Then, based on the estimated depth information for each delamination defect, the graph-quantification process is carried out to characterize the transverse location and size of each delamination defect by the local peak-to-peak stratified imaging method. The estimated size accuracy for the delamination defects can reach more than 95%. Overall, the proposed strategy can quantitatively and simultaneously characterize the axial and transverse information of delamination defects internal GFRP composites with high resolution.