Nondestructive Measurement of Multilayer Coating Thickness on Interwoven CFRP Using Terahertz Time-Domain Polarimetric Imaging

Coating thickness measurement and inspection of defects beneath optically opaque surfaces are among the most promising commercial applications of the terahertz (THz) imaging technology. However, there are two main sources of complexity in THz spectral measurements that have resulted in reduced accuracy and high uncertainty in the determination of coating thicknesses. These factors include: The need for a priori knowledge of the complex index of refraction of each coating layer, and the complex and polarization-sensitive reflectivity of samples with carbon-fiber-reinforced polymer substrates. In this article, we propose a combined hardware and software solution to address these two limitations. We employ the polarimetric version of our portable handheld spectral reflection scanner and use a novel training model on time-of-flight measurements obtained by sparse deconvolution of the THz time-domain pulses. Our method does not depend on separate measurements of the index of refraction of the coating layers; rather, it relies on a physics-based linear model, trained using a small subset of the sample data. We show that thickness measurements with mean square error of about 10 \mum and accuracy of more than 90% can be achieved when the useful bandwidth of the scanner is limited to about 1.5 THz using photoconductive antenna emitters and detectors. Finally, we simulated the spectral response of samples with slight variations in the refractive index of the coating sublayers to evaluate the validity of the linear model in extracting the optical properties of individual layers. We show that the anisotropic response of the carbon fiber bundles and the weave pattern structure of the substrate can significantly influence the accuracy of the coating measurement, and therefore, a polarimetric imaging approach should be used for inspection of similar samples.