Thin Coatings Thickness Detection of CFRP Based on Terahertz Time-Domain Spectroscopy
Yu-lei Huang,Wei-xing Li,Mei-qiang Zhu,Nan Zhang,Lin Ke,Liang Zou
DOI: https://doi.org/10.3964/j.issn.1000-0593(2024)10-2753-08
2024-01-01
Spectroscopy and spectral analysis
Abstract:To address the challenges of substrate anisotropy, multiple reflections, echo overlap, and environmental noise in measuring thin coatings on CFRP, a coating thickness measurement method based on terahertz time domain spectroscopy and sparse decomposition is proposed. This method has been successfully validated on simulated waveforms and real samples, CFRP composites are multilayer, cach layer is typically composed of carbon fibers firmly bound with cpoxy matrix, CFRP composites are anisotropic, with variable reflection at the same location in different incidence directions and locations in the same incidence direction. As a result, the measurement accuracy of coating thickness needs to be improved when the optimization algorithm based on the theoretical model is used. In the case of thin coatings, direct differentiation of coating interfaces in the time domain is challenging due to noise and multiple reflections. Therefore, the conventional time-of-flight method(TOF) alone can't work well. To address this, sparse decomposition is utilized to locate the interfaces of thin coatings, and then the TOF is applied to calculate the thickness, Initially, the relationship between terahertz reflection signals on multilayer structures and sparse decomposition models is analyzed, and a sparse dictionary is constructed using reference signals. Then, with the prior knowledge that the propagation path of terahertz signals only changes at coating interfaces, the number of nonzero pulses in the sparse decomposition of the terahertz reflected in the number of coating layers determines the signal. Based on the above facts. the LAOMP algorithm is selected to solve the sparse representation of the terahertz-reflected signal. The LAOMP algorithm specifies the sparsity directly, facilitating the selection of sparse pulses to locate the layer interface. After obtaining the pulses corresponding to each coating interface, the thickness can be solved by TOF. Finally, this proposed method is validated for both simulated data and the samples with two coating layers. The total thickness of CFRP samples is 102 and 66 mu m, respectively. The method also is compared with the classical spectral projection gradient-based L1 (SPG1.1), The results show that the LAOMP algorithm can effectively detect the thin coating of 35 pm on CFRP with a small refractive index difference of adjacent coatings, and the total error of the double costing sample is less than 11% when the total thickness is 66 am, Compared with SPGLI, the parameter tuning of the LAOMP is casier, and the decomposed sparse impulses are more stable.