Establishment and Optimization of Temperature Compensation Model for Benzoic Acid Detection Based on Terahertz Metamaterial
Jun Hu,Yin Li,Rui Chen,Yong He,Yande Liu
DOI: https://doi.org/10.1016/j.infrared.2022.104101
IF: 2.997
2022-01-01
Infrared Physics & Technology
Abstract:To regulate the use of food additives is of great significance to human health, for the excessive use of food additives can cause damage to human body so serious as organ failure and even cancers. Benzoic acid (BA) is a widely used food preservative, and it is of great importance to achieve high-sensitive detection of it. Conventional THz detection method can obtain the fingerprint spectrum of the sample in the terahertz band, it can only detect a sample to be detected at a milligram level, and it is not suitable for the micro or trace analysis of the sample. In this paper, an enhanced terahertz (THz) micro-nano structure of metamaterial based on electro-magnetic theory was designed to achieve enhanced detection of benzoic acid additives by THz technology. THz spectroscopy is more sensitive to temperature changes than other spectroscopic detection techniques, so it is particularly important to find out a method of spectral temperature compensation to minimize the inference caused by temperature variations in the detection of samples with variable temperature. To this end, based on the enhanced technology of THz metamaterial, an optimized compensation method for external temperature variations was explored in this paper. Firstly, the THz spectra of seven samples with different concentrations under six different temperatures were collected. Secondly, smoothing and baseline correction methods were used to preprocess the spectra to study the better spectral pretreatment methods. To reduce the influence of the change of external temperature on the detection, the External Parameter Orthogonalization (EPO) and Generalized Least Squares Weighting (GLSW) methods were introduced to compensate the temperature change. These two aforementioned methods were compared with the method without temperature compensation, respectively. Then, Uninformative Variable Elimination (UVE), principal component analysis (PCA), successive projections algorithm (SPA) and Competitive Adaptive Reweighted Sampling (CARS) were used to explore the optimal terahertz band selection method. Finally, Partial Least Squares (PLS) and Least square support vector machine (LS-SVM) models were established, respectively, to realize the enhanced detection of benzoic acid additives. The LS-SVM model combined with SPA and GLSW has the best effect, with the R-p(2) of 0.9892, the RMSEP of 3.8412 x 10(-5) and the LOD of 1.158 x 10(-4) g/mL. The LOD of the model is nearly three times higher than that of the mixed global model. The results show that the GLSW method exhibited a better compensation effect on the THz spectra of benzoic acid solution samples at different temperatures. Therefore, this study can be applied to other temperature-compensated models and can also provide reference for the detection of samples with variable temperature.