AbstractNear infrared spectroscopy analysis technology, as an emerging analytical technique, has the advantages of no pre-treatment, fast detection speed, no damage to samples, no use of solvents, and the ability to simultaneously detect multiple components. It has been widely used in fields such as food, pharmaceuticals, agriculture, petrochemicals, and the environment.

Based on the micro near-infrared spectrometer developed by the Microsystem Center of Chongqing University, this paper studies its application in the analysis and classification of alcohol components, which is of great significance for achieving rapid detection of alcohol. Based on the near-infrared spectra of Baijiu collected by the micro near-infrared spectrometer, this paper adopts chemometrics methods such as principal component analysis, partial zui small binary method, zui small binary support vector machine, and combines relevant pattern recognition technology to establish qualitative and quantitative analysis models for liquor detection, so as to achieve the analysis and classification of liquor components. The main research contents are as follows: ① Built a near-infrared spectroscopy analysis experimental platform with a miniature near-infrared spectrometer as the core, and developed a feasible overall experimental plan. Alcohol solution and six kinds of Baijiu in the market were selected as experimental objects, and near-infrared spectral data were collected. In order to improve the stability and applicability of subsequent models, corresponding preprocessing algorithms were studied based on the characteristics of the collected spectral signals. ② In response to the large amount of data in near-infrared spectroscopy, this paper uses principal component analysis and partial least squares algorithm to extract feature information from spectral data, and applies pattern recognition methods (such as SIMCA, Mahalanobis distance method) to first establish a linear qualitative analysis model. Considering the possible nonlinear relationship between spectra and quality parameters, a least squares support vector machine was used to establish a nonlinear qualitative analysis model.

The results indicate that the least squares support vector machine has a high classification accuracy. ③ The quantitative analysis model of Baijiu alcohol content based on the spectral data of alcohol solution was studied. The results showed that the correction standard deviation of the quantitative analysis model established by the least squares support vector machine was 0.0040, and the prediction standard deviation was 0.2912, which was superior to the principal component analysis model and the biased least squares model.