本研究使用拉曼光谱分析技术采集不同产地和不同酒龄的黄酒样品指纹信息，对比判别分析（DA）和最小二乘支持向量机（LS-SVM）所建黄酒品质快速模型性能，确定最优模型以实现快速准确地评价黄酒品质。本研究在全波段范围利用主成分分析对拉曼光谱数据降维，计算降维谱图间马氏距离，基于ward’s算法建立判别分析模型；将全波段范围作为最小二乘支持向量机的输入量，选择出能较好处理非线性问题的RBF为核函数，同时采用交互验证方式优化RBF核函数参数，基于优化RBF核函数，建立最小二乘支持向量机鉴别模型。研究结果表明：拉曼光谱结合最小二乘支持向量机鉴别模型对黄酒产地和酒龄的鉴别正确率均为100%；拉曼光谱结合判别分析鉴别模型对嘉善、绍兴和上海黄酒的鉴别正确率分别为100%、80%和80%，对黄酒酒龄的鉴别正确率均为100%；最小二乘支持向量机模型性能优于判别分析模型。拉曼光谱结合化学计量学方法可快速、准确评价黄酒品质。

In this study, Raman spectra analysis technology was used to rapidly and accurately evaluate various Chinese rice wines. Fingerprint information was collected from Chinese rice wine samples from different sources and of different ages, followed by a comparison of the prediction performances of quality evaluation models established by discriminant analysis (DA) and least-squares support vector machine (LS-SVM) in order to select the optimal model. Principal component analysis was used in the entire wavelength range to reduce the dimension of Raman spectroscopy data. The Mahalanobis distance between reduced-dimension spectra was calculated and discrimination analysis models were established based on ward’s calculation. The entire wavelength range was used as the input of LS-SVM, to select a kernel radial basis function (RBF) that could alleviate non-linear issues. The kernel RBF was optimized at the same time via interactive authentication. The LS-SVM discrimination model was established based on the optimized kernel RBF function. The result showed that Raman spectroscopy coupled with LS-SVM could 100% accurately identify both production origin and wine age. The correction rates of Raman spectroscopy coupled with DA to discriminate Chinese rice wines from among Jiashan, Shaoxing, and Shanghai origins were 100%, 80%, and 80%, respectively, while that to determine wine age was 100% for all samples. Thus, the LX-SVM model was superior to the DA model. It was concluded that Raman spectroscopy in combination with chemometrics could be used for the rapid and accurate evaluation of Chinese rice wines.

国家自然科学基金资助项目（21105065）；全国优秀博士学位论文作者专项资金资助项目（201059）