白酒基酒等级的准确评判对白酒的质量控制至关重要, 是白酒分级储存和勾兑的重要依据。 目前白酒基酒分级是在生产车间班组工人初步分级基础上, 通过专业评酒人员的人工感官审评结合气相色谱法测定主要酯类物质含量最终确定白酒基酒的等级。 人工感官审评和气相色谱分析方法繁琐, 耗时费力, 很难实现实时快速检测。 红外光谱技术具有快速分析、 无损检测及灵敏度高和重现性好等优点, 在食品品质检测领域得到广泛应用。 运用傅里叶变换红外光谱（FTIR）和衰减全反射（ATR）技术结合化学计量学方法对不同等级白酒基酒及其主要酯类化合物进行快速定性和定量分析。 采用线性判别（LDA）和误差反向传播人工神经网络（BPANN）分析模型对不同等级白酒基酒进行判别, 线性判别分析训练集和测试集总体识别率均达到100%, BPANN分析训练集和测试集总体识别率均在95%以上。 采用区间偏最小二乘法（siPLS）对四种主要酯类化合物进行定量分析, 己酸乙酯、 乳酸乙酯、 乙酸乙酯和丁酸乙酯含量的训练集模型相关系数分别为0.986 4, 0.991 5, 0.970 2和0.951 4, 测试集模型的相关系数分别为0.982 4, 0.961 9, 0.905 2和0.808 0。 结果表明, 傅里叶变换红外光谱技术结合化学计量学方法能有效实现不同等级白酒基酒的准确判别, 同时基酒中主要酯类化合物的快速检测定量模型效果良好, 能满足白酒生产中的分析检测要求, 为白酒基酒等级的快速判别提供一种客观而准确的分析方法, 有效提升白酒的智能化生产水平。

The judgment of base liquor grades is of vital importance in quality control, it’s also an important basis for liquor classified storage and blending. At present, the grades of base liquor are identified by the artificial sensory evaluation of liquor assessment professionals combined with the determination of the main esters by gas chromatography based on the preliminary classification of the workers in workshop according to their experience. However, the artificial sensory evaluation and gas chromatography are cumbersome and time-consuming, which cannot provide real-time information and rapid detection. In recent years, infrared spectroscopy has been widely used in the field of food identification because of its quick analysis, nondestructive testing, high sensitivity and good reproducibility. In this paper, the Fourier transform infrared (FTIR) spectroscopy method combined with attenuated total reflectance (ATR) were applied for the effective qualitative and quantitative analysis of base liquor. FTIR-ATR spectrum in the mid-IR region was used for classifying different grades of base liquor and quantitative analysis of the main esters combined with chemometrics. The results showed that the FTIR-ATR combined with liner discriminant analysis (LDA) could distinguish four different grades of base liquor. The accuracy of the training set and test set both were 100%. The FTIR-ATR combined with back propagation artificial neural network (BPANN) could also achieve an accurate determination of different grades of base liquor effectively, and the correct recognition rate of the training set and test set both were over 95%. The results showed a comparatively good recognition performance of the two methods. The FTIR-ATR combined with synergy interval partial least squares (siPLS) models for predicting the content of the ester compounds were realized with good results. The correlation coefficients of the training set of the ethyl hexanoate, ethyl lactate, ethyl acetate, and ethyl butyrate in the siPLS model were 0.986 4, 0.991 5, 0.970 2 and 0.951 4, respectively. And the correlation coefficients of the test set were 0.982 4, 0.961 9, 0.905 2, and 0.808 0, respectively. These results corroborate the hypothesis that the FTIR-ATR can effectively achieve accurate determination of different grades of base liquor combined with chemometrics, and the quantitative model for rapidly detecting the main esters in base liquor is good, which can meet testing requirements during liquor production. FTIR-ATR combined with chemometrics provide a fast and accurate method for the determination of the grades of base liquor which can effectively enhance the level of intelligent production of liquor.