为实现酱香型白酒基酒关键指标的快速检测，该研究以气相色谱法、滴定法为参考方法，利用近红外光谱技术结合偏最小二乘算法建立了酱香型白酒中7种关键成分的分析模型。通过间隔偏最小二乘法筛选出各物质的特征波段，并在12种单一或组合的方法中筛选出最优谱图处理方式。优化后的乙酸、丙酸、乙酸乙酯、乳酸乙酯、糠醛、乙酸异戊酯、总酸模型交叉验证决定系数R2Val为0.8542~0.9638，交叉验证均方根误差RMSECV为0.0038~0.5158；外部验证预测标准偏差SEP为0.0028~0.4785；RPD为2.41~6.43，模型预测准确性与稳健性良好。该研究表明近红外方法可快速检测酱香型白酒基酒中乙酸、丙酸、乙酸乙酯、乳酸乙酯、糠醛、乙酸异戊酯、总酸含量，极大提升检测效率，实现生产过程中的快速反馈，并为在线近红外光谱技术在智能酿造中的应用提供了方法基础。

In order to achieve the rapid detection of the key indexes of Jiangxiangxing Baijiu, analytical models of 7 key components in base liquor were established by near infrared spectroscopy (NIR) combined with partial least squares algorithm, with gas chromatography and titration as the reference methods. The characteristic bands of each substance were screened by interval partial least squares, and the optimal pretreatments were selected from 12 methods (alone or in combination). The optimized models of acetic acid, propionic acid, ethyl acetate, ethyl lactate, furfural, isoamyl acetate, and total acid were good, which R2Val were 0.8542~0.9638, and RMSECV were 0.0038~0.5158. External validation was executed, which SEP were 0.0028~0.4785, and the RPD were 2.41~6.43. It is demonstrated that the NIR method could rapidly detect the contents of acetic acid, propionic acid, ethyl acetate, ethyl lactate, furfural, isoamyl acetate, and total acid in Jiangxiangxing base liquor with good accuracy, robustness, and predictive performance. This method could greatly improve the efficiency of detection, realize rapid feedback in the production process, and provide a basis for the application of online NIR in intelligent brewing.

贵州省科技计划项目(黔科合成果[2023]一般149)；贵州省科技计划项目(黔科合成果[2023]一般150)；贵州省工信厅发展专项资金科技创新项目(202209)