摘要
为构建白酒基酒中甲酸的近红外预测模型,达到快速预测甲酸含量的目的,采用气相色谱法测定白酒基酒中甲酸的含量作为近红外预测模型的输入量,并通过分析白酒基酒样品的近红外光谱图,比较不同光谱数据预处理方法对建模效果的影响。结果表明,白酒基酒中甲酸对近红外有特异吸收,最佳预处理方法与最优波段为:一阶导数+减去一条直线,谱区区间选择在5813～5740 cm^(-1)和8482～8260 cm^(-1)。甲酸校正集样品的化学值与近红外的预测值的决定系数为99.25%,交叉验证均方根误差为4.26 mg/100 m L,相对分析误差为11.6;进一步对甲酸进行验证和评价,甲酸验证集样品的决定系数为97.6%,均方根误差为5.53 mg/100 mL,相对分析误差为9.68。结果表明,所建立的模型效果良好,能满足白酒生产中甲酸的快速检测需求。
The near infrared spectrum-based prediction model of formic acid in base Baijiu had been constructed for rapid prediction of formic acid content in Baijiu. The content of formic acid in base Baijiu was determined by GC and then used as the input values for the model. Through the analysis of the near infrared spectrums of base Baijiu samples, the effects of different pretreatment methods of spectral data on the modeling were compared. The results showed that, formic acid in base Baijiu exhibited specific near infrared absorption, and the best pretreatment method was first derivative combined with deduction of a straight line, the best spectral ranges were 5813 cm-1 to 5740 cm-1 and 8482 cm^-1 to 8260 cm^-1. A calibration model had been established using partial least square regression and the internal cross-validation method. Through the optimization of the model, the coefficients of determination (R2) between chemically measured values and NIR predicted values in calibration set for formic acid were 99.25 %, the cross-validation root mean square deviation was 4.26 mg/100 mL, and relative error of analysis was 11.6. The prediction model for formic acid was validated, the R2 values for formic acid in validation set was 97.6 %, root mean square deviation was 5.53 mg/100 mL, and the relative error of analysis was 9.68. The constructed prediction model could meet rapid measurement requirements of formic acid in Baijiu.
作者
陈林
庹先国
张贵宇
孙安
田万春
CHEN Lin;TUO Xianguo;ZHANG Guiyu;SUN An;TIAN Wanchun(School of Automation & Information Engineering,Sichuan University of Science & Engineering,Zigong,Sichuan 643000;School of Information Engineering,Southwest University of Science and Technology,Mianyang,Sichuan 621010;Sichuan Key Lab of Artificial Intelligence,Zigong,Sichuan 643000,China)
出处
《酿酒科技》
2019年第4期30-35,共6页
Liquor-Making Science & Technology
基金
四川省科技计划重点研发项目(17ZDYF3058)
关键词
近红外光谱
白酒基酒
甲酸
快速检测
near infrared spectroscopy
base Baijiu
formic acid
rapid measurement
