食品科学 ›› 2018, Vol. 39 ›› Issue (2): 281-286.doi: 10.7506/spkx1002-6630-201802044

• 安全检测 • 上一篇    下一篇

白酒基酒中典型醇的近红外预测模型构建

刘建学1,2,杨国迪1,韩四海1,2,李璇1,2,李佩艳1,2,徐宝成1,2   

  1. (1.河南科技大学食品与生物工程学院,河南?洛阳 471023;2.河南省食品原料工程技术研究中心,河南?洛阳 471023)
  • 出版日期:2018-01-25 发布日期:2018-01-05
  • 作者简介:刘建学,杨国迪,韩四海,李璇,李佩艳,徐宝成
  • 基金资助:
    国家自然科学基金面上项目(31471658);河南省重点科技攻关项目(152102110025)

Prediction Model for Typical Alcohols in Base Liquor Based on Near Infrared Spectroscopy

LIU Jianxue1,2, YANG Guodi1, HAN Sihai1,2, LI Xuan1,2, LI Peiyan1,2, XU Baocheng1,2   

  1. (1. College of Food and Bioengineering, Henan University of Science and Technology, Luoyang 471023, China;2. Henan Engineering Research Center of Food Material, Luoyang 471023, China)
  • Online:2018-01-25 Published:2018-01-05

摘要: 采用气相色谱法测定白酒基酒中的正丙醇、正丁醇、正戊醇和异戊醇的含量作为建立近红外预测模型的化学值,将近红外光谱图结合偏最小二乘法和内部交互验证法建立基酒中典型醇的快速检测模型,并进一步优化模型。确定了最优光谱预处理方法和最佳谱区,正丙醇、正丁醇、正戊醇和异戊醇的校正集样品的真实值与近红外预测值的决定系数(R2)分别为0.952、0.981、0.963和0.981,内部交互验证均方根误差分别为0.27、0.49、0.101?mg/100?mL和0.67?mg/100?mL;验证集的决定系数(R2)分别为0.947、0.980、0.928和0.952,预测均方根误差分别为0.40、0.81、0.49?mg/100?mL和1.35?mg/100?mL。结果表明建立的典型醇近红外快速检测模型的准确度、稳定性及预测性能均呈现良好,为白酒基酒的醇类物质品质分析方法研究提供了新的思路。

关键词: 近红外光谱, 白酒基酒, 典型醇, 偏最小二乘法

Abstract: In this paper, the contents of n-propanol, n-butanol, amyl alcohol and isoamyl alcohol in base liquor were determined by gas chromatography and used as chemical values for the establishment of calibration and validation?sets for a rapid predictive model based on near infrared spectroscopy (NIR) to measure typical alcohols in base liquor. The model was developed using partial least squares (PLS) regression with internal cross validation and optimized. The optimal spectral pretreatment method and the optimal spectral region were determined. The coefficients of determination (R2) between the actual and the NIR predicted values of n-propanol, n-butanol, amyl alcohol and isoamyl alcohol for the calibration set were 0.952, 0.981, 0.963 and 0.981, and the root mean square error of cross-validation (RMSECV) were 0.27, 0.49, 0.101 and 0.67 mg/100 mL, respectively; the R2 values for the validation set were 0.947, 0.980, 0.928 and 0.952, and RMSEPs were 0.40, 0.81, 0.49 and 1.35 mg/100 mL, respectively. Results showed that the predictive model exhibited good accuracy, stability and prediction performance and could provide a new approach for the analysis of alcohols in base liquor.

Key words: near infrared spectroscopy, base liquor, typical alcohols, partial least squares

中图分类号: