摘要
采用近红外光谱(NIRS)分析技术对白酒基酒中的总酯含量进行定量分析,通过偏最小二乘法(PLS)法建立分析模型,同时后向间隔偏最小二乘法(BiPLS)对整个谱区进行光谱特征波段筛选。用决定系数(R2)、校正均方根误差(RMSEC)以及预测均方根误差(RMSEP)对模型进行评价。结果表明:特征波段筛选能够对基酒总酯模型起到显著的优化作用,模型的决定系数R2从0.484提升至0.937,RMSEC及RMSEP值分别从0.490、0.476降低至0.172和0.177,在减少模型复杂程度的同时,有效地提高了模型的稳定性与准确度,经过基酒盲样验证,说明波段优化所建立的模型有较为准确的预测结果。
The total ester content in base liquor of Baijiu(Chinese liquor)was quantitatively analyzed by near-infrared spectroscopy(NIRS),and the analysis model was established by partial least square(PLS)method.Meanwhile,the spectral characteristic bands in the whole spectral region were screened by the backward interval partial least squares method(BIPLS).The model was evaluated by determination coefficient(R2),root mean square error of correction(RMSEC)and root mean square error of prediction(RMSEP).The results showed that the characteristic of band screening could significantly optimize the total ester model of base liquor,the R2 of model increased from 0.484 to 0.937,the RMSEC and RMSEP value decreased to 0.172 and 0.177 from 0.490 and 0.476,respectively,while reducing model complexity,the stability and accuracy of the model was effectively improved.Through blind validation of base liquor,results showed that the optimized model had more accurate prediction results of the proposed model.
作者
高畅
张宇飞
辛颖
李艳敏
魏金旺
朱婷婷
李子文
孙海波
GAO Chang;ZHANG Yufei;XIN Ying;LI Yanmin;WEI Jinwang;ZHU Tingting;LI Ziwen;SUN Haibo(Beijing Shunxin Agriculture Co.,Ltd.,Niulanshan Distillery,Beijing 101301,China;China National Research Institute of Food&Fermentation Industries Corporation,Beijing 100015,China)
出处
《中国酿造》
CAS
北大核心
2021年第4期155-158,共4页
China Brewing
基金
国家重点研发计划项目(2018YFE0196600)。
关键词
白酒
基酒
近红外光谱技术
定量分析
波段筛选
Baijiu
basic liquor
near-infrared spectroscopy
quantitative analysis
wave band screening
