为研究馥郁香型白酒瓶贮年份的真实性,通过气相色谱直接进样内标测定馥郁香型白酒样品中48 种主要风味组分的含量,结合偏最小二乘回归(partial least squares regression,PLSR)多元统计学方法,建立了馥郁香型成品白酒瓶贮年份PLSR预测模型。该模型相关系数(R2)为0.953 4,校正标准偏差(root mean square error of calibration,RMSEC)为0.215 9,预测标准偏差(root mean square error of prediction,RMSEP)为0.8384,相对分析误差(relative percent deviation,RPD)为4.631 3,模型回归拟合效果好,预测精度高。采用jack-knife方法对PLSR模型回归系数进行显著性检验,筛选出对预测模型有显著影响的特征标记物,发现己酸乙酯、乙醛、辛酸乙酯和正丁醇与馥郁香型白酒瓶贮年份显著相关,为馥郁香型白酒瓶贮年份PLSR预测模型的特征标记物。
李宜聪
,
樊双喜
,
吉鑫
,
李春扬
,
辛鹏
,
谌松强
,
张泽辉
,
刘凤晖
,
钟其顶
. 偏最小二乘回归法筛选馥郁香型白酒瓶贮年份特征标记物[J]. 食品与发酵工业, 2020
, 46(22)
: 14
-20
.
DOI: 10.13995/j.cnki.11-1802/ts.024113
Direct injection gas chromatography technology was used to quantify the 48 characteristic flavor components in fragrant Baijiu in order to authenticate the bottled storage year of fragrant Baijiu.Partial least squares regression (PLSR) combined with multivariate statistical methods was used to establish the prediction model of bottled storage year of fragrant Baijiu.The R2 of the model was 0.953 4,the root mean square error of calibration (RMSEC) was 0.215 9,the root mean square error of prediction (RMSEP) was 0.838 4,and the relative percent deviation (RPD) was 4.631 3.The research results showed that both the regression fitting effect and the prediction accuracy of the model were good.The jack-knife method was used to test the significance of the regression coefficient of the PLSR model and screen out the characteristic markers of bottled storage year of fragrant Baijiu.The research results showed that ethyl hexanoate,acetaldehyde,ethyl caprylate and normal butanol were significantly correlated with the bottled storage year of fragrant Baijiu.
[1] 刘建新,刘晓方,刘凤晖,等.酒鬼酒馥郁香型风格的研究[J].酿酒科技,2005(10):98-104.
[2] 常宇桐,罗云敬,钱承敬,等.全二维气质技术解析馥郁香型白酒风味与质量品评关系[J].食品科学技术学报,2019,37(6):64-70;93.
[3] 徐占成.中国白酒贮存年份科学鉴别技术的研究——"挥发系数鉴别年份酒的方法"[J].酿酒科技,2009(1):96-98.
[4] 庄名扬.谈谈年份酒与鉴别方法[J].四川食品与发酵,2008,44(3):28-31.
[5] 杨涛,李国友,庄名扬.中国白酒"年份酒"鉴别方法的研究[J].科学实验,2008,35(5):33-38.
[6] 王国祥.基于Raman光谱和支持向量机回归的白酒年份鉴别方法研究[D].南京:南京财经大学,2015.
[7] 徐瑞煜.浓香型白酒的荧光光谱特性及其鉴别方法研究[D].无锡:江南大学,2016.
[8] 李彪.基于指纹图谱的白酒类别定性研究[D].武汉:武汉轻工大学,2015.
[9] 蔡心尧,尹建军,胡国栋.毛细管柱直接进样法测定白酒香味组分的研究[J].色谱,1997,15(5):367-371.
[10] R CORE TEAM.R:A language and environment for statistical computing.R foundation for statistical computing.Vienna,Austria[DB/OL] https://www.R-project.org/,2020.
[11] 葛翔宇,丁建丽,王敬哲,等.基于竞争适应重加权采样算法耦合机器学习的土壤含水量估算[J].光学学报,2018,38(10):393-400.
[12] 齐琛,方秋莲.偏最小二乘建模在R软件中的实现及实证分析[J].数学理论与应用,2013,33(2):103-111.
[13] 蒋辉,邓伟民,陈晓青.基于Pearson系数与多元核支持向量分类的葡萄酒分析[J].农业机械学报,2014,45(1):203-208.
[14] 林万福,李福玉,李振林.谈乙醛、乙缩醛在白酒中的含量及其量比关系[J].科学实验,2002,29(3):42-43.
[15] 周娴,熊旺平,周翔,等.基于R语言的PLS在中药数据处理的研究与分析[J].科技传播,2015,7(24):200-201;213.
[16] 张华东,阮陆宁.偏最小二乘回归在R软件中的实现及其优缺点剖析[J].科技广场,2015(11):12-17.
[17] GUILLÉN D A,PALMA M,NATERA R,et al.Determination of the age of sherry wines by regression techniques using routine parameters and phenolic and volatile compounds[J].J Agric Food Chem,2005,53(7):2 412-2 417.
[18] ZHY Y W,CHEN X Y,WANG S M,et al.Simultaneous measurement of contents of liquirtin and glyeyrrhizic acid in liquorice based on near infrared spectroscopy[J].Spectrochim Acta A,2018,196:209-214.
[19] ZHENG hong,LU Hongfei.A least-squares support vector machine (LS-SUM) based on fractal analysis and CILLAB parameters for the detection of browning degree on mango (Mangifera indica L.)[J].Computers and Electronics in Agriculture,2012,83:47-51.
[20] GHAEDI M,GHAEDI A M,HOSSAINPOUR M,et al.Least square-support vector (LS-SVM) method for modeling of methylene blue dye adsorption using copper oxide loaded on activated carbon:Kinetic and isotherm study[J].Journal of Industrial and Engineering Chemistry,2014,20(4):1 641-1 649.
[21] 常宇桐,罗云敬,钱承敬,等.顶空固相微萃取-气相色谱质谱法测定馥郁香型白酒中的挥发性香气成分[J].食品安全质量检测学报,2018,9(17):4 618-4 627.
[22] 程娇娇.不同年份赊店酒香气模型的建立及其对比分析[D].新乡:河南科技学院,2018.
[23] 朱梦旭,范文来,徐岩.我国白酒蒸馏过程及原酒、白酒中乙醛的研究[J].食品与发酵工业,2016,42(4):6-11.
[24] 黄跃勇,张国强,孙靖茹.新蒸馏白酒贮存过程中乙醛和乙缩醛的变化[J].酿酒,2009,36(3):25-26.
