三种名优白酒的GC-MS和ICP-OES分析及其鉴别模型研究

发布时间：2018-05-09 08:51

本文选题：白酒鉴别 + 模式识别　；参考：《成都理工大学》2017年硕士论文

【摘要】：白酒是中国人在社交、庆典、婚礼等活动中不可缺少的常见饮品,有五千多年的悠久历史和深厚的文化内涵,具有独特的风格特征。传统的白酒鉴别主要依靠品酒师的感官品评,情感、环境等主客观因素会影响品评准确性,而且白酒产量、数量庞大,感官品鉴专家数量相对不足,难以应对市场监督鉴别需求。本研究选择水井坊井台装,五粮液水晶装,红花郎10年三种具有代表性的名优白酒,采用GC-MS和ICP-OES测定白酒样品中的有机组分和无机元素,利用安捷伦Mass Profiler Professional,Matlab分析软件,结合主成分分析,聚类分析,线性判别分析,偏最小二乘判别分析,支持向量机等建模技术,建立三种名优白酒的分类预测模型,探索名优白酒的真伪鉴别方法。(1)优化了GC-MS测定白酒中有机成分的仪器条件,最佳的色谱条件为:载气流速1.0 mL/min,分流比5:1,升温程序为:起始温度40°C,保持5 min,5°C/min升至100°C,保持5 min,10°C/min升至190°C保持20 min。最佳的质谱条件为:电子轰击离子源(EI),电子能量70 eV;离子源温度230°C;扫描质量范围(m/z):35~500。在优化的GC-MS条件下,内标乙酸正戊酯保留时间的RSD(n=6)为0.1%,峰面积的RSD(n=6)为3.0%,方法精密度良好。(2)对ICP-OES测定白酒中Ca、Na、K、Mg、Cu、Fe、Mn、Zn、Sr等9种元素进行了方法学验证,各元素的标准曲线的相关系数介于0.998~1.000之间,加标回收率介于95%~110%,相对标准偏差(RSD,n=6)均低于4.7%,检出限低于0.005 mg/L,9个元素测定的线性好,准确度、精密度、检出限均符合分析化学痕量分析的要求。(3)用ICP-OES测定了175组白酒样品中的9种金属元素浓度,并分析了三种名优白酒的金属元素浓度特征。分析结果表明,同品系白酒的金属元素浓度分布相对集中,其差异小于品系间差异,三种名优白酒的金属元素浓度各具特征。五粮液水晶装样品中Ca、Mg浓度最低,其平均浓度分别低至2.10 mg/L和0.32mg/L。水井坊井台装样品中的Ca、Mg、Sr的浓度明显高于红花郎样品,K、Mn、Cu、Fe的浓度范围相近。红花郎10年、15年,青花郎20年三个品系样品的9种金属元素浓度差异小。(4)对16组五粮液水晶装与11组同厂、13组同产地白酒的GC-MS和ICP-OES数据的主成分分析,结果表明,所有五粮液水晶装样品聚集于一个小区域内,与同厂样品有部分交叉,说明16组五粮液水晶装样品的有机组分和无机金属浓度特征高度一致,且与部分同厂样品相似。基于GC-MS数据的聚类分析结果表明,16组五粮液水晶装分成了两部分,其中14组与2组同厂样品聚在一起,另外2组与1组宜宾样品单独聚在一起。五粮液水晶装的GC-MS和ICP-OES分析结果一致。(5)对20组水井坊井台装与18组同厂、23组同产地白酒样品的GC-MS和ICP-OES数据的主成分分析,结果表明,水井坊井台装样品的有机组分和无机金属特征具有一致性,与同厂样品有部分交叉,且与成都产的白酒有显著差异,水井坊井台装与成都的白酒样品的ICP-OES数据的差异不及GC-MS数据。聚类分析结果显示20组水井坊井台装样品分为四部分,分别与相似的同厂、同产地样品聚在一起。由于水井坊井台装样品的生产日期跨度较大,因此水井坊井台装样品的一致性不及五粮液水晶装。(6)对19组红花郎10年与20组同厂样品、11组同产地白酒样品的的GC-MS和ICP-OES数据的主成分分析结果表明,红花郎10年样品的有机组分和无机金属特征具有一致性,与红花郎15年和青花郎20年相似度较高,与其它古蔺产的酱香型白酒区分明显。基于GC-MS数据的聚类分析结果表明,19组红花郎10年样品中的13组聚在一起,其余6组与各自相似的样品聚在一起。聚类分析结果与主成分分析结果一致。红花郎10年与红花郎15年、青花郎20年同属于郎酒的年份酒系列,具有高度相似的生产原料、酿造工艺、发酵方式、勾兑方法,因此其有机组分和无机金属特征具有极高的相似度。(7)分别建立了基于GC-MS、ICP-OES和GC-MS-ICP-OES相结合的数据的三种名优白酒分类预测模型,并用未参与建模的样品进行验证。结果表明,结合名优品系白酒的GC-MS和ICP-OES数据建立的名优品系白酒分类预测模型比分别基于GC-MS和ICP-OES的模型有更高的交叉验证正确率和整体预测正确率,可以很好的对名优白酒进行区分,并对其它未知样品具有一定的鉴别能力。五粮液水晶装的W-Combin-LDA模型的交叉验证正确率和整体预测正确率分别为94.1%和94.6%,水井坊井台装的S-Combin-LR模型的交叉验证正确率和整体预测正确率分别为90.4%和99.1%,红花郎10年的L-Combin-SVM的交叉验证正确率和整体预测正确率分别为93.18%和95.2%。
[Abstract]:Liquor is a common drink of Chinese people in social, celebration, wedding and other activities. It has a long history and profound cultural connotation for more than five thousand years. It has unique style characteristics. The traditional liquor identification mainly depends on the sensory evaluation of the wine taster, emotion, environment and other subjective and objective factors will affect the accuracy of the evaluation, and the production of liquor. The quantity is huge, the number of sensory evaluation experts is relatively insufficient, it is difficult to cope with the market supervision and identification needs. This study selected three representative famous liquor of Shuijingfang well platform, Wuliangye crystal and Honghua Lang for 10 years. The GC-MS and ICP-OES were used to determine the organic and inorganic elements in the liquor samples, and the Agilent Mass Profiler Profes was used. Sional, Matlab analysis software, combined with principal component analysis, cluster analysis, linear discriminant analysis, partial least squares discriminant analysis, support vector machine and other modeling techniques, set up the classification prediction model of three kinds of famous liquor, and explore the true and false identification method of famous liquor. (1) optimize the instrument conditions for the determination of organic components in liquor by GC-MS, the best color The spectrum conditions are as follows: the air flow velocity of 1 mL/min and the shunt ratio 5:1, the starting temperature is 40 degree C, 5 min, 5 C/min to 100 degrees, 5 min, 10 degree C/min to 190 C to maintain 20 min.: electronic bombardment ion source (EI), electronic energy 70 eV; ion source temperature 230 degrees; scanning mass range Under MS conditions, RSD (n=6) of the retention time of n-amyl acetate was 0.1%, the peak area RSD (n=6) was 3%, and the precision was good. (2) the determination of 9 elements, such as Ca, Na, K, Mg, Cu, Fe, Mg, Cu, Ca, K, Mg, and so on, was verified by ICP-OES. The relative standard deviation (RSD, n=6) were lower than 4.7%, the detection limit was less than 0.005 mg/L, the determination of 9 elements was well linear, accuracy, precision, detection limit were all in line with the requirements of analytical chemical trace analysis. (3) the concentration of 9 metal elements in 175 groups of liquor samples was measured by ICP-OES, and the metal element concentration characteristics of three kinds of famous liquor were analyzed. The results show that the concentration distribution of metal elements in the same strain liquor is relatively concentrated, and the difference is less than the difference between lines. The concentration of metal elements in three kinds of famous liquor is different. The concentration of Ca and Mg in Wuliangye crystal samples is the lowest, the average concentration is as low as 2.10 mg/L, and the concentration of Ca, Mg, Sr in the 0.32mg/ L. Shuijingfang well stand samples is clear. The concentration range of K, Mn, Cu, Fe is similar to that of Honghua Lang. 10 years, 15 years, 15 years, and 9 kinds of metal elements in 20 years of Qinghua Lang. (4) the principal component analysis of 16 groups of Wuliangye crystals and 11 groups, the GC-MS and ICP-OES data of the 13 groups with the origin of Baijiu, and the results show that all the Wuliangye crystal samples are gathered together. In a small area, it is partially intersecting with the same factory samples. It shows that the 16 groups of Wuliangye crystal samples are highly consistent with the concentration characteristics of the organic and inorganic metals, and similar to some of the samples from the same factory. The results of cluster analysis based on GC-MS data show that 16 groups of Wuliangye crystals are divided into two parts, of which 14 groups are together with 2 groups in the same factory. Together, the other 2 groups were together with 1 groups of Yibin samples. The results of GC-MS and ICP-OES in Wuliangye crystal were consistent. (5) principal component analysis on the data of GC-MS and ICP-OES of 20 groups of Shuijingfang well sets and 18 groups, and the samples of white spirits from the same origin, the results showed that the features of the Shuijingfang well set samples had the characteristics of organic and inorganic metals. The ICP-OES data of the Shuijingfang well set and Chengdu liquor samples were not the same as that of the same factory samples. The difference of the ICP-OES data between the Shuijingfang well set and Chengdu liquor samples was less than that of the GC-MS data. The cluster analysis results showed that the 20 groups of Shuijingfang well set samples were divided into the same factory and together with the producing samples. The production date of the Shuijingfang well set sample is more span, so the consistency of the Shuijingfang well platform sample is not as good as the Wuliangye crystal. (6) the principal component analysis of the 19 groups of Honghua Lang 10 and 20 sets of samples, the GC-MS and ICP-OES data of the 11 groups of liquor samples from the same origin show that the organic and inorganic gold of the 10 year samples of Honghua Lang are of the organic composition and the inorganic gold. The characteristics are consistent with 15 years of Honghua Lang and 20 years of Qinghua Lang. It is distinct from other Gulin Maotai liquor. The results of cluster analysis based on GC-MS data show that 13 groups of the 19 groups of Honghua Lang are gathered together in 10 years, and the other 6 groups are together with their similar samples. The results are consistent. 10 years with Honghua Lang and Honghua Lang 15 years, Qinghua Lang 20 years belong to the wine series of Lang wine, with highly similar production materials, brewing technology, fermentation mode, blending method, so it has a very high similarity between the characteristics of organic and inorganic metals. (7) a combination of GC-MS, ICP-OES and GC-MS-ICP-OES was established, respectively. Three kinds of famous liquor classification prediction models were used, and the model samples were used to verify the model. The results showed that the combination of GC-MS and ICP-OES data, which was established by the famous liquor, had higher accuracy rate of cross validation and overall prediction accuracy than the model based on GC-MS and ICP-OES respectively. The W-Combin-LDA model of Wuliangye crystal has a correct rate of 94.1% and 94.6% respectively, and the correct rate of cross validation and the overall prediction accuracy of the S-Combin-LR model of Shuijingfang well set are 90, respectively. The cross validation accuracy and overall prediction accuracy of.4% and 99.1%, Honghua Lang 10 years were 93.18% and 95.2%. respectively, and L-Combin-SVM

【学位授予单位】：成都理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TS262.3;O657

【参考文献】