基于金属纳米复合材料修饰电极阵列的黄酒酒龄、品牌和地域鉴别

发布时间：2018-08-19 18:33

【摘要】：黄酒营养丰富,市场发展前景广阔,并享有"国酒"美誉,但因质量评价手段较少,导致诸如虚报酒龄、乱贴知名品牌标签和以次充好冒充地理标志产品等非法行为时常发生,严重侵犯了正规厂家利益和消费者权益。复合材料修饰电极因表面能够固定化学性质优良的纳米材料和聚合物等,克服了单一材料修饰电极和裸电极反应微弱、灵敏度低等缺点,近年来已在酒类酒龄、地域和品牌鉴别中有所应用。基于此,本文首次将自主研制的金属纳米复合材料修饰电极阵列用于黄酒质量真伪鉴别,结合模式识别方法成功实现了不同酒龄、品牌和地域黄酒的区分和预测,从而为净化黄酒市场、维护消费者和正规厂家的利益提供了一套全新可靠的解决方案。本文具体研究内容、方法和结论如下:(1)采用自制的PACBK/Au/GCE、PABSA/Au/GCE和PASP/Pt/GCE组成修饰玻碳电极阵列成功实现了对3年陈、5年陈、8年陈、10年陈、15年陈和20年陈6种酒龄绍兴古越龙山黄酒的区分和预测。本部分选取在黄酒陈酿过程中含量变化较大的三种呈味物质:维生素C(维生素,酸味)、酪氨酸(氨基酸,涩味)和葡萄糖(糖类,甜味),采用循环伏安法(CV)和电流时间法(i-t)等对应制备了 PACBK/Au/GCE、PABSA/Au/GCE和PASP/Pt/GCE三种聚合物/金属纳米复合材料修饰电极,并在优化pH、扫速和缓冲液浓度等检测条件下实现了对三种呈味物质在一系列浓度梯度溶液中的定量测定,通过对比三种物质在电极上的检测限和在黄酒中的含量,发现物质在黄酒中含量远大于检测限,这说明黄酒样品满足了修饰电极的检测条件,即证明了修饰电极鉴别黄酒的有效性。基于电极有效性的基础上,在6种酒龄黄酒样品中采用复频多电位阶跃法作为激发信号施加于电极阵列,选取响应电流信号曲线与时间轴所围区域的面积值作为特征值,结合主成分分析(PCA)、局部保留投影(LPP)、线性判别分析(LDA)和支持向量机(LSSVM、LIBSVM)等模式识别方法对黄酒酒龄进行区分和预测,PCA、LPP和LDA三种区分模型中,LDA区分效果最好,在二维图和三维图中,6种酒龄黄酒都能够明显分开;LSSVM和LIBSVM两种酒龄回归模型中,LIBSVM预测效果要优于LSSVM,特别是均方差更小。(2)采用自制的PACBK/Au/GCE、PABSA/Au/GCE和PGA/Cu/GCE组成修饰玻碳电极阵列成功实现了对3种古越龙山、3种塔牌和会稽山总计7种品牌绍兴黄酒的区分和预测。本部分选取在各品牌黄酒间含量差异相对较大的三种呈味物质:维生素C(维生素,酸味)、酪氨酸(氨基酸,涩味)和没食子酸(酚类,苦味),采用循环伏安法等电化学方法对应制备了 PACBK/Au/GCE、PABSA/Au/GCE和PGA/Cu/GCE三种聚合物/金属纳米复合材料修饰电极,在优化pH、扫速、富集电位和时间等检测条件下应用线性扫描伏安法(LSV)和差分脉冲伏安法(DPV)等实现了三种呈味物质在一系列浓度梯度中的定量测定,并将三种电极对维生素C、酪氨酸和没食子酸的检测限与三种呈味物质在黄酒中实际含量进行了比较,结果显示实际含量均远大于检测限,说明基于修饰电极阵列对黄酒品牌进行鉴定是可行的。之后在7种品牌黄酒中采用方波和梯形波两种多电位阶跃法施加于电极阵列得到响应电流信号曲线,选取电流曲线与时间轴所包围区域面积值作为特征值,结合PCA、LPP、LDA、LIBSVM、ELM(极限学习机)和BPNN(BP神经网络)等模式识别方法对黄酒品牌进行了区分和预测,区分模型显示PCA、LPP和LDA三种区分效果均存在3种塔牌黄酒类间距过小的问题,ELM区分效果也不太理想,但LIBSVM模型效果较佳,训练集和测试集区分正确率分别为100%和99.05%;预测模型显示ELM和LIBSVM预测效果一般,而BPNN效果较好,预测准确率达到了 97.14%。(3)采用自制的 SMWCNT/Au/GCE、PABSA/Au/GCE 和 PGA/Cu/GCE 组成修饰玻碳电极阵列成功实现了对江苏丹阳(镇江)、青岛即墨(青岛)、浙江汾湖(嘉兴)、浙江同康(台州)和古越龙山(绍兴)5种地域黄酒的区分和预测。本部分选取在各地域黄酒中比较有代表性的三种呈味物质:5'-GMP(添加剂,鲜味)、酪氨酸(氨基酸,涩味)和没食子酸(酚类,苦味),采用滴涂法和循环伏安法等制备了 SMWCNT/Au/GCE、PABSA/Au/GCE和PGA/Cu/GCE三种金属纳米复合材料修饰电极,在优化的最佳pH和扫速等条件下应用LSV和DPV等电化学方法实现了三种呈味物质在一系列浓度梯度溶液中的定量测定,并将三种物质在对应电极上的检测限与其在黄酒中的含量进行了比较,结果显示含量均远大于检测限,说明黄酒中的三种物质含量满足修饰电极响应条件,电极的有效性符合要求。之后在5种地域黄酒中采用复频多电位阶跃法施加于电极阵列获得响应电流信号曲线,选取电流曲线与时间轴所包围区域的面积值作为特征值,结合PCA、LPP、LDA、LIBSVM和ELM等模式识别方法对黄酒地域进行了区分和预测,区分结果显示PCA、LPP和LDA均存在浙江汾湖和浙江同康黄酒类间距过小以及古越龙山黄酒样本点较分散等问题,ELM和LIBSVM区分效果较好,训练集和预测集区分正确率都较高;预测结果显示ELM和LIBSVM两种回归模型表现较好,特别是决定系数R~2较大。
[Abstract]:Yellow rice wine has abundant nutrition, broad market prospects and enjoys the reputation of "national wine". However, due to the lack of quality evaluation methods, illegal behaviors such as false declaration of wine age, labeling of well-known brands and faking geographical indications as inferior products often occur, which seriously infringe the interests of regular manufacturers and consumers. Surface can immobilize nano-materials and polymers with good chemical properties, overcome the weakness and low sensitivity of single material modified electrode and bare electrode. In recent years, it has been used in wine age, region and brand identification. The identification of authenticity and falsity of yellow rice wine quality and the identification and prediction of different wine ages, brands and regions have been successfully realized by pattern recognition method, thus providing a new and reliable solution for purifying the yellow rice wine market and safeguarding the interests of consumers and regular producers. U/GCE, PABSA/Au/GCE and PASP/Pt/GCE modified glassy carbon electrode arrays were successfully used to distinguish and predict six kinds of Shaoxing Guyuelongshan yellow wine aged 3 years, 5 years, 8 years, 10 years, 15 years and 20 years. PACBK/Au/GCE, PABSA/Au/GCE and PASP/Pt/GCE polymer/metal nanocomposites modified electrodes were prepared by cyclic voltammetry (CV) and current-time method (i-t) with acid (amino acid, astringent taste) and glucose (sugar, sweet taste). The electrode was modified by PACBK/Au/GCE, PABSA/Au/GCE and PASP/Pt/GCE. Under the conditions of optimizing pH, sweeping rate and buffer concentration, the three flavoring substances were detected. Quantitative determination in a series of concentration gradient solutions showed that the content of the three substances in yellow rice wine was much higher than the detection limit by comparing the detection limit of the three substances on the electrode and the content in yellow rice wine. On this basis, the complex frequency multi-potential step method was applied to the electrode array as excitation signal in six kinds of rice wine samples. The response current signal curve and the area around the time axis were selected as eigenvalues, combined with principal component analysis (PCA), partial retention projection (LPP), linear discriminant analysis (LDA) and support vector machine (LSSVM, LIBSVM). Pattern recognition method was used to distinguish and predict the age of yellow rice wine. Among the three models, PCA, LPP and LDA, LDA was the best. In the two-dimensional and three-dimensional charts, six kinds of yellow rice wine could be separated obviously; LSSVM and LIBSVM were better than LSSVM in predicting the age of yellow rice wine, especially the mean square deviation was smaller. BK/Au/GCE, PABSA/Au/GCE, PGA/Cu/GCE and PGA/Cu/GCE modified glassy carbon electrode arrays were successfully used to distinguish and predict three kinds of Shaoxing rice wine from three kinds of Guyue Longshan, three kinds of Tapai and a total of seven brands of Jishan. In this part, three flavoring substances, vitamin C (vitamin, acid), tyrosine (amino acid), with relatively large content difference among the different brands of rice wine were selected. Polymer/metal nanocomposites modified electrodes PACBK/Au/GCE, PABSA/Au/GCE and PGA/Cu/GCE were prepared by cyclic voltammetry and gallic acid (phenols, bitters) respectively. Linear sweep voltammetry (LSV) and differential pulse voltammetry (DPV) were used to optimize pH, sweep rate, enrichment potential and time. The quantitative determination of three flavoring substances in a series of concentration gradients was realized by DPV. The detection limits of vitamin C, tyrosine and gallic acid were compared with the actual contents of three flavoring substances in yellow rice wine. The results showed that the actual contents were far greater than the detection limits, indicating that the brand of yellow rice wine was based on modified electrode array. The response current signal curve was obtained by applying square wave and trapezoidal wave multi-potential step method to the electrode array in seven brands of yellow rice wine. The current curve and the area around the time axis were selected as the eigenvalues, combined with PCA, LPP, LDA, LIBSVM, ELM (Extreme Learning Machine) and BPNN (BP Neural Network) etc. The discriminant model showed that there were three kinds of tower rice wine with too small spacing among PCA, LPP and LDA, and the ELM discriminant effect was not ideal, but the LIBSVM model had better effect. The correct discriminant rate of training set and test set was 100% and 99.05% respectively. SVM has a good prediction effect, but BPNN has a good prediction accuracy rate of 97.14%. (3) Using self-made SMWCNT / Au / GCE, PABSA / Au / GCE and PGA / Cu / GCE modified glassy carbon electrode arrays, we have successfully realized five regions: Jiangsu Suanyang (Zhenjiang), Qingdao Jimo (Qingdao), Zhejiang Fenhu (Jiaxing), Zhejiang Tongkang (Taizhou) and Guyue Longshan (Shaoxing). In this part, three typical flavoring substances, 5'-GMP (additive, delicious), tyrosine (amino acid, astringent taste) and gallic acid (phenolic, bitter taste), were selected to prepare SMWCNT/Au/GCE, PABSA/Au/GCE and PGA/Cu/GCE metal nanocomposites by trickling and cyclic voltammetry. The modified electrode was used to quantitatively determine three flavoring substances in a series of concentration gradient solutions by electrochemical methods such as LSV and DPV under the optimum pH and sweeping speed conditions. The detection limits of the three substances on the corresponding electrode were compared with the contents in rice wine. The results showed that the contents were far greater than the detection limits. The content of the three substances in rice wine satisfies the response condition of the modified electrode and the validity of the electrode meets the requirements.Then the response current signal curve was obtained by applying the complex frequency multi-potential step method to the electrode array in the five regional rice wine.The current curve and the area around the time axis were selected as the eigenvalues and combined with PCA, LPP, LDA, LIB as the eigenvalues. SVM and ELM pattern recognition methods were used to distinguish and predict the regions of yellow rice wine. The results showed that PCA, LPP and LDA had the problems of too small distance between Fenhu and Tongkang yellow rice wine in Zhejiang Province, and scattered sample points of Guyuelongshan yellow rice wine. ELM and LIBSVM had better distinguishing effect, and the training set and prediction set had higher accuracy. It shows that the two regression models of ELM and LIBSVM perform well, especially the determination coefficient R~2 is larger.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TS261.7

【相似文献】