掺假蜂蜜的鉴别方法研究

发布时间：2018-01-25 11:10

本文关键词： 蜂蜜掺假识别近红外流变二维相关分析主成分分析神经网络　出处：《南京农业大学》2016年博士论文　论文类型：学位论文

【摘要】：我国作为世界养蜂大国,其蜂蜜生产、消费、出口都位居世界前列。但蜂蜜品质,尤其是屡禁不止的蜂蜜掺假现象一直制约着中国蜂业的健康发展。有必要大力发展蜂蜜品质及掺假鉴别技术,尤其是廉价便捷的现代无损鉴别技术。本文研究掺假蜂蜜的鉴别方法,深入探究近红外(NIR)、流变(RHE)分析方法及其融合在掺假蜂蜜识别上的应用,可以推进我国蜂蜜检测技术发展,维护百姓食品安全,为中国蜂业健康发展及增强国产蜂蜜竞争力作出贡献。本研究丰富了真假蜂蜜在NIR和RHE分析方法上的基础数据,所实现的多种分析方法下的掺假蜂蜜二维(2D)指纹谱图不仅可以对真假蜂蜜进行模式识别,还为掺假蜂蜜提供了直观上的指纹谱图,为多维度鉴别蜂蜜掺假提供了技术支持。本论文以拓展传统的NIR和RHE方法的分析维度,实现多种分析方法下的掺假蜂蜜二维指纹谱图,并建模识别掺假蜂蜜为基本目的。研究揭示了掺假蜂蜜的NIR和RHE特性,包括温度、水分、掺假对蜂蜜相关特性的影响。在成功实现蜂蜜的NIR和RHE检测方法及近红外-流变(NIR-RHE)融合技术的二维指纹谱图同时对掺假蜂蜜进行了建模识别的比较研究。主要研究内容和结论如下:1、运用主成分分析和二维相关分析相结合的方法研究了温度、水、掺假对蜂蜜短波NIR光谱影响,给出了脱水蜜和掺水蜜的二维-近红外(2D-NIR)指纹谱图,展示出水分在蜂蜜中的光学指纹。在NIR及2D-NIR谱图的特征提取,相关分析和主成分分析基础上,对不同类型样本用BP-神经网络进行了真假蜂蜜模式识别和掺假蜂蜜掺假程度预测的数学建模。基本结论为:(1)利用廉价便捷的短波近红外技术实现蜂蜜掺假识别在技术上具有可行性。难以区分的蜂蜜差异可在温度扰动下被放大。2D-NIR同步相关谱图可揭示微扰中特征波段之间的关联变化,展示不同蜂蜜的光学指纹,从直观上以指纹谱图方式展现蜂蜜差异。(2)2D-NIR同步相关谱图可通过升温过程中吸收峰峰位和峰强的改变放大水合差异,并呈现出清晰的光学指纹,据此可以区分不同脱水和掺水程度蜂蜜。(3)基于近红外分析和BP神经网络的真假蜂蜜模式识别效果好,掺假蜂蜜掺假程度预测则绩效不高。2、在实验基础上确立了真假蜂蜜样本的流变模型和黏度变化经验公式,并对黏度经验公式中的两常数:温度活化能和水分活化能予以理论解释。运用二维-流变(2D-RHE)谱图新方法研究了掺假对蜂蜜的影响。在RHE特征及2D-RHE谱图特征提取,相关分析和主成分分析基础上,用BP神经网络对不同输入变量进行了真假蜂蜜模式识别和掺假蜂蜜掺假程度预测的数学建模。基本结论为:(1)蜂蜜掺假未能改变蜂蜜的Newton体和无触变的流变特性,也未能改变黏温、黏水及黏度随温度水分两者的指数衰减规律。常数U0和活化能Ea既随原蜜样本也随掺假度不同而不同,但不随掺假度单调变化。(2)温度活化能和水分活化能可通过度量流体有效流动的阈能差异,来折射流体内部结构组成的微观差异。(3 )蜂蜜的2D-RHE谱图能反映温度微扰中各样本在外力剪切下的细节差异,可直观展示温度微扰下剪切率递增情况下表观黏度变化情况。(4)基于流变分析和BP神经网络的真假蜂蜜模式识别可行,掺假蜂蜜掺假程度预测则绩效低。3、运用二维近红外-流变(2D NIR-RHE)融合相关谱图新方法,揭示温度扰动下NIR和RHE两分析方法下特征的关联变化。对近红外-流变(NIR-RHE)和2D NIR-RHE融合特征建立了真假蜂蜜模式识别和掺假蜂蜜掺假程度预测的数学模型,探讨了多信息融合技术进行蜂蜜掺假识别的可能性。基本结论为:(1)蜂蜜的2D NIR-RHE谱图指纹特征明显,能直观展示温度微扰下NIR和RHE两种分析方法对应特征的关联变化,可用来反映掺假样本细节特征。(2)基于融合技术和BP神经网络的真假蜂蜜模式识别可行,掺假蜂蜜掺假程度预测绩效不高。4、提出引进包括模型结果范围、平均值、50%阈值和90%阈值这些统计参数在内的网络绩效综合评判方法。比较研究发现:蜂蜜真假的模式识别以近红外方法最优,总识别率平均值可达92.06%,50%阈值93.9%,90%阈值83.3%;流变方法和融合方法总识别率均值可达83%,50%阈值85.2%,90%阈值81.5%。二维分析则以2D NIR-RHE方法略胜,其总识别率均值可达83.7%,50%阈值85.2%,90%阈值 79.6%。
[Abstract]:China as the world's largest country of beekeeping, its honey production, consumption, exports are among the highest in the world. But the quality of honey, especially honey adulteration phenomenon repeated has been restricting the healthy development of China apiculture. It is necessary to vigorously develop and adulteration of honey quality technology, especially modern non-destructive identification technology of cheap and convenient identification method. This paper explores the adulteration of honey, near infrared (NIR), rheological (RHE) analysis method and its application in the identification of the fusion of adulterated honey, can promote the development of detection technology of Chinese honey, people maintain food safety, contribute to the healthy development and enhance the competitiveness of domestic honey bee China. This study enriches the basic data true honey in the NIR and RHE analysis methods, a variety of the two-dimensional analysis method under the adulteration of honey (2D) fingerprint can not only be used in pattern recognition of the true and false honey Also, provide a visual on fingerprint for the adulteration of honey, which provides technical support for multi dimension detecting honey adulteration. Analysis of this paper to expand the dimension NIR and the traditional RHE method, to achieve a variety of fingerprint analysis method under the adulteration of honey 2D spectra, and identify the adulteration of honey for the basic purpose of modeling. The study reveals the NIR and RHE properties, adulteration of honey including temperature, moisture, effect of honey adulteration related characteristics. After the successful implementation of NIR and RHE for detection of honey and the near infrared - rheological (NIR-RHE) 2-D fingerprint fusion technology spectrum also makes a comparative study on identification of adulteration of honey modeling. The main research contents and conclusions are as follows: 1, using the method of principal component analysis and two-dimensional correlation analysis on the combination of temperature, water, influence of honey adulteration of HF NIR spectra, given honey and honey water dehydration by two dimensional infrared (2D-NIR) refers to Pattern spectrum shows water optical fingerprint in honey. In the NIR and 2D-NIR spectrum feature extraction, correlation analysis and principal component analysis on the basis of different types of samples by BP- neural network for pattern recognition and false honey adulteration of honey adulteration level prediction mathematical modeling. The main conclusions are as follows: (1) using cheap and convenient short near infrared technology honey adulteration identification is feasible in technology. It is difficult to distinguish the difference of honey in the temperature perturbations are amplified.2D-NIR synchronous spectrum can reveal the relation between the characteristics of the perturbation in band change, show the optical fingerprint of different kinds of honey, from the visual to fingerprint way honey. (2) 2D-NIR synchronous amplification hydration difference peak position and peak intensity of the absorption spectra can be through the heating process, and showing a clear optical fingerprint, which can distinguish the different off Water and water level of honey. (3) based on the true and false pattern recognition results of near infrared analysis and BP neural network, the adulteration of honey adulteration to predict the extent is not high performance.2, on the basis of experiment established the rheological model and the viscosity change of empirical formula of true and false honey samples, and the viscosity in the empirical formula of two constants: the activation energy and water temperature can be activated theoretically. By using two-dimensional flow (2D-RHE) was studied on the spectrum of adulterated honey. In the new method of RHE and 2D-RHE spectrum feature extraction, correlation analysis and principal component analysis based on mathematical modeling and pattern recognition and honey adulteration of honey adulteration degree on forecast different input variables by using BP neural network. The main conclusions are as follows: (1) honey adulteration of honey and the body does not change Newton without thixotropic rheological properties, also failed to change the viscosity temperature, viscosity and viscosity of water with temperature and water The attenuation law of both constant and activation energy index. U0 Ea with both the original honey samples with different degrees of adulteration, but not with the degree of adulteration varies. (2) the temperature activation energy and water activation energy by measuring effective flow threshold difference, to reflect the internal structure of the micro fluid group differences honey. (3) the 2D-RHE spectrum can reflect the details of each sample perturbation in shear force under temperature, can display the temperature perturbation under the shear rate increasing under the condition of apparent viscosity changes. (4) based on the true and false honey rheological analysis pattern recognition and BP neural network is feasible, the adulteration of honey adulteration degree the prediction performance of low.3, using two-dimensional near-infrared (2D NIR-RHE) - rheological fusion correlation spectra method, reveal the change of temperature perturbation associated NIR and RHE two analysis method. Characteristics of near infrared (NIR-RHE) and 2D - rheological characteristics of the establishment of NIR-RHE fusion The mathematical model of the true and false honey pattern recognition and adulteration of honey adulteration level prediction, discusses the multi information fusion technology for the possibility of honey adulteration identification. The main conclusions are as follows: (1) 2D NIR-RHE honey fingerprints characteristic is obvious, can change related visual display two analysis method of temperature perturbation NIR and RHE corresponding features that can be used to reflect the sample adulteration minutiae. (2) based on the true and false pattern recognition fusion technology and BP neural network is feasible, the adulteration of honey adulteration prediction performance is not high degree of.4, proposed the introduction including the model results, the average, the 50% threshold and the 90% threshold of these statistical parameters, comprehensive evaluation of network performance comparison research method. Found: the optimal pattern recognition method of near infrared and honey, the average value of the total recognition rate can reach 92.06%, 50% 93.9% 90% 83.3% threshold, threshold; rheological method and fusion method of the total recognition rate The average value is 83%, 50% threshold 85.2%, 90% threshold 81.5%., two-dimensional analysis is slightly better than 2D NIR-RHE method, the total recognition rate is 83.7%, 50% threshold 85.2%, 90% threshold 79.6%..

【学位授予单位】：南京农业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：S896.1

【参考文献】