紫外可见吸收光谱结合主成分-反向传播人工神经网络鉴别真假蜂蜜  

Application of UV-Visible Absorption Spectroscopy and Principal Components-Back Propagation Artificial Neural Network to Identification of Authentic and Adulterated Honeys

作　　者：欧文娟[1] 孟耀勇[1] 张小燕[1] 孔猛[1] 

机构地区：[1]华南师范大学生物光子学研究院,广州510631

出　　处：《分析化学》2011年第7期1104-1108,共5页Chinese Journal of Analytical Chemistry

基　　金：国家自然科学基金(No.60411130595);广东省中医药局项目(No.2008233)资助

摘　　要：研究紫外-可见吸收光谱技术结合化学计量学方法鉴别真假蜂蜜。根据蜂蜜中果糖和葡萄糖的典型质量比1.2:1.0,配制与真蜂蜜相近的掺假溶液,并以5%～20%的比例掺入真蜂蜜中。获取纯正蜂蜜和掺假蜂蜜的紫外-可见吸收光谱,选择最佳敏感波段250～400 nm的吸光度值进行主成分分析(PCA),优选主成分作为反向传播人工神经网络(BPANN)的输入向量。输出结果显示,校准集和预测集的准确鉴别率均为100%;对应的均方根误差分别为8.523×10-3和8.961×10-3。研究结果表明,基于PCA-BPANN的紫外-可见吸收光谱技术能够方便、快速、准确地鉴别真假蜂蜜,为食品质量的快速检测提供可靠参考。UV-visible（UV-vis） absorption spectroscopy in combination with chemometrics was used to identify the authentic and adulterated honeys.Adulterant solution prepared using D-fructose and D-glucose following the mass ratio of typical of honey composition（1.2∶1.0） was close to the real honey and added to individual honeys at levels of 5%,10%,15% and 20%.Absorption spectra of authentic and adulterated honeys in the wavelength range of 220-750 nm were acquired.The absorbance values of the best-sensitive band（250-400 nm） were selected to build models.The optimal identification model was developed with principal component analysis in combination with back propagation artificial neural network（PCA-BP-ANN）.The scores of optimal principal components were used as the input vectors of model.The output results showed that the correct identification rates were 100% for both the calibration and prediction sets and the corresponding root-mean-square errors were 8.523×10-3（RMSEC） and 8.961×10-3（RMSEP）,respectively.The study demonstrates that UV-vis absorption spectroscopy based on PCA and BP-ANN can be used as a convenient,rapid and accurate technique for identification of authentic and adulterated honeys

关 键 词：蜂蜜 掺假 紫外-可见吸收光谱 反向传播人工神经网络 主成分-反向传播人工神经网络 

分 类 号：S896.1[农业科学—特种经济动物饲养]